Doctoral thesis of Philosophy: Computational studies of neuroactive peptides for treating neurological disorders and disease

Computational Studies of Neuroactive Peptides for Treating Neurological

A thesis submitted in fulfilment of the requirements for the degree of Doctor of Philosophy

Jierong Wen

Master of Biotechnology with Distinction, RMIT University

School of Science

College of Science, Technology, Engineering and Maths

RMIT University

December 2020

Disorders and Disease

Declaration

I certify that except where due acknowledgement has been made, the work is that of the

author alone; the work has not been submitted previously, in whole or in part, to qualify for

any other academic award; the content of the thesis is the result of work which has been

carried out since the official commencement date of the approved research program; any

editorial work, paid or unpaid, carried out by a third party is acknowledged; and, ethics

procedures and guidelines have been followed.

Jierong Wen

09th December 2020

Acknowledgements

The present work cannot be made without the help and support from many people.

Thanks to my supervisory team, Andrew and David, you always help and support me during

my Ph.D. study and for my future career.

Andrew, you inspired me to step into the field of computational biology since my Master's

program. You guided me to pave the way for learning the drug discovery and design through

computational methods with your knowledge and patience, whenever I needed. I learned

that professional success was not only achieved via hardworking but also enthusiasm for the

study. You let me know that thinking critically and creatively is the most important thing

during research. Also, I appreciated your financial support for the publication and my

international academic conference travel.

David, you were happy to share the up-to-date experimental outcomes without exception,

providing a crucial basis for my computational research interests. You have proved that the

great success was based upon passion, vision and attention to even minor errors. I also want

to thank you for your generosity in providing me with the opportunities for joining in the

cooperation work that I was interested in, and your support when I was confronted with

problems during study.

The computational study in this thesis was designed based on the experimental results. Many

thanks to Dr. Jeffrey McArthur and colleagues at the University of Wollongong for conducting

the experiments and providing the fundamental knowledge that this thesis can reply on.

Thanks to the insightful suggestions regarding Chapter 4 of this thesis provided by Dr. Quentin

Kaas, Dr. Han-Shen Tae and Assoc. Prof. Rilei Yu. The computational resources in this thesis

were provided by the Pawsey Supercomputing Centre (Australia), National Computational

Infrastructure (NCI) and CSCS Swiss National Supercomputing Centre (Switzerland) via the

PRACE Project Access scheme.

I was grateful for the help from Assoc. Prof. Samantha Richardson, SEH Research Training

Services, School of Graduate Research (SGR) and RMIT Scholarship team for supporting me

during the COVID19 pandemic. I also appreciated the subsidy from the SGR team and the

School of Science for my international conference travel.

My parents invariably support me during my study. They encourage me with their efforts and

let me know that I should never stop pursuing my dream. Thus, I can be more confident and

focus on my study.

Financial support

This research was supported by the Australian Research Council (DP150103990).

III

Table of Contents Chapter 1 : Introduction – Nicotinic Acetylcholine Receptors and Voltage-Gated Ion Channels Targeted

by Conotoxins ......................................................................................................................................... 3

1.1 Overview ....................................................................................................................................... 4

1.2 nAChRs .......................................................................................................................................... 4

1.2.1 Structure of nAChRs .............................................................................................................. 5

1.2.2 nAChR Stoichiometries .......................................................................................................... 8

1.2.3 nAChR Multiple Binding Interfaces ....................................................................................... 8

1.3 Voltage-Gated Calcium Channels ................................................................................................. 9

1.3.1 Structure of VGCCs .............................................................................................................. 10

1.3.2 VGCCs Multiple Binding Sites .............................................................................................. 12

1.4 Conotoxins .................................................................................................................................. 13

1.5 Conotoxins as Versatile Tools ..................................................................................................... 14

1.5.1 Conotoxin Interactions with nAChRs................................................................................... 14

1.5.2 Conotoxin Interactions with Voltage-Gated Ion Channels .................................................. 19

1.5.3 The Potential of Conotoxins as Analgesics .......................................................................... 24

1.6 Computational Approaches Assist in Drug Design: General Observations and Overview ......... 25

1.7 Aim .............................................................................................................................................. 26

1.8 References .................................................................................................................................. 28

: Methodology ...................................................................................................................... 43

2.1 Homology Modelling .................................................................................................................. 44

2.1.1 Background.......................................................................................................................... 44

2.1.2 Searching for Sequence Similarity ....................................................................................... 45

2.1.3 Model Building .................................................................................................................... 46

2.1.4 Model Evaluation ................................................................................................................ 46

2.1.5 Examples of Application ...................................................................................................... 47

2.2 Molecular Dynamics Simulation ................................................................................................. 50

2.2.1 Background.......................................................................................................................... 50

2.2.2 Force Field Potential Energy Function and Parameters ...................................................... 52

2.2.3 Equation of Motion ............................................................................................................. 55

2.2.4 Time Step ............................................................................................................................. 56

2.2.5 Solvation .............................................................................................................................. 56

2.2.6 Periodic Boundary Conditions ............................................................................................. 58

2.2.7 Initial Configuration ............................................................................................................. 59

2.2.8 Treatment of Pressure and Temperature ........................................................................... 60

2.2.9 Examples of Application ...................................................................................................... 63

2.3 MMPBSA ..................................................................................................................................... 64

2.4 References .................................................................................................................................. 66

Chapter 3 : Effects of C-terminal Carboxylation on α-Conotoxin LsIA Interactions with Human α7

Nicotinic Acetylcholine Receptor .......................................................................................................... 71

3.1 Introduction ................................................................................................................................ 72

3.2 Methods ...................................................................................................................................... 75

3.2.1 Homology Modelling ........................................................................................................... 75

3.2.2 Molecular Dynamics Simulations ........................................................................................ 76

3.2.3 Structural Visualization and Analysis .................................................................................. 77

3.3 Results......................................................................................................................................... 78

3.3.1 Homology Modelling of LsIA and LsIA# Anchoring to Human α7 nAChR ............................ 78

3.3.2 Stability and Flexibility of the LsIA and LsIA# Bound α7 nAChR Complexes ....................... 78

3.3.3 Residues Undermine the Binding of LsIA at α7 Complementary Face versus LsIA# ........... 83

3.3.4 LsIA Favors Hydrogen Bonding and Cation-π Interactions between R10 and the α7(−)

Subunit ......................................................................................................................................... 84

3.3.5 LsIA# Favors Hydrophobic Interactions with α7 Subunit Residues via P7 and V11 ............ 88

3.3.6 α7-Y217 Forms Close Contacts with Both LsIA and LsIA# ................................................... 89

3.3.7 An Intra-molecular Salt-bridge in LsIA# is Retained upon Binding to α7 ............................ 90

3.3.8 Changes in Inter-subunit Contacts on the Five Interfaces of Adjacent α7 Subunits ........... 91

3.4 Discussion ................................................................................................................................... 95

3.4.1 LsIA Binding Causes Higher Fluctuations in C-loop and the Juxtamembrane Cys-Loop

Regions ......................................................................................................................................... 95

3.4.2 Carboxylation at LsIA C-terminus Causes Indirect Loss of Contacts between Key Residue R10

and the α7 Complementary Face ................................................................................................. 96

3.4.3 Key Pairwise Interactions Favor the Binding of LsIA at α7 nAChR versus LsIA# ................. 96

3.4.4 The Hydrophobic Interactions Undermine the Binding of LsIA at α7 nAChR versus LsIA# . 97

3.4.5 Intra-molecular Interactions Play a Role in the Loss of Key Toxin-receptor Contacts ........ 97

3.4.6 Inter-subunit Interactions Play a Role in the Loss of Key Toxin-receptor Contacts ............ 97

3.5 Conclusion .................................................................................................................................. 98

3.6 References ................................................................................................................................ 100

Chapter 4 : Effects of C-terminal Carboxylation of LsIA on Interactions with α3β2 Nicotinic

Acetylcholine Receptor Interfaces ...................................................................................................... 105

4.1 Introduction .............................................................................................................................. 106

4.2 Methods .................................................................................................................................... 110

4.2.1 Homology Modelling ......................................................................................................... 110

4.2.2 Molecular Dynamics Simulation ........................................................................................ 111

4.3 Results....................................................................................................................................... 113

4.3.1. Stability and Flexibility of the LsIA-α3β2 Complexes ....................................................... 113

4.3.2. LsIA and LsIA# Toxin-receptor Contacts ........................................................................... 118

4.3.3. Important Pairwise Interactions of LsIA with Residues at α3(+)β2(−) Interfaces ............ 120

4.3.4. Important Pairwise Interactions of LsIA with Residues at β2(+)α3(−) Interfaces ............ 122

4.3.5. Understanding Changes in Electrostatic Potential by C-terminal Carboxylation of LsIA at

α3(+)β2(−) and β2(−)α3(+) Interfaces ........................................................................................ 124

4.3.6. Inter-subunit Contacts at the α3(+)β2(−) and β2(+)α3(−) Interfaces .............................. 126

4.4 Discussion ................................................................................................................................. 129

4.4.1. Different Combinations of the Interfaces in Rat α3β2 may Affect the Distinct Stability and

Flexibility of the Receptor and LsIA ............................................................................................ 129

4.4.2. C-terminal Carboxylation of LsIA Marginally Enhances Overall α3β2 nAChR Contacts Versus

LsIA ............................................................................................................................................. 130

4.4.3. C-Terminal Carboxylation of LsIA Mainly Enhances the Contacts with Residues on the

Complementary Subunit of both α3(+)β2(−) and β2(+)α3(−) Interfaces ................................... 131

4.4.4 Inter-subunit Contacts of α3β2 nAChR Interfaces may Facilitate the Interactions with LsIA#

Versus LsIA ................................................................................................................................. 135

4.5 Conclusion ................................................................................................................................ 135

4.6 References ................................................................................................................................ 137

Chapter 5 : Differentiation of Cav3.2 Interactions by μ-Conotoxins as Potential Analgesics ............. 143

5.1 Introduction .............................................................................................................................. 144

5.2 Methodology ............................................................................................................................ 148

5.2.1 Homology Modelling ......................................................................................................... 149

5.2.2 Molecular Dynamics Simulations of Apo Cav3.2 ............................................................... 149

5.2.3 Docking of μ-Conotoxins to the Ion Channel. ................................................................... 152

5.2.4 Molecular Dynamics Simulations of Cav3.2 and μ-Conotoxin Complexes ........................ 153

5.3 Results....................................................................................................................................... 153

5.3.1 Flexibility and Stability of Cav3.2 and the Toxins upon Binding at the Membrane Ion Channel.

.................................................................................................................................................... 153

5.3.2 Binding Free Energy of μ-Conotoxins upon Anchoring to Cav3.2 Relative to GIIIA. .......... 159

5.3.3 Important Pairwise Contacts between μ-Conotoxins and Cav3.2 ..................................... 163

5.4 Discussion ................................................................................................................................. 173

5.4.1 Determinants Influencing the Flexibility of the Protein-peptide Complex ....................... 173

5.4.2 Key Residues on Cav3.2 Affecting the Binding Affinity of μ-Conotoxins versus GIIIA ....... 174

5.4.3 Important Pairwise Interactions of μ-Conotoxin Bound to Cav3.2 Complexes ................. 176

5.4.4 Prediction of Binding Conformations of μ-Conotoxins at Cav3.2 ...................................... 180

5.5 Conclusion ................................................................................................................................ 180

5.6 References ................................................................................................................................ 182

Chapter 6 : Summary and Future Work .............................................................................................. 190

6.1 Summary ................................................................................................................................... 191

6.2 Future work .............................................................................................................................. 193

VII

Appendix 1 : Supplementary Materials - Effects of C-terminal Carboxylation of LsIA on Interactions

with α3β2 Nicotinic Acetylcholine Receptor Interfaces ..................................................................... 195

Appendix 2 : Supplementary Materials - Differentiation of Cav3.2 Interactions by μ-Conotoxins as

Potential Analgesics ............................................................................................................................ 208

Appendix 3 : List of Publications and Conference Abstracts .............................................................. 221

VIII

List of Figures

Figure 1.1. 3-D structure of nAChR and its binding site targeted by α-conotoxins. ............................... 7

Figure 1.2. The topology of rabbit Cav1.1 channel (PDB ID: 5GJV) [93]. ............................................... 11

Figure 1.3. Structure of α-conotoxin LsIA (PDB ID: 5T90). .................................................................... 15

Figure 1.4. Structure of μ-conotoxin GIIIA (PDB ID: ITCG). ................................................................... 20

Figure 2.1. The sequence alignment outcome of human (h) α7 nAChR subunit and Ac-AChBP (PDB ID:

2BR8) at the extracellular domain using ClustalW. .............................................................................. 47

Figure 2.2. The sequence alignment outcome of rat (r) α3 and β2 nAChR subunits and Ac-AChBP (PDB

ID: 2BR8) at the extracellular domain using ClustalW. ......................................................................... 48

Figure 2.3. Homology model of human(h) Cav3.2 built upon cryo-EM structure of hCav3.1 α1 subunit

(PDB ID: 6KZO) [37] via SWISS-MODEL [15]. ......................................................................................... 49

Figure 2.4. One example of the sequence alignment results for Cav3.2 obtained from SWISS-MODEL.

.............................................................................................................................................................. 50

Figure 2.5. The simulation complex of LsIA bound to rat α3β2 nAChR. ............................................... 52

Figure 2.6. The conformation of TIP3P water molecules applied to the current study in this thesis. . 57

Figure 2.7. The side view of Cav3.2 centred in a grid box. .................................................................... 59

Figure 3.1. Ribbon conformation of LsIA binding to human α7 nAChR. ............................................... 74

Figure 3.2 Analysis of stability of human α7 nAChR over molecular dynamics (MD) simulations in the

LsIA and LsIA# bound types. ................................................................................................................. 82

Figure 3.3. A total number of contacts and the variations between LsIA and LsIA# bound types. ...... 83

Figure 3.4. Side view of binding conformation of LsIA-R10 to the ligand binding domain (LBD) of human

α7 nAChR. ............................................................................................................................................. 86

Figure 3.5. Binding model of LsIA-P14, R10 and C17 of amidated (wt) LsIA and carboxylated (mut) LsIA

into the binding pocket on the complementary (−) face of human α7 nAChR. ................................... 87

Figure 3.6. Side view of binding conformation of LsIA into the binding pocket with the interaction

between human α7(+)-W171 and LsIA-P7, A8 and V11, respectively. ................................................. 88

Figure 3.7. Top view of binding conformation of amidated (wt) and carboxylated (mut) LsIAs at the

principal (+) face of human α7 nAChR. ................................................................................................. 90

Figure 3.8. The intramolecular interactions within C-terminal carboxylated (mut) and amidated (wt)

LsIAs by binding to human α7 nAChR. .................................................................................................. 91

Figure 3.9. Comparison of differences in the interactions between principal (+) and complementary

(−) interfaces of human α7 nAChR after binding of LsIA and LsIA#. ..................................................... 93

Figure 4.1. Binding mode of -conotoxin LsIA at rat α3β2 nAChR. .................................................... 109

Figure 4.2. Root-mean-square deviation (RMSD) (Mean ± SD) plots for MD simulation of LsIA and its

carboxylated analogue (LsIA#) binding at α3β2 pentameric nicotinic acetylcholine receptor (nAChR)

complexes based upon the template 2BR8. ....................................................................................... 115

Figure 4.3. Differences between RMSF plots of LsIA and LsIA# binding at α3β2 nAChR.................... 117

Figure 4.4. Summary of the number of contacts in LsIA and LsIA# bound forms. ............................. 119

Figure 4.5. Important interactions formed by LsIA-P14 and C17 with the receptor residues on the

complementary (−) face at the α3(+)β2(−) interface. ........................................................................ 121

Figure 4.6. Important interactions formed by P7 with the receptor residues on the complementary (−)

face at α3(+)β2(−) interface. ............................................................................................................... 122

Figure 4.7. Important interactions formed by LsIA-P7, R10, V11, P14 and C17 with the receptor

residues on the complementary (−) face at β2(+)α3(−) interface. ..................................................... 124

Figure 4.8. Electrostatic potential energy on the surface of the α3β2 interfaces bound by LsIA/LsIA#

shown in PyMOL [73]. ......................................................................................................................... 126

Figure 4.9. Inter-subunit interactions between the principal and the accessory faces of α3β2 nAChR

interfaces bound by LsIA. .................................................................................................................... 128

Figure 5.1. Conformation of homology model Cav3.2 in apo form and bound by μ-conotoxin GIIIA. 145

Figure 5.2. Methodology flowchart involved in the present study of Cav3.2 using Microsoft Visio [66].

............................................................................................................................................................ 149

Figure 5.3. Molecular dynamics simulation system of Cav3.2 in apo form. ........................................ 151

Figure 5.4. The size of the docking grid set up using MGLTools 1.5.6 (www.mgltools.scripps.edu). . 152

Figure 5.5. Root-mean-square deviation (RMSD) plot of Cav3.2 α1 subunit in apo form and bound by

five μ-conotoxins, namely GIIIA, PIIIA, GIIIB, KIIIA and GIIIC (Mean ± SD).......................................... 154

Figure 5.6. Differences in RMS Fluctuation (Diff in RMSF) depicted for four (I, II, III and IV) domains of

Cav3.2 α1 subunit bound by other μ-conotoxins compared with that of GIIIA bound form. . ........... 156

Figure 5.7. Relative free binding energy (G) of μ-conotoxins PIIIA, GIIIB, KIIIA and GIIIC versus GIIIA

at human Cav3.2 via MMPBSA. ........................................................................................................... 160

Figure 5.8. Changes in the number of contacts between the coupled residues of Cav3.2 bound by μ-

conotoxins, PIIIA (green), GIIIB (yellow), GIIIC (red) and KIIIA (light purple) compared with Cav3.2/GIIIA

complex. .............................................................................................................................................. 164

Figure 5.9. Significant pairwise interactions between PIIIA and Cav3.2 compared with Cav3.2/GIIIA

complex. .............................................................................................................................................. 168

Figure 5.10. Significant pairwise interactions between GIIIB and Cav3.2 compared with Cav3.2/GIIIA

complex. .............................................................................................................................................. 170

Figure 5.11. Significant pairwise interactions between KIIIA and Cav3.2 compared with Cav3.2/GIIIA

complex. .............................................................................................................................................. 171

Figure 5.12. Significant pairwise interactions between GIIIC and Cav3.2 compared with Cav3.2/GIIIA

complex. .............................................................................................................................................. 172

List of Tables

Table 1.1. Sequence alignment of common α-conotoxins in the globular form (cysteine connectivity I-

III and II-IV) and their target nAChR subtypes. ..................................................................................... 16

Table 1.2. Sequence alignment of ω-conotoxins in globular form (cysteine connectivity I-IV, II-V and

III-VI) and their target voltage-gated calcium channel (VGCC) subtypes, as discussed in this chapter.

.............................................................................................................................................................. 21

Table 1.3. Amino acid sequence of µ-conotoxins in globular form (cysteine connectivity I-IV, II-V and

III-VI) and their target voltage-gated ion channels (VGICs), as discussed in this chapter. ................... 23

Table 2.1. The compatible water molecule models with widely used force fields. .............................. 58

Table 5.1. The important residues contribute to the changes of structural flexibility of Cav3.2 bound by

μ-conotoxins with statistical significance (p < 0.05), compared with the GIIIA bound form. ............ 158

XII

Abstract

Since the first synthesized conotoxin analgesic, ω-conotoxin MVIIA (ziconotide), was

approved by the FDA in 2004, the potential of venom peptides has been highlighted by many

studies. α-Conotoxins are the most widely studied subfamily due to their potency for nicotinic

acetylcholine receptors (nAChRs), the abnormal function of which is related to neurological

disease and disorders. On the other hand, voltage-gated ion channels, such as the T-type

(Cav3) calcium channels, are also involved in cell signalling. There is at present a lack of potent

inhibitors or selective probes to further investigate their specific roles in neural physiology

and pathology, such as chronic pain. Furthermore, owing to the diversity of this cocktail of

conopeptides, the selectivity of the conotoxins as inhibitors of neuronal proteins needs to be

further understood and improved. Therefore, it is important to identify the determinants that

may affect the sensitivity of the toxins to act as inhibitors of nAChRs and voltage-gated

calcium channels through computational methods at the atomic level.

In Chapter 1, the scientific literature is summarized within the scope of this study.

Additionally, the literature also refined the aim of this thesis is discussed in this chapter.

Chapter 2 provides a comprehensive overview of the principles of the applied computational

approaches in this study.

Once the background and methodology involved in the present work are understood, one can

apply this knowledge and skills to investigate specific research questions outlined in the

subsequent chapters.

Chapter 3 identifies the important residues of the LsIA/α7 nAChR complex that affect the

interactions between the toxin and the membrane protein due to C-terminal carboxylation of

LsIA. α7 nAChR is involved in Alzheimer’s disease, Parkinson’s disease, inflammation and

other disorders. The intramolecular salt-bridge established by R10 and C17 of carboxylated

LsIA (LsIA#) mainly affects the reduced interactions at R10 and C17 by LsIA# with the residues,

W77 and Q79, on α7 nAChR via cation- and hydrogen bond interactions.

The effects of C-terminal modification of LsIA on interactions with α3β2 nAChR, involved in

cardiovascular diseases, were also investigated in Chapter 4. This chapter highlights the

differences in interactions between neuropeptides and nAChR inter-monomer interfaces

composed of different subunits. For instance, (α3)2(β2)3, as a heteropentameric nAChR,

possesses three types of binding interfaces targeted by LsIA, namely α3(+)β2(−), β2(+)α3(−)

and β2(+)β2(−) binding sites. This chapter demonstrates the important role of LsIA#-C17 in

enhancing binding interactions with α3β2 nAChR due to its attraction with positively charged

residues, (−)K187 and (−)K82, at α3(+)β2(−) and β2(+)α3(−) interfaces respectively, rather than

those on the β2(+)β2(−)binding site. However, the effect of LsIA/LsIA#-R10 on interactions

with all α3β2 interfaces is minor, possibly due to the electrostatic repulsion between R10 and

the residues on the binding site in proximity to R10 of the toxins.

Recently, the role of Cav3.2 in pain sensation has been proposed, and analgesic molecules

targeting this ion channel have been explored as possible novel therapies for alleviating such

pain. Chapter 5 provides basic insights into the binding site of Cav3.2 and the residues

interacting with μ-conotoxins, which may account for the sensitivity of certain toxins, such as

GIIIA and PIIIA, to the Cav channel, in contrast to GIIIB, KIIIA and GIIIC which show fewer

inhibitory effects on Cav3.2. In general, these μ-conotoxins bind to the Cav3.2 via pore-

blocking mechanism. However, GIIIA binds more deeply at the outer vestibule of Cav3.2,

whilst PIIIA demonstrates a relatively superficial binding conformation. The partially different

binding conformations may be attributed to the acidic residue, D382, on the ion channel, that

makes contact with the residues GIIIA-K11 and K16 via electrostatic interactions.

Finally, Chapter 6 summarises the main findings in this thesis and provides further research

directions that need to be focused on.

Chapter 1 : Introduction – Nicotinic Acetylcholine

Receptors and Voltage-Gated Ion Channels Targeted

by Conotoxins

1.1 Overview

Neuropathic pain afflicts millions of population in Australia, whereas the effective analgesic

with few toxicant side effects to these chronic pain is still in demand [1]. It has been suggested

that neuronal nicotinic acetylcholine receptors (nAChRs) and voltage-gated calcium channels

(VGCCs) are indicated to be involved in neuropathic pain [2, 3], namely α7 nAChR [4], Cav3.2

[5] and others. However, the physiological, structural and dynamical properties of some

targets of chronic pain are still not fully characterised.

One promising class of molecules which has served as useful molecular probes of neuronal

receptor function are conotoxins, which are known to bind to ion channel subtypes and

nAChRs with high potency and specificity. They are considered potential candidate analgesics

for chronic pain and selective probes for dissection of the function of nAChRs and voltage-

gated ion channels. These conopeptides, extracted from the venom of marine cone snail, are

rich in disulphide bonds. The unique nature of structures renders these neuropeptides

promising in drug lead design and development. One example is the commercially available

intrathecal analgesic ziconotide [6], which is a synthetic compound including ω-conotoxin

MVIIA, for resistant cancer and AIDS pain treatment [7]. Therefore, it would be intriguing to

investigate the possibility of other conotoxins for treating neurological disorders, disease and

neuropathic pain. This chapter will introduce the general background of the nAChRs, voltage-

gated calcium channels, conotoxins, and their current antagonization mechanism at the

nAChRs and the ion channels.

1.2 nAChRs

nAChRs are predominantly situated in the central (CNS) and peripheral (PNS) nervous

systems, as well as non-neuronal cells, such as muscle cells [8]. These receptors are described

as ligand-gated ion channels belonging to the family of Cys-loop receptors [9]. To date, the

neuronal nAChRs can be generally subdivided into 2 categories, the homopentamers and

heteropentamers, which differ in their composition of unique nAChR subunits [10]. The

former consists of five identical subunits, whilst the latter is constituted by two or more

distinct subunits [11] with different combinations. The subunits consist of α1-10, β1-4, γ , δ

and ε subunit [10]. Additionally, the subunits including α1, δ, ε, β1, and γ are mainly expressed

in neuromuscular junction [12, 13].

These ligand-gated ion channels mainly contribute to the cholinergic pathways [14], including

synaptic transmission, inflammation and cognitive pathway via mediating ion flux through the

channel [15, 16]. Neurotransmitters, such as acetylcholine (ACh), can activate the nAChR

resulting in channel opening [17]. Therefore, the dysfunction or changes in numbers of

nAChRs are suggested to be related to various neurological diseases and disorders. The

subunits α7 and α9 included nAChRs have been suggested to be involved in inflammation and

neuropathic pain [12, 18-21]. Also, the homomeric α7 nAChR is related to Alzheimer's and

Parkinson’s disease [22-25]. The α3β4 nAChR and α5-contained nAChRs are highly implicated

in nicotine and drug abuse [26-29], as well as lung cancer [30, 31]. The α4β2 nAChR subtype

is a therapeutical target for neurodegenerative dementia treatment, whereas α3 and α6-

contained nAChRs may be potentially related to Parkinson’s disease and dementia with Lewy

bodies [32]. The muscle-type nAChR subunit α1 is the main target for the binding of

autoantibody and may cause myasthenia gravis [33-35], a neuromuscular dysfunction

disease. Therefore, potential inhibitors/activators targeting certain nAChR subtypes with

specificity would serve as potentially good candidates of drug leads for treating nAChR-

involved disease and pain. However, as the specific role and function of these nicotinic

receptors in neurological disease are ambiguous, it is also essential to develop selective

subtype probes to outline their physiological and pathological functions.

1.2.1 Structure of nAChRs

The crystal structures of nAChRs provide insights into the structural conformations of these

ligand-gated receptors. Although there is a relatively small group of nAChR structures that

has been resolved, the crystallized structures of these receptors facilitate the elucidation of

the conformations of the nicotinic receptor in three dimensions. Before 2001, most

experiments have been carried out in attempts to obtain a full-receptor structure, but failed

due to the low solubility of expressed nAChRs [36]. This obstacle was overcome by Unwin in

2005 [13], who determined the first 3-D structure of nAChR at 4 Å resolution (PDB ID: 2BG9)

via electron microscopy. These groundbreaking findings provide a basic insight into the

structure and function of nicotinic receptors. In general, pentameric nAChRs consist of 3

domains, the extracellular domain (ECD) possessing an N-terminus (N-T), transmembrane

domain (TMD) with 4 transmembrane helices, and the intracellular domain (ICD) linking with

a short C-terminus (C-T) [37, 38]. The interfaces between the ECD of subunits contain the

binding site for ligands, such as acetylcholine (ACh). The ECD may play an important role in

cation selectivity and permeation [37, 39]. Recently, other 3-D structures of ECD of nAChR

bound by agonists, such as α7 nAChR (PDB ID: 3SQ9 in 3.10 Å resolution) [40] and two cryo-

EM structures of α4β2 nAChR stoichiometries (PDB ID: 6CNJ in 3.70 Å resolution, 6CNK in 3.90

Å resolution) [41] illustrate the conformation of the receptor with small-molecule binding.

Furthermore, the whole structure of α4β2 nAChR and nicotine complex (PDB ID: 6PV7 in 3.34

Å resolution) [42] was identified, enabling exploration of the binding mode of the agonist at

nAChR and understanding the ion conduction mechanism through the channel of nAChR. In

addition, the X-ray structure of α4β2 nAChR (PDB ID: 5KXI in 3.94 Å resolution ) [43] in apo

form may serve as a target for theoretical study on ligand binding modes.

The structural topology of nAChRs is shown in Figure 1.1A. Figure 1.1B demonstrates the

binding domain of nAChR targeted by most of the known ligands, which situate at the

interface (close to C-loop) between the adjacent subunits. Each binding interface is composed

of 2 adjacent subunits, such as α3 and β2 subunit as the principal (+) and complementary (−)

faces, respectively.

Figure 1.1. 3-D structure of nAChR and its binding site targeted by α-conotoxins. (A) The topology of nAChR. The α-helices are presented by the purple color, while the β strands and loops are shown in yellow and cyan colors, respectively. (B) The common binding domain of ligands at one nAChR interface at the extracellular domain. (C) The potential binding sites of α-conotoxins at heterogenous (left) and homogenous (right) nAChRs, respectively. The green colored stars represent the toxins.

The structures of acetylcholine binding protein (AChBP) co-crystallized with α-conotoxin

initiated an alternative way of tackling studies of the binding approach of conotoxins. Since

there are few available crystal structures of nAChR bound by α-conotoxin complex, the

characterized 3-D structure of ECD of AChBP/α-conotoxin complexes, with relatively high

resolution compared with those obtained for the nAChRs, has been applied in a large number

of studies as the scaffold for the arrangement of nAChR subunits occupied by the

conopeptides [44, 45]. These templates include Aplysia californica acetylcholine binding

protein (Ac-AChBP) with PnIA mutant (PDB ID: 2BR8), ImI (PDB ID: 2BYP, 2CT9), TxIA [A10L]

(PDB ID: 2UZ6), GIC (PDB ID: 5CO5), PeIA (PDB ID: 5JME) and LvIA [44-52], respectively.

Additionally, the crystal structure of Lymnaea stagnalis (Ls-) AChBP bound by α-cobratoxin

(PDB ID: 1YI5) [53] and α3β4 subtype with LsIA complex (PDB ID: 5T90) [54] also contributes

to the research of α-conotoxin targeting nAChR for therapeutic purpose. Furthermore, the

nomenclature of conotoxins is set up according to the genes and cysteine framework of these

neuropeptides. For LsIA, the letter LS is the abbreviation of the organism, Lymnaea stagnalis,

of conotoxin A (gene) superfamily, whilst the letters I and A represent the cysteine framework

(I) of the peptide and the order (A as the first conotoxin identified in this organism) of the

discovery of toxins in such species.

1.2.2 nAChR Stoichiometries

Various nAChR stoichiometries have been determined and show distinct binding affinity to

their ligands, such as allosteric modulators, agonists and antagonists. The nAChR subunits

assemble in a distinct ratio in nAChR stoichiometries, such as (α7)4(β2)1 and (α7)3(β2)2 nAChR

isoforms uncovered by Milena and colleagues [55]. The results have suggested that both

isoforms exhibited similar inhibitory effects by antagonists as for α7 nAChR. However, this is

not the typical case among all nAChR subtypes. The most common α3β2 nAChR stoichiometry

is (α3)2(β2)3, whereas the existence of the (α3)3(β2)2 assembly may confer different binding

affinity of the ligand. This notion is supported by the uncovered distinct sensitivity of agonists,

inhibitors and allosteric modulators between (α4)3(β2)2 and (α4)2(β2)3, as well as between

(α9)3(α10)2 and (α9)2(α10)3, owing to the unique binding interface of α4(+)α4(−) and

α9(+)α9(−) of (α4)3(β2)2 and (α9)3(α10)2, respectively [56-59]. In addition, the divergent

arrangements of nAChR stoichiometry may confer differentiated pharmacophysicological

properties between them [60].

1.2.3 nAChR Multiple Binding Interfaces

Various binding interfaces can affect the potential of anchoring by the ligands, due to

different nAChR stoichiometries, resulting from various arrangements of nAChR subunits.

Also, the species of nAChR may influence the binding potency of the ligand [61]. The

selectivity of agonists, such as ACh and NS3573, was substantially reduced at the α10(+)α10(−)

interface of rat (α9)2(α10)3 and the α4(+)α4(−) interface of (α4)3(β2)2 nAChR versus other

binding interfaces of the nAChR isoforms [58, 62]. In contrast, both chicken α10(+)α10(−) and

α9(+)α9(−) interfaces of (α9)2(α10)3 and the β2(+)β2(−) interface of (α3)2(β2)3 showed

enhanced frequency of occupation by agonist, compared with other heteromeric binding

interfaces, including the α9(+)α10(−) and α10(+)α9(−) interfacial sites of (α9)2(α10)3 nAChR

and the α3(+)β2(−) and β2(+)α3(−) interfaces of (α3)2(β2)3 nAChR [62, 63]. Nevertheless,

further basic insights into the effects of differential binding sites on the binding affinity of

conotoxins are required to fully understand these differences at the molecular level.

1.3 Voltage-Gated Calcium Channels

VGCCs (Cav channels), a family of transmembrane proteins, conduct ions influx into cells

under the depolarization membrane potential of excitable cells [64]. This group of

mammalian ion channels is classified into ten subgroups due to the functional α1 subunit

being encoded by the CACNA1x genes with distinct cellular signal conduction functions [65,

66]. In general, there are three main subfamilies of VGCCs, namely Cav1 (L-type), Cav2(N, P/Q-

type) and Cav3 (T-type). The Cav1 and Cav2 types are activated upon high transmembrane

depolarising potential whereas the Cav3 class conducts ions under low depolarization, close

to the resting membrane potential [67 , 68].

The VGCCs are composed of a pore constituted by the α1 subunit and other key auxiliary

subunits (accessories), namely the α2δ, β and γ subunits [69-71]. Comparable to the diverse

dynamical properties of nAChRs, the different assemblies of subunits also affect the distinct

physiological and pathological properties of VGCCs, which require effective subtype-selective

drug leads. These channels include an α1 protein, conducting ion currents, which is the target

for several drugs [64, 72]. For example, Cav2.2 has been indicated as a promising target for

pain treatment, alcohol abuse and anxiety [73, 74], due to their important role in

neurotransmitter and nociception transmission [75, 76]. The FDA has approved the

intrathecal analgesic, ω-conotoxin MVIIA[77, 78], which is used for treating severe pain,

mainly based on its antagonising effects at Cav2.2 via a pore-blocking binding mechanism [78].

The accessory proteins, α2δ and β, are highly responsible for VGCCs trafficking, and are

associated with the function of pore gating (of α1 subunit) in Cav1 and Cav2 channel subtypes

[79, 80]. The α2δ-1 and α2δ-2 subunits are exemplified as the epileptic target for gabapentin

and pregabalin [81-83]. By contrast, the pore-forming α1 subunit of the Cav3 class may

independently function without assistance from auxiliary subunits [84, 85]. These T-type

calcium channels have been suggested to be implicated in multiple psychiatric and

neurological diseases including visceral, neuropathic and inflammatory pain, epilepsy and

Parkinson’s disease [86-92].

1.3.1 Structure of VGCCs

Due to the complexity of the structure of VGCCs and their locations, the intact 3-D structure

of human VGCC has yet to be fully resolved. However, the cryo-EM structure of rabbit Cav1.1

(PDB ID: 5GJV) [93], obtained at 3.6 Å resolution, provides an insight into the complete

structure of the L-type calcium channel. There are four repeats of the calcium channel in the

TMD as shown in Figure 1.2. In each repeat, the inactivated Cav1.1 pore of α1 protein with a

closed gate is surrounded by a voltage-sensing domain (VSD) that is coupled with the pore

gating [94]. The extracellular loops of the calcium channel, situated above the channel

vestibule, constitute a potential binding site for α2δ-1, together with the 2 caches and the von

Willebrand factor domain A (VWA) of α2δ-1 subunit in the ECD. The other auxiliary subunits

are the γ and β1 proteins, attached to VSD-IV of the calcium channel and below the VSD-II,

respectively. Also, the α1-interaction domain (AID) situates between the intracellular α-helix

linker I-II, which serves as a bridge linking VSD-II and the β1a subunit at the ICD and is

positioned at the cleft of β1a [95]. The interface between β1a and AID presents an AID-binding

site (ABS), which is responsible for ion channel activation [96, 97]. This cryo-EM structure was

used as a template for the construction of inactivated human Cav1.1 α1, rabbit Cav1.2 α2δ-1,

rat Cav1.2 α1 subunit, mouse Cav-α2δ-2 protein and human Cav1.2 α1 and β subunit and their

mutants [94, 98-103], which indicates a promising future for the use of recently identified

cryo-EM structure in understanding the physiological properties of the channels for drug

design. In this case, the cryo-EM structures of human Cav3.1 bound by Z944 blocker and in its

apo form (PDB ID: 6KZP in 3.10 Å resolution and 6KZO in 3.30 Å resolution) [104] may also be

exploited using computational methods for understanding the dynamic features of T-type

calcium channels and their respective binding mechanisms by ligands.

Figure 1.2. The topology of rabbit Cav1.1 channel (PDB ID: 5GJV) [93]. The pore forming α1 subunit is constituted by four domains, DI (light grey), DII (green), DIII (ice blue) and DIV (dark grey). The accessory subunits, α2δ-1, γ and β1a, are shown in lime, pink and yellow colors, respectively.

In addition, the experimentally-determined TMD conformation of bacteria CavAb (PDB ID:

4MVQ and 4MS2[105]) serves as the initial structure for investigation of ion conduction

through bacteria calcium channel [106], as well as providing a template to build homology

models of human calcium channels for evaluating the binding affinity and potency of 1,4-

Dihydropyridine (DHP) at the L-type calcium channel [107].

Due to the lack of the appropriate calcium channel 3-D structure in open form or bound by

ligand, the determined structure of sodium channel (Nav) and potassium (Kv) channel were

extensively used as the scaffolds to construct a comparative model of calcium channels [108].

The structure of the TMD of calcium channels is comparable with that of the sodium and

potassium channels [105, 109], therefore some studies employed the 3-D structures of these

channels to build the comparative model of Cav channels. Some examples of these templates

involved in comprehensive computational studies on calcium channels are described in the

following.

The mammalian Kv1.2 (PDB ID: 2A79 [110]) was applied to model the structure of rabbit Cav2.1

[111] and Cav1.2 with the assistance of a bacteria NaChBac structure [112, 113]. The

comparative model of Cav1.2 in apo form or with DHP occupied was constructed with

potassium channel KcsA (PDB ID: 2A9H [114]) in complex with charybdotoxin and KvAP bound

by Fab (PDB ID: 1ORS [115]), representing the closed and opened state of the pore for the

channel [116, 117]. The selectivity filter of Cav1.2 was built using the Nav1.4 P-loop domain

[118] by Tikhonov and Zhorov [117], due to the divergent structure of P-loop in Cav channel

from the Kv channel [105]. The reported NavAb channel (PDB ID: 5VB8) was used to model the

3-D structure of open gated human Cav1.2 [99].

1.3.2 VGCCs Multiple Binding Sites

The study on conotoxins' interactions with VGCCs is sparse. However, the VGCCs consist of

multiple binding sites as the target for various animal toxins [119]. The ω-conotoxins, such as

GVIA and MVIIA, impede the pore (formed between S5 and S6) at DIII of N-type calcium

channel [120, 121], and result in an almost irreversible inhibition [122]. The latter (MVIIA) has

been approved by the FDA to serve as a cancer pain relief medication [123], which will be

further discussed in section 1.5.2. Voltage gating modifiers can also mediate the gating via

anchoring to the extracellular region of VSDs [124], such as ω-agatoxin IVA and other spider

toxins targeting P/Q-type calcium channels [125]. Additionally, it has been suggested that the

α2δ subunit of Cav2 may be a potential pharmacophore for ω-conotoxins [126], however,

further study is required to understand the binding mechanism of ω-conotoxins at the α1

subunit of Cav2 channels co-expressed with the α2δ accessory protein.

Compared with Cav2 channels, the binding site of toxins anchoring to Cav3 remains to be

explored. Generally, the VSD at DIV was described as the main binding domain of T-type

calcium channels. δ-Conotoxin TxVIA was shown to bind at the S3-4 linker of Cav3.1 at DIV via

molecular docking methods [127]. Apart from conotoxins, other animal toxins have also been

demonstrated as gating modifiers, targeting the T-type calcium channels. For instance,

scorpion toxin, kurtoxin, was found to be the first known toxin that alters the gate opening of

Cav3.1, via anchoring to the S3-4 loop of VSD in DIV of the ion channel [128]. This binding

domain of Cav3.1 and Cav3.2 has been demonstrated as a potential therapeutical target by

spider toxins, ProTxI and ProTxII [129], respectively, similar to the impeding functions of these

two toxins at sodium channels [130].

1.4 Conotoxins

Marine cone snails are venomous mollusks, which generate and express their venom for

hunting via paralyzing their prey, such as worms, fish, and some marine invertebrate species.

Also, the high efficiency and diversity of the toxins (from hundreds to thousands of types in a

single species [131, 132]) produced by cone snails, make them one of the most poisonous

animals to humans [133]. The toxins extracted from the venom are called conopeptides, and

may be broadly classified into two groups: conotoxins with more than two disulphide bonds;

and disulphide-poor conopeptides with one or no disulphide linkage [134].

Conotoxins are one category of conopeptides that are rich in disulphide bonds (Figure 1.3).

They are small peptides, generally from 10 to 30 amino acids in length. Nevertheless,

understanding of the specific function of these neuropeptides is still an active area of

research. Although almost 100,000 different conotoxins have been predicted [135], less than

3% of them have been characterized, as listed in the Conoserver web-server

(http://www.conoserver.org/). These neurotoxins are basically categorized using three

taxonomic methods, according to gene superfamilies (genetic variations during evolution),

cysteine frameworks (the disulphide bond between conserved cysteine residues) and

pharmacological families (the binding potency at a particular target) [134]. The most

prevailing schemes are cysteine frameworks and pharmacological families in drug design. For

further details regarding these classification schemes, one can refer to the classification tables

on the Conoserver web-server [136]. The structure and amino acid sequence of conotoxins

involved in this work will be demonstrated in the main chapters in this thesis.

1.5 Conotoxins as Versatile Tools

In 2004, the FDA approved the analgesic ziconotide, based on the ω-conotoxin MVIIA

conopeptide isolated from Conus magus [137], which stunningly underlines the prospect for

intractable pain relief by conotoxins through intrathecal surgical injection [6, 77]. Regarding

the experimental studies, it has been illustrated that these conotoxins possess versatile

activities at Nav, Cav and Kv channels as well as nAChRs, mainly via antagonistic effects [38,

118, 124, 138-140]. In this section, the α-conotoxins will be introduced systematically with

respect to their pharmacological targets.

1.5.1 Conotoxin Interactions with nAChRs

α-Conotoxins are the largest group among the identified conotoxins. They were widely

exploited due to their specific inhibitory effects at nAChRs [141-144]. Similar to some other

conotoxins, α-conotoxins generally contain post-translational modifications, with C-terminal

amidation being amongst the most common [134]. Also, these neuropeptides are stabilized

by the formation of two disulphide bonds connecting four conserved cysteine residues (CICII-

CIII-CIV). The globular form of α-conotoxins encompasses cysteine connectivity I-III and II-IV

[145]. This isomer confers 2 loops (usually termed as loop 1 and loop 2 or loops m/n [134])

with distinct combinations of the residues within each loop, namely 3/4, 3/5, 4/3, 4/4, 4/5,

4/6, 4/7 and 5/5 [146-148]. LsIA epitomizes the 4/7 α-conotoxins, as shown in Figure 1.3. In

addition, other disulphide linking forms, such as I-II and III-IV and I-IV and II-III [149, 150],

named as ribbon and bead α-conotoxin isoforms, respectively, were also discovered in α-

conotoxins.

Figure 1.3. Structure of α-conotoxin LsIA (PDB ID: 5T90). The yellow bonds represent the disulphide bonds (I-III, II-IV) formed in LsIA.

• α-Conotoxins targeting α3* nAChRs

Experimental studies have demonstrated that native α-conotoxins ImI, PnIA, MII, PeIA, GIC,

OmIA, AnIA and AnIB, have remarkable selectivity at α3β2 versus other nAChR subtypes

(Table 1.1). In this section, the inhibitory effects of these potential α3β2 selective probes will

be summarised, based on previous experimental studies.

α-Conotoxin ImI shows inhibitory potency around 217 nM at human α7 with 8-fold lower IC50

at α9 nAChR [151, 152], yet demonstrating much higher antagonistic activity at the human

α3β2 nAChR subtype (IC50 ~ 41 nM). The determinants of ImI bound to α7 subtype are R7,

W10, P6 and D5 via interactions with corresponding hydrophobic residues on the receptor

[152-155]. Similar to ImI, the rat α3β2 is far more sensitive to blockade by PnIA (IC50 ~ 10nM)

versus that to α7 (IC50 ~ 256 nM) subtype [156, 157]. The PnIA mutant, PnIA[N11S], also shows

less binding preference to α7 nAChR versus its native type [158]. Compared with α3β4 (in

hundreds of nM) [159], MII selectively binds at rat α3β2 nAChR with IC50 of 0.5 ~ 3.7 nM,

whilst N5, P6 and H12 are the main residues ensuring occupation of the binding pocket by the

toxin [160, 161].

Table 1.1. Sequence alignment of common α-conotoxins in the globular form (cysteine connectivity I-III and II-IV) and their target nAChR subtypes. The conserved cysteine residues are highlighted in bold form.

α-

Cysteine

Amino acid sequence

Target nAChR

References

Conotoxins

framework

---GCCSDPRCAWR----C*---

α3β2 > α7 > α9

[151, 152]

ImI

4/3

---ACCSDRRCRWR----C*---

α7

[152, 154]

ImII

---GCCSDPRCRYR----CR#--

α9α10 > α3*,

RgIA

[162]

α6/α3β2β3, α7

---GCCSYPPCFATNPD-C*---

α3β4 > α2* ≈ α3β2,

AuIB

4/6

α4*, muscle-type

[163]

nAChR

---GCCSDPRCNYDHPEIC*---

α9α10 > α6* > α3β2

[164, 165]

Vc1.1

---GCCSLPPCAANNPDŸC*---

α3β2 > α7

[156, 157]

PnIA

---GCCSLPPCALSNPDŸC*---

α7 > α3β2

[156]

PnIB

---GCCSHPACSVNHPELC*---

α9α10 ≈ α3β2 ≈

PeIA

α6/α3β2β3 > α3β4

[166]

> α7

---GCCSNPVCHLEHSNLC*---

α3β2 > α3β4

[160, 161]

MII

4/7

---GCCSHPACAGNNQHIC*---

α3β2 > α3β4 ≈ α4β2

GIC

>> muscle-type

[167]

nAChRs

---GCCSHPACNVNNPHICG*--

α3β2 ≈ α7 > α6*

[168, 169]

OmIA

---GCCSYPPCFATNSDYC*---

AuIA

α3β4 > α2* ≈ α3β2,

[163]

α4*, muscle nAChR

---GCCSYPPCFATNSGYC*---

AuIC

----CCSHPACAANNQDŸC*---

α3β2

[170]

AnIA

---GCCSHPACAANNQDŸC*---

α3β2 > α7

AnIB

IRDECCSNPACRVNNPHVCRRR#

α7 ≈ α3β2

ArIA

[171]

--DECCSNPACRVNNPHVCRRR#

α7 ≈ α6/α3β2β3 >

ArIB

α3β2

IRDĚCCSNPACRVNNOHVC#---

α7 ≈ α3β2 > α4β2

[141, 143,

GID

172]

Note: * C-terminal amid; # C-terminal carboxylate; Ě: γ-carboxyglutamic acid; Ÿ: Sulfated tyrosine; O:

Hydroxyproline

PeIA exhibits higher binding potency at rat α3β2 (IC50 23 nM) and chimeric α6/α3β2β3 (IC50

19.8 nM) nAChRs versus the α3β4 (IC50 480 nM) and α7 nAChRs (IC50 1.8 μM) [166]. Mutations

on S9 (to histidine) and V10 (to alanine) of PeIA dramatically enhanced the selectivity at both

α3β2 and α6/α3β2β3 nAChR subtypes. By contrast, the N11 of the toxin was identified as the

key residue to be replaced with an arginine or certain unnatural amino acids for

differentiation between α3β2 and α6/α3β2β3 via mutagenesis scheme [173, 174].

Conotoxin GIC is remarkably selective to human α3β2 (IC50 1.1 nM) versus α3β4 and α4β2

with IC50 in the hundreds of nM, but has no detectable effects on muscle nAChRs [167],

possibly because of its unique residue, glutamine, at position 13 of the toxin versus that of

other α-conotoxins (also antagonizing at the α3β2) with relatively short side chains [175]. The

binding potencies of OmIA at α7 (IC50 27.1 nM) and α3β2 (IC50 11 nM) are similar, but is

reduced by 20-fold at α6* nAChR [168, 169]. It indicates that the OmIA is a potential probe

distinguishing between the anatomic location of α3β2 and α6* nAChR. In addition, some

other conotoxins, AuIA, AuIB and AuIC, are highly selective to α3β4 (IC50 ~ 750 nM) nAChR

versus α2*, α3β2, α4* and muscle type nAChRs [163]. Conotoxin AnIA and AnIB potently

inhibit ACh-evoked currents at α3β2 with IC50 of 5.84nM and 0.28 nM, respectively [170]. AnIB

also antagonizes the α7 (IC50 76 nM) nAChR subtype; on the other hand, the post-translational

modification, such as sulfotyrosine at position 16, influences the reduced sensitivity of α7

nAChR to AnIB by approximately 10-fold.

• α-Conotoxins targeting α7 nAChR

Certain conotoxins potently block α7 nAChR, such as PnIB, PnIA mutants, ImII, GID, ArIA and

ArIB. The determinants of these ligands and their blockade effects will be generally reviewed

in this section.

The native PnIB, from the same Conus species as PnIA, preferentially blocks α7 (IC50 ~ 61.3

nM) versus α3β2 (IC50 ~ 1.97 μM) [156], which may be accounted for the S4, L5, P6, P7, A9,

and L10 of the toxin via forming interactions with residues of α7 subtype [176]. The mutation

on alanine in position 10 of PnIA to leucine substantially shifted the sensitivity from α3β2 (IC50

~ 99 nM) to α7 (IC50 ~ 13 nM) nAChR [156]. This mutation also improved the binding potency

at chicken α7 nAChR [177], whilst PnIA [A9R A10L] reinforced the potency at α7 nAChR versus

its native type [178]. Based on the study on ImII, this toxin shows distinct preference from

ImI, and was shown to bind selectively to α7 nAChR, possibly endowed by its different binding

site against ImI [152, 154].

GID, with an extended N-terminus (but no amidated C-terminus), is highly selective to both

α7 (IC50 4.5 or 5.1 nM) and α3β2 (IC50 3.1 ~ 3.6 nM) nAChRs, and also inhibits α4β2 nAChR

(IC50 128.6 nM ~ 4.8 μM) [141, 143, 172]. The P9, A10, R12, V13, N14 and O16 of GID play an

important role in interacting with α7 nAChR; the mutation on R12 with an alanine reduced

the IC50 value of GID inhibition by 10-fold and 3-fold at α7 and α3β2 nAChRs, respectively; and

the first four N-terminal residues are important for the exertion of inhibitory effects on all

three nAChR subtypes [48, 143, 172]. R2, A10, N15 and V18 of the toxin are the key residues

for the sensitivity of α3β2 nAChR by GID [141].

Both ArIA and ArIB preferentially blockade the rat α7 (IC50 6.02 nM by ArIA and 1.81 nM by

ArIB) and α3β2 nAChRs (IC50 18 nM by ArIA and 60.1 nM by ArIB) expressed in Xenopus oocyte,

whilst ArIB is also selective to α6/α3β2β3 nAChR with IC50 value of 6.45 nM identified by

Whiteaker and colleagues [171]. They also found the ArIB[V11L,V16D] substantially reduced

its blockade to α3β2 (no detected effects > 10 μM) and α6/α3β2β3 (IC50 828 nM), yet retains

its potency at α7 (IC50 1.09 nM), which endows the ArIB [V11L,V16D] as one of the most

potent selective probes to α7 nAChR [179].

• α-Conotoxins targeting α9α10 nAChR

α-Conotoxin Vc1.1 (ACV1) specifically binds at nAChRs outside the CNS [180], which

differentiates it from other conotoxins targeting nAChRs in the CNS. This peptide with 16

amino acids from the venom of Conus victoriae shows inhibition at the rat α9α10 nAChR (IC50

19 nM) subtype, and moderately antagonizes rat α6* (in hundreds of nM) and α3* nAChRs

(in μM) with low potency [164, 165]. However, clinical trials of Vc1.1 have been suspended at

phase 2A owing to its markedly lower affinity at human α9α10 nAChR versus the rat nAChR

[181]. The different potencies of Vc1.1 between rat and human α9α10 nAChRs possibly result

from the absence of hydrogen bonds established by T59 of α9 subunit and the amidated C-

terminus of Vc1.1 , and the asparagine in position 9 of the toxin [182]. In addition, an

experimental study showed that the application of cyclic Vc1.1 (linking the N and C-terminus

of Vc1.1) renders the chronic visceral pain in mice with reduced pain sensation and improved

the oral bioavailability of the toxin as well [183, 184].

The RgIA with IC50 value of around 5nM at rat α9α10 nAChR, shows its highly distinct

specificity to this nAChR subtype from α2*, α3*, α4*, α6/α3β2β3 and α7 nAChRs [162],

possibly due to the Y10 of RgIA versus other α-conotoxins with high sequence similarity, such

as ImI and ImII targeting α7 and α3β2 nAChRs. Additionally, residues, D5, P6, R7 and R9 of

RgIA are suggested to play an important role in α9α10 nAChR sensitivity [185]. PeIA has similar

inhibitory activity at rat α9α10 (IC50 ~ 7 nM) relative to RgIA [166]. MiIIA [∆1,M2R, M9G, N10K,

H11K] also shows potential to bind at α9α10 (IC50 511 nM) and to selectively probe the α1*

muscle-type nAChRs (IC50 25 nM at α1β1δε and 16 nM at α1β1γδ) versus its native type [186].

Additionally, a recently characterized αO-conotoxin GeXIVA, from Conus generalis, presents

strong inhibition at the rat α9α10 nAChR subtype [150, 187].

1.5.2 Conotoxin Interactions with Voltage-Gated Ion Channels

Both μ-conotoxins and ω-conotoxins are widely known as potent inhibitors targeting voltage-

gated ion channels [78, 138, 188]. Specifically, μ-conotoxins preferentially bind at Nav

channels, whereas ω-conotoxins are sensitive to Cav2 channels. Distinct from α-conotoxins,

the μ-conotoxins generally possess 3 disulphide bonds linking 6 conserved cysteine residues,

one example being μ-conotoxin GIIIA (Figure 1.4). The common cysteine framework is CICII-

CIII-CIV-CVCVI, belonging to framework III of the conotoxins [134], such as GIIIA, GIIIB, GIIIC,

PIIIA, SxIIIB and others. The μ-conotoxins PnIVA and PnIVB, extracted from Conus pennaceus

venom, also exhibit a distinct cystine framework, composed of CICII-CIII-CIV-CV-CVI (framework

IV) [189]. As a result, both cysteine frameworks render μ-conotoxins with three loops. For ω-

conotoxins, there are also 6 conserved cystines, providing a four-loop framework of the

neuropeptides [189]. In this section, the distinct interactions by these conopeptides with their

corresponding preferentially targeted ion channels are discussed.

Figure 1.4. Structure of μ-conotoxin GIIIA (PDB ID: ITCG). The yellow bonds represent the disulphide bonds (I-IV, II-V, III-VI) formed in GIIIA (above). The black bold lines represent the disulphide bonds of the linear structure of GIIIA, while the blue lines show the three loops rendered by the disulphide linkage between conserved cysteine residues of the toxin.

