Báo cáo khoa học: ˚ Crystal structure at 3 A of mistletoe lectin I, a dimeric type-II ribosome-inactivating protein, complexed with galactose

doi:10.1046/j.1432-1033.2003.03646.x

Eur. J. Biochem. 270, 2739–2749 (2003) (cid:1) FEBS 2003

Crystal structure at 3 A˚ of mistletoe lectin I, a dimeric type-II ribosome-inactivating protein, complexed with galactose

Hideaki Niwa1, Alexander G. Tonevitsky2, Igor I. Agapov3, Steve Saward1, Uwe Pfu¨ ller4 and Rex A. Palmer1 1School of Crystallography, Birkbeck College, University of London, Malet Street, London WC1E 7HX, UK; 2Institute of Transplantology and Artificial Organs, Moscow, Russia; 3Institute of Genetics and Selection of Microorganisms, Moscow, Russia; 4Institut fu¨r Phytochemie, Universita¨t Witten/Herdecke, Witten, Germany

N-terminal site 2-OH forms hydrogen bonds with Asp27 and Lys41, and at the C-terminal site 3-OH and 6-OH undergo water-mediated interactions and C5 has a hydro- phobic contact. MLI is a galactose-specific lectin and shows little affinity for N-acetylgalactosamine. The reason for this is discussed. Structural differences among the RIPs investi- gated in this study (their quaternary structures, location of sugar-binding sites, and fine sugar specificities of their B-chains, which could have diverged through evolution from a two-domain protein) may affect the binding sites, and consequently the cellular transport processes and biological responses of these toxins.

Keywords: lectin; mistletoe (Viscum album); ribosome-inac- tivating protein; b)trefoil.

The X-ray structure of mistletoe lectin I (MLI), a type-II ribosome-inactivating protein (RIP), cocrystallized with galactose is described. The model was refined at 3.0 A˚ resolution to an R-factor of 19.9% using 21 899 reflections, with Rfree 24.0%. MLI forms a homodimer (A–B)2 in the crystal, as it does in solution at high concentration. The dimer is formed through contacts between the N-terminal domains of two B-chains involving weak polar and non- polar interactions. Consequently, the overall arrangement of sugar-binding sites in MLI differs from those in monomeric type-II RIPs: two N-terminal sugar-binding sites are 15 A˚ apart on one side of the dimer, and two C-terminal sugar- binding sites are 87 A˚ apart on the other side. Galactose binding is achieved by common hydrogen bonds for the two binding sites via hydroxy groups 3-OH and 4-OH and hydrophobic contact by an aromatic ring. In addition, at the

In type-I RIPs there is a single polypeptide chain which shares sequence and structural homology with the A-chains of type-II RIPs. The low cytotoxicity of type-I RIPs is considered to be due to the lack of the cell internalizing facility exerted by the B-chains of type-II RIPs [2–4].

After endocytosis of a type-II RIP, it is transported to the Golgi network and then to the endoplasmic reticulum [5]. The toxin is finally translocated across the endoplasmic reticulum membrane into the cytosol to exert its catalytic action on ribosomes. Studies using MLI [6,7] indicate that the disulfide bond that links the two chains is reduced and the A-chain is unfolded before translocation into the cytosol.

The use of the mistletoe plant (Viscum album), well known in ancient religious ceremonies, is thought to have extended to more medicinal purposes since ancient times [1]. More recently, particularly since the beginning of the 20th century, extracts of the plant have been applied in the treatment of cancer, especially in continental Europe, although its efficacy in this respect is not fully understood (http:// www.cancer.gov/cancerinfo/pdq/cam/mistletoe). It is now considered that mistletoe lectin I (MLI) is the most important component in this respect. MLI is a type-II ribosome-inactivating protein (RIP) which consists of two chains [1]: a catalytic A-chain, which inactivates protein synthesis, and a lectin B-chain, which binds to carbohydrate moieties of cell surfaces, triggering the internalization of MLI into the cell. The two chains are linked by a disulfide bond. Other type-II RIPs include the plant toxins ricin from Ricinus communis and abrin from Abrus precatorius.

The A-chain inhibits protein synthesis by cleaving the N-glycosidic bond in adenosine A4324 in 28S eukaryotic rRNA by hydrolysis (EC 3.2.2.22) [8]. The structure of the rRNA loop where this adenine exists has been studied using synthetic nucleotides [9], and the location of the loop in a ribosome was identified in a 5.0-A˚ electron density map for the crystal structure of a bacterial 50S ribosome subunit [10]. How the specific adenine is recognized by an RIP, the mechanism of the catalytic action itself, and why the depurination of the single adenine arrests protein synthesis are not properly understood. It has also been demonstrated that RIPs not only release adenine from rRNA, but in vitro, both type-I and type-II RIPs release adenine from DNA, and many type-I RIPs can release adenine from poly(A) [11].

Three type-II RIPs or mistletoe lectins, MLI, MLII and MLIII, have been isolated from mistletoe extract [12], MLI

Correspondence to H. Niwa, School of Crystallography, Birkbeck College, University of London, Malet Street, London WC1E 7HX, UK. Fax: + 44 20 76316803, Tel.: + 44 20 76316800, E-mail: h.niwa@mail.cryst.bbk.ac.uk Abbreviations: MLI, MLII, MLIII, mistletoe lectin I, II, III; RIP, ribosome-inactivating protein; MLA, MLI A-chain; RTA, ricin toxin A-chain; ABA, abrin-a A-chain; MLB, MLI B-chain; RTB, ricin toxin B-chain; ABB, abrin-a B-chain. (Received 11 March 2003, revised 27 April 2003, accepted 30 April 2003)

2740 H. Niwa et al. (Eur. J. Biochem. 270)

(cid:1) FEBS 2003

on station 7.2, with wavelength 1.488 A˚ and a 30-cm Mar Research image plate detector. Before being mounted on a loop, the crystal was soaked in a cryoprotectant solution containing about 30% glycerol and was flash cooled to 100 K for data collection. Each image was collected with a 1.5 (cid:2) step and a 900-s exposure time. A total of 16 images for the sweep range 24.0 (cid:2) were collected.

being the most abundant. MLI exists as a noncovalently associated dimer (A–B)2 in high concentration [13,14] while ricin, abrin, MLII and MLIII are monomeric toxins. The importance of the fact that MLI is cytotoxic as a dimeric type-II RIP has been emphasized, as it is known that some type-II RIPs exist as dimers or tetramers and are not cytotoxic to a whole cell system, even though their A-chains can inhibit protein synthesis in a cell-free system [15]. These include dimeric R. communis agglutinin [16], an isolectin of ricin, and Sambucus nigra agglutinin V [17] from the elder plant, and also a tetrameric S. nigra agglutinin I [18].

Data processing was carried out with DENZO and SCALE- PACK [28], and subsequently with SCALA in the CCP4 suite [29]. The space group was P6522 as previously determined [24], with unit cell parameters of a ¼ b ¼ 107.65 A˚ , c ¼ 311.92 A˚ . The solvent content was calculated to be 71% for one molecule of MLI with molecular mass 62 kDa per asymmetric unit [30]. The data were processed to 2.9 A˚ , providing 68 126 measured reflections, which reduced to 24 175 unique reflections. The overall completeness and multiplicity (redundancy) to 2.9 A˚ were 98.0% and 2.8, respectively, and the overall R-merge was 6.9%.

Type-II RIPs are usually Gal/GalNAc-specific lectins, S. nigra agglutinin I alone having specific affinity for terminal sialic acid sequences [19]. MLI also has some affinity for terminal sialic acid sequences, which is another major difference from ricin [20]. Among Gal/GalNAc lectins, affinity for Gal and GalNAc differ. MLI is Gal specific, showing little affinity for GalNAc [21], whereas MLIII is GalNAc specific and MLII has similar affinity for both [12]. Whereas ricin shows similar affinity for both Gal and GalNAc, R. communis agglutinin exhibits little affinity for GalNAc [22]. S. nigra agglutinin V has higher affinity for GalNAc than Gal [17]. Why these differences occur in homologous proteins or lectins is a subject for further study.

Molecular replacement was performed with AMORE [31] using the previous partial 3.7-A˚ model [24], and the more complete ricin model [32] (pdb 2aai) was used independ- ently for checking purposes. The structure was refined using X-PLOR [33] and in the later stages the CNS packages [34], with manual intervention on graphics using O [35]. Refinement was carried out using data up to 3.0 A˚ . (The R-merge between 3.12 A˚ and 3.0 A˚ was 28.1%.) There was a total of 21 899 reflections, of which 1117 (5%) were kept separate to calculate Rfree. All reflections were used for the refinement with a bulk-solvent correction proce- dure. Individual isotropic B-factors were subsequently employed. This reduced the R-factor and Rfree by 2.1% and 1.9%, respectively, without introducing any unrea- sonable B values. An anisotropic overall B-factor was also refined.

Previous papers have described the crystallization of MLI [23] and subsequently the structure of MLI at 3.7 A˚ , showing it to be a dimer in the solid state [24]. In this paper, we present the structure of MLI with noncovalently bound galactose. The present structure, including bound galactose molecules, was refined at 3.0 A˚ . The sugar- binding sites are clearly defined and are discussed in detail. The biological importance of MLI as a dimeric cytotoxic type-II RIP is also discussed and details of the dimer interface are analysed. The availability of the known structure of MLI without a specific sugar [25] (pdb 1ce7, 2mll) has enabled us to make a comparison with the present structure.

Weak diffraction is often associated with high solvent content and a large unit cell. However, Krauspenhaar et al. [26] recently reported the structure of MLI with adenine bound in the A-chain active site. These crystals, grown in a microgravity environment, exhibited improved diffraction quality over other MLI complexes. The method of crystal- lization may have contributed to the improvement in diffracting power.

Materials and methods

The C-terminal residues after Gly248 of the A-chain could not be located because the electron density was too weak, presumably because of disorder. A total of six glycosylating sugar units was included in the final model as follows. For the four putative glycosylation sites of type Asn-X-Thr/Ser in MLI (Asn112 in the A-chain, Asn61, Asn96 and Asn136 in the B-chain), it was possible to model the first glycosylating sugar as GlcNAc to all four sites and in addition a second GlcNAc at Asn96 and Asn136 in the B-chain. A model including four water molecules was refined to an R-factor 20.6% with Rfree 25.8% [36]. This model was further refined to include 47 water molecules, which were selected from the peaks above 3.5 r in the Fo–Fc map by examining the geometry and the electron density. The final R-factor and Rfree were 19.9% and 24.0%, respectively. Ramachandran plots were calculated using PROCHECK [37]. For nonglycine and nonproline residues, 87.6% were in the most favoured regions, and 12.4% were in additional or generously allowed regions.

Extraction and purification of MLI has been described previously [7]. MLI crystals were obtained by the hanging drop method. The protein solution at 18 mgÆmL)1 concen- tration contained 0.1 M galactose and 0.01 M acetate buffer at pH 4.0. The reservoir solution contained 0.9 M ammo- nium sulfate and 0.1 M glycine buffer at pH 3.4. The droplet consisted of 1 lL of the protein solution and 1 lL of the reservoir solution. Hexagonal crystals grew to about 0.2 mm in a few weeks. Sequence data were obtained as described previously [24] and in [27].

The X-ray data were collected at the Synchrotron Radiation Source (SRS) at the Central Laboratory of the Research Councils (CLRC) Daresbury Laboratory, UK,

Other important refinement statistics are summarized in Table 1. Hydrogen bonds and hydrophobic contacts between sugar molecules and protein, and between dimer molecules were analysed using HBPLUS [38]. Figures 1 and 6–8 were drawn with MOLSCRIPT [39], and Figs 4 and 5 were drawn with SETOR [40]. The refined coordinates and structure factors have been deposited at PDB with the accession code 1OQL.

Mistletoe lectin I with galactose (Eur. J. Biochem. 270) 2741

(cid:1) FEBS 2003

Table 1. Refinement statistics.

Parameter Value

50.0–3.0 21899, 1117 Resolution (A˚ ) No. of reflections (total, test set) No. of atoms

(pdb 1abr), and when the current structure was super- imposed on to the other type-II RIPs, the number of matched Ca atoms for MLI and ricin was 469 (91.8% out of 511) with rmsd 0.92 A˚ , and 479 (93.7%) with rmsd 1.05 A˚ for MLI and abrin-a, using a 3.0-A˚ cut-off distance. As differences exist between the current and already published sugar-free MLI [25] in terms of sequence and detailed structure as described below, a structure-based sequence alignment of the A-chains of MLI (MLA), ricin (RTA) and abrin-a (ABA), together with a type-I RIP momordin [42] (pdb 1mom) is shown in Fig. 2.

Total, proteins Covalently bound sugars Ligands (galactose) Solvent (water)

4132, 3977 (511 residues) 84 (6 GlcNAc) 24 (2 Gal) 47 19.9 24.0 R (%) Rfree (%) Average B-factors (A˚ 2)

All atoms A-chain B-chain Covalently bound sugars Ligands (galactose) Solvent (water) 40.8 45.2 35.7 60.3 65.0 30.0

0.011 1.72 Rmsd from ideal values Bond lengths (A˚ ) Bond angles ((cid:2))

Results

Although the structure of MLA determined here clearly has some features in common with other RIPs, important differences do occur, for example in solvent-exposed regions, particularly in the region running from strand e to helix B (residues 91–100 of MLA). In this region MLA and ABA are structurally similar, whereas in RTA and momordin this region is more extensive, with additional amino-acid residues and consequent differences in the structures. This region has been proposed [43] as being an antigenic epitope site in RIPs, and, in fact, this is the site where a monoclonal antibody against MLA has been found to bind [7]. Another notably labile region is the sheet g–h. Structural lability in this region is commonly observed in both type-I RIPs and the A-chains of type-II RIPs, and is the region in the present structure with the most consistently high B-factors.

Overview and the A-chain

Figure 1 shows a ribbon diagram of the structure of MLI. The overall structure of the MLI monomer is similar to two other type-II RIPs, ricin [32] (pdb 2aai) and abrin-a [41]

The active site is located in a prominent, centrally located cleft of the A-chain. Six residues conserved among all RIPs (Fig. 2; Tyr76, Tyr115, Glu165, Arg168, Trp199 and Ser203) are located here in a structurally highly conserved hydrogen-bond network. Tyr76 alone in this region exhibits various conformations in known RIP structures [42], and its conformational change is the main difference in active-site geometry compared with the A-chain of ricin, as reported in sugar-free MLI [25]. In MLA, there is a glycosylation site at Asn112 at the edge of the active-site cleft, which is discussed further in the Discussion section.

B-chain

The B-chain of MLI (MLB) consists of two homologous globular domains. Each domain has a diameter of (cid:2) 30 A˚ and consists of three repetitive subdomains, which form a pseudo-threefold symmetry around a hydrophobic core. The fold of one such domain has been classified as the b-trefoil fold [44]. Sequence alignment of the six subdomains with those of ricin (RTB) and abrin-a (ABB) is shown in Fig. 3. Three B-chains can be aligned without insertion or deletion except for the N-terminal region. It should be noted that the assignment of subdomains here differs from that described for RTB [45]: in the RTB description, 1a1 (the suffix indicates a strand number in a subdomain shown in Fig. 3) was not included in the repetitive subdomain, and one unit consists of, for the first one for example, 1a2, 1a3, 1a4 and 1b1, if the strand designation in Fig. 3 is used, resulting in 1c and 2c units being one strand short. Three disulfide bonds are conserved in MLB, but one in the 1a subdomain is lost as the result of a mutation from Cys to Ser39. There are two glycosylation sites in RTB and ABB, which are conserved in MLB. In addition, MLB has

Fig. 1. Ribbon representation of the structure of MLI. The A-chain is located above and the B-chain is below. The disulfide bond between the two chains is shown in yellow. The glycosylating sugars included in the final structure are shown in brown. Galactose molecules are depicted in ball-and-stick. Tyr76 and Tyr115 in the active site of the A-chain, and Asp23 in the N-terminal sugar-binding site of the B-chain are shown in red.

2742 H. Niwa et al. (Eur. J. Biochem. 270)

(cid:1) FEBS 2003

another glycosylation site at Asn61. The structure of one subdomain may be summarized as: strand 1 fi turn fi b-sheet (strands 2 + 3) fi W-loop (including a 310-helical region at the end) fi strand 4. Strands 1 and 4 also form a b-sheet. An interesting feature of the fold is provided by the repetitive Gln-X-Trp (QxW) sequences (Fig. 3) [46].

Sugar-binding sites

actions in RTB are also retained in these (and only these) two subdomains in MLB (Fig. 3). Clear electron density for binding sugar molecules was observed at each site (Fig. 4). b-Galactose was modelled into both sites, the electron density being complete for the C-terminal site, but lacking for 1-OH at the N-terminal site. The electron density for the side chain of Lys254 in the C-terminal site was also weak and tenuous, indicating disorder. Water molecules bound to galactose, three in the C-terminal site, were included in an attempt to complete all possible sugar-binding interactions. Each sugar-binding site exists in a shallow cleft formed by

Sugar-binding sites in RTB exist in the subdomains 1a and 2c [45], and the residues involved in protein–sugar inter-

Fig. 2. Sequence alignment of MLA, RTA, ABA and momordin (MOM). The secondary-structure designation follows that of ricin [32]. The C-terminal residues not included in the refined structure are indicated in lower case. Identically conserved residues among these four proteins are shown in bold: those that are also identically conserved among all RIPs are darkly shaded, and those highly conserved are lightly shaded. Conserved residues in RIPs may be obtained from the Pfam database. Possible glycosylation sites are underlined.

Mistletoe lectin I with galactose (Eur. J. Biochem. 270) 2743

(cid:1) FEBS 2003

Fig. 3. Sequence alignment of six subdomains in MLB, RTB and ABB. The strand regions are indicated by horizontally striped rectangles with the strand number in the top row, and the common 310-helical (kink) regions are indicated by obliquely striped rectangles. Identically conserved Ile and Trp are shown in dark shading and the conserved hydrophobic residues in light shading. Cysteines that make disulfide bonds are in yellow, the key residues involved in sugar binding are in magenta and possible glycosylation sites are marked in cyan. Repetitive QxW sequences are underlined.

contiguous stretches in the protein chain with some 20 residues in one subdomain, from strand 2, a 310-helical kink, strand 3 and an W-loop with a 310-helical kink at the end, as indicated in Figs 3 and 5.

provides a stereoview of the N-terminal sugar-binding site in MLB. Protein–galactose interactions that are common in both the N-terminal and C-terminal sugar-binding sites of RTB [45] are also common to the two sites of MLB. These are: (a) via hydroxy groups 3-OH and 4-OH of galactose, donating hydrogen bonds to Od2 and Od1 of aspartate (Asp23 in the N-terminal site, Asp235 in the C-terminal

Possible interactions between MLB and galactose are summarized in Fig. 5 together with those between RTB and lactose using the same criteria for comparison, and Fig. 6

Fig. 4. Stereoview of an electron density map of the C-terminal MLB sugar-binding site. Cyan: 2Fo–Fc + 1.25 r; blue: Fo–Fc + 3.0 r; red: Fo–Fc ) 3.0 r. Purple, Fo–Fc + 3.0 r, was calculated by deleting coordinates of galactose and water molecules.

2744 H. Niwa et al. (Eur. J. Biochem. 270)

(cid:1) FEBS 2003

Fig. 5. Schematic drawings of the sugar-binding sites of MLB and RTB. (A) MLB N-terminal site. (B) MLB C-terminal site. (C) RTB N-terminal site. (D) RTB C-terminal site. Key residues, hydrogen bonds that are formed with bound sugar, and secondary-structure elements are shown.

Fig. 6. Stereoview of the N-terminal MLB sugar-binding site showing a hydrogen-bond network.

Mistletoe lectin I with galactose (Eur. J. Biochem. 270) 2745

(cid:1) FEBS 2003

binding sites (Fig. 3). A hydrogen bond that corresponds to the one from the valine N to asparagine O is correspond- ingly observed in any subdomain of the RIP B-chain.

Dimer structure

site); (b) Nd2 of asparagine (Asn47, Asn256) donating a hydrogen bond to 3-OH of galactose; (c) an aromatic ring (Trp38, Tyr249) making hydrophobic contact by stacking its ring approximately parallel to the C3–C4–C5 plane of galactose. These aspartates, asparagines and aromatic rings are identically conserved residues in all B-chain sugar- binding sites.

In the MLI dimer, shown in Fig. 7, two A–B monomers are related by crystallographic twofold symmetry and face each other at the N-terminal domain of the B-chain, where the three hairpin loops in the b-trefoil fold make major contacts with those of the other in the following way: a « c¢, b « b¢, c « a¢ [24]. When viewed perpendicularly to the twofold axis, two oblong-shaped MLB molecules are seen to make an angle of (cid:2) 160 (cid:2). The dimensions of the dimer are 157 · 63 · 48 A˚ 3, and the contact area is 755 A˚ 2 per monomer. The existence of a dimer structure is consistent with an earlier electron microscope study of MLI [14], which corroborates the idea that this is a real dimer form, not an artefact of the crystal packing. The distances between sugar- binding sites in one dimer, calculated as straight distances between corresponding O4 atoms of galactose, are shown in Fig. 7.

A number of both polar and nonpolar contacts are observed at the dimer interface, as shown in Fig. 8. Residues in the 1b subdomain (the side chain of Tyr68, the main- chain atoms of Ala72 and Gly73, and Val74) make hydrophobic contact with the equivalent region in the other molecule. In addition, Og of Tyr68 makes a hydrogen bond with the side chain of Gln122 in the 1c¢ subdomain. Several hydrogen bonds are formed between residues in the 1a and 1c¢ subdomains, some of which are water mediated. In addition, Ile114 makes hydrophobic contacts with carbon atoms in the side chains of Arg25 and Asn26. Among the residues that make polar interactions, Gln34, which makes hydrogen bonds with the main-chain atoms of Thr118, and Ser111, which makes water-mediated hydrogen bonds with the O of Gly32, are not conserved in RTB and ABB. However, Oc of Ser111 is located in a similar position to Oc1 of Thr110 of RTB. Among the hydrophobic contacts, Ilel14 of MLB is mutated to an asparagine in RTB and a serine in ABB.

Comparison of MLI structures

In addition to these common interactions, specific contacts are made in each site, as described below in comparison with RTB. In the N-terminal site, Nf of Lys41 interacts with 2-OH and 3-OH of galactose, and 2-OH donates a hydrogen bond to Od2 of Asp27. In RTB, although the lysine is conserved (Lys40), because of the distance between the probable hydrogen of Nf and 2-OH or 3-OH being slightly greater than the criterion used (2.5 A˚ ), it was not selected as a hydrogen bond here. Asp27 in MLB corresponds to Gly26 in RTB, which consequently does not form a hydrogen bond. In RTB, 6-OH forms hydrogen bonds with Ne2 of Gln35 and a main-chain N, whereas in MLB 6-OH is not involved in interactions with the protein. The glutamine is conserved in the two proteins, however, and differences in orientation of the CH2OH groups in galactose are responsible for the changes in these interac- tions. The main-chain N of Asp26 in MLB donates a hydrogen bond to 4-OH, but in RTB a bond between N of the corresponding residue (Asp25) and 6-OH is probable, as mentioned above. In the C-terminal site of MLB, a water- bridged hydrogen bond is made between 3-OH of galactose and Oc1 of Thr252, which is equivalent to Lys41 at the N-terminal site. In RTB, Thr252 is mutated to His251, of which Ne2 directly donates a hydrogen bond to 3-OH. In MLB, positive electron density was observed extending from 6-OH of galactose to Arg245, where two hydrogen- bonding water molecules were located (Fig. 4). Ile247, which is equivalent in position to Gln36 of the N-terminal site, makes hydrophobic contact with C6 of galactose, as is also observed in RTB. The main-chain N of Gln238 donates a hydrogen bond to 4-OH, as equivalent interactions are observed in the N-terminal site of MLB (Asp26–4-OH, as described above) and in the C-terminal site of RTB (Ala237–4-OH). Although specific interactions with bound sugar are not established, Asp26 in MLB with galactose undergoes conformational change (v1 ¼ )164(cid:2)) as com- pared with Asp25 in RTB with lactose (v1 ¼ )65(cid:2)).

Hydrogen-bond networks of the sugar-binding sites

(Fig. 6). The Ne2 and Oe2 of

Two MLI structures are currently available in the PDB, namely 1ce7 and 2mll [25]. The coordinate sets for these are identical and will be referred to only as 2MLL, which is MLI without lectin-bound sugar. 2MLL was refined at 2.7 A˚ to an R-factor ¼ 25.1% and Rfree ¼ 31.9%, accord- ing to the PDB file. The published model comprises 241 residues in the A-chain, 255 residues in the B-chain, 3 glycosylating GlcNAc sugars, and 215 water molecules. Excluding the C-terminus of the A-chain, where residues could not be located in either the current structure or 2MLL, and the N-terminus of the B-chain, where the structure is not defined well in the current structure and residues were not located in 2MLL, the numbers of differences in sequence between the two structures are 39 in the A-chain and 31 in the B-chain. These differences arise from: (a) differences in original sequences determined by two methods [27,47]; (b) the fact that 2MLL contains truncated residues in solvent-exposed loops and also a total

Oc2 of the identically conserved aspartate (Asp23, Asp235), which accepts a hydrogen bond from 3-OH of galactose, accepts another hydrogen bond from Ne2 of glutamine (Gln48, Gln257) this glutamine form hydrogen bonds with the main-chain atoms in the W-loop. Nd2 of identically conserved asparagine (Asn47, Asn256), which donates a hydrogen bond to 3-OH of galactose, also donates a hydrogen bond to the main- chain O of valine (Val24, Val236), while the main-chain N of this valine donates a hydrogen bond to the main-chain O of the asparagine, thus forming a bridge between the region from strand 2 to the helical kink (left-hand side of a sugar- binding site in Figs 5 and 6) and the helical kink at the end of the binding site (right-hand side). The above valine and glutamine are identically conserved in all B-chain sugar-

2746 H. Niwa et al. (Eur. J. Biochem. 270)

(cid:1) FEBS 2003

Fig. 7. Ribbon representations of MLI dimer. (A) and (B) show orthogonal views of the whole dimer with bound galactose molecules. (C) is the top view of (A), depicting only B-chains, where in the N-terminal domains only the hairpin loops (strand 2 and 3) of three subdomains are depicted and their design- ations are shown in the same colour. Distances between galactose molecules (straight distan- ces between two O4 atoms) are also shown.

of 11 deleted residues compared with its original sequence [47].

there is no significant discrepancy in the active site, including the conformation of Tyr76 (Tyr75 in 2MLL), which differs from that of ricin.

In the B-chains, three points relating to sugar binding are: (a) in the N-terminal sugar-binding site, the locations of Oe1 and Ne2 of a glutamine (Gln36 in the current structure and Gln32 in 2MLL) are interchanged; (b) in the C-terminal site, 2MLL has deletions at Gln238 and Ala239 of the current

The A-chains of the present structure and 2MLL can be superimposed, with rmsd 0.56 A˚ for 232 matched residues, and the B-chains with 0.45 A˚ for 253 matched residues with 3.0-A˚ cut-off distance. In the A-chains, structural differ- ences exist in solvent-exposed regions, and there is a shift by one residue in the strand a at the N-terminus. However,

Fig. 8. Stereoview of the MLI dimer interface. Three hairpin loops (strand 2 and 3) of the N-terminal domains of the B-chains are shown. The dimer molecules are related by twofold symmetry and the view is along the twofold axis. Residues that make hydrophobic contacts are in darker grey.

Mistletoe lectin I with galactose (Eur. J. Biochem. 270) 2747

(cid:1) FEBS 2003

structure and, as a result, the position of Ala239 in the current structure is occupied by Asn232 in 2MLL, which corresponds to Asn240 of MLI; (c) the conformation of a lysine (Lys254 in the current structure and Lys246 in 2MLL) is different in the two structures; however, as mentioned previously the electron density of this side chain is weak in the current structure (and possibly also in 2MLL) and this point does not justify further discussion.

Discussion

As to the C-terminal site, there is a possibility that the disordered side chain of Lys254 causes steric hindrance with the N-acetyl group of GalNAc. However, because of the disorder, whether it completely hinders sugar binding is not conclusive. When a GalNAc is located at the C-terminal sugar-binding site of RTB so that its pyranose ring is superimposed on that of the galactose of bound lactose, it is found that the oxygen in the acetyl group could form a hydrogen bond with the hydroxy group of Ser238. In MLB Ser238 in RTB is mutated to Ala239 and cannot make a hydrogen bond with a sugar. It is proposed that this serine in RTB may contribute to GalNAc binding.

In the present MLI structure, the electron density of galactose in the N-terminal site of MLI was less pronounced than that in the C-terminal site. This may suggest that, of the two sites, the N-terminal site has the lower affinity for galactose, as is also the case for ricin.

The refined structure of MLI complexed with galactose is presented. The molecular structure shares common features with other RIPs without extra or shortened main-chain loops. MLA has a glycosylation site at the rim of the active- site cleft, which is unique among RIPs. On superimposition of the structure of the RTA–ApG complex [48] on to MLA, a glycosylating MLA sugar at Asn112 was seen to occupy (at least partially) the guanine-binding site of the RTA substrate analogue. However, glycosylation appears not to affect the catalytic activity because it is known that recombinant MLA shows similar activity to that of plant- derived MLA [27]. As molecular dynamics studies also suggest [48,49], it is possible that, when the RNA substrate loop binds to an A-chain of a type-II RIP, it adopts a different conformation from that of the dinucleotide substrate analogue. In fact, the conformation of the bound ApG in ricin differs from that in the structure of the ribosomal loop determined by NMR [50] (pdb 1scl) or by X-ray crystallography [9] (pdb 430d, 483d).

The two sugar-binding sites of MLB exhibit common features sometimes observed in sugar-binding sites in proteins other than RIP B-chains: an extensive hydrogen- bond network and hydrophobic stacking [51]. Sugar-affinity studies have shown that 4-OH and then 3-OH are the hydroxy groups of galactose that strongly affect sugar binding for MLI [21] and ricin [52], and they are involved in the common binding mode for the two sugar-binding sites of both proteins.

MLI crystallizes readily from ammonium sulfate in acidic conditions (pH 3.4 for the actual crystal used for the current structure analysis). Sugar affinity decreases under acidic conditions for MLI [55] and ricin [56], therefore the electron density of galactose shown in Fig. 4 may be regarded as corresponding to a partially bound sugar. The cause of this decrease in affinity is considered to be either due to protonation of the aspartate in the sugar-binding site, which can be a hydrogen-bond acceptor, or conformational change. As there is little conformational change among the crystal structures of MLI (pH 3.4), ricin (pH 4.75) [32] and abrin-a (pH 8.0) [41], the decrease in sugar affinity in acidic conditions is probably associated with protonation of the identically conserved aspartate, which interacts with 3-OH and 4-OH of bound galactose in these sugar-binding sites. The MLI dimer involves several polar and hydrophobic contacts through the N-terminal domains of two B-chains. As the formation of the dimer is concentration dependent [13,14], the overall interaction is weak. In the 1a subdomain of the B-chains of some type-II RIPs, the hairpin loop is stabilized by a disulfide bond, as in RTB and ABB. However, this S-S bond does not exist in MLB because of the mutation from a cysteine to Ser40, and the loop is more flexible. Therefore, it may be speculated that the increased flexibility caused by the loss of the disulfide bond in MLB may play a crucial role in dimer formation [24]. As little shift of the 1a loop is observed in the superimposition of MLB, RTB and ABB, it appears that at least permanent dislocation of the loop is not a requisite for dimer formation. It is not possible to conclude, however, from this study alone how important each of the various interactions described in the Results section is individually for the dimer formation. Although hydrophobic 1b hairpin loops do in fact contact each other, this region is also hydrophobic in monomeric RTB and ABB, as seen in the sequence in Fig. 3, therefore this hydrophobic contact alone is not enough to cause MLI dimer formation. Among the residues that make interactions, Gln34 and Ile114 are unique to MLB. Mutation studies of these residues may reveal their role in the dimer formation.

It is known that GalNAc binds only to the C-terminal sugar-binding site of ricin, which is the high-affinity site [53]. However, experimental results on the difference in sugar specificity between the two binding sites are not available for MLB. In interpreting sugar-binding specificity assays of lectins, consideration should be paid to the actual number of binding sites in each protein. The two sugar-binding sites in type-II RIP B-chains are structurally and chemically very similar, but not identical. As a dimer, MLI in fact possesses four sugar-binding sites, and this property, in view of the novel interbinding-site distances resulting from this dimeri- zation, is likely to affect the toxin’s ability to bind to cell surface sugars. Ambiguity exists when the sugar specificity of a single binding site is examined using biochemical data that are unavoidably from all of the available binding sites. With respect to the specificity of MLI for GalNAc, however, evidence derived from solid-phase assay [54] indicates that each binding site in fact does not have high GalNAc specificity.

In contrast with some type-II RIPs mentioned in the Introduction, the dimeric structure of MLI does not interfere with its toxic activity. It may be, however, that the quaternary structure does affect other aspects of the biological processes. One factor that may be involved is the relative disposition of the sugar-binding positions.

There is insufficient space for the N-acetyl group of GalNAc to be accommodated in the N-terminal site of MLB, because it is blocked by the two residues that form hydrogen bonds with 2-OH, namely Asp27 and Lys41.

2748 H. Niwa et al. (Eur. J. Biochem. 270)

(cid:1) FEBS 2003

5. Sandvig, K. & van Deurs, B. (1996) Endocytosis, intracellular transport, and cytotoxic action of Shiga toxin and ricin. Physiol. Rev. 76, 949–966.

The distance between two sugar-binding sites in a monomer is 47 A˚ , whereas in the MLI dimer two N-terminal sugar- binding sites are 15 A˚ apart on one side of the dimer and two C-terminal sites are 87 A˚ apart on the other side of the dimer (Fig. 7). All sugar-binding sites are centrally located in the MLI dimer. Unless some quaternary structural change occurs, the dimer must bind to the cell surface by laying the whole (A–B)(B–A) structure along the cell surface so that the binding sites are close enough to cell-surface oligosaccharides.

6. Agapov, I.I., Tonevitsky, A.G., Moysenovich, M.M., Mal- uchenko, N.V., Weyhenmeyer, R. & Kirpichnikov, M.P. (1999) Mistletoe lectin dissociates into catalytic and binding subunits before translocation across the membrane to the cytoplasm. FEBS Lett. 452, 211–214.

7. Agapov, I.I., Tonevitsky, A.G., Maluchenko, N.V., Moysenovich, M.M., Bulah, Y.S. & Kirpichnikov, M.P. (1999) Mistletoe lectin A-chain unfolds during the intracellular transport. FEBS Lett. 464, 63–66. 8. Endo, Y. & Tsurugi, K. (1987) RNA N-glycosidase activity of ricin A-chain. J. Biol. Chem. 262, 8128–8130.

9. Correll, C.C., Wool, I.G. & Munishkin, A. (1999) The two faces of the Escherichia coli 23S rRNA sarcin/ricin domain: the structure at 1.11 A˚ resolution. J. Mol. Biol. 292, 275–287.

10. Ban, N., Nissen, P., Hansen, J., Capel, M., Moore, P.B. & Steitz, T.A. (1999) Placement of protein and RNA structures into a 5 A˚ - resolution map of the 50S ribosomal subunit. Nature (London) 400, 841–847.

As there is no significant difference in catalytic activity between recombinant MLA and plant-origin MLA, it may be argued that differences in cytotoxicity between MLI and ricin are mainly influenced by the B-chains [27]. The actual binding sites of the toxins on cells, subsequent transport processes, and biological responses are probably affected by differences in fine sugar specificity and/or quaternary structure with associated differences in the arrangement of the sugar-binding sites. Recent studies highlight differences in the membrane-binding sites of MLI and ricin [57]. It has been suggested [58] that ricin may in fact dimerize on binding to cell surface receptors. However, the study was carried out on ricin with chemically blocked sugar-binding sites, and the actual dimer form was not specified.

11. Barbieri, L., Valbonesi, P., Bonora, E., Gorini, P., Bolognesi, A. & Stirpe, F. (1997) Polynucleotide: adenosine glycosidase activity of ribosome-inactivating proteins: effect on DNA, RNA and poly (A). Nucleic Acids Res. 25, 518–522.

12. Franz, H., Ziska, P. & Kindt, A. (1981) Isolation and properties of three lectins from mistletoe (Viscum album L.). Biochem. J. 195, 481–484.

13. Olsnes, S., Stirpe, F., Sandvig, K. & Pilh, A. (1982) Isolation and characterization of viscumin, a toxic lectin from Viscum album L. (mistletoe). J. Biol. Chem. 257, 13263–13270.

14. Lutsch, G., Noll, F., Ziska, P., Kindt, A. & Franz, H. (1984) Electron microscopic investigations on the structure of lectin I from Viscum album L. FEBS Lett. 170, 335–338.

15. Citores, L., Ferreras, J.M., Iglesias, R., Carbajales, M.L., Arias, F.J., Jime´ nez, P., Rojo, M.A. & Girbe´ s, T. (1993) Molecular mechanism of inhibition of mammalian protein synthesis by some four-chain agglutinins: proposal of an extended classification of plant ribosome-inactivating proteins (rRNA N-glycosidases). FEBS Lett. 329, 59–62.

There is evidence to suggest that RIP B-chains evolved from a primordial peptide of about 40 residues: (a) by gene duplications and fusions into a three-subdomain protein (a b-trefoil domain); and (b) by further duplication into a two-domain protein [59]. The fact that some bacteria have proteins with a b-trefoil domain sharing sequence and structural homology with RIP B-chains [60] suggests that b-trefoil fold proteins existed in early evolutionary proces- ses. When the six subdomains of MLB are aligned, it is seen that the sequence identity between subdomains and also between two domains is about 20%. However, overall sequence identity between MLB, RTB and ABB is more than 50% [27]. Hence, it is reasonable to speculate that RIP B-chains diverged evolutionarily from a precursor protein that already had a two-domain structure.

16. Sweeney, E.C., Tonevitsky, A.G., Temiakov, D.E., Agapov, I.I., Saward, S. & Palmer, R.A. (1997) Preliminary crystallographic characterization of ricin agglutinin. Proteins 28, 586–589.

Acknowledgements

17. Van Damme, E.J.M., Barre, A., Rouge´ , P., Van Leuven, F. & Peumans, W.J. (1996) Characterization and molecular cloning of Sambucus nigra agglutinin V (nigrin b), a GalNAc-specific type-2 ribosome-inactivating protein from the bark of elderberry (Sambucus nigra). Eur. J. Biochem. 237, 505–513.

We thank members of the School of Crystallography at Birkbeck College for their valuable discussions, and colleagues at Daresbury Laboratory, UK for their help and support during the course of this work.

18. Van Damme, E.J.M., Barre, A., Rouge´ , P., Van Leuven, F. & Peumans, W.J. (1996) The NeuAc (a2,6) Gal/GalNAc-binding lectin from elderberry (Sambucus nigra) bark, a type-2 ribosome- inactivating protein with an unusual specificity and structure. Eur. J. Biochem. 235, 128–137.

References

1. Franz, H. (1985) Inhaltsstoffe der Mistel (Viscum album L.) als potentielle Arzneimittel. Pharmazie 40, 97–104. 19. Shibuya, N., Goldstein, I.J., Broekaert, W.F., Nsimba-Lubaki, M., Peeters, B. & Peumans, W.J. (1987) The elderberry (Sambucus nigra L.) bark lectin recognizes the Neu5Ac (a2–6) Gal/GalNAc sequence. J. Biol. Chem. 262, 1596–1601. 2. Barbieri, L., Battelli, M.G. & Stirpe, F. (1993) Ribosome- inactivating protein from plants. Biochim. Biophys. Acta 1154, 237–282. 20. Wu, A.M., Song, S.-C., Hwang, P.-Y., Wu, J.H. & Pfu¨ ller, U. (1995) Interaction of mistletoe toxic lectin-I with sialoglycopro- teins. Biochem. Biophys. Res. Commun. 214, 396–402. 21. Lee, R.T., Gabius, H.-J. & Lee, Y.C. the galactose-specific toxic

(1994) The sugar- combining area of lectin of mistletoe extends beyond the terminal sugar residue: comparison with a homologous toxic lectins, ricin. Carbohydr. Res. 254, 269–276. 3. Van Damme, E.J.M., Hao, Q., Chen, Y., Barre, A., Vandenbus- sche, F., Desmyter, S., Rouge´ , P. & Peumans, W.J. (2001) Ribo- some-inactivating proteins: a family of plant proteins that do more than inactivating ribosomes. Crit. Rev. Plant Sci. 20, 395–465. 4. Peumans, W.J., Hao, Q. & Van Damme, E.J.M. (2001) Ribo- some-inactivating proteins from plants: more than RNA N-gly- cosidases? FASEB J. 15, 1493–1506.

Mistletoe lectin I with galactose (Eur. J. Biochem. 270) 2749

(cid:1) FEBS 2003

41. Tahirov, T.H., Lu, T.-H., Liaw, Y.-C., Chen, Y.-L. & Lin, J.-Y. (1995) Crystal structure of abrin-a at 2.14 A˚ . J. Mol. Biol. 250, 354–367. 22. Nicholson, G.L. & Blaustein, J. (1972) The interaction of Ricinus communis agglutinin with normal and tumor cell surface. Biochim. Biophys. Acta 266, 543–547.

42. Husain, J., Tickle, I.J. & Wood, S.P. (1994) Crystal structure of momordin, a type I ribosome inactivating protein from the seeds of Momordica charantia. FEBS Lett. 342, 154–158. 23. Sweeney, E.C., Palmer, R.A. & Pfu¨ ller, U. (1993) Crystallization of the ribosome inactivating protein MLI from Viscum album (mistletoe) complexed with b-D-galactose. J. Mol. Biol. 234, 1279–1281.

43. Lebeda, F.J. & Olson, M.A. (1999) Prediction of a conserved, neutralizing epitope in ribosome-inactivating proteins. Int. J. Biol. Macromol. 24, 19–26.

24. Sweeney, E.C., Tonevitsky, A.G., Palmer, R.A., Niwa, H., Pfueller, U., Eck, J., Lentzen, H., Agapov, I.I. & Kirpichnikov, M.P. (1998) Mistletoe lectin I forms a double trefoil structure. FEBS Lett. 431, 367–370.

44. Murzin, A.G., Lesk, A.M. & Chothia, C. (1992) b-Trefoil fold: pattern of structure and sequence in the KunitZ inhibitors, inter- leukins-1b and 1a and fibroblast growth factors. J. Mol. Biol. 223, 531–543. 45. Rutenber, E. & Robertus, J.D. (1991) Structure of ricin B-chain at 2.5 A˚ resolution. Proteins 10, 260–269. 25. Krauspenhaar, R., Eschenburg, S., Perbandt, M., Kornilov, V., Konareva, N., Mikailova, I., Stoeva, S., Wacker, R., Maier, T., Singh, T., Mikhailov, A., Voelter, W. & Betzel, C. (1999) Crystal structure of mistletoe lectin I from Viscum album. Biochem. Biophys. Res. Commun. 257, 418–424. 46. Hazes, B. (1996) The (QxW)3 domain: a flexible lectin scaffold. Protein Sci. 5, 1490–1501.

47. Soler, M.H., Stoeva, S. & Voelter, W. (1998) Complete amino acid sequence of the B chain of mistletoe lectin I. Biochem. Biophys. Res. Commun. 246, 596–601.

48. Monzingo, A.F. & Robertus, J.D. (1992) X-ray analysis of sub- strate analogs in the ricin A-chain active site. J. Mol. Biol. 227, 1136–1145.

26. Krauspenhaar, R., Rypniewski, W., Kalkura, N., Moore, K., DeLucas, L., Stoeva, S., Mikhailov, A., Voelter, W. & Betzel, Ch. (2002) Crystallisation under microgravity of mistletoe lectin I from Viscum album with adenine monophosphate and the crystal structure at 1.9 A˚ resolution. Acta Crystallogr. D58, 1704–1707. 27. Eck, J., Langer, M., Mo¨ ckel, B., Baur, A., Rothe, M., Zinke, H. & Lentzen, H. (1999) Cloning of the mistletoe lectin gene and characterization of the recombinant A-chain. Eur. J. Biochem. 264, 775–784. 49. Olson, M.A. (1997) Ricin A-chain structural determination for binding substrate analogues: a molecular dynamics simulation analysis. Proteins 27, 80–95. 50. Szewczak, A.A. & Moore, P.B. (1995) The sarcin/ricin loop, a 28. Otwinowski, Z. & Minor, W. (1997) Processing of X-ray diffrac- tion data collected in oscillation mode. Methods Enzymol. 276, 307–326. modular RNA. J. Mol. Biol. 247, 81–98. 51. Quiocho, F.A. (1989) Protein–carbohydrate interactions: basic molecular features. Pure Appl. Chem. 61, 1293–1306. 29. Collaborative Computational Project, 4 (1994) The CCP4 suite: programs for protein crystallography. Acta Crystallogr. D50, 760–763. 30. Matthews, B.W. (1968) Solvent content of protein crystals. J. Mol. Biol. 33, 491–497. 31. Navaza, J. (1994) AMoRe: an automated package for molecular 52. Solı´ s, D., Ferna´ ndez, P., Dı´ az-Maurin˜ o, T., Jime´ nez-Barbero, J. & Martı´ n-Lomas, M. (1993) Hydrogen-bonding pattern of methyl b-lactoside binding to the Ricinus communis lectins. Eur. J. Biochem. 214, 677–683. replacement. Acta Crystallogr. A50, 157–163.

53. Hatakeyama, T., Yamasaki, N. & Funatsu, G. (1986) Evidence tryptophan residue in the low-affinity for involvement of saccharide binding site of ricin D. J. Biochem. (Tokyo) 99, 1049–1056. 32. Rutenber, E., Katzin, B.J., Ernst, S., Collins, E.J., Mlsna, D., Ready, M.P. & Robertus, J.D. (1991) Crystallographic refinement of ricin to 2.5 A˚ . Proteins 10, 240–250.

33. Bru¨ nger, A.T. (1992) X-PLOR, Version 3.1. A System for X-Ray Crystallography and NMR. Manual. Yale University press, New Haven, CT.

54. Galanina, O.E., Kaltner, H., Khraltsova, L.S., Bovin, N.V. & Gabius, H.-J. (1997) Further refinement of description of the ligand-binding characteristics for the galactose-binding mistletoe lectin, a plant agglutinin with immunomodulatory potency. J. Mol. Recogn. 10, 139–147.

55. Lee, R.T., Gabius, H.-J. & Lee, Y.C. (1992) Ligand binding characteristics of the major mistletoe lectin. J. Biol. Chem. 267, 23722–23727. 34. Bru¨ nger, A.T., Adams, P.D., Clore, G.M., DeLano, W.L., Gros, P., Grossekunstleve, R.W., Jiang, J.-S., Kuszewki, J., Nilges, M., Pannu, N.S., Read, R.J., Rice, L.M., Simonson, T. & Warren, G.L. (1998) Crystallography & NMR System: a new software suite for macromolecular structure determination. Acta Crystal- logr. D54, 905–921.

56. Yamasaki, N., Hatakeyama, T. & Funatsu, G. (1985) Ricin D–saccharide interaction as studied by ultraviolet difference spectroscopy. J. Biochem. (Tokyo) 98, 1555–1560.

57. Moisenovich, M., Tonevitsky, A., Agapov, I., Niwa, H., Schewe, H. & Bereiter-Hahn, J. (2002) Differences in endocytosis and intracellular sorting of ricin and viscumin in 3T3 cells. Eur. J. Cell Biol. 81, 529–538. 35. Jones, T.A., Zou, J.-Y. & Cowan, S.W. (1991) Improved methods for building protein models in electron density maps and the location of errors in these models. Acta Crystallogr. A47, 110–119. 36. Niwa, H. (2001) Crystal structure of a ribosome inactivating protein from mistletoe (Viscum album) and a homologous lectin domain from elder (Sambucus nigra). PhD Thesis, University of London, UK.

58. Venkatesh, Y.P. & Lambert, J.M. (1997) Galactose-induced dimerization of blocked ricin at acidic pH: evidence for a third galactose-binding site in ricin B-chain. Glycobiology 7, 329–335.

37. Laskowski, R.A., MacArthur, M.W., Moss, D.S. & Thornton, J.M. (1993) PROCHECK: a program to check the stereochemical quality of protein structures. J. Appl. Crystallogr. 26, 283–291. 38. McDonald, I.K. & Thornton, J.M. (1994) Satisfying hydrogen- bonding potential in proteins. J. Mol. Biol. 238, 777–793.

59. Rutenber, E., Ready, M. & Robertus, J.D. (1987) Structure and evolution of ricin B chain. Nature (London) 326, 624–626. 60. Fujimoto, Z., Kuno, A., Kaneko, S., Yoshida, S., Kobayashi, H., Kusakabe, I. & Mizuno, H. (2000) Crystal structure of Strepto- myces olivaceoviridis E-86 b-xylanase containing xylan-binding domain. J. Mol. Biol. 300, 575–585. 39. Kraulis, P.J. (1991) MOLSCRIPT: a program to produce both detailed and schematic plots of protein structures. J. Appl. Crystallogr. 24, 946–950.

40. Evans, S.V. (1993) SETOR: hardware lighted three-dimensional solid model representations of macromolecules. J. Mol. Graph. 11, 134–138.