Crystal Structure of the Complex between the Fab′ Fragment of the Cross-Neutralizing Anti-HIV-1 Antibody 2F5 and the Fab Fragment of Its Anti-idiotypic Antibody 3H6

Crystal Structure of the Complex between the Fab′ Fragment of the Cross-Neutralizing Anti-HIV-1 Antibody 2F5 and the Fab Fragment of Its Anti-idiotypic Antibody 3H6

doi:10.1016/j.jmb.2008.07.057 J. Mol. Biol. (2008) 382, 910–919 Available online at www.sciencedirect.com Crystal Structure of the Complex between ...

2MB Sizes 0 Downloads 54 Views

doi:10.1016/j.jmb.2008.07.057

J. Mol. Biol. (2008) 382, 910–919

Available online at www.sciencedirect.com

Crystal Structure of the Complex between the Fab′ Fragment of the Cross-Neutralizing Anti-HIV-1 Antibody 2F5 and the Fab Fragment of Its Anti-idiotypic Antibody 3H6 Steve Bryson 1,5 , Jean-Philippe Julien 1 , David E. Isenman 1,2 , Renate Kunert 6 , Hermann Katinger 6,7 and Emil F. Pai 1,3,4,5 ⁎ 1

Department of Biochemistry, University of Toronto, 1 King's College Circle, Toronto, Ontario, Canada M5S 1A8 2

Department of Immunology, University of Toronto, 1 King's College Circle, Toronto, Ontario, Canada M5S 1A8 3

Department of Medical Biophysics, University of Toronto, 101 College Street, Toronto, Ontario, Canada M5G 1L7 4

Department of Molecular Genetics, University of Toronto, 1 King's College Circle, Toronto, Ontario, Canada M5S 1A8

The monoclonal antibody 2F5 neutralizes a broad range of human immunodeficiency virus-1 isolates via a conserved epitope on the viral glycoprotein gp41. The conformation of the principal epitope is a type I βturn centered on gp41 residues 664DKW666; in addition, binding studies indicate that residues N- and C-terminal to this core as well as structurally more distant parts of gp41 also contribute to the interaction. Ab2/3H6 is an anti-idiotypic antibody that inhibits the interaction between 2F5 and gp41 and as such, Ab2/3H6 may, in principle, possess a paratope that mimics the gp41 epitope. To establish the potential of Ab2/3H6 to serve as a guide for the design of vaccine components against human immunodeficiency virus, we investigated the crystal structure of the heterodimeric complex of Ab2/ 3H6 Fab and 2F5 Fab′. Ab2/3H6 Fab binds to 2F5 Fab′ via a helix-like protrusion formed by residues 58(H)RYSPSLNTRL67(H) of the 2F5 Fab′ variable domain and proximal to but not overlapping with the gp41 664 DKW666 epitopebinding pocket. This defines Ab2/3H6 as an anti-idiotypic antibody of the Ab2γ class, i.e., an antigen-inhibitable idiotype that does not carry the internal image of the linear primary gp41 662ELDKWAS668 epitope. © 2008 Elsevier Ltd. All rights reserved.

5

Ontario Cancer Institute/ Princess Margaret Hospital, Division of Cancer Genomics and Proteomics, MaRS/TMDT, 101 College Street, Toronto, Ontario, Canada M5G 1L7

6

Institute of Applied Microbiology, University of Natural Resources and Applied Life Sciences, Vienna, Muthgasse 18, A-1190 Vienna, Austria

7

Polymun Scientific Immunbiologische Forschung GmbH, Nußdorfer Lände 11, A-1190 Vienna, Austria Received 25 February 2008; received in revised form 16 July 2008; accepted 18 July 2008

*Corresponding author. Department of Biochemistry, University of Toronto, 1 King's College Circle, Toronto, Ontario, Canada M5S 1A8. E-mail address: [email protected]. Abbreviations used: HIV, human immunodeficiency virus; Ab, antibody; nmAb, neutralizing monoclonal antibody; CDR, complementarity-determining region; HRP, horseradish peroxidase; TBS, Tris-buffered saline. 0022-2836/$ - see front matter © 2008 Elsevier Ltd. All rights reserved.

Crystal Structure of Ab2/3H6 Fab-2F5 FabVComplex

911

Received 25 February 2008; received in revised form 16 July 2008; accepted 18 July 2008 Available online 27 July 2008 Edited by I. Wilson