• ω-Conotoxins targeting Cav2 channels

ω-Conotoxins show analgesic effects via their inhibition of VGCCs, and the ω-MVIIA conotoxin

(ziconotide) is representative of these toxins which favourably bind at the VGCCs [77, 137,

190]. This subfamily of neuropeptides, ranging in size from 24 to 29 residues, have three

disulphide bonds connecting six conserved cysteine residues, which confer four loops on the

toxins [191]. Apart from MVIIA, other ω-conotoxins also demonstrate inhibitory effects on

Cav2.2, such as ω-conotoxin SO-3. However the distinct recovery time from inhibition by SO-

3 [192], may indicate a different binding mechanism of this toxin at Cav2.2 versus that of

MVIIA, despite the comparable ability to inhibit the ion channel [193, 194]. The intrathecal

ziconotide targets the pore of N-type calcium channel (Cav2.2) (pIC50 10.65 or IC50 0.068 nM),

and has been shown to be effective at alleviating the server pain related to late-stage cancer

and AIDS [6, 195-197]. The other potential non-opioid analgesic is CVID (AM336), which also

binds at Cav2.2 (pIC50 10.42) via a similar pore blockade mechanism as ziconotide. The Y13

affiliated with loop 2 plays an important role in the sensitivity of MVIIA and CVID to Cav2.2

[191, 198], whereas the lysine in position 10 of native CVID and MVIIA [R10K] result in the

inhibition of the evoked transmitter release being augmented 100-fold versus the native

MVIIA [199]. Also, residues Y22 and K24 of GVIA (IC50 0.078 nM) exert great impact on

inhibitory activity of the toxin against N-type calcium channels [121, 122, 200].

Table 1.2. Sequence alignment of ω-conotoxins in globular form (cysteine connectivity I-IV, II-V and III-VI) and their target voltage-gated calcium channel (VGCC) subtypes, as discussed in this chapter. The conserved cysteine residues are highlighted in bold form.

ω-

Cysteine

Amino acid sequence

Target

References

Conotoxins

arrangement

channels

MVIIC

CKGKGAPCRKTMYDCCSGSCGR-RGKC*-

Cav2.1

[123]

MVIIA

CKGKGAKCSRLMYDCCTGSC-R-SGKC*-

[78]

C-C-CC-C-C

CVID

CKSKGAKCSKLMYDCCSGSCSGTVGRC*-

Cav2.2

[191, 198]

GVIA

CKSOGSSCSOTSYNCCR-SCNOYTKRCY*

[121, 122]

Note: * C-terminal amid; O: Hydroxyproline

Two recently identified ω-conotoxins, MoVIA (IC50 0.0035 nM) and MoVIB (IC50 0.365 nM),

showed commensurately drastic blockade effects on Cav2.2 expressed from rat dorsal root

ganglia (DRG) neurons relative to other Cav2.2-targeting conotoxins [197]. The different

residues on position 13 between the vermivorous MoVIB (with R13) and most other

piscivorous ω-conotoxins (with Y13) may be sub-species-dependent in terms of their effective

antagonism of the N-type calcium channel. In addition, although previous experiments

demonstrated that MVIIC can inhibit Cav2.1 (P/Q-type calcium channel) [123], the lethality of

the toxin on mammals makes it currently challenging for pharmacological development [201].

• μ-Conotoxins targeting voltage-gated sodium and/or potassium channels

μ-Conotoxins were widely known as constituents of the cocktail of venom for antagonising

voltage-gated sodium and potassium channels [139, 202-204], which are responsible for

mediating the influx of sodium ions in both excitable and non-excitable cells [205]. The

abnormal function of sodium channels may be involved in neuropathic and diabetic pain, long

QT syndrome, and other diseases [205-207]. Thus, the sodium channels are treated as a

prospective pharmacophore. The antagonists selective to sodium channel subtypes have

been extensively studied. PIIIA has been shown to potently inhibit bacterial sodium channels

(IC50 ~ 5 pM), NaChBac [208]. KIIIA shows moderate binding affinity at Nav1.7 (IC50 363 nM),

similar to spider toxin PaurTx3 [209], whereas the wild type KIIIA inhibits the human Nav1.7

and rat Nav1.4 with IC50 of 37 nM and 97 nM, respectively [210]. An experimental study

showed that the R14 of KIIIA is the key residue that enhanced the selectivity at rat Nav1.7

versus Nav1.4 [210], which, however, was not affected by Nav1.7 [I1410D] mutant,

demonstrating an enhanced selectivity by KIIIA at Nav1.7 compared with Nav1.4. Furthermore,

the residue, tryptophan, at position 8 of KIIIA contributes to the specificity at skeletal Nav1.4

versus other neuronal sodium channels expressed from oocytes, identified via

electrophysiological experiments [211]. Other μ-conotoxins, SmIIIA, CnIIIA and MIIIA, also

demonstrate their binding affinity with subtype selectivity to sodium channels [188]. SmIIIA

inhibits Nav1.6 with IC50 of 160 nM [212]. CnIIIA exhibits inhibition at Nav1.4 ( IC50 270 nM)

[213]. In that study, it was inferred that G6, G9 and N8 of CnIIIA may represent the key factors

influencing the selectivity towards TTX resistant sodium channels. Both PIIIA and SxIIIA

antagonize Nav1.1 with IC50 of 120 and 370 nM [213, 214], respectively. TsIIIA marginally

inhibits human sodium channels, hNav1.8 (IC50 2.11 μM) [215]. However, as TsIIIA has no

inhibitory effects on hNav1.1-7 up to 10μM, it may serve as a potential selective probe for

distinguishing Nav1.8 from other Nav1 channels via mutagenesis studies.

Table 1.3. Amino acid sequence of µ-conotoxins in globular form (cysteine connectivity I-IV, II-V and III-VI) and their target voltage-gated ion channels (VGICs), as discussed in this chapter. The conserved cysteine residues are highlighted in bold form.

µ-

Cysteine

Amino acid sequence

Target channels References

Conotoxins

arrangement

SIIIA

ZNCCNG--GCSSKWCRDHARCC*--

Kv1.1, Kv1.6 >>

[216]

kv1.2-5, kv2.1

SmIIIA

ZRCCNGRRGCSSRWCRDHSRCC*--

Nav1.6

[212]

CC-C-C-CC

CnIIIA

GRCCDVPNACSGRWCRDHAQCC*--

Nav1.4

[213]

TsIIIA

-GCCRWP--CPSR-CGMARCCSS

Nav1.8 >>

[215]

Nav1.1-7

Note: * C-terminal amid; Z: Pyroglutamic acid

For sodium channels, one of the binding determinants by μ-conotoxins is the conserved

positively charged residues in loop 2 for most μ-conotoxins [213]. In a previous experimental

study, PIIIA-R12 and R14 were suggested to be involved in interactions with the residues on

the S5 loop and the selective filter of NaChBac, respectively, via formation of salt bridges

[208]. Similarly, conserved arginines at positions 7 and 13 of KIIIA and SmIIIA, respectively,

also modified the inhibitory effects at sodium channels [210, 212]. Apart from the conserved

residues, the subtype specificity by μ-conotoxins may be conferred by non-conserved

residues at other positions, such as those at the N-T and the 1st loop of the toxins.

Potassium channels are targeted by μ-conotoxins with specificity as well. The voltage-gated

potassium channels regulate the efflux of potassium ions (K+) and the membrane potential

and therefore affect the cell excitability of muscle and neurons [216]. Neurological diseases

related to potassium channel dysfunction include schizophrenia, Alzheimer’s disease,

paroxysmal, choreoathetosis, episodic ataxia and others [217]. SIIIA and PIIIA target Kv1.6

and Kv1.1 with high binding potency ( in nanomolar) [216], respectively, which was identified

via electrophysiological recordings. However, Kv1.2-5 and Kv2.1 are not sensitive to these two

conopeptides up to 10 μM of the toxins. Apart from SIIIA and PIIIA, there is little evidence

demonstrating the potential potency of μ-conotoxins at voltage-gated potassium channels.

1.5.3 The Potential of Conotoxins as Analgesics

Although peptide is an emerging drug market, they can be easily synthesized for specifically

binding at pharmaceutical target possibly with less side effects versus the small-molecule

drugs, such as insulin and glucagon-like peptide 1 agonists (GLP-1) [218]. Since the FDA

approved analgesic, ω-MVIIA, has been commercialised, several conotoxins, such as α-RgIA4

(KCP-400), Mini-Ins (insulin analog), Contulakin-G (CGX-1160), have also been put through the

pre-clinical or clinical trials particularly for treating neuropathic pain and diabetes [219], while

the analgesic efficacy of some of these conotoxins needs to be verified [220]. Nonetheless,

the stability, efficacy, relatively low oral bioactivity and toxicity issues obstacle the way to the

success of conotoxins as potential drug leads [184, 201] for neuropathic pains, neurological

disease and others. For instance, the toxicity and safety issues resulted in a discontinuation

of the χ-MrIA (Xen2174) and CGX-1160 at clinical trial [221], while the low efficacy of Vc1.1 at

human α9α10 nAChR compared with the rat α9α10 nAChR at the pre-clinical phase, also

prevents the development of such conotoxin as analgesic drug compound [222, 223].

However, due to unmet medical needs to non-addictive drug candidates versus the opioid-

based medication, studies on conotoxins have made progress for improving the stability and

oral bioavailability of the toxins [218], such as the cyclization on Vc1.1 backbone making oral

administration applicable to the toxin [184]. In addition, since the synergistic effects of

conotoxins with certain opioid medicines and neurotoxin have been investigated [224], it may

be intriguing to investigate the potential of conotoxins combined with other toxins/opioids in

treating corresponding diseases with efficacy, oral bioavailability, and fewer side effects.

1.6 Computational Approaches Assist in Drug Design: General Observations and Overview

Computational methods have facilitated the design of new drug leads and selective probes

for pharmacological purposes, due to the advancement of bioinformatics, computational

technology and the explosion of data resources [225]. More importantly, the extremely high

expense and time consumption for drug design and discovery, drive the emergence and

development of computation-aided drug discovery. To successfully produce a drug for a

targeted disease, it can cost from 550 million to 3.8 billion AUD and take approximately 14

years [226-228], for which mutagenesis and synthesis account for a large portion of the costs

and time [227]. However, it may fail at any stage of this procedure. Thus, computational

methods and effective use of data repositories serve as a theoretical checkpoint for

facilitating the drug design procedure and reducing the substantial expenses, rather than

conducting experiments relying on hypotheses without prior rationales.

To date, early drug design research has involved substantial use of computational approaches,

such as molecular modelling [229], docking [230], and molecular dynamics (MD) simulation

[100] to generate and develop the models of proteins, peptides or other molecules of interest

with unknown structures. The 3-D model of the target receptor can be subsequently

processed via docking and/or virtual screening for selecting the optimal binding ligands which

may potentially act as antagonists/agonists [142, 231], with the association of big data

repositories, such as the ZINC or PubChem database [232]. The obtained receptor-ligand

complex may then be subjected to further studies via MD simulations to predict the

movement of the receptor calculated under appropriate conditions and parameters (using

MD simulation packages and their associated force fields), to determine the stability of

receptor-ligand complexes obtained after docking. In addition to predictions of receptor-

ligand structures and dynamics, numerous computational approaches have also been applied

to improve the potential inhibition/activation propensity of the ligand anchored to the

receptor, such as modifications of certain residues. Strategies for enhancing conotoxin

activity at nAChRs have demonstrated the effective use of molecular docking, together with

MD simulations, in novel peptide design [231, 233-235]. In addition, artificial intelligence (AI),

especially for machine learning also helps drug discovery. It has been shown that machine

learning and deep learning can improve the accuracy of virtual screening in structure-based

drug design [236] and the homology modelling of proteins [237].

1.7 Aim

To date, extensive studies have focused on the conotoxin interactions with nAChRs or VGICs

involved in neurological disorders and disease [49, 233, 238-240]. However, much remains to

be discovered regarding the molecular-level binding mechanism of certain conotoxins at the

nAChRs or T-type calcium channels involved in such diseases, as well as their interactions. The

central aim of the work described in this thesis is to understand the interactions between α-

conotoxins LsIA/LsIA# and nAChRs as well as the contacts formed between μ-conotoxins and

T-type calcium channels. The details of the specific aims and the corresponding chapter

address them, as well as a brief summary of the key findings, are as follows:

1) Elucidate the binding mechanism of an LsIA analogue (LsIA with a C-terminal carboxyl

group) via computational methods, and identify the key factors affecting the changes

in binding potency at human α7 nAChRs compared with wild-type LsIA.

This work is explored in Chapter 3. The LsIA analogue comprises a free negative charge at C-

terminus which forms an intramolecular contact with LsIA-R10 via forming a salt bridge.

However, this salt-bridging interaction was absent in the wild-type LsIA. The computational

findings show the reduced contacts between LsIA analogue (LsIA#) and α7 nAChR, such as

R10-W77 via cation- interaction, R10-L141 and C17-Q79 via forming hydrogen bonds,

compared with LsIA. The limited number of changes in interactions by LsIA and its analogue

with α7 nAChR, is commensurate with the experimental study by Inserra and colleagues [241],

demonstrating a 3-fold reduction of binding affinity at α7 nAChR due to the LsIA

carboxylation.

2) Determine the interactions by LsIA and its analogue with α3β2 nAChR interfaces to

show the role of different combinations of interfaces for heterogenous nAChRs in the

binding of the neurotoxin.

Distinct from human α7 nAChR, rat (α3)2(β2)3 nAChR is a heteropentameric nAChR, which

consists of two α3 and three β2 subunits (Figure 1.1). Thus, there are three types of interface

composition for α3β2 nAChR. The electrophysiological study by Inserra et al. also shows the

similar binding affinity of α-conotoxin LsIA (with an amidated C-terminus) at human α7 and

rat α3β2 nAChRs [241], while the augmented inhibition at α3β2 nAChR was observed due to

C-terminal carboxylation, with reduced inhibition by the LsIA analogue at α7 nAChR as

mentioned previously, compared with native LsIA. Therefore, differences in potency between

α7 nAChR and α3β2 nAChRs could be up to 9-fold. However, few studies show the effects of

individual interfaces of certain nAChR isoforms on conotoxin binding, therefore it merits

attention to investigate the contribution from α3β2 nAChR interfaces to the binding of LsIA

and its analogue.

Chapter 4 discusses the marginally enhanced interactions by LsIA analogue with α3β2 nAChR

versus the native LsIA. Among all the α3β2 interfaces, β2(+)α3(−) and α3(+)β2(−) interfaces

account for the enhanced contacts with LsIA analogue, while this was undermined at the

β2(+)β2(−) interface upon binding by LsIA# versus LsIA.

3) Identify the binding mechanism of μ-conotoxins with their distinct interactions with

Cav3.2.

Apart from the potential drug leads for treating neurological disorders and cancer by

abnormal function of α7 and α3β2 nAChRs detailed in Chapters 3 and 4, the conotoxins are

also treated as a potent analgesic for neuropathic pain related to overexpression and

dysfunction of Cav3.2 [242, 243] (Chapter 5). Nonetheless, few conotoxins showed robust

antagonization at T-type calcium channels, for which the binding mechanism of these

conopeptides is presently unclear. Computational approaches were applied to understand

the binding mechanism of five μ-conotoxins at Cav3.2 and to compare their binding affinities.

The results show enhanced binding affinity by PIIIA and KIIIA compared to several other μ-

conotoxins, with a similar pore-blocking mechanism as those targeting the Nav channels [138,

215, 244]. The electro-positively charged residues of the μ-conotoxins mainly contribute to

the different interactions with Cav3.2 around the pore vestibule and extracellular loops. This

chapter proposes several hypotheses regarding conopeptide interactions with Cav3.2 which

may be explored by further experimental study probing the role of binding site residues on

the channel in conotoxin binding.

1.8 References

Harriott AM, Gold MS. Contribution of primary afferent channels to neuropathic pain.

Dib-Hajj SD, Cummins TR, Black JA, Waxman SG. Sodium channels in normal and

Bourinet E, Altier C, Hildebrand ME, Trang T, Salter MW, Zamponi GW. Calcium-

McGivern JG. Ziconotide: a review of its pharmacology and use in the treatment of

Gotti C, Clementi F. Neuronal nicotinic receptors: from structure to pathology.

Tsetlin V, Kuzmin D, Kasheverov I. Assembly of nicotinic and other Cys-loop receptors.

Hurst R, Rollema H, Bertrand D. Nicotinic acetylcholine receptors: From basic science

Hone AJ, McIntosh JM. Nicotinic acetylcholine receptors in neuropathic and

Unwin N. Refined structure of the nicotinic acetylcholine receptor at 4Å resolution.

de Jonge WJ, Ulloa L. The alpha7 nicotinic acetylcholine receptor as a pharmacological

1. Current pain and headache reports. 2009;13(3):197-207. 2. pathological pain. Annual Review of Neuroscience. 2010;33(1):325-47. 3. permeable ion channels in pain signaling. Physiological Reviews. 2014;94(1):81-140. Dolphin AC. Voltage-gated calcium channels: their discovery, function and importance as drug targets. Brain and Neuroscience Advances. 2018 Oct 2;2:2398212818794805. doi: 10.1177/2398212818794805. PMID: 30320224; PMCID: PMC6179141. 4. Bagdas D, Gurun MS, Flood P, Papke RL, Damaj MI. New insights on neuronal nicotinic acetylcholine receptors as targets for pain and inflammation: a focus on α7 nAChRs. Current neuropharmacology. 2018;16(4):415-25. 5. Bourinet E, Alloui A, Monteil A, Barrère C, Couette B, Poirot O, et al. Silencing of the Cav3.2 T-type calcium channel gene in sensory neurons demonstrates its major role in nociception. The EMBO Journal. 2005;24(2):315-24. 6. pain. Neuropsychiatric Disease and Treatment. 2007;3(1):69-85. 7. Staats PS, Yearwood T, Charapata SG, Presley RW, Wallace MS, Byas-Smith M, et al. Intrathecal ziconotide in the treatment of refractory pain in patients with cancer or AIDS: a randomized controlled trial. JAMA. 2004;291(1):63-70. 8. Progress in Neurobiology. 2004;74(6):363-96. 9. Journal of Neurochemistry. 2011;116(5):734-41. 10. to therapeutics. Pharmacology & Therapeutics. 2013;137(1):22-54. 11. Wu J, Lukas RJ. Naturally-expressed nicotinic acetylcholine receptor subtypes. Biochemical Pharmacology. 2011;82(8):800-7. 12. inflammatory pain. FEBS Letters. 2018;592(7):1045-62. 13. Journal of Molecular Biology. 2005;346(4):967-89. Buckingham SD, Jones AK, Brown LA, Sattelle DB. Nicotinic acetylcholine receptor 14. signalling: roles in Alzheimer's disease and amyloid neuroprotection. Pharmacological Reviews. 2009;61(1):39-61. Xu S, Zhang T, Kompella SN, Yan M, Lu A, Wang Y, et al. Conotoxin αD-GeXXA utilizes 15. a novel strategy to antagonize nicotinic acetylcholine receptors. Scientific Reports. 2015;5:14261. 16. target for inflammation. British Journal of Pharmacology. 2007;151(7):915-29. 17. Turner MW, Marquart LA, Phillips PD, McDougal OM. Mutagenesis of α-conotoxins for enhancing activity and selectivity for nicotinic acetylcholine receptors. Toxins. 2019;11(2):113. Di Cesare Mannelli L, Cinci L, Micheli L, Zanardelli M, Pacini A, McIntosh JM, et al. α- 18. conotoxin RgIA protects against the development of nerve injury-induced chronic pain and prevents both neuronal and glial derangement. Pain. 2014;155(10):1986-95.

D'Andrea MR, Nagele RG. Targeting the alpha 7 nicotinic acetylcholine receptor to in Alzheimer's disease pyramidal neurons. Current

Pohanka M. Alpha7 nicotinic acetylcholine receptor is a target in pharmacology and

Rose JE. Multiple brain pathways and receptors underlying tobacco addiction.

Berrettini W, Yuan X, Tozzi F, Song K, Francks C, Chilcoat H, et al. α-5/α-3 Nicotinic for heavy smoking. Molecular Psychiatry. increase risk

Improgo MRD, Scofield MD, Tapper AR, Gardner PD. From smoking to lung cancer: the

Loram LC, Harrison JA, Chao L, Taylor FR, Reddy A, Travis CL, et al. Intrathecal injection 19. of an alpha seven nicotinic acetylcholine receptor agonist attenuates gp120-induced mechanical allodynia and spinal pro-inflammatory cytokine profiles in rats. Brain, Behavior, and Immunity. 2010;24(6):959-67. 20. Wang H, Yu M, Ochani M, Amella CA, Tanovic M, Susarla S, et al. Nicotinic acetylcholine receptor α7 subunit is an essential regulator of inflammation. Nature. 2003;421(6921):384-8. 21. Shytle RD, Mori T, Townsend K, Vendrame M, Sun N, Zeng J, et al. Cholinergic modulation of microglial activation by α7 nicotinic receptors. Journal of Neurochemistry. 2004;89(2):337-43. 22. reduce amyloid accumulation Pharmaceutical Design. 2006;12(6):677-84. 23. Oz M, Petroianu G, Lorke DE. α7-Nicotinic acetylcholine receptors: new therapeutic avenues in Alzheimer’s disease. In: Li MD, editor. Nicotinic Acetylcholine Receptor Technologies. New York, NY: Springer New York; 2016. p. 149-69. 24. Kalkman HO, Feuerbach D. Modulatory effects of α7 nAChRs on the immune system and its relevance for CNS disorders. Cellular and Molecular Life Sciences. 2016;73(13):2511- 30. 25. toxicology. International Journal of Molecular Sciences. 2012;13(2):2219-38. Frahm S, Ślimak Marta A, Ferrarese L, Santos-Torres J, Antolin-Fontes B, Auer S, et al. 26. Aversion to nicotine is regulated by the balanced activity of β4 and α5 nicotinic receptor subunits in the medial habenula. Neuron. 2011;70(3):522-35. 27. Biochemical Pharmacology. 2007;74(8):1263-70. 28. Toll L, Zaveri NT, Polgar WE, Jiang F, Khroyan TV, Zhou W, et al. AT-1001: a high affinity and selective α3β4 nicotinic acetylcholine receptor antagonist blocks nicotine self- administration in rats. Neuropsychopharmacology. 2012;37(6):1367-76. 29. receptor subunit alleles 2008;13(4):368-73. Hung RJ, McKay JD, Gaborieau V, Boffetta P, Hashibe M, Zaridze D, et al. A 30. susceptibility locus for lung cancer maps to nicotinic acetylcholine receptor subunit genes on 15q25. Nature. 2008;452(7187):633-7. 31. CHRNA5/A3/B4 connection. Oncogene. 2010;29(35):4874-84. 32. Gotti C, Moretti M, Bohr I, Ziabreva I, Vailati S, Longhi R, et al. Selective nicotinic acetylcholine receptor subunit deficits identified in Alzheimer's disease, Parkinson's disease and dementia with Lewy bodies by immunoprecipitation. Neurobiology of Disease. 2006;23(2):481-9. 33. MacLennan C, Beeson D, Buijs A-M, Vincent A, Newsom-Davis J. Acetylcholine receptor expression in human extraocular muscles and their susceptibility to myasthenia gravis. Annals of Neurology. 1997;41(4):423-31. 34. Kalamida D, Poulas K, Avramopoulou V, Fostieri E, Lagoumintzis G, Lazaridis K, et al. Muscle and neuronal nicotinic acetylcholine receptors. The FEBS Journal. 2007;274(15):3799- 845.

Hille B. Ion channels of excitable membranes. 3rd ed. ed. Sunderland, Mass.: Sinauer;

ion permeation Gharpure A, Teng J, Zhuang Y, Noviello CM, Walsh RM, Jr., Cabuco R, et al. Agonist in the α3β4 ganglionic nicotinic receptor. Neuron.

Tzartos S, Hochschwender S, Vasquez P, Lindstrom J. Passive transfer of experimental 35. autoimmune myasthenia gravis by monoclonal antibodies to the main immunogenic region of the acetylcholine receptor. Journal of Neuroimmunology. 1987;15(2):185-94. Grant MA, Gentile LN, Shi Q-L, Pellegrini M, Hawrot E. Expression and spectroscopic 36. analysis of soluble nicotinic acetylcholine receptor fragments derived from the extracellular domain of the α-subunit. Biochemistry. 1999;38(33):10730-42. 37. Wu Z-s, Cheng H, Jiang Y, Melcher K, Xu HE. Ion channels gated by acetylcholine and serotonin: structures, biology, and drug discovery. Acta pharmacologica Sinica. 2015;36(8):895-907. 38. McIntosh JM, Yoshikami D, Mahe E, Nielsen DB, Rivier JE, Gray WR, et al. A nicotinic acetylcholine receptor ligand of unique specificity, alpha-conotoxin ImI. Journal of Biological Chemistry. 1994;269(24):16733-9. 39. 2001. Li S-X, Huang S, Bren N, Noridomi K, Dellisanti CD, Sine SM, et al. Ligand-binding 40. domain of an α7-nicotinic receptor chimera and its complex with agonist. Nature Neuroscience. 2011;14(10):1253-9. 41. Walsh RM, Jr., Roh S-H, Gharpure A, Morales-Perez CL, Teng J, Hibbs RE. Structural principles of distinct assemblies of the human α4β2 nicotinic receptor. Nature. 2018;557(7704):261-5. 42. selectivity and 2019;104(3):501-11.e6. 43. Morales-Perez CL, Noviello CM, Hibbs RE. X-ray structure of the human α4β2 nicotinic receptor. Nature. 2016;538(7625):411-5. 44. Celie PHN, Kasheverov IE, Mordventisev DY, Hogg RC, van Nierop P, van Elk R, et al. Crystal structure of nicotinic acetylcholine receptor homolog AChBP in complex with an α- conotoxin PnIA variant. Nature Structural and Molecular Biology. 2005;12. 45. Ulens C, Hogg RC, Celie PH, Bertrand D, Tsetlin V, Smit AB, et al. Structural determinants of selective α-conotoxin binding to a nicotinic acetylcholine receptor homolog AChBP. Proceedings of the National Academy of Sciences. 2006;103(10):3615-20. Dutertre S, Ulens C, Büttner R, Fish A, van Elk R, Kendel Y, et al. AChBP-targeted α- 46. conotoxin correlates distinct binding orientations with nAChR subtype selectivity. The EMBO Journal. 2007;26(16):3858-67. Grishin AA, Cuny H, Hung A, Clark RJ, Brust A, Akondi K, et al. Identifying key amino 47. acid residues that affect α-conotoxin AuIB inhibition of α3β4 nicotinic acetylcholine receptors. Journal of Biological Chemistry. 2013;288(48):34428-42. Suresh A, Hung A. Molecular simulation study of the unbinding of α-conotoxin 48. [ϒ4E]GID at the α7 and α4β2 neuronal nicotinic acetylcholine receptors. Journal of Molecular Graphics and Modelling. 2016;70:109-21. 49. Xu M, Zhu X, Yu J, Yu J, Luo S, Wang X. The crystal structure of Ac-AChBP in complex with α-conotoxin LvIA reveals the mechanism of its selectivity towards different nAChR subtypes. Protein & Cell. 2017:1-11. 50. Zouridakis M, Papakyriakou A, Ivanov IA, Kasheverov IE, Tsetlin V, Tzartos S, et al. Crystal structure of the monomeric extracellular domain of α9 nicotinic receptor subunit in complex with α-conotoxin RgIA: molecular dynamics insights into RgIA binding to α9α10 nicotinic receptors. Frontiers in Pharmacology. 2019;10(474).

pharmacological characterization. Molecular

Catterall WA. Voltage-gated calcium channels. Cold Spring Harbor Perspectives in

Hansen SB, Sulzenbacher G, Huxford T, Marchot P, Taylor P, Bourne Y. Structures of 51. Aplysia AChBP complexes with nicotinic agonists and antagonists reveal distinctive binding interfaces and conformations. The EMBO journal. 2005;24(20):3635-46. Ulens C, Hogg RC, Celie PH, Bertrand D, Tsetlin V, Smit AB, et al. Structural 52. determinants of selective alpha-conotoxin binding to a nicotinic acetylcholine receptor homolog AChBP. Proceedings of the National Academy of Sciences. 2006;103(10):3615-20. 53. Bourne Y, Talley TT, Hansen SB, Palmer T, Marchot P. Crystal structure of a Cbtx-AChBP complex reveals essential interactions between snake α-neurotoxins and nicotinic receptors. EMBO Journal. 2005;24(8):1512-22. 54. Abraham N, Healy M, Ragnarsson L, Brust A, Alewood PF, Lewis RJ. Structural mechanisms for α-conotoxin activity at the human α3β4 nicotinic acetylcholine receptor. Scientific Reports. 2017;7:45466. 55. Moretti M, Zoli M, George AA, Lukas RJ, Pistillo F, Maskos U, et al. The novel α7β2- nicotinic acetylcholine receptor subtype is expressed in mouse and human basal forebrain: Pharmacology. and biochemical 2014;86(3):306-17. 56. Wang Z-J, Deba F, Mohamed TS, Chiara DC, Ramos K, Hamouda AK. Unraveling amino acid residues critical for allosteric potentiation of (α4)3(β2)2-type nicotinic acetylcholine receptor responses. The Journal of Biological Chemistry. 2017;292(24):9988-10001. 57. Mazzaferro S, Benallegue N, Carbone A, Gasparri F, Vijayan R, Biggin PC, et al. Additional acetylcholine (ACh) binding site at α4/α4 interface of (α4β2) 2α4 nicotinic receptor influences agonist sensitivity. The Journal of Biological Chemistry. 2011;286(35):31043-54. 58. Harpsøe K, Ahring PK, Christensen JK, Jensen ML, Peters D, Balle T. Unraveling the high- and low-sensitivity agonist responses of nicotinic acetylcholine receptors. The Journal of Neuroscience. 2011;31(30):10759-66. 59. Indurthi DC, Pera E, Kim H-L, Chu C, McLeod MD, McIntosh JM, et al. Presence of multiple binding sites on α9α10 nAChR receptors alludes to stoichiometric-dependent action of the α-conotoxin, Vc1.1. Biochemical pharmacology. 2014;89(1):131-40. 60. Moroni M, Bermudez I. Stoichiometry and pharmacology of two human α4β2 nicotinic receptor types. Journal of Molecular Neuroscience. 2006;30(1-2):95-6. 61. Yu J, Zhu X, Zhang L, Kudryavtsev D, Kasheverov I, Lei Y, et al. Species specificity of rat and human α7 nicotinic acetylcholine receptors towards different classes of peptide and protein antagonists. Neuropharmacology. 2018;139:226-37. 62. Boffi JC, Marcovich I, Gill-Thind JK, Corradi J, Collins T, Lipovsek MM, et al. Differential contribution of subunit interfaces to α9α10 nicotinic acetylcholine receptor function. Molecular Pharmacology. 2017;91(3):250-62. 63. Sambasivarao SV, Roberts J, Bharadwaj VS, Slingsby JG, Rohleder C, Mallory C, et al. Acetylcholine promotes binding of α-conotoxin MII at α3β2 nicotinic acetylcholine receptors. ChemBioChem. 2014;15(3):413-24. 64. Marais E, Klugbauer N, Hofmann F. Calcium Channel α2δ Subunits—structure and gabapentin binding. Molecular Pharmacology. 2001;59(5):1243-8. 65. Biology. 2011;3(8):a003947-a. 66. Catterall WA, Perez-Reyes E, Snutch TP, Striessnig J. International Union of Pharmacology. XLVIII. nomenclature and structure-function relationships of voltage-gated calcium channels. Pharmacological Reviews. 2005;57(4):411-25.

Simms Brett A, Zamponi Gerald W. Neuronal voltage-gated calcium channels:

Bean BP. Two kinds of calcium channels in canine atrial cells. Differences in kinetics,

in neurological and Zamponi GW. Targeting voltage-gated calcium channels

Gribkoff VK. The role of voltage-gated calcium channels in pain and nociception.

Adams DJ, Berecki G. Mechanisms of conotoxin inhibition of N-type (Cav2.2) calcium

Dolphin AC. Voltage‐gated calcium channels and their auxiliary subunits: physiology

Klugbauer N, Lacinová L, Marais E, Hobom M, Hofmann F. Molecular diversity of the

67. structure, function, and dysfunction. Neuron. 2014;82(1):24-45. 68. selectivity, and pharmacology. The Journal of General Physiology. 1985;86(1):1-30. 69. Mikami A, Imoto K, Tanabe T, Niidome T, Mori Y, Takeshima H, et al. Primary structure and functional expression of the cardiac dihydropyridine-sensitive calcium channel. Nature. 1989;340(6230):230-3. 70. Mori Y, Friedrich T, Kim M-S, Mikami A, Nakai J, Ruth P, et al. Primary structure and functional expression from complementary DNA of a brain calcium channel. Nature. 1991;350(6317):398-402. Varadi G, Lory P, Schultz D, Varadi M, Schwartz A. Acceleration of activation and 71. inactivation by the β subunit of the skeletal muscle calcium channel. Nature. 1991;352(6331):159-62. 72. psychiatric diseases. Nature Reviews Drug Discovery. 2016;15(1):19-34. 73. Newton PM, Orr CJ, Wallace MJ, Kim C, Shin H-S, Messing RO. Deletion of N-type calcium channels alters ethanol reward and reduces ethanol consumption in mice. Journal of Neuroscience. 2004;24(44):9862-9. 74. Saegusa H, Kurihara T, Zong S, Kazuno A, Matsuda Y, Nonaka T, et al. Suppression of inflammatory and neuropathic pain symptoms in mice lacking the N-type Ca2+ channel. The EMBO journal. 2001;20(10):2349-56. 75. Winquist RJ, Pan JQ, Gribkoff VK. Use-dependent blockade of Cav2.2 voltage-gated calcium channels for neuropathic pain. Biochemical Pharmacology. 2005;70(4):489-99. 76. Seminars in Cell & Developmental Biology. 2006;17(5):555-64. 77. de la Calle Gil AB, Peña Vergara I, Cormane Bornacelly MA, Pajuelo Gallego A. Intrathecal ziconotide and morphine for pain relief: a case series of eight patients with refractory cancer pain, including five cases of neuropathic pain. Neurology and Therapy. 2015;4(2):159-68. 78. channels. Biochimica et Biophysica Acta (BBA) - Biomembranes. 2013;1828(7):1619-28. 79. and pathophysiology and pharmacology. The Journal of Physiology. 2016;594(19):5369-90. 80. Witcher D, De Waard M, Sakamoto J, Franzini-Armstrong C, Pragnell M, Kahl S, et al. Subunit identification and reconstitution of the N-type Ca2+ channel complex purified from brain. Science. 1993;261(5120):486-9. 81. calcium channel alpha2delta subunit. Journal of Neuroscience. 1999;19(2):684-91. 82. Gong HC, Hang J, Kohler W, Li L, Su TZ. Tissue-specific expression and gabapentin- binding properties of calcium channel α2δ subunit subtypes. Journal of Membrane Biology. 2001;184(1):35-43. Dissanayake VU, Gee NS, Brown JP, Woodruff GN. Spermine modulation of specific 83. [3H]-gabapentin binding to the detergent-solubilized porcine cerebral cortex alpha 2 delta calcium channel subunit. British Journal of Pharmacology. 1997;120(5):833-40. Li J, Stevens L, Klugbauer N, Wray D. Roles of molecular regions in determining 84. differences between voltage dependence of activation of CaV3.1 and CaV1.2 calcium channels. The Journal of Biological Chemistry. 2004;279(26):26858-67.

Arias JM, Murbartián J, Vitko I, Lee J-H, Perez-Reyes E. Transfer of β subunit regulation

Hildebrand ME, Smith PL, Bladen C, Eduljee C, Xie JY, Chen L, et al. A novel slow- ion channel modulator attenuates neuropathic pain. Pain.

Kim D, Park D, Choi S, Lee S, Sun M, Kim C, et al. Thalamic control of visceral

Buraei Z, Yang J. The β subunit of voltage-gated Ca2+ channels. Physiological Reviews.

Chen Y-h, Li M-h, Zhang Y, He L-l, Yamada Y, Fitzmaurice A, et al. Structural basis of the

85. from high to low voltage-gated Ca2+ channels. FEBS Letters. 2005;579(18):3907-12. 86. Bladen C, McDaniel SW, Gadotti VM, Petrov RR, Berger ND, Diaz P, et al. Characterization of novel cannabinoid based T-type calcium channel blockers with analgesic effects. ACS Chemical Neuroscience. 2015;6(2):277-87. 87. inactivation-specific 2011;152(4):833-43. 88. nociception mediated by T-type Ca 2+ channels. Science. 2003;302(5642):117-9. Yang Y-C, Tai C-H, Pan M-K, Kuo C-C. The T-type calcium channel as a new therapeutic 89. target for Parkinson’s disease. Pflügers Archiv - European Journal of Physiology. 2014;466(4):747-55. 90. Dragicevic E, Schiemann J, Liss B. Dopamine midbrain neurons in health and Parkinson’s disease: emerging roles of voltage-gated calcium channels and ATP-sensitive potassium channels. Neuroscience. 2015;284:798-814. 91. Matsunami M, Kirishi S, Okui T, Kawabata A. Chelating luminal zinc mimics hydrogen sulfide-evoked colonic pain in mice: possible involvement of T-type calcium channels. Neuroscience. 2011;181:257-64. 92. Souza IA, Gandini MA, Zhang F-X, Mitchell WG, Matsumoto J, Lerner J, et al. Pathogenic Cav3.2 channel mutation in a child with primary generalized epilepsy. Molecular Brain. 2019;12(1):86. 93. Wu J, Yan Z, Li Z, Qian X, Lu S, Dong M, et al. Structure of the voltage-gated calcium channel Cav1.1 at 3.6 Å resolution. Nature. 2016;537:191. Bourdin B, Briot J, Tétreault M-P, Sauvé R, Parent L. Negatively charged residues in the 94. first extracellular loop of the L-type CaV1.2 channel anchor the interaction with the Ca(V)α2δ1 auxiliary subunit. The Journal of Biological Chemistry. 2017;292(42):17236-49. Van Petegem F, Clark KA, Chatelain FC, Minor DL, Jr. Structure of a complex between 95. a voltage-gated calcium channel beta-subunit and an alpha-subunit domain. Nature. 2004;429(6992):671-5. 96. 2010;90(4):1461-506. 97. α1–β subunit interaction of voltage-gated Ca2+ channels. Nature. 2004;429(6992):675-80. Kotev M, Pascual R, Almansa C, Guallar V, Soliva R. Pushing the limits of computational 98. structure-based drug design with a cryo-EM structure: the Ca2+ channel α2δ-1 subunit as a test case. Journal of Chemical Information and Modeling. 2018;58(8):1707-15. Korkosh VS, Kiselev AM, Mikhaylov EN, Kostareva AA, Zhorov BS. Atomic mechanisms 99. of timothy syndrome-associated mutations in calcium channel Cav1.2. Frontiers in Physiology. 2019;10:335. 100. Feng T, Kalyaanamoorthy S, Ganesan A, Barakat K. Atomistic modeling and molecular dynamics analysis of human CaV1.2 channel using external electric field and ion pulling simulations. Biochimica et Biophysica Acta (BBA) - General Subjects. 2019;1863(6):1116-26. 101. Geisler S, Schöpf CL, Stanika R, Kalb M, Campiglio M, Repetto D, et al. Presynaptic α2δ- 2 calcium channel subunits regulate postsynaptic GABAA receptor abundance and axonal wiring. The Journal of Neuroscience. 2019;39(14):2581-605.

Zhao Y, Huang G, Wu Q, Wu K, Li R, Lei J, et al. Cryo-EM structures of apo and

Feng T, Kalyaanamoorthy S, Barakat K. L-type calcium channels: structure and

Long SB, Campbell EB, MacKinnon R. Crystal structure of a mammalian voltage-

Shafrir Y, Durell SR, Guy HR. Models of the structure and gating mechanisms of the

Stary A, Shafrir Y, Hering S, Wolschann P, Guy HR. Structural model of the CaV1.2 pore.

Jiang Y, Lee A, Chen J, Ruta V, Cadene M, Chait BT, et al. X-ray structure of a voltage-

Tikhonov DB, Zhorov BS. Structural model for dihydropyridine binding to L-type

Tikhonov DB, Zhorov BS. Modeling P-loops domain of sodium channel: homology with

Stevens M, Peigneur S, Tytgat J. Neurotoxins and their binding areas on voltage-gated Jiang X, Raju PK, D'Avanzo N, Lachance M, Pepin J, Dubeau F, et al. Both gain-of- 102. function and loss-of-function de novo CACNA1A mutations cause severe developmental epileptic encephalopathies in the spectrum of Lennox-Gastaut syndrome. Epilepsia. 2019;60(9):1881-94. 103. Bayle M, Neasta J, Dall'Asta M, Gautheron G, Virsolvy A, Quignard J-F, et al. The ellagitannin metabolite urolithin C is a glucose-dependent regulator of insulin secretion through activation of L-type calcium channels. British Journal of Pharmacology. 2019;176(20):4065-78. 104. antagonist-bound human Cav3.1. Nature. 2019;576(7787):492-7. Tang L, Gamal El-Din TM, Payandeh J, Martinez GQ, Heard TM, Scheuer T, et al. 105. Structural basis for Ca2+ selectivity of a voltage-gated calcium channel. Nature. 2014;505(7481):56-61. 106. Adiban J, Jamali Y, Rafii-Tabar H. Modeling ion permeation through a bacterial voltage- gated calcium channel CaVAb using molecular dynamics simulations. Molecular BioSystems. 2017;13(1):208-14. 107. El-Moselhy TF, Sidhom PA, Esmat EA, El-Mahdy NA. Synthesis, Docking simulation, biological evaluations and 3D-QSAR study of 1,4-dihydropyridines as calcium channel blockers. Chemical and Pharmaceutical Bulletin. 2017;65(10):893-903. 108. functions. Ion Channels in Health and Sickness: IntechOpen; 2018. 109. Walsh CP, Davies A, Butcher AJ, Dolphin AC, Kitmitto A. Three-dimensional structure of CaV3.1: comparison with the cardiac L-type voltage-gated calcium channel monomer architecture. Journal of Biological Chemistry. 2009;284(33):22310-21. 110. dependent shaker family K+ channel. Science. 2005;309(5736):897-903. 111. Bruhova I, Zhorov BS. A homology model of the pore domain of a voltage-gated calcium channel is consistent with available SCAM data. The Journal of general physiology. 2010;135(3):261-74. 112. pore domain of the NaChBac ion channel. Biophysical Journal. 2008;95(8):3650-62. 113. Channels. 2008;2(3):210-5. 114. Yu L, Sun C, Song D, Shen J, Xu N, Gunasekera A, et al. Nuclear magnetic resonance structural studies of a potassium channel−charybdotoxin complex. Biochemistry. 2005;44(48):15834-41. 115. dependent K+ channel. Nature. 2003;423(6935):33-41. 116. Saddala MS, Kandimalla R, Adi PJ, Bhashyam SS, Asupatri UR. Novel 1, 4- dihydropyridines for L-type calcium channel as antagonists for cadmium toxicity. Scientific Reports. 2017;7(1):45211. 117. calcium channels. The Journal of Biological Chemistry. 2009;284(28):19006-17. 118. potassium channels and interaction with ligands. Biophysical Journal. 2005;88(1):184-97. 119. sodium channels. Frontiers in Pharmacology. 2011;2:71-.

Ellinor PT, Zhang J-F, Horne WA, Tsien RW. Structural determinants of the blockade of

Lee CW, Bae C, Lee J, Ryu JH, Kim HH, Kohno T, et al. Solution structure of kurtoxin: a for Cav3 voltage-gated Ca2+ Channels. Biochemistry. selective

120. N-type calcium channels by a peptide neurotoxin. Nature. 1994;372(6503):272-5. 121. Feng Z-P, Doering CJ, Winkfein RJ, Beedle AM, Spafford JD, Zamponi GW. Determinants of Inhibition of Transiently Expressed Voltage-gated Calcium Channels by ω- Conotoxins GVIA and MVIIA. Journal of Biological Chemistry. 2003;278(22):20171-8. 122. Feng Z-P, Hamid J, Doering C, Bosey GM, Snutch TP, Zamponi GW. Residue Gly1326 of the N-type calcium channel α1B subunit controls reversibility of ω-conotoxin GVIA and MVIIA block. Journal of Biological Chemistry. 2001;276(19):15728-35. 123. Wang F, Yan Z, Liu Z, Wang S, Wu Q, Yu S, et al. Molecular basis of toxicity of N-type calcium channel inhibitor MVIIA. Neuropharmacology. 2016;101:137-45. Zamponi GW, Striessnig J, Koschak A, Dolphin AC. The physiology, pathology, and 124. pharmacology of voltage-gated calcium channels and their future therapeutic potential. Pharmacological Reviews. 2015;67(4):821-70. 125. Mintz IM, Venema VJ, Swiderek KM, Lee TD, Bean BP, Adams ME. P-type calcium channels blocked by the spider toxin ω-Aga-IVA. Nature. 1992;355(6363):827-9. 126. Mould J, Yasuda T, Schroeder CI, Beedle AM, Doering CJ, Zamponi GW, et al. The α2δ Auxiliary subunit reduces affinity of ω-conotoxins for recombinant N-type (Cav2.2) calcium channels. The Journal of Biological Chemistry. 2004;279(33):34705-14. 127. Wang D, Himaya SWA, Giacomotto J, Hasan MM, Cardoso FC, Ragnarsson L, et al. Characterisation of δ-conotoxin TxVIA as a mammalian T-Type calcium channel modulator. Marine Drugs. 2020;18(7):343. 128. gating modifier 2012;51(9):1862-73. 129. Bladen C, Hamid J, Souza IA, Zamponi GW. Block of T-type calcium channels by protoxins I and II. Molecular Brain. 2014;7:36-. 130. Middleton RE, Warren VA, Kraus RL, Hwang JC, Liu CJ, Dai G, et al. Two tarantula peptides inhibit activation of multiple sodium channels. Biochemistry. 2002;41(50):14734-47. 131. Gao B, Peng C, Yang J, Yi Y, Zhang J, Shi Q. Cone snails: a big store of conotoxins for novel drug discovery. Toxins. 2017;9(12):397. 132. Olivera BM. Conus venom peptides: reflections from the biology of clades and species. Annual Review of Ecology and Systematics. 2002;33(1):25-47. 133. Holmes D. Conotoxins: how a deadly snail could help ease pain. The Lancet Neurology. 2014;13(9):867-8. 134. Kaas Q, Westermann J-C, Craik DJ. Conopeptide characterization and classifications: an analysis using ConoServer. Toxicon. 2010;55(8):1491-509. 135. Lobo-Ruiz A, Tulla-Puche J. 2 - Synthetic approaches of naturally and rationally designed peptides and peptidomimetics. In: Koutsopoulos S, editor. Peptide Applications in Biomedicine, Biotechnology and Bioengineering: Woodhead Publishing; 2018. p. 23-49. 136. Kaas Q, Westermann J-C, Craik D. Conopeptide characterization and classifications: an analysis using ConoServer. Toxicon : official journal of the International Society on Toxinology. 2010;55 8:1491-509. 137. Miljanich GP. Ziconotide: neuronal calcium channel blocker for treating severe chronic pain. Current Medicinal Chemistry. 2004;11(23):3029-40. 138. Wakamatsu K, Kohda D, Hatanaka H, Lancelin JM, Ishida Y, Oya M, et al. Structure- activity relationships of mu-conotoxin GIIIA: structure determination of active and inactive

interactions between neuronal α7 ImI. The Journal of Biological Chemistry. sodium channel blocker peptides by NMR and simulated annealing calculations. Biochemistry. 1992;31(50):12577-84. 139. Dudley SC, Todt H, Lipkind G, Fozzard HA. A mu-conotoxin-insensitive Na+ channel mutant: possible localization of a binding site at the outer vestibule. Biophysical Journal. 1995;69(5):1657-65. 140. Catterall WA, Cestèle S, Yarov-Yarovoy V, Yu FH, Konoki K, Scheuer T. Voltage-gated ion channels and gating modifier toxins. Toxicon. 2007;49(2):124-41. 141. Banerjee J, Yongye AB, Chang YP, Gyanda R, Medina?Franco JL, Armishaw CJ. Design and synthesis of α-conotoxin GID analogues as selective α4β2 nicotinic acetylcholine receptor antagonists. Peptide Science. 2014;102(1):78-87. 142. Molgó J, Aráoz R, Benoit E, Iorga BI. Physical and virtual screening methods for marine toxins and drug discovery targeting nicotinic acetylcholine receptors. Expert Opinion on Drug Discovery. 2013;8(10):1203-23. 143. Nicke A, Loughnan ML, Millard EL, Alewood PF, Adams DJ, Daly NL, et al. Isolation, structure, and activity of GID, a novel α4/7-conotoxin with an extended N-terminal sequence. The Journal of Biological Chemistry. 2003;278(5):3137-44. 144. McIntosh JM, Azam L, Staheli S, Dowell C, Lindstrom JM, Kuryatov A, et al. Analogs of α-conotoxin MII are selective for α6-containing nicotinic acetylcholine receptors. Molecular Pharmacology. 2004;65(4):944-52. 145. Gray WR, Luque A, Olivera BM, Barrett J, Cruz LJ. Peptide toxins from Conus geographus venom. The Journal of Biological Chemistry. 1981;256(10):4734-40. 146. Luo S, Zhangsun D, Schroeder CI, Zhu X, Hu Y, Wu Y, et al. A novel α4/7-conotoxin LvIA from Conus lividus that selectively blocks α3β2 vs. α6/α3β2β3 nicotinic acetylcholine receptors. FASEB J. 2014;28(4):1842-53. El Hamdaoui Y, Wu X, Clark RJ, Giribaldi J, Anangi R, Craik DJ, et al. Periplasmic 147. expression of 4/7 α-conotoxin TxIA analogs in E. coli favors ribbon isomer formation – suggestion of a binding mode at the α7 nAChR. Frontiers in Pharmacology. 2019;10(577). 148. Nicke A, Wonnacott S, Lewis RJ. α-Conotoxins as tools for the elucidation of structure and function of neuronal nicotinic acetylcholine receptor subtypes. European Journal of Biochemistry. 2004;271(12):2305-19. 149. Wu X, Tae H-S, Huang Y-H, Adams DJ, Craik DJ, Kaas Q. Stoichiometry dependent inhibition of rat α3β4 nicotinic acetylcholine receptor by the ribbon isomer of α-conotoxin AuIB. Biochemical Pharmacology. 2018;155:288-97. 150. Xu P, Kaas Q, Wu Y, Zhu X, Li X, Harvey PJ, et al. Structure and activity studies of disulphide-deficient analogues of αO-conotoxin GeXIVA. Journal of Medicinal Chemistry. 2020. 151. Johnson DS, Martinez J, Elgoyhen AB, Heinemann SF, McIntosh JM. alpha-Conotoxin ImI exhibits subtype-specific nicotinic acetylcholine receptor blockade: preferential inhibition of homomeric alpha 7 and alpha 9 receptors. Molecular Pharmacology. 1995;48(2):194-9. 152. Ellison M, Gao F, Wang H-L, Sine SM, McIntosh JM, Olivera BM. α-Conotoxins ImI and ImII target distinct regions of the human α7 nicotinic acetylcholine receptor and distinguish human nicotinic receptor subtypes. Biochemistry. 2004;43(51):16019-26. 153. Quiram PA, Jones JJ, Sine SM. Pairwise acetylcholine receptors and α-conotoxin 1999;274(28):19517-24.

Ellison M, McIntosh JM, Olivera BM. α-Conotoxins ImI and ImII: similar α7 nicotinic 154. receptor antagonists act at different sites. The Journal of Biological Chemistry. 2003;278(2):757-64. 155. Dutertre S, Nicke A, Tyndall JDA, Lewis RJ. Determination of α-conotoxin binding modes on neuronal nicotinic acetylcholine receptors. Journal of Molecular Recognition. 2004;17(4):339-47. 156. Luo S, Nguyen TA, Cartier GE, Olivera BM, Yoshikami D, McIntosh JM. Single-residue alteration in α-conotoxin pnia switches its nAChR subtype selectivity. Biochemistry. 1999;38(44):14542-8. 157. Everhart D, Reiller E, Mirzoian A, McIntosh JM, Malhotra A, Luetje CW. Identification of residues that confer α-conotoxin-PnIA sensitivity on the α3 subunit of neuronal nicotinic acetylcholine receptors. Journal of Pharmacology and Experimental Therapeutics. 2003;306(2):664-70. 158. Hogg RC, Miranda LP, Craik DJ, Lewis RJ, Alewood PF, Adams DJ. Single amino acid substitutions in α-conotoxin pnia shift selectivity for subtypes of the mammalian neuronal nicotinic acetylcholine receptor. The Journal of Biological Chemistry. 1999;274(51):36559-64. 159. Cuny H, Yu R, Tae H-S, Kompella SN, Adams DJ. α-Conotoxins active at α3-containing nicotinic acetylcholine receptors and their molecular determinants for selective inhibition. British Journal of Pharmacology. 2017;175(11):1855-68. 160. Harvey SC, Mcintosh JM, Cartier GE, Maddox FN, Luetje CW. Determinants of specificity for α-conotoxin MII on α3β2 neuronal nicotinic receptors. Molecular Pharmacology. 1997;51(2):336-42. 161. Cartier GE, Yoshikami D, Gray WR, Luo S, Olivera BM, McIntosh JM. A new α-conotoxin which targets α3β2 nicotinic acetylcholine receptors. Journal of Biological Chemistry. 1996;271(13):7522-8. 162. Ellison M, Haberlandt C, Gomez-Casati ME, Watkins M, Elgoyhen AB, McIntosh JM, et al. α-RgIA: A novel conotoxin that specifically and potently blocks the α9α10 nAChR. Biochemistry. 2006;45(5):1511-7. 163. Luo S, Kulak JM, Cartier GE, Jacobsen RB, Yoshikami D, Olivera BM, et al. Alpha- conotoxin AuIB selectively blocks α3β4 nicotinic acetylcholine receptors and nicotine-evoked norepinephrine release. Journal of Neuroscience. 1998;18(21):8571-9. 164. Vincler M, Wittenauer S, Parker R, Ellison M, Olivera BM, McIntosh JM. Molecular mechanism for analgesia involving specific antagonism of α9α10 nicotinic acetylcholine receptors. Proceedings of the National Academy of Sciences. 2006;103(47):17880-4. 165. Clark RJ, Fischer H, Nevin ST, Adams DJ, Craik DJ. The Synthesis, structural characterization, and receptor specificity of the α-conotoxin Vc1.1. Journal of Biological Chemistry. 2006;281(32):23254-63. 166. McIntosh JM, Plazas PV, Watkins M, Gomez-Casati ME, Olivera BM, Elgoyhen AB. A novel α-conotoxin, PeIA, cloned from conus pergrandis, discriminates between rat α9α10 and α7 nicotinic cholinergic receptors. Journal of Biological Chemistry. 2005;280(34):30107-12. 167. McIntosh JM, Dowell C, Watkins M, Garrett JE, Yoshikami D, Olivera BM. α-Conotoxin GIC from Conus geographus, a novel peptide antagonist of nicotinic acetylcholine receptors. Journal of Biological Chemistry. 2002;277(37):33610-5. 168. Talley TT, Olivera BM, Han K-H, Christensen SB, Dowell C, Tsigelny I, et al. α-Conotoxin OmIA is a potent ligand for the acetylcholine-binding protein as well as α3β2 and α7 nicotinic acetylcholine receptors. Journal of Biological Chemistry. 2006;281(34):24678-86.

receptors and α-conotoxin PnIB.

receptors and the

169. Chi S-W, Kim D-H, Olivera BM, McIntosh JM, Han K-H. Solution conformation of a neuronal nicotinic acetylcholine receptor antagonist α-conotoxin OmIA that discriminates α3 vs. α6 nAChR subtypes. Biochemical and Biophysical Research Communications. 2006;345(1):248-54. 170. Loughnan ML, Nicke A, Jones A, Adams DJ, Alewood PF, Lewis RJ. Chemical and functional identification and characterization of novel sulfated α-conotoxins from the cone snail Conus anemone. Journal of Medicinal Chemistry. 2004;47(5):1234-41. 171. Whiteaker P, Christensen S, Yoshikami D, Dowell C, Watkins M, Gulyas J, et al. Discovery, synthesis, and structure activity of a highly selective α7 nicotinic acetylcholine receptor antagonist. Biochemistry. 2007;46(22):6628-38. 172. Millard EL, Nevin ST, Loughnan ML, Nicke A, Clark RJ, Alewood PF, et al. Inhibition of neuronal nicotinic acetylcholine receptor subtypes by α-conotoxin GID and analogues. Journal of Biological Chemistry. 2009;284(8):4944-51. 173. Hone AJ, Fisher F, Christensen S, Gajewiak J, Larkin D, Whiteaker P, et al. PeIA-5466: a novel peptide antagonist containing non-natural amino acids that selectively targets α3β2 nicotinic acetylcholine receptors. Journal of Medicinal Chemistry. 2019;62(13):6262-75. 174. Hone AJ, Ruiz M, Scadden Ml, Christensen S, Gajewiak J, Azam L, et al. Positional scanning mutagenesis of α-conotoxin PeIA identifies critical residues that confer potency and selectivity for α6/α3β2β3 and α3β2 nicotinic acetylcholine receptors. Journal of Biological Chemistry. 2013;288(35):25428-39. 175. Chi S-W, Kim D-H, Olivera BM, McIntosh JM, Han K-H. Solution conformation of alpha- conotoxin GIC, a novel potent antagonist of alpha3beta2 nicotinic acetylcholine receptors. Biochemical Journal. 2004;380(Pt 2):347-52. 176. Quiram PA, McIntosh JM, Sine SM. Pairwise interactions between neuronal α7 Journal of Biological Chemistry. acetylcholine 2000;275(7):4889-96. 177. Hogg RC, Hopping G, Alewood PF, Adams DJ, Bertrand D. α-Conotoxins PnIA and [A10L]PnIA stabilize different states of the α7-L247T nicotinic acetylcholine receptor. The Journal of Biological Chemistry. 2003;278(29):26908-14. 178. Kasheverov IE, Chugunov AO, Kudryavtsev DS, Ivanov IA, Zhmak MN, Shelukhina IV, et al. High-affinity α-conotoxin PnIA analogs designed on the basis of the protein surface topography method. Scientific Reports. 2016;6:36848. 179. Grau V, Richter K, Hone AJ, McIntosh JM. Conopeptides [V11L;V16D]ArIB and RgIA4: powerful tools for the identification of novel nicotinic acetylcholine receptors in monocytes. Frontiers in Pharmacology. 2019;9(1499). 180. Grant MA, Shanmugasundaram K, Rigby AC. Conotoxin therapeutics: a pipeline for success? Expert Opinion on Drug Discovery. 2007;2(4):453-68. 181. McIntosh JM, Absalom N, Chebib M, Elgoyhen AB, Vincler M. Alpha9 nicotinic acetylcholine treatment of pain. Biochemical Pharmacology. 2009;78(7):693-702. 182. Yu R, Kompella SN, Adams DJ, Craik DJ, Kaas Q. Determination of the α-conotoxin Vc1.1 binding site on the α9α10 nicotinic acetylcholine receptor. Journal of Medicinal Chemistry. 2013;56(9):3557-67. 183. Castro J, Grundy L, Deiteren A, Harrington AM, O'Donnell T, Maddern J, et al. Cyclic analogues of α-conotoxin Vc1.1 inhibit colonic nociceptors and provide analgesia in a mouse model of chronic abdominal pain. British Journal of Pharmacology. 2018;175(12):2384-98.

Lu B, Yu F, Wang J, Zhao S, Zhao D, Dai Q, et al. New O-superfamily conotoxins from Island. Chinese Science Bulletin. inhabited near Chinese Hainan

Lee S. Pharmacological inhibition of voltage-gated Ca2+ channels for chronic pain relief.

function of voltage-gated calcium

184. Clark RJ, Jensen J, Nevin ST, Callaghan BP, Adams DJ, Craik DJ. The engineering of an orally active conotoxin for the treatment of neuropathic pain. Angewandte Chemie International Edition. 2010;49(37):6545-8. Julie LD, David JC. Alpha-conotoxins: nicotinic acetylcholine receptor antagonists as 185. pharmacological tools and potential drug leads. Current Medicinal Chemistry. 2001;8(4):327- 44. 186. Peigneur S, Devi P, Seldeslachts A, Ravichandran S, Quinton L, Tytgat J. Structure- function elucidation of a new α-conotoxin, MiIIA, from Conus milneedwardsi. Marine Drugs. 2019;17(9):535. 187. Luo S, Zhangsun D, Harvey PJ, Kaas Q, Wu Y, Zhu X, et al. Cloning, synthesis, and characterization of αO-conotoxin GeXIVA, a potent α9α10 nicotinic acetylcholine receptor antagonist. Proceedings of the National Academy of Sciences. 2015;112(30):E4026-E35. 188. Gilchrist J, Olivera BM, Bosmans F. Animal toxins influence voltage-gated sodium channel function. In: Ruben PC, editor. Voltage Gated Sodium Channels. Berlin, Heidelberg: Springer Berlin Heidelberg; 2014. p. 203-29. 189. Fainzilber M, Nakamura T, Gaathon A, Lodder JC, Kits KS, Burlingame AL, et al. A new cysteine framework in sodium channel blocking conotoxins. Biochemistry. 1995;34(27):8649- 56. 190. Wallace MS, Charapata SG, Fisher R, Byas-Smith M, Staats PS, Mayo M, et al. Intrathecal ziconotide in the treatment of chronic nonmalignant pain: a randomized, double- blind, placebo-controlled clinical trial. Neuromodulation: Technology at the Neural Interface. 2006;9(2):75-86. 191. Nielsen KJ, Adams DA, Alewood PF, Lewis RJ, Thomas L, Schroeder T, et al. Effects of chirality at Tyr13 on the structure−activity relationships of ω-conotoxins from Conus magus. Biochemistry. 1999;38(21):6741-51. 192. Conus striatus 2000;45(5):432-5. 193. Dong M, Wang F, Yan Z, Yu S, Wei J, Wu Q, et al. Structure–activity analysis of N-Type calcium channel inhibitor SO-3. Biochemistry. 2018;57(44):6349-55. 194. Dai Q, Liu F, Zhou Y, Lu B, Yu F, Huang P. The Synthesis of SO-3, a conopeptide with high analgesic activity derived from Conus striatus. Journal of Natural Products. 2003;66(9):1276-9. 195. Current Neuropharmacology. 2013;11(6):606-20. 196. Ramírez D, Gonzalez W, Fissore R, Carvacho I. Conotoxins as Tools to understand the physiological (CaV) channels. Marine Drugs. 2017;15(10):313. 197. Sousa SR, McArthur JR, Brust A, Bhola RF, Rosengren KJ, Ragnarsson L, et al. Novel analgesic ω-conotoxins from the vermivorous cone snail Conus moncuri provide new insights into the evolution of conopeptides. Scientific Reports. 2018;8(1):13397. 198. Kim JI, Takahashi M, Ogura A, Kohno T, Kudo Y, Sato K. Hydroxyl group of Tyr13 is essential for the activity of omega-conotoxin GVIA, a peptide toxin for N-type calcium channel. Journal of Biological Chemistry. 1994;269(39):23876-8. 199. Adams DJ, Smith AB, Schroeder CI, Yasuda T, Lewis RJ. ω-Conotoxin CVID Inhibits a Pharmacologically Distinct Voltage-sensitive Calcium Channel Associated with Transmitter

Yarotskyy V, Elmslie KS. Omega-conotoxin GVIA alters gating charge movement of N-

ion channels. Missouri Medicine. Enkvetchakul D. Genetic disorders of

Lewis S. Metabolite targets sodium channels in diabetic pain. Nature Reviews

Yang M, Zhou M. μ-conotoxin TsIIIA, a peptide inhibitor of human voltage-gated

Release from Preganglionic Nerve Terminals. The Journal of Biological Chemistry. 2003;278(6):4057-62. 200. type (CaV2.2) calcium channels. Journal of Neurophysiology. 2009;101(1):332-40. 201. Nielsen KJ, Adams D, Thomas L, Bond T, Alewood PF, Craik DJ, et al. Structure-activity relationships of ω-conotoxins MVIIA, MVIIC and 14 loop splice hybrids at N and P/Q-type calcium channels11Edited by P. E. Wright. Journal of Molecular Biology. 1999;289(5):1405-21. Shon K-J, Olivera BM, Watkins M, Jacobsen RB, Gray WR, Floresca CZ, et al. μ- 202. Conotoxin PIIIA, a new peptide for discriminating among tetrodotoxin-sensitive Na channel subtypes. The Journal of Neuroscience. 1998;18(12):4473-81. 203. Nielsen KJ, Watson M, Adams DJ, Hammarström AK, Gage PW, Hill JM, et al. Solution structure of μ-conotoxin PIIIA, a preferential inhibitor of persistent tetrodotoxin-sensitive sodium channels. Journal of Biological Chemistry. 2002;277(30):27247-55. 204. Kimball IH, Nguyen PT, Olivera BM, Sack JT, Yarov-Yarovoy V. Molecular determinants of μ-conotoxin KIIIA interaction with the voltage-gated sodium channel Nav1.7. bioRxiv. 2019:654889. 205. Cardoso FC, Lewis RJ. Sodium channels and pain: from toxins to therapies. British Journal of Pharmacology. 2018;175(12):2138-57. 206. 2010;107(4):270-5. 207. Neuroscience. 2012;13(7):448-9. 208. Finol-Urdaneta RK, McArthur JR, Korkosh VS, Huang S, McMaster D, Glavica R, et al. Extremely potent block of bacterial voltage-gated sodium channels by µ-conotoxin PIIIA. Marine Drugs. 2019;17(9):510. 209. Tran HNT, Tran P, Deuis JR, Agwa AJ, Zhang AH, Vetter I, et al. Enzymatic ligation of a pore blocker toxin and a gating modifier toxin: creating double-knotted peptides with improved sodium channel NaV1.7 inhibition. Bioconjugate Chemistry. 2020;31(1):64-73. 210. McArthur JR, Singh G, McMaster D, Winkfein R, Tieleman DP, French RJ. Interactions of Key charged residues contributing to selective block of neuronal sodium channels by μ- conotoxin KIIIA. Molecular Pharmacology. 2011;80(4):573-84. 211. Van Der Haegen A, Peigneur S, Tytgat J. Importance of position 8 in μ-conotoxin KIIIA for voltage-gated sodium channel selectivity. The FEBS Journal. 2011;278(18):3408-18. 212. West PJ, Bulaj G, Garrett JE, Olivera BM, Yoshikami D. μ-Conotoxin SmIIIA, a Potent inhibitor of tetrodotoxin-resistant sodium channels in amphibian sympathetic and sensory neurons. Biochemistry. 2002;41(51):15388-93. 213. Zhang M-M, Fiedler B, Green BR, Catlin P, Watkins M, Garrett JE, et al. Structural and Functional diversities among μ-conotoxins targeting TTX-resistant sodium channels. Biochemistry. 2006;45(11):3723-32. 214. Walewska A, Skalicky JJ, Davis DR, Zhang M-M, Lopez-Vera E, Watkins M, et al. NMR- based mapping of disulphide bridges in cysteine-rich peptides: application to the μ-conotoxin SxIIIA. Journal of the American Chemical Society. 2008;130(43):14280-6. 215. sodium channel hNav1.8. Toxicon. 2020;186:29-34. Leipold E, Ullrich F, Thiele M, Tietze AA, Terlau H, Imhof D, et al. Subtype-specific block 216. of voltage-gated K+ channels by μ-conopeptides. Biochemical and Biophysical Research Communications. 2017;482(4):1135-40.

Layer RT, McIntosh JM. Conotoxins: Therapeutic potential and application. Marine

Lusher SJ, McGuire R, van Schaik RC, Nicholson CD, de Vlieg J. Data-driven medicinal

Sliwoski G, Kothiwale S, Meiler J, Lowe EW, Jr. Computational methods in drug

217. Benatar M. Neurological potassium channelopathies. QJM: An International Journal of Medicine. 2000;93(12):787-97. 218. Pennington MW, Czerwinski A, Norton RS. Peptide therapeutics from venom: current status and potential. Bioorganic & Medicinal Chemistry. 2018;26(10):2738-58. 219. Bjørn-Yoshimoto WE, Ramiro IBL, Yandell M, McIntosh JM, Olivera BM, Ellgaard L, et al. Curses or cures: a review of the numerous benefits versus the biosecurity concerns of conotoxin research. Biomedicines. 2020;8(8):235. 220. Drugs. 2006;4(3):119-42. 221. Harvey AL. Toxins and drug discovery. Toxicon. 2014;92:193-200. 222. Nevin ST, Clark RJ, Klimis H, Christie MJ, Craik DJ, Adams DJ. Are α9α10 nicotinic acetylcholine receptors a pain target for α-conotoxins? Molecular Pharmacology. 2007;72(6):1406-10. 223. Azam L, McIntosh JM. Molecular basis for the differential sensitivity of rat and human α9α10 nAChRs to α-conotoxin RgIA. Journal of Neurochemistry. 2012;122(6):1137-44. 224. Scott DA, Wright CE, Angus JA. Actions of intrathecal ω-conotoxins CVID, GVIA, MVIIA, and morphine in acute and neuropathic pain in the rat. European Journal of Pharmacology. 2002;451(3):279-86. 225. chemistry in the era of big data. Drug Discovery Today. 2014;19(7):859-68. 226. Ou-Yang S-s, Lu J-y, Kong X-q, Liang Z-j, Luo C, Jiang H. Computational drug discovery. Acta Pharmacologica Sinica. 2012;33(9):1131-40. 227. discovery. Pharmacological Reviews. 2013;66(1):334-95. 228. Romano JD, Tatonetti NP. Informatics and computational methods in natural product drug discovery: a review and perspectives. Frontiers in Genetics. 2019;10(368). 229. Wang NX, Watson GB, Loso MR, Sparks TC. Molecular modeling of sulfoxaflor and neonicotinoid binding in insect nicotinic acetylcholine receptors: impact of the Myzus β1 R81T mutation. Pest Management Science. 2016;72(8):1467-74. 230. Trott O, Olson AJ. AutoDock Vina: improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading. Journal of Computational Chemistry. 2010;31(2):455-61. 231. Leffler AE, Kuryatov A, Zebroski HA, Powell SR, Filipenko P, Hussein AK, et al. Discovery of peptide ligands through docking and virtual screening at nicotinic acetylcholine receptor homology models. Proceedings of the National Academy of Sciences. 2017;114(38):E8100-E9. 232. Sterling T, Irwin JJ. ZINC 15 – Ligand Discovery for Everyone. Journal of Chemical Information and Modeling. 2015;55(11):2324-37. Yu R, Craik DJ, Kaas Q. Blockade of neuronal α7-nAChR by α-conotoxin ImI explained 233. by computational scanning and energy calculations. PLoS computational biology. 2011;7(3):e1002011-e. 234. Tabassum N, Yu R, Jiang T. Computational determination of the binding mode of α- conotoxin to nicotinic acetylcholine receptor. Journal of Ocean University of China. 2016;15(6):1027-33. 235. Mansbach RA, Travers T, McMahon BH, Fair JM, Gnanakaran S. Snails in silico: a review of computational studies on the conopeptides. Marine Drugs. 2019 2019/03//; 17(3). Available from: https://doi.org/10.3390/md17030145.

Furini S, Domene C. Ion-triggered selectivity in bacterial sodium channels. Proceedings

236. Adeshina YO, Deeds EJ, Karanicolas J. Machine learning classification can reduce false positives in structure-based virtual screening. Proceedings of the National Academy of Sciences. 2020;117(31):18477. 237. Qin Z, Wu L, Sun H, Huo S, Ma T, Lim E, et al. Artificial intelligence method to design and fold alpha-helical structural proteins from the primary amino acid sequence. Extreme Mechanics Letters. 2020;36:100652. 238. Christensen SB, Hone AJ, Roux I, Kniazeff J, Pin J-P, Upert G, et al. RgIA4 potently blocks mouse α9α10 nAChRs and provides long lasting protection against oxaliplatin-induced cold allodynia. Frontiers in Cellular Neuroscience. 2017;11(219). 239. Chen F, Huang W, Jiang T, Yu R. Determination of the μ-conotoxin PIIIA specificity against voltage-gated sodium channels from binding energy calculations. Marine Drugs. 2018;16(5):153. 240. of the National Academy of Sciences of the United States of America. 2018;115(21):5450-5. Inserra MC, Kompella SN, Vetter I, Brust A, Daly NL, Cuny H, et al. Isolation and 241. characterization of α-conotoxin LsIA with potent activity at nicotinic acetylcholine receptors. Biochemical Pharmacology. 2013;86(6):791-9. 242. Rose KE, Lunardi N, Boscolo A, Dong X, Erisir A, Jevtovic-Todorovic V, et al. Immunohistological demonstration of CaV3.2 T-type voltage-gated calcium channel expression in soma of dorsal root ganglion neurons and peripheral axons of rat and mouse. Neuroscience. 2013;250:263-74. 243. Tomita S, Sekiguchi F, Deguchi T, Miyazaki T, Ikeda Y, Tsubota M, et al. Critical role of Cav3.2 T-type calcium channels in the peripheral neuropathy induced by bortezomib, a proteasome-inhibiting chemotherapeutic agent, in mice. Toxicology. 2018. 244. Mahdavi S, Kuyucak S. Systematic study of binding of µ-conotoxins to the sodium Channel NaV1.4. Toxins. 2014;6(12):3454-70.

: Methodology

Homology modelling and molecular dynamics (MD) simulations are widely implemented in

drug design. To date, some methods, particularly for homology modelling, have been

developed together with user-friendly interfaces and conducted via web servers, which

enables their use beyond experts in this field of study. However, understanding the principles

of these computational techniques can facilitate one to address the problems which occur

during construction of comparative models and MD simulation runs. For example, collapse of

the initial homology modelled protein structure in the course of MD simulations may require

one to assess simulation conditions and parameters to ensure system stability (e.g.

decreasing the time step in equilibration stages). Thus, the fundamental theory of these

computational approaches is presented in this chapter for a better conception and

appreciation of the methods, including their advantages and drawbacks.

2.1 Homology Modelling

2.1.1 Background

The focus of this thesis is on the interactions between conotoxins and ion channel proteins,

made possible by the availability of experimental structures of both types of biomolecules.

Although an increased number of 3-D structures of conotoxins, extracted from marine cone

snails, have been determined through NMR or crystallization methods, the models of certain

ligand-gated and voltage-gated ion channels are still in deficit, such as α7 nAChR and T-type

calcium channel, Cav3.2 [1 , 2], whereas a large amount of protein sequence data is available

[3]. Due to their important involvement in neurological diseases and disorders, homology

modelling methods are widely used in silico via prediction of uncharacterized ion channels for

a better understanding of their physiological and pharmacological properties. The main

principle of homology modelling is based on the fact that the structures of the protein are

much more conserved than protein sequences during evolution [4]; and there is estimated to

be up to ~10,000 protein folds [5 , 6 , 7].

Searching for a template of the protein sequence of interest, selecting the template and

construction of the model are the most significant factors involved in homology modelling.

Currently available modelling approaches are the Protein Model Portal, Modeller/ModWeb,

SWISS-MODEL [8], Phyre2 [6], PRIMO [9] and so on. The general procedure of most packages

involves three main steps. First, the protein sequence of target with an unknown structure is

obtained and compared with the repository of protein sequences in the database via

sequence alignment [6]. Then, one or several templates with high identity will be selected

through a particular ranking algorithm, the conserved domains of which identified in the

previous procedure will be utilized for constructing a rough backbone of the protein in

question [10]. Subsequently, the loops and side-chains will be added for an integrated

homology model [6]. Each step will be explained in detail in the following parts.

2.1.2 Searching for Sequence Similarity

Homology modelling approaches require a template as a scaffold that shares high sequence

similarity with the sequence of the targeted protein. This template structure plays a critical

role in homology modelling, as well as in the following MD simulation for accurately

mimicking the protein system. Nevertheless, more and more experimental structures of

members in protein families have been published and deposited in the Protein Data Bank

(PDB) [11] and it is, therefore, feasible to rationally build a model for an increasing number of

structurally uncharacterized proteins of interest.

To date, there are generally two groups of sequence alignment methods, namely pairwise

(PSA) and multiple sequence alignment (MSA). When sequence identity is high (>40%), both

sequence alignment methods can reliably provide homologs. Pairwise sequence alignment

approaches, such as BLAST - Basic Local Alignment Search Tool [12], FASTA [13] and SSEARCH

- Smith-Waterman Search, can identify the important homologs due to their various and

precise algorithms for calculation of statistical significance in sequence similarities [7 , 13]. On

the contrary, multiple sequence alignment approaches can provide more homologs and

identify conserved domains within protein families, which can contribute to inferring the

functional properties of the sequence for the query protein [7]. Commonly applied MSA

methods include Position-Specific Iterated (PSI) BLAST [12] and Hidden Markov Model

(HMM) based method, HMMER3. Furthermore, multiple sequence alignment methods may

favor the determination of homologs even with relatively weak sequence similarity [12].

2.1.3 Model Building

After sequence alignment and choosing the optimal homolog(s), the 3-D structure of the

query sequence is constructed. The most widely and first used model building programs were

designed based on rigid-body assembly [14], including SWISS-MODEL [15], Builder, 3D-

JIGSAW [16] and NEST [17]. The main principle involved in these methods is the construction

of the frame of the targeted protein based on the aligned Cα atoms. Initially, the average

structure of the template is generated. The conserved residues of which determined

previously will be subsequently manipulated for depicting the rigid body of the query protein

by incorporation of the aligned Cα atoms with those of the template [18]. The non-conserved

domain will be constructed via superimposing the model onto the template. For the sake of

building loops and side chains, different methods are applied differently [14]. Other methods

for construction of comparative models are segment matching and modelling by spatial

restraints [19 , 20] used in SegMod/DNCAD [21] and Modeller, respectively [22].

2.1.4 Model Evaluation

It is essential to ensure the accuracy of the model via model evaluation methods, especially

when similarity between the query and template sequence is less than 30% [14]. A number

of methods, in essence, evaluate the 3-D profile of the protein model by assessing the

compatibility of the model with its amino acid sequence, such as VERIFY3D [23], rather than

relying on the R-factor. Other evaluation programs, such as PROCHECK [24], verify the model

structure through stereochemistry parameters calculated from coordinates of protein

conformation [25], embracing global (hydrogen bond energies, ,  and 1 torsion angles)

and local parameters (such as proline  angles, 3 torsion angles and disulphide bond lengths)

[26].

The quality of the theoretical model of the protein can also be evaluated using the Protein

Structure Analysis (ProSA) program, or Qmean scoring function [27] using z-score, which

includes all the experimentally available conformations of the protein in the PDB [28]. The

quality of the homologous model therefore can be determined via comparing its z-score with

those of the proteins in the same family.

2.1.5 Examples of Application

• Modeller

The homology model of extracellular domain (ECD) of human α7 and (α3)2(β2)3 nAChRs was

constructed using Modeller9v6 [22] in Chapters 3 and 4, respectively. The amino acid

sequence of α7 (UniProtKB: P36544), α3 (UniProtKB: P04757) and β2 (UniProtKB: P12390)

nAChR subunits at ECD was obtained via UniProtKB [29] (www.uniprot.org), while the

sequence of α-conotoxin LsIA was obtained from PDB (PDB ID: 5T90). The co-crystalized

conotoxin PnIA variant (PDB ID: 2BR8) [30] was selected as the template for both nAChRs.

structure of Aplysia californica acetylcholine binding protein (Ac-AChBP) bound by α-

In general, the sequence of the targeted neuronal nAChR (α7 and α3β2 nAChRs) with α-

conotoxin LsIA, was aligned with that of the template (2BR8 in this study) respectively using

ClustalW of Clustral Omega (Figure 2.1, 2.2). The sequence alignment outcomes will be

subsequently used to build the 3-D model of α7 and α3β2 nAChRs bound by LsIA, respectively.

For LsIA with C-terminal carboxylation (LsIA#), the modification at the C-terminus of LsIA was

deployed using Deep View [31].

Figure 2.1. The sequence alignment outcome of human (h) α7 nAChR subunit and Ac-AChBP (PDB ID: 2BR8) at the extracellular domain using ClustalW.

Figure 2.2. The sequence alignment outcome of rat (r) α3 and β2 nAChR subunits and Ac-AChBP (PDB ID: 2BR8) at the extracellular domain using ClustalW.

• SWISS-MODEL

In Chapter 5, the automated web server SWISS-MODEL was used to construct a homology

model of the human Cav3.2 channel (Figure 2.3) [32]. This homology modelling method

comprises the step of searching highly relevant templates and model alignment [3]. The input

data of the query sequence can be an amino acid or nucleic acid sequence. The entry of amino

acid sequence of human Cav3.2 at the transmembrane domain (TMD) was also obtained from

UniProt (UniProtKB: O95180). The UniProtKB/Swiss-Prot database can also be examined to

shed light on the functional determinants of the query sequence, including valuable

information regarding the sequences’ active domains and studies on mutation and variation

of the query protein [7]. SWISS-MODEL possesses a searching database regarding single

sequence alignment, such as BLAST [10], and HHblits. HHblits is a HMM-based multiple

sequence alignment method with half to one-fold increasing sensitivity and precision [8], over

the PSI-BLAST method [33]. Multiple templates are generated in descending order according

to Global Model Quality Estimate (GMQE) [32] and Quaternary Structure Quality Estimate

(QSQE) scores [34], as shown in Figure 2.4. The model of protein will subsequently be

established by fitting to individual templates via the OpenStructure software [35]

accompanied by ProMod3 as an engine for generating coordinates of the protein structure

[36], if more than one of templates are selected. In this study, the cryo-EM structure of human

Cav3.1 (PDB ID: 6KZO) was selected as explained in Chapter 5. Once the comparative model

of the target protein is constructed, the quality of the model will be assessed via QMEAN

scoring function [10].

Figure 2.3. Homology model of human(h) Cav3.2 built upon cryo-EM structure of hCav3.1 α1 subunit

(PDB ID: 6KZO) [37] via SWISS-MODEL [15]. Top view (A) and lateral view (B) of Cav3.2. The figures of

conformation were produced via VMD [38]. Disordered loops at the intracellular region were removed

before MD simulation.

Figure 2.4. One example of the sequence alignment results for Cav3.2 obtained from SWISS-MODEL. The sequence similarity between the α1 subunit of the target protein and template (human Cav3.1, PDB ID: 6KZO) is 47%.

2.2 Molecular Dynamics Simulation

The homology models of the targeted protein can be refined via MD simulation protocols for

mimicking the dynamics of the target within a model solvated environment. MD simulation

provides a trajectory detailing the movements of each atom (or particle) modelled within the

system, which may be subjected to visualisation and further analysis. These molecular

motions are identified by calculating the forces acting on each atom with respect to time,

which is determined by the force field (FF) of the simulation protocol and integration of

Newton’s classical equations of motion.

2.2.1 Background

From the first MD simulation implemented in the late 1970s [39], this method has been widely

used for predicting the thermodynamic and structural properties of protein and complexes

[40]. A general simulation system for a membrane-bound macromolecule comprises a

membrane bilayer (that mimics the lipid barrier in cells), protein/ protein-ligand complex

(solute), water molecules and ions within a cell of an initially specified size and shape. A cubic

simulation box is a commonly applied example. The vectors of the unit cell are commonly set

up at the very beginning of simulation according to the size of protein and the membrane

bilayer complex [41].

Specifically, the protein in apo form, or bound with a ligand as part of the receptor complex,

obtained from X-ray crystallography, NMR, cryo-EM or theoretical models, will be inserted

into the lipid bilayers and orientated with the bilayer, typically assuming maximum coverage

of the hydrophobic TMD of the proteins by the non-polar tails of the lipids. Water molecules

serve as the solvent to immerse the solute. Cations and anions will be added subsequently

into the system for neutralizing the charges of protein complex (with or without the lipid

bilayer) and providing a specified level of ionic concentration. One example of the simulation

ensembles without lipid molecules is shown in Figure 2.5. Afterwards, a force field will be

selected for determining the potential energy of the system, thereby enabling the mimicking

of the dynamic properties of particles with respect to time. A time-dependent trajectory will

be generated for visualization and further analysis. MD simulations can be performed using a

number of different ensembles, in which various parameters of the system are kept constant

or varied. For example, early MD simulations were conducted under a constant number of

particles (N), energy (E) and volume (V) [40]. However, rather than maintaining constant

energy, experimental studies are typically performed at constant temperature and pressure.

Thus, modern applications of MD simulations are typically performed under conditions that

are intended to mimic more realistic physiological or experimental conditions.

Figure 2.5. The simulation complex of LsIA bound to rat α3β2 nAChR. The Na+ and Cl- ions are added to achieve 0.15 M concentration. LsIAs are shown in green color, while the two α3 and three β2 nAChR subunits are depicted with blue and yellow colors, respectively.

There are also various determinants for improving the accuracy of MD simulation results,

including the initial configuration, force field potential energy function and parameters,

solvation, time step, boundary conditions and treatment of pressure and temperature. These

important factors are discussed in detail in the following parts.

2.2.2 Force Field Potential Energy Function and Parameters

Force fields play a fundamental role in simulation, the accurate description of the energy,

𝐸(𝑅), of a system that is a function of the coordinates of the particles [42]. The potential

energy, 𝑈, requires specification of the parameters for atoms and bonds, which are obtained

from experimental data, or quantum mechanical calculations [43].

In general, a force field includes the potential energy, which can be described by

bonded/intramolecular contributions and nonbonded interactions (equation 2.1). The former

interactions consist of bond stretching, angle bending and bond rotation/torsion, while the

electrostatic and van der Waals interactions are related to the latter (equation 2.2). The

equations of the FF potential may be summarised as follows:

(2.1) 𝐸(𝑅) = 𝐸𝑏𝑜𝑛𝑑𝑒𝑑 + 𝐸𝑛𝑜𝑛−𝑏𝑜𝑛𝑑𝑒𝑑 + 𝐸𝑜𝑡ℎ𝑒𝑟𝑠

(2.2) 𝐸 = E𝑠𝑡𝑟 + E𝑏𝑒𝑛𝑑 + E𝑡𝑜𝑟𝑠 + E𝑣𝑑𝑤 + 𝐸𝑒𝑙 + 𝐸𝑐𝑟𝑜𝑠𝑠

𝑏𝑜𝑛𝑑𝑠

𝑎𝑛𝑔𝑙𝑒𝑠

𝑡𝑜𝑟𝑠𝑖𝑜𝑛𝑠

(2.3) 𝑈 = ∑ + ∑ + ∑ [1 + cos( 𝑛∅ − 𝛿)] 𝑘𝑏(𝑟 − 𝑟0)2 𝑘𝑎(𝜃 − 𝜃0)2 𝑉𝑛 2 1 2 1 2

12 𝜎𝑖𝑗 12 − 𝑟𝑖𝑗

6 𝜎𝑖𝑗 6 ) 𝑟𝑖𝑗

𝐿𝐽

𝑒𝑙𝑒𝑐

+ ∑ + ∑ 4𝜖𝑖𝑗( 𝑞𝑖𝑞𝑗 𝑟𝑖𝑗 + ∑ 𝑉𝑖𝑚𝑝 𝑖𝑚𝑝𝑟𝑜𝑝𝑒𝑟

Both bond stretching and angle bending are represented with harmonic potentials/Hooke’s

Law [44]. In bond stretching potential as shown in equation 2.3, the first term is adopted from

Hooke’s law formula that simplified the Morse potential. The reference bond length, 𝑟0, can

be calculated from experimental methods, such as X-ray diffraction, and the spring constant

𝑘𝑏 showing the stiffness of spring [45], can be obtained from infrared or Raman spectra [46].

Similar to the angle bending, the bending energy potential is presumed to quadratically

change with the angle, when 𝜃0 represents the reference angle. The stiffness of the angel is

described by angle force constant 𝑘𝑎. Furthermore, the force constants are much lower than

the spring constants due to less energy required by twisting an angle compared with

distorting a bond [40]. It should be noted that a rough approximation of the stretching and

bending potentials could be adopted in MD simulation as they do not affect the properties of

interest in most cases of biomolecular simulations under modelled physiological conditions.

However, torsion potentials are much more important, which exert great impact on stability

of conformations and structural changes with time [42]. Torsion energy terms are required

when four or more atoms are existing in a molecule. This energy term is usually expressed

with cosine series. The 𝑉𝑛 represents the qualitative height of barrier, and ∅ is the torsion

angle. The 𝛿 and 𝑛 (multiplicity) are referred to as the phase and number of minimum or

maximum value between 0 and 2 [42]. However, as suggested in the previous studies, some

torsional terms in the AMBER force field only consist of single torsional expansion, whereas

OPLS and MM2 force fields contain three terms from cosine expansion [47, 48], the equation

of which is shown in the following (equation 2.4). Additionally, the improper torsion term is

involved in demonstrating the energy of torsion that are out-of-plane for preserving the

planarity [49].

𝑡𝑜𝑟𝑠𝑖𝑜𝑛𝑠

(2.4) (1 + cos ∅) + (1 − cos 2∅) + (1 + cos 3∅) 𝑈𝑡𝑜𝑟𝑠 = ∑ 𝑘0 + 𝑘2 2 𝑘3 2 𝑘1 2

The non-bonded interactions, van der Waals (vdW) and electrostatic interactions, are

described by Lennard-Jones potential and Coulombic interactions, respectively, between two

atoms with partial charges [50]. The vdW interactions describe the repulsive and attractive

contacts between atoms, which are elicited by the temporal fluctuation of electronic charge

[51]. The most widely used potential function for vdW is the Lennard-Jones potential or 12-6

potential. The 𝑖 and 𝑗 represent 2 interactive atoms, and the corresponding 𝜖𝑖𝑗 is the well

depth describing the strength of attraction between two particles. The 𝜎 is the distance

between two particles when the intermolecular potential is 0, while 𝑟 represents the distance

of separation between the centre of two particles. Another important potential describing

the non-bonded interactions between paired atoms, is electrostatic interactions potential.

The terms, 𝑞𝑖 and 𝑞𝑗, represent the charge of atoms 𝑖 and 𝑗, respectively. These charges are

attributed to partial atomic charges, which are restrained to the nuclear centres of atoms.

The potential energy function of molecules is parameterized for individual force fields,

however which may share similar parameters regarding bond stretching and angle bending

that are considered ‘hard’ (low variation) degrees of freedom. However, problems, such as

clashes or unreliable results of simulations, may be due to the unmodified geometries and

energy potentials of torsion, vdW and electrostatic interactions, in consideration of the

procedures employed in parameter development for the respective force field [52, 53]. In

addition, it may require enormous effort and time for developing the parameters for a ‘soft’

(or highly varying) degree of freedom.

Furthermore, force fields also vary in terms of the property and scale of the molecules that

can be studied. For example, CHARMM [54], AMBER [55] and OPLS force fields, are well suited

for the simulation of large molecules, such as proteins and polymers [56]. By contrast, OPLS

model is more suitable for liquid simulations [57]. It should be noted that the parameters of

certain novel molecules may not be included in all types of force fields, such as amino acid

hydroxyproline (O), which is parameterized in GROMACS G54A7FF force field rather than in

CHARMM force fields. Thus, care should be taken when selecting a force field for conducting

the corresponding simulation system.

2.2.3 Equation of Motion

As mentioned before, MD simulation will generate a trajectory including the information of

velocity, acceleration and position of the particles that change with time [40]. However, these

data are calculated and estimates are made based on the principles of Newton’s laws of

motion [54]. According to Newton’s second law, 𝐹 = 𝑚𝑎, the motion of particle is described

by equations 2.5 and 2.6:

(2.5) = 𝑑2𝑟 𝑑𝑡2 𝐹𝑗 𝑚𝑗

(2.6) 𝐹𝑗 = − 𝑑𝐸 𝑑𝑟

Where the 𝑚𝑗 is the mass of particle along one coordinate (𝑟) with the potential force applied

on the particle due to the potential energy 𝐸 in that direction, when there are 3N dimensional

coordinates of the particle in an MD simulation.

However, it may not be feasible to calculate every micro-moment of trajectory, which is great

time consuming and requires large computational resources. Therefore, simulation

algorithms are used which enable rapid evaluation of Newton’s Law. One of the most

advantageous integrated equations of motion uses the Verlet algorithm [52 , 58], which can

estimate the dynamic characteristics of particles for a very long time with accuracy and

efficiency in accordance with equation 2.7 [56].

(2.7) 𝑟(1 + ∆𝑡) = 2𝑟(𝑡) − 𝑟(𝑡 − ∆𝑡) + ∆𝑡2𝑎(𝑡)

The positions and accelerations at time 𝑡 and (𝑡 − ∆𝑡) are used to calculate the new position

at time 𝑡 + ∆𝑡 without using explicit velocities. This algorithm (equation 2.7) with a modest

storage requirement enables precision of calculation.

For the calculation of velocities with time, the leap-frog algorithm is deployed (equations 2.8

and 2.9).

(2.8) 𝑣(𝑡) = 𝑣 (𝑡 + ∆𝑡) + 𝑣(𝑡 − ∆𝑡) 1 2 1 2 1 2

(2.9) 𝑟(𝑡 + ∆𝑡) = 𝑟(𝑡) + 𝑣∆𝑡(𝑡 + ∆𝑡) 1 2

the velocity at time (𝑡 +

In these two algorithms, the position of the particle at time (𝑡 + ∆𝑡) is calculated based upon

∆𝑡), which allows the velocities and positions to leap over each

other. The velocities and positions of an individual particle, therefore, are not calculated at

the same time step.

2.2.4 Time Step

To achieve a continuous and stable simulation, a very short time step should be applied and

combined for the integration of motion [53], however as the efficiency of the simulation

should also be considered, the time step should be reasonably short rather than too short

(i.e. the simulated system should be at least fully equilibrated before produced useful

trajectory for analysis). Also, there is a trade-off between accuracy and computational

efficiency; time steps should be selected to be smaller than the frequency of motion of the

fastest degrees of freedom. Time steps (2 fs to 10 fs) are commonly used, depending on the

scale of the systems.

2.2.5 Solvation

Solvation can exert a great impact on the structural and functional properties of molecules

[59]. The coordinates and parameters of the solvent are generally included in the bank of

force field data. However, the selection of force field can also affect the types of solvation

models; most force fields are developed with aqueous models in mind, whereas sometimes

cations and anions need to be introduced for neutralization of the charged system and/or

adding a co-solvent of a particular concentration to the system. The aqueous models have

been approved to mimic the thermodynamic and structural properties of water [60], which

include but are not limited to the commonly applied SPC [61], SPC/E [62], TIP3P, TIP4P [60],

TIP4P-Ew [63], and TIP5P [64] models. The molecular structure of TIP3P is shown in Figure2.6

as an example.

Figure 2.6. The conformation of TIP3P water molecules applied to the current study in this thesis.

Implicit aqueous models are also commonly used solvation methods in simulation. The free

energy of water models varies due to Lennard-Jones and Coulombic terms [65]. These

energies contribute to the total potential energy of the modelled system, and should be

chosen so as to be compatible with the force field selected for the simulation. Common water

molecule models and their compatible force fields are listed in Table 2.1. However, certain

force fields may optimally suit a particular type of aqueous model. For instance, TIP3P may fit

CHRAMM and GROMOS force fields [65], while SPC/E is best compatible with AMBER99,

GROMOS 53A6 and OPLS-AA force fields [66].

Table 2.1. The compatible water molecule models with widely used force fields.

Force field

SPC

SPC/E

TIP3P

TIP4P

TIP4P-

TIP5P

Reference

Ew

CHARMM

[65, 67, 68]

√

OPLS

√

[66, 67, 69]

√

AMBER

√

[66-68, 70]

√

GROMOS

√

[66, 68]

2.2.6 Periodic Boundary Conditions

The proper boundary treatment is required to ensure the macroscopic properties of the

molecules, which are calculated through relatively small numbers of particles during

simulation, are sufficiently accurate [40]. For this reason, periodic boundary conditions are

employed. A simple example is the replicate cubic box, in which particles will be copied in all

directions for a periodic arrangement (Figure 2.7). One cubic box is embraced by eight same

neighbours in a 3-D space. If a particle of the system leaves the box, it will then be replaced

by another particle that enters the box from the opposite edge. This enables the removal of

surface effects, which may affect the accuracy of the water structure around the solute

biomolecule, allowing for an approximation of a bulk water environment.

The cubic cell may be sufficient for a simple periodic boundary system. However, some

simulations may need a different shape of cell; other cell shapes may also be more efficient

for the respective simulation systems, requiring smaller number of water molecules to be

included while still allowing for sufficient separation between image solute molecules. In the

system of an intermolecular complex immersed by the solvent molecules, for instance, the

motion of the solute (complex) is commonly of more interest over the solvent molecules far

away from it. In this case, the other cell shapes can also achieve the purpose, namely

truncated octahedron, rhombic dodecahedron, elongated dodecahedron and hexagonal

prism [40]. One can select these five periodic cell shapes, depending on which best suits the

geometry of the system.

As mentioned above, if a particle of the system passes through any edge of the box, it will be

removed and reappears on the opposite edge with the same velocity. However, the

interaction of the original system with its images must be avoided. Thus, a system should be

placed in a box of appropriate dimensions or shape, to obtain the appropriate results, but still

allowing for increased computational efficiency. Within the boundary conditions, the particles

of the system may also be placed in a bulk fluid to avoid edge effects.

Figure 2.7. The side view of Cav3.2 centred in a grid box. (A) The periodic boundary conditions for Cav3.2. (B) One cubic box with a Cav3.2 model.

2.2.7 Initial Configuration

Before the simulation, it is crucial to choose the starting configuration for generating more

realistic properties of the protein and increasing the likelihood of success of simulation in

predicting properties [71]. Apart from this, high-energy interactions can be removed by

conducting energy minimization via the steepest descent or other methods [72]. Additionally,

for some inhomogeneous systems, such as membranes, the solvent may intrude into the

middle of the bilayer, thus it may be better to remove such water molecules in the starting

structure to ensure a proper arrangement of particles. Methods for generating novel initial

configurations, such as PACKMOL [73], can contribute to providing a well-packed system

before simulation.

2.2.8 Treatment of Pressure and Temperature

To further improve the accuracy of MD simulations that describe the structural properties of

a canonical ensemble in experiment, equilibration steps can be incorporated apart from

energy minimization. These equilibrium ensembles include constant temperature, volume

and number of particles NVT (optional) and pressure NPT (constant temperature, pressure

and number of particles) ensemble simulations. NVT and NPT ensembles will not only improve

the flexibility of MD simulation to model different external conditions of the system, but also

ensure the stabilization of the geometry and orientation of solvent packed around the solute

until the temperature and density (pressure) achieving appropriate values.

• NVT ensembles

The NVT ensembles with a thermostat are designed to investigate the particular behaviour of

a system under certain temperature and prevent the steady energy drift from happening.

Velocity scaling is a simple way of controlling variations in temperature (equation 2.11).

2 ∑ 𝑚𝑖𝑣𝑖 𝑖

(2.10) 𝑁𝑘𝐵𝑇 = 3 2 1 2

(2.11) 𝜆 = √(𝑇 𝑇𝑖)⁄

The factor 𝜆 can be calculated using the equation 2.11 shown above, here, 𝑇𝑖 is the associated

temperature calculated from the kinetic energy, while 𝑇 is the desired temperature.

However, this scaled velocity thermostat may not be applicable for a canonical ensemble as

it does not include the fluctuation in temperature.

A weaker formulation of the Berendsen thermostat is used via coupling to an external heat

bath according to equation 2.12 [74].

(2.12) 𝑇0 𝜆2 = 1 + { − 1} 𝛿𝑡 𝜏 𝑇(𝑡 − ) 𝛿𝑡 2

𝜏 is the empirical parameter to determine how tightly the coupling of the system should be

𝛿𝑡

to the external heat bath. The strength of coupling decreases with an increasing value of 𝜏,

which means the system will achieve its desired temperature in a longer time. The 𝑇(𝑡 − )

term is introduced due to the leap-frog algorithm, which is used for time integration.

However, it should be noted that the Berendsen thermostat does not produce correct

canonical ensembles [75], which therefore may be more suitable for relaxing the system to

the target temperature during equilibrium.

Once the system reaches its target temperature, it is essential to apply a thermostat method

that enables sampling from a canonical ensemble. The most widely used method is invented

by Nosé and then modified by Hoover [76, 77]. In general, this thermostat is treated as an

integral with the system, via incorporating a friction force ξ (equation 2.13).

(2.13) − ξ 𝑑2𝑟𝑖 𝑑𝑡2 = 𝑑𝑟𝑖 𝑑𝑡 𝐹𝑖 𝑚𝑖

The equation of motion of ξ can be calculated using equation 2.14, in which the constant 𝑄

controls the coupling strength of the reservoir and the real system [78], so that the fluctuation

of temperature is considered [77]. The 𝑇 represents the current temperature, while 𝑇0 is the

reference temperature.

(2.14) = (𝑇 − 𝑇0) 𝑑ξ 𝑑𝑡 1 𝑄

• NPT ensembles

The NPT ensembles couple the system to an external pressure bath. Weak coupling is

represented by Berendsen pressure coupling, whereas an extended coupling approach is the

Parrinello-Rahman pressure barostat. The pressure changes between the first order

relaxation of pressure (𝑃) and the reference pressure (𝑃0) is shown by equation 2.15,

(2.15) ) = 𝑑P ( 𝑑𝑡 𝑃0 − 𝑃 𝜏𝑝

where 𝜏𝑝 (equation 2.15) is the time constant for the coupling. The isotropic pressure (𝑃) can

be calculated using equation 2.16,

(2.16) 𝑃 = (𝐸𝐾 −𝐸) 2 3𝑉

where the tensor P (distinct from pressure 𝑃) is calculated by the Clausius virial theorem via

𝑉

the (𝐸𝐾 −𝐸) term. 𝐸𝐾 is the total kinetic energy of the system, while 𝐸 represents the

potential energy, calculated by the sum of all potential energies of pairs of particles.

Similar to the Berendsen thermostat in which the velocity was scaled by factor 𝜆, a scaling

factor, 𝜇0, is applied for each dimension (equation 2.17),

(2.17) 𝜇0 = 1 − (𝑃0 −𝑃) 𝛽∆𝑡 3𝜏𝑃

where 𝛽 is the isothermal compressibility that has a ratio with 𝜏𝑃.

Parrinello-Rahman pressure coupling allows the fluctuation of volume and shape by inclusion

of an extra degree of freedom, which is similar with the Nosé-Hoover temperature algorithm

[76, 79 ]. According to the matrix equation of motion, the box vectors are represented by the

matrix 𝑏 (equation 2.18).

(2.18)

𝑑𝑏2 𝑑𝑡2 = 𝑉𝑊−1𝑏′−1 − (𝑃 − 𝑃𝑟𝑒𝑓)

where 𝑊determines the strength of the coupling, while 𝑉 represents the volume of the box.

The present and reference pressures are denoted by matrices 𝑃 and 𝑃𝑟𝑒𝑓, respectively.

The equations of motion change due to the fluctuation of box vectors and the modified

Parrinello-Rahman algorithm, as shown in the equation 2.19.

(2.19) − (𝑀 ) 𝑑2𝑟𝑖 𝑑𝑡2 = 𝑑𝑟𝑖 𝑑𝑡 𝐹𝑖 𝑚𝑖

𝑀 denotes the mass matrix [80], which can be calculated using equation 2.20.

(2.20) 𝑀 = 𝑏−1 [𝑏 + 𝑏֙] 𝑏֙−1 𝑑𝑏֙ 𝑑𝑡 𝑑𝑏 𝑑𝑡

One can implement Parrinello-Rahman combined with Nosé-Hoover thermostat, however,

this is quite time consuming for the MD simulation. In periodic boundary conditions, a

constant NPT ensemble is often sufficient for describing the canonical ensemble which

accurately models experimental conditions.

2.2.9 Examples of Application

The general methodology of MD simulation applied in this thesis are described in this section.

The details of the method will be explained in the following main chapters (Chapters 3,4 &5)

in this thesis.

The GROMACS 4 (v4.6.5) was applied for simulation of nAChR and LsIA systems (Chapter 3,4),

while GROMACS v2016 was used as the molecular dynamics package for Cav3.2 (Chapter 5)

and µ-conotoxin systems. The parameters and coordination of molecules were computed

using CHARMM27 force field [81] for all the simulation systems for consistency. The models

of the simulation systems that obtained from homology models and docking methods were

used to generate the topology file of each simulation system. Whilst a grid box in appropriate

dimension was set up for each system, the clashes between the atoms were eliminated due

to periodic boundary conditions. The initial structure in the grid box will be immersed with

water molecules and neutralized by cations (Na+/Ca2+) and anion (Cl-) for achieving 0.15 M

concentration. Once the systems were energy minimized, the NPT ensemble was applied for

nAChR and LsIA systems, while the Cav3.2 and µ-conotoxin systems went through both NVT

and NPT equilibrium. The MD production run was conducted for generating the simulation

trajectory for further analysis.

The major differences between the methodology of simulating the neuronal nAChRs at ECD,

α7 (in Chapter 3) and α3β2 (in Chapter 4), and neuronal ion channel pore forming subunit

(Cav3.2 in Chapter 5) are mainly due to the necessity of including a lipid membrane for the

latter system. The TMD of Cav3.2 α1 subunit was selected to explore the binding mechanism

of µ-conotoxins of GIIIA, GIIIB, GIIIC, PIIIA and KIIIA at the ion channels. Therefore, the bilayer

membrane is essential for inclusion in the simulation, as the membrane environment is crucial

to the structural stability of the complex, which enables the appropriate function of the ion

dioleoyl-sn-glycero-3-phosphocholine

channel, and also affects the binding of ligands [82, 83]. In Chapter 5, the molecule of 1,2-

(DOPC) bilayer was retrieved from Lipid-book

(https://lipidbook.org/lipid/) for Cav3.2 in apo form or bound by µ-conotoxins. The individual

ligand binding complexes were embedded within the DOPC bilayer using VMD, whilst the

bilayer molecules of DOPC within 4 Å of the protein-ligand complex were removed for

ensuring the functional integrity of the protein. By contrast, since the ligand binding domain

of nAChR is mainly located between the adjacent subunits at the ECD, the membrane bilayer

was excluded in the nAChR/LsIA systems in the present study (Chapter 3, 4). Additionally, the

dynamics changes of nAChRs/ion channels upon bound by the conotoxins were investigated

in Chapter 3, however such topic needs further exploration in the future.

2.3 MMPBSA

The Molecular Mechanics Poisson-Boltzmann Surface Area (MMPBSA) approach is widely

used to predict the binding free energy of the ligand to the targeted receptor with efficiency

depending on the systems [84]. It has been demonstrated that MMPBSA provides more

reliable binding affinity results versus the scoring methods, such as docking, and is less time-

consuming with a reasonable price versus experiments [85]. In the present study, the

principles of MMPBSA will be simply explained in the following.

The MMPBSA approaches commonly adopt the algorithm by Kollman et al. [86]. The equation

is shown below (equation 2.21). In this equation, the binding free energy is calculated via the

complex binding free energy subtracting the sum of the energy of the protein and ligand. Each

free energy term comprises several contributions, as shown in equation 2.22,

(2.21) ∆𝐺𝑏𝑖𝑛𝑑 =< 𝐺𝑃𝐿 >−< 𝐺𝑃 > −< 𝐺𝐿 >

(2.22) 𝐺 = 𝐸𝑐𝑜𝑣𝑎𝑙𝑒𝑛𝑡 + 𝐸𝑒𝑙 + 𝐸𝑣𝑑𝑊 + 𝐺𝑝𝑜𝑙 + 𝐺𝑛𝑜𝑛−𝑝𝑜𝑙 − 𝑇𝑆

in which, the Molecular Mechanics (MM) energy consists of covalent (angle, bond and torsion,

shown in Equation 2.23), electrostatic and vdW contacts. The polar and non-polar interactions

contributing to the solvation free energy are represented by 𝐺𝑝𝑜𝑙 and 𝐺𝑛𝑜𝑛−𝑝𝑜𝑙, respectively.

The entropy change of the conformation upon the binding of the ligand is predicted by

including the 𝑇𝑆 term in each of the free energy terms in ∆𝐺𝑏𝑖𝑛𝑑. The energies arising from

the bond stretching, angle bending, and torsion energies, together comprise the covalent

energy:

(2.23) 𝐸𝑐𝑜𝑣𝑎𝑙𝑒𝑛𝑡 = 𝐸𝑏𝑜𝑛𝑑 + 𝐸𝑎𝑛𝑔𝑙𝑒 + 𝐸𝑡𝑜𝑟𝑠𝑖𝑜𝑛

Notably, three individual simulations assist in the computation of the respective binding free

energies, namely protein-ligand complex, ligand only and receptor in the apo form [85].

Nonetheless, in most MMPBSA methods, such as g_mmpbsa compatible with GROMACS, the

binding energy is commonly predicted from the complex simulation alone [87, 88].

Furthermore, the accuracy of MMPBSA may also be affected by the calculation of the entropy

of the conformation [89].

2.4 References

Dolphin AC. Voltage‐gated calcium channels and their auxiliary subunits: physiology

Rodriguez R, Chinea G, Lopez N, Pons T, Vriend G. Homology modeling, model and

Koonin EV, Wolf YI, Karev GP. The structure of the protein universe and genome

Schaeffer RD, Daggett V. Protein folds and protein folding. Protein Engineering Design

Pearson WR. An introduction to sequence similarity ("homology") searching. Current

Björn W. Protein structure prediction: mdoel building and quality assessment

Schwede T, Kopp Jr, Guex N, Peitsch MC. SWISS-MODEL: an automated protein

1. and pathophysiology and pharmacology. The Journal of Physiology. 2016;594(19):5369-90. 2. Inserra MC, Kompella SN, Vetter I, Brust A, Daly NL, Cuny H, et al. Isolation and characterization of α-conotoxin LsIA with potent activity at nicotinic acetylcholine receptors. Biochemical Pharmacology. 2013;86(6):791-9. 3. software evaluation: three related resources. Bioinformatics. 1998;14(6):523-8. 4. Illergård K, Ardell DH, Elofsson A. Structure is three to ten times more conserved than sequence—a study of structural response in protein cores. Proteins: Structure, Function, and Bioinformatics. 2009;77(3):499-508. 5. evolution. Nature. 2002;420:218. 6. & Selection. 2011;24(1-2):11-9. 7. Protocols in Bioinformatics. 2013;Chapter 3:Unit3.1-Unit3.1. 8. Arnold K, Bordoli L, Kopp J, Schwede T. The SWISS-MODEL workspace: a web-based environment for protein structure homology modelling. Bioinformatics. 2006;22(2):195-201. Hatherley R, Brown DK, Glenister M, Tastan Bishop Ö. PRIMO: An interactive 9. homology modeling pipeline. PLOS ONE. 2016;11(11):e0166698. 10. Waterhouse A, Bertoni M, Bienert S, Studer G, Tauriello G, Gumienny R, et al. SWISS- MODEL: homology modelling of protein structures and complexes. Nucleic Acids Research. 2018;46(W1):W296-W303. 11. Berman H, Henrick K, Nakamura H, Markley JL. The worldwide Protein Data Bank (wwPDB): ensuring a single, uniform archive of PDB data. Nucleic Acids Research. 2007;35(suppl_1):D301-D3. 12. Altschul SF, Madden TL, Schäffer AA, Zhang J, Zhang Z, Miller W, et al. Gapped BLAST and PSI-BLAST: a new generation of protein database search programs. Nucleic acids research. 1997;25(17):3389-402. Pearson WR, Lipman DJ. Improved tools for biological sequence comparison. 13. Proceedings of the National Academy of Sciences of the United States of America. 1988;85(8):2444-8. 14. [Doctoral Thesis]. Sweden: Stockholm University 2005. 15. homology-modeling server. Nucleic Acids Research. 2003;31(13):3381-5. Bates PA, Kelley LA, MacCallum RM, Sternberg MJE. Enhancement of protein modeling 16. by human intervention in applying the automatic programs 3D-JIGSAW and 3D-PSSM. Proteins: Structure, Function, and Bioinformatics. 2001;45(S5):39-46. Petrey D, Xiang Z, Tang CL, Xie L, Gimpelev M, Mitros T, et al. Using multiple structure 17. alignments, fast model building, and energetic analysis in fold recognition and homology modeling. Proteins: Structure, Function, and Bioinformatics. 2003;53(S6):430-5. Suresh A. Computational design and studies of neuro-active peptides and the effects 18. of calcium on nicotinic acetylcholine receptors in relation to neurodegenerative disorders. RMIT University; 2016.

Abagyan RA, Batalov S. Do aligned sequences share the same fold?11Edited by F. E.

Levitt M. Accurate modeling of protein conformation by automatic segment matching.

Šali A, Blundell TL. Comparative protein modelling by satisfaction of spatial restraints.

Eisenberg D, Lüthy R, Bowie JU. VERIFY3D: Assessment of protein models with three-

Benkert P, Biasini M, Schwede T. Toward the estimation of the absolute quality of

in three-dimensional

Guex N, Peitsch MC. SWISS-MODEL and the Swiss-Pdb Viewer: an environment for

Remmert M, Biegert A, Hauser A, Söding J. HHblits: lightning-fast iterative protein

Bienert S, Waterhouse A, de Beer TAP, Tauriello G, Studer G, Bordoli L, et al. The functionality. Nucleic acids research. features and 19. Wallner B, Elofsson A. All are not equal: a benchmark of different homology modeling programs. Protein Science. 2005;14(5):1315-27. 20. Cohen. Journal of Molecular Biology. 1997;273(1):355-68. 21. Journal of Molecular Biology. 1992;226(2):507-33. 22. Journal of Molecular Biology. 1993;234(3):779-815. 23. dimensional profiles. Methods in Enzymology. 277: Academic Press; 1997. p. 396-404. Laskowski RA, MacArthur MW, Moss DS, Thornton JM. PROCHECK: a program to check 24. the stereochemical quality of protein structures. Journal of Applied Crystallography. 1993;26:283-91. 25. Laskowski RA, Rullmann JAC, MacArthur MW, Kaptein R, Thornton JM. AQUA and PROCHECK-NMR: Programs for checking the quality of protein structures solved by NMR. Journal of Biomolecular NMR. 1996;8(4):477-86. 26. Morris AL, MacArthur MW, Hutchinson EG, Thornton JM. Stereochemical quality of protein structure coordinates. Proteins: Structure, Function, and Bioinformatics. 1992;12(4):345-64. 27. individual protein structure models. Bioinformatics. 2011;27(3):343-50. 28. Wiederstein M, Sippl MJ. ProSA-web: interactive web service for the recognition of errors structures of proteins. Nucleic Acids Research. 2007;35(suppl_2):W407-W10. Quintaje SB, Orchard S. The annotation of both human and mouse kinomes in 29. UniProtKB/Swiss-Prot. One Small Step in Manual Annotation, One Giant Leap for Full Comprehension of Genomes. 2008;7(8):1409-19. Celie PHN, Kasheverov IE, Mordventisev DY, Hogg RC, van Nierop P, van Elk R, et al. 30. Crystal structure of nicotinic acetylcholine receptor homolog AChBP in complex with an α- conotoxin PnIA variant. Nature Structural and Molecular Biology. 2005;12. 31. comparative protein modeling. ELECTROPHORESIS. 1997;18(15):2714-23. 32. Biasini M, Bienert S, Waterhouse A, Arnold K, Studer G, Schmidt T, et al. SWISS-MODEL: modelling protein tertiary and quaternary structure using evolutionary information. Nucleic Acids Research. 2014;42(W1):W252-W8. 33. sequence searching by HMM-HMM alignment. Nature Methods. 2011;9:173. Bertoni M, Kiefer F, Biasini M, Bordoli L, Schwede T. Modeling protein quaternary 34. structure of homo- and hetero-oligomers beyond binary interactions by homology. Scientific Reports. 2017;7(1):10480. 35. Biasini M, Schmidt T, Bienert S, Mariani V, Studer G, Haas J, et al. OpenStructure: an integrated software framework for computational structural biology. Acta crystallographica Section D, Biological crystallography. 2013;69(Pt 5):701-9. 36. SWISS-MODEL Repository-new 2017;45(D1):D313-D9.

Zhao Y, Huang G, Wu Q, Wu K, Li R, Lei J, et al. Cryo-EM structures of apo and

Humphrey W, Dalke A, Schulten K. VMD: visual molecular dynamics. Journal of

Leach AR. Molecular modelling : principles and applications. 2nd ed. ed. Harlow:

Hofsäss C, Lindahl E, Edholm O. Molecular dynamics simulations of phospholipid

González MA. Force fields and molecular dynamics simulations. École thématique de

Polanski J. 4.26 - Chemoinformatics. In: Brown S, Tauler R, Walczak B, editors.

Bassichis WH. A simple model of initial tension in springs. The Physics Teacher.

Lim T-C. Spring Constant analogy for estimating stiffness of a single polyethylene

Tadayon M, Amini S, Masic A, Miserez A. The mantis shrimp saddle: a biological spring

Pu M, Garrahan JP, Hirst JD. Comparison of implicit solvent models and force fields in the PB1 domain. Chemical Physics Letters.

Hospital A, Goñi JR, Orozco M, Gelpí JL. Molecular dynamics simulations: advances and

37. antagonist-bound human Cav3.1. Nature. 2019;576(7787):492-7. 38. Molecular Graphics. 1996;14(1):33-8. 39. McCammon JA, Gelin BR, Karplus M. Dynamics of folded proteins. Nature. 1977;267:585. 40. Prentice Hall; 2001. 41. bilayers with cholesterol. Biophysical journal. 2003;84(4):2192-206. 42. la Société Française de la Neutronique. 2011;12:169-200. 43. Comprehensive Chemometrics (Second Edition). Oxford: Elsevier; 2020. p. 635-76. 44. 2019;57(6):400-2. 45. molecule. Journal of Mathematical Chemistry. 2003;34(3):151-61. 46. combining stiffness and flexibility. Advanced Functional Materials. 2015;25(41):6437-47. Jorgensen WL, Maxwell DS, Tirado-Rives J. Development and testing of the OPLS all- 47. atom force field on conformational energetics and properties of organic liquids. Journal of the American Chemical Society. 1996;118(45):11225-36. 48. Cho SG, Unwalla RJ, Cartledge FK, Profeta Jr. S. Chlorosilanes: development and application of MM2 force field parameters. Journal of Computational Chemistry. 1989;10(6):832-49. 49. Joodaki F, Martin LM, Greenfield ML. Planarity and out-of-plane vibrational modes of tryptophan and tyrosine in biomolecular modeling. Physical chemistry chemical physics : PCCP. 2019;21(43):23943-65. 50. molecular dynamics simulations of 2011;515(4):283-9. 51. Kawai S, Foster AS, Björkman T, Nowakowska S, Björk J, Canova FF, et al. Van der Waals interactions and the limits of isolated atom models at interfaces. Nature Communications. 2016;7:11559. 52. Grubmüller H, Heller H, Windemuth A, Schulten K. Generalized verlet algorithm for efficient molecular dynamics simulations with long-range interactions. Molecular Simulation. 1991;6(1-3):121-42. 53. applications. Advances and Applications in Bioinformatics and Chemistry. 2015;8:37-47. 54. MacKerell AD, Bashford D, Bellott M, Dunbrack RL, Evanseck JD, Field MJ, et al. All- atom empirical potential for molecular modeling and dynamics studies of proteins. The Journal of Physical Chemistry B. 1998;102(18):3586-616. 55. Weiner SJ, Kollman PA, Case DA, Singh UC, Ghio C, Alagona G, et al. A new force field for molecular mechanical simulation of nucleic acids and proteins. Journal of the American Chemical Society. 1984;106(3):765-84.

Berendsen HJC, Grigera JR, Straatsma TP. The missing term in effective pair potentials.

Hu Z, Jiang J. Assessment of biomolecular force fields for molecular dynamics

Nguyen TT, Viet MH, Li MS. Effects of water models on binding affinity: evidence from neuraminidase. simulation A/H5N1 binding tamiflu to of of

Cheng QH, Lu C, Lee HP. A molecular dynamics simulation of hexagonal solid platinum 56. Michael PA. Introduction to molecular dynamics simulation. Computational Soft Matter: From Synthetic Polymers to Proteins. 2004. p. 1-28. 57. Jorgensen WL, Tirado-Rives J. The OPLS [optimized potentials for liquid simulations] potential functions for proteins, energy minimizations for crystals of cyclic peptides and crambin. Journal of the American Chemical Society. 1988;110(6):1657-66. 58. Levesque D, Verlet L, Kürkijarvi J. Computer "experiments" on classical fluids. IV. transport properties and time-correlation functions of the Lennard-Jones liquid near its triple point. Physical Review A. 1973;7(5):1690-700. 59. Jalili S, Amani P. Molecular dynamics simulation study of solvation effects of water and trifluoroethanol on gamma-aminobutyric acid (GABA). Journal of Molecular Liquids. 2014;197:27-34. 60. Jorgensen WL, Chandrasekhar J, Madura JD, Impey RW, Klein ML. Comparison of simple potential functions for simulating liquid water. The Journal of Chemical Physics. 1983;79(2):926-35. Berendsen HJC, Postma JPM, van Gunsteren WF, Hermans J. Interaction models for 61. water in relation to protein hydration. In: Pullman B, editor. Intermolecular Forces: Proceedings of the Fourteenth Jerusalem Symposium on Quantum Chemistry and Biochemistry Held in Jerusalem, Israel, April 13–16, 1981. Dordrecht: Springer Netherlands; 1981. p. 331-42. 62. The Journal of Physical Chemistry. 1987;91(24):6269-71. 63. Horn HW, Swope WC, Pitera JW, Madura JD, Dick TJ, Hura GL, et al. Development of an improved four-site water model for biomolecular simulations: TIP4P-Ew. The Journal of Chemical Physics. 2004;120(20):9665-78. 64. Mahoney MW, Jorgensen WL. A five-site model for liquid water and the reproduction of the density anomaly by rigid, nonpolarizable potential functions. The Journal of Chemical Physics. 2000;112(20):8910-22. 65. Mark P, Nilsson L. Structure and dynamics of the TIP3P, SPC, and SPC/E water models at 298 K. The Journal of Physical Chemistry A. 2001;105(43):9954-60. 66. simulations in a protein crystal. Journal of Computational Chemistry. 2010;31(2):371-80. Piana S, Donchev AG, Robustelli P, Shaw DE. Water Dispersion Interactions Strongly 67. Influence Simulated Structural Properties of Disordered Protein States. The Journal of Physical Chemistry B. 2015;119(16):5113-23. 68. all-atom TheScientificWorldJournal. 2014;2014:536084-. Hess B, van der Vegt NFA. Hydration Thermodynamic properties of amino acid 69. analogues: a systematic comparison of biomolecular force fields and water models. The Journal of Physical Chemistry B. 2006;110(35):17616-26. 70. Caldwell JW, Kollman PA. Structure and properties of neat liquids using nonadditive molecular dynamics: water, methanol, and N-methylacetamide. The Journal of Physical Chemistry. 1995;99(16):6208-19. 71. Wang B-B, Wang X-D, Chen M, Xu J-L. Molecular dynamics simulations on evaporation of droplets with dissolved salts. Entropy. 2013;15:1232-46. 72. nanowires. Molecular Simulation. 2005;31(4):289-93.

Berendsen HJC, Postma JPM, Gunsteren WFv, DiNola A, Haak JR. Molecular dynamics

Nosé S, Klein ML. Constant pressure molecular dynamics for molecular systems.

Evans DJ, Holian BL. The Nose–Hoover thermostat. The Journal of Chemical Physics.

in GROMACS. Fernández-Pendás M, Escribano B, Radivojević T, Akhmatskaya E. Constant pressure Journal of Molecular Modeling.

Sapay N, Tieleman DP. Combination of the CHARMM27 force field with united-atom

Genheden S, Ryde U. The MM/PBSA and MM/GBSA methods to estimate ligand-

Kumari R, Kumar R, Lynn A. g_mmpbsa—A GROMACS tool for high-throughput MM-

73. Martínez L, Andrade R, Birgin EG, Martínez JM. PACKMOL: a package for building initial configurations for molecular dynamics simulations. Journal of Computational Chemistry. 2009;30(13):2157-64. 74. with coupling to an external bath. The Journal of Chemical Physics. 1984;81(8):3684-90. 75. Mudi A, Chakravarty C. Effect of the Berendsen thermostat on the dynamical properties of water. Molecular Physics. 2004;102(7):681-5. 76. Molecular Physics. 1983;50(5):1055-76. 77. 1985;83(8):4069-74. 78. Hünenberger PH. Thermostat algorithms for molecular dynamics simulations. In: Dr. Holm C, Prof. Dr. Kremer K, editors. Advanced Computer Simulation: Approaches for Soft Matter Sciences I. Berlin, Heidelberg: Springer Berlin Heidelberg; 2005. p. 105-49. Yu T-Q, Alejandre J, López-Rendón R, Martyna GJ, Tuckerman ME. Measure-preserving 79. integrators for molecular dynamics in the isothermal–isobaric ensemble derived from the Liouville operator. Chemical Physics. 2010;370(1):294-305. 80. hybrid Monte Carlo simulations 2014;20(12):2487. 81. lipid force fields. Journal of Computational Chemistry. 2011;32(7):1400-10. 82. Qiu W, Fu Z, Xu GG, Grassucci RA, Zhang Y, Frank J, et al. Structure and activity of lipid bilayer within a membrane-protein transporter. Proceedings of the National Academy of Sciences. 2018;115(51):12985. 83. Agwa AJ, Peigneur S, Chow CY, Lawrence N, Craik DJ, Tytgat J, et al. Gating modifier toxins isolated from spider venom: modulation of voltage-gated sodium channels and the role of lipid membranes. Journal of Biological Chemistry. 2018;293(23):9041-52. 84. Wang C, Nguyen PH, Pham K, Huynh D, Le T-BN, Wang H, et al. Calculating protein- ligand binding affinities with MMPBSA: method and error analysis. Journal of Computational Chemistry. 2016;37(27):2436-46. 85. binding affinities. Expert Opinion on Drug Discovery. 2015;10(5):449-61. 86. Srinivasan J, Cheatham TE, Cieplak P, Kollman PA, Case DA. Continuum solvent studies of the stability of DNA, RNA, and phosphoramidate−DNA helices. Journal of the American Chemical Society. 1998;120(37):9401-9. 87. PBSA calculations. Journal of Chemical Information and Modeling. 2014;54(7):1951-62. 88. Miller BR, McGee TD, Swails JM, Homeyer N, Gohlke H, Roitberg AE. MMPBSA.py: an efficient program for end-state free energy calculations. Journal of Chemical Theory and Computation. 2012;8(9):3314-21. 89. Wang C, Greene DA, Xiao L, Qi R, Luo R. recent developments and applications of the MMPBSA method. Frontiers in Molecular Biosciences. 2018;4(87).

Chapter 3 : Effects of C-terminal Carboxylation on α-

Conotoxin LsIA Interactions with Human α7 Nicotinic

Acetylcholine Receptor

3.1 Introduction

nAChRs are situated in the plasma membranes of certain neurons in the central and

peripheral nervous systems [1]. These ligand-gated ion channels are involved in signal

transmission, neuronal integration and cell excitability through activation by acetylcholine

(ACh) binding [1]. There are two general types of nAChRs, heteropentamers and

homopentamers, both forming a cylindrical helical bundle around the pore of the receptors

[2]. The homomeric nAChRs are constituted by 5 identical subunits, whereas

heteropentameric nAChRs consist of distinct subunits which may include combinations of α1-

10, β1-4, γ, δ and ε subunits [3]. The various combinations of subunits also render distinct

nAChRs with unique physiological and pharmacological properties [4].

Human (h) α7 nAChR subtype, as an example of homopentameric nAChRs, has been

extensively studied due to its important function in cognitive processes and an implicated

involvement in the cholinergic anti-inflammatory pathway [5]. The dysfunction of α7 nAChR,

therefore, is related to many neurological diseases and disorders, such as inflammation,

schizophrenia [6, 7], Alzheimer’s disease and Parkinson’s disease [1, 8-10]. The determination

of the crystal structures of acetylcholine binding protein (AChBP) complex to conotoxins has

enabled the construction of molecular models of human α7 nAChR bound to conotoxins [6,

11, 12], although detailed structural information on toxin-bound nAChRs is relatively scarce.

Even if the similarities between sequences of AChBP [13] and the extracellular domain (ECD)

(the region for ACh binding [14]) of human α7 nAChR are less than 30% [15, 16], up to 52% of

its sequence at the ligand binding domain (LBD) is identical to that of AChBP [17, 18]. This

suggests that AChBPs and nAChRs share similar functional and structural properties [13, 19].

Thus, AChBPs may serve as useful structural templates for comparative modelling of the ECDs

of nAChRs.

Conotoxins, derived from marine cone snail, are peptides rich in disulphide bonds connecting

conserved cysteine residues. The synthesis and manipulation of these neuroactive peptides

as potential pharmacophores antagonists and molecular probes have attracted attention

because of their relatively short length, well-defined backbone structure, and specificity to a

wide range of ion channels and nAChRs with high potency [1, 20-22]. Furthermore, with many

possible variations of amino acid residues apart from conserved cysteines, conotoxins vary

greatly in their ability to target different neuronal/muscular acetylcholine receptors [23]. The

most studied conotoxin subfamily is α-conotoxins, which are known to naturally and

selectively bind at nAChR subtypes [24]. α-Conotoxins commonly contain fewer than 20

amino acid residues with a compact structure including 2 disulphide bridges [25]. These

peptides are widely utilized to probe the structure-function relationships of nAChR binding

sites [26], for differentiating variable nAChR subtypes [5], and in the development of potential

pharmacological tools [27].

α-Conotoxin LsIA (17 amino acids in length), is a neuropeptide discovered from Conus limpusi,

which shows distinct high binding affinity (IC50 10.1 nM) to hα7 nAChR subtype relative to

α3β4 [28], α3β2 and α3α5β2 nAChR subtypes, as determined in previous experiments [29].

Similar to other α-conotoxins, LsIA possesses 2 disulphide bonds connecting 4 conserved

cysteine residues [30], and the disulphide bridges produce a unique 2-loop structure (4/7)

[23] with a naturally amidated C-terminus (shown in Figure 3.1A). The unique motif in loop 1

of certain α-conotoxins was implicated in subtype-selective binding to nAChRs via the highly

conserved proline [31], such as Lo1a, GID and LsIA (in position 7 of LsIA) [1, 29, 32]. Other

studies also pinpointed the importance of residues in the loop 2 on the specificity of α-

conotoxins binding at nAChRs [33]. Moreover, it has been demonstrated that the C-terminus,

R10 and connective asparagines (in position 12 and 13) of LsIA are involved in distinguishing

between neuronal subtypes, namely, α3β4 [28], α3β2 and α7 nAChR [29], with mutations

based on these residues showing distinct efficacies. It has also been shown that carboxylation

of the C-terminus of α-conotoxin LsIA (LsIA#) reduces potency to hα7 nAChR relative to

naturally LsIA, but enhances potency at rat α3β2 nAChR [29], highlighting the potential

efficacy of C-terminal modification to alter nAChR subtype selectivity. Additionally, apart from

most α-conotoxins, some peptides from other conotoxin subfamilies also retain a naturally

amidated C-terminus, such as μ-conotoxins GIIIA [34] KIIIA [35] and ω-conotoxins CVID [36]

and MVIIA [37] that target voltage-gated ion channels and other proteins. Among them,

MVIIA (ziconotide), is a strong analgesic for treating severe chronic pains, which has been

approved by the US Food and Drug Administration (FDA) [38], meanwhile other conotoxins,

such as ω-CVID [36] and α-Vc1.1 [39] for treatment of neuronal and cancer-related pains [40]

are under clinical trial, however which has been discontinued for Vc1.1 due to species

selectivity issue [41, 42]. The possibility of altering the selectivity or potency via C-terminal

carboxylation of these conotoxins, of known therapeutic importance, provides an additional

exciting avenue for the development of novel conotoxin derivatives which may serve as

molecular probes or new treatments.

Figure 3.1. Ribbon conformation of LsIA binding to human α7 nAChR. (A) The conformation of LsIA with four conserved cysteines. (B) Top view of LsIA (green) binding at the ligand binding domain of five interfaces between adjacent α7 subunits based upon the template (2BR8), where the subunits are shown in cyan color. (C) Side view of LsIA anchored to one interface of two adjacent α7 subunits.

Previous studies have mainly provided insights into determinants that affect the binding of

some α-conotoxins to hα7 nAChR, in order to better understand the binding mechanisms of

ligands [23, 43], and structure-function relationships of α7 nAChR and α-conotoxins [1, 44].

Additionally, much work has focused on mutating and synthesizing novel neuropeptides

binding at the receptors with improved efficacy and specificity, experimentally or proposed

via computational predictions [20, 26, 29, 45].

It is widely accepted that α-conotoxins bind at the interface between two adjacent α7

subunits [18, 46-48], where aromatic pockets are present for ACh targeting [49]. One example

of binding conformation of α-conotoxin LsIA to hα7 nAChR is shown in Figure 3.1B,C.

Compared with AChBP, highly conserved amino acids of α7 nAChR are contained within 7

loops, for which 3 loops (A–C) are located on the principal face, also termed the (+) face, and

4 loops (I–IV) on the complementary face, denoted as the (−) face [43], constituting the

binding site at the ECD of α7 nAChR [50]. The key residues of these loops, therefore, play

significant roles in interactions with key residues of α-conotoxins. The most important residue

pairs that affect the stabilization of hα7 bound by α-ImI and PnIB complexes, respectively, are

ImI-R7 and α7(+)-Y217, and PnIB-L10 and α7(+)-W171 [23, 43], via cation-π and hydrophobic

interactions respectively. In contrast, residues on the (−) face of rat α7 subunit exert a pivotal

impact on the binding affinity of α-conotoxin TxIA (A10L) [51], and residues at the (−) face of

human α7 nAChR are also implicated in the determination of selectivity of the toxin to the

receptor [23]. Furthermore, in recent studies, Abraham et al. determined the co-crystal

structure of Lymnaea stagnalis (Ls) AChBP and LsIA complex (PDB ID: 5T90) [28], which was

used as the template for building comparative models of LsIA bound to human α7, α3β2 and

α3β4 complexes. However, a knowledge gap exists regarding the important role of the C-

terminal end of α-conotoxins in their selectivity and stabilization to nAChRs at the atomic

level, which needs further investigation to understand and fully exploit the potential of C-

terminal modification in designing novel conopeptides for nAChR as selective probes and

potential drug leads.

In the present study, the interaction mechanism of both LsIA and LsIA# with hα7 was

elucidated using molecular dynamics (MD) simulations together with a number of

computational analysis methods. The unique intramolecular interactions within LsIA favored

particular conformational motifs for LsIA, and thereby facilitated unique interactions with

respective residues at the LBD of α7 nAChR which are absent in LsIA#. Key pairs of residues

have been investigated which are known to be highly implicated in the enhancement of

binding potency of LsIA targeting α7 nAChR, compared with its LsIA analogue bound form. An

inter-residue contact network analysis approach was applied for determination of possible

disruptions or changes in inter-subunit interactions due to binding of LsIA and LsIA#, and

proposed additional possible structural characteristics which may be related to the higher

potency of LsIA relative to LsIA#.

3.2 Methods

3.2.1 Homology Modelling

The sequence alignment of hα7 nAChR (UniProtKB: P36544) to Ac-AChBP was determined

using multiple sequence alignment of human nAChR subunits α4-10 and β2-4 with that of

Aplysia californica acetylcholine binding protein (Ac-AChBP). Subsequently, homology models

of LsIA bound to the pentameric ECD of hα7 were generated using Modeller9v6 [52]. The

coordinates of Ac-AChBP co-crystallized with the double mutant α-conotoxin

PnIA[A10L,D14K] (PDB ID: 2BR8) [6] serve as a template. One hundred models were

generated, after which the model with the top DOPE [53] score was selected for simulations.

The modification of the C-terminus of LsIA was conducted using Deep View (Swiss-PdbViewer,

v4.1, Swiss Institute of Bioinformatics, Lausanne, Switzerland) [54] for changing the naturally

amidated C-terminus to the carboxylated one.

The overall quality of the homology model of α7 nAChR is comparable to that of the original

structure (2BR8), given that the overall quality factor identified via ERRAT [55] for α7 nAChR

bound by LsIA paradigm and the template is 86.1 and 89.1, respectively. The main structural

violation of α7 nAChR model dominated in regions of Cys-loop (residue D153 to Q162) and

α1-β1 loop (residue N39 to N46), which are not located at the LBD of α7 nAChR targeted by

LsIA.

3.2.2 Molecular Dynamics Simulations

The homology models of LsIA and LsIA# anchored to human α7 nAChR complexes, were

centered in rectangular boxes with an initial size of 99.593 × 99.593 × 99.593 Å3 and 99.603 ×

99.603 × 99.603 Å3, respectively, for fully submerging the α7-LsIA complex structures in water.

Both of the systems were energy minimised using GROMACS 4 (v4.6.5, Science For Life

Laboratory, Stockholm University and KTH, Stockholm, Sweden; and Biomedical Centre,

Uppsala, Sweden) [56] with the CHARMM27 force field. Both systems were solvated with

water by TIP3P water models [57]. Na+ and Cl− were added for achieving a concentration of

0.15 M and neutralizing the charged amino acid residues in the two systems. For LsIA-α7

complex, 80 Na+ and 70 Cl− were added, whereas 85 Na+ and 70 Cl− were introduced in the

LsIA# bound complex. Both systems were energy minimized for up to 1000 time steps utilizing

the steepest descent minimisation algorithm. The temperature was regulated by the velocity-

rescale method [58, 59], increasing from 0 to 300 K over 100 ps in NPT ensembles with

pressure under 1 atm. In order to restrain all protein molecules in the systems, the LINCS

algorithm [60] was implemented with a time step of 2 fs. The cutoff of van der Waals (vdW)

interaction was set to 12 Å with a smooth switch of 10 Å, meanwhile, the electrostatic energy

was calculated via Particle Mesh Ewald (PME) method [61]. Subsequently, the production

runs under constant NPT conditions were conducted for at least 30 ns with restraints

removed. Production simulations of both LsIA and LsIA#-α7 complexes were performed using

different random seeds to set random initial particle velocities, with 12 independent

simulations for each system. Since α7 nAChR contains five equivalent interfaces, the total

amount of trajectory collected for analysis of α7(+)α7(−) interfaces were at least 3.6 µs (2

systems × 12 independent simulations × 30 ns × 5 interfaces). Analyses were performed either

on data obtained as an average over the independent simulation trajectories, or over

independent trajectories as well as interfaces, as noted in the Results and Discussion. In this

study, the frames were saved every 10 ps for analyses for all trajectories.

3.2.3 Structural Visualization and Analysis

Conformations of the LsIA-α7 complexes obtained via MD simulations, were identified and

analysed using a number of computational methods. Visual Molecular Dynamics (VMD) [62]

was used for observing MD simulation results and production of figures of binding

conformations. Root-mean-square deviation (RMSD) and RMS fluctuation (RMSF) were

calculated using the GROMACS suite of analysis tools; and were used for evaluation of

structure stability, flexibility and rigidity of the protein complexes. The average structures

generated from the rmsf command implemented in GROMACS were examined using

Cytoscape [63], which is a program for visualization and comparison of inter-residue

interaction networks in protein complexes. The mindist command implemented in GROMACS

was used for determining numbers and distances of contacts between interactive residues of

ligand and receptor with a cutoff of 4.5 Å. In all Results and Discussion sections, the five

subunits of α7 nAChR are labelled as chain A to E, while the five ligands of LsIA/LsIA# binding

at the interfaces between the α7(+)α7(−) subunits are labelled as chain F to J.

RMSD, RMSF and gmx_mindist analyses are based upon different manipulations of MD

simulation trajectories. RMSD of each nAChR subunit/ligand was the average based on 12

seeds of MD simulation trajectories with each seed running for around 40 ns with the segment

beyond 30ns cut off. Therefore, the total amount of time shown in Figure 3.2A to D is 30 ns.

The RMSF of each subunit/ligand was calculated based upon the original MD simulation data,

where 7a and 7c are short for LsIA and LsIA# bound form, respectively. The gmx_mindist of

each paired residues was calculated based upon the whole trajectory, with the first 10 ns of

each individual seed being cut off.

3.3 Results

3.3.1 Homology Modelling of LsIA and LsIA# Anchoring to Human α7 nAChR

The starting homology models of α7/LsIA were constructed using 2BR8 [6] as a template, and

subsequently simulated. Remarkably, a recent co-crystal structure is available of Ls-AChBP

bound to amidated α-conotoxin LsIA (PDB code: 5T90 [28]), which showed minor structural

differences in AChBP compared to all other conotoxin-Ac-AChBP crystal structures to date,

possibly due to species differences between Ls- and Ac-AChBP. The authors constructed

homology models of LsIA bound to α7 nAChR based on Ls-AChBP, and comprehensively

elaborated and tested key interactions identified in the model in a number of elegant

experiments. Nonetheless, the minor structural differences between Ls- and Ac-AChBP, and

the availability of co-crystal structures of 4/7-conotoxins bound to both species isoforms,

afford an opportunity to explore an alternative, possible conformation of LsIA-α7 complex

based on Ac-AChBP, which shares a comparable (though somewhat lower) sequence

similarity to hα7. In this chapter, the present LsIA-α7 models based on Ac-AChBP were

compared and contrasted to those based on Ls-AChBP where relevant.

3.3.2 Stability and Flexibility of the LsIA and LsIA# Bound α7 nAChR Complexes

Time series plots of RMSD, calculated for the backbone atoms and averaged over 12

simulations, for LsIA and LsIA#-α7 nAChR complexes, are shown for the α7 subunits in Figure

3.2A,B, and for the toxins (LsIA and LsIA#) in Figure 3.2C,D. These RMSD plots are qualitatively

similar with those of previous simulations of nAChR-toxin complexes [14, 18, 64]. The

structure of LsIA (Figure 3.2C) and hα7 nAChR (Figure 3.2A) complex becomes stabilized, on

average, after 20 ns, while the LsIA# binding at the ECD of the receptor does not achieve its

stabilization until 27ns (Figure 3.2B,D), and appears to continue to exhibit structural drift

towards the end of the simulation period. Thus, there is greater variation in the RMSD plots

for the LsIA#-α7 system compared to LsIA, suggesting that the LsIA#-α7 complex might be

somewhat less stable. This demonstrates a potentially more stable structure of LsIA binding

to α7 nAChR complex, which will be discussed further in the sections below.

However, the RMSD results show an asymmetric pattern in chain H in LsIA (Figure 3.2C), which

demonstrates a divergent pattern with higher RMSD values (around 0.425 nm) versus the

other 4 chains of LsIA, whereas there is no such pattern present in the carboxylated

counterpart (Figure 3.2D). Thus, the LsIA-bound form provides a more steady structure over

its analogue binding complex, while the rigidity of protein structure varies in distinct portions

of the protein complex (Figure 3.2C,D). The RMSD plot of chain H (ligand) in LsIA is different

from the other chains of the LsIA (Figure 3.2C). This may suggest a potentially distinct

interaction between chain H and corresponding LBD of α7 nAChR compared with the others.

By contrast, the substantial differences between the five LsIA# toxins (Figure 3.2D) were not

observed, nor is there any substantial difference among the chains of α7 nAChR bound by

LsIA and LsIA# (Figure 3.2A,B). Overall, the RMSD results indicate that LsIA-α7 reach stability

earlier, with less fluctuation in RMSD per chain compared to LsIA#-α7 complex. However, for

the amidated complex, there is some asymmetry in the RMSD, with one of the LsIA monomers

undergoing substantially higher structural deviation compared to the other toxins (chains)

bound at α7 nAChR interfaces. RMS fluctuation (RMSF) plots are shown for the LsIA and LsIA#-

bound α7 receptor residues in Figure 3.2E,F, and for the toxin residues in Figure 3.2G,H,

respectively. The absolute RMSF values for the α7 nAChR (Figure 3.2E,F) for both LsIA and

carboxylated bound forms are consistent with previous studies on hα7. These plots

demonstrate the greatest flexibility in the Cys-loop, and are similar to the RMSF results of

several previous studies, such as that of α-ImI bound form, by Yu et al. via dynamic simulations

[18]. Yu and colleagues found that the Cys-loop RMSF of hα7 increased upon bound by ImI,

compared with the apo hα7 nAChR. There is also relatively high flexibility of β1/β2 loop and

β10 strand, similar to previous studies [65].

In order to clearly show the differences in α7 nAChR residue RMSF when bound to LsIA

compared to LsIA# binding, Figure 3.2I shows “Diff in RMSF”, taken by subtracting the

receptor RMSF values of the carboxylated complex from those of the amidated complex.

Thus, for example, residues with positive “Diff in RMSF” values indicate higher flexibility for

amidated relative to LsIA#-α7 complex. The main difference is that there is substantially

higher flexibility in the Cys-loop region of α7 nAChR subunit for LsIA bound complex compared

to LsIA# bound form. This region forms the main contact between the ECD and

transmembrane domain (TMD; though absent in the present simulation) in the intact α7

nAChR structure. Another region that exhibits marginally higher RMSF for LsIA-bound α7 is

the C-loop region, which forms part of the canonical agonist binding site and, in the α-subunits

of nAChRs, is composed of a β hairpin with a twin Cys motif at the edge of the loop.

There are also slightly higher RMSF values in regions of the β1/β2 loop and C-terminus of β10

strand that are related to pore gating in the TMD. Both regions are related to the contacts

between ECD and TMD for mediating the open and closed states of the pore in TMD, but via

distinct pathways. The increased flexibility of the former that interacts with M2–M3 linker

may cause a conformational change of M2 α-helices in the TMD [66-68], which is a known

role of the Cys-loop in pore gating of Cys-loop receptors [69, 70]. The latter is associated with

the N-terminus of M1 in the TMD via covalent interactions [71].

Additionally, there is reduced flexibility in the N-terminal α1 helices of α7 nAChR bound by

LsIA relative to LsIA#. This region has been demonstrated to form a key Fab35 antibody

binding region for muscle-type nAChR [72], but its role in neuronal nAChRs is presently less

clear. The impact of conotoxin binding on nAChR-antibody linkage is an intriguing future

avenue of research. The absolute RMSF plots for the toxin residues of LsIA and LsIA# are

shown in Figures 3.2G and H respectively. Both exhibit similar patterns of RMSF, with

especially high fluctuation at N- and C-terminal residues. As with the RMSD behavior

described above, for LsIA, there is some asymmetry in RMSF, with one monomer (labelled as

chain H in Figure 3.2G) exhibiting substantially higher fluctuation in the N-terminus compared

to all four of the other monomers. Thus, the source of asymmetry appears to be localized

mainly to the disordered and highly flexible loop 1 residues, S1 and G2.

To directly compare the effects of carboxylation, the “Diff in RMSF” plot for the LsIA residues

is shown in Figure 3.2J. It can be seen that residues A8 and C9 are more flexible for the LsIA,

with positive “Diff in RMSF” values. However, the main impact is that LsIA has a more rigid

structure in residues of loop 2 (except for N12) and in the N-terminus of LsIA (Figure 3.2G,H,J).

Figure 3.2 Analysis of stability of human α7 nAChR over molecular dynamics (MD) simulations in the LsIA and LsIA# bound types. Cα atoms in β strand RMSD of five α7 subunits in the LsIA (A) and LsIA# (B) bound forms, respectively. Cα atoms in β strand RMSD of five ligands (LsIA) binding in the ligand binding domain (LBD) in LsIA (C) and LsIA# (D), respectively. Cα RMSF of five α7 subunits in the LsIA (E) and LsIA# (F) bound forms, respectively. Cα RMSF of five ligands (LsIA) binding in LBD in LsIA (G) and LsIA# (H), respectively. The black curves demonstrate the mean values of RMSD/RMSF of the five interfaces/ligands in the LsIA binding complexes. (I) Differences of Cα RMSF between LsIA and LsIA# bound α7 subunits. (J) Differences of Cα RMSF between LsIA and LsIA# (LsIA mean value subtracts that of LsIA#).

3.3.3 Residues Undermine the Binding of LsIA at α7 Complementary Face versus LsIA#

To explain differences in RMSD and RMSF, how amidation and carboxylation change the

extent of residue pairwise contacts between LsIA and α7 was examined. Figure 3.3A shows

the absolute total number of contacts formed by each residue of LsIA (green) and LsIA# (pink)

with α7 nAChR, as well as differences in pairwise interactions (amidated minus carboxylated,

Figure 3.3B). The most significant interactions formed by LsIA and LsIA# with α7 subunit,

which predominantly occurred in R10 and V11 in loop 2 and P7 in loop 1 of LsIA (Figure 3.3A),

which is also known to exert a great impact on selectivity to nAChR by the ligand [28,61].

Other minor interactions are established by LsIA-C17 and the residues on the receptor,

significantly contributing to the receptor contacts by LsIA (Figure 3.3A,B). In contrast,

although LsIA-A8 makes substantial contacts with the receptor, its impact on selectively

binding to the target is much lower when compared with other residues of LsIA, as explained

below.

Figure 3.3. A total number of contacts and the variations between LsIA and LsIA# bound types. (A) The total number of contacts for amidated (green) and carboxylated (pink) LsIAs with human α7

nAChR. (B) The significant changes (amidated minus carboxylated) in the number of contacts between LsIA and LsIA# bound types. (C) The interactive residues of the LsIA and α7 nicotinic receptor complex.

Figure 3.3B shows differences in the number of α7 nAChR interactions with the LsIA residues.

The effects of gaining contacts for LsIA, relative to LsIA#, outweighed those of losing contacts

at the LBD of hα7 nAChR subtype in total, as also suggested by Inserra et al. in their

experimental results [29]. However, there were no significant changes appearing in loop 1 of

LsIA, while most of the variations of contacts took place in loop 2 and C-terminus of the toxin

(Figure 3.3B). Since loop 2 makes contacts mostly with α7(−) (Figure 3.3B,C), this pattern

indicates that the C-terminal amidation/carboxylation state mainly affects interactions with

the α7(−) face. Enhancement in toxin-receptor contacts for the amidated form is qualitatively

consistent with this toxin’s higher potency compared to the LsIA#. Two residues, namely R10

and C17, in loop 2 of LsIA show substantially higher contacts for the amidated form (Figure

3.3B). This is consistent with the predicted importance of LsIA-R10 in forming close hydrogen

bonds with the hydrophilic pocket of α7 in previous studies [28]. However, compared with

the LsIA# bound form, although LsIA-R10 shows a higher absolute number of contacts with

α7 nAChR, in relative (percentage) terms, LsIA-C17 has the greatest enhancement, with an

increase of 45% higher contact number compared with the corresponding residue of LsIA#

(not shown). On the contrary, the most significantly impaired interaction was attributed to

G2 with around 20% decrease (not shown). Additionally, P7, V11 and N12 of LsIA show minor

reductions for LsIA and α7 nAChR complex, but the contacts of asparagine and proline in

positions 15 and 14 of the toxin with the receptor are slightly enhanced in the amidated

ligand-bound form.

3.3.4 LsIA Favors Hydrogen Bonding and Cation-π Interactions between R10 and the α7(−)

Subunit

The specific toxin-receptor pairwise contacts were examined, which may play an important

role in improving the selectivity of LsIA binding to human α7 nAChR, namely LsIA-C17 with

α7(−)-Q79 and Q139, and LsIA-R10 with α7(−)-L141 and W77 (Figure 3.3C). In contrast, P7 and

V11 of LsIA# possess preferential contacts with residues on the principal (+) face of α7. The

interactions which favor the LsIA will be firstly discussed, followed by those that favor the

binding of LsIA# at α7 nAChR interfaces.

For LsIA, a significant interaction is established between LsIA-R10 and α7(−)-W77 and L141

via cation-π and hydrogen bond interactions, respectively. Compared with the carboxylated

analogue and α7 nAChR complex, a unique sandwich-like conformation [62], composed by

LsIA-R10 and α7(−)-W77 (3.9 Å) and L141 (1.7 Å/2.9 Å) (Figure 3.4A,B), was determined. The

preferential interactions involving hydrogen bonding between R10 and the hydrophilic

environment of the α7(−) subunit are in qualitative agreement with the model proposed by

Abraham et al. [28].

A perpendicular (T-shaped) position of the LsIA-R10 sidechain with respect to α7(−)-W77 was

observed in both LsIA bound conformations (Figure 3.4C,D), wherein both cation-π

interactions are likely to be relatively weak in this arrangement (4.0 Å in LsIA; 5.0 Å in LsIA#).

The intramolecular interactions between C4 and R10 may account for this situation, as this T-

shaped motif often occurred when LsIA-C4 lost its persistent contacts with the guanidinium

moiety of R10. Thus, this perpendicular conformation is relatively rare in the LsIA bound form,

due to the tight intramolecular interactions.

Figure 3.4. Side view of binding conformation of LsIA-R10 to the ligand binding domain (LBD) of human α7 nAChR. (A) The sandwich-like motif of LsIA-R10 with α7(−)-L141 and W77 and (B) binding conformation of LsIA# with corresponding residues on α7(−) face. Perpendicular-like binding conformation of LsIA-R10 with α7(−)-L141 and W77 in amidated (C) and carboxylated (D) LsIA bound forms, respectively. The purple and green dashed lines indicate the residues of LsIA with impaired and improved interactions with respective residues of the receptor, whereas the receptor subunits bound by LsIA (green) and LsIA# (pink) are shown in transparent cyan and silver colors, respectively. (E) Distances (Å) of the corresponding interactions between residues of LsIA/LsIA# and the receptor in ligand-bound forms were predicted via gmx_mindist.

Other, more minor interactions identified in the present simulations include those between

the LsIA-C17 with (−)Q139 and (−)Q79 (Figure 3.5A,B) as well as (−)H137 (not shown), of α7

subunit, via hydrogen bond interactions. Notably, interactions between C17 and the

respective glutamine residues of the receptor, are amongst the most significantly enhanced

contacts, compared with its carboxylated counterpart bound complex. However, it should be

noted that although the distance between LsIA-C17 and (−)Q139 is decreased, the interaction

between them is likely to be relatively weak, as a hydrogen bond interaction is not possible

due to a long distance (8.6 Å) between LsIA-C17 and (−)Q139. The proximate packing of LsIA

(Figure 3.5C) allowed a minor persistent hydrogen bond interaction between contacts of P14

and (−)Q139 (1.8 Å /3.3 Å) (Figure 3.5B,D), whereas a relatively weak vdW interaction is

established with the corresponding residues in the LsIA# binding form (Figure 3.5A).

Figure 3.5. Binding model of LsIA-P14, R10 and C17 of amidated (wt) LsIA and carboxylated (mut) LsIA into the binding pocket on the complementary (−) face of human α7 nAChR. Interactions

established by R10, P14 and C17 with corresponding residues in amidated (green) (A) and carboxylated (pink) (B) LsIA binding forms. (C) Side view of the binding conformation of the LsIA and LsIA# to α7 nAChR subunit. (D) Distances (Å) of the corresponding interactions between residues of LsIA/LsIA# and the receptor in ligand-bound forms were predicted via gmx_mindist.

3.3.5 LsIA# Favors Hydrophobic Interactions with α7 Subunit Residues via P7 and V11

The description in the above section noted the loss of LsIA contacts between LsIA-P7 and A8

with α7(+)-W171, as well as V11 with α7(−)-L131/141 (Figure 3.3B) which are, conversely,

favored by the LsIA# binding. This may be related to the shifting of the balance in LsIA

interactions with the principal (+) and complementary (−) subunits mentioned above.

In addition to α7(+)-W171 having higher contacts with LsIA#-P7 and A8 (Figure 3.6), it should

be noted that the interaction between (+)W171 and LsIA-P7 also shows the highest absolute

number of contacts in both LsIA and LsIA# bound receptors (not shown). Notably, proline in

position 7 of LsIA, which is amongst the conserved amino acids of LsIA, plays a critical role in

the formation of the α-helix structure of α-conotoxins, by strongly contributing to the binding

of α-conotoxins at nAChR through making various contacts with residues on both sides of the

binding site [1, 43, 73].

Figure 3.6. Side view of binding conformation of LsIA into the binding pocket with the interaction between human α7(+)-W171 and LsIA-P7, A8 and V11, respectively. (A) The conformation of LsIA anchored to the principal (+) face of α7 subunit. (B) The conformation of LsIA# anchored to the (+) face of α7 subunit.

3.3.6 α7-Y217 Forms Close Contacts with Both LsIA and LsIA#

As the principal determinant for binding of ImI is the interaction between aromatic ring of

α7(+)-Y217 and ImI-R6 [43], the interaction between α7(+)-Y217 and respective residues of

LsIA was investigated in this study. However, there is no potential interaction between R10

of LsIA or LsIA# and the corresponding residue of α7 nAChR; nor are apparent variations of

interactions between these two residues detected, due to the position of LsIA-R10 being

relatively far away from α7(+)-Y217 (Figure 3.7A,B). Nonetheless, several contacts were

identified in the present simulations, between α7(+)-Y217 and LsIA-A8, LsIA-C9 and the two

asparagine residues in loop 2 of the toxin, which were observed in both LsIA and LsIA# bound

forms. Among them, the most notable interaction is established by LsIA-C9 with α7(+)-Y217.

C9 is in proximity with (+)Y217, which allowed vdW interaction that could be an important

determinant for binding LsIA at α7 nAChR (Figure 3.7A,B). Additionally, hydrophobic

interaction formed by α7(+)-Y217 and LsIA-A8 may also be related to the anchoring of LsIA

and its variant to hα7 nAChR. However, there are no obvious changes in contacts between

(+)Y217 and C9 regarding the C-terminus carboxylation of LsIA, whereas the degree of

hydrophobic interaction between LsIA-A8 and (+)Y217 (3.9 Å) is slightly higher in LsIA receptor

complex compared with the LsIA# bound type (6.2 Å) (Figure 3.7). In contrast, the vdW

interaction established between LsIA-G2 and α7(+)-Y210 is reduced, possibly owing to the

LsIA being more deeply inserted into ligand binding pocket, together with a more flexible

backbone of the N-terminus of LsIA, relative to its carboxylated analogue bound form.

Figure 3.7. Top view of binding conformation of amidated (wt) and carboxylated (mut) LsIAs at the principal (+) face of human α7 nAChR. (A) The conformation of LsIA anchored to the (+) face of α7 subunit. (B) The conformation of LsIA# anchored to the (+) face of α7 subunit. The black dashed lines demonstrate the interaction occurred in both LsIA bound forms. (C) Distances (Å) of the interactive residues of LsIA and α7 nAChR complexes.

3.3.7 An Intra-molecular Salt-bridge in LsIA# is Retained upon Binding to α7

The present simulations indicate that the LsIA#-C17 forms a stable salt bridge (2.0 Å) with R10

(Figure 3.8A), the interaction of which was also discovered by Inserra and colleagues in

experiments [29] in the solution NMR structure of the free solvated LsIA. Thus, our present

simulation confirms that this C17-R10 salt bridge remains intact when LsIA is bound to α7.

In addition to this salt bridge interaction, our simulations identified a number of additional

persistent hydrogen bonds which may further differentiate LsIA from LsIA#. Another

persistent hydrogen bond formed by LsIA-C17 and the backbone carbonyl group of N15 in

both types of LsIA bound to α7 was also observed in the present simulations (Figure 3.8). This

intramolecular interaction is likely involved in drawing the C-terminus closer to loop 2 of α-

helix of LsIAs. Furthermore, the LsIA-N15 also established hydrogen bonds with the backbone

of C17, with distances of 2.7 Å and 3.2 Å (Figure 3.8B). Finally, a persistent intramolecular

hydrogen bond was also formed by R10 with the backbone of C4 (2.4 Å) in both LsIA and LsIA#.

Figure 3.8. The intramolecular interactions within C-terminal carboxylated (mut) and amidated (wt) LsIAs by binding to human α7 nAChR. (A) The conformation of LsIA# aligned with LsIA. The conformation of LsIA# is shown in transparent green color. (B) The conformation of LsIA aligned with LsIA#, which is shown in transparent pink color. (C) Distances (Å) of intramolecular interactions between residues of LsIA and LsIA#.

3.3.8 Changes in Inter-subunit Contacts on the Five Interfaces of Adjacent α7 Subunits

The binding of LsIA and LsIA# to hα7 nAChR also elicits variations of contacts between

residues at the interfaces between α7 subunits, which were examined using the Cytoscape

[63] molecular analysis method. It is hereby proposed that structural disruptions to the

integrity of inter-subunit contacts may have implications for the potency of conotoxins

against nAChRs. Figure 3.9 shows the network representations of changes in inter-subunit

contacts for the nodes (boxes) that indicate interface residues, numbered and labelled by

receptor chain (A–E). Each sub-figure shows residues at a different interface (e.g. Figure 3.9A

shows the chain A–B interface, while Figure 3.9B shows the chain B–C interface). The edges

(lines) connecting the nodes indicate possible contacts between the interface residues; red

dotted lines indicate contacts that are closer for the carboxylated form, green dashed lines

indicate contacts that are closer for the amidated form, while black lines indicate contacts

that are similar for both forms. The nodes are color coded as follows: residues are colored red

if they mainly have closer contacts with opposing interface residues at the carboxylated

bound form, green if they mainly have closer contacts with opposing interface residues at the

amidated bound form, and gray if their overall contacts with opposing residues are

unchanged between the two forms.

Overall, the predominance of grey nodes and edges shows that LsIA binding generally caused

similar inter-subunit contacts of α7 nAChR with the LsIA# bound form. There is a slightly closer

association between the interfaces for the amidated form, since all interfaces contain some

residues that have closer association at the amidated form indicated by the presence of green

nodes in Figure 3.9A to E; but the interface between chains B and C does not have any residues

that associate more closely at the carboxylated form, indicated by the absence of red nodes

in Figure 3.9B. Nonetheless, certain interactions within interfaces of the carboxylated bound

form which are weakened may have implications for LsIA binding to the LBD of the receptor.

On the other hand, interface contacts weakened by binding of LsIA may also have implications

for the inhibitory potency of this form of the toxin. Below, more details regarding which

residues were involved in these disruptions in inter-subunit contacts, are discussed.

Figure 3.9. Comparison of differences in the interactions between principal (+) and complementary (−) interfaces of human α7 nAChR after binding of LsIA and LsIA#. (A,B,C,D, and E) Determination of interactive residues at five interfaces of α7 nAChR via Cytoscape program. (F) Conformation of changes in the secondary structure of (+) face of LsIA bound receptors, where LsIA bound face is shown in transparent color.

In general, acidic residues (aspartic and glutamic acids) at the inter-subunit interfaces have

higher contacts with other residues for the LsIA bound form. This can be demonstrated by

considering certain residues situated in the N-terminal α1 helix(−) which interacted with the

residues on loop between α-helix and β1(+) strand, and resulted in pulling the N-terminal

helices into a closer proximity in both LsIA and LsIA# forms, and the gap between interfaces,

therefore, was slightly shortened by toxin binding. For example, α7(−)-Q26, were dominantly

involved in interacting with both (+)V44 (chains A&B and chains E&A) (Figure 3.9A,E) and with

(+)E41 (chains B&C and chains C&D) (Figure 3.9B,C), respectively. Interestingly, repulsive

electrostatic interaction established by (+)E120 and (−)D123 (chains C&D and chains E&A) in

carboxylated bound type, may elicit separation of the conjunctive interfaces (Figure 9C,E). In

this case, residues on LsIA# probably lost contacts with the receptor due to the proximity of

several negative charges contributed by LsIA as well as the acidic residues of the two

interfaces of the receptor. Additionally, on the boundary surface between chain D(+) and E(−)

(Figure 3.9D), (−)G194, (−)E195 and (−)I191 formed more interactions with (+)Q70 and

(+)N116 on β2 and β4 strands, respectively, in carboxylated bound form.

Finally, another trend exhibited by examination of the network diagrams shows that several

clusters of hydrophobic residues formed closer contacts with each other in the amidated

form. Specifically, (−)P143 on β6, established hydrophobic interaction with (+)W171 (chains

B&C and chains D&E) on B-loop(+), as shown in Figure 3.9B,D, which may result in a longer

distance between (+)W171 and (+)S172 with P7 and A8 of LsIA by triggering the shifting of B-

loop(+) toward β6(−). Other unique hydrophobic interactions were determined between

(+)A118 and (−)M63 and (−)I145 respectively, on chain A and B interface (Figure 3.9A). Among

them, methionine residue is situated close to β1/β2 loop, that was implicated in the

conformational changes at the TMD of AChBP [74].

Taken together, the conclusions that can be drawn are that (1) binding of LsIA favors inter-

subunit interactions involving aspartic acid and glutamic acid residues near the agonist-

binding site; and (2) binding of LsIA# favors a closer association between hydrophobic

residues centred on (−)P143 and (+)A118, near the Cys-loop regions at the juxtamembrane

domain. However, further studies are required for exploring the effects of the inter-α7

subunit contacts on stabilization and specificity of LsIA binding to the receptor.

3.4 Discussion

3.4.1 LsIA Binding Causes Higher Fluctuations in C-loop and the Juxtamembrane Cys-Loop

Regions

The LsIA-α7 complex achieved stability earlier than LsIA# bond form (Figure 3.2). However,

the asymmetric pattern of N-terminus of LsIA was observed in the amidated bound complex,

possibly due to the high fluctuation at this region of the toxin relative to LsIA#. A similar

asymmetric pattern also occurred in the RMSD plot of apo chicken α7 nAChR, determined by

Yi and colleagues [15]. Notably, the rigidity in loop 2 of LsIA was also identified, possibly due

to differences in both toxin-receptor complexes, as well as intramolecular LsIA contacts, as

detailed below.

For the α7 nAChR, the enhanced flexibility in the Cys-loop regions could be related to changes

in α7 nAChR gating activity, and could provide an indirect measure of the potential inhibitory

effects of conotoxins bound to nAChR ECDs. Further work involving the whole α7 nAChR

bound to conotoxins will likely be required to examine the direct impact of toxin binding on

the pore region. Also, the minorly enhanced rigidity of C-loop at α7 nAChR was also observed

upon bound by LsIA versus LsIA#. This enhanced fluctuation is in accordance with previous

studies on nAChRs, including those for apo α7 and α7-antagonist-bound complexes [65].

Nonetheless, the higher RMSF of both C-loop and the juxtamembrane Cys-loop regions, for

LsIA, is consistent with the present notion that movements in these two areas are strongly

coupled, and is required for transmission of structural effects due to agonist (or antagonist)

binding from the ECD to the transmembrane pore. The higher RMSF for both regions for LsIA

bound α7 nAChR suggests that this natural form of the conotoxin may have a greater impact

on the ligand binding and subsequent TMD coupling mechanism of α7 subtype compared to

the less potent, carboxylated form.

3.4.2 Carboxylation at LsIA C-terminus Causes Indirect Loss of Contacts between Key

Residue R10 and the α7 Complementary Face

In general, the LsIA shows marginally enhanced contacts with α7 nAChR versus the LsIA#

bound form. For the LsIA, an increase in the number of contacts at R10 and the C-terminus

(C17) with α7 residues (Figure 3.3B) is likely responsible for its higher rigidity in loop 2 (Figure

3.2J, RMSF plots) compared to the carboxylated form. Since R10 is known to be a key residue

responsible for LsIA inhibition of nAChRs [28], our present simulation results indicate that

introducing a negative charge via carboxylation at the C-terminus not only causes reduction

in α7 nAChR contacts directly with C17, but also an indirect loss of contact at R10, and this

latter disruption may play the key role in reduced potency of the LsIA# at α7 nAChR.

3.4.3 Key Pairwise Interactions Favor the Binding of LsIA at α7 nAChR versus LsIA#

The sandwich-like cation-π interactions formed by LsIA-R10 with α7(−)-W77 and L141 might

serve to stabilize the binding of LsIA at α7(−) subunit, leading to a deeper burial of LsIA into

the binding pocket. A similar ‘sandwich’ motif has also been determined by Grishin and

colleagues through computational methods [63], demonstrating a particular arrangement for

AulB-F9 and residues, K61 and W59, on the β4 subunit of α3β4 nAChR. However, the

interactions involved in the present study also include hydrogen bond formation.

Furthermore, the cation-π interaction has also been demonstrated between R7 of ImI and

aromatic residues, such as (+)Y217, (+)W171 and (+)Y115, on the binding pocket of hα7 nAChR

through computational docking approach [64]. In addition, as (−)W77 significantly affects the

specific binding of ImI to hα7 nAChR and was shown to be crucial for modulating the response

of rat α7 nAChR to the existence of agonist [43,65], further investigation on binding of LsIA at

hα7 nAChR may help to explicate the important role of (−)W77 in LsIA binding. Additionally,

the dramatically enhanced hydrogen bond interactions between LsIA-C17 and (−)Q79 also

reinforce the binding of the toxin at α7 nAChR versus the binding of LsIA#.

3.4.4 The Hydrophobic Interactions Undermine the Binding of LsIA at α7 nAChR versus LsIA#

From the inter-residue contact results, the highest number of contacts are formed between

α7(+)-W171 and LsIA#-P7 due to hydrophobic interaction between (+)W171 and the imino

ring of P7, the interaction of which has also been identified between α7(+)-W171 and PnIB-

P6 and P7 in previous experimental studies [4, 23]. Furthermore, V11 of LsIA formed

hydrophobic contacts with the residues (−)L131/141, further contributing to the enhanced

hydrophobic interactions by LsIA# relative to LsIA. By contrast, α7(+)-W171 also formed a

vdW interaction with LsIA-A8 versus LsIA#, however which is very weak.

3.4.5 Intra-molecular Interactions Play a Role in the Loss of Key Toxin-receptor Contacts

While C-terminal amidation is known to be important for the folding of α-conotoxin ImI

through its effects on disulphide pairing [75], it may also directly or indirectly affect the

binding stability of LsIA to hα7 nAChR via intramolecular interactions. The intra-molecular

interactions between LsIA#-R10 and C17 may substantially interfere the nicotinic interactions

by LsIA# versus LsIA. By contrast, the additional hydrogen bonding pattern of LsIA-N15 and

C17 may play a role in the greater rigidity of C17 and N15 shown in the RMSF plots for LsIA

over its carboxylated analogue (Figure 3.2J). At last, the hydrogen bond formed between C4

and R10 may exert an impact on maintaining both LsIA and LsIA# in their characteristic fold.

3.4.6 Inter-subunit Interactions Play a Role in the Loss of Key Toxin-receptor Contacts

As mentioned previously, the acidic residues, (+)E41, (+)E120 and (−)D123, at the inter-

subunit interfaces may favor the binding of LsIA to α7 nAChR, versus LsIA#, associated with

the hydrophobic interactions of (+)W171-(−)P143, (+)A118-(−)M63 and (+)A118-(−)I145 at

LsIA and α7 nAChR complex (Figure 3.9). By contrast, the vdW and hydrogen bond interactions

formed by residues on F-loop with those at β2 and β4 strands of α7 may affect the reduced

contacts with LsIA#. These observed interactions of LsIA#/α7 complex may be related to a

motion which may be described as an upward swing of the F-loop and a downward shift of

the C-loop of chain E (Figure 3.9F), which has been suggested to promote agonist binding at

apo chicken α7 nAChR [15]. Furthermore, β6 strand and C-loop have been suggested as part

of key determinants for binding of ACh to α7 nAChR [18].

3.5 Conclusion

In this chapter, to understand the importance of C-terminus of LsIA for selectively binding to

hα7 nAChR at the atomic level, MD simulations were deployed to propose the determinants

for the differential activity between LsIA and LsIA# at the α7 nicotinic receptor. Simulations

suggest that LsIA-bound α7 exhibits higher fluctuations in C-loop (at the canonical agonist

binding site), as well as the juxtamembrane Cys-loop region. This variation in dynamic

fluctuations in both regions together may be related to differences in the capability of

structural changes to be transmitted from the ECD down to the TMD when amidated, but not

carboxylated, LsIA is bound to α7 nAChR. Our findings also confirm a persistent intra-

molecular salt bridge between R10 and the C-terminus of LsIA# when bound to α7 nAChR,

and may be responsible for preventing the interactions between R10 and key residues on α7

subunit, such as R10 with (−)W77 and (−)L141, via cation-π and hydrogen bond interactions

in LsIA bound form, and therefore, affects the preferential selectivity of LsIA to α7 nAChR.

This model is qualitatively consistent with the previous experimental findings which proposed

a key role for R10 in forming interactions within the hydrophilic environment of the α7 binding

pocket. Overall, carboxylation of the C-terminus reduces the selectivity of LsIA to α7 nAChR

versus its native type.

Additionally, the impacts of binding by LsIA and LsIA# on the overall structure and inter-

subunit contacts of α7 nAChR were examined using an inter-residue network analysis

approach, suggesting a clockwise tilting of C and F loops upon binding to LsIA#, which is absent

for LsIA binding. The network analysis also indicates that proximate acidic residues, aspartic

and glutamic acids, are possible at interfaces of α7 nAChR bound by the LsIA, whereas the

carboxylated counterpart disrupts these same-charge contacts while, instead, promoting

closer contacts between interfacial hydrophobic residues. This approach may facilitate the

determination of structural variations in residue interactions at inter-subunit interfaces of the

receptor upon bound by conotoxins.

Further exploration may be required for understanding specific impacts of residues of LsIA-

α7 nAChR complexes on selectivity and stability of the ligands and the influence of inter-

subunit contacts on native LsIA, via free binding energy calculation and experimental

mutagenesis studies to probe the effects of the key residues, at both LsIA and α7 nAChR,

which are proposed in this present work. Furthermore, the carboxylation of C-terminus of

conotoxins with known activity at receptors may need further investigation, such as studies

on their binding potency and interactions with nAChRs or structural stability. The predicted

molecular mechanism of LsIA binding to the neuronal nAChR, and in particular the possible

roles of the C-terminal charge state, may facilitate a better understanding of the effect of C-

terminal amidation/carboxylation on the binding potency of conotoxins at neuronal nAChRs,

as most α-conotoxins and certain neuroactive peptides in other conotoxin subfamilies also

possess a naturally amidated C-terminus in their structures. More broadly, these findings may

assist in the design of novel peptides for nAChR and selective probes with high potency for

understanding the structural and pharmacological features of nAChRs.

3.6 References

Song C, Corry B. Role of acetylcholine receptor domains in ion selectivity. Biochimica

Hurst R, Rollema H, Bertrand D. Nicotinic acetylcholine receptors: from basic science

Pohanka M. Alpha7 nicotinic acetylcholine receptor is a target in pharmacology and

D'Andrea MR, Nagele RG. Targeting the alpha 7 nicotinic acetylcholine receptor to in Alzheimer's disease pyramidal neurons. Current

1. Lebbe EKM, Peigneur S, Maiti M, Devi P, Ravichandran S, Lescrinier E, et al. Structure- function elucidation of a new α-conotoxin, Lo1a, from Conus longurionis. Journal of Biological Chemistry. 2014;289(14):9573-83. 2. et Biophysica Acta (BBA) - Biomembranes. 2009;1788(7):1466-73. 3. to therapeutics. Pharmacology & Therapeutics. 2013;137(1):22-54. 4. Hogg RC, Hopping G, Alewood PF, Adams DJ, Bertrand D. α-Conotoxins PnIA and [A10L]PnIA stabilize different states of the α7-L247T nicotinic acetylcholine receptor. Journal of Biological Chemistry. 2003;278(29):26908-14. 5. toxicology. International Journal of Molecular Sciences. 2012;13(2):2219-38. 6. Celie PHN, Kasheverov IE, Mordventisev DY, Hogg RC, van Nierop P, van Elk R, et al. Crystal structure of nicotinic acetylcholine receptor homolog AChBP in complex with an α- conotoxin PnIA variant. Nature Structural and Molecular Biology. 2005;12. Kalkman HO, Feuerbach D. Modulatory effects of α7 nAChRs on the immune system 7. and its relevance for CNS disorders. Cellular and Molecular Life Sciences. 2016;73(13):2511- 30. 8. reduce amyloid accumulation Pharmaceutical Design. 2006;12(6):677-84. Taly A, Corringer P-j, Guedin D, Lestage P, Changeux J-p. Nicotinic receptors: allosteric 9. transitions and therapeutic targets in the nervous system. Nature Reviews Drug Discovery. 2009;8(9):733-50. 10. Söderman A, Mikkelsen JD, West MJ, Christensen DZ, Jensen MS. Activation of nicotinic α7 acetylcholine receptor enhances long term potentation in wild type mice but not in APPswe/PS1ΔE9 mice. Neuroscience Letters. 2011;487(3):325-9. 11. Ulens C, Hogg RC, Celie PH, Bertrand D, Tsetlin V, Smit AB, et al. Structural determinants of selective α-conotoxin binding to a nicotinic acetylcholine receptor homolog AChBP. Proceedings of the National Academy of Sciences. 2006;103(10):3615-20. 12. Bourne Y, Talley TT, Hansen SB, Palmer T, Marchot P. Crystal structure of a Cbtx-AChBP complex reveals essential interactions between snake α-neurotoxins and nicotinic receptors. EMBO Journal. 2005;24(8):1512-22. 13. Brejc K, Sbrevea, Dijk WJV, Klaassen RV, Schuurmans M, Oost JVD, et al. Crystal structure of an ACh-binding protein reveals the ligand-binding domain of nicotinic receptors. Nature. 2001;411(6835):269-76. 14. Yi M, Tjong H, Zhou H-X. Spontaneous conformational change and toxin binding in α7 acetylcholine receptor: Insight into channel activation and inhibition. Proceedings of the National Academy of Sciences. 2008;105(24):8280-5. 15. Maricq AV, Peterson AS, Brake AJ, Myers RM, Julius D. Primary structure and functional expression of the 5HT3 receptor, a serotonin-gated ion channel. Science. 1991;254(5030):432. Harel M, Kasher R, Nicolas A, Guss JM, Balass M, Fridkin M, et al. The Binding site of 16. acetylcholine receptor as visualized in the X-ray structure of a complex between α- bungarotoxin and a mimotope peptide. Neuron. 2001;32(2):265-75.

100

Hansen SB, Talley TT, Radić Z, Taylor P. Structural and

Olivera BM, Rivier J, Clark C, Ramilo CA, Corpuz GP, Abogadie FC, et al. Diversity of

receptors and α-conotoxin PnIB. Quiram PA, McIntosh JM, Sine SM. Pairwise Interactions between neuronal α7 Journal of Biological Chemistry.

Dutertre S, Lewis RJ. Toxin insights into nicotinic acetylcholine receptors. Biochemical

Quiram PA, Sine SM. Structural elements in α-conotoxin ImI essential for binding to

Dutertre S, Nicke A, Tyndall JDA, Lewis RJ. Determination of α-conotoxin binding 17. modes on neuronal nicotinic acetylcholine receptors. Journal of Molecular Recognition. 2004;17(4):339-47. Yu R, Craik DJ, Kaas Q. Blockade of neuronal α7-nAChR by α-conotoxin ImI explained 18. by computational scanning and energy calculations. PLoS Computational Biology. 2011;7(3):e1002011-e. 19. ligand recognition characteristics of an acetylcholine-binding protein from Aplysia californica. Journal of Biological Chemistry. 2004;279(23):24197-202. 20. Kasheverov IE, Zhmak MN, Khruschov AY, Tsetlin VI. Design of new α-conotoxins: from computer modeling to synthesis of potent cholinergic compounds. Marine Drugs. 2011;9(10). 21. Azam L, McIntosh JM. Alpha-conotoxins as pharmacological probes of nicotinic acetylcholine receptors. Acta Pharmacologica Sinica. 2009;30(6):771-83. 22. Conus neuropeptides. Science. 1990;249(4966):257-63. 23. acetylcholine 2000;275(7):4889-96. 24. Hoggard MF, Rodriguez AM, Cano H, Clark E, Tae H-S, Adams DJ, et al. In vivo and in vitro testing of native α-conotoxins from the injected venom of Conus purpurascens. Neuropharmacology. 2017;127:253-9. 25. Mcintosh JM, Santos AD, Olivera BM. Conus peptides targeted to specific nicotinic acetylcholine receptor subtypes. Annual Review of Biochemistry. 1999;68(68):59. 26. Suresh A, Hung A. Molecular simulation study of the unbinding of α-conotoxin [ϒ4E]GID at the α7 and α4β2 neuronal nicotinic acetylcholine receptors. Journal of Molecular Graphics and Modelling. 2016;70:109-21. 27. Armishaw CJ. Synthetic α-conotoxin mutants as probes for studying nicotinic acetylcholine receptors and in the development of novel drug leads. Toxins. 2010;2(6):1471. 28. Abraham N, Healy M, Ragnarsson L, Brust A, Alewood PF, Lewis RJ. Structural mechanisms for α-conotoxin activity at the human α3β4 nicotinic acetylcholine receptor. Scientific Reports. 2017;7:45466. Inserra MC, Kompella SN, Vetter I, Brust A, Daly NL, Cuny H, et al. Isolation and 29. characterization of α-conotoxin LsIA with potent activity at nicotinic acetylcholine receptors. Biochemical Pharmacology. 2013;86(6):791-9. Chi S-W, Kim D-H, Olivera BM, McIntosh JM, Han K-H. Solution conformation of alpha- 30. conotoxin GIC, a novel potent antagonist of alpha3beta2 nicotinic acetylcholine receptors. Biochemical Journal. 2004;380(Pt 2):347-52. 31. Pharmacology. 2006;72(6):661-70. 32. Nicke A, Loughnan ML, Millard EL, Alewood PF, Adams DJ, Daly NL, et al. Isolation, structure, and activity of GID, a novel α4/7-conotoxin with an extended N-terminal sequence. Journal of Biological Chemistry. 2003;278(5):3137-44. 33. neuronal α7 receptors. Journal of Biological Chemistry. 1998;273(18):11007-11. 34. Wakamatsu K, Kohda D, Hatanaka H, Lancelin JM, Ishida Y, Oya M, et al. Structure- activity relationships of mu-conotoxin GIIIA: structure determination of active and inactive

101

from preganglionic nerve

Rajendra W, Armugam A, Jeyaseelan K. Toxins in anti-nociception and anti-

Gao B, Peng C, Yang J, Yi Y, Zhang J, Shi Q. Cone snails: a big store of conotoxins for

Quiram PA, Jones JJ, Sine SM. Pairwise ImI. receptors and α-conotoxin interactions between neuronal α7 Journal of Biological Chemistry.

Lester HA, Dibas MI, Dahan DS, Leite JF, Dougherty DA. Cys-loop receptors: new twists

sodium channel blocker peptides by NMR and simulated annealing calculations. Biochemistry. 1992;31(50):12577-84. 35. Zhang M-M, Green BR, Catlin P, Fiedler B, Azam L, Chadwick A, et al. Structure/function characterization of μ-conotoxin KIIIA, an analgesic, nearly irreversible blocker of mammalian neuronal sodium channels. Journal of Biological Chemistry. 2007;282(42):30699-706. 36. Adams DJ, Smith AB, Schroeder CI, Yasuda T, Lewis RJ. ω-Conotoxin CVID inhibits a pharmacologically distinct voltage-sensitive calcium channel associated with transmitter release Journal of Biological Chemistry. terminals. 2003;278(6):4057-62. Kohno T, Kim JI, Kobayashi K, Kodera Y, Maeda T, Sato K. Three-dimensional structure 37. in solution of the calcium channel blocker omega-conotoxin MVIIA. Biochemistry. 1995;34(32):10256-65. 38. inflammation. Toxicon. 2004;44(1):1-17. 39. Livett BG, Sandall DW, Keays D, Down J, Gayler KR, Satkunanathan N, et al. Therapeutic applications of conotoxins that target the neuronal nicotinic acetylcholine receptor. Toxicon. 2006;48(7):810-29. 40. Satkunanathan N, Livett B, Gayler K, Sandall D, Down J, Khalil Z. Alpha-conotoxin Vc1.1 alleviates neuropathic pain and accelerates functional recovery of injured neurones. Brain Research. 2005;1059(2):149-58. 41. novel drug discovery. Toxins. 2017;9(12):397. 42. Wu C-H, Lee C-H, Ho Y-S. Nicotinic acetylcholine receptor-based blockade: applications of molecular targets for cancer therapy. Clinical Cancer Research. 2011;17(11):3533-41. 43. acetylcholine 1999;274(28):19517-24. 44. Eddins D, Sproul AD, Lyford LK, McLaughlin JT, Rosenberg RL. Glutamate 172, essential for modulation of L247T α7 ACh receptors by Ca2+, lines the extracellular vestibule. American Journal of Physiology-Cell Physiology. 2002;283(5):C1454-C60. Kasheverov IE, Chugunov AO, Kudryavtsev DS, Ivanov IA, Zhmak MN, Shelukhina IV, et 45. al. High-affinity α-conotoxin PnIA analogs designed on the basis of the protein surface topography method. Scientific Reports. 2016;6:36848. Barrantes FJ. The nicotinic acetylcholine receptor current views and future trends. 46. Barrantes FJe, editor. Berlin, Heidelberg: Springer Berlin Heidelberg : Imprint: Springer; 1998. 47. Lebbe EKM, Peigneur S, Wijesekara I, Tytgat J. Conotoxins targeting nicotinic acetylcholine receptors: an overview. Marine Drugs. 2014;12(5):2970-3004. 48. Boffi JC, Marcovich I, Gill-Thind JK, Corradi J, Collins T, Lipovsek MM, et al. Differential contribution of subunit interfaces to α9α10 nicotinic acetylcholine receptor function. Molecular Pharmacology. 2017;91(3):250-62. 49. and turns. Trends in Neurosciences. 2004;27(6):329-36. 50. Tsigelny I, Sugiyama N, Sine SM, Taylor P. A model of the nicotinic receptor extracellular domain based on sequence identity and residue location. Biophysical journal. 1997;73(1):52-66.

102

Šali A, Blundell TL. Comparative protein modelling by satisfaction of spatial restraints.

Shen M-Y, Sali A. Statistical potential for assessment and prediction of protein

Guex N, Peitsch MC. SWISS-MODEL and the Swiss-PdbViewer: an environment for

Colovos C, Yeates TO. Verification of protein structures: patterns of nonbonded atomic

Quigley D, Probert MIJ. Langevin dynamics in constant pressure extended systems.

Bussi G, Donadio D, Parrinello M. Canonical sampling through velocity rescaling. The

Hess B, Bekker H, Berendsen HJC, Fraaije JGEM. LINCS: a linear constraint solver for

Essmann U, Perera L, Berkowitz ML, Darden T, Lee H, Pedersen LG. A smooth particle

Humphrey W, Dalke A, Schulten K. VMD: visual molecular dynamics. Journal of

Hilf RJC, Dutzler R. Structure of a potentially open state of a proton-activated

Lee WY, Sine SM. Principal pathway coupling agonist binding to channel gating in

Unwin N, Fujiyoshi Y. Gating movement of acetylcholine receptor caught by plunge-

Hibbs RE, Gouaux E. Principles of activation and permeation in an anion-selective Cys-

Dutertre S, Ulens C, Büttner R, Fish A, van Elk R, Kendel Y, et al. AChBP-targeted α- 51. conotoxin correlates distinct binding orientations with nAChR subtype selectivity. The EMBO Journal. 2007;26(16):3858-67. 52. Journal of Molecular Biology. 1993;234(3):779-815. 53. structures. Protein Science. 2006;15(11):2507-24. 54. comparative protein modeling. Electrophoresis. 1997;18(15):2714-23. 55. interactions. Protein Science. 1993;2(9):1511-9. 56. Hess B, Kutzner C, van der Spoel D, Lindahl E. GROMACS 4: algorithms for highly efficient, load-balanced, and scalable molecular simulation. Journal of Chemical Theory and Computation. 2008;4(3):435-47. 57. Mark P, Nilsson L. Structure and dynamics of the TIP3P, SPC, and SPC/E water models at 298 K. The Journal of Physical Chemistry A. 2001;105(43):9954-60. 58. The Journal of Chemical Physics. 2004;120(24):11432-41. 59. Journal of Chemical Physics. 2007;126(1):014101. 60. molecular simulations. Journal of Computational Chemistry. 1997;18(12):1463-72. 61. mesh Ewald method. The Journal of Chemical Physics. 1995;103(19):8577-93. 62. Molecular Graphics. 1996;14(1):33-8. 63. Shannon P, Markiel A, Ozier O, Baliga NS, Wang JT, Ramage D, et al. Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Research. 2003;13(11):2498-504. 64. Sahu BS, Obbineni JM, Sahu G, Singh PK, Sonawane PJ, Sasi BK, et al. Molecular interactions of the physiological anti-hypertensive peptide catestatin with the neuronal nicotinic acetylcholine receptor. Journal of Cell Science. 2012;125(9):2323-37. 65. Peng W, Ding F. Biomolecular recognition of antagonists by α7 nicotinic acetylcholine receptor: antagonistic mechanism and structure–activity relationships studies. European Journal of Pharmaceutical Sciences. 2015;76:119-32. 66. pentameric ligand-gated ion channel. Nature. 2009;457(7225):115-8. 67. nicotinic receptors. Nature. 2005;438(7065):243-7. 68. freezing. Journal of Molecular Biology. 2012;422(5):617-34. 69. loop receptor. Nature. 2011;474:54. 70. Lummis SCR, Beene DL, Lee LW, Lester HA, Broadhurst RW, Dougherty DA. Cis-trans isomerization at a proline opens the pore of a neurotransmitter-gated ion channel. Nature. 2005;438(7065):248-52.

103

Baenziger JE, Hénault CM, Therien JPD, Sun J. Nicotinic acetylcholine receptor–lipid 71. interactions: mechanistic insight and biological function. Biochimica et Biophysica Acta (BBA) - Biomembranes. 2015;1848(9):1806-17. Noridomi K, Watanabe G, Hansen MN, Han GW, Chen L. Structural insights into the 72. molecular mechanisms of myasthenia gravis and their therapeutic implications. eLife. 2017;6:e23043. 73. Ellison M, Gao F, Wang H-L, Sine SM, McIntosh JM, Olivera BM. α-Conotoxins ImI and ImII target distinct regions of the human α7 nicotinic acetylcholine receptor and distinguish human nicotinic receptor subtypes. Biochemistry. 2004;43(51):16019-26. 74. Gao F, Bren N, Burghardt TP, Hansen S, Henchman RH, Taylor P, et al. Agonist- mediated conformational changes in acetylcholine-binding protein revealed by simulation and intrinsic tryptophan fluorescence. Journal of Biological Chemistry. 2005;280(9):8443-51. 75. Kang TS, Vivekanandan S, Jois SDS, Kini RM. Effect of C-Terminal Amidation on Folding and Disulphide-Pairing of α-Conotoxin ImI. Angewandte Chemie International Edition. 2005;44(39):6333-7.

104

Chapter 4 : Effects of C-terminal Carboxylation of LsIA

on Interactions with α3β2 Nicotinic Acetylcholine

Receptor Interfaces

105

4.1 Introduction

Neuronal nicotinic acetylcholine receptors (nAChRs) are ligand-gated ion channels [1] that

play important roles in mediating signalling transmission between neurons. Therefore, they

are related to numerous neurological disorders and diseases, such as Parkinson’s disease,

Alzheimer’s disease, dementia, schizophrenia and addiction [2-4]. These receptors exist as

five subunit combinations of α subunit (α2-10) only (homomeric α7 and α9 subtypes), or

heteromeric combinations of α and β (β2-4) subunits [5]. Each subunit comprises an N-

terminal extracellular domain (ECD), transmembrane domain (TMD) and an intracellular

domain (ICD). The ECD is of particular importance due to the presence of the canonical ligand-

binding domain (LBD) at the interface of two adjacent subunits [6, 7]. It has been

demonstrated that the ECD of α subunit can facilitate the binding of agonists [8], such as

acetylcholine (ACh), at the LBD [1], resulting in the opening of the channel in the TMD.

Moreover, neuronal nAChRs are widely recognised as a target of α-conotoxins, such as LsIA,

showing high potency at α3β2 (IC50 10.3 nM) and α7 (IC50 10 nM) nAChR subtypes [9].

The α3β2 subtype is expressed in the central and peripheral nervous systems, including the

cerebellum, spinal cord, and autonomic ganglion neurons [10, 11]. Various stoichiometries of

the α3 and β2 subunits are known to exist, with the most common consisting of two α3 and

three β2 subunits ((α3)2(β2)3) [12]. A recent study has suggested that the (α3)3(β2)2

stoichiometry may also exist at the motoneuron-Renshaw cell synapse in the spinal cord [13].

In this stoichiometry, a pair of β2(+)α3(−) and α3(+)β2(−) interfaces are commensurate with

those constituting the (α3)2(β2)3 isoform. However, an α3(+)α3(−) interface in the former,

(α3)3(β2)2, and a β2(+)β2(−) interface in the latter, (α3)2(β2)3, cause the main structural

differences between the α3β2 isoforms. Moreover, the physiological and pharmacological

properties of various nAChR isoforms may be affected by the ratio of α to β subunits [14, 15].

In addition, previous studies have also demonstrated the existence of alternative

stoichiometries of nAChRs, such as α4β2 ((α4)3(β2)2 and (α4)2(β2)3), α3β4 ((α3)3(β4)2 and

(α3)2(β4)3) and α9α10 nAChR ((α9)3(α10)2 and (α9)2(α10)3) [14-19].

More importantly, a previous study by Boffi et al. has demonstrated that the different

combinations of subunit interfaces of α9α10 nAChR showed distinct contributions of α9 and

α10, depending on which serves as the (−) face to the binding of ACh [18]. Furthermore, the

106

presence of α-conotoxins targeting multiple interfaces of nAChRs has attracted attention for

investigating the binding determinants of the successful binding of the ligand at the receptors

[15, 19]. However, the role of the different interfaces in the binding of LsIA to nAChRs remains

unclear.

The combinations of an α subunit as the principal (+) face with one β subunit as

complementary (−) face form interfaces, which are commonly accepted as the canonical

region binding agonists and antagonists. However, other different interfaces of α3β2 nAChR,

namely β2(+)β2(−), β2(+)α3(−) and α3(+)α3(−), may also contribute to the binding of ligands,

such as ACh and α-conotoxin MII [20]. Further, the residues on β subunits have been

suggested to be involved in the nAChR subtype selectivity of α-conotoxins [21-23]. Other

studies emphasized the important role of the β subunit as a structural component, which

serves as an accessory binding face for α subunits [18, 21, 24, 25]. Therefore, it is intriguing

to explore the role of the β2 subunit in interacting with novel α-conotoxins at the different

binding interfaces of the (α3)2(β2)3 nAChR isoform.

LsIA, isolated from the venom of Conus limpusi, is known to inhibit rat α3β2 nAChR with

nanomolar potency [9]. Similar to a range of α-conotoxins, native LsIA possesses an amidated

C-terminus (C-T) and two disulphide bridges contributed by four conserved cysteine residues

(I-III, II-IV). These disulphide bridges also confer it with two structural loops consistent with

the 4/7 subclass of α-conotoxins, characterized by four amino acid residues in the first loop

(loop 1) and seven residues in the second loop (loop 2) [21, 26].

These structural characteristics may contribute to the high binding affinity of α-conotoxins to

nAChR subtypes. For example, LsIA has been shown to selectively inhibit human α7 and rat

α3β2 nAChRs expressed in Xenopus laevis oocytes [9, 27], whereas the LsIA[R10F] mutant

preferentially binds at human α3β4 nAChR [27]. Furthermore, the carboxy-terminal (C-

terminal) amidation, together with the sulfated tyrosine, are crucial for the activity of AnIB

on α7, rather than α3β2 [28]. Similar to native LsIA, most 4/7 α-conotoxins possess an

amidated C-T in their native form, namely α-conotoxins Vc1.1, MII, GIC, AnIB, and TxIA.

However, an amidated C-T does not exist in certain identified α-conotoxins, such as GID and

RgIA with a C-terminal carboxylate, that also bind at α3β2, α7 and α9α10 nAChRs with high

potency [26, 29]. Notably, previous experimental results have shown that the C-terminal

107

carboxylation enhanced the binding affinity of LsIA at rat α3β2 nAChR by three-fold, yet this

modification (C-terminal carboxylation of LsIA) resulted in a three-fold reduced affinity at

human α7 nAChR. In the present context, “C-terminal carboxylation” or “carboxylated C-

terminus” refers to the presence of a single COO- functional group that terminates the peptide

backbone, compatible with the terminology adopted by Inserra et al. [9] Although these

differences are marginal, the investigation of the effects of C-T of α-conotoxins on receptor

interface interactions may nonetheless shed light on the possibilities for improvements and

design of α-conotoxins with high selectivity to individual nAChR subtypes [30].

The determined crystal structure of the acetylcholine binding protein (AChBP) has facilitated

in silico studies of conotoxin-nAChR complexes. Although the sequence similarity between rat

α3, β2 and α7 and Aplysia californica acetylcholine binding protein (Ac-AChBP) at the ECD is

approximately 27%, 24% and 27%, respectively, the AChBPs share high structural similarity to

nAChRs at this region [31-34]. The highly resembled structure between the ECD of AChBPs

and nAChRs may suggest the closely related pharmacological natures of the AChBPs [34],

compared with those of nAChR. As a result, the characterized structure of AChBP can provide

reliable homology models of nAChRs for molecular dynamics (MD) simulations and

subsequent computational analyses [33, 35]. Previous studies have employed the molecular

modelling of nAChRs based on the crystal structures of AChBP [31, 36] and α1 nAChR subunit

[37]. Moreover, numerous studies have produced comparative models of conotoxins binding

at nAChRs complex, built using the co-crystal structures of Ac-AChBP with α-conotoxins

TxIA[A10L], PnIA[A10L, D14K], ImI and LvIA [32, 34, 38-42], respectively, as well as Lymnaea

stagnalis (Ls-) AChBP bound by α-cobratoxin [43] and LsIA [27], respectively.

In the present study, MD simulations were employed to predict the determinants for the

differences in receptor interactions with amidated (LsIA) and carboxylated (LsIA#) analogue

of the toxin at α3(+)β2(−), β2(+)α3(−) and β2(+)β2(−) interfaces (see Figure 4.1), which may

help to explain the differential inhibition of α3β2 by these two toxin isoforms. These results

predict differences in the structures, motions and intra-molecular contacts between LsIA#

and LsIA, as to elucidate the toxin-receptor interactions which are increased for LsIA#

compared to LsIA at certain types of interfaces, which may thereby be proposed to be key

determinants for the increased inhibitory activity of LsIA# at α3β2 versus LsIA. The

electrostatic property that affects the binding of LsIA# at α3β2 versus its native type was also

108

explored via Adaptive Poisson–Boltzmann Solver (APBS). Furthermore, in addition to toxin-

receptor contacts and their physical characteristics, we examined the structural influences

exerted on the receptor by the two LsIA isoforms and employed a molecular network

approach to compare the effects that each LsIA isoform may exert on contacts between

adjacent nAChR subunits at α3(+)β2(−) and β2(+)α3(−) interfaces. The consideration of inter-

subunit contacts to predict the potential bioactivity of α-conotoxins bound at specific nAChR

interfaces is worthy of further investigation.

Figure 4.1. Binding mode of -conotoxin LsIA at rat α3β2 nAChR. (A) Top view of LsIA anchored to the receptor. The five subunits of the α3β2 and the LsIAs anchoring to five interfaces of the nicotinic receptor are labelled with their nomenclature names and the chain numbers used to identify subunits in this work. The α3 and β2 subunits are represented by blue and yellow colors, respectively, in NewCartoon drawing form; and LsIAs bound at α3(+)β2(−), β2(+)α3(−) and β2(+)β2(−) sites are shown in green, pink and red colors, respectively. (B) Side view of LsIA binding at the α3(+)β2(−) interface. The homology model of the LsIA/α3β2 complex was constructed via homology modelling (Modeller9v6) method [44], and was further refined via molecular dynamics (MD) simulations (GROMACS v4.6.5 [45]). A detailed description of the methods is provided in the Methods section. (C) The 3-D structure of LsIA with 4 conserved cysteine residues shown in CPK form. The yellow bonds

109

represent two disulphide bonds of the toxin. The amino acid sequence of LsIA is shown on the right- hand side of the figure. The asterisk (*) denotes the C-terminal amidation.

4.2 Methods

In the present study, the complex of -conotoxin LsIA- and LsIA#-α3β2 nAChR was first

established via homology modelling with an appropriate template. The MD simulation was

subsequently applied for refining the comparative models and providing the trajectory,

including the geometries, velocities and energies of the atoms [46] for further analysis.

4.2.1 Homology Modelling

The comparative model of LsIA- and LsIA#-α3β2 nAChR complexes were built according to the

process described in our previous study [47]. Briefly, the sequence of LsIA bound to the ECD

of rat α3 (UniProtKB: P04757) and β2 subunits (UniProtKB: P12390) was aligned to that of α-

conotoxin PnIA[A10K, D14K] bound to Ac-AChBP (PDB ID: 2BR8) complex [38], with the known

crystal structure. The structure with the best discrete optimized protein energy (DOPE) score

was selected from 100 models built via Modeller9v6 [44]. The DOPE score is calculated based

on the distance-dependent energy function applied in the homology modelling approach,

Modeller, and was used to evaluate the homology models at the atomic level [48]. The lower

the DOPE, the better the model. The top view of the homology model of the LsIA-bound α3β2

nAChR interfaces is shown in Figure 4.1A, whilst Figure 4.1B demonstrates the side view of

the motif of LsIA anchored to α3(+)β2(−) interface. The strategy, in which the model was

constructed for LsIA binding at five interfaces of α3β2 nAChR, was also widely employed in

other studies on α-conotoxin bound heteropentameric nAChR complexes [20, 27, 49]. It

should be noted that the commonly assumed order of (α3)2(β2)3 for the homology modelling

of the nicotinic receptor is α3−β2−α3−β2−β2 [19, 24]. Although there is a lack of studies of

concatemeric constructs of α3β2 nAChR, classic concatemer studies of α4β2 nAChR by

Carbone et al. [50] indicated that functional receptor assemblies do not involve triplets of the

same subunit. Thus, the αβαββ assembly motif is assumed to constitute the functional form

of (α3)2(β2)3 as well.

110

As in our previous study on LsIA-α7 complexes [23], a recent co-crystal structure is available

of Ls-AChBP, bound to amidated α-conotoxin LsIA (PDB code: 5T90 [29]), which shows minor

structural differences in AChBP compared to all other conotoxin-Ac-AChBP crystal structures

to date, possibly due to species differences between Ls- and Ac-AChBP. The authors

constructed homology models of LsIA bound to α7 nAChR or apo α7 nAChR based on the Ls-

AChBP crystal structure, and comprehensively elaborated and tested key interactions

identified in the model in many elegant experiments [27, 51]. In both cases, minor structural

differences between Ls- and Ac-AChBP, and the availability of co-crystal structures of 4/7-

conotoxins bound to both species isoforms, allow the exploration of an alternative, possible

conformation of LsIA-α3β2 complex based on Ac-AChBP, which shares a comparable (though

somewhat lower) sequence similarity to rat α3 and β2 subunits.

In addition, the recently revealed X-ray structure of human α4β2 nAChR co-crystallized with

nicotine (PDB ID:5KXI) [52] putatively supports the homology modelling of (α4)3(β2)2 nAChR

stoichiometry [53] and molecular docking studies on α4β2 nAChR [54-56] for the small

molecules, such as nicotine. Additionally, this determined structure companied with the

complex of Ac-AChBP co-crystallized with the PnIA variant, benefits the comparative

modelling of (α3)3(β4)2 and (α3)2(β4)3 nAChRs bound by α-conotoxin AuIB [15]. However, the

nicotine bound α4β2 nAChR (PDB ID: 5KXI) [52] was only used for the construction of

(α3)3(β4)2/ribbon isomer of AuIB (linking cysteine residues I-IV, II-III) complex [15], whereas

the globular AuIB (linking cysteine residues I-III, II-IV) anchoring to (α3)2(β4)3 was based on

the same template (PDB ID: 2BR8) [38] as in the present study.

4.2.2 Molecular Dynamics Simulation

• MD simulations on LsIA-α3β2 complexes MD simulations were carried out on the complexes using GROMACS v4.6.5 [45] and the

CHARMM27 force field. The refined protein complexes were analyzed by following the

process described in Chapter 3 [47]. To improve conformational sampling, each LsIA-α3β2

complex was run using 28 independent simulations, each initiated using a different random

seed, up to at least 27ns. Unless otherwise stated, all subsequent results are reported as

averages over the independent trajectories. Both systems were solvated with TIP3P water

111

models [57]. For the LsIA-α3β2 complex, 70 mM Cl− and 110 mM Na+ were added, to achieve

an approximately 0.15 M salt concentration, whereas for the LsIA#-α3β2 complex, 70 Cl− and

115 Na+ were added. Simulation trajectory frames were recorded every 10 ps for data

analysis. The inter-subunit contacts were visualized through Cytoscape [58], which was also

performed in our previous study on α7 nAChR [23]. All structure graphics were generated via

visual molecular dynamics (VMD) [59]. Other simulation conditions and parameters are the

same as those applied in our previous study in Chapter 3 [60]. The cluster analyses were

performed over the whole trajectories for both LsIA and LsIA# bound forms via GROMACS

packages. The trajectory frame interval was extracted every 10 frames, therefore around

7634 frames were acquired for clustering. The cutoff for LsIA- and LsIA#-α3β2 complexes was

set up as 3.8 Å. As a result, the number of identified clusters for LsIA and LsIA# bound forms

was 8 and 10, respectively. The top 1 ranked conformations (top 1 cluster) represent 92.4%

(LsIA) and 91.4% (LsIA#) for the overall structural population for each protein complex, and

were selected for APBS analysis [61] using the PyMOL APBS plugin [62] (APBS Tools 2.1).

The overall quality factor of homology model of α3β2 nAChR bound by LsIA is 83.1 using

ERRAT [63], which is slightly lower than that of the original structure (2BR8) with the quality

factor of 89.1. The minorly reduced quality of structure mainly appeared at the Cys- and F-

loop at α3 (residue V132 to C142) and β2 (residue H160 to F163) nAChR subunits. However,

such violation of structure was improved after the MD simulations of α3β2 bound by

LsIA/LsIA# complexes, given that the quality factor was increased to 88.2 and 86.7 for

α3β2/LsIA and α3β2/LsIA paradigms, respectively. In addition, the region with similar

structural violation between the α3β2 nAChR model and the template (2BR8) situated at the

loop between α1 helix and β1 as well as certain residues on β9 and β10 strands, which are

minorly related to the LBD of α3β2 nAChR interfaces.

• MD simulations on LsIA only To compare the intramolecular interactions within the LsIA/LsIA# in solution with their

nAChR-bound forms, simulations were deployed only on LsIA and LsIA#, respectively, for 300

ns each. Consistent with the simulations on the receptor-ligand complexes, the result showed

a proximate distance between R10 and C17 of LsIA# (3.44 Å ± 1.4) versus that of the LsIA (3.38

Å ± 0.96). For achieving 0.15 M concentration, the LsIA and LsIA# were solvated with 3 mM

Na+ and 5 mM Cl−, and 3 mM Na+ and 4 mM Cl−, respectively, as well as water molecule models

112

(TIP3P). The other parameters for simulation in this step are consistent with those for LsIA-

α3β2 complex simulations.

4.3 Results

4.3.1. Stability and Flexibility of the LsIA-α3β2 Complexes

The root-mean-square deviation (RMSD) time series plots for the LsIA- and LsIA#-α3β2

complexes are shown in Figure 4.2A–D, each time series of which was taken as an average

over 28 independent simulations. The fitting was performed on the Cα atoms of the entire

receptor-ligand complex and RMSD values were calculated for the respective individual

chains, discussed below. These plots demonstrate the stability of the backbone atoms of the

protein complexes [64], all of which approach a plateau in their respective RMSD plots by

approximately 15 ns. The patterns of the RMSD plots in the present study are qualitatively

comparable to those of the RMSD results based on simulations of other nAChRs-toxins models

[65-67]. Despite the similarity of the RMSD plots of the α3β2 nAChR subunits between LsIA

and LsIA# bound forms, there are, however, some differences in the RMSD behavior observed

for both of the ligands. LsIA and LsIA# show an asymmetric pattern in their RMSD plots,

depending on the targeted α3β2 interfaces bound by the toxins, as shown by Figure 4.2C, D.

The figures demonstrate two distinct types of RMSD values. Chain G (bound at the β2(+)α3(−)

interface), chain I (β2(+)β2(−)) and chain J (β2(+)α3(−)) exhibited high values of RMSD

(approaching 0.5 nm), whereas chain F (bound at α3(+)β2(−)) and H (α3(+)β2(−)) exhibited

much lower RMSD values (approaching 0.3 nm). The RMSD results indicate that LsIA, when

bound at interfaces involving β2(+) as the principal face, went through substantially greater

structural drift from the initial homology model compared to those bound at the α3(+)

interfaces. This asymmetry may be because the β2(+)α3(−), and β2(+)β2(−) interfaces are less

preferable for the binding of LsIAs, compared with the α3(+)β2(−) interfaces.

To obtain further insights into the regions of LsIA responsible for the interface-specific

differences in RMSD as described above, absolute RMSF plots were obtained for the LsIA-

(Figure S1.1A,C in Appendix 1) and LsIA#-α3β2 complexes (Figure S1.1B,D). To confirm that

the structural and dynamical properties of the LsIA-nAChR complexes were reasonably

113

converged within the simulation time frame, the sum of squared differences (SSD) were

plotted at 3, 6, 9, 12, 15, 18, 21, 24 and 27 ns RMSFs for the average of the trajectories (Figure

S1.2), as well as individual trajectories (not shown). These plots indicate that 21 ns is sufficient

for convergence, at least within the timescale of the present simulations (Figure S1.2).

In the RMSF results for LsIA and LsIA# (Figure S1.1C,D), a high fluctuation in the N-terminus

(N-T) of LsIA was observed. For both LsIA and LsIA#, the values of the RMSF followed the same

pattern as the RMSD, with chains G, I and J (binding at β2(+) interfaces) exhibiting distinctly

higher RMSF compared to chains F and H (binding at α3(+) interfaces), particularly in the N-T

and N15 regions. Thus, the N-T of both LsIA and LsIA# appears to undergo far greater flexibility

at β2(+)α3(−) compared to those bound at α3(+)β2(−) interfaces. The differences, mainly

regarding the N-T of LsIAs, may suggest important variations in toxin-receptor interactions,

depending on the interface arrangement.

114

Figure 4.2. Root-mean-square deviation (RMSD) (Mean ± SD) plots for MD simulation of LsIA and its carboxylated analogue (LsIA#) binding at α3β2 pentameric nicotinic acetylcholine receptor (nAChR) complexes based upon the template 2BR8. (A) RMSD plots of LsIA bound at α3β2 nAChR. (B) RMSD plots of LsIA# bound to α3β2 nAChR. ‘Amid’ represents the amidated LsIA bound to α3β2 nAChR, whereas ‘Carb’ represents the carboxylated LsIA (LsIA#) bound α3β2 nAChR subunits. RMSD plots of (C) LsIA and (D) LsIA# upon anchored to the receptor.

The differences in RMSF results between LsIA- and LsIA#-α3β2 complexes were compared, by

subtracting the RMSF values of LsIA# bound α3β2 nAChR from those of the LsIA bound

complex (Figure 4.3A). Positive values represent enhanced flexibility of the backbone of the

LsIA bound complex, whereas lower flexibility (relative to LsIA#) is demonstrated by negative

values. Only residues that show differences in RMSF values greater than 0.5 Å were included,

as this value has been used as a cut-off point in previous MD simulation studies comparing

similar proteins in different states [68, 69].

For α3β2 nAChR subunits (Figure 4.3A), there is an increased RMSF value in some regions,

including Cys-loop and C-loop of the LsIA-bound α3 subunit versus the LsIA# complex,

115

illustrating the enhanced flexibility in this region. Greater C-loop flexibility may infer a lower

capacity for LsIA to lock the ligand-binding site in an open-loop, inhibited conformation, or

may be indicative of weaker receptor-toxin interactions for LsIA leading to higher flexibility.

In contrast, for the β2 subunit, the opposite result was observed, showing that the rigidity of

these regions is higher for the LsIA bound subunit compared to the LsIA#. In contrast to the

observation regarding the C-loop of the α3 subunit, the enhanced rigidity of the β2 subunit

bound by LsIA, may suggest that LsIA has less effect on suppressing Cys-loop and C-loop

movements at α3(+) rather than β2(+) interfaces. However, the increased flexibility of

residues in the Cys-loop and C-loop of the α3 subunit was not significant at the 5% level,

except for the two residues, α3-Y215 (P-value 0.031) and α3-C217 (P-value 0.016), which

showed statistically significant differences from that of such residues on α3 subunit of α3β2

nAChR bound by LsIA#. Further, the α3-Y215 played an important role in ACh binding at

α3(+)β2(−) interfaces, by forming aromatic interactions with the neurotransmitter [20].

For LsIA (Figure 4.3B), the differences in RMSF values are shown for the different interfaces

of the α3β2 nAChR bound by the toxins. At α3(+)β2(−) interfaces, LsIA# was less flexible than

LsIA at most positions, especially for the residues, N13, P14, N15, I16 of loop 2 and the C-T

(beige bars, Figure 4.3B). This suggests that at α3(+)β2(−), LsIA# may form a higher number

of intermolecular contacts with the receptor, or may form internal intra-molecular contacts,

which can result in its significantly greater rigidity. At the β2(+)α3(−) interfaces, LsIA# was,

however, less flexible than LsIA in only loop 1, whereas LsIA# exhibited similar flexibility to

LsIA in loop 2 (yellow bars, Figure 4.3B), in contrast to the observed RMSF behavior for

α3(+)β2(−) above. Finally, at the β2(+)β2(−) interface, a different RMSF behavior was

observed; aside from LsIA-G1 and S2, LsIA# exhibited higher flexibility for both loops 1 and 2

compared to LsIA (green bars, Figure 4.3B). Thus, at both mixed subunit interfaces, the

structure of LsIA# was generally more rigid than LsIA, but at β2(+)β2(−), LsIA# was only more

rigid at residues in position 1 and 2 of the toxin, but more flexible elsewhere in both loops 1

and 2. Therefore, it can be proposed that the enhanced rigidity of LsIA# compared to LsIA at

α3(+)β2(−) interfaces is qualitatively consistent with the expectation that LsIA# should have

greater toxin-receptor contacts than LsIA (and therefore less mobile), given the known higher

potency of LsIA# at α3β2. There is also an indication that at the β2(+)α3(−) interfaces, LsIA#

might be marginally more rigid than LsIA, although caution must be exercised in this

116

interpretation, due to lack of statistical significance at the 5% level for the results at this type

of interfaces. It is plausible that considering both these mixed interfaces may help elucidate

the toxin’s differential potency. However, the same cannot be said for β2(+)β2(−), for which

there was generally higher flexibility for LsIA#, suggesting fewer toxin-receptor contacts.

Thus, the consideration of β2(+)β2(−) is less likely to have a predictive value, and in

subsequent sections, this chapter focused on the discussion mainly on α3(+)β2(−) and

β2(+)α3(−) interfaces, with the majority of the β2(+)β2(−) results provided in Appendix 1. To

elucidate the determinants affecting the possible differences at α3β2 interfaces’ binding

potency by LsIA and its analogue, the basis for the distinctions in RMSF behavior between LsIA

and LsIA# at the α3(+)β2(−) and β2(+)α3(−) interfaces, was further investigated by examining

specific intra- and inter-molecular contacts between LsIA and the α3β2 receptor discussed

below.

Figure 4.3. Differences between RMSF plots of LsIA and LsIA# binding at α3β2 nAChR. (A) Averaged diff in RMSF plot of α3β2 subunits bound by LsIA. The ’Amid’ label on the positive side of the y-axis

117

represents the LsIA bound subunits to show higher flexibility versus the LsIA# bound form, whereas the ’Carb’ label on the negative side of the axis shows that the LsIA# bound form gains higher flexibility over the LsIA bound subunit. (B) Diff in RMSF for ligands anchoring to the nicotinic receptor interfaces. The differences of RMSF between LsIA and LsIA# (The RMSF value of LsIA# was subtracted from that of LsIA) are demonstrated, with different colors of the bars for the LsIA binding at different interfaces. The diff in RMSF of LsIA binding at α3(+)β2(−), β2(+)α3(−), and β2(+)β2(−) interfaces are shown with beige, yellow and green, colors respectively, whilst the averaged diff in RMSF of LsIA is shown in pink color. The statistical significance of changes in the RMSF of α3 and β2 subunits bound by LsIA/LsIA# and the individual residues of LsIA/LsIA# (binding at different interfaces), was calculated over 28 individual seeds (* p < 0.05; ** p < 0.001).

4.3.2. LsIA and LsIA# Toxin-receptor Contacts

The total number of contacts for each residue of LsIA formed with α3β2 is depicted in Figure

4.4A, whereas Figures 4.4B,C and S1.3 show the important pairwise interactions influencing

receptor interactions by LsIAs anchored to the α3(+)β2(−) (Figure 4.4B), β2(+)α3(−) (Figure

4.4C) and β2(+)β2(−) (Figure S1.3) interfaces.

Overall, the LsIA# with α3β2 nAChR complex had higher inter-residue contacts between the

ligand and the receptor versus the LsIA bound complex, as can be seen in Figure 4.4A. The

total number of receptor interactions made by LsIA was mainly contributed by P7 on loop 1,

and with R10 and V11 on loop 2 for both LsIA and LsIA#. The higher toxin-receptor contacts

may partially explain the higher rigidity of LsIA# at both α3(+)β2(−) and β2(+)α3(−) interfaces,

compared with the LsIA bound complex.

In terms of the specific determinants affecting the total receptor interactions by LsIAs,

pairwise interactions between the receptor and the LsIAs were further investigated via

gmx_mindist with a cutoff of 4.5 Å. To make the differences in per-residue contacts between

LsIA and LsIA# at the α3(+)β2(−) interfaces clear, the number of individual pairwise contacts

of LsIA# was subtracted by that of LsIA. Thus, a negative value represents relatively higher

LsIA# contacts, whereas the positive value shows higher contacts for LsIA with the

corresponding residues on α3β2 nAChR, which is indicated within the bars for each LsIA

residue. The absolute value of 4 for the differences in the number of contacts for coupled

residues was selected as a threshold for including the potentially important pairwise contacts

affecting the receptor interactions. Additionally, for ease of interpretation, pairwise

interactions were excluded if the original number of contacts, between LsIA and LsIA# bound

complexes, was less than 15, indicating a low level of contacts formed by the coupled

118

residues. The statistical significance was evaluated for the selected pairwise interactions over

28 individual simulations. It is shown that P7, P14 and C17 formed the most significant

variations in several contacts between LsIA and LsIA# binding at both α3(+)β2(−) and

β2(+)α3(−) interfaces (Figure 4.4B,C), and fewer differences occurred at the β2(+)β2(−)

interface (Figure S1.3). The interactions involving each of the above LsIA residues at the

α3(+)β2(−) and β2(+)α3(−) interfaces are discussed in turn in the following sections.

Figure 4.4. Summary of the number of contacts in LsIA and LsIA# bound forms. (A) A total number of receptor interactions by LsIA and LsIA# (Mean ± SD). The mean values were calculated over five α3β2 interfaces, and the error bars represent the standard deviation (SD). The significant changes of pairwise contacts for LsIA anchored to α3(+)β2(−) (B) and β2(+)α3(−) (C) interfaces of α3β2 nAChR by C-terminal carboxylation. The residues on the principal (+) face of binding interfaces of α3β2 are shown with a black line border. The statistical significance of the differences between the number of pairwise interactions was calculated over 28 individual seeds (* p < 0.05; ** p < 0.001).

119

4.3.3. Important Pairwise Interactions of LsIA with Residues at α3(+)β2(−) Interfaces

The increased contacts between LsIA#-C17 and β2(−)-K187 likely exerted the greatest impact

on improved binding affinity of LsIA# versus LsIA at α3(+)β2(−) interfaces, together with P7

and P14 forming hydrophobic interactions with corresponding aromatic and hydrophobic

residues, F143, L145 and W81, on the β2 subunit. The dominant enhanced contacts occurred

between LsIA#-C17 and β2(−)-K187 via salt-bridge/hydrogen bonds interactions, which were

absent in the LsIA/α3β2 complex (Figure 4.5,S1.4). Additionally, the proximity between P7

and β2(−)-W81 and β2(−)-L145 and the enhanced pairwise contacts between P14 and β2(−)-

F143 (Figure 4.5,4.6) contributed to the augmented hydrophobic interactions in LsIA# versus

the LsIA bound form.

Furthermore, the salt-bridge interaction between C17 and R10 of LsIA# (Figure S1.5) directly

facilitated the LsIA#-R10 contact with β2(−)-L145, via forming van der Waals interactions

(Figure 4.6), by placing the toxin in a favorable conformation for such interaction. Although

the interactions formed by LsIA#-R10 and β2(−)-W81 were reduced against the LsIA bound

form, the distance between the coupled residues was relatively large in the latter complex.

Thus, the carboxylation of LsIA probably causes a switch in the favorable interaction of R10

with β2(−)-W81 (for LsIA) to β2(−)-L145 (for LsIA#). Nevertheless, as the van der Waals

interaction between R10 and β2(−)-L145 was relatively weak, this “switch” may only result in

a slightly overall increased number of contacts between the key residue R10 and the receptor

at the α3(+)β2(−) interfaces by LsIA#. Consequently, the ability to form a salt bridge/hydrogen

bond between the C-T of LsIA# and β2(−)-K187 is likely to be a key factor explaining the higher

affinity of LsIA# compared to that of the amidated form LsIA at α3(+)β2(−) interfaces. This

interaction may also play a critical role in helping produce an overall higher number of

contacts between α3β2 and LsIA# compared to LsIA bound form, by drawing the toxin closer

to the (−) face of the receptor, indirectly stimulating the significantly augmented contacts at

other residues, such as R10 and P14, as discussed above.

120

Figure 4.5. Important interactions formed by LsIA-P14 and C17 with the receptor residues on the complementary (−) face at the α3(+)β2(−) interface. Binding mode of LsIA (transparent green) (A) and LsIA# (transparent orange) (B) at the (−) face with the pairwise interactions which may substantially affect the binding affinity. The key residues involved in pairwise interactions on LsIAs are shown in CPK form, whereas the corresponding residues on the receptor are depicted with the Licorice form (a drawing format of graphic representation in visual molecular dynamics (VMD)). The red dashed line represents the contacts that are much stronger in this form of LsIA/α3β2 complex. Snapshots were chosen by visual inspection for illustrative purposes. (C) The distance (Å) between the coupled residues of bound forms. Moreover, 32a represents the LsIA bound type, whereas 32c denotes the LsIA# bound complex. The colored dashed lines represent the median distance of the pairwise residue. The y-axis shows the probability density function of the distance between the coupled residues, which is involved in all the distance (Å) plots in this study.

121

Figure 4.6. Important interactions formed by P7 with the receptor residues on the complementary (−) face at α3(+)β2(−) interface. Binding mode of LsIA (A) and LsIA# (B) at the β2(−) face with the pairwise interactions dramatically influences the binding affinity. (C) The distance (Å) between the coupled residues of bound forms that affect the changes in α3(+)β2(−) interface interactions by the toxin.

However, contrary to the β2 subunit at α3(+)β2(−) interfaces, the changes in receptor

contacts at α3(+), between LsIA and LsIA# bound interfaces were minor, apart from

significantly enhanced contact between α3(+)-Y215 and LsIA#-C3 versus the LsIA bound form

(Figure S1.6). Other interactions, such as N6-(+)Y22, P7-(+)W174 and G2-(+)Y215 only slightly

affected the binding affinity of the toxin due to C-terminal carboxylation (Figure S1.7).

4.3.4. Important Pairwise Interactions of LsIA with Residues at β2(+)α3(−) Interfaces

Carboxylation of the C-T of LsIA also results in markedly enhanced pairwise interactions with

residues on the α3(−) subunit of β2(+)α3(−) interfaces, such as C17-(−)K82 and V11-(−)L134,

via salt-bridge/hydrogen bond and hydrophobic interactions, respectively (Figure 4.4C,4.7).

The increased number of receptor contacts for LsIA# was even more pronounced at

β2(+)α3(−) than that of the α3(+)β2(−) interfaces, except for the weakened contacts between

122

LsIA#-P7 and α3(−)-W80 versus the LsIA bound form. This is shown in Figure 4.4C, in which

LsIA#, occupying the β2(+)α3(−) sites, had a greater relative number of contacts at the key

residues demonstrated previously, such as P7, A8, R10, V11, P14 and C17 (with contact

number differences of up to ~ −30), compared to that at α3(+)β2(−) (with contact number

differences of only up to ~ −15). This suggests that the carboxylation of C-T may cause a

greater enhancement in binding affinity at β2(+)α3(−) compared to that at α3(+)β2(−) binding

sites, such that the interactions at this type of interfaces may also be used to predict LsIA

activity. Nevertheless, it should be noted that the changes in the number of contacts with

statistical significance were commensurate with those at α3(+)β2(−), except for the additional

substantially enhanced van der Waals contacts between R10 and I144 at β2(+)α3(−) bound

by LsIA#.

Similar to the α3(+)β2(−) interfaces, the C-terminal carboxylation may directly enhance the

interactions established by LsIA#-C17 with a lysine residue at the (−) face, namely α3(−)-K82.

Figures 4.4C and 4.7 illustrate that LsIA# had greater contacts between the negatively charged

C-T and α3(−)-K82 on β2 strand of α3 subunit, which was also confirmed in the ligand

interaction 2-D diagram (Figure S1.4B,E). However, it appears that the proximity between

α3(−)-K82 and LsIA#-C17 may not facilitate enhanced interactions by R10 and α3(−)-W80. At

both β2(+)α3(−) and α3(+)β2(−) binding sites, the intramolecular interactions of R10-C17 in

LsIA# may only interfere with the enhanced interactions between LsIA# and α3(−)-W80 as

well as β2(−)-W81, via mainly favoring the binding of LsIA over LsIA# to the binding interfaces.

Further, few variations of interactions occurred at β2(+) subunit (Figure S1.8), resembling the

α3(+)β2(−) interfaces.

In contrast, at β2(+)α3(−), R10 made more increased contacts with three residues for LsIA#

relative to LsIA, namely, α3(−)-I144, K82 and W80 (Figure 4.4). Particularly, the α3(−)-I144

formed substantially enhanced contacts with LsIA#-R10, shown in Figure 4.7. However, at

α3(+)β2(−), LsIA#-R10 had higher contacts only with two residues, β2(−)-L145 and β2(−)-F143.

Apart from this, LsIA#-P14 made enhanced contacts with two receptor residues, α3(−)-K82

and T142 (Figure 4.7), relative to LsIA; this is in contrast to the α3(+)β2(−) interfaces, for which

LsIA#-P14 established markedly enhanced contacts with a single residue, β2(−)-F143.

123

Figure 4.7. Important interactions formed by LsIA-P7, R10, V11, P14 and C17 with the receptor residues on the complementary (−) face at β2(+)α3(−) interface. Binding mode of LsIA (transparent pink) (rotated ~ 180⁰ of the y-axis) (A and C) and LsIA# (transparent gray) (rotated ~ 180⁰ of the y-axis) (B and D) at the (−) face with the pairwise interactions which may substantially affect the binding affinity. For better visualization, the unrelated residues in the figures are shown in transparent color and labelled with smaller size font. (E) The distance (Å) between the coupled residues of bound forms.

4.3.5. Understanding Changes in Electrostatic Potential by C-terminal Carboxylation of LsIA

at α3(+)β2(−) and β2(−)α3(+) Interfaces

In order to understand the effects of C-terminal carboxylation on the electrostatic properties

of LsIA and α3β2 nAChR, which may shed further light on the differences in LsIA and LsIA#

124

activities, the APBS [61] analysis was conducted on LsIAs binding at α3(+)β2(−) and β2(+)α3(−)

via PyMOL APBS plugin [62] (APBS Tools 2.1). PyMOL with an APBS plugin facilitates the

visualization of electrostatic potential 3-D surface molecules after calculation [70]. The

Protein Data Bank (PDB) format file of molecules was pre-processed via pdb2pqr [71]

software for assigning parameters of the atoms, such as radius and charges, from various

force fields [72].

The results demonstrated that the surface of C-terminal residues, C17 and P14, of LsIA obtain

a larger negatively charged surface versus that of LsIA# (Figure 4.8A,B). Additionally, the

pocket formed by β2(−)-F143, L145 and K187 on the β2 subunit of α3(+)β2(−) interfaces

bound by LsIA were predominantly negatively charged. However, a relatively

neutral/positively charged cavity formed by the corresponding residues on LsIA# bound

interfaces was identified. Therefore, electrostatic repulsion between LsIA and α3(+)β2(−)

interfaces may substantially intervene in the α3β2 nAChR binding affinity versus the LsIA#

bound form. For β2(+)α3(−), the surface of the binding cavity around α3(−)-L134 and K82 was

dominated by neutral/positive charge when bound by LsIA# versus the LsIA anchoring form

(Figure 4.8C,D), which resembled the α3(+)β2(−) binding site. An electrostatic interaction

between α3(−)-K82 with LsIA#-R10 (Figure 4.8D) was also observed,which may confer a

stronger repulsion between them. This type of charge-charge interactions was also

determined in previous studies, such as the interactions between α-conotoxin ImI-R11 and

R148 and H146 on LBD of Ls-AChBP [60], as well as R10-K57 on the β4(−) subunit of α3β4

nAChR bound by LsIA [27]. Both of the electrostatic repulsion interactions are associated with

the low binding affinity of ImI and LsIA at these nAChRs, compared to other nAChR subtypes,

such as α7 and α3β2 nAChRs.

125

Figure 4.8. Electrostatic potential energy on the surface of the α3β2 interfaces bound by LsIA/LsIA# shown in PyMOL [73]. (A and C) LsIA binds at α3(+)β2(−) and β2(+)α3(−) interfaces, respectively. (B and D) LsIA# binds at α3(+)β2(−) and β2(+)α3(−) interfaces of α3β2 nAChR. The electrostatic potential is within a range from -2 kT/e (red) to +2 kT/e (blue). The white color represents the neutral (0 kT/e) electrostatic potential.

4.3.6. Inter-subunit Contacts at the α3(+)β2(−) and β2(+)α3(−) Interfaces

To determine the effects of inter-subunit contacts between residues on the binding

interfaces, a network interaction analysis was employed via the software Cytoscape [58].

Differences in inter-subunit contacts help reveal changes in the relative orientations of the

subunits at toxin-bound interfaces, which may be related to the capacity of the toxins to

inhibit nAChR function [39, 74-76].

At the α3(+)β2(−) interfaces, α3(+)-S152 (on the Cys-loop) formed close hydrogen bond

interactions with β2(−)-Q65 and β2(+)-A63, when bound with LsIA. These interactions were

disrupted when LsIA# is bound, which is associated with an increased distance between the

126

interfaces in the lower region (Figure 4.9A). In contrast, the van der Waals interactions formed

by α3(+)-G123 and β2(−)-S129 were increased for LsIA# relative to LsIA, and were associated

with a reduction in the distances between the interfaces in the middle part of the LsIA#-bound

interfaces.

There was also a relative increase in distance between α3(+)-W174 and β2(−)-W81 for LsIA#-

bound α3(+)β2(−) interfaces compared to LsIA. This corresponds to a shorter distance

between these residues for LsIA-bound interfaces. The closeness between α3(+)W174 (on the

β7/β8 loop) and β2(−)W81 (which lies on the β2 strand) when LsIA bound (Figure 4.9A,S1.10),

may have implications for how LsIA interacts with α3β2 interfaces. In particular, these

contacts might lead to an upward shifting of (−)W81, which benefits the enhanced

interactions between β2(−)-W81 and LsIA-R10, and yet indirectly affect the changes in

contacts in other important pairwise interactions. For example, LsIA-P7 and P14 formed

reduced hydrophobic interactions with β2(−)-W81, L145 and F143 (on the β5 strand),

associated with minor decreased contacts between the LsIA-R10 and β2(−)-L145, versus the

LsIA# bound complex.

Overall, at α3(+)β2(−) interfaces, LsIA# binding caused the greater separation between the

α3(+) and β2(−) subunits compared to LsIA at specific regions, including α3(+)-S152 on the

Cys-loop with β2 (−)Q65 and (−)A63, respectively, and α3(+)-W174 with β2(−)-W81, while

reducing the separation at (+)G123 and (−)S129.

127

Figure 4.9. Inter-subunit interactions between the principal and the accessory faces of α3β2 nAChR interfaces bound by LsIA. Contacts at α3(+)β2(−) (A) and β2(+)α3(−) (B) interfaces bound by LsIA and LsIA#. The region where differences in interactions occur due to anchoring of LsIA and LsIA# is shown with green and pink background colors, respectively. The pairwise interactions showing significant divergence in both complexes are shown in bold form.

At the β2(+)α3(−) interfaces, the relatively strong hydrophobic interactions between β2(+)-

W175 (also on the β7/β8 loop) and α3(−)-I144 (on β5 strand) for the LsIA-bound form (Figure

4.9B,S1.10) were identified, which may influence the differentiation between LsIA and LsIA#

bound complex on β2(+)α3(−) interfaces, demonstrating greater contacts for the LsIA# and

residues on the β7/β8 loop and (−)I144 by R10. In summary, LsIA binding caused greater

contacts between the subunits, drawing them closer, especially via the β2(+)-W175 and

α3(−)-I144 residues. This is consistent with the contact difference plots (Figure 4.4), showing

that at the β2(+)α3(−) interfaces, the binding of LsIA# resulted in a much greater increase in

contacts relative to LsIA, when compared to that at the α3(+)β2(−). Part of this may be

because LsIA# less effectively draws together the β2(+) and α3(−) subunits, which favors the

interactions with the toxin relative to LsIA.

128

4.4 Discussion

4.4.1. Different Combinations of the Interfaces in Rat α3β2 may Affect the Distinct Stability

and Flexibility of the Receptor and LsIA

Simulations revealed differences in the overall dynamics and structure of the toxin-receptor

complex when bound with either LsIA# or LsIA. In addition, there are also variations at the

α3(+)β2(−) compared to β2(+)α3(−) interfaces.

The examination of the RMSD shows interface-dependent differences. For the ligands (LsIA

and LsIA#), a distinct pattern of the toxin RMSD values is observed at β2(+)α3(−) (chains G

and J) and β2(+)β2(−) (chain I) interfaces, different from those binding at α3(+)β2(−)

interfaces (chains F and H), for both LsIA and LsIA# bound complexes (Figure 4.2C,D). Although

a negative charge at the C-terminal carboxylate of LsIA# may affect the conformation of the

ligand as well, this asymmetry pattern arising at both LslA and LsIA# RMSD plots may clarify

doubt about the conformational changes induced by C-terminal carboxylation. These

differences in RMSD suggest that both LsIA and LsIA# may adopt different conformations at

the α3(+)β2(−) versus β2(+)α3(−) and β2(+)β2(−) interfaces, particularly at the N-T of the

ligand for both LsIA and LsIA#. Thus, these results may highlight the potential importance of

nAChR interfaces in influencing the preferred conformations of bound conotoxins.

To explore the molecular basis of the differentiation between LsIA# and LsIA, as well as

between the interfaces, the differences in RMSF value are also compared in this study, as

shown in Figures 4.3A and B, indicating the different flexibility of the α3β2, and the toxins, for

the LsIA and LsIA#-α3β2 complexes. The Cys-loop and C-loop of LsIA bound α3 subunit in the

LsIA bound complex showed higher flexibility compared to the LsIA# bound subunit (Figure

4.3A). This appears consistent with the reduced rigidity of Cys-loop observed in several

computational studies on α7 nAChR bound by antagonists versus the apo type [65, 66].

However, to avoid the over-interpretation of the RMSF data, the statistical significance of the

RMSF differences between LsIA and LsIA# bound forms was calculated as well. As a result,

only two residues, Y215 and C217, on the C-loop of α3 subunit, showed significantly enhanced

rigidity upon binding by LsIA#.

129

For LsIA, the averaged differences in RMSF resemble those in LsIA bound at the human α7

nAChR, illustrating a more rigid structure of the loop 2 in the ligand, and relatively flexible

backbone characterizing the loop 1 for LsIA# compared to LsIA. At both mixed subunits,

α3(+)β2(−) and β2(+)α3(−) interfaces, the structure of the loop 2 of LsIA# is generally more

rigid than LsIA. However, this observation (increased rigidity of loop 2 of LsIA#) is only critical

at α3(+)β2(−) binding sites. The remarkably enhanced rigidity of LsIA# compared to LsIA at

α3(+)β2(−) interfaces is qualitatively consistent with the expectation that LsIA# should have

greater toxin-receptor contacts (and therefore less mobile) than LsIA does, given the known

higher potency of LsIA# at α3β2. It is plausible that both of the mixed interfaces may need to

be studied, particularly for the LsIA/α3(+)β2(−) interfaces, which show significant rigidity

differences between the LsIA and LsIA# at loop 2 and C-T. The toxin’s differential potency was

elucidated in regard to their interactions with α3β2 in the following sections.

4.4.2. C-terminal Carboxylation of LsIA Marginally Enhances Overall α3β2 nAChR Contacts

Versus LsIA

Overall, the total number of contacts formed by the LsIA# with the α3β2 nAChR marginally

surpasses that of the LsIA (Figure 4.4A), which is consistent with the experimental study,

concluding that the binding potency of LsIA at α3β2 is improved by three-fold via the C-

terminal carboxylation of the LsIA [9], as mentioned previously. In particular, LsIA#-C17 and

P14 are two key residues that were identified to form remarkably greater contacts with the

receptor compared to LsIA. The enhanced contacts of the LsIA# may be induced by the

potential intra-molecular salt bridges or hydrogen bonds established by the C-T and the R10

(Figure S1.5), which enhanced the rigidity of the residues in loop2, namely V11, N13, P14,

N15, I16 and C-T, rather than R10. This increased rigidity was also suggested in a previous

experimental study using an electrophysiological recording approach and predicted in our

computational study on LsIA/α7 nAChR systems [9, 47]. Therefore, the C-terminal

carboxylation not only directly results in the enhanced receptor interactions by the C-T, but

also improves the receptor interactions by P14 of the LsIA#.

However, the relatively reduced receptor contacts at C3, A8, C9, N12 and I16 may partly offset

this slightly increased number of α3β2 interactions by LsIA# over the LsIA. Furthermore,

130

among all the residues of LsIA/LsIA# strongly interacting with α3β2 nAChR, C3, P7, R10, N12

and P14 show larger variations in terms of the number of receptor contacts over five binding

interfaces. This may suggest that the interactions by the LsIA/LsIA# with an individual type of

α3β2 interfaces, namely α3(+)β2(−), β2(+)α3(−) and β2(+)β2(−), may differ from each other.

Therefore, the identification of the statistically significant changes in receptor contacts by

individual residues was implemented for the prediction of the key residues involved in

enhanced α3β2 interactions by LsIA# versus LsIA, among all α3β2 binding interfaces. It is also

of interest to explore if different α3β2 nAChR interfaces may affect the slightly reduced

binding affinity of LsIA versus its C-terminal carboxylation form.

4.4.3. C-Terminal Carboxylation of LsIA Mainly Enhances the Contacts with Residues on the

Complementary Subunit of both α3(+)β2(−) and β2(+)α3(−) Interfaces

The key important pairwise interactions include that formed by C17 and β2(−)-K187 via

forming hydrogen bonds/salt bridge, which is reinforced in the LsIA# bound complex,

together with a dramatically enhanced hydrophobic interaction between P14 and β2(−)-F143

on α3(+)β2(−) interfaces of α3β2 nAChR versus the LsIA-bound complex (Figure 4.5). The

geometry of LsIA#-C17 confers strong hydrogen bonds or salt-bridges established by the

carboxyl group of C17 with β2(−)-K187. In our previous study, similar hydrogen bonds

between LsIA#-C17 and (−)Q79 of human α7 nAChR were determined [47]. Interestingly, salt

bridges/hydrogen bonds formed by other α-conotoxins and the receptor are involved in

anchoring the inhibitors to nAChRs. For example, β2(−)-K187 at α3(+)β2(−) interfaces of α3β2

nAChR was implicated in charge-charge interactions with E11 of the α-conotoxin MII [20],

whereas human β2(−)-D171 (D196 as numbered from first Met) forms relatively strong

contacts with G1 of α-conotoxin RegIIA via salt-bridge [77], when the identical aspartic acid

residue is situated in position 195 of the rat β2 subunit. Further, the proximity between LsIA#-

C3 and α3(+)-Y215 confers substantially enhanced hydrogen bond contacts by LsIA# with

α3β2 nAChR (Figure S1.6). The (+)Y215 of α3 subunit has been suggested to be important for

the binding of ACh to α3(+)β2(−) binding sites via aromatic interactions [20]. This is also the

only residue on the (+) face at both α3(+)β2(−) and β2(+)α3(−) interfaces showing markedly

different effects on the binding between LsIA and LsIA#.

131

At the (−) subunit on the α3(+)β2(−) interfaces of α3β2 nAChR, other important differential

interactions involve those formed by P7 and P14 with aromatic and hydrophobic residues,

β2(−)-W81, F143 and L145, via hydrophobic interactions, which were significantly enhanced

in the LsIA# bound complex versus the LsIA (Figure 4.5,4.6). The hydrophobic interactions

between residues on α-conotoxins and aromatic residues of nAChRs were also identified in

other α-conotoxin bound complexes. For instance, a hydrophobic interaction is formed

between L15 of LtIA with F119 (F143 from first Met) on the rat β2 subunit of α3β2 nAChR

[36]. This interaction favors the selectivity of LtIA to α3β2 nAChR versus the α3β4 subtype.

On the contrary, the contacts formed by LsIA#-R10 with β2(−)-W81 and L145 slightly affect

the different receptor interactions between LsIA and LsIA#, by compensating the changes in

contacts between each other. More importantly, the stacking interactions (cation- and

− interactions) by R10 were not observed for either LsIA or LsIA# bound interfaces. This

finding differs from the previous studies, illustrating the important role of stacking contacts

in α-conotoxins anchored to nAChRs. The stacking interactions formed by the antagonists

with aromatic residues on nAChRs were determined in numerous previous studies on α-

conotoxins and ACh binding at nAChRs [6, 30, 38]. The proximity of R7 of α-conotoxin ImI with

the aromatic residues, namely Y195, W149 and Y92, on the (+) face of the human α7 nAChR,

may form cation- interactions for stabilizing the ligand and receptor complex [1, 31]. Similar

− interactions were found between α-conotoxin MII-H9 and H12 and (−)F119 at α3(+)β2(−)

interface and the (−)W55 at β2(+)α3(−) interfaces of rat α3β2 nAChR by Sambasivarao and

colleagues [20]. Other cation- and − interactions were determined in the F9 of AuIB with

W59 and K61 on β4(−) subunit of α3β4, and LsIA-R10 with (−)W77 of human α7 nAChR,

respectively [39, 47].

The residues on position 5 of α-conotoxins may contribute to the potency of toxins at nAChRs

via establishing polar interactions with negatively charged residues. In this study, the

persistent hydrogen bond interactions between S5 and β2(−)-D195 stabilized the toxins

bound to α3β2 nAChR in both LsIA and LsIA# complexes (Figure S1.4A, D). It has been

suggested that this aspartic acid on rat β2 subunit also established salt-bridge interactions

with other α-conotoxin at position 5. For example, the R5 of LtIA, associated with S168 [36],

renders the toxin with a shallow binding position versus other α-conotoxins targeting nAChRs.

132

In addition, the aspartic acids in position 197 and 195 at the α3(+) subunit of rat α3β2 and Ac-

AChBP, respectively, also play an important role in the increased sensitivity of TxIA[A10L] to

the receptors via electrostatic interactions with R5 of the toxin [32]. Taken together, our

results support the important role of the charged residues on the F-loop of the β2 subunit of

α3β2 nAChR in selectivity by α-conotoxins.

At β2(+)α3(−) interfaces, for LsIA#, a salt bridge/hydrogen bond interaction exists between

α3(−)-K82 and C17, together with other major hydrophobic interactions, such as LsIA#-V11

and α3(−)-L134 (Figure 4.7 and S1.4), for increasing the binding affinity by the toxin. Similar

to α3(+)β2(−) interfaces, the LsIA#-C17 formed a substantially enhanced interaction with

α3(−)-K82 via electrostatic interactions (Figure 4.8). The hydrogen bond/salt bridge

interaction exerts the most critical impact on enhancing the binding affinity of LsIA# at α3β2

nAChR, associated with another substantially increased pairwise contacts between LsIA-V11#

and α3(−)-L134 via hydrophobic interactions. The α3(−)-L134 residue is also a homologous

residue at position 107 of the human α7 subunit, in which the (−)L107 of α7 subunits forms

hydrophobic interactions with ImI-W10 [31]. A proximal distance between LsIA#-R10 and

α3(−)-I144 was also observed for forming van der Waals interactions versus the LsIA bound

form, with a resemblance to the enhanced van der Waals interactions between LsIA#-R10 and

β2(−)-L145 at α3(+)β2(−). LsIA# binding generally causes a higher association between the

subunits at the Cys-loop region compared to LsIA at α3(+)β2(−) (Figure 4.9). At β2(+)α3(−)

interfaces, the binding of LsIA# binding causes slightly reinforced contacts between the

subunits, drawing them closer together, except for the β2(+)-W175 and α3(−)-I144 residues.

This is consistent with the contact difference plots (Figure 4.4), which show that at both

interfaces, LsIA# shows a higher number of contacts with α3β2 compared to LsIA. This may,

in part, be due to LsIA# being more able to effectively draw together the subunits.

Additionally, α3(−)-I144 is also homologous with (−)L141 on α7 subunit of human α7 nAChR,

which was determined to form an enhanced interaction with LsIA-R10 of LsIA versus the

carboxylated analogue [47]. In summary, the hydrophobic residues, isoleucine and leucine, at

this homologous position, may contribute to the enhanced interactions by LsIA# at α7 and

α3β2 nAChRs versus the wild type toxin.

133

Nevertheless, the interactions between LsIA#-P7 and α3(−)-W80 were reduced, in contrast to

the binding of LsIA via establishing hydrophobic interactions with the pyrrolidine ring of P7.

Thus, the role of α3(−)-W80 and β2(−)-W81 putatively favors the binding of LsIA compared

with LsIA# at both α3β2 interfaces via hydrophobic or van der Waals interactions, despite

enhanced interactions between P7 and β2(−)-W81.

The electrostatic repulsion may undermine the contacts formed by LsIA-R10 with aromatic

residues at the β2(+)α3(−) interface via stacking interactions. The loss of aromatic/stacking

interactions between LsIA-R10 and corresponding aromatic residues probably bears a close

resemblance to that on α3(+)β2(−) interfaces. The closer distance between LsIA#-R10 and C17

via intramolecular salt bridge interaction may induce the electrostatic repulsion between

LsIA-R10 and K82 (Figure 4.8), which forms dramatically enhanced interactions with LsIA#-

C17. A previous study showed that the LsIA [R10F, N12L] facilitates the selectivity of LsIA

binding at human α3β4 over α7, Ls-AChBP and Ac-AChBP [27]. Thus, further study may require

the mutagenesis of LsIA#-R10, possibly substituted by hydrophobic residue, for increasing the

binding affinity of toxins at α3β2. Other studies also suggested the important role of

hydrophobic residues, such as alanine and leucine, at position 10 of toxins, for the augmented

inhibition of nAChRs by PnIA, PnIB and PeIA [S9H,V10A], respectively. A10 of PnIA and L10 of

PnIB are critical for the inhibitory effects by the peptides at α7, α3β4, and α3β4 nAChRs,

respectively [32, 78, 79]. The replacement of alanine by leucine in position 10 of PnIA reduced

the sensitivity to α3β2 [79]. Multiple amino acid substitutions, including position 10 of certain

α-conotoxins, can also affect the inhibition of nAChR subtypes. PeIA [S9H,V10A], for example,

enhanced the selectivity of the toxin to α3β2 and α6/α3β2β3 [80].

The findings in this computational study may provide data for further experimental study. The

in vitro cells, such as Xenopus laevis oocytes [32], human embryonic kidney (HEK-293) cells

[81] and SH-SY5Y cells [82], are widely applied in the functional characterization of nAChRs

and their sensitivity to their potential antagonists (e.g. α-conotoxins and cocaine) [13, 20, 21,

30, 36, 77, 81, 83, 84], namely, α3β2, α3β4, α7, α7β2, and AChBP for the crystallization of the

structure. Among these cell lines, the most commonly used cell is the Xenopus oocyte, which

has been demonstrated to successfully express different α3β4 and α9α10 nAChR

stoichiometries [16, 18].

134

4.4.4 Inter-subunit Contacts of α3β2 nAChR Interfaces may Facilitate the Interactions with

LsIA# Versus LsIA

The previous study has shown an open C-loop (10 to 14 Å) of α3β2 nAChR interfaces may

favor the binding of α-conotoxin MII versus the close C-loop region (7.4 to 10.1 Å) possibly

due to less steric hindrance of the open C-loop region versus the closed one [20]. The

open/closed C-loop region is denoted by the distance between the C-loop of the primary face

and the corresponding complementary face at a binding interface of nAChR. In the present

study, the fewer contacts between LsIA may preferentially ease the inter-subunit contacts at

the binding interfaces of α3β2 nAChR, whilst which was intervened upon bound by LsIA#

possibly due to the increased interactions between the toxin and the residues at α3β2 nAChR

interfaces. Furthermore, the studies on ligands binding at nAChRs and AChBPs have

demonstrated the changes in conformation of the ligand-binding domain of the receptors due

to the binding of ligand [32]. Such conformational changes at the binding domain of nAChR

might assist in the channel function at the transmembrane domain [32, 38, 85]. Though the

studies on nAChRs in this thesis mainly focus on the extracellular binding domain of the

receptors, fewer inter-subunit contacts at α3β2 nAChR interfaces bound by LsIA# might exert

an impact on the preferential binding by LsIA# versus LsIA. However, such assumption needs

to be validated in consideration of both α3β2 nAChR LBD and TMD in future studies.

4.5 Conclusion

In the present study, the molecular models of rat α3β2 nAChR bound by LsIA and LsIA# were

constructed for an investigation of the effects of the C-terminal carboxylation of LsIA on the

receptor interactions versus the wild type LsIA. The findings may provide new insights into

elucidating the effects of different α3β2 binding interfaces, mainly between α3(+)β2(−) and

β2(+)α3(−) interfaces, on the interactions by the LsIA and the residues on α3β2 nAChR.

An examination of the RMSD plots of LsIA at the α3(+)β2(−), β2(+)α3(−) and β2(+)β2(−)

interfaces suggests that the toxins adopt remarkably different conformations at the latter

interface, possibly indicating that the β2(+)β2(−) is unsuited to facilitate LsIA binding and

inhibition. Subsequent discussions were, therefore, largely focused on the mixed-subunit

interfaces. At both α3(+)β2(−) and β2(+)α3(−) interfaces, the carboxylated C-T (C17) formed

135

strong hydrogen bonds/salt bridges with lysines at positions 187 (at α3(+)β2(−) interfaces)

and 82 (at β2(+)α3(−) interfaces), which occur together with the significantly enhanced

interactions by P14 and P7 with hydrophobic residues (L145, W81 and F143) on the (−) face

of α3(+)β2(−) interfaces. In addition, homologous residues, leucine and isoleucine, at

positions 145 and 144 of β2(−) and α3(−) subunits, respectively, may contribute to the

enhanced binding affinity of LsIA# at both interfaces via forming van der Waals interactions

with LsIA#-R10. However, at β2(+)α3(−) interfaces, only LsIA#-V11 was observed to establish

markedly augmented hydrophobic contacts with α3(−)-L134, compromised by the reduced

hydrophobic contacts between LsIA#-P7 and α3(−)-W80. On the (+) face, β2(+)-Y215 exhibited

enhanced hydrogen bond contacts with C3 of LsIA# versus the LsIA bound form at α3(+)β2(−)

binding sites. By contrast, the significantly reduced contacts, for LsIA#, between β2(+)-Y177

and N12 at β2(+)β2(−) interfaces, decreased the total number of receptor contacts by LsIA#.

Combined with the enhanced rigidity of the residues (from position 13 to 17) of LsIA# over

LsIA, it can be proposed that the significant changes due to C-terminal carboxylation may

occur predominantly at α3(+)β2(−)and β2(+)α3(−) interfaces, rather than β2(+)β2(−) sites, via

hydrophobic and salt-bridge/hydrogen bond interactions at the complementary faces of

binding interfaces.

The determinants predicted in this study may provide information for the improvement of

selectivity and specificity of LsIA to the nAChRs. The experimental studies employing the

mutagenesis or synthesis of chemically modified -conotoxins may be conducted in the

future to improve the binding affinity of the toxins targeting nicotinic receptors. Additionally,

further experimental studies on nAChRs in association with computational approaches may

help improve the effectiveness of mediating nicotinic receptor activity by α-conotoxins as

molecular drugs for novel therapies.

136

4.6 References

Hurst R, Rollema H, Bertrand D. Nicotinic acetylcholine receptors: from basic science

Dineley KT, Pandya AA, Yakel JL. Nicotinic ACh receptors as therapeutic targets in CNS

Paterson D, Nordberg A. Neuronal nicotinic receptors in the human brain. Progress in

Lamotte d'Incamps B, Zorbaz T, Dingova D, Krejci E, Ascher P. Stoichiometry of the Journal of Neuroscience. the Renshaw cell.

1. Ellison M, Gao F, Wang H-L, Sine SM, McIntosh JM, Olivera BM. α-Conotoxins ImI and ImII target distinct regions of the human α7 nicotinic acetylcholine receptor and distinguish human nicotinic receptor subtypes. Biochemistry. 2004;43(51):16019-26. 2. Gotti C, Moretti M, Bohr I, Ziabreva I, Vailati S, Longhi R, et al. Selective nicotinic acetylcholine receptor subunit deficits identified in Alzheimer's disease, Parkinson's disease and dementia with Lewy bodies by immunoprecipitation. Neurobiology of Disease. 2006;23(2):481-9. 3. to therapeutics. Pharmacology & Therapeutics. 2013;137(1):22-54. 4. disorders. Trends in pharmacological sciences. 2015;36(2):96-108. 5. Neurobiology. 2000;61(1):75-111. 6. Everhart D, Cartier GE, Malhotra A, Gomes AV, McIntosh JM, Luetje CW. Determinants of potency on α-conotoxin MII, a peptide antagonist of neuronal nicotinic receptors. Biochemistry. 2004;43(10):2732-7. 7. McIntosh JM, Azam L, Staheli S, Dowell C, Lindstrom JM, Kuryatov A, et al. Analogs of α-conotoxin MII are selective for α6-containing nicotinic acetylcholine receptors. Molecular Pharmacology. 2004;65(4):944-52. 8. Gao F, Mer G, Tonelli M, Hansen SB, Burghardt TP, Taylor P, et al. Solution NMR of acetylcholine binding protein reveals agonist-mediated conformational change of the C-loop. Molecular Pharmacology. 2006;70(4):1230-5. 9. Inserra MC, Kompella SN, Vetter I, Brust A, Daly NL, Cuny H, et al. Isolation and characterization of α-conotoxin LsIA with potent activity at nicotinic acetylcholine receptors. Biochemical Pharmacology. 2013;86(6):791-9. Cox BC, Marritt AM, Perry DC, Kellar KJ. Transport of multiple nicotinic acetylcholine 10. receptors in the rat optic nerve: high densities of receptors containing α6 and β3 subunits. Journal of Neurochemistry. 2008;105(5):1924-38. 11. Millar NS, Gotti C. Diversity of vertebrate nicotinic acetylcholine receptors. Neuropharmacology. 2009;56(1):237-46. 12. Miles TF, Dougherty DA, Lester HA. The 5-HT3AB receptor shows an A3B2 stoichiometry at the plasma membrane. Biophysical Journal. 2013;105(4):887-98. 13. heteromeric nicotinic receptors of 2018;38(21):4943-56. 14. Moroni M, Bermudez I. Stoichiometry and pharmacology of two human α4β2 nicotinic receptor types. Journal of Molecular Neuroscience. 2006;30(1-2):95-6. 15. Wu X, Tae H-S, Huang Y-H, Adams DJ, Craik DJ, Kaas Q. Stoichiometry dependent inhibition of rat α3β4 nicotinic acetylcholine receptor by the ribbon isomer of α-conotoxin AuIB. Biochemical Pharmacology. 2018;155:288-97. 16. Mazzaferro S, Benallegue N, Carbone A, Gasparri F, Vijayan R, Biggin PC, et al. Additional acetylcholine (ACh) binding site at α4/α4 interface of (α4β2) 2α4 nicotinic receptor influences agonist sensitivity. Journal of Biological Chemistry. 2011;286(35):31043-54.

137

Dutertre S, Nicke A, Lewis RJ. β2 subunit contribution to 4/7 α-conotoxin binding to

receptor β the nicotinic acetylcholine Shiembob DL, Roberts RL, Luetje CW, McIntosh JM. Determinants of α-conotoxin BuIA subunit. Biochemistry.

Harpsøe K, Ahring PK, Christensen JK, Jensen ML, Peters D, Balle T. Unraveling the 17. high- and low-sensitivity agonist responses of nicotinic acetylcholine receptors. The Journal of Neuroscience. 2011;31(30):10759-66. Boffi JC, Marcovich I, Gill-Thind JK, Corradi J, Collins T, Lipovsek MM, et al. Differential 18. contribution of subunit interfaces to α9α10 nicotinic acetylcholine receptor function. Molecular Pharmacology. 2017;91(3):250-62. 19. Indurthi DC, Pera E, Kim H-L, Chu C, McLeod MD, McIntosh JM, et al. Presence of multiple binding sites on α9α10 nAChR receptors alludes to stoichiometric-dependent action of the α-conotoxin, Vc1.1. Biochemical Pharmacology. 2014;89(1):131-40. 20. Sambasivarao SV, Roberts J, Bharadwaj VS, Slingsby JG, Rohleder C, Mallory C, et al. Acetylcholine Promotes Binding of α-Conotoxin MII at α3β2 Nicotinic Acetylcholine Receptors. ChemBioChem. 2014;15(3):413-24. 21. the nicotinic acetylcholine receptor. Journal of Biological Chemistry. 2005;280(34):30460-8. 22. selectivity on 2006;45(37):11200-7. Zhangsun D, Zhu X, Wu Y, Hu Y, Kaas Q, Craik DJ, et al. Key residues in the nicotinic 23. acetylcholine receptor β2 subunit contribute to α-conotoxin LvIA binding. Journal of Biological Chemistry. 2015;290(15):9855-62. 24. Walsh RM, Jr., Roh S-H, Gharpure A, Morales-Perez CL, Teng J, Hibbs RE. Structural principles of distinct assemblies of the human α4β2 nicotinic receptor. Nature. 2018;557(7704):261-5. 25. Di Cesare Mannelli L, Cinci L, Micheli L, Zanardelli M, Pacini A, McIntosh JM, et al. α- conotoxin RgIA protects against the development of nerve injury-induced chronic pain and prevents both neuronal and glial derangement. Pain. 2014;155(10):1986-95. 26. Nicke A, Loughnan ML, Millard EL, Alewood PF, Adams DJ, Daly NL, et al. Isolation, structure, and activity of GID, a novel α4/7-conotoxin with an extended N-terminal sequence. Journal of Biological Chemistry. 2003;278(5):3137-44. 27. Abraham N, Healy M, Ragnarsson L, Brust A, Alewood PF, Lewis RJ. Structural mechanisms for α-conotoxin activity at the human α3β4 nicotinic acetylcholine receptor. Scientific Reports. 2017;7:45466. 28. Loughnan ML, Nicke A, Jones A, Adams DJ, Alewood PF, Lewis RJ. Chemical and functional identification and characterization of novel sulfated α-conotoxins from the cone snail Conus anemone. Journal of Medicinal Chemistry. 2004;47(5):1234-41. 29. Armishaw CJ. Synthetic α-conotoxin mutants as probes for studying nicotinic acetylcholine receptors and in the development of novel drug leads. Toxins. 2010;2(6):1471. Cartier GE, Yoshikami D, Gray WR, Luo S, Olivera BM, McIntosh JM. A new α-conotoxin 30. which targets α3β2 nicotinic acetylcholine receptors. Journal of Biological Chemistry. 1996;271(13):7522-8. 31. Dutertre S, Nicke A, Tyndall JDA, Lewis RJ. Determination of α-conotoxin binding modes on neuronal nicotinic acetylcholine receptors. Journal of Molecular Recognition. 2004;17(4):339-47. 32. Dutertre S, Ulens C, Büttner R, Fish A, van Elk R, Kendel Y, et al. AChBP-targeted α- conotoxin correlates distinct binding orientations with nAChR subtype selectivity. The EMBO Journal. 2007;26(16):3858-67.

138

Hildebrand PW, Rose AS, Tiemann JKS. Bringing molecular dynamics simulation data

Rucktooa P, Haseler CA, van Elk R, Smit AB, Gallagher T, Sixma TK. Structural 33. characterization of binding mode of smoking cessation drugs to nicotinic acetylcholine receptors through study of ligand complexes with acetylcholine-binding protein. The Journal of Biological Chemistry. 2012;287(28):23283-93. 34. Xu M, Zhu X, Yu J, Yu J, Luo S, Wang X. The crystal structure of Ac-AChBP in complex with α-conotoxin LvIA reveals the mechanism of its selectivity towards different nAChR subtypes. Protein & Cell. 2017:1-11. SMIT AB, BREJC K, SYED N, SIXMA TK. Structure and function of AChBP, homologue of 35. the ligand-binding domain of the nicotinic acetylcholine receptor. Annals of the New York Academy of Sciences. 2003;998(1):81-92. Luo S, Akondi KB, Zhangsun D, Wu Y, Zhu X, Hu Y, et al. Atypical α-conotoxin LtIA from 36. Conus Litteratus targets a novel microsite of the α3β2 nicotinic receptor. Journal of Biological Chemistry. 2010;285(16):12355-66. 37. Dellisanti CD, Yao Y, Stroud JC, Wang Z-Z, Chen L. Crystal structure of the extracellular domain of nAChR α1 bound to α-bungarotoxin at 1.94 Å resolution. Nature Neuroscience. 2007;10:953. 38. Celie PHN, Kasheverov IE, Mordventisev DY, Hogg RC, van Nierop P, van Elk R, et al. Crystal structure of nicotinic acetylcholine receptor homolog AChBP in complex with an α- conotoxin PnIA variant. Nature Structural and Molecular Biology. 2005;12. 39. Grishin AA, Cuny H, Hung A, Clark RJ, Brust A, Akondi K, et al. Identifying key amino acid residues that affect α-conotoxin AuIB inhibition of α3β4 nicotinic acetylcholine receptors. Journal of Biological Chemistry. 2013;288(48):34428-42. 40. Suresh A, Hung A. Molecular simulation study of the unbinding of α-conotoxin [ϒ4E]GID at the α7 and α4β2 neuronal nicotinic acetylcholine receptors. Journal of Molecular Graphics and Modelling. 2016;70:109-21. 41. Ulens C, Hogg RC, Celie PH, Bertrand D, Tsetlin V, Smit AB, et al. Structural determinants of selective α-conotoxin binding to a nicotinic acetylcholine receptor homolog AChBP. Proceedings of the National Academy of Sciences of the United States of America. 2006;103(10):3615-20. 42. Zouridakis M, Papakyriakou A, Ivanov IA, Kasheverov IE, Tsetlin V, Tzartos S, et al. Crystal structure of the monomeric extracellular domain of α9 nicotinic receptor subunit in complex with α-conotoxin RgIA: molecular dynamics insights into RgIA binding to α9α10 nicotinic receptors. Frontiers in Pharmacology. 2019;10(474). 43. Bourne Y, Talley TT, Hansen SB, Palmer T, Marchot P. Crystal structure of a Cbtx-AChBP complex reveals essential interactions between snake α-neurotoxins and nicotinic receptors. EMBO Journal. 2005;24(8):1512-22. 44. Eswar N, Webb B, Marti-Renom MA, Madhusudhan MS, Eramian D, Shen M-Y, et al. Comparative protein structure modeling using Modeller. Current Protocols in Bioinformatics. 2006;Chapter 5:Unit-5.6. 45. Hess B, Kutzner C, van der Spoel D, Lindahl E. GROMACS 4: algorithms for highly efficient, load-balanced, and scalable molecular simulation. Journal of Chemical Theory and Computation. 2008;4(3):435-47. 46. into view. Trends in Biochemical Sciences. 2019;44(11):902-13. 47. Wen J, Hung A. Effects of C-terminal carboxylation on α-conotoxin LsIA interactions with human α7 nicotinic acetylcholine receptor: molecular simulation studies. Marine Drugs. 2019;17(4):206.

139

Shen M-Y, Sali A. Statistical potential for assessment and prediction of protein

Ferreira LG, Santos RNd, Oliva G, Andricopulo AD. Molecular docking and structure-

Humphrey W, Dalke A, Schulten K. VMD: visual molecular dynamics. Journal of

Lerner M, Carlson H. APBS plugin for PyMOL. Ann. Arbor.: University of Michigan.;

Colovos C, Yeates TO. Verification of protein structures: patterns of nonbonded atomic

Kufareva I, Abagyan R. Methods of protein structure comparison. Methods in

48. structures. Protein Science. 2006;15(11):2507-24. 49. Leffler AE, Kuryatov A, Zebroski HA, Powell SR, Filipenko P, Hussein AK, et al. Discovery of peptide ligands through docking and virtual screening at nicotinic acetylcholine receptor homology models. Proceedings of the National Academy of Sciences. 2017;114(38):E8100-E9. 50. Carbone AL, Moroni M, Groot-Kormelink PJ, Bermudez I. Pentameric concatenated (α4)2(β2)3 and (α4)3(β2)2 nicotinic acetylcholine receptors: subunit arrangement determines functional expression. British Journal of Pharmacology. 2009;156(6):970-81. 51. Nemecz A, Taylor P. Creating an α7 nicotinic acetylcholine recognition domain from the acetylcholine-binding protein: crystallographic and ligand selectivity analyses. Journal of Biological Chemistry. 2011;286(49):42555-65. 52. Morales-Perez CL, Noviello CM, Hibbs RE. X-ray structure of the human α4β2 nicotinic receptor. Nature. 2016;538(7625):411-5. 53. based drug design strategies. Molecules. 2015;20(7):13384-421. 54. Newcombe J, Chatzidaki A, Sheppard TD, Topf M, Millar NS. Diversity of nicotinic acetylcholine receptor positive allosteric modulators revealed by mutagenesis and a revised structural model. Molecular Pharmacology. 2018;93(2):128-40. 55. Gonzalez-Gutierrez JP, Hodar M, Viscarra F, Paillali P, Guerra-Díaz N, Pessoa-Mahana H, et al. Minimal structural changes determine full and partial nicotinic receptor agonist activity for nicotine analogues. Molecules. 2019;24(15):2684. 56. Norleans J, Wang J, Kuryatov A, Leffler A, Doebelin C, Kamenecka TM, et al. Discovery of an intrasubunit nicotinic acetylcholine receptor binding site for the positive allosteric modulator Br-PBTC. Journal of Biological Chemistry. 2019. 57. Mark P, Nilsson L. Structure and dynamics of the TIP3P, SPC, and SPC/E water models at 298 K. The Journal of Physical Chemistry A. 2001;105(43):9954-60. 58. Shannon P, Markiel A, Ozier O, Baliga NS, Wang JT, Ramage D, et al. Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Research. 2003;13(11):2498-504. 59. Molecular Graphics. 1996;14(1):33-8. Talley TT, Olivera BM, Han K-H, Christensen SB, Dowell C, Tsigelny I, et al. α-Conotoxin 60. OmIA is a potent ligand for the acetylcholine-binding protein as well as α3β2 and α7 nicotinic acetylcholine receptors. Journal of Biological Chemistry. 2006;281(34):24678-86. Baker NA, Sept D, Joseph S, Holst MJ, McCammon JA. Electrostatics of nanosystems: 61. Application to microtubules and the ribosome. Proceedings of the National Academy of Sciences. 2001;98(18):10037. 62. 2006. 63. interactions. Protein Science. 1993;2(9):1511-9. 64. Molecular Biology. 2012;857:231-57. 65. Yi M, Tjong H, Zhou H-X. Spontaneous conformational change and toxin binding in α7 acetylcholine receptor: Insight into channel activation and inhibition. Proceedings of the National Academy of Sciences. 2008;105(24):8280-5.

140

Dilip A, Lešnik S, Štular T, Janežič D, Konc J. Ligand-based virtual screening interface

DeLano W. The PyMOL Molecular Graphics System [Available from:

Gharpure A, Teng J, Zhuang Y, Noviello CM, Walsh RM, Jr., Cabuco R, et al. Agonist in the α3β4 ganglionic nicotinic receptor. Neuron. ion permeation

Yu R, Craik DJ, Kaas Q. Blockade of neuronal α7-nAChR by α-conotoxin ImI explained 66. by computational scanning and energy calculations. PLoS Computational Biology. 2011;7(3):e1002011-e. Sahu BS, Obbineni JM, Sahu G, Singh PK, Sonawane PJ, Sasi BK, et al. Molecular 67. interactions of the physiological anti-hypertensive peptide catestatin with the neuronal nicotinic acetylcholine receptor. Journal of Cell Science. 2012;125(9):2323-37. 68. Dong Y-w, Liao M-l, Meng X-l, Somero GN. Structural flexibility and protein adaptation to temperature: Molecular dynamics analysis of malate dehydrogenases of marine molluscs. Proceedings of the National Academy of Sciences. 2018;115(6):1274-9. 69. Meher BR, Wang Y. Exploring the drug resistance of V32I and M46L mutant HIV-1 protease to inhibitor TMC114: flap dynamics and binding mechanism. Journal of Molecular Graphics Modelling. 2015;56:60-73. 70. between PyMOL and LiSiCA. Journal of Cheminformatics. 2016;8(1):46. Dolinsky TJ, Nielsen JE, McCammon JA, Baker NA. PDB2PQR: an automated pipeline 71. for the setup of Poisson-Boltzmann electrostatics calculations. Nucleic Acids Research. 2004;32(Web Server issue):W665-7. Dolinsky TJ, Czodrowski P, Li H, Nielsen JE, Jensen JH, Klebe G, et al. PDB2PQR: 72. expanding and upgrading automated preparation of biomolecular structures for molecular simulations. Nucleic acids research. 2007;35(Web Server issue):W522-W5. 73. http://www.pymol.org. 74. Kouvatsos N, Giastas P, Chroni-Tzartou D, Poulopoulou C, Tzartos SJ. Crystal structure of a human neuronal nAChR extracellular domain in pentameric assembly: Ligand-bound α2 homopentamer. Proceedings of the National Academy of Sciences. 2016;113(34):9635-40. 75. Henchman RH, Wang H-L, Sine SM, Taylor P, Andrew McCammon J. Ligand-induced conformational change in the α7 nicotinic receptor ligand binding domain. Biophysical Journal. 2005;88(4):2564-76. 76. selectivity and 2019;104(3):501-11.e6. Kompella SN, Cuny H, Hung A, Adams DJ. Molecular basis for differential sensitivity of 77. α-conotoxin RegIIA at rat and human neuronal nicotinic acetylcholine receptors. Molecular Pharmacology. 2015;88(6):993-1001. Hogg RC, Miranda LP, Craik DJ, Lewis RJ, Alewood PF, Adams DJ. Single amino acid 78. substitutions in α-conotoxin pnia shift selectivity for subtypes of the mammalian neuronal nicotinic acetylcholine receptor. Journal of Biological Chemistry. 1999;274(51):36559-64. Luo S, Nguyen TA, Cartier GE, Olivera BM, Yoshikami D, McIntosh JM. Single-residue 79. alteration in α-conotoxin pnia switches its nAChR subtype selectivity. Biochemistry. 1999;38(44):14542-8. 80. McIntosh JM, Plazas PV, Watkins M, Gomez-Casati ME, Olivera BM, Elgoyhen AB. A novel α-conotoxin, PeIA, cloned from conus pergrandis, discriminates between rat α9α10 and α7 nicotinic cholinergic receptors. Journal of Biological Chemistry. 2005;280(34):30107-12. 81. Thomsen MS, Zwart R, Ursu D, Jensen MM, Pinborg LH, Gilmour G, et al. α7 and β2 nicotinic acetylcholine receptor subunits form heteromeric receptor complexes that are expressed in the human cortex and display distinct pharmacological properties. PloS One. 2015;10(6):e0130572-e.

141

Kovalevich J, Langford D. Considerations for the use of SH-SY5Y neuroblastoma cells

82. in neurobiology. Methods in Molecular Biology. 2013;1078:9-21. 83. Cuny H, Yu R, Tae H-S, Kompella SN, Adams DJ. α-Conotoxins active at α3-containing nicotinic acetylcholine receptors and their molecular determinants for selective inhibition. British Journal of Pharmacology. 2017;175(11):1855-68. 84. Ma Z-g, Jiang N, Huang Y-b, Ma X-k, Brek Eaton J, Gao M, et al. Cocaine potently blocks neuronal α3β4 nicotinic acetylcholine receptors in SH-SY5Y cells. Acta Pharmacologica Sinica. 2020;41(2):163-72. 85. Ulens C, Hogg RC, Celie PH, Bertrand D, Tsetlin V, Smit AB, et al. Structural determinants of selective alpha-conotoxin binding to a nicotinic acetylcholine receptor homolog AChBP. Proceedings of the National Academy of Sciences of the United States of America. 2006;103(10):3615-20.

142

Chapter 5 : Differentiation of Cav3.2 Interactions by μ-

Conotoxins as Potential Analgesics

143

5.1 Introduction

The calcium (Cav) channels regulate the calcium ion influx through the membrane in response

to membrane depolarization [1]. This calcium channel subfamily is different from sodium

(Nav) and potassium (Kv) channels in that they selectively mediate the calcium ion influx across

the cell membrane versus sodium and potassium ions, based upon the fact that the

concentration of both sodium and potassium ions are remarkably higher than the calcium

ions, respectively, within plasma and cytosol of cells. Therefore, the unique physiological

natures of Cav channels have attracted tremendous attention to explore their functional

kinetics and mechanism of action of individual calcium channel isoforms [2-5].

In general, these calcium channels were sub-categorized into three main groups, being Cav1

(including Cav1.1-1.4), Cav2 (Cav2.1-2.3) and Cav3(Cav3.1-3.3), depending on the gene

subfamily of α1 subunit [6, 7]. This subunit forms the pore and the voltage sensing domains

(VSDs) of Cav channels, with four repeats (DI-IV) at the transmembrane domain (TMD) (Figure

5.1). Each repeat is constituted by six segments (S1-6), in which S1-4 forms the VSD

surrounding the S5 and S6 helices, which construct the pore of Cav channels [8]. Above the

vestibule of the pore (S5-6), there are two P helices (P1 and P2) connecting S5 and S6 together

with the extracellular loops. Between these two helices, a selectivity filter (SF) ring formed by

four negatively charged residues, EEEE for Cav1 and Cav2 and EEDD for Cav3 [9, 10],

demarcates the narrowest region of the channel as a bottleneck. Importantly, the EEEE or

EEDD ring determines the calcium ion selectivity for corresponding Cav channels [11]. In the

extracellular domain (ECD), the extracellular loops between S5 and P1 (S5 loop), as well as P2

and S6 (S6 loop), form contacts with the α2 moiety of the α2δ auxiliary, in which the Von

Willebrand factor (VWF) A domain with a metal-ion-dependent adhesion site (MIDAS)

stabilizes the ligand targeting α2δ accessory protein of Cav channels, such as gabapentin [12].

144

Figure 5.1. Conformation of homology model Cav3.2 in apo form and bound by μ-conotoxin GIIIA. (A) Top view of 3-D structure of apo Cav3.2 with 4 domains shown in yellow, red, blue and grey colors, respectively. (B) Conformation of μ-conotoxin GIIIA and its sequence alignment with other μ- conotoxins, PIIIA, GIIIB, GIIIC and KIIIA. Top (C) and side (D) view of binding conformation by GIIIA at Cav3.2. The extracellular loops of each domain were removed for clarifying the main structural components of Cav3.2 model based on the template 6KZO.

On the other hand, the calcium channels can also be classified as high (H), intermediate (I)

and low (L) voltage activated (VA) channels depending on the physiological function of the α1

145

subunit. Cav1 (L-type) is classified into the group of HVA channels with slower inactivation and

large single-channel conductance, while Cav3 (T-type) exemplifies the LVA channels that

provide swifter inactivation and small single-channel conductance [13-15]. The N-type

calcium channel, Cav2.2, represents the IVA channels, for which the conductance and the

inactivation kinetics are between those for L- and T-type calcium channels [16, 17].

T-type calcium channels are suggested to be involved in the intrinsic neuronal oscillations [18,

19], neuronal firing and excitability [20, 21], cell survival, proliferation and differentiation [22-

24]. Thus, the abnormal function of such calcium channels may result in numerous

pathophysiologies, namely epilepsy, hypertension, cardiovascular diseases, neuropathic

pains, cancers and other pathologies [25-30]. In particular, Cav3.2 is related to inflammation,

neuropathic pains, chronic itch, children absence epilepsy and idiopathic generalized

epilepsies putatively due to its unique role in neuronal excitability via the nociception

pathway [26, 31-34]. However, knowledge about the specific functional mechanism and

physiological properties of Cav3.2 is sparse, and studies to develop subtype selective probes

of Cav3.2 are currently ongoing. Furthermore, previous studies illustrated the upregulated

expression of Cav3.2 induced chronic seizures for animals [21], whilst the silencing of Cav3.2

interrupted the nociception pathway and lessened the calcium currents through the ion

channel [33]. Therefore, the discovery and development of drug leads antagonizing the Cav3.2

are in demand.

The conotoxins are disulphide rich peptides (10 to 40 amino acid in length), extracted from

the marine cone snail venoms [35]. The diverse species of the cone snails make them potent

antagonists/agonists targeting nAChRs, calcium, sodium and potassium ion channels with

selectivity [36-42], through producing numerous conotoxins with distinct paralysis

mechanisms by each species [43]. To date, the intrathecal ziconotide (ω-conotoxin MVIIA) has

been approved by the FDA as a potent analgesic, which acts via blocking the pore of N-type

calcium channel [44-46]. Many other ω-conotoxins, therefore, are widely studied for

improving the selectivity and binding affinity at the Cav2.2 [47, 48]. Conversely, few

conotoxins have been demonstrated to selectively inhibit Cav3.2, which plays a role in chronic

pain, as mentioned previously. A recent publication illustrated the inhibitory effects of δ-

conotoxin TxVIA to human(h) Cav3.2 (IC50 240 nM) via binding to the S3-4 linker at DIV of the

calcium channel as gating modifier [36]. The binding mechanism of TxVIA resembles that of

146

other animal venoms, such as spider and scorpion toxins targeting the T-type calcium channel

[49-51].

Nonetheless, the potential antagonists targeting Cav3.2 should not be limited to the toxins

from the sub-family that have been suggested to inhibit the channel. A recent experimental

study has demonstrated the selectivity of μ-conotoxin GIIIA to Cav3.2 versus the other T-type

calcium channels (not published). μ-Conotoxins possess three disulphide bonds linking six

conversed cysteine residues (Figure 5.1), which may be critical for their functional features.

These disulphide bonds also render the toxins with 3 loops. Notably, the previous studies have

illustrated the pore blocking mechanism of μ-conotoxins at TTX-sensitive ion channels [37,

52-56]. With the highly similar construction of the pore forming α1 subunit being comparable

to that of the sodium channels [51, 57, 58], Cav3.2 might be a novel pharmacophore targeted

by some other μ-conotoxins, such as PIIIA, GIIIB, GIIIC and KIIIA, compared with GIIIA. Also, it

would be interesting to further explore the binding mechanism of these μ-conotoxins at

Cav3.2.

The characterized α1 subunit of hCav3.1 facilitates computational studies on T-type calcium

channels [7, 36]. The cryo-EM structure of Cav3.1 bound by ligands (PDB ID: 6KZP) and in its

apo form (PDB ID: 6KZO) was uncovered by Zhao et al [7]. The determined structure of Cav3.1

prompts the computational studies on investigation of binding mechanism and determinants

of the toxins at T-type calcium channels. For instance, a recent study on δ-conotoxin TxVIA

has revealed the binding pocket at Cav3.2 and the interactive residues playing important roles

in the binding of the toxin [36]. By contrast, few previous studies have investigated the

binding conformation and important pairwise interactions between the toxins and T-type

calcium channels via computational techniques [50, 59]. This was due to the low sequence

identity (less than 22%) of the T-type calcium channels with Cav1 and Cav2 channels [7, 13],

though the structure of rabbit Cav1.1 has been characterized in experiments [60].

In the present study, the binding affinity of GIIIA at hCav3.2 was explored, which will be

compared with that of other μ-conotoxins bound complexes, including PIIIA, GIIIB, GIIIC and

KIIIA via Molecular Mechanics Poisson-Boltzmann Surface Area (MMPBSA) approach [61-63].

GIIIA is the prototypical -conotoxin experimentally demonstrated to exhibit inhibitory

activity against Cav3.2 via patch clamp experiments. Subsequently, a number of other -

147

conotoxins have also been characterised, with varying degrees of activity (unpublished). In

particular, PIIIA also inhibited the hCav3.2 with high potency as GIIIA, whilst the other

conotoxins, GIIIB, GIIIC and KIIIA demonstrated weak inhibitory effects on T-type calcium

channels, including Cav3.1, Cav3.2 and Cav3.3 (D.J. Adams, personal communication).

Molecular modelling and dynamics simulations are presently used to predict the factors that

may be responsible for the binding affinity of GIIIA at Cav3.2. Additionally, simulations of the

other conotoxins, with varying activities, may also enable the delineation of the key

interactions required for inhibition. The interactions by the toxins with Cav3.2 will be

subsequently simulated using the GROMACS packages [62, 64]. This will provide insight into

the key pairwise interactions of Cav3.2, and a number of toxins of varying inhibitory potencies,

that exert prominent impacts on differentiating the binding affinity of the toxins from that of

GIIIA. The results show that GIIIA and PIIIA predominantly occlude the channel pore at the

outer vestibule of Cav3.2, whereas GIIIB is accommodated by the channel at a shallow binding

location. By contrast, the toxin GIIIC might be less sensitive to Cav3.2 due to the loss of

important salt-bridge or hydrogen bond interactions with E1504 and K1809 on the calcium

channel. The findings may shed light for further study for developing Cav3.2 specific analgesics

with enhanced potency.

5.2 Methodology

In this study, the homology model of hCav3.2 was constructed using the cryo-EM structure of

Cav3.1 (PDB ID: 6KZO) in apo form [7] as the template (Figure 5.2). Molecular dynamics (MD)

simulations were performed on the comparative model of Cav3.2 for structural refinement,

and the top three most highly populated structures of the protein were selected through

cluster analysis for the subsequent docking of the μ-conotoxins, PIIIA, GIIIA, GIIIB, GIIIC and

KIIIA, to the calcium channel. The five docking complexes then went through MD simulations

for structural relaxation and refinement of the initial predicted binding poses. After the

simulations, the output trajectories were analysed via MMPBSA [63] for predicting the

binding free energy of each μ-conotoxin at Cav3.2 using the GROMACS packages [65].

148

Figure 5.2. Methodology flowchart involved in the present study of Cav3.2 using Microsoft Visio [66]. The homology model of Cav3.2 was constructed using the template of the cryo-EM structure of Cav3.1 in apo form (PDB ID: 6KZO). The cluster analysis of MD simulation trajectory of apo Cav3.2 was used to determine the top ranked cluster models of the calcium channel, which will be bound by the µ- conotoxins (PIIIA, GIIIA, GIIIB, GIIIC and KIIIA) via docking approach. The best binding models were then selected for further refinement of the protein-ligand complexes via molecular dynamics simulations. Further analysis, including MMPBSA, was subsequently applied to the simulation trajectory of µ-conotoxin bound Cav3.2 complexes.

5.2.1 Homology Modelling

The apo Cav3.1 (PDB ID: 6KZO) was selected as the template for building a comparative model

of Cav3.2 with a sequence identity of around 51% [7]. The amino acid sequence of Cav3.2

(UniProtKB: O95180) was obtained from the Uniprot website (https://www.uniprot.org/). In

general, the sequence of the interested calcium channel, Cav3.2, may be aligned with that of

Cav3.1 for constructing the model of the ion channel in interest [67]. Such a process can be

implemented via homology modelling software or webserver [68-70]. In this study, the cryo-

EM structure of Cav3.1 and the sequence of Cav3.2 were submitted to the homology modelling

web server, SWISS-MODEL [68, 71, 72], for building the model of Cav3.2, which was also

conducted in the previous studies [67, 73].

5.2.2 Molecular Dynamics Simulations of Apo Cav3.2

The comparative model of apo Cav3.2 built via homology modelling was subsequently

processed via MD simulations. The Cav3.2 was situated in the centre of a grid box with the

initial dimensions of 117.931 × 118.697 × 119.735 Å3. The 1,2-dioleoyl-sn-glycero-3-

149

phosphocholine (DOPC) bilayer (359 DOPC molecules) was introduced to embed the model

of Cav3.2 (Figure 5.3). However, lipid bilayer was not included for the nAChR and LsIA systems,

since the focus there is on the ECD of the receptors only, as explained in section 2.2.9 in

Chapter 2. The complex (Cav3.2 and DOPC) was fully solvated by 34022 water molecules

(TIP3P) [74] and 201 Ca2+ and 416 Cl- were added to produce an ionic concentration of

approximately 0.15 M. The whole system was energy minimized in GROMACS (v2016) [75]

with a maximum of 5,000,000 steepest descent steps. The parameters and coordination of

the atoms in the CHARMM27 force field [65, 76] were used to compute the dynamics of the

calcium channel through GROMACS. The LINCS algorithm was used to constrain covalent

bonds [77]. All simulations employed in 2 fs time steps. The van de Waals cut-off is set up at

10 Å. Other parameters are commensurate with those in a previous study by the author of

the present thesis [78]. The system was put through several equilibration steps, in which the

non-hydrogen atoms of the protein were positionally restrained using a force constant of

1000 kJ mol-1nm-2. Firstly, in the NVT ensemble, the temperature (300 K) was kept constant

by the V-rescale thermostat [79], with the equilibration run of 100ps in duration. Secondly,

the velocity rescaling was also implemented in the NPT ensemble to control the temperature

up to 300 K over 25 ns of the subsequent equilibration step. The constraint was removed in

the final constant NPT equilibration simulation, which ran for 250 ns in order to reach

equilibrium of the system. The production run MD simulation for apo Cav3.2 was conducted

with 8 independent simulations for 1000 ns each, with the initial velocities of each simulation

randomly assigned using a different random seed. Consequently, the total simulation time for

the system is 8000 ns (1000 ns × 8). The trajectories were saved every 10 ps for further

analyses.

150

Figure 5.3. Molecular dynamics simulation system of Cav3.2 in apo form. The Cav3.2 model is shown in grey color, and which was inserted into the DOPC bilayer. DOPC and Cav3.2 complex was subsequently immersed in water within a rectangular box with dimensions of 117.931 × 118.697 × 119.735 Å3. The Ca2+ and Cl- were added to neutralize the system and achieve 0.15 M concentration.

The simulation trajectories were processed for cluster analysis using the GROMACS packages

[64]. The interval of the trajectory frame was set as 10. Thus, the frame was extracted every

1ns from the trajectories for 1000 frames in total within a single 1000ns trajectory. The cut-

off of cluster analysis is 3.5 Å with up to 20 models generated. The cluster analysis plots are

shown in Figure S2.1. The three highest ranking models, model 1-3, are the protein targets

bound by μ-conotoxins via docking methods, accounting for around 50% of the whole

trajectory.

Compared with the original template (6KZO) structure, the overall quality factor of the

comparative model of Cav3.2 is 86.8 using ERRAT [80], which is higher than that of the

template (83.2). The quality of structure of Cav3.2 model was improved particularly at the

extracellular loops at DIII (residue K1455 to P1462) and P1 helix at IV (residue N1787 to

R1797), compared with the original structure (6KZO). The structural violation that occurred

at extracellular loops at DI of 6KZO was reduced at the Cav3.2 model, which was further

improved via MD simulation of the Cav3.2 in apo form, giving quality factors of these three

cluster models as 90.2% (model1), 90.2% (model2) and 90.5% (model3), respectively.

151

5.2.3 Docking of μ-Conotoxins to the Ion Channel.

Docking approaches are widely used in predictions of protein-ligand binding structures using

scoring functions which may also provide an estimate of binding affinities [81-84]. To date,

ToxDock (web-form of Rosetta) [87], HADDOCK [88] and Maestro [89]. In the present study,

many docking protocols have been developed, such as AutoDock Vina [81], Rosetta [85, 86],

AutoDock Vina was applied [81] for scoring the binding of individual μ-conotoxins and

producing the putative binding conformations of the protein-ligand complexes [84]. The 3-D

structures of PIIIA (PDB ID: 1R9I), GIIIA (PDB ID: 1TCG), GIIIB (PDB ID: 1GIB), GIIIC (PDB ID:

6MJD) and KIIIA (PDB ID: 2LXG) were obtained from the PDB (www.rcsb.org), and docked into

the top three ranked cluster models of Cav3.2 respectively. The grid box of docking was set

up with dimensions of 60 × 60 × 60 Å3 to cover the entire pore of Cav3.2, as shown in Figure

5.4. Flexible rotatable bonds were assigned to the toxin sidechains, with the toxin backbones

and protein atoms kept rigid. After the docking, the binding score of the top bound complexes

are shown in Table S2.1. The lowest binding affinity represents the highest binding score.

Figure 5.4. The size of the docking grid set up using MGLTools 1.5.6 (www.mgltools.scripps.edu). The number of points in x (red), y (green) and z (blue) dimensions show the grid box dimensions (60 × 60 × 60 Å3) in this study. The cluster model of Cav3.2 was shown in white color with NewCartoon drawing method.

152

5.2.4 Molecular Dynamics Simulations of Cav3.2 and μ-Conotoxin Complexes

The docking complexes with the highest binding scores and the most favourable binding

conformations above the pore of Cav3.2 were obtained using the AutoDock Vina [81] protocol

for each conotoxin bound form. These protein-ligand complexes were used as the initial

structures for MD simulations of Cav3.2/μ-conotoxin complexes. The composition of each

Cav3.2/μ-conotoxin system is listed in Table S2.2. Similar to the apo Cav3.2 system, the

protein-peptide complexes were also energy minimised within a maximum of 5,000,000

steepest descent steps. The systems ran for 100 ps and 50 ns in NVT and NPT ensembles,

respectively, with constraint on all protein-ligand complexes. The production simulation for

each system produces three individual trajectories with approximately 128 ns duration for

each. The parameters and the force field for μ-conotoxins bound Cav3.2 complexes

simulations are commensurate with those in the apo Cav3.2 system. Subsequently, the MD

simulation trajectories were analysed via MMPBSA [61, 62, 90] calculations performed on

frames extracted from the last 10ns of the trajectories, whilst inter-molecular contacts were

calculated using the gmx_mindist code over the whole trajectory for each system.

5.3 Results

5.3.1 Flexibility and Stability of Cav3.2 and the Toxins upon Binding at the Membrane Ion

Channel.

The stability of Cav3.2 bound by μ-conotoxins and in its apo form was predicted via C-α based

root-mean-square deviation (RMSD) analysis (Figure 5.5). For the μ-conotoxin-bound Cav3.2

complexes, the mean values of RMSD, were calculated over 3 individual simulations with a

length of 128 ns for each as shown in Figure 5.5A,B. Figure 5.5A shows the C-α RMSD plots of

Cav3.2 bound by five μ-conotoxins identified and characterised for the activities against Cav3.2

in previous experimental studies, GIIIA, PIIIA, GIIIB, KIIIIA and GIIIC. It can be seen that all

forms of Cav3.2 bound by μ-conotoxins achieved its stability around 30 ns, while the final

structure of KIIIA and GIIIA bound ion channels are more similar to its starting structure

among five μ-conotoxins bound Cav3.2 complexes, due to a smaller RMSD value. Therefore,

regarding the stability of these conotoxins, KIIIA and GIIIA retain a more stable structure

during MD simulation versus that of other μ-conotoxins upon anchoring to Cav3.2 (Figure

153

5.5B). For apo Cav3.2, the RMSD was calculated as the mean of RMSD values over 8

trajectories, with the structure being stabilized by around 80 ns (Figure 5.5C).

Figure 5.5. Root-mean-square deviation (RMSD) plot of Cav3.2 α1 subunit in apo form and bound by five μ-conotoxins, namely GIIIA, PIIIA, GIIIB, KIIIA and GIIIC (Mean ± SD). RMSD plots are depicted for (A) Cav3.2 bound by μ-conotoxins, (B) the toxins when anchoring to Cav3.2 and (C) Cav3.2 in apo form. The solid lines represent the mean value of the RMSD (nm) over 8 and 3 individual simulations for apo Cav3.2 and Cav3.2 bound by μ-conotoxins complexes, respectively, while the corresponding transparent colors with ribbon drawing method show the standard deviation (SD).

154

To further interpret the dynamics of Cav3.2 bound by μ-conotoxins and in its apo form, the

RMS fluctuation (RMSF) analysis was used for evaluating the flexibility of four domains (DI,

DII, DIII and DIV) of the membrane ion channel with a DIII-IV linker, as shown in Figure 5.6.

155

Figure 5.6. Differences in RMS Fluctuation (Diff in RMSF) depicted for four (I, II, III and IV) domains of Cav3.2 α1 subunit bound by other μ-conotoxins compared with that of GIIIA bound form. RMSF values of GIIIA bound Cav3.2 were subtracted from those of individual μ-conotoxins bound forms, respectively. The positive value represents the increased flexibility of Cav3.2 bound by corresponding

156

μ-conotoxins versus GIIIA, whereas the negative values show the reduced flexibility of Cav3.2 anchored by these toxins, when compared with GIIIA binding to Cav3.2 complex.

To elucidate the differences between the toxin binding, an absolute value of differences in

RMSF was set up at 0.15 nm as a threshold. The difference in RMSF was calculated for each

complex relative to the RMSF calculated for GIIIA bound Cav3.2, which serves as the

prototypical -conotoxin demonstrated to exhibit inhibition against Cav3.2. Thus, only the

changes of RMSF which were either above or lower than 0.15 nm (including 0.15 nm) were

treated as considerable changes in RMSF, for comparing the structural rigidity of Cav3.2 bound

by the individual toxin with the GIIIA binding. The statistical significance was subsequently

conducted for the selected ‘Diff in RMSF’ values. Consequently, the residues which show

difference in rigidity for Cav3.2 bound by GIIIA compared to the other μ-conotoxins are shown

in Table 5.1.

Basic residues in the extracellular loops exhibited marked differences in RMSF between GIIIA

and the other toxins. Compared with the GIIIA and Cav3.2 complex, residue R334 (at S5 loop)

of Cav3.2 is more flexible at DI of the calcium channel bound by PIIIA, GIIIB and KIIIA. Notably,

the KIIIA-bound Cav3.2 embraces the most extent of enhanced flexibility versus the binding

of GIIIA, particularly at the extracellular loop of all four domains, among all the toxin-bound

complexes. Apart from R334 of Cav3.2, other positively charged residues, such as K944 on DII

and R1781 on DIV of Cav3.2 also contributed to the flexibility at the extracellular loop of Cav3.2

upon binding by KIIIA, compared to GIIIA. Furthermore, non-polar residues also exhibited

substantial differences in flexibility. In particular, A332 and A333 at DI and L943 at DII, also

contributed to the increased flexibility of the extracellular loop of Cav3.2 bound by the several

toxins compared to GIIIA bound calcium channel. Conversely, the enhanced rigidity of DIII-IV

linker of Cav3.2, mostly resulted from the remarkably reduced flexibility at R1585, D1596,

Y1597, T1600 and R1601 on Cav3.2 in PIIIA, GIIIB, KIIIA and GIIIC bound forms versus the GIIIA

bound protein.

Thus, it was concluded that the positively charged residues, namely R344, K944 and R1781 on

the extracellular loops of Cav3.2 at domains I, II and IV, show enhanced rigidity of Cav3.2

bound by GIIIA compared with the rest of the μ-conotoxins in the present study. Conversely,

several residues on the DIII-IV linker showed reduced rigidity of Cav3.2 bound to GIIIA versus

157

the other conopeptides, which are associated with negatively charged (D1596), aromatic

(Y1597) and hydroxyl (T1600) amino acids at DIII-IV linker. It is also noted that there were only

moderate changes in RMSF identified between the GIIIC and GIIIA bound complexes.

In order to further explore whether the increased/decreased flexibility will affect the binding

affinity of μ-conotoxins anchoring to Cav3.2, MMPBSA and important intermolecular

interactions identification calculations were conducted, which will be discussed in the

following parts.

Table 5.1. The important residues contribute to the changes of structural flexibility of Cav3.2 bound by μ-conotoxins with statistical significance (p < 0.05), compared with the GIIIA bound form.

Toxins

PIIIA

GIIIB

KIIIA

GIIIC

Structure

Increased

DI-R334(S5 loop)

DI-R334(S5

DI-A332,

DI-

Flexibility

loop)

A333, DI-R334

(S5

loop), DII-

L943, DII-K944

(S5 loop), E1566

(DIII-IV

linker),

DIV-R1781

(S5

loop)

Decreased

DII-R788 (loop of

R1585(DIII-IV

R1585, T1600,

R1585(DIII-IV

S1),R1585,

linker)

R1601

(DIII-IV

linker)

Flexibility

D1596,Y1597(DIII-

linker)

IV linker)

158

5.3.2 Binding Free Energy of μ-Conotoxins upon Anchoring to Cav3.2 Relative to GIIIA.

The MMPBSA method [63, 91] can predict the relative free binding energy (G) of μ-

conotoxins compared with GIIIA at Cav3.2, via the g_mmpbsa package [62, 92]. The MMPBSA

method has been widely applied in drug design, such as the development of anticancer drug

leads and neurological disease medications [93, 94], via calculation of the binding affinity of

peptide/protein-protein complexes [61, 91]. The total free binding energy (G) of GIIIA

anchoring to Cav3.2 was selected as the control, which was subtracted from the binding

energy of other individual μ-conotoxins. As a result, the G represents the free binding

energy of the toxin relative to that of GIIIA targeting Cav3.2. The positive value of G means

the neuropeptide shows the relatively weaker binding affinity upon binding to Cav3.2,

compared with GIIIA [95]. In contrast, a higher binding affinity of the toxin is indicated by the

negative values of G.

The results show that two μ-conotoxins, PIIIA and KIIIA, are predicted to have greater binding

affinity at Cav3.2 versus GIIIA (Figure 5.7A). By contrast, compared with the anchoring of GIIIA,

the GIIIB toxin demonstrates a dramatically lower binding affinity to Cav3.2, while there are

only minor changes in the affinity upon binding by GIIIC versus GIIIA. However, there were no

statistically significant differences identified between the total G values of these μ-

conotoxins, over 3 independent simulations for each μ-conotoxin and Cav3.2 complex,

involved in the present study.

159

Figure 5.7. Relative free binding energy (G) of μ-conotoxins PIIIA, GIIIB, KIIIA and GIIIC versus GIIIA at human Cav3.2 via MMPBSA. (A) The relative free binding energy of each μ-conotoxin at Cav3.2. PIIIA, GIIIB, KIIIA and GIIIC are colored by lime, yellow, light purple and red colors, respectively. (B) The decomposed binding energy contributed by individual residues of Cav3.2 bound by μ-conotoxins (* p < 0.05).

Since the total relative binding affinity of the toxins provides few significant details, it might

be helpful to investigate the contribution of the individual residues on the receptors to the

binding of GIIIA versus other individual μ-conotoxins. The relative binding energy contributed

by each residue was calculated via a Python script developed for the g_mmpbsa ensemble

[61, 62]. The process for obtaining the individual contributions from residues of Cav3.2 to the

160

relative free binding energy (G) of the toxins resembled the process for calculating the

total relative binding affinity described above. The free binding energy values of individual

residues on Cav3.2 bound by GIIIA were subtracted from those of the compared μ-conotoxins

bound residues. To underline the significant changes in the binding affinity, only residues for

which the absolute value of G lies above 4 kcal/mol were considered for subsequent

statistical analyses. The outcomes are depicted in Figure 5.7B. The results in this figure are

described separately below for each -conotoxin, compared with the behaviour of GIIIA.

• PIIIA

The aspartic acid at position 1504 of Cav3.2 contributed most to the reduced binding affinity

by PIIIA versus GIIIA at Cav3.2. Notably, this negatively charged residue is one of the four SF

residues (EEDD) of Cav3.2 [96], which are responsible for the calcium ions selectivity [97].

However, the other SF residues, E378, E974 and D1802, negligibly influenced the changes in

affinity between PIIIA and GIIIA upon anchoring to Cav3.2. Residues close to the SF, such as

D382, show significantly increased binding affinity of PIIIA versus GIIIA. The increased affinity

was, however, compensated by the reduced affinity at D1460 on the S5 loop at ECD.

Nonetheless, D975, next to E974 (SF) at DII, did not affect the binding by PIIIA compared with

GIIIA at Cav3.2. Similarly, marginal differences were observed in binding affinity of other

residues on the extracellular loops of Cav3.2 between PIIIA and GIIIA bound forms, such as

E282, E283, E286, D953, D1512, E1777 and E1814. In summary, the substantially greater

binding affinity at D382 on Cav3.2 was undermined by the reduced affinity at E1460 and

D1504 upon binding by PIIIA relative to GIIIA.

• GIIIB

Commensurate with PIIIA, the SF residue at DIII, D1504, exerted substantial impact on the

lower binding affinity of GIIIB versus GIIIA at Cav3.2. By contrast, E378 (SF) contributed to the

enhanced binding affinity of GIIIB relative to GIIIA. For the residues close to the SF of Cav3.2,

D382 showed increased binding affinity at Cav3.2 anchored by GIIIB compared to GIIIA,

however this enhancement was drastically compensated for the reduction of affinity at E283

on the extracellular loop of the ion channel bound by GIIIB. In addition, E974 and D1802 of

Cav3.2 minorly affected the difference in binding affinity of GIIIB compared with GIIIA

anchored to the calcium channel.

161

• KIIIA and GIIIC

For KIIIA, the important role of D1504 was observed as well, given by a significantly lower

binding affinity at this position of Cav3.2 upon binding by the toxin relative to GIIIA. Another

residue that contributed to the lower binding affinity of KIIIA versus GIIIA, is K1809 on P2 of

Cav3.2 at DIV. Nonetheless, only the residues, namely D382 and D975, which are close to the

SF of Cav3.2 at domains I and II dramatically reduced the binding affinity of GIIIC versus GIIIA,

rather than the SF residues themselves. This compensating changes in affinity of GIIIC versus

GIIIA, due to D382 and D975, may cause the trivial differences between the overall binding

affinities of GIIIC and GIIIA at Cav3.2.

Finally, the SF residue, D1504, of Cav3.2 at DIII, plays a critical role in the different binding

affinity of GIIIA versus the other μ-conotoxins. The other residues on the SF, such as E378,

only resulted in an enhanced binding affinity at this position via the binding of GIIIB versus

GIIIA. The residues close to the SF, such as D382, principally contributed to the greater binding

affinity of PIIIA, but provided a relatively unfavourable contribution towards GIIIB from

anchoring to Cav3.2, compared with GIIIA. However, apart from GIIIC, the binding of other

neuropeptides showed unfavourable energetic contributions by the residues at the

extracellular loop and P2 helix, such as E283 at DI, E1460 at DIII and K1809 on P2 helix at DIV.

The significant changes in G contributed by individual residues of Cav3.2 account for the

increased and reduced total binding affinities upon anchoring to PIIIA and GIIIB, respectively,

as elucidated via the MMPBSA approach. Also, the subtle difference in the total binding free

energy of GIIIC versus GIIIA is attributed to the compensation between the enhanced and

reduced binding affinity at certain residues. However, the higher binding affinity of KIIIA at

Cav3.2 is not fully explained by consideration of the decomposed free binding energy results.

This apparent inconsistency may reflect limitations in the free energy decomposition

methodology applied to receptor-peptide complexes. Thus, it may be useful to supplement

this free energy analysis and directly consider inter-residue contacts between the toxins and

Cav3.2, which will provide further information about the protein-peptide interactions at the

atomic level. The results obtained from gmx_mindist will also be compared with the MMPBSA

outcomes in the following part.

162

5.3.3 Important Pairwise Contacts between μ-Conotoxins and Cav3.2

• Significant differences in the number of contacts between other μ-conotoxins and

Cav3.2 compared with GIIIA bound form

The number of contacts between the determinant residues of Cav3.2 (identified in section

5.3.2) and the individual residues of the toxins for each μ-conotoxin bound complex was

calculated via gmx_mindist in this study. This facilitates finding the significant changes in the

pairwise contacts between GIIIA and other μ-conotoxins bound forms, including PIIIA, GIIIB,

KIIIA and GIIIC. Therefore, the key determinants of the coupled interactions that may

contribute to differences in the free binding affinity of the neuropeptides to Cav3.2, will be

identified and compared with the MMPBSA outcomes. These findings may shed light on

further experimental mutagenesis studies which may help develop selective peptides probing

Cav3.2. The filter condition applied to identify close inter-residue contacts was defined as

follows.

Generally, gmx_mindist analysis produces two sets of computational results, which are the

minimum distance between one group, determinant residues of Cav3.2 identified via

MMPBSA free energy decomposition method, and the other group, individual residues of

each μ-conotoxin including GIIIA, with respect to time, as well as the number of contacts

between the atoms of these two groups [98]. Atoms from different residues are considered

to be in contact if the distance between them falls within 4.5 Å in the present work. The

pairwise contacts by GIIIA with Cav3.2 were compared with those of the individual μ-

conotoxins bound complexes. The number of contacts between the pairwise interactions of

GIIIA bound complex was subtracted from those of other toxins anchoring to Cav3.2. Thus,

the negative and positive values demonstrate the increased and decreased number of

contacts by GIIIA, respectively, versus the compared conotoxin groups, with each Cav3.2

residue, as shown in Figure 5.8. For the changes in the number of contacts output, 1) a value

of 11 was selected as the threshold for selecting the number of interactions formed by the

coupled residues for either GIIIA or other μ-conotoxin-bound complexes in a comparison of

the coupled residues on both forms; and 2) the absolute number of differential pairwise

contacts by Cav3.2 and GIIIA from those of the other toxins bound complexes, was selected

to be above 4 for ease of analysis and visualisation. These filter conditions could remove the

trivial pairwise interactions and the marginal changes in the interactions between GIIIA and

163

the other toxins binding at Cav3.2. The selected data set was further processed for statistical

significance analysis. The coupled residues with a significantly different number of contacts

between the toxin bound complexes will be discussed in this study.

Figure 5.8. Changes in the number of contacts between the coupled residues of Cav3.2 bound by μ- conotoxins, PIIIA (green), GIIIB (yellow), GIIIC (red) and KIIIA (light purple) compared with Cav3.2/GIIIA complex. The interacting residues of Cav3.2 and toxin complexes were labelled with the name and number of μ-conotoxins and Cav3.2 residues, e.g. K17 (PIIIA)-D382 (Cav3.2) for each colored bar.

164

In Figure 5.8, PIIIA and GIIIB bound forms provided the highest number of coupled residues

with substantial differences from the control group (GIIIA bound Cav3.2 ensemble). Although

the figure shows a moderate number of pairwise interactions in the KIIIA bound complex

versus GIIIA, all the differences in contacts of the coupled residues are above 20 or below -

20, compared with the changes in the contacts of other toxin bound complexes. In addition,

among the Cav3.2 complexed with the conopeptide models, the changes in contacts of

coupled residues only demonstrated relatively minor differences between GIIIA and GIIIC

bound ensembles. These results are compatible with those of MMPBSA, illustrating KIIIA,

GIIIB, PIIIA and GIIIC in descending order of the differences in binding affinity to Cav3.2 versus

GIIIA.

For each comparison between the interactions by GIIIA and the other neuropeptides with

Cav3.2, half of the residues on Cav3.2 with significant differences in interactions with GIIIA,

are the same as those responsible for the variation trend of the residues determined via the

MMPBSA approach. These residues include D1504 and K1809 in KIIIA bound form, as well as

D382 and D975 in GIIIB and GIIIC bound complexes (Figure 5.7B,5.8). Direct consideration of

inter-residue contacts identified additional residues to those obtained from MMPBSA, such

as the SF residues, E378 and D1504, of PIIIA and GIIIB bound Cav3.2. Compared with GIIIA,

D1504 on Cav3.2 showed higher number of interactions with PIIIA-R14 and K17, though there

is a reduced number of contacts between PIIIA-Q15 and D1504. By contrast, this residue

(D1504) contributed to a lower binding affinity of PIIIA at Cav3.2, with a positive G versus

GIIIA. Similarly, the substantially higher contacts formed by D1504 with R13 counteracted its

reduced pairwise interactions with R14 in the GIIIB bound Cav3.2 complex, relative to the

GIIIA, when MMPBSA results showed a lower affinity contributed by D1504 for GIIIB binding

at Cav3.2. Another SF residue, E378 on Cav3.2 at DI, is also worth consideration. The MMPBSA

results demonstrated that the E378 on Cav3.2 contributed to reduction in the binding affinity

of GIIIB at this position versus GIIIA, whilst consideration of inter-residue contacts indicates a

reduction of contacts between E378 and K11 of GIIIB.

Apart from the SF residues, other negatively charged residues, such as E283, D382 and E1460,

are of interest with regards to comparison between MMPBSA predictions of free energy

contributions, and direct measurement of the number of toxin-receptor contacts. For the

PIIIA-bound Cav3.2 complex, D382 (close to SF) and E1460 (on extracellular loops), formed

165

decreased contacts with PIIIA-R12 and K17 relative to GIIIA, whereas a higher binding affinity

by the toxin at D382 was determined via MMPBSA. By contrast, E1460 substantially

reinforced the interactions with F7 and K9 of PIIIA, whereas the MMPBSA results suggested a

reduced binding affinity of PIIIA at this position of Cav3.2 when compared with GIIIA anchoring

to the ion channel. In addition, E283 formed higher number of interactions with GIIIB-R1 and

D2, while GIIIB exhibited reduced binding affinity at E283 as determined via MMPBSA

analysis.

In summary, it is instructive to compare the binding affinity contributed by individual residues

on Cav3.2, determined by MMPBSA, with direct measures of inter-residue contact between

residues of PIIIA, GIIIB, GIIIC and KIIIA versus GIIIA upon anchoring to the ion channel. The

important differences in number of pairwise contacts of Cav3.2/μ-conotoxin bound

complexes against GIIIA bound complex, are largely consistent with the total binding free

energy of PIIIA, GIIIB and GIIIC computed via MMPBSA, except for KIIIA showing a substantially

reduced number of contacts with Cav3.2. However, since there are no substantial differences

in total binding affinity of toxins determined via MMPBSA method, the focus is chiefly on the

differences in the important pairwise interactions that may affect the binding conformation

of the toxins at Cav3.2 in the following Discussion part.

• Important pairwise interactions between μ-conotoxins and Cav3.2

To further interpret the differences in pairwise contacts between GIIIA and other μ-

conotoxins upon anchoring to Cav3.2, a snapshot of the binding conformation of μ-conotoxins

at Cav3.2 via VMD was used to present the variation of pairwise interactions, as shown in

Figure 5.8. The differences between the important pairwise interactions of GIIIA and

individual μ-conotoxins will be illustrated one by one.

o PIIIA versus GIIIA

The configuration of PIIIA and GIIIA anchoring to Cav3.2 is shown in Figure 5.9, which

demonstrates the varied interactions by the toxins with D382 and D1504 (Figure 5.9A,B) at

the pore vestibule and E1460 (Figure 5.9C,D) on the extracellular loop of Cav3.2. D1504, on

the SF of Cav3.2 at DIII, formed strong contacts with PIIIA-R14 and K17 via salt-bridge or

hydrogen bond interactions, whereas such interactions are much weaker between D1504 and

R13 (homologous to PIIIA-R14) as well as K16 (homologous to PIIIA-K17) of GIIIA, due to the

166

relatively long distance between the residues. By contrast, PIIIA-Q15 lost contacts with

D1504, which instead established a hydrogen bond interaction with GIIIA-Q14. Also, it can be

observed that D382 weakly formed interactions with PIIIA-K17 and R12 due to long distances

between these residues. However, this aspartic acid residue established strong salt-bridge

contacts with GIIIA-K16, assisted by a van der Waals interaction with GIIIA-K11 (homologous

to PIIIA-R12). From Figures 5.9B and C, it is shown that PIIIA-F7 and K9 are in proximity to

E1460 at the S5 loop of Cav3.2 at the ECD of DIII. Thus, van der Waals and electrostatic

interactions by E1460 with F7 and K9, respectively, are present. On the contrary, such

interactions were not found for the GIIIA bound Cav3.2 ensemble, probably because GIIIA is

situated further from the ECD of DIII versus PIIIA (Figure S2.2). In summary, the contacts by

E1460 and D1504 with PIIIA mainly occur with the charged residues of the toxin, K9, R14 and

K17 via salt-bridge and hydrogen bond interactions, which are also associated with a van der

Waals interaction with the pair of F7-D1504. However, these predominant pairwise

interactions were absent at GIIIA, except for a hydrogen bond formed between Q14 and

D1504. By contrast, D382 of Cav3.2 preferentially interacted with two lysine residues of GIIIA,

namely K11 and K16, via electrostatic interactions, which were absent in the binding of PIIIA.

167

Figure 5.9. Significant pairwise interactions between PIIIA and Cav3.2 compared with Cav3.2/GIIIA complex. The coupled residues between Cav3.2 and PIIIA (A,C) as well as GIIIA (B,D). The interacting residues, D382, E1460 and D1504 on Cav3.2 are shown with light purple, mauve and blue colors, respectively. The residues of GIIIA and PIIIA are highlighted by the same color scheme as the corresponding interactive residues. For the residues of the toxins contacting with more than 2 residues (including two residues) on Cav3.2, they are shown with magenta color. The reduced and increased distances between the residues of toxins and Cav3.2 are depicted with red and black dotted lines, respectively. This coloring nomenclature was applied to the following figures illustrating the significant interactions between coupled residues of Cav3.2 and μ-conotoxins complexes. Binding conformations were produced via VMD for interpreting the important changes in G results between GIIIA and other μ-conotoxins upon anchoring to Cav3.2 in this study.

168

o GIIIB versus GIIIA

Amongst the five conotoxins studied, Cav3.2 showed the highest number of interacting

residues with GIIIB which are different from those of GIIIA, involving the residues E283, E378,

D382 and E1504. There are two SF residues of Cav3.2, E378 (at DI) and E1504 (at DIII), which

show substantial differences in Cav3.2 contacts by GIIIB compared with GIIIA (Figure 5.10A,B).

The binding affinity of GIIIB was reduced possibly due to the shallow binding site of GIIIB

relative to GIIIA (Figure S2.3). This binding position, therefore, particularly counterpoised the

interactions formed by GIIIB-K11 and R14 with E378 and D1504, respectively, with high

separation distances between the coupled residues. By contrast, GIIIA-K11 and Q14

(homologous to GIIIB-R14) established salt-bridge/hydrogen bond interactions with the SF

residues of Cav3.2, which contributes to greater binding affinity of GIIIA at the ion channel

relative to the GIIIB. Furthermore, the proximity between GIIIB-R13 and D1504 stabilized the

binding of GIIIB, to a less extent, via forming salt bridges. Similarly, D382 of Cav3.2 significantly

reinforced the interactions with GIIIA-K11 and K16 versus those at GIIIB bound complex at the

vestibule of the pore. In contrast, regarding the residue at the extracellular loop of Cav3.2,

E283 established notably enhanced interactions with the N-terminal residues of GIIIB, namely

R1 and D2 via salt bridge/hydrogen bond interactions (Figure 5.10C,D). On the contrary, such

contacts were not presented at GIIIA when anchoring to Cav3.2.

In summary, the negatively charged residues at the vestibule of the pore of Cav3.2 primarily

strengthened the binding of GIIIA via forming more persistent salt-bridge and hydrogen bond

interactions with K11, K16 and Q14 of the toxin, compared with the homologous residues of

GIIIB. However, as E283 of Cav3.2 at the extracellular loop (DI) is located in a proximal position

to GIIIB-R1 and R2, it may result in a shallow binding of GIIIB versus GIIIA, the latter of which

binds more deeply into the pore of Cav3.2.

169

Figure 5.10. Significant pairwise interactions between GIIIB and Cav3.2 compared with Cav3.2/GIIIA complex. The coupled residues between Cav3.2 and GIIIB (A,C) as well as GIIIA (B,D). The interacting residues, E283, E378, D382 and D1504, on Cav3.2 are shown with orange, ochre, light purple and blue colors, respectively. The residues of GIIIA and GIIIB are highlighted by the same color scheme as the corresponding interactive residues. For the residues of the toxins contacting with more than 2 residues (including two residues) on Cav3.2, they are shown with magenta color.

170

o KIIIA versus GIIIA

The μ-conotoxin KIIIA also superficially binds at Cav3.2 above the outer vestibule of the pore,

as GIIIB does (Figure S2.4). However, distinct from GIIIB and PIIIA, this conopeptide formed

few salt-bridge interactions with the determinants identified via MMPSA, namely D1504 (at

DIII) and K1809 (at DIV), which could have enabled KIIIA to anchor more strongly to the ion

channel as shown in Figure 5.11A. The lack of such interactions by KIIIA probably results from

the unavailable positive charged residues at the N-terminus or the loop 1 of the toxin,

although the van der Waals interactions may contribute to the KIIIA binding at Cav3.2 to a

lesser extent, such as interaction pairs involving N3-K1809 (not shown) and S5-D1504. By

contrast, the accommodation of GIIIA at Cav3.2 facilitates the interactions by D1504 and

K1809 with Q18 and Q14 of the toxin, respectively (Figure 5.11B).

Figure 5.11. Significant pairwise interactions between KIIIA and Cav3.2 compared with Cav3.2/GIIIA complex. The coupled residues between Cav3.2 and KIIIA (A) as well as GIIIA (B). The interacting residues, D1504 and K1809, on Cav3.2 are shown in blue and tan colors, respectively. The residues of GIIIA and KIIIA are highlighted by the same color scheme as the corresponding interactive residues. For the residues of the toxins contacting with more than two residues (including two residues) on Cav3.2, they are shown with magenta color.

171

o GIIIC versus GIIIA

The differences between the ion channel contacts with GIIIC and GIIIA at Cav3.2 are significant

but with moderate changes in the number of contacts by the two aspartic acids at positions

382 and 975 with residues of the μ-conotoxins (Figure 5.12). K9 and K16 of both μ-conotoxins

formed persistent salt-bridge interactions with D975 and D382, respectively, at the vestibule

of Cav3.2. The reduced interactions between GIIIC-K11 and D975 on Cav3.2 counteracted the

increased contacts by D975 with R14, when compared with GIIIA bound complex. The

similarity in the binding domain between GIIIC and GIIIA probably accounts for the minor

differences in their interactions with Cav3.2 (Figure S2.5).

Figure 5.12. Significant pairwise interactions between GIIIC and Cav3.2 compared with Cav3.2/GIIIA complex. The coupled residues between Cav3.2 and GIIIC (A) as well as GIIIA (B). The interacting residues, D382 and D975, on Cav3.2 are shown in light purple and red colors, respectively. The residues of GIIIA and GIIIC are highlighted by the same color scheme as the corresponding interactive residues. For the residues of the toxins contacting with more than two residues (including two residues) on Cav3.2, they are shown with magenta color.

172

5.4 Discussion

5.4.1 Determinants Influencing the Flexibility of the Protein-peptide Complex

The C-α RMSD plots demonstrate the stability of the protein-peptide complexes during time.

Figure 5.5A shows that the stability of the μ-conotoxin-bound Cav3.2 is achieved around 30ns,

as do the individual toxins (Figure 5.5B). However, a distinct pattern in the RMSD plot for PIIIA

is noteworthy, with a consistently higher RMSD value during simulation time (~ 128ns),

compared with the other μ-conotoxins, namely GIIIA, GIIIB, KIIIA and GIIIC. In addition, a

fluctuation of the RMSD plot may be observed for KIIIA and GIIIB from the simulation starting

point to 30ns, possibly due to the short amino acid sequences of KIIIA (16 amino acid in length)

and an extra pyroglutamic acid at the N-terminus of GIIIB (Figure 5.1).

To explore the flexibility of Cav3.2 upon bound by GIIIA versus the other μ-conotoxins, RMSF

analysis was employed for toxin bound Cav3.2 complexes. The key residues affecting the

flexibility of the ion channel structure are shown in Table 5.1. The positively charged residue

R334 played an important role in the overall flexibility of Cav3.2 at the S5 loop at DI, upon

binding by KIIIA, GIIIB and GIIIC, compared with GIIIA. Furthermore, this residue is particularly

high in flexibility for GIIIB-bound Cav3.2 (Figure 5.6,S2.6), which may therefore influence the

flexibility of the whole S5 loop at DI. The differences in flexibility may be associated with the

lower binding affinity of GIIIB versus GIIIA at this domain. Such positive residue also markedly

enhanced the flexibility of the extracellular loop of Cav3.2 bound by KIIIA, associated with the

other charged residues such as K944 at DII, and non-polar residues, A332, A333 (DI), L943 (DII)

and R1781 (DIV). By contrast, the rigidity of Cav3.2 at the extracellular loop is similar upon

bound by GIIIC and GIIIA.

In addition, the rigidity of the DIII-IV linker was primarily enhanced due to the binding of these

μ-conotoxins versus GIIIA at Cav3.2. It has been suggested the DIII-IV linker serves as a critical

factor that differentiates the structure and function of T-type (LVA) calcium channels from

the HVA ones [99, 100], via mediating the inactivation of the ion channels. Furthermore, the

DIII-IV linker also affected the activation kinetics of human and rat Cav3.2 [101-103], probably

via altering the anatomy of the ion channel [104]. The extracellular loops at domains I, II and

III are in proximity to the von Willebrand A and Cache1 domains of α2 accessory subunit of

173

Cav1.1 [60]. The latter domain contains the binding site of pregabalin, which is used to treat

neuropathic pains and fibromyalgia [60, 105].

5.4.2 Key Residues on Cav3.2 Affecting the Binding Affinity of μ-Conotoxins versus GIIIA

The MMPBSA results show the increased binding affinity of KIIIA and PIIIA versus GIIIA at

Cav3.2, whereas GIIIB demonstrates a reduced binding affinity (Figure 5.7A). The binding

affinity of GIIIC is similar to GIIIA targeting Cav3.2. However, since there are few significant

differences between GIIIA and the other conotoxins in terms of the total binding affinity, the

binding free energy contributions from individual residues on Cav3.2 were also calculated to

elucidate any distinctions between the toxins at the individual amino acid level. The key

residues, E283 D382, D975, D1504 and K1809, affecting the variations of binding affinity

between the studied toxins and GIIIA upon anchoring to Cav3.2 are shown in Figure 5.7B,

which will be discussed in the following:

• D1504

D1504 plays an important role in differentiating the binding affinity of PIIIA, GIIIB and KIIIA

versus the GIIIA bound complex. The binding affinity was significantly lower at this aspartic

acid residue on Cav3.2 for all the three conotoxins mentioned above, when compared with

GIIIA. The SF residue, D1504, at DIII of Cav3.2, is responsible for regulating the inactivation of

the calcium channels [106]. In a previous experimental study, D1504, situated next to K1503,

was identified as one of the critical residues favoring the binding of antagonist, Z944, at T-

type calcium channels [7]. By contrast, D1504 on Cav3.2 does not alter the binding of GIIIC at

this position versus the GIIIA bound form, possibly due to lack of changes in flexibility of the

ion channel. In previous experimental studies, it has been demonstrated that the mutations

(EEEE) at SF of Nav1.2 (SF motif: DKEA) can adapt its feature of sodium selectivity to a channel

with calcium ion sensitivity [57, 58]. In addition, this modification also reduced the sensitivity

of Nav1.2 to TTX antagonism at the ion channel [107].

• D382, D975 and K1809

The aspartic acid residue at position 382, close to the SF of Cav3.2 at DI, contributes to the

changes in binding affinity of PIIIA, GIIIB and GIIIC versus GIIIA. This residue is conserved

among all the human voltage-gated calcium channels except for Cav2.3 [7]. Both GIIIB and

174

GIIIC show significantly lower affinity at D382 on Cav3.2, whereas this negatively charged

residue facilitates the binding of PIIIA at the ion channel, when compared with GIIIA bound

complex. However, the binding of KIIIA was not affected by D382 of Cav3.2, due to the

different binding conformation of KIIIA versus the other conotoxins at the ion channel. This

assumption is further supported by the lower binding affinity at K1809 on the P2 helix in DIV

of Cav3.2. Previous experimental studies suggested that the single mutants of rat skeletal

muscle (μ1) sodium channels, D762K and E765K, intervened with the electrostatic

interactions with GIIIB [108]. Thus, it is of interest to conduct mutation of K1809 to an acidic

residue which might switch the sensitivity of KIIIA to Cav3.2.

D975 exerts little effect on binding of most of the conotoxins, except for the binding of GIIIC

relative to GIIIA at Cav3.2. This aspartic acid residue substantially favors the binding of GIIIC

at Cav3.2 versus GIIIA, which is the only residue modifying the binding affinity of the toxin at

DII of Cav3.2. The reduced changes in rigidity of the ion channel upon anchoring to GIIIC versus

GIIIA compared with other μ-conotoxins as mentioned above, might account for the role of

D975 in differentiating the GIIIC from the other neuropeptides. Importantly, there is a striking

role of D975 in determining the structural and functional properties among Cav channels [7].

Additionally, it has been suggested that E974 plays a critical role in calcium-dependent

inactivation of HVA channels [106]. Thus, the unique differences in binding affinity of GIIIC

from GIIIA invite further study to investigate the mechanical function of Cav3.2 upon bound

by GIIIC, when compared with the other μ-conotoxins.

• E283

GIIIB demonstrates a greatly reduced binding affinity at E283 (S5 loop at the ECD) on Cav3.2

at DI, which is the only residue within the extracellular loop of Cav3.2 that modifies the binding

affinity of GIIIB versus GIIIA. This might be related to the observed highest increase of

flexibility at R334 at the extracellular loop of Cav3.2. The enhanced flexibility may support the

GIIIB anchoring to the ion channel at this position. In addition, the binding of PIIIA to Cav3.2

might be impaired by the lower binding affinity at E1460 (S5 loop on Cav3.2 at DIII).

In summary, the SF residue, D1504, at DIII substantially reduced the binding affinity of PIIIA,

GIIIB and KIIIA at Cav3.2, while such a residue at DI (E378) favors the binding of GIIIB versus

GIIIA bound complex. Similarly, residues in proximity to the SF at DI (D382) and DIV (K1809)

175

of Cav3.2 also interfered with the binding of GIIIB, KIIIA and GIIIC to the ion channel. However,

D382 and D975 (at DII) facilitate the binding of GIIIA and GIIIC, respectively, to Cav3.2. The

binding of PIIIA and GIIIB to Cav3.2 was undermined by the reduced binding affinity at the

residues on extracellular loops, namely, E1460 and E283 on the S5 loop at DI and DIII,

respectively. It appears that most of the residues affecting the binding affinity of the toxins

versus GIIIA are negatively charged, with certain residues, such as D382 and D1504 that

distinctly affect the binding of GIIIB and GIIIC versus the GIIIA bound complex. Thus, the

unique difference in binding affinity of KIIIA at K1809 of Cav3.2 among all the bindings of

toxins, is intriguing, and further investigation was performed to examine the intermolecular

interactions between the toxins and these determinant residues identified via MMPBSA

approach, compared with the GIIIA bound complex.

5.4.3 Important Pairwise Interactions of μ-Conotoxin Bound to Cav3.2 Complexes

The changes in number of contacts formed by the residues on Cav3.2 that were predicted to

affect the binding affinity of GIIIA versus the other conotoxins, as elucidated via MMPBSA

(discussed above), are discussed in this section. A fewer number of residues on Cav3.2 shows

substantial differences between the conotoxins compared to the MMPBSA results.

The difference highlights the need to assess the applicability of MMPBSA and the

implementation of such method to studies of binding between biomacromolecules. It has

been demonstrated that MMPBSA analysis gave a higher performance over many scoring and

docking methods, but is, however, inferior to more computationally intensive

theoretical average precision (AP) calculation in a previous study on factor Xa inhibitors by

Genheden and colleagues [109]. They concluded that the AP approach produces a

comparable mean absolute deviation but a better R2 score versus that of MMPBSA, and is

more efficient, when compared with the experimental result [109, 110]. Furthermore, the

g_mmpbsa approach applied in this work only takes account of the interaction energies

obtained from several independent simulations of complexes [109, 111]. However, the

traditional MMPBSA approach is based upon three types of simulation for the binding

complex, the ligand only and the receptor in apo form [112]. The different methodology of

g_mmpbsa versus the classical MMPBSA may in particular affect the accuracy of MMPBSA

176

approach for the absolute binding free energy [62, 109, 113, 114]. In addition, the binding

affinity of the ligand also depends on electric charges of the binding domain and the ligand

[63, 90, 109, 115]. Therefore, the highly charged μ-conotoxins with various charges may also

affect the precision of MMPBSA in this study.

Given the factors which may affect the MMPBSA binding free energy predictions, it is

worthwhile considering direct measures of inter-residue contacts which may also provide

predictions of important interactions responsible for the binding affinity between conotoxins.

The contacts which show significant differences between the conotoxins relative to GIIIA are

discussed in the following:

• Important residues on μ-conotoxins interacting with D1504 and E378

The SF residue, D1504, of Cav3.2 at DIII exerts considerable impact on the interactions with

residues of PIIIA, GIIIB and KIIIA, relative to GIIIA. In the PIIIA, GIIIB and KIIIA bound Cav3.2

ensembles, D1504 substantially enhanced contacts with PIIIA-R14 (homologous residue to

R13 of GIIIA and GIIIB and GIIIC, and K7 of KIIIA) and K17, GIIIB-R13 and KIIIA-S5 versus GIIIA

(Figure 5.9-5.11). By contrast, the binding of PIIIA-Q15, GIIIB-R14 and KIIIA-W8, respectively,

are substantially reduced at D1504, versus the strong interactions between the GIIIA-Q14 and

D1504 on Cav3.2. Interestingly, these three residues and GIIIA-Q14 are homologous among

the toxins (Figure 5.1). However, the dominant interactions between PIIIA-R14 and K17 offset

the reduced contacts of Q15-D1504 upon binding to Cav3.2, whereas such enhancement of

interactions cannot compensate for the dramatically decreased contacts at GIIIB and KIIIA

with D1504. Thus, these findings may suggest the critical role of D1504 in favoring the binding

of PIIIA to Cav3.2 versus GIIIA via interactions with R14 and K17 of PIIIA, while this residue

prevents GIIIB and KIIIA from interacting with the SF of Cav3.2 at DIII. Previous experimental

and computational studies have shown that GIIIA-R13 is critical for the selectivity to Nav1.4

via interactions with the EEDD ring of the sodium channel at DI and DIV [37, 116-119], and

the substitution of Q14 of GIIIA with aspartic acid substantially affects the sensitivity of

skeletal muscle Na+ channel (μ1) to the toxin [54]. Also, R14 of PIIIA facilitates the binding of

the neuropeptide at Nav1.4 via interacting with E180 at DI identified through the

computational methods, associated with the interactions of K17-E180 at DII [120]. In addition,

GIIIA[R13A] showed a slightly different secondary structure, compared with the native GIIIA

at the disulphide bond between C4 and C20 [121]. Based on all the findings in this study and

177

the previous results, it may suggest the important role of PIIIA-R14 and K17 in enhancing the

interactions by the toxin with D1504 of Cav3.2 versus the other homologous residues of GIIIA

(R13), GIIIB (R13) and KIIIA (K7). Thus, it would be intriguing to investigate the role of PIIIA-

R14/GIIIA-R13 in inhibitory effects on Cav3.2 in the future.

Another SF residue at DI of Cav3.2 is E378, which has fewer interactions with GIIIB-K11, but

higher contacts with GIIIA-K11 via forming salt bridge/hydrogen bond interactions (Figure

5.10). This may account for a shallow binding of GIIIB versus GIIIA, as shown in Figure S2.3. In

previous mutant cycle and electrophysiological studies, the mutation of K11 to alanine

significantly reduced the inhibition of GIIIA mutant at Nav1.4, compared with the native GIIIA

showing strong interactions with the negatively charged residues on Nav1.4 [118]. Thus, the

present finding may suggest other residues, rather than K11 of the toxins, are critical for the

differential selectivity by GIIIB versus GIIIA for Cav3.2.

• Important residues on μ-conotoxins interacting with D382, E975 and K1809

The binding of PIIIA and GIIIB was significantly lower at D382 of Cav3.2 versus the GIIIA,

whereas for GIIIC this was only moderately lower. GIIIA has substantially higher contacts with

D382 via K11 and K16, which are the key factors that accommodate the PIIIA at Nav1.4

determined through computational methods [117]. The enhanced interactions established by

GIIIA-K11 and K16 with D382 can help stabilize the binding of GIIIA at the outer vestibule of

Cav3.2 at DI for forming favourable contacts with SF residue, E378. By contrast, the GIIIB has

few contacts with the residues on this region due to lack of interactions with E378 and D382.

In addition, PIIIA also has lower interactions at D382 with the homologous residues at

positions 12 and 17. In summary, D382 is the only residue that has lower contacts with the

first positively charged residues on loops 2 and 3 of PIIIA, GIIIB and GIIIC, versus similar

residues of GIIIA in which salt-bridge/hydrogen bond interactions are formed with D382 at DI

of Cav3.2. This finding may putatively suggest a different binding domain of GIIIA versus the

other μ-conotoxins at Cav3.2 with a preference to bind at the outer vestibule of the pore at

DI of the ion channel.

PIIIA and GIIIB are mainly different from GIIIA in their interaction with the residues at the

outer vestibule of Cav3.2 at domain I or III. However, GIIIC also demonstrates moderate

changes in contacts with residue D975 at DII of Cav3.2 with statistical significance, when

178

compared with GIIIA bound complex (Figure 5.12,S2.5). D975 locates next to the SF residue,

E974, at DII. A previous computational study has suggested the important role of D180,

proximal to the DEKA ring of Nav1.4 at DII, in the binding of GIIIA via forming salt-bridge

interactions with K17 of the toxin [120].

Compared with GIIIA, the toxin demonstrating the most unique differences in interactions

with residues at the outer pore at DIV is KIIIA. This is related to the different binding

orientation of KIIIA at Cav3.2 versus the other conotoxins (Figure S2.4), due to loss of contacts

at W8 and H12. Previous experimental and modelling studies have underlined the critical role

of W8 on KIIIA in inhibitory effects on human and rat Nav1.2 as well as human Nav1.7 [55,

122]. In particular, KIIIA-W8 formed persistent - interactions with Y362 on P2 helix (DI) of

Nav1.7, which constitutes the key residues critical for the sensitivity of Nav1.7 to KIIIA with

D919 and D923 on P2 helix at DII [55, 123]. Overall, the binding of KIIIA at Cav3.2 was lower

than GIIIA, due to the loss of contacts at W8 and H12 with D1504. Further study is required

to test the binding kinetics of KIIIA at Cav3.2.

• Important residues on μ-conotoxins interacting with E283 and E1460

Both PIIIA and GIIIB show higher numbers of contacts with residues on the extracellular loops

of Cav3.2 at DIII and DI, respectively, when compared with GIIIA bound complex (Figure

5.9,5.10). For PIIIA, the proximity between K9 and E1460 allowed persistent electrostatic

interactions, assisted by van der Waals interactions of F7-E1460 on Cav3.2 at DIII. R1 and D2

of GIIIB likewise formed salt-bridge contacts with another glutamic acid at position 283 (DI)

on Cav3.2. Such interactions may explain the dramatically reduced interactions of PIIIA and

GIIIB with D382 and/or D1504 on Cav3.2. PIIIA preferentially binds to DIII of Cav3.2 including

residue E1460 and D1504, whereas DI of Cav3.2 favors the anchoring of GIIIB at D283, when

compared with GIIIA. This pattern of binding may also be applied to GIIIC, which shows

enhanced interactions with D975 at DII but reduces contacts at D382 of Cav3.2. A previous

molecular modelling study highlighted the fundamental effects of PIIIA-K9 on selectivity to

Nav1.4 via interacting with SF residue, D177, at DI [124]. This may suggest a distinct binding

conformation of PIIIA at Nav1.4 versus Cav3.2.

In summary, the pairwise interactions of K9-E1460 and R14/K17-D1504 enhance the contacts

by PIIIA at DIII of Cav3.2, whilst R1/D2-E283 of GIIIB bound complex accommodates the toxin

179

to the calcium channel. In GIIIC bound form, the coupled residues of R14/K9-D975 reinforce

the binding of GIIIC at DII of Cav3.2 versus GIIIA. The GIIIA-K16 formed persistent electrostatic

interactions with D382 (DI) on Cav3.2. However, KIIIA marginally established stronger

interactions with residues on Cav3.2 over the pairwise interactions of GIIIA bound complex.

Future research may facilitate evaluation of the determinants identified on both toxins and

the calcium channel via mutagenesis and electrophysiological experimental techniques.

5.4.4 Prediction of Binding Conformations of μ-Conotoxins at Cav3.2

All five conotoxins, GIIIA, GIIIB, GIIIC, PIIIA and KIIIA, target the Cav3.2 via pore blocking

mechanism. Nonetheless, although the toxins binding to Cav3.2 partially overlap with each

other, the distinct differences in Cav3.2 interactions by the toxins versus GIIIA result in various

orientations of the toxins upon binding to the calcium channel. GIIIA anchors to the outer

vestibule at DIII of Cav3.2, while the N-terminus of GIIIB tilts to the extracellular loop of DI

(Figure S2.11). On the other hand, GIIIC shares a similar binding domain as GIIIA. However,

the toxin preferentially protrudes into the outer EEDD ring of Cav3.2 at DII apart from GIIIA,

which accommodates close to the outer vestibule at DI of the calcium channel.

When compared with sodium channels, the binding pocket of the μ-conotoxins at Cav3.2 is

occupied by acidic residues, which resembles that of sodium channels bound by μ-conotoxins

[108, 120, 124]. In addition, the S5 loop at DIII is also important for the binding of ω-

conotoxins at N-type calcium channel [125-127], such as GIVA and CVID. However, the binding

conformation of the μ-conotoxins at Cav3.2, such as PIIIA, may partially vary from those toxins

anchoring to the TTX-sensitive sodium channels and N-type calcium channels.

5.5 Conclusion

This chapter describes predictions of the binding orientations, dynamics, and important

pairwise residue contacts between a number of -conotoxins and Cav3.2. Differences in

binding modes and specific contacts are proposed to account for differences in their inhibitory

activities against the channel. Comparisons were made relative to GIIIA, the prototypical -

conotoxin experimentally demonstrated to inhibit the Cav channel. The increased flexibility of

180

R334 on Cav3.2 at DI particularly contributes to the greater flexibility of the S5 loop in GIIIB

versus the GIIIA bound complex. Consequently, the enhanced electrostatic interactions

between GIIIB-R1 and the E283 at the extracellular loop make the N-terminus of the GIIIB tilt

into the loop, which results in a superficial binding conformation by the toxin versus GIIIA.

The PIIIA enhanced its interactions with E1460 and D1504 at positions K9, R14 and R17 via

forming salt-bridge/hydrogen bond interactions at DIII. The GIIIC shares the most similar

binding conformation versus GIIIA, except for loop 3 that projects to the outer vestibule of

Cav3.2 at DII. The Cav3.2 may be less sensitive to KIIIA due to lack of enhanced electrostatic

interactions compared with GIIIA. For GIIIA, the key residues are K11, Q14 and K16 that form

salt-bridge and hydrogen bonds for accommodating the GIIIA in proximity to D382 at DI and

D1504 at DIII of Cav3.2. These findings can facilitate further study on mutagenesis of the

coupled residues to improve the selectivity and binding potency of the toxins to Cav3.2.

181

5.6 References

Darszon A, Nishigaki T, Beltran C, Treviño CL. Calcium channels in the development,

Catterall WA. Calcium channels. In: Squire LR, editor. Encyclopedia of Neuroscience.

Zhao Y, Huang G, Wu Q, Wu K, Li R, Lei J, et al. Cryo-EM structures of apo and

Buraei Z, Yang J. Structure and function of the β subunit of voltage-gated Ca²⁺ channels.

Catterall WA. Voltage-gated calcium channels. Cold Spring Harbor Perspectives in

Nowycky MC, Fox AP, Tsien RW. Three types of neuronal calcium channel with

Tsien RW, Lipscombe D, Madison DV, Bley KR, Fox AP. Multiple types of neuronal

Huguenard JR. Low-threshold calcium currents in central nervous system neurons. 1. Veselovskii NS, Fedulova SA. Effects of substituting barium for calcium ions during research into inward currents in mammalian neurons. Neurophysiology. 1986;18(3):227-31. 2. Bean BP. Two kinds of calcium channels in canine atrial cells. Differences in kinetics, selectivity, and pharmacology. The Journal of General Physiology. 1985;86(1):1-30. 3. maturation, and function of spermatozoa. Physiological Reviews. 2011;91(4):1305-55. 4. Ellis SB, Williams ME, Ways NR, Brenner R, Sharp AH, Leung AT, et al. Sequence and expression of mRNAs encoding the α1 and α2 subunits of a DHP-sensitive calcium channel. Science. 1988;241(4873):1661-4. 5. Mikami A, Imoto K, Tanabe T, Niidome T, Mori Y, Takeshima H, et al. Primary structure and functional expression of the cardiac dihydropyridine-sensitive calcium channel. Nature. 1989;340(6230):230-3. 6. Oxford: Academic Press; 2009. p. 543-50. 7. antagonist-bound human Cav3.1. Nature. 2019;576(7787):492-7. 8. Flucher BE. Specific contributions of the four voltage-sensing domains in L-type calcium channels to gating and modulation. The Journal of general physiology. 2016;148(2):91-5. 9. Li J, Stevens L, Klugbauer N, Wray D. Roles of molecular regions in determining differences between voltage dependence of activation of CaV3.1 and CaV1.2 calcium channels. Journal of Biological Chemistry. 2004;279(26):26858-67. 10. Biochim Biophys Acta. 2013;1828(7):1530-40. 11. Sanchez-Sandoval AL, Herrera Carrillo Z, Díaz Velásquez CE, Delgadillo DM, Rivera HM, Gomora JC. Contribution of S4 segments and S4-S5 linkers to the low-voltage activation properties of T-type CaV3.3 channels. PLOS ONE. 2018;13(2):e0193490. 12. Canti C, Nieto-Rostro M, Foucault I, Heblich F, Wratten J, Richards MW, et al. The metal-ion-dependent adhesion site in the Von Willebrand factor-A domain of α2δ subunits is key to trafficking voltage-gated Ca2+ channels. Proceedings of the National Academy of Sciences. 2005;102(32):11230-5. 13. Biology. 2011;3(8):a003947-a. 14. Tyson JR, Snutch TP. Molecular nature of voltage-gated calcium channels: structure and species comparison. Wiley Interdisciplinary Reviews: Membrane Transport and Signaling. 2013;2(5):181-206. 15. McGivern JG. Targeting N-type and T-type calcium channels for the treatment of pain. Drug Discovery Today. 2006;11(5):245-53. 16. different calcium agonist sensitivity. Nature. 1985;316(6027):440-3. 17. calcium channels and their selective modulation. Trends Neurosci. 1988;11(10):431-8. 18. Annu Rev Physiol. 1996;58:329-48.

182

Iftinca MC, Zamponi GW. Regulation of neuronal T-type calcium channels. Trends in

Cain SM, Snutch TP. T-type calcium channels in burst-firing, network synchrony, and

Li Y, Liu S, Lu F, Zhang T, Chen H, Wu S, et al. A role of functional T-type Ca2+ channel

Song I, Kim D, Choi S, Sun M, Kim Y, Shin H-S. Role of the α1G T-Type calcium channel in mutant mice. The Journal of Neuroscience.

Vassort G, Talavera K, Alvarez JL. Role of T-type Ca2+ channels in the heart. Cell Calcium.

Choi S, Na HS, Kim J, Lee J, Lee S, Kim D, et al. Attenuated pain responses in mice

Gadotti VM, Kreitinger JM, Wageling NB, Budke D, Diaz P, Zamponi GW. Cav3.2 T-type

Lu A, Yang L, Xu S, Wang C. Various conotoxin diversifications revealed by a venomic

Lambert RC, McKenna F, Maulet Y, Talley EM, Bayliss DA, Cribbs LL, et al. Low-voltage- 19. activated Ca2+ currents are generated by members of the CavT subunit family (alpha1G/H) in rat primary sensory neurons. The Journal of Neuroscience. 1998;18(21):8605-13. 20. Pharmacological Sciences. 2009;30(1):32-40. 21. epilepsy. Biochimica et Biophysica Acta (BBA) - Biomembranes. 2013;1828(7):1572-8. Ohkubo T, Yamazaki J. T-type voltage-activated calcium channel Cav3.1, but not 22. Cav3.2, is involved in the inhibition of proliferation and apoptosis in MCF-7 human breast cancer cells. Int J Oncol. 2012;41(1):267-75. Bertolesi GE, Shi C, Elbaum L, Jollimore C, Rozenberg G, Barnes S, et al. The Cav2+ 23. channel antagonists mibefradil and pimozide inhibit cell growth via different cytotoxic mechanisms. Molecular Pharmacology. 2002;62(2):210-9. 24. in hepatocellular carcinoma cell proliferation. Oncol Rep. 2009;22(5):1229-35. 25. Jagodic MM, Pathirathna S, Nelson MT, Mancuso S, Joksovic PM, Rosenberg ER, et al. Cell-specific alterations of T-type calcium current in painful diabetic neuropathy enhance excitability of sensory neurons. The Journal of Neuroscience. 2007;27(12):3305-16. 26. Heron SE, Khosravani H, Varela D, Bladen C, Williams TC, Newman MR, et al. Extended spectrum of idiopathic generalized epilepsies associated with CACNA1H functional variants. Annals of Neurology. 2007;62(6):560-8. 27. in spontaneous absence seizures 2004;24(22):5249-57. 28. 2006;40(2):205-20. 29. Hansen PBL. Functional importance of T-type voltage-gated calcium channels in the cardiovascular and renal system: news from the world of knockout mice. American Journal of Physiology-Regulatory, Integrative and Comparative Physiology. 2015;308(4):R227-R37. 30. Santoni G, Santoni M, Nabissi M. Functional role of T-type calcium channels in tumour growth and progression: prospective in cancer therapy. British journal of pharmacology. 2012;166(4):1244-6. 31. Souza IA, Gandini MA, Zhang F-X, Mitchell WG, Matsumoto J, Lerner J, et al. Pathogenic Cav3.2 channel mutation in a child with primary generalized epilepsy. Molecular Brain. 2019;12(1):86. 32. lacking Ca(V)3.2 T-type channels. Genes, Brain, and Behavior. 2007;6(5):425-31. Bourinet E, Alloui A, Monteil A, Barrère C, Couette B, Poirot O, et al. Silencing of the 33. Cav3.2 T-type calcium channel gene in sensory neurons demonstrates its major role in nociception. The EMBO Journal. 2005;24(2):315-24. 34. calcium channels control acute itch in mice. Molecular Brain. 2020;13(1):119. 35. study of Conus flavidus. Molecular & Cellular Proteomics. 2014;13(1):105-18. 36. Wang D, Himaya SWA, Giacomotto J, Hasan MM, Cardoso FC, Ragnarsson L, et al. Characterisation of δ-conotoxin TxVIA as a mammalian t-type calcium channel modulator. Marine Drugs. 2020;18(7):343.

183

channel sensitivity Cummins TR, Aglieco F, Dib-Hajj SD. Critical molecular determinants of voltage-gated to μ-conotoxins GIIIA/B. Molecular Pharmacology.

Rucktooa P, Smit AB, Sixma TK. Insight in nAChR subtype selectivity from AChBP crystal

Dave K, Lahiry A. Conotoxins: Review and docking studies to determine potentials of in medicinal chemistry.

Chuang RSI, Jaffe H, Cribbs L, Perez-Reyes E, Swartz KJ. Inhibition of T-type voltage-

Bladen C, Hamid J, Souza IA, Zamponi GW. Block of T-type calcium channels by

37. sodium 2002;61(5):1192-201. Gilchrist J, Olivera BM, Bosmans F. Animal toxins influence voltage-gated sodium 38. channel function. In: Ruben PC, editor. Voltage Gated Sodium Channels. Berlin, Heidelberg: Springer Berlin Heidelberg; 2014. p. 203-29. 39. structures. Biochemical Pharmacology. 2009;78(7):777-87. 40. Franco A, Kompella SN, Akondi KB, Melaun C, Daly NL, Luetje CW, et al. RegIIA: an α4/7-conotoxin from the venom of Conus regius that potently blocks α3β4 nAChRs. Biochemical Pharmacology. 2012;83(3):419-26. 41. Shon K-J, Stocker M, Terlau H, Stühmer W, Jacobsen R, Walker C, et al. κ-Conotoxin Pviia is a peptide inhibiting the Shaker K+ channel. Journal of Biological Chemistry. 1998;273(1):33-8. Cordeiro S, Finol-Urdaneta RK, Köpfer D, Markushina A, Song J, French RJ, et al. 42. Conotoxin κM-RIIIJ, a tool targeting asymmetric heteromeric Kv1 channels. Proceedings of the National Academy of Sciences. 2019;116(3):1059-64. 43. conotoxin as an anticancer drug molecule. Current topics 2012;12:845-51. Staats PS, Yearwood T, Charapata SG, Presley RW, Wallace MS, Byas-Smith M, et al. 44. Intrathecal ziconotide in the treatment of refractory pain in patients with cancer or AIDS: a Randomized Controlled Trial. JAMA. 2004;291(1):63-70. 45. Wallace MS, Charapata SG, Fisher R, Byas-Smith M, Staats PS, Mayo M, et al. Intrathecal ziconotide in the treatment of chronic nonmalignant pain: a randomized, double- blind, placebo-controlled clinical trial. Neuromodulation: Technology at the Neural Interface. 2006;9(2):75-86. de la Calle Gil AB, Peña Vergara I, Cormane Bornacelly MA, Pajuelo Gallego A. 46. Intrathecal ziconotide and morphine for pain relief: a case series of eight patients with refractory cancer pain, including five cases of neuropathic pain. Neurology Therapy. 2015;4(2):159-68. Berecki G, Daly NL, Huang YH, Vink S, Craik DJ, Alewood PF, et al. Effects of arginine 10 47. to lysine substitution on ω-conotoxin CVIE and CVIF block of Cav2.2 channels. British Journal of Pharmacology. 2014;171(13):3313-27. Feng Z-P, Hamid J, Doering C, Bosey GM, Snutch TP, Zamponi GW. Residue Gly1326 of 48. the N-type Calcium Channel α1B Subunit Controls Reversibility of ω-Conotoxin GVIA and MVIIA Block. Journal of Biological Chemistry. 2001;276(19):15728-35. 49. gated calcium channels by a new scorpion toxin. Nature Neuroscience. 1998;1:668. 50. Lawrence N, Wu B, Ligutti J, Cheneval O, Agwa AJ, Benfield AH, et al. Peptide- membrane interactions affect the inhibitory potency and selectivity of spider toxins ProTx-II and GpTx-1. ACS Chemical Biology. 2019;14(1):118-30. 51. protoxins I and II. Molecular Brain. 2014;7:36-. Shon K-J, Olivera BM, Watkins M, Jacobsen RB, Gray WR, Floresca CZ, et al. μ- 52. Conotoxin PIIIA, a new peptide for discriminating among tetrodotoxin-sensitive na channel subtypes. The Journal of Neuroscience. 1998;18(12):4473-81.

184

Heinemann SH, Terlau H, Stühmer W, Imoto K, Numa S. Calcium channel the sodium channel by single mutations. Nature.

Kumari R, Kumar R, Lynn A. g_mmpbsa—a GROMACS tool for high-throughput MM-

Chen WL, Xie SQ, Zeng FF, Li BM. A new process knowledge representation approach

Nielsen KJ, Watson M, Adams DJ, Hammarström AK, Gage PW, Hill JM, et al. Solution 53. structure of μ-conotoxin PIIIA, a Preferential inhibitor of persistent tetrodotoxin-sensitive sodium channels. Journal of Biological Chemistry. 2002;277(30):27247-55. Dudley SC, Jr., Chang N, Hall J, Lipkind G, Fozzard HA, French RJ. μ-Conotoxin GIIIA 54. interactions with the voltage-gated Na+ channel predict a clockwise arrangement of the domains. Journal of General Physiology. 2000;116(5):679-90. 55. Kimball IH, Nguyen PT, Olivera BM, Sack JT, Yarov-Yarovoy V. Molecular determinants of μ-conotoxin KIIIA interaction with the voltage-gated sodium channel Nav1.7. bioRxiv. 2019:654889. 56. Hill JM, Alewood PF, Craik DJ. Three-dimensional solution structure of μ-conotoxin GIIIB, a specific blocker of skeletal muscle sodium channels. Biochemistry. 1996;35(27):8824- 35. 57. characteristics conferred on 1992;356(6368):441-3. 58. Stephens RF, Guan W, Zhorov BS, Spafford JD. Selectivity filters and cysteine-rich extracellular loops in voltage-gated sodium, calcium, and NALCN channels. Frontiers in Physiology. 2015;6(153). 59. Ohkubo T, Yamazaki J, Kitamura K. Tarantula toxin ProTx-I differentiates between human t-type voltage-gated Ca2+ channels Cav3.1 and Cav3.2. Journal of Pharmacological Sciences. 2010;112(4):452-8. 60. Wu J, Yan Z, Li Z, Qian X, Lu S, Dong M, et al. Structure of the voltage-gated calcium channel Cav1.1 at 3.6 Å resolution. Nature. 2016;537:191. 61. Miller BR, McGee TD, Swails JM, Homeyer N, Gohlke H, Roitberg AE. MMPBSA.py: an efficient program for end-state free energy calculations. Journal of Chemical Theory and Computation. 2012;8(9):3314-21. 62. PBSA calculations. Journal of Chemical Information and Modeling. 2014;54(7):1951-62. 63. Wang C, Nguyen PH, Pham K, Huynh D, Le T-BN, Wang H, et al. Calculating protein- ligand binding affinities with MMPBSA: method and error analysis. Journal of computational chemistry. 2016;37(27):2436-46. Hess B, Kutzner C, van der Spoel D, Lindahl E. GROMACS 4: algorithms for highly 64. efficient, load-balanced, and scalable molecular simulation. Journal of Chemical Theory and Computation. 2008;4(3):435-47. Bjelkmar P, Larsson P, Cuendet MA, Hess B, Lindahl E. Implementation of the 65. CHARMM force field in GROMACS: analysis of protein stability effects from correction maps, virtual interaction sites, and water models. Journal of Chemical Theory and Computation. 2010;6(2):459-66. 66. using parameter flow chart. Computers in Industry. 2011;62(1):9-22. 67. Huang D, Shi S, Liang C, Zhang X, Du X, An H, et al. Delineating an extracellular redox- sensitive module in T-type Ca2+ channels. Journal of Biological Chemistry. 2020;295(18):6177- 86. 68. Arnold K, Bordoli L, Kopp J, Schwede T. The SWISS-MODEL workspace: a web-based environment for protein structure homology modelling. Bioinformatics. 2006;22(2):195-201. 69. Hatherley R, Brown DK, Glenister M, Tastan Bishop Ö. PRIMO: an interactive homology modeling pipeline. PLOS ONE. 2016;11(11):e0166698.

185

features and Bienert S, Waterhouse A, de Beer TAP, Tauriello G, Studer G, Bordoli L, et al. The functionality. Nucleic Acids Research.

Hess B, Bekker H, Berendsen HJC, Fraaije JGEM. LINCS: A linear constraint solver for

Bussi G, Donadio D, Parrinello M. Canonical sampling through velocity rescaling. The

Colovos C, Yeates TO. Verification of protein structures: patterns of nonbonded atomic

Ferreira LG, Santos RNd, Oliva G, Andricopulo AD. Molecular docking and structure-

Shamsara J. CrossDocker: a tool for performing cross-docking using Autodock Vina.

Rohl CA, Strauss CEM, Misura KMS, Baker D. Protein Structure Prediction Using

Trellet M, Melquiond ASJ, Bonvin AMJJ. A unified conformational selection and 70. Webb B, Sali A. Comparative protein structure modeling using MODELLER. Current Protocols in Bioinformatics. 2016;54:5.6.1-5.6.37. 71. SWISS-MODEL Repository-new 2017;45(D1):D313-D9. 72. Waterhouse A, Bertoni M, Bienert S, Studer G, Tauriello G, Gumienny R, et al. SWISS- MODEL: homology modelling of protein structures and complexes. Nucleic Acids Research. 2018;46(W1):W296-W303. 73. Stringer RN, Jurkovicova-Tarabova B, Huang S, Haji-Ghassemi O, Idoux R, Liashenko A, et al. A rare CACNA1H variant associated with amyotrophic lateral sclerosis causes complete loss of Cav3.2 T-type channel activity. Molecular Brain. 2020;13(1):33. 74. Mark P, Nilsson L. Structure and dynamics of the TIP3P, SPC, and SPC/E water models at 298 K. The Journal of Physical Chemistry A. 2001;105(43):9954-60. 75. Berendsen HJC, van der Spoel D, van Drunen R. GROMACS: a message-passing parallel molecular dynamics implementation. Computer Physics Communications. 1995;91(1):43-56. 76. Sapay N, Tieleman DP. Combination of the CHARMM27 force field with united-atom lipid force fields. Journal of Computational Chemistry. 2011;32(7):1400-10. 77. molecular simulations. Journal of Computational Chemistry. 1997;18(12):1463-72. 78. Wen J, Hung A. Effects of C-terminal carboxylation on α-conotoxin LsIA interactions with human α7 nicotinic acetylcholine receptor: molecular simulation studies. Marine Drugs. 2019;17(4):206. 79. Journal of Chemical Physics. 2007;126(1):014101. 80. interactions. Protein Science. 1993;2(9):1511-9. 81. Trott O, Olson AJ. AutoDock Vina: improving the speed and accuracy of docking with a new scoring function, efficient optimization, and multithreading. Journal of Computational Chemistry. 2010;31(2):455-61. 82. Macindoe G, Mavridis L, Venkatraman V, Devignes M-D, Ritchie DW. HexServer: an FFT-based protein docking server powered by graphics processors. Nucleic Acids Research. 2010;38(Web Server issue):W445-W9. 83. based drug design strategies. Molecules. 2015;20(7):13384-421. 84. SpringerPlus. 2016;5(1):344. 85. Rosetta. Methods in Enzymology. 383: Academic Press; 2004. p. 66-93. 86. Schaap MG, Leij FJ, van Genuchten MT. Rosetta: a computer program for estimating soil hydraulic parameters with hierarchical pedotransfer functions. Journal of Hydrology. 2001;251(3):163-76. Lyskov S, Chou F-C, Conchúir SÓ, Der BS, Drew K, Kuroda D, et al. Serverification of 87. molecular modeling applications: the Rosetta online server that includes everyone (ROSIE). PLOS ONE. 2013;8(5):e63906. 88. induced fit approach to protein-peptide docking. PLOS ONE. 2013;8(3):e58769.

186

Fogolari F, Brigo A, Molinari H. Protocol for MM/PBSA Molecular Dynamics

Simms Brett A, Zamponi Gerald W. Neuronal voltage-gated calcium channels:

Giardina SF, Werner DS, Pingle M, Feinberg PB, Foreman KW, Bergstrom DE, et al. 89. Novel, self-assembling dimeric inhibitors of human β tryptase. Journal of Medicinal Chemistry. 2020;63(6):3004-27. 90. Wang C, Greene DA, Xiao L, Qi R, Luo R. Recent developments and applications of the MMPBSA method. Frontiers in Molecular Biosciences. 2018;4:87-. 91. Simulations of Proteins. Biophysical Journal. 2003;85(1):159-66. Baker NA, Sept D, Joseph S, Holst MJ, McCammon JA. Electrostatics of nanosystems: 92. Application to microtubules and the ribosome. Proceedings of the National Academy of Sciences. 2001;98(18):10037. Kurt BZ, Gazioglu I, Dag A, Salmas RE, Kayık G, Durdagi S, et al. Synthesis, 93. anticholinesterase activity and molecular modeling study of novel carbamate-substituted thymol/carvacrol derivatives. Bioorganic & Medicinal Chemistry. 2017;25(4):1352-63. 94. Zhao F-L, Yang G-H, Xiang S, Gao D-D, Zeng C. In silico analysis of the effect of mutation on epidermal growth factor receptor in non-small-cell lung carcinoma: from mutational analysis to drug designing. Journal of Biomolecular Structure and Dynamics. 2017;35(2):427- 34. Samsudin F, Parker JL, Sansom MSP, Newstead S, Fowler PW. Accurate prediction of 95. ligand affinities for a proton-dependent oligopeptide transporter. Cell Chemical Biology. 2016;23(2):299-309. Schaller D, Gündüz MG, Zhang FX, Zamponi GW, Wolber G. Binding mechanism 96. investigations guiding the synthesis of novel condensed 1,4-dihydropyridine derivatives with L-/T-type calcium channel blocking activity. European Journal of Medicinal Chemistry. 2018;155:1-12. Tang L, Gamal El-Din TM, Payandeh J, Martinez GQ, Heard TM, Scheuer T, et al. 97. Structural basis for Ca2+ selectivity of a voltage-gated calcium channel. Nature. 2014;505(7481):56-61. Swetha RG, Ramaiah S, Anbarasu A. R521C and R521H mutations in FUS result in weak 98. binding with Karyopherinβ2 leading to Amyotrophic lateral sclerosis: a molecular docking and dynamics study. Journal of Biomolecular Structure and Dynamics. 2017;35(10):2169-85. 99. structure, function, and dysfunction. Neuron. 2014;82(1):24-45. 100. Hamid J, Peloquin JB, Monteil A, Zamponi GW. Determinants of the differential gating properties of Cav3.1 and Cav3.3 T-type channels: a role of domain IV? Neuroscience. 2006;143(3):717-28. 101. Lee J-H, Kim E-G, Park B-G, Kim K-H, Cha S-K, Kong ID, et al. Identification of T-Type α1H Ca2+ Channels (Cav3.2) in major pelvic ganglion neurons. The Journal of Neurophysiology. 2002;87(6):2844-50. 102. Jagannathan S, Punt EL, Gu Y, Arnoult C, Sakkas D, Barratt CLR, et al. Identification and localization of T-type voltage-operated calcium channel subunits in human male germ cells: expression of multiple isoforms. Journal of Biological Chemistry. 2002;277(10):8449-56. 103. Perez-Reyes E. Molecular physiology of low-voltage-activated T-type calcium channels. Physiological Reviews. 2003;83(1):117-61. 104. Lin W-H, Wright DE, Muraro NI, Baines RA. Alternative splicing in the voltage-gated sodium channel DmNav regulates activation, inactivation, and persistent current. The Journal of Neurophysiology. 2009;102(3):1994-2006.

187

in the selectivity Schlief T, Schönherr R, Imoto K, Heinemann SH. Pore properties of rat brain II sodium filter domain. European Biophysics Journal.

Singh N, Warshel A. Absolute binding free energy calculations: on the accuracy of

Field MJ, Cox PJ, Stott E, Melrose H, Offord J, Su T-Z, et al. Identification of the α2-δ-1 105. subunit of voltage-dependent calcium channels as a molecular target for pain mediating the analgesic actions of pregabalin. Proceedings of the National Academy of Sciences. 2006;103(46):17537-42. 106. Abderemane-Ali F, Findeisen F, Rossen ND, Minor DL. A selectivity filter gate controls voltage-gated calcium channel calcium-dependent inactivation. Neuron. 2019;101(6):1134- 49.e3. 107. channels mutated 1996;25(2):75-91. Li RA, Ennis IL, Vélez P, Tomaselli GF, Marbán E. Novel structural determinants of mu- 108. conotoxin (GIIIB) block in rat skeletal muscle (mu1) Na+ channels. Journal of Biological Chemistry. 2000;275(36):27551-8. 109. Genheden S, Ryde U. The MM/PBSA and MM/GBSA methods to estimate ligand- binding affinities. Expert Opinion on Drug Discovery. 2015;10(5):449-61. 110. Genheden S, Nilsson I, Ryde U. Binding affinities of factor Xa inhibitors estimated by thermodynamic integration and MM/GBSA. Journal of Chemical Information and Modeling. 2011;51(4):947-58. 111. Kollman PA, Massova I, Reyes C, Kuhn B, Huo S, Chong L, et al. Calculating structures and free energies of complex molecules: combining molecular mechanics and continuum models. Accounts of Chemical Research. 2000;33(12):889-97. 112. Swanson JMJ, Henchman RH, McCammon JA. Revisiting free energy calculations: a theoretical connection to MM/PBSA and direct calculation of the association free energy. Biophysical Journal. 2004;86(1 Pt 1):67-74. 113. Huang K, Luo S, Cong Y, Zhong S, Zhang JH, Duan L. An accurate free energy estimator: based on MM/PBSA combined with interaction entropy for protein-ligand binding affinity. Nanoscale. 2020. 114. computational scoring of protein-ligand interactions. Proteins. 2010;78(7):1705-23. 115. Sun H, Duan L, Chen F, Liu H, Wang Z, Pan P, et al. Assessing the performance of MM/PBSA and MM/GBSA methods. 7. entropy effects on the performance of end-point binding free energy calculation approaches. Physical Chemistry Chemical Physics. 2018;20(21):14450-60. 116. Chang NS, French RJ, Lipkind GM, Fozzard HA, Dudley S. Predominant interactions between μ-conotoxin Arg-13 and the skeletal muscle Na+ channel localized by mutant cycle analysis. Biochemistry. 1998;37(13):4407-19. 117. Mahdavi S, Kuyucak S. Molecular dynamics study of binding of µ-conotoxin GIIIA to the voltage-gated sodium channel Nav1.4. PLOS ONE. 2014;9(8):e105300. 118. Xue T, Ennis IL, Sato K, French RJ, Li RA. Novel interactions identified between μ- conotoxin and the Na+ Channel domain I P-loop: implications for toxin-pore binding geometry. Biophysical Journal. 2003;85(4):2299-310. 119. Hui K, Lipkind G, Fozzard HA, French RJ. Electrostatic and steric contributions to block of the skeletal muscle sodium channel by μ-conotoxin. The Journal of General Physiology. 2002;119(1):45-54. 120. Chen R, Robinson A, Chung S-H. Mechanism of μ-conotoxin PIIIA binding to the voltage-gated Na+ channel NaV1.4. PLOS ONE. 2014;9(3):e93267.

188

Ellinor PT, Zhang J-F, Horne WA, Tsien RW. Structural determinants of the blockade of

function of voltage-gated calcium 121. Wakamatsu K, Kohda D, Hatanaka H, Lancelin JM, Ishida Y, Oya M, et al. Structure- activity relationships of mu-conotoxin GIIIA: structure determination of active and inactive sodium channel blocker peptides by NMR and simulated annealing calculations. Biochemistry. 1992;31(50):12577-84. 122. Van Der Haegen A, Peigneur S, Tytgat J. Importance of position 8 in μ-conotoxin KIIIA for voltage-gated sodium channel selectivity. The FEBS Journal. 2011;278(18):3408-18. 123. Kimball IH, Nguyen PT, Sack JT, Yarov-Yarovoy V. Assessing the Structural basis of µ- conotoxin KIIIA inhibition of the voltage-gated sodium channel Nav1.7. Biophysical Journal. 2018;114(3):635a. 124. Patel D, Mahdavi S, Kuyucak S. Computational Study of Binding of μ-conotoxin GIIIA to bacterial sodium channels NaVAb and NaVRh. Biochemistry. 2016;55(12):1929-38. 125. Adams DJ, Berecki G. Mechanisms of conotoxin inhibition of N-type (Cav2.2) calcium channels. Biochimica et Biophysica Acta (BBA) - Biomembranes. 2013;1828(7):1619-28. 126. N-type calcium channels by a peptide neurotoxin. Nature. 1994;372(6503):272-5. 127. Ramírez D, Gonzalez W, Fissore R, Carvacho I. Conotoxins as tools to understand the physiological (CaV) channels. Marine Drugs. 2017;15(10):313.

189

Chapter 6 : Summary and Future Work

190

6.1 Summary

Computational molecular modelling and simulation methods have facilitated the early-drug

design process with fewer expenses and time-consumption versus experiments alone. Since

some of the neurological disorders, including chronic pain, require treatment and/or

medications with efficacy, and which do not have the addictive potential of many current

analgesics, the potential of conotoxins as drug leads inhibiting the ligand-gated and voltage-

gated ion channels were investigated in this project. In this thesis, the findings help to

understand the imperative role of conotoxins in interactions with nAChRs and T-type calcium

channels for potential therapeutic purposes.

While many studies have attempted to clarify the binding mechanism of α-conotoxins at

nAChRs, the understanding behind the selectivity of certain conotoxins, such as LsIA targeting

both α7 and (α3)2(β2)3 nAChRs, needs to be improved. Furthermore, the emerging

understanding of the role of Cav3.2, a member of the T-type calcium channel subfamily, in

neuropathic pain also merits the discovery and development of potential drug leads for the

treatment of pain. Additionally, as the pharmacological and physiological properties of α7 and

α3β2 nAChRs, as well as Cav3.2, need further exploration, subtype-selective probes with

isoform specificity are also in demand for understanding the possible relationship between

the binding of the toxins and their effects on the activity of the receptors/ion channels.

The present study demonstrates the potential of conotoxins as drug leads or selective probes

to nAChRs and Cav3.2, as shown in the following:

A previous experimental study has shown that the C-terminal carboxylation of LsIA can

differentiate the α7 from α3β2 nAChR by around 9-fold. This thesis helps further

understanding of the binding mechanism of LsIA at both nAChRs. The toxin anchors to the

interfaces between the adjacent nAChR subunits, at the canonical binding site of nAChRs for

most α-conotoxins.

Via molecular dynamics (MD) simulations, the key factors that affect the differences in

interactions with corresponding nAChR subtypes, as a function of C-terminal charge state,

were also investigated. For α7 nAChR, C-terminal carboxylation directly results in substantially

reduced interactions by LsIA#-C17 with (−)Q79 on α7 nAChR versus the LsIA bound form via a

191

hydrogen bond. Notably, such decreased pairwise interactions mainly contribute to the

reduced nAChR interactions in total with LsIA#, when compared with LsIA. Another pair of the

highlighted coupled residues are LsIA-R10 and α7-W77, which also plays a crucial role in the

reduction of nAChR interactions with LsIA#, possibly due to the reduced cation- interactions

between them. In addition, the clockwise tilting of C- and F-loop of α7 nAChR may be

responsible for the unfavorable binding of LsIA# to the nAChR.

On the contrary, the binding of LsIA# was favored at α3β2 nAChR, particularly at α3(+)β2(−)

and β2(+)α3(−) interfaces. For α3β2 nAChR, salt-bridge/hydrogen bond and hydrophobic

interactions mainly contribute to the enhanced interactions with LsIA# versus LsIA through

the addition of a negative charge to the C-terminus of LsIA. The LsIA#-C17 forms increased

electrostatic interactions with (−)K187 and (−)K82 at α3(+)β2(−) and β2(+)α3(−) binding sites,

respectively. The substantially enhanced hydrophobic interactions formed by LsIA#-P7 and

P14 with L145, W81 and F143 also benefit the binding of LsIA# compared with LsIA at

α3(+)β2(−) interfaces. However, few changes in interactions between LsIA#-R10 and residues

on α3β2 nAChR were observed. The reduced interactions by LsIA with α3β2 nAChR possibly

result from the electrostatic repulsion between the residues of LsIA and nAChR, determined

via adaptive Poisson–Boltzmann solver (APBS) calculations.

The beauty of conotoxins is their unique disulphide-bonded structures with different various

combinations of residues that confer the toxin with the potential to modulate the functions

of distinct targets. For instance, the μ-conotoxins possess 3 disulphide bonds linking 6

conserved cysteines, which potently inhibit voltage-gated sodium (Nav) channels via blocking

the pore of the ion channel. Since the voltage-gated calcium (Cav) channels share a similar

pore structure with that of the Nav channels, docking calculations, MD simulations and

MMPBSA were applied for exploring the prospective binding mechanisms of certain μ-

conotoxins at Cav3.2, using Nav as a conceptual model of pore block.

The binding mechanism of μ-conotoxins, GIIIA, PIIIA, GIIIB, GIIIC and KIIIA at Cav3.2 resembles

that of the toxins targeting Nav channels via blocking the pore of the Cav channel. However,

key residues affecting the differences in Cav3.2 interactions among different μ-conotoxin

bound forms may influence the binding conformation of the neuropeptides at Cav3.2. For

instance, the reduced interactions between D382 on Cav3.2 with PIIIA, GIIIB may account for

192

a shallow binding position of these toxins versus GIIIA, which anchors to the outer vestibule

of the ion channel. Notably, Cav3.2 may be less sensitive to KIIIA versus the other toxins, due

to the loss of important electrostatic interactions by D1504 and K1809 with the peptide. By

contrast, GIIIC exhibits marginally fewer interactions with D382 of Cav3.2, and shares the most

common binding conformation with GIIIA at the ion channel, when compared with the other

three toxins.

Overall, the findings in this thesis may facilitate further mutagenesis study, such as the

mutation on the LsIA-P7 and V11, LsIA#-R10 as well as PIIIA-Q15, GIIIA/GIIIB-R14 and KIIIA-

W8 targeting human α7 nAChR, rat α3β2 nAChR and Cav3.2, respectively, for developing small

peptides with enhanced inhibitory effects on nAChR and voltage-gated ion channels related

to neurological disorders and chronic pain.

6.2 Future work

This thesis employed computational methods to elucidate the determinants that affect the

differential binding of conotoxins at nAChRs and Cav3.2 channels. However, further study is

needed via mutation of single or multiple residues that were suggested to play an important

role in the sensitivity of the toxins to the nAChRs and other ion channels.

As mentioned previously, conotoxins are short peptides rich in disulphide bonds and may go

through different post-translational modifications in most conotoxin subfamilies. For instance,

LsIA obtains a naturally amidated C-terminus in its native type. While the modification of C-

terminus of LsIA can affect the binding of the toxin at α7 and α3β2 nAChRs, this modification

may also influence the sensitivity of conotoxins with a similar amidated C-terminus to bind to

nAChRs and voltage-gated ion channels, such as the μ-conotoxins, PIIIA and GIIIA, targeting

Cav3.2.

In addition, the key residues of both conotoxins and their targeted membrane proteins

described in this thesis were predicted to affect their interactions with nAChRs and Cav3.2. It

would be intriguing to implement mutation at these determinants of conotoxin bound

complexes to further elucidate their role in the binding of the toxins to the nAChRs and Cav

channels via computational and experimental approaches.

193

Another interesting perspective of study on nAChRs is the sensitivity of different nAChR

isoforms to the conotoxins. For instance, an assumption that LsIA# may preferentially bind to

(α3)2(β2)3 rather than (α3)3(β2)2 stoichiometry can be proposed. This hypothesis can be

investigated via further studies employing homology modelling, MD simulations and free

energy calculations, such as umbrella sampling methods. On the other hand, the LsIA and μ-

conotoxins can also possess ribbon and bead forms via alternative arrangements of disulphide

bonds between the conserved cysteines, apart from the globular form studied in this thesis.

Therefore, the interactions by other isoforms of these conopeptides may also influence their

interactions with their corresponding targeted nAChRs and ion channels, as well as the

binding mechanisms of the conotoxins at the targeted proteins.

In conclusion, the full potential of conotoxins is worthwhile to be deeply explored. The

present study is expected to contribute to the ongoing study on conotoxins for therapeutic

purposes.

194

Appendix 1 : Supplementary Materials - Effects of C-

terminal Carboxylation of LsIA on Interactions with

α3β2 Nicotinic Acetylcholine Receptor Interfaces

195

Figure S1.1. RMS Fluctuation (RMSF) plots of -conotoxin LsIA/LsIA# binding at α3β2 nAChR. RMSF plots of LsIA (A) and LsIA# (B) bound forms. RMSF plots of LsIA (C) and LsIA# (D) upon anchoring to the nicotinic acetylcholine receptor.

196

Figure S1.2. Convergence study via calculating the sum of squared differences (SSD) between 3, 6, 9, 15, 18, 21, 24 and 27 ns RMSFs with the 27 ns simulations. SSD of (A) LsIAs and (B) α3β2 nAChR bound by LsIA and LsIA#. The higher SSD values indicate the greater overall difference from the 27 ns RMSF results. The figures show that there is an exponential decay in the difference in RMSF with increasing blocks of time, with the difference approaching zero from the 21 ns, which shows the RMSF from 21 ns is relatively similar to that of the 27 ns.

197

Figure S1.3. Significant changes of pairwise contacts for LsIA anchoring to β2(+)β2(−) interface of α3β2 nAChR. The residues on principal (+) face are labelled with a black line border. The statistical significance of the difference between the number of pairwise interactions was calculated over 28 individual seeds (* p < 0.05).

198

199

Figure S1.4. The α3β2 nAChR interface interactions with LsIA and LsIA#, respectively, according to the statistically significant variations of receptor contacts by C-terminal carboxylation. The top- ranking cluster (cluster #1) was chosen from the whole trajectory for each interface for ligand interaction analyses via Maestro (Schrödinger Release 2020-2) [1]. The structure of LsIA (A, B and C) /LsIA# (D, E and F) was selected based on cluster analysis in GROMACS packages. The observed hydrogen bond, hydrophobic and stacking interactions of LsIA/LsIA# with corresponding residues of α3β2 subtype are shown in the figure.

200

• LsIA and LsIA# intra-molecular interactions

To further explain the effects of the C-terminal amidation/carboxylation of LsIA on its

interactions with α3β2 nAChR, we calculated the number of inter-atomic contacts and the

atomic distances between the C-terminal of LsIA with other residues within the toxin, as well

as between each residue of LsIA and the receptor at the α3(+)β2(−) and β2(+)α3(−) binding

interfaces.

At α3(+)β2(−) interfaces, for LsIA#, the short distances (2.7 Å and 3.4 Å) between carboxylated

C-T (C17) and LsIA#-R10, allows an intramolecular salt bridge formed by C17 and R10 between

them (Figure S1.5), whereas this kind of interaction is absent in LsIA. Figure S1.5A shows an

illustrative graphic, whereas Figure S1.5B shows histograms of the C17-R10 distance,

indicating the lower separation for LsIA#. These strong contacts also exist in LsIA# binding at

β2(+)α3(−) interfaces, but are slightly reduced at β2(+)β2(−) interfaces (not shown), providing

further support for the distinctiveness of the β2(+)β2(−) interface compared to the mixed-

subunit interfaces, as noted in previous sections. In summary, this finding supports the

distinct pattern at RMSF and “diff in RMSF” plots of chain I (β2(+)β2(−)), indicating higher

rigidity of LsIA# versus LsIA may be partly due to enhanced internal salt bridges which restrict

the toxin’s mobility.

Figure S1.5. Intra-molecular contacts between R10 and C17 of LsIA# versus LsIA binding at α3(+)β2(−) interfaces. (A) Significant intra-molecular contacts formed within the LsIA# (transparent orange color) versus the LsIA (transparent green color), upon binding at α3(+)β2(−) interface. Important residues regarding intra-molecular interactions are shown in CPK form. The red dashed line represents the contacts that are much stronger in this form of LsIA/α3β2 complex. (B) The probability density function of distance (Å) between the coupled residues of LsIAs, and the colored dashed line demonstrates the median distance of the pairwise interactions.

201

Figure S1.6. Interactions formed by LsIA-G2 and C3 with the receptor residues on the principal (+) face at α3(+)β2(−) interface of α3β2 nAChR. Binding conformation of LsIA (A) and LsIA# (B) at the (+) face with the pairwise interactions, which may substantially affect the binding affinity. The key residues involved in pairwise interactions on LsIAs are shown in CPK form, while the corresponding residues on the receptor are depicted with the Licorice form (a drawing format of graphic representation in VMD).

The pairwise interactions in both receptor-ligand complexes, namely S5-(+)D195, P7-(+)W174, and G2-

(+)Y215, mainly contribute to stabilizing the binding of LsIA/LsIA# to the α3(+)β2(−) interfaces. These

aromatic residues have also been suggested to play an important role in the binding of α-conotoxin

ImI and ACh to human α7 and rat α3β2 nAChR, respectively, via forming cation- interactions with

ACh and ImI-R7 [2-4]. The persistent interactions between the aromatic residues on the α3(+) face and

the conserved residues, C9 and P7, of LsIA can be observed in both LsIA and LsIA# anchored forms,

such as P7-(+)W174 on the α3(+)β2(−) interfaces (Figures S1.4A, D and S1.7). The conserved proline

exists in most α-conotoxins in loop 1 [5,6].

• Pairwise interactions weakly affect the binding affinity at α3(+)β2(−) interfaces

202

Figure S1.7. Interactions formed by LsIA-P7 and C9 with the receptor residues on the principal (+) face at α3(+)β2(−) interface. Binding mode of LsIA (A) and LsIA# (B) at the (+) face with the pairwise interactions which weakly affect the binding affinity. (C) The probability density function of distance (Å) between the coupled residues of bound forms. 32a represents the LsIA bound type, whilst 32c denotes the LsIA# bound complex.

• Pairwise interactions weakly affect the binding affinity at β2(+)α3(−) interfaces

At the principal β2(+) face, the conserved residues, P7 and C9 (not shown), and polar residue

N6 of LsIA interact (Figure S1.8) in a similar manner to those at α3(+)β2(−) (Figure S1.7).

Specifically, P7 exhibits higher contacts with aromatic residues, (+)W175 and (+)Y220 on β2(+)

face for LsIA# compared to LsIA, whereas which establish relatively enhanced contacts with

LsIA-A8. However, the differential β2(+) contacts by N6, P7, A8 and C9 slightly affect the

enhanced contacts by LsIA# at β2(+)α3(−) interfaces.

203

Figure S1.8. Interactions formed by LsIA-P7 and A8 with the receptor residues on the principal (+) face at β2(+)α3(−) interface of α3β2 nAChR. Binding conformation of LsIA (A and C) and LsIA# (B and D) at the (+) face with the pairwise interactions which weakly affect the binding affinity. The LsIA and LsIA# binding at the interfaces are shown in transparent pink and grey colors, respectively. (E) The probability density function of distance (Å) between the coupled residues of bound forms.

204

• Pairwise interactions at β2(+)β2(−) interface of α3β2 nAChR bound by LsIA

At the β2(+)β2(−) interface, the hydrogen bonds formed between β2(+)-Y177 and the non-

conserved residue N12 of LsIA, reinforce the contacts with amidated LsIA bound complex

relative to the LsIA# (Figure S1.3,S1.9). Apart from this, the contacts formed by β2(+)-Y220 on

the β10 strand of β2 subunit with LsIA#-C9, are also substantially reduced versus LsIA bound

form in terms of van der Waals interactions. Notably, the homologous residues of α7 nAChR

subunit corresponding to Y177 and Y220 of β2, have been suggested to form polar

interactions with ImI-R7 via aromatic interactions [2,3] and related to the binding of α-

conotoxin [7]. In addition, a hydrogen bonding contact formed by (+)D178 and N12 and the

hydrophobic interactions between (+)W175 and P7, are slightly improved in the LsIA bound

complex. Nevertheless, few pairwise contacts with significant enhancement were identified,

resulting from the C-terminal carboxylation of LsIA at this interface, possibly due to the loss

of contacts with R10 and C17 of LsIA#. Interestingly, on the (−) face of the β2(+)β2(−)

interface, the only apparent change in interactions occurs between P14 and (−)F143,

compared with other pairwise interactions at α3(+)β2(−) and β2(+)α3(−) sites. The few

variations between LsIA and LsIA# in contact with residues on (−) face at this site probably

result from the absence of key distinguished receptor interactions by LsIA#-C17 and R10.

205

Figure S1.9. Important interactions formed by LsIA-C9 and N12 with the relative residues on β2(+)β2(−) interface of α3β2 nAChR. Binding conformation of LsIA (A) and LsIA# (B) at the (+) face with the pairwise interactions drastically affecting the binding affinity. Binding conformation of LsIA (C) and LsIA# (D) at the (−) face with the pairwise interactions. The LsIA and LsIA# binding at the interfaces are shown in transparent red and ice-blue colors, respectively. (E) The probability density function of distance (Å) between the important pairwise interactions of bound forms.

206

Figure S1.10. The distance of the inter-subunit interactions between the principal (+) and the accessory (−) face of α3β2 nAChR interfaces bound by LsIAs.

Schrödinger Maestro (Version 2020-2), Schrodinger LLC: New York, NY, 2020.

Dutertre, S.; Nicke, A.; Tyndall, J.D.A.; Lewis, R.J. Determination of α-conotoxin binding modes on neuronal nicotinic acetylcholine receptors. J. Mol. Recognit. 2004, 17, 339-347, doi:10.1002/jmr.683.

Ellison, M.; Gao, F.; Wang, H.-L.; Sine, S.M.; McIntosh, J.M.; Olivera, B.M. α-Conotoxins ImI and ImII target distinct regions of the human α7 nicotinic acetylcholine receptor and distinguish human nicotinic receptor subtypes. Biochemistry 2004, 43, 16019-16026, doi:10.1021/bi048918g.

Sambasivarao, S.V.; Roberts, J.; Bharadwaj, V.S.; Slingsby, J.G.; Rohleder, C.; Mallory, C.; Groome, J.R.; McDougal, O.M.; Maupin, C.M. Acetylcholine Promotes Binding of α-Conotoxin MII at α3β2 Nicotinic Acetylcholine Receptors. ChemBioChem 2014, 15, 413-424, doi:10.1002/cbic.201300577.

References

Armishaw, C.; Jensen, A.A.; Balle, T.; Clark, R.J.; Harpsøe, K.; Skonberg, C.; Liljefors, T.; Strømgaard, K. Rational Design of α-Conotoxin Analogues Targeting α7 Nicotinic Acetylcholine Receptors: Improved antagonistic activity by incorporation of proline derivatives. J. Biol. Chem. 2009, 284, 9498-9512, doi:10.1074/jbc.M806136200.

Cartier, G.E.; Yoshikami, D.; Gray, W.R.; Luo, S.; Olivera, B.M.; McIntosh, J.M. A new α- conotoxin which targets α3β2 nicotinic acetylcholine receptors. J. Biol. Chem. 1996, 271, 7522- 7528, doi:10.1074/jbc.271.13.7522.

Dutertre, S.; Lewis, R.J. Computational approaches to understand α-conotoxin interactions at neuronal nicotinic receptors. European Journal of Biochemistry 2004, 271, 2327-2334, doi:10.1111/j.1432-1033.2004.04147.x.

207

Appendix 2 : Supplementary Materials - Differentiation

of Cav3.2 Interactions by μ-Conotoxins as Potential

Analgesics

208

o N

e d o M

21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0

1000

2000

3000

4000

5000

6000

7000

8000

Simulation time (ns)

Figure S2.1. The cluster analysis over the apo Cav3.2 trajectories. The simulation run contains 8 individual trajectories with 1000ns for each.

Table S2.1. The binding affinity (kcal/mol) of five μ-conotoxins docking into Cav3.2 with 4 repetitions. The perspective binding affinity with the best binding conformation of the conotoxins above the vestibule of Cav3.2 is highlighted with bold and italic forms.

Repetition PIIIA GIIIA KIIIA GIIIB GIIIC

Repeat 1st

Repeat 2nd

Repeat 3rd

Cluster models model1 model2 model3 model1 model2 model3 model1 model2 model3 model1 model2 model3 Repeat 4th -6.5 -7.5 -6.5 -6 -7.3 -6.2 -5.9 -6.9 -6.2 -5.8 -7.2 -6.1 -7.5 -5.9 -5.9 -7.2 -6 -5.9 -7.3 -5.7 -6.1 -5.9 -5.8 -6 -8.1 -7.9 -6.9 -8.2 -7.4 -7 -7.2 -7.5 -7 -7.4 -7.4 -6.9 -5.9 -6.7 -5.3 -6 -6.6 -5.6 -5.2 -6.9 -5.4 -4 -6.9 -5.8 -6.8 -6.5 -7.1 -6.7 -6.4 -6.9 -6.5 -6.3 -6.7 -6.2 -6.5 -7.1

Table S2.2. The μ-conotoxin bound Cav3.2 systems for molecular dynamics (MD) simulations involved in this study. The number of molecules for the DOPC bilayer, water molecules (TIP3P), Ca2+ and Cl- constitute each conotoxin bound complex.

209

GIIIA

PIIIA

GIIIB

KIIIA

GIIIC

DOPC TIP3P(water) Ca2+ Cl-

375 38452 201 422

384 39659 201 421

385 38898 201 423

371 38465 201 419

375 39697 201 423

Figure S2.2. Binding conformation of μ-conotoxin PIIIA and GIIIA at Cav3.2. PIIIA (A) and GIIIA (B) are shown with green and cyan colours, respectively. (C,D) Density of distance between the coupled residues among μ-conotoxins bound complexes. The y-axis represents the probability density functions of the distance between the coupled residues, which was applied in all the density of distance plots in the present study.

210

Figure S2.3. Binding conformation of μ-conotoxin GIIIB and GIIIA at Cav3.2. GIIIB (A) and GIIIA (B) are shown with yellow and cyan colors, respectively. (C,D) Density of distance between the coupled residues among μ-conotoxins bound complexes.

211

Figure S2.4. Binding conformation of μ-conotoxin KIIIA and GIIIA at Cav3.2. KIIIA (A) and GIIIA (B) are shown with dark blue and cyan colors, respectively. (C,D) Density of distance between the coupled residues among μ-conotoxins bound complexes.

212

Figure S2.5. Binding conformation of μ-conotoxin GIIIC and GIIIA at Cav3.2. GIIIC (A) and GIIIA (B) are shown with pink and cyan colors, respectively. (C,D) Density of distance between the coupled residues among μ-conotoxins bound complexes.

213

Figure S2.6. The residues affect the flexibility of Cav3.2 at DI and III (A) and DII and IV (B), bound by toxins versus GIIIA. The positively charged residues, arginine, are colored with blue color, while the hydrophobic residues, alanine, are colored with white.

Figure S2.7. The overlayed binding conformation of GIIIA and PIIIA. The toxins and the Cav3.2 are depicted with Cartoon and NewCartoon forms in VMD, respectively. The ion channel bound by GIIIA

214

and its determinant residues are shown in transparent silver color. The voltage-sensing domains (VSDs) were removed for clarification. The last frame of whole trajectory was selected for each conotoxin bound complex, which was applied for all the figures shown below.

Figure S2.8. The overlayed binding conformation of GIIIA and GIIIB. The toxins and the Cav3.2 are depicted with Cartoon and NewCartoon forms in VMD, respectively. The ion channel bound by GIIIA and its determinant residues are shown in transparent silver color. The VSDs were removed for clarification.

215

Figure S2.9. The overlayed binding conformation of GIIIA and KIIIA. The binding conformation of the toxins at (A) domains I and III as well as (B) domains II and IV. The toxins and the Cav3.2 are depicted with Cartoon and NewCartoon forms in VMD, respectively. The ion channel bound by GIIIA and its determinant residues are shown in transparent silver color. The VSDs were removed for clarification.

216

Figure S2.10. The overlayed binding conformation of GIIIA and GIIIC. The binding conformation of the toxins at (A) domains I and III as well as (B) domains II and IV. The toxins and the Cav3.2 are depicted with Cartoon and NewCartoon forms in VMD, respectively. The ion channel bound by GIIIA and its determinant residues are shown in transparent silver color. The VSDs were removed for clarification.

217

Figure S2.11. The overlayed binding conformation of all five μ-conotoxins at Cav3.2 in the present study. The top (A) and side (B) view of the binding conformations of the toxins. For clarification, the ion channel is shown with QuickSurf drawing method in VMD, whilst the conotoxins are shown in Ribbon form.

218

219

Figure S2.12. Sequence alignment using ClustalW between human Cav3.1 (PDB ID: 6KZO) and human Cav3.2 (UniProtKB: O95180).

220

Appendix 3 : List of Publications and Conference

Abstracts

Journal publications

Wen, J. & Hung, A. Effects of C-Terminal Carboxylation on α-Conotoxin LsIA Interactions with

Human α7 Nicotinic Acetylcholine Receptor: Molecular Simulation Studies. Marine Drugs.

2019;17(4):206. DOI:10.3390/md17040206

Wen, J., Adams, D.J. & Hung, A. Interactions of the α3β2 Nicotinic Acetylcholine Receptor

Interfaces with α-Conotoxin LsIA and its Carboxylated C-terminus Analogue: Molecular

Dynamics Simulations. Marine Drugs. 2020;18(7):349. DOI: 10.3390/md18070349

Other conference abstracts

Wen, J. & Hung, A. Molecular Simulation Studies of C-Terminal Carboxylation impacts on α-

Conotoxin LsIA Interactions with Human Nicotinic Acetylcholine Receptors. Oral and poster

presentations at: the 92nd Annual Meeting of the Japanese Biochemical Society; Sep 2019;

Yokohama, JP

Wen, J. & Hung, A. The Role of C-Terminal Carboxylation In α-Conotoxin LsIA Interactions with

Human α7 Nicotinic Acetylcholine Receptor in Silico. Oral and poster presentations at: the

57th Annual Meeting of the Biophysical Society of Japan; Sep 2019; Miyazaki, JP

221

Doctoral thesis of Philosophy: Computational studies of neuroactive peptides for treating neurological disorders and disease

Tags: