REVIEW ARTICLE
Structure and potential function of c-aminobutyrate
type A receptor-associated protein
Jeannine Mohrlu
¨der
1
, Melanie Schwarten
1,2
and Dieter Willbold
1,2
1 Institut fu
¨r Strukturbiologie und Biophysik (ISB-3), Forschungszentrum Ju
¨lich, Germany
2 Institut fu
¨r Physikalische Biologie, Heinrich-Heine-Universita
¨tDu
¨sseldorf, Germany
Introduction
Human c-aminobutyrate type A (GABA
A
) receptor-
associated protein (GABARAP) was initially identified
as a ligand of the c-subunit of the GABA
A
receptor
[1], which is a ligand-gated ion channel that mediates
rapid inhibitory synaptic transmission in the central
nervous system (CNS) and serves as a target for multi-
ple neuroactive drugs. Further studies revealed that
GABARAP is also involved in the GABA
A
receptor
trafficking to the plasma membrane [2–4]. GABARAP
and its homologs are found in a broad range of
organisms, from plants to mammals. Mammalian
forms of GABARAP share 100% identity at the amino
acid level, suggesting that the function of GABARAP
is essential or advantageous in mammals.
Currently, knowledge obtained from the large
number of GABARAP binding proteins strongly
suggest that it participates in multiple biological pro-
cesses, such as general vesicular transport and fusion
events, autophagy and apoptosis. However, the
precise cellular function of GABARAP remains to
Keywords
apoptosis; autophagy; GABA
A
receptor;
GABARAP; protein structure; protein–protein
interaction; trafficking
Correspondence
J. Mohrlu
¨der or D. Willbold, ISB-3,
Forschungszentrum Ju
¨lich, 52425 Ju
¨lich,
Germany
Fax: +49 2461 61 2023
Tel: +49 2461 61 2100
E-mail: j.mohrlueder@fz-juelich.de;
d.willbold@fz-juelich.de
(Received 14 May 2009, revised 8 July
2009, accepted 9 July 2009)
doi:10.1111/j.1742-4658.2009.07207.x
The c-aminobutyrate type A receptor-associated protein (GABARAP) is a
ubiquitin-like modifier, and is implicated in a variety of membrane traffick-
ing and fusion events that are crucial to synaptic plasticity, autophagy and
apoptosis. However, important aspects of GABARAP function and regula-
tion remain poorly understood. We review the current state of knowledge
about GABARAP, highlighting newly-identified GABARAP ligands, and
discuss the possible physiological relevance of each ligand interaction.
Abbreviations
Atg, autophagy-related protein; BDV, borna disease virus; CHC, clathrin heavy chain; CNS, central nervous system; CRT, calreticulin; DDX47,
DEAD box polypeptide 47; ER, endoplasmic reticulum; GABA
A
, 4-aminobutyrate type A; GABARAP, GABA
A
receptor-associated protein;
GABARAPL, GABARAP-like; GATE-16, Golgi-associated ATPase enhancer of 16 kDa; GEC1, initially isolated in guinea-pig endometrial cells,
estrogen-induced 1.8 kb RNA coded protein; GluR, glutamate receptor; GRIP1, glutamate receptor-interacting protein 1; hp, hydrophobic
pocket; LGG, LC3, GABARAP and GATE-16 family; MAP1 LC3, microtubule-associated protein 1 light chain 3; Nix, Nip-like protein x; NSF,
N-ethylmaleimide sensitive factor; PDZ, postsynaptic density protein ⁄discs large ⁄zonula occludens-1; PE, phosphatidylethanolamine; PI,
phosphoinositide; PRIP-1, phospholipase C-related inactive protein type 1; PS, phosphatidylserine; RAB24, ras-related protein 24; TfR,
transferrin receptor; UBL, ubiquitin-like; Ubq, ubiquitin; ULK1, unc-51-like kinase 1.
FEBS Journal 276 (2009) 4989–5005 ª2009 The Authors Journal compilation ª2009 FEBS 4989
be elucidated in more detail, as will be discussed
below.
GABARAP is reversibly coupled to
membranes in a ubiquitin (Ubq)-like
manner
The 14 kDa GABARAP protein belongs to the Ubq-
like modifiers and is enzymatically coupled to a target
moiety in a Ubq-like manner [5]. Ubq is covalently
attached to its target protein via a three-step mecha-
nism (Fig. 1A). It is initially activated by an E1 Ubq-
activating enzyme, resulting in a thioester linkage
between its carboxy-terminal glycine and the E1 cyste-
ine sulfhydryl group. This process requires ATP as an
energy source. Ubq is subsequentially transferred to
the active site cysteine of E2 and is finally coupled to a
lysine residue of the target protein via an isopeptide
bond. This step is mediated by one of the hundreds of
E3 Ubq-protein ligases.
Unlike Ubq, GABARAP does not covalently attach
to a protein, but rather to the phospholipids phospha-
tidylethanolamine (PE) or phosphatidylserine (PS) [6,7]
(Fig. 1C). The cysteine protease autophagy-related
E1
C
Ub GG
E1
C
Ub GG
E3
Targets
Ub
Ub
Ub
Ub
GG
K
A Ubiquitin (Ub)
E2
Ub GG
C
ATP AMP+PPi
B Atg12
Atg12 G
Atg7
C
Atg12 G
Atg12 G
K
Atg10
2
1
g
tA G
C
ATP AMP+PPi
Atg7
C
Atg5
Atg12-Atg5
Atg16L
Atg16L
Atg4B
GRP GL
GRP G
Atg7
C
L
GRP G
C GABARAP (GRP)
?
Atg7
C
GRP G
Atg3
C
GRP G
ATP AMP+PPi
Atg4B
GRP G
PE
Atg12-Atg5
Atg16L
Atg12-Atg5
Atg16L
Atg12-Atg5
Atg16L
Atg12-Atg5
Atg16L
Fig. 1. GABARAP (GRP) and Atg12 are modified via an Ubq-like conjugation system. (A) Ubq processing is initially mediated by E1 (ubiqutin-
activating enzyme), which results in a thioester linkage between the C-terminal glycine of Ubq and the cysteine sulfhydryl group. It is subse-
quently transferred to the active site cysteine of E2 and is finally attached to lysine side chains of target proteins via isopeptide bonds. The
latter step is mediated by hundreds of specific E3 Ubq-protein ligases. Further ubiquitin molecules can be attached to target-bound ubiquitin,
thus forming a poly-ubiquitin chain. (B) Atg12 conjugation is initiated by Atg7, an E1-like enzyme. Atg12 is then transferred to Atg10, an
E2-like conjugating enzyme, and is finally coupled to Atg5 through an isopeptide bond. The Atg12–Atg5 conjugate associates with Atg16L to
form huge protein complexes. (C) The first processing step of GABARAP is carried out by the cysteine protease Atg4B, leading to a trun-
cated form (GABARAP-I) with a C-terminal glycine residue. It is subsequentially activated by Atg7 and then transferred to the E2-like enzyme
Atg3. GABARAP is finally attached to phospholipids (PE or PS), and this is considered to be mediated by Atg12–Atg5 ⁄Atg16 multimers.
Delipidation is mediated by Atg4B as well.
Structure and function of GABARAP J. Mohrlu
¨der et al.
4990 FEBS Journal 276 (2009) 4989–5005 ª2009 The Authors Journal compilation ª2009 FEBS
protein 4B (Atg4B) processes GABARAP into GABA-
RAP-I with an exposed C-terminal glycine [8,9]. Cata-
lyzed by the E1-like enzyme Atg7 and the E2-like
enzyme Atg3, GABARAP-I is subsequentially trans-
formed into GABARAP-PE or GABARAP-PS
(GABARAP-II) [10,11], which are considered to be
membrane-associated. GABARAP-II delipidation is
also mediated by Atg4B [8]. No E3-like enzyme has
yet been identified for the GABARAP modification.
Nevertheless, based on current knowledge about the
processing of the GABARAP homologous proteins
Atg8 and MAP LC3 [12–15], the approximately
800 kDa Atg16L protein complex can also be specu-
lated to assume E3-like activity for GABARAP lipid-
ation. This complex consists of homo-oligomerized
conjugates of Atg12, Atg5 and Atg16 (Atg16L in
mammals). It is part of the second cellular Ubq-like
protein modification process required for autophago-
some formation, namely the Atg12 Ubq-like system
[16–18] (Fig. 1B).
GABARAP is known to be a cytosolic protein
that localizes to membrane structures, such as trans-
port vesicles, golgi network and the endoplasmic
reticulum (ER) [19,20]. It remains a matter of con-
troversy as to whether the subcellular distribution
of GABARAP depends on its modification state.
Tanida et al. [5] reported that GABARAP localizes
to membrane compartments even prior to its lipida-
tion. By contrast, Kabeya et al. [6] found that
GABARAP is localized at autophagosomal mem-
branes after lipidation.
GABARAP belongs to the
GABARAP-like ⁄microtubule-associated
protein 1 light chain 3 (MAP1 LC3)
protein family
GABARAP is classified as a protein of the GABARAP-
like family (structural classification of proteins;
SCOP, http://scop.mrc-lmb.cam.ac.uk/scop/). Sequence
based protein classification (UniProt, http://www.
uniprot.org/) also lists it as a member of the MAP1
LC3 protein family. This protein family encompasses
GABARAP-like 1 (GABARAPL1) ⁄initially isolated in
guinea pig endometrial cells, estrogen-induced 1.8 kb
RNA coded protein (GEC1); GABARAP-like 2
(GABARAPL2) ⁄Golgi-associated ATPase enhancer of
16 kDa (GATE-16); GABARAP-like 3 (GABARAPL3);
microtubule-associated protein 1 light chain 3 (MAP1
LC3); Atg8 and LC3, GABARAP and GATE-16 family
(LGG) (Fig. 2 and Table 1). These proteins are evolu-
tionary conserved in eukaryotic cells from plants to
mammals. Regarding sequence conservation, human
GABARAP exhibits 100% identity with other mamma-
lian organisms, thus demonstrating complete evolution-
ary conservation.
Because GABARAP knockout mice are viable and
fertile and display no obvious alterations of behaviour
and health compared to wild-type animals, GEC1 is
considered to functionally substitute for GABARAP
because both proteins display 86% sequence identity.
However, the expression level of GEC1 and GATE-16
in GABARAP knockout mice is unaffected [21,22].
Human GABARAP ------------MKFVYKEEHPFEKRRSEGEKIRKKYPDRVPVIVEKAPKAR-IGDLDKKKYLVPSDLTVGQFYFLI 64
Human GABARAPL1 ------------MKFQYKEDHPFEYRKKEGEKIRKKYPDRVPVIVEKAPKAR-VPDLDKRKYLVPSDLTVGQFYFLI 64
Human GABARAPL2 ------------MKWMFKEDHSLEHRCVESAKIRAKYPDRVPVIVEKVSGSQ-IVDIDKRKYLVPSDITVAQFMWII 64
Human GABARAPL3 ------------MKFQYKEVHPFEYRKKEGEKIRKKYPDRVPLIVEKAPKAR-VPDLDRRKYLVPSDLTDGQFYLLI 64
Human MAP1LC3A MPSDR----------PFKQRRSFADRCKEVQQIRDQHPSKIPVIIERYKGEKQLPVLDKTKFLVPDHVNMSELVKII 67
Human MAP1LC3B MPSEK----------TFKQRRTFEQRVEDVRLIREQHPTKIPVIIERYKGEKQLPVLDKTKFLVPDHVNMSELIKII 67
Human MAP1LC3C MPPPQK----IPSVRPFKQRKSLAIRQEEVAGIRAKFPNKIPVVVERYPRETFLPPLDKTKFLVPQELTMTQFLSII 73
Yeast Atg8 ------------MKSTFKSEYPFEKRKAESERIADRFKNRIPVICEKAEKSD-IPEIDKRKYLVPADLTVGQFVYVI 64
C.elegans LGG-1 ------------MKWAYKEENNFEKRRAEGDKIRRKYPDRIPVIVEKAPKSK-LHDLDKKKYLVPSDLTVGQFYFLI 64
C.elegans LGG-2 MSGNRGGSYISGIVPSFKERRPFHERQKDVEEIRSQQPNKVPVIIERFDGERSLPLMDRCKFLVPEHITVAELMSIV 77
C.elegans LGG-3 ---------------METETATTPTGNTEPTAAASAEPPKSDKVTVRLRNIADAPVLKNKKMVVNPTDTVASFILKL 62
Human GABARAP RKRIHLRAEDALFFFVN-NVIPP---TSATMGQLYQEHHEEDFFLYIAYSDESVYGL-------------------- 117
Human GABARAPL1 RKRIHLRPEDALFFFVN-NTIPP---TSATMGQLYEDNHEEDYFLYVAYSDESVYGK-------------------- 117
Human GABARAPL2 RKRIQLPSEKAIFLFVD-KTVPQ---SSLTMGQLYEKEKDEDGFLYVAYSGENTFGF-------------------- 117
Human GABARAPL3 RKRIHLRPEDALFFFVN-NTIPP---TSATMGQLYEDSHEEDDFLYVAYSNESVYGK-------------------- 117
Human MAP1LC3A RRRLQLNPTQAFFLLVNQHSMVS---VSTPIADIYEQEKDEDGFLYMVYASQETFGF-------------------- 121
Human MAP1LC3B RRRLQLNANQAFFLLVNGHSMVS---VSTPISEVYESEKDEDGFLYMVYASQETFGMKLSV---------------- 125
Human MAP1LC3C RSRMVLRATEAFYLLVNNKSLVS---MSATMAEIYRDYKDEDGFVYMTYASQETFGCLESAAPRDGSSLEDRPCNPL 147
Yeast Atg8 RKRIMLPPEKAIFIFVN-DTLPP---TAALMSAIYQEHKDKDGFLYVTYSGENTFGR-------------------- 117
C.elegans LGG-1 RKRIQLRPEDALFFFVN-NVIPQ---TMTTMGQLYQDHHEEDLFLYIAYSDESVYGGEVEKKE-------------- 123
C.elegans LGG-2 RRRLQLHPQQAFFLLVNERSMVS---NSMSMSNLYSQERDPDGFVYMVYTSQPAFG--------------------- 130
C.elegans LGG-3 RKLLNIQANNSLFLYIDNTFAPSPDTTFETLSRCYSVKITDKEILELQYSITPAYG--------------------- 118
α4
β3β4
α2
α1
α3
β2
β1
Fig. 2. Sequence alignment of GABARAP-like proteins. Amino acids are shown in single-letter code. Alignment was performed using
CLUSTALW 2.0.10. The enumerated secondary structure elements of GABARAP are depicted above the respective amino acid position of
the GABARAP sequence.
J. Mohrlu
¨der et al. Structure and function of GABARAP
FEBS Journal 276 (2009) 4989–5005 ª2009 The Authors Journal compilation ª2009 FEBS 4991
GABARAP and its mammalian homologs are
expressed ubiquitously in all tissues that have been
investigated with moderate variations in their respec-
tive expression patterns [20,22–25]. GABARAPL1
shows higher expression in the CNS, whereas GABA-
RAP predominates in the endocrine glands [22,23].
MAP LC3 protein family members Atg8, GATE16,
and LC3 undergo a Ubq-like modification similar to
GABARAP lipidation [6,9–11,24,26–28]. So far, this
has not been investigated for GABARAPL3 and
LGG.
Atg8 is the only member of the MAP1 LC3 family
present in yeast and thus has been studied extensively
because it exhibits model system properties. Atg8 is a
scaffold protein, supporting membrane expansion by
mediating tethering and hemifusion of membranes
upon oligomerization [29]. During autophagy, it con-
trols the expansion of the autophagosome precursor,
the phagophore [16,30,31]. When autophagy is
induced, most Atg8 is converted to Atg8-PE [32]. The
amount of Atg8 regulates the level of autophagy in
yeast by specifically modulating the size of the auto-
phagosomes, without affecting the number of auto-
phagosomes. Autophagosome formation comprises a
cycle of Atg8 trafficking in which Atg8 is initially
recruited to an expanding phagophore structure and
is subsequently released from it [31]. It was demon-
strated that the release of Atg8 is required for the
completion of autophagosome formation and is medi-
ated by PE-deconjugation [31]. Atg8 knockout cells
are unable to accumulate autophagic vesicles inside
the vacuole. Furthermore, maturation of the amino-
peptidase I precursor, starvation-induced protein
breakdown and the survival rate during starvation are
drastically affected by Atg8 knockout [30,33]. Apart
from its function in autophagy, Atg8 also participates
in further membrane trafficking events. It substitutes
for GATE-16 activity in intra-Golgi transport in vitro,
indicating that both proteins share a similar conserved
function [34].
Among the mammalian GABARAP homologs,
MAP LC3 is the most studied protein. It was origi-
nally identified as one of three light chains (LC1, LC2
and LC3) that copurifies with MAP1A and MAP1B
[35]. Three human homologs of LC3 (LC3A, LC3B
and LC3C) have been reported. They exhibit distinct
expression patterns in different human tissues, as well
as different patterns of subcellular localization [36].
Thus, it thus can be speculated that these three ortho-
logs may differ in their physiological function. LC3
undergoes a post-translational modification analogous
to the above-mentioned GABARAP-like lipidation
procedure (Fig. 1) [7,10,11,28]. Induction of autophagy
by different stress signals such as starvation stimulates
the conversion of LC3-I to LC3-II and the upregula-
tion of LC3 expression [37]. Depending on form-II
formation, LC3 localizes on the membrane of complete
spherical autophagosomes, as well as on isolation
membranes, establishing this protein as a reliable auto-
phagosomal marker [6,38,39].
GABARAP structures
In addition to the post-translational modification,
GABARAP and Ubq also show intriguing structural
similarities, although their sequences exhibit only 7%
identity. Thus, the major part of GABARAP displays
a ubiquitin-like (UBL) fold. Ubq and GABARAP resi-
dues P30 to D111 superimpose with an rmsd of 2.5 A
˚
over 72 core a-carbon atoms (Fig. 3A). The UBL core
domain contains two parallel b-strands (b1 and b4),
which are framed by one anti-parallel b-strand (b2 and
b3) on each side, and two a-helices (a3 and a4) on the
concave side of the b-sheet. In addition to the UBL
core domain, GABARAP possesses two additional
a-helices at its N-terminus (a1 and a2). Those helices
appear to be important for tubulin binding and oligo-
merization [40].
Efforts to elucidate the 3D structure of GABARAP
yielded three crystals and two solution structures [40–
44]. Overall, these structures are very similar, although
there are remarkable differences at the N-terminal resi-
dues (M1-R14) and the loops connecting b1, b2 (P37-
K46) and a3, b3 (K66-A75), as well as the C-terminus
(S113-L117). Interestingly, Stangler et al. [43] observed
that residues within the N-terminal a-helix demon-
strate broadened and split resonances in their NMR
spectra, indicating the existence of at least two slightly
different conformations, whereas Kouno et al. [44]
Table 1. Sequence identities among GABARAP-like proteins.
Protein Species
Sequence
identity
with human
GABARAP (%)
GABARAPL1 ⁄GEC1 Homo sapiens 86
GABARAPL2 ⁄GATE-16 Homo sapiens 57
GABARAPL3 Homo sapiens 82
MAP1 LC3A Homo sapiens 29
MAP1 LC3B Homo sapiens 30
MAP1 LC3C Homo sapiens 37
LGG-1 Caenorhabditis elegans 82
LGG-2 Caenorhabditis elegans 32
LGG-3 Caenorhabditis elegans 21
Atg8 Saccharomyces cerevisiae 54
Structure and function of GABARAP J. Mohrlu
¨der et al.
4992 FEBS Journal 276 (2009) 4989–5005 ª2009 The Authors Journal compilation ª2009 FEBS
could not determine the structure of the N-terminal
region. Coyle et al. [40] also showed that, in crystals,
GABARAP can adopt at least two conformations: one
so-called ‘closed’ conformation with the N-terminal
a-helix connected to the core region, and an oligomeric
conformation of GABARAP termed the ‘open’ confor-
mation. In the latter, the N-terminus of one GABA-
RAP molecule projects away from the core domain in
the direction of a neighboring GABARAP molecule.
An oligomer formation was also described by Naka-
togawa et al. [29] for the yeast homolog of GABARAP
Atg8 upon conjugation with PE.
Ligand binding properties of GABARAP
Phage display screens against GABARAP with ran-
domized dodecapeptide libraries have revealed that
most of the GABARAP binding peptides contain at
least one tryptophan residue. Indeed, NMR spectro-
scopic studies of GABARAP revealed two hydropho-
bic pockets with remarkable affinity to indol
derivatives on its surface, termed hp1 and hp2 [45].
These hydrophobic pockets are located on one face of
its core structure. In addition, this site is highly con-
served at the amino acid level compared to other
GABARAP homologs; however the opposite side
shows significant divergence.
1
H-
15
N heteronuclear
single quantum coherence titration experiments with
15
N-GABARAP showed that hp1 is formed by E17,
I21, P30, K48, L50 and F104, whereas hp2 is formed
by Y49, V51, P52, L55, F60, L63 and I64 (Fig. 4).
Two crystal structures of GABARAP-peptide
complexes confirmed that these hydrophobic pockets
determined with the use of small ligands are also rele-
vant for protein–protein interactions. GABARAP in
complex with a high-affinity artificial peptide identified
by phage display screening was the first 3D structure
of GABARAP in complex with a ligand. Over its
entire length, the peptide (DATYTWEHLAWP) is in
close contact with GABARAP. The mainly hydropho-
bic interactions within the complex are mediated by a
short central 3
10
helix (WEHL) and an additional
a1
a2
a3
a4
a1
b1
b2
b4
b3
a3
a4
A
B
a2
W183
L186
Fig. 3. (A) Comparison of GABARAP (blue, Protein Data Bank code:
1KOT) [43] and ubiquitin (yellow, Protein Data Bank code: 1UBI)
[143]. (B) Overlay of free GABARAP (blue) and GABARAP in com-
plex with CRT residues 178–188 (red, Protein Data Bank code:
3DOW) [47].
hp1
hp2
a4A
B
a3
a1
a2
b2
b1
b3
b4
Fig. 4. Illustration of prominent hydrophobic pockets on the surface
of GABARAP. Ribbon (A) and surface diagrams (B) of GABARAP
are shown. Residues forming hp1 and hp2 are colored red and
blue, respectively.
J. Mohrlu
¨der et al. Structure and function of GABARAP
FEBS Journal 276 (2009) 4989–5005 ª2009 The Authors Journal compilation ª2009 FEBS 4993