Keywords: HIV-1; gp41; antibody; anti-idiotypic; crystal structure

Introduction Vaccines represent one of our best weapons in our fight against infectious diseases. One such disease that unfortunately is quite recalcitrant to this approach, with no prophylactic and/or therapeutic vaccine known, is AIDS. Since its discovery in 1981, this disease has killed more than 40 million people worldwide.1 Attempts at creating a vaccine targeting human immunodeficiency virus (HIV) have been strongly focused on the virus' highly glycosylated envelope (Env) protein, which mediates the attachment and fusion process between HIV and its host's CD4 T lymphocytes, dentritic cells and cells of the macrophage lineage.2 Env consists of a trimer of dimers, the two non-covalently associated subunits gp120, which initiates the CD4 receptor interaction, and gp41, which is responsible for the subsequent steps of membrane fusion.3–5 A few conserved regions of Env that can elicit neutralizing antibody responses in naturally infected individuals, e.g., the CD4 binding site and the membrane-proximal region, have been the targets of several experimental vaccines. 6 Although usually large amounts of antigen-specific antibodies could be obtained, they failed to effectively neutralize HIV.7–13 A major difficulty in designing a molecule capable of eliciting neutralizing antibodies against HIV's Env protein is to correctly mimic the three-dimensional structure of the antigen as recognized in vivo. According to the network theory of immunoregulation proposed by Jerne14,15 and revised by Bona and Kohler,16 the lymphocytes' ability to recognize any molecular shape should make it possible to elicit an anti-idiotypic antibody (Ab2) that interacts with an idiotope on an anti-HIV-1 neutralizing antibody (Ab1).2 Then, if Ab2's paratope presents the same features as Ab1's idiotope, it bears the internal image of the antigen and thus could be a vaccine candidate. The type of anti-idiotypic antibody, which carries the internal image of the antigen, is termed Ab2β.17 Other categories of anti-idiotypic antibodies include Ab2α, which recognizes a nonantigen-inhibitable idiotype, and Ab2γ, which recognizes an antigen-inhibitable idiotype without carrying an internal image of the antigen.18 Internal image antibodies bear several advantages over conventional vaccine approaches. First and foremost, other vaccine candidates, such as recombinant protein antigens, do not always reproduce conformational epitopes of the native molecule, and they might differ in their degree of glycosylation.2 Antiidiotype-based strategies circumvent these short-

comings by accurately reflecting those structures within their binding sites.19–23 Furthermore, vaccines based on a monoclonal antibody are safer by avoiding exposure of patient and medical personnel to live or attenuated virus; they can also be conveniently produced on a large scale using welldeveloped technology.2 In 2002, Kunert et al. reported the generation and characterization of the murine antibody Ab2/3H6, an anti-idiotypic antibody against the monoclonal, broadly neutralizing anti-HIV-1 antibody 2F5.24 This discovery was particularly interesting because it seemed to open the possibility of using an anti-idiotypic antibody, Ab2/3H6, to elicit the production of the broadly neutralizing monoclonal antibody (nmAb) 2F5, which interacts with an epitope located at the membrane-proximal region on HIV's gp41 and thereby inhibits membrane fusion between HIV and the host cell.25–27 The monoclonal antibody Ab2/ 3H6 blocked the binding of nmAb 2F5 to its synthetic epitope, hence suggesting that Ab2/3H6 was either an Ab2β or an Ab2γ class anti-idiotypic antibody.24 A mouse/human Ab2/3H6 chimera F ab was expressed to high levels in methylotrophic yeast and in Chinese hamster ovary cells; the primary structure was determined and it was immunochemically characterized. Some key candidates for amino acids crucial for 3H6/2F5 binding were identified.28 Here, we report the crystal structure of the Fab fragment of the anti-idiotypic mouse antibody Ab2/ 3H6 in complex with the anti-HIV-1 nmAb 2F5 Fab′ fragment. Our study of the Ab2/3H6 Fab–Ab1/2F5 Fab′ immunological complex clearly shows that Ab2/3H6 belongs to the Ab2γ class with respect to the primary gp41 epitope. Its paratope does not resemble the structure of the linear 662ELDKWAS668 antigen and it does not bind to the main 2F5 paratope but still interferes with the latter's binding to the gp41 epitope.

Results Although standard purification methods, such as ion exchange, or protein A affinity chromatography are very reliable in separating Fab and Fc fragments, they failed in the case of Ab2/3H6. Anion-exchange chromatography's usual success is based on the more basic nature of Fab fragments when compared to their Fc counterparts. The rare acidic nature of Ab2/3H6 Fab was confirmed by cellulose acetate electrophoresis at pH 8.6 (Fig. 1a). Ab2/3H6 Fab had a significantly more anodal mobility than the basic 2F5 Fab′. Protein A affinity chromatography also

912

Fig. 1. Characterization of 3H6 Fab and its complex with 2F5 Fab′. (a) A cellulose acetate membrane was soaked in 90 mM sodium diethyl barbiturate buffer pH 8.6. Samples of Ab2/3H6 Fab, 2F5 Fab′, an equimolar mixture of Ab2/3H6 Fab and 2F5 Fab′ prior to crystallization, and a crystal that was washed and redissolved, were applied to the membrane. The gel was run at 100 V for 40 min before the membrane was stained with Amido Black protein stain. (b) SDS-PAGE and concanavalin A blot analysis of purified Ab2/3H6 Fab. Ab2/3H6 Fab was applied to an SDS-PAGE gel under reducing (R) and non-reducing (NR) conditions. An identical gel was blotted and probed with concanavalin A–HRP conjugate (right). The arrows point to molecules of higher molecular weight due to glycosylation.

failed to achieve separation of the Fab and Fc fragments of Ab2/3H6 because both fragments bound to the column at pH 8.0, but the binding was weak and both were eluted at pH 6.5. The approach that proved successful involved coupling an Fc-specific antibody to an Affi-Gel 10 support in order to immunodeplete the Fc fragment. Commassie-stained SDS-PAGE gels of the purified Ab2/3H6 Fab showed an additional protein band at a higher

Crystal Structure of Ab2/3H6 Fab-2F5 FabVComplex

molecular weight than the expected bands for both the light and heavy chains under reducing conditions, and the Ab2/3H6 Fab band under nonreducing conditions (Fig. 1b). Investigation of the amino acid sequence† suggested a potential glycosylation site at Asn58(H-3H6)‡ on the complementarity-determining region (CDR)-2 heavy-chain loop. The bands of higher molecular weight seen on the SDS gel were shown to be glycosylated by blotting the protein with concanavalin A–horseradish peroxidase (HRP) (Fig. 1b), confirming results obtained with the humanized form of the antibody.30,31 As the glycosylated portion of the sample was small and availability of the sample limited, crystallization trials were performed without attempting further purification. Large prism-shaped crystals (300 μm × 100 μm × 100 μm) were grown by the hanging-drop vapour diffusion method. Several hits identified in commercial screens were optimized and 0.1 M sodium citrate buffer, pH 5.0, 15–18% polyethylene glycol (PEG) 8000 and 0.2 M ammonium sulfate as precipitant solution produced the best results. Matthews coefficient analysis32 indicated two Fabsized molecules occupying the asymmetric unit. The presence in the crystal of both Ab2/3H6 Fab and 2F5 Fab′ was confirmed by applying a washed and dissolved crystal to a cellulose acetate electrophoresis membrane at pH 8.6 (Fig. 1a). Following data collection, molecular replacement techniques placed both the 2F5 Fab′ and Ab2/3H6 Fab molecules in the crystal's unit cell. In addition to the coordinates of the constant and variable domains of 2F5 Fab′, a model of the variable domain of Ab2/3H6 Fab [GenBank accession number EF512553 (heavy chain); EF512554 (light chain)28] was used in the search. The latter was generated by incorporating the coordinates of the variable Fab domain of the mouse antibody F124 [Protein Data Bank (PDB) ID 1F11], the antibody with the highest level of sequence identity to 3H6 of all antibodies represented in the RCSB PDB. Figure 2 shows the interaction between Ab2/3H6 Fab and 2F5 Fab′. Residues from all three heavy-chain CDR loops of Ab2/3H6 Fab make contact with the variable domain of the heavy chain of 2F5 Fab′, centered on residues 58(H)RYSPSLNTRL67(H). These amino acids are the core of the 2F5 epitope recognized by Ab2/3H6. They adopt a helix-like conformation at the base of the CDR-H2 loop (Fig. 3 and Supplementary Fig. 1a). This motif is composed of two consecutive constrained type I β-turns (amino acids 60 to 63 and 63 to 66, respectively) whereby the carbonyl of the first residue (i) in both β-turns forms hydrogen bonds with the peptide nitrogen atoms of the i + 3 and i + 4 residues. This helix-like motif projects from the 2F5 Fab′ protein surface, while Ab2/3H6 Fab heavy-chain residues, contributed by all CDR loops, form a cleft that † http://www.cbs.dtu.dk/services/NetNGlyc ‡ Amino acids are numbered according to Kabat et al.29

Crystal Structure of Ab2/3H6 Fab-2F5 FabVComplex

913

Fig. 2. Features of the 3H6/2F5 binding site. (a) Stereo surface-rendered representation of the complex between 3H6 Fab and 2F5 Fab′. 3H6 Fab (heavy chain, cyan; light chain, green); 2F5 Fab′ (heavy chain, purple; light chain, rose). The regions of contact between 3H6 Fab and 2F5 Fab′ are coloured blue and red, respectively. All interactions between Ab2/3H6 Fab and 2F5 Fab′ are mediated by heavy-chain residues.

complements the shape of the protruding epitope (Fig. 3). A total surface area of 1231 Å2 is buried in the formation of the complex between Ab2/3H6 Fab and 2F5 Fab′.33 The closest contacts involve the side chains of Asn64(H-2F5) and Thr65(H-2F5), residues located at the i + 1 and i + 2 positions, respectively, of the second type I β-turn. These side chains insert directly into the Ab2/3H6 Fab binding cleft (Fig. 3) and form hydrogen bonds as well as van der Waals interactions with residues of the CDR-H1 and CDRH3 loops of Ab2/3H6 Fab (Table 1). Two arginine residues, Arg58(H-2F5) and Arg66 (H-2F5), flank the 2F5 Fab′ epitope. The guanidinium group of Arg58(H-2F5) points towards the phenol ring of Ab2/3H6 Fab residue Tyr99(H-3H6), forming a π–cation interaction. The methylene carbon atoms of the Arg66(H-2F5) side chain are stacked against the phenol ring of Ab2/3H6 Fab residue Tyr53(H3H6). In addition, Tyr53(H-3H6) stacks against the 2F5 Fab′ residue Arg82B(H-2F5). In return, Ab2/3H6 Fab CDR-H3 residues Ile97(H-3H6), Gly98(H-3H6) and Tyr99(H-3H6) at the loop apex insert into a cleft at the base of the 2F5 Fab′ CDR-H2 loop. They form

hydrogen bonds and van der Waals contacts with several 2F5 Fab′ amino acids (Table 1). The formation of the crystal lattice also gives rise to a molecular interaction that can clearly be identified as a crystal artifact: Leu100A(H-2F5) and Phe100B(H-2F5) at the apex of the long CDR-H3 loop of 2F5 Fab′ insert into a hydrophobic pocket at the variable/constant domain interface of an adjacent, symmetry-related Ab2/3H6 Fab (Supplementary Fig. 2) but without influencing the contacts at the recognition interface between 2F5 Fab′ and Ab2/ 3H6 Fab.

Discussion The low pI of the Ab2/3H6 Fab fragment made its separation from the Fc fragment challenging. Alignment of the Ab2/3H6 amino acid sequence with variable κ (Vκ) residues in the IMGT database§ § http://imgt.cines.fr/

Crystal Structure of Ab2/3H6 Fab-2F5 FabVComplex

914

Fig. 3. The linear sequence 58(H)RYSPSLNTRL67(H) is the 2F5 epitope recognized by Ab2/3H6; it is part of the variable region of the 2F5 Fab′ heavy chain (purple) and assumes a helix-like motif (magenta), which protrudes from the Fab′ surface. 2F5 Fab′ residues Asn64(H-2F5) and Thr65(H-2F5) insert directly into the Ab2/3H6 Fab binding cleft (blue).

identified a germ line sequence with a high level of sequence identity.31 Allele bt20 (IGKV17–121*01) is 96.6% identical to the mature Ab2/3H6 Vk sequence. The calculated isoelectric point (pI) of this allele is 3.90, which is the lowest pI value of all the 97 functional mouse germ line Vκ sequences listed (pI range from 3.90 to 9.12). Clearly, the acidity of this germ line sequence, combined with only minor mutational changes, especially in the light chain Vregion, contributed to the low pI of Ab2/3H6 Fab. SDS-PAGE and concanavalin A–HRP blotting analysis confirmed the presence of a glycosylated and an aglycosylated species in the Ab2/3H6 Fab

samples used for crystallization (Fig. 1). Inspection of the potential N-glycosylation site Asn58(H-3H6) in the electron density map of the Ab2/3H6 Fab–2F5 Fab′ complex did not reveal any density extending beyond the side chain (data not shown) and there was no indication that the glycan would participate in the interaction with 2F5 Fab′. This confirms previous data that showed that the glycosylation status of Ab2/3H6 Fab does not change its binding affinity to 2F5.30 The contacts between Ab2/3H6 Fab and 2F5 Fab′ are formed predominantly between the heavy chains of the two molecules. Residues from all three heavy-chain CDR loops on Ab2/3H6 Fab

Table 1. Contacts between residues of 2F5 Fab′ and Ab2/3H6 Fab 2F5 Fab′ Arg 58(H) Tyr59(H) Tyr59(H) Pro61(H) Asn64(H) Thr65(H) Arg66(H) Arg82B(H)

3H6 Fab

Contact type

Contact details

Tyr99(H) Gly100(H) Gly98(H) Ile97(H) Asp31(H) Gly98(H) Gly100(H) Thr30(H) Phe33(H) Tyr53(H) Tyr53(H)

π–cation van der Waals Hydrogen bond van der Waals Hydrogen bond Hydrogen bond Hydrogen bond Hydrogen bond van der Waals Stacking Stacking

Arg guanidinium group/Tyr phenol ring Tyr δ- and ε-carbon/Gly α-carbon Tyr peptide nitrogen/Gly carbonyl oxygen Pro β- and γ-carbon/Ile δ-, γ1-, and γ2-carbon Asn δ-nitrogen/Asp carbonyl oxygen Asn δ-oxygen/Gly peptide nitrogen Asn δ-oxygen/Gly peptide nitrogen Thr γ-hydroxyl/Thr carbonyl oxygen Thr γ-carbon/Phe ε2-carbon Arg δ- and γ-carbon/Tyr δ- and ε-carbon Arg β- and γ-carbon/Tyr ζ- and ε-carbon

Crystal Structure of Ab2/3H6 Fab-2F5 FabVComplex

interact with a single, linear ten amino acid sequence on the surface of 2F5 Fab′. The lowest B-factors are observed in residues immediately surrounding the site of contact, indicating that the binding stabilizes the respective conformations. Sequence alignments combined with inspection of a large number of Fab structures show the backbone conformation of the epitope recognized by Ab2/3H6, a protruding ten amino acid helix-like motif on the 2F5 Fab′ surface, to be a widely conserved feature in the heavy-chain variable domains of antibodies. Nevertheless, previous experiments characterized the Ab2/3H6–2F5 interaction as specific relative to the alternate antiHIV antibody 2G12 and an unspecified mouse IgG.24 Closer examination of the electron density map covering the contact area between the two fragments reveals the structural basis for this specificity. The two 2F5 Fab′ side chains Asn64(H2F5) and Thr65(H-2F5) point directly into the Ab2/ 3H6 Fab binding cleft and form several hydrogen bonds (Table 1). Arg82B(H-2F5) is located close to Thr65(H-2F5) on the surface of 2F5 and in contact with Tyr53(H-3H6), further extending the area of contact. During antibody maturation, all three of these residues have been mutated. The first two changed from amino acids quite common in these positions (Lys64 is found in 76% of all human variable domain sequences in the Kabat database and Ser65 in 22%) to those that occur only rarely; threonine and asparagine are found in 0.52% and 1.0%, respectively.34,35 At position H82B, the germ line asparagine (4.6%) changes to the even rarer Arg (2.9%). The combination of`64N-T-(X)18-R82B- seems to be unique. As described, the specific interactions between the paratope of Ab2/3H6 and its antigen are not with the CDR loops of 2F5 but with amino acids on a spatially adjacent conserved helix-like motif. This, together with the fact that Asn64(H-2F5), Thr65(H-2F5) and Arg82B(H-2F5), the key residues recognized by Ab2/3H6, are found only with low frequency in human antibodies makes it tempting to speculate that they might play a so far unrecognized role in gp41 binding. Crystal structures of 2F5 with longer gp41 peptides36–38 show that their extended Nterminal portions, although not long enough to directly interact with Asn64(H-2F5) and Thr65(H2F5), point at the helix-like motif. Moving further towards the N-terminus, the gp41 sequence enters its helical, heptad repeat part,5 which consequently could harbour a 2F5 binding site in addition to the well-characterized major 662ELDKWAS668 β-turn epitope, a scenario that would explain why extending gp41 peptides beyond the core epitope significantly improves binding to 2F5.39,40 It is obvious, however, that further studies will be required to determine the validity of such a hypothesis. Comparing the interactions between 2F5 Fab′ and the longer gp41 epitope peptides 654EKNEQELLELDKWASLW670 and 659ELLELDKWASLWN671 (PDB IDs 1TJI37 and 2P8M,38 respectively) with the contacts formed when 2F5 Fab′ is part of the anti-idiotypic complex with Ab2/3H6 Fab reveals that only one 2F5

915 Fab′ residue will bind to both the epitope peptide as well as Ab2/3H6 Fab. In the peptide complex, Arg58 (H-2F5) forms a salt bridge with the amino acid corresponding to Glu662 of gp41, while it undergoes a π–cation interaction with Tyr99(H-3H6). As the contact site between Ab2/3H6 Fab and 2F5 Fab′ lies immediately adjacent to the site of gp41 binding (Fig. 4), the anti-idiotypic antibody will prevent access of gp41 to 2F5 by simple steric repulsion, consistent with in vitro HIV-1 neutralization assays, in which Ab2/3H6 decreased the potency of 2F5.24 A comparison of the entire variable light chain sequence of Ab2/3H6 to its germ line equivalent showed only three mutations, all of a conservative nature. Inspection of the structure reveals that one of these mutations, Glu50Asp(L-3H6) of the CDR-L2 loop, stabilizes the Ab2/3H6 CDR-H3 loop. The aspartate side chain forms a tight hydrogen bond with the phenol hydroxyl of the CDR-H3 apex residue Tyr99(H-3H6). The somatic mutation from glutamate to aspartate with its shorter side chain optimizes this interaction (Supplementary Fig. 3). Stabilizing Tyr99(H-3H6) is important for the π– cation interaction with 2F5 Fab′ residue Arg58(H2F5) (Table 1). In addition, fixing its conformation influences the two flanking glycine residues whose backbone atoms interact directly with the important 2F5 Fab′ epitope residue Asn64(H-2F5) (Table 1). The importance of Tyr99(H-3H6) is highlighted by the fact that its mutation to alanine, which will lead to a destabilization of the apex of the CDR-H3 loop, completely abolishes the ability of Ab2/3H6 Fab to bind to 2F5.28 As the majority of the direct contacts to 2F5 are mediated by the Ab2/3H6 CDR-H3 loop apex, the loss of binding capacity upon alanine mutation is fully consistent with our structural data. When three more residues of the Ab2/3H6 CDRH3 loop (Tyr102(H-3H6), Pro101(H-3H6) and Pro100C(H-3H6)) were mutated to alanine, only a minor additional reduction in binding capacity was observed (from 28.7% to 8.7%).28 Tyr102(H-3H6) is situated at the C-terminal base of the CDR-H3 loop and participates in a series of stacking interactions, beginning with residue Arg94(H-3H6), the Nterminal bases of the CDR-H3 loop, followed by CDR-H2 loop residues Tyr32(H-3H6) and Tyr27(H3H6), and CDR-H3 loop residue Ile97(H-3H6), which forms a direct contact with 2F5 Fab′ epitope residue Pro61(H-2F5). The adjacent amino acid Pro101(H-3H6) supports the Tyr102(H-3H6) conformation through van der Waals contacts. Finally, Pro100C(H-3H6) and its neighbour Pro100D(H-3H6) are linked by a cis-peptide bond. The Pro100C(H-3H6) side chain makes tight van der Waals contact with light-chain residue Tyr96(L-3H6), supporting the CDR-H3 loop conformation, and its partner Pro100D(H-3H6) contacts the phenol ring of the important Tyr99(H-3H6) residue, again via van der Waals interactions (Supplementary Fig. 3). Consistent with the results of binding experiments, these data suggest a supportive, but not critical, role for Tyr102(H-3H6), Pro101(H-3H6) and Pro100C(H3H6). The stability of the apex of the Ab2/3H6

916

Crystal Structure of Ab2/3H6 Fab-2F5 FabVComplex

Fig. 4. Surface rendering of the Ab2/3H6 Fab–2F5 Fab′ complex with superimposed gp41 peptide. The crystal structure of 2F5 Fab′ variable domain with bound 13-mer peptide (ELLELDKWASLNW), representing gp41 residues 659–671, was superimposed onto the 2F5 Fab′ variable domain from the Ab2/3H6 Fab–2F5 Fab′ complex to show the proximity of Ab2/ 3H6 Fab to the 2F5 Fab′ paratope. Ab2/3H6 Fab (heavy chain, cyan; light chain, green). 2F5 Fab′ (heavy chain, purple; light chain, rose). gp41 peptide (yellow). The contact regions of Ab2/3H6 Fab and 2F5 Fab′ are coloured blue and red, respectively.

CDR-H3 loop, assured via Tyr99(H-3H6), appears to be central to the antibody's binding to 2F5. In the present crystal structure, Tyr67(L-3H6) of Ab2/3H6, at the end of a non-CDR loop, comes close to the extended 2F5 CDR-H3 loop; however, the conformation of the latter is dictated by the interaction between its apex residues Leu100A(H2F5) and Phe100B(H-2F5) and an adjacent Ab2/3H6 Fab molecule. Symmetry-related, crystal packing stabilization of the CDR-H3 loop of 2F5 has also been observed in complexes with various gp41 peptides.36,37 However, when crystallized in an alternate space group, where it is not influenced by neighbours in the crystal lattice, this CDR-H3 loop does not adopt a single stable conformation.38

Conclusion The crystal structure of the complex between Ab2/ 3H6 Fab and 2F5 Fab′ was determined at 3.0 Å resolution. It demonstrates that Ab2/3H6 binds to 2F5 via a protrusion on the surface of the light chain's variable domain, a conserved helix-like loop that does

not interact with the primary 2F5 paratope, which binds the ELDKWAS peptide epitope. It is especially CDR-H3 of Ab2/3H6 that forms contacts to 2F5 via hydrogen bonds, van der Waals forces and π–cation interactions, although almost all other loops of the heavy chain and parts of the light chain of Ab2/3H6 contribute, too. Ab2/3H6 achieves high specificity for this region mainly by targeting a pair of amino acids, Asn64(H-2F5) and Thr65(H-2F5), located on the surface of the epitope protrusion. A third residue, Arg82B (H-2F5), lines up with the two amino acids, extending the interaction site between 2F5 and Ab2/3H6. There is only limited overlap, centered on Lys662 of gp41, between the parts of 2F5 recognized by Ab2/3H6 and those interacting with peptides derived from the linear gp41 epitope. Overlap, however, is sufficient to lead to steric competition between Ab2/3H6 and gp41 (or peptides representing the core ELDKWAS epitope). Together with the clear absence of structural features on Ab2/3H6 resembling the conformation assumed by the core peptide when bound to the 2F5 paratope, this characterizes Ab2/3H6 as a γ-class anti-idiotypic antibody with respect to the primary gp41 epitope centered on 664DKW666. Expanding our

Crystal Structure of Ab2/3H6 Fab-2F5 FabVComplex

structural and mechanistic knowledge of complexes between 2F5 and larger parts of gp41, preferably representing both the non-fusogenic and fusogenic states, will be necessary to better define the whole interaction interface between these two proteins.

Materials and Methods Production and purification of antibody fragments The mouse Ab2/3H6 IgG was produced and purified as published.24 It was fragmented into its Fab and Fc portions using papain. Ab2/3H6 IgG was dissolved in 50 mM phosphate buffer, pH 6.0, containing 2 mM ethylenediaminetetraacetic acid and 10 mM L-cysteine. Papain was added at a 1:100 (w/w) ratio with respect to the IgG substrate and incubated at 37 °C for 4 h. The reaction was stopped by adding 50 mM iodoacetamide. Fab and Fc fragments were separated using a rabbit anti-mouse Fcγ-specific IgG antibody from Jackson ImmunoResearch, Inc. linked to an Affi Gel-10 support from BioRad. Briefly, the Affi Gel-10 was incubated with a solution of rabbit anti-mouse Fcγ-specific IgG antibody dissolved in 0.1 M Mops buffer, pH 7.5, at 4 °C overnight. Non-reacted gel was blocked with ethanolamine. The IgGcoupled matrix was then incubated with the papain-digested mixture of 3H6 Fab and Fc at room temperature for 1 h. The unbound Ab2/3H6 Fab was collected in the flow-through fraction. An Ab2/3H6 Fab glycoform was identified by an enzyme-linked concanavalin A blotting procedure. Proteins subjected to SDS-PAGE were blotted onto a nitrocellulose membrane, the membrane was blocked with 2% Tween 20 in Tris-buffered saline (TBS), and probed with a 5-μg/ml solution of concanavalin A labeled with HRP (Sigma-Aldrich) dissolved in TBS containing 1 mM each of MgCl2, CaCl2 and MnCl2. Following a 16-h incubation at room temperature, unbound concanavalin A–HRP was washed away with TBS and bound concanavalin A–HRP was detected using standard peroxidase substrates. As the intensity of the glycoform band was low (see Fig. 1b), the protein was used in crystallization trials without further purification. 2F5 Fab′ was obtained according to Bryson et al.41

917 Table 2. Statistics for data collection and model refinement Crystal parameters Unit cell a = 67.4 Å, b = 98.3 Å, c = 154.3 Å Space group P212121 Asymmetric unit One 2F5 Fab′/3H6 Fab heterodimer Data collection Resolution range (Å) 50.0–3.00 (3.05–3.00) Redundancy 3.5 (3.5) Unique reflections 19,786 Completeness (%) 93.6 (94.2) I/σ〈I〉 19.6 (3.4) 6.4 (39.8) Rsym (%) Refinement 26.1/29.9 Rcryst/Rfree (%) RMSD bonds (Å) 0.0041 RMSD angles (°) 0.996 2 49.2 Average B-factor (Å ) (2F5) 42.9 Average B-factor (Å2) (3H6) Ramachandran statistics Most favored (%) 87.5 Allowed (%) 11.4 Generously allowed (%) 1.0 Disallowed (%) 0.1 Values in parentheses represent the highest-resolution bin.

The structure of the complex was modeled and refined using CNS and Coot.46 Rigid body, real-space, positional and B-factor refinements were used to generate a final model. Due to the limited resolution of the diffraction data, water molecules were not included in the final model. Statistics for data collection and model refinement are given in Table 2. All figures were generated using PyMOL software.47 Protein Data Bank accession code Coordinates and structure factors for the Ab2/3H6 Fab– Ab1/2F5 Fab′ complex were deposited in the RCSB PDB under access code 3BQU.

Crystallization and data collection

Acknowledgements

For crystallization, equal amounts of 2F5 Fab′ and Ab2/ 3H6 Fab were mixed to a total concentration of 8 mg/ml. The protein concentration was determined by the Bradford method.42 Crystals were grown by the hanging-drop method with a well solution containing 0.10 M sodium citrate buffer, pH 5.0, 15–18% PEG 8000 and 0.20 M ammonium sulfate. Two microliters of protein solution was mixed with 2 μl of the corresponding well solution. Before being flash-frozen in a stream of boiling nitrogen, the crystals were soaked in well solution containing 20% glycerol. Data were collected on a Rigaku FR-C rotating copper anode with Xenocs optics and a Mar Research detector (Mar345) at 110 K. They were processed using DENZO, SCALEPACK43 and XDS.44

This work was supported by a University of Toronto Research Program grant sponsored by Sanofi-Pasteur Canada, the Canada Research Chairs Program (E.F.P.) and the Fonds québécois de la recherche sur la nature et les technologies (J.-P.J.). We are grateful to Ms. A. Cunningham for the preparation and crystallization of Ab2/3H6 Fab and 2F5 Fab′.

Structure determination, model building and refinement The structure was determined by molecular replacement methods using CNS.45 Rotation and translation solutions were found using as search models 2F5 Fab′ (PDB ID 2F5B) and mouse antibody F124 Fab (PDB ID 1F11) fragments separated into variable and constant domains.

Supplementary Data Supplementary data associated with this article can be found, in the online version, at doi:10.1016/ j.jmb.2008.07.057

References 1. UNAIDS. (2006). Overview of the Global AIDS Epidemic. 2006 Report on the global AIDS epidemic.

918 2. Wilks, D. & Dalgleish, A. G. (1992). Anti-idiotypic therapeutic strategies in HIV infection. Mol. Cell. Biol. Hum. Dis. Ser. 1, 283–308. 3. Center, R. J., Leapman, R. D., Lebowitz, J., Arthur, L. O., Earl, P. L. & Moss, B. (2002). Oligomeric structure of the human immunodeficiency virus type 1 envelope protein on the virion surface. J. Virol. 76, 7863–7867. 4. Trkola, A., Dragic, T., Arthos, J., Binley, J. M., Olson, W. C., Allaway, G. P. et al. (1996). CD4-dependent, antibody-sensitive interactions between HIV-1 and its co-receptor CCR-5. Nature, 384, 184–187. 5. Chan, D. C., Fass, D., Berger, J. M. & Kim, P. S. (1997). Core structure of gp41 from the HIV envelope glycoprotein. Cell, 89, 263–273. 6. Burton, D. R., Desrosiers, R. C., Doms, R. W., Koff, W. C., Kwong, P. D., Moore, J. P. et al. (2004). HIV vaccine design and the neutralizing antibody problem. Nat. Immunol. 5, 233–236. 7. Ho, J., Uger, R. A., Zwick, M. B., Luscher, M. A., Barber, B. H. & MacDonald, K. S. (2005). Conformational constraints imposed on a pan-neutralizing HIV-1 antibody epitope result in increased antigenicity but not neutralizing response. Vaccine, 23, 1559–1573. 8. Luo, M., Yuan, F., Liu, Y., Jiang, S., Song, X., Jiang, P. et al. (2006). Induction of neutralizing antibody against human immunodeficiency virus type 1 (HIV-1) by immunization with gp41 membraneproximal external region (MPER) fused with porcine endogenous retrovirus (PERV) p15E fragment. Vaccine, 24, 435–442. 9. Zhang, H., Huang, Y., Fayad, R., Spear, G. T. & Qiao, L. (2004). Induction of mucosal and systemic neutralizing antibodies against human immunodeficiency virus type 1 (HIV-1) by oral immunization with bovine papillomavirus–HIV-1 gp41 chimeric viruslike particles. J. Virol. 78, 8342–8348. 10. Joyce, J. G., Hurni, W. M., Bogusky, M. J., Garsky, V. M., Liang, X., Citron, M. P. et al. (2002). Enhancement of alpha-helicity in the HIV-1 inhibitory peptide DP178 leads to an increased affinity for human monoclonal antibody 2F5 but does not elicit neutralizing responses in vitro. Implications for vaccine design. J. Biol. Chem. 277, 45811–45820. 11. Xiao, Y., Zhao, Y., Lu, Y. & Chen, Y. H. (2000). Epitopevaccine induces high levels of ELDKWA-epitopespecific neutralizing antibody. Immunol. Invest. 29, 41–50. 12. Coeffier, E., Clement, J. M., Cussac, V., KhodaeiBoorane, N., Jehanno, M., Rojas, M. et al. (2000). Antigenicity and immunogenicity of the HIV-1 gp41 epitope ELDKWA inserted into permissive sites of the MalE protein. Vaccine, 19, 684–693. 13. Eckhart, L., Raffelsberger, W., Ferko, B., Klima, A., Purtscher, M., Katinger, H. & Rüker, F. (1996). Immunogenic presentation of a conserved gp41 epitope of human immunodeficiency virus type 1 on recombinant surface antigen of hepatitis B virus. J. Gen. Virol. 77, 2001–2008. 14. Jerne, N. K. (1974). Towards a network theory of the immune system. Ann. Immunol. 125C, 373–389. 15. Jerne, N. K., Roland, J. & Cazenave, P. A. (1982). Recurrent idiotopes and internal images. EMBO J. 1, 243–250. 16. Bona, C. A. & Kohler, H. (1984). Anti-idiotypic antibodies and internal images in monoclonal and anti-idiotypic antibodies. In Monoclonal and Antidiotypic Antibodies: Probes for Receptor Structure and Function (Venter, J. C., Fraser, C. M. & Linstrom, J., eds), pp. 141–149, A. R. Liss, New York.

Crystal Structure of Ab2/3H6 Fab-2F5 FabVComplex 17. Dalgleish, A. G. & Kennedy, R. C. (1988). Antiidiotypic antibodies as immunogens: idiotype-based vaccines. Vaccine, 6, 215–220. 18. Zhou, E. M., Lohman, K. L. & Kennedy, R. C. (1990). Administration of noninternal image monoclonal antiidiotypic antibodies induces idiotype-restricted responses specific for human immunodeficiency virus envelope glycoprotein epitopes. Virology, 174, 9–17. 19. Chanh, T. C., Dreesman, G. R. & Kennedy, R. C. (1987). Monoclonal anti-idiotypic antibody mimics the CD4 receptor and binds human immunodeficiency virus. Proc. Natl Acad. Sci. USA, 84, 3891–3895. 20. Chang, C. C., Hernandez-Guzman, F. G., Luo, W., Wang, X., Ferrone, S. & Ghosh, D. (2005). Structural basis of antigen mimicry in a clinically relevant melanoma antigen system. J. Biol. Chem. 280, 41546–41552. 21. Mosolits, S., Campbell, F., Litvinov, S. V., Fagerberg, J., Crowe, J. S., Mellstedt, H. & Ellis, J. H. (2004). Targeting human Ep-CAM in transgenic mice by anti-idiotype and antigen based vaccines. Int. J. Cancer, 112, 669–677. 22. Stein, K. E. & Soderstrom, T. (1984). Neonatal administration of idiotype or anti-idiotype primes for protection against Escherichia coli K13 infection in mice. J. Exp. Med. 160, 1001–1011. 23. Garcia, K. C., Ronco, P. M., Verroust, P. J., Brünger, A. T. & Amzel, L. M. (1992). Three-dimensional structure of an angiotensin II–Fab complex at 3 Å: hormone recognition by an anti-idiotypic antibody. Science, 257, 502–507. 24. Kunert, R. E., Weik, R., Ferko, B., Stiegler, G. & Katinger, H. (2002). Anti-idiotypic antibody Ab2/3H6 mimics the epitope of the neutralizing anti-HIV-1 monoclonal antibody 2F5. AIDS, 16, 667–668. 25. Muster, T., Steindl, F., Purtscher, M., Trkola, A., Klima, A., Himmler, G. et al. (1993). A conserved neutralizing epitope on gp41 of human immunodeficiency virus type 1. J. Virol. 67, 6642–6647. 26. Purtscher, M., Trkola, A., Gruber, G., Buchacher, A., Predl, R., Steindl, F. et al. (1994). A broadly neutralizing human monoclonal antibody against gp41 of human immunodeficiency virus type 1. AIDS Res. Hum. Retroviruses, 10, 1651–1658. 27. Root, M. & Steger, H. K. (2004). HIV-1 gp41 as a target for viral entry inhibition. Curr. Pharm. Des. 10, 1805–1825. 28. Gach, J. S., Quendler, H., Strobach, S., Katinger, H. & Kunert, R. (2008). Structural analysis and in vivo administration of an anti-idiotypic antibody against mAb 2F5. Mol. Immunol. 45, 1027–1034. 29. Kabat, E. A., Wu, T. T., Perry, H. M., Gottesman, K. S. & Foeller, C. (1991). Sequences of proteins of immunological interest. Department of Health and Human Services, Washington, DC. 30. Gach, J. S., Maurer, M., Hahn, R., Gasser, B., Mattanovich, D., Katinger, H. & Kunert, R. (2007). High level expression of a promising anti-idiotypic antibody fragment vaccine against HIV-1 in Pichia pastoris. J. Biotechnol. 128, 735–746. 31. Gach, J. S., Quendler, H., Weik, R., Katinger, H. & Kunert, R. (2007). Partial humanization and characterization of an anti-idiotypic antibody against monoclonal antibody 2F5, a potential HIV vaccine? AIDS Res. Hum. Retroviruses, 23, 1405–1415. 32. Matthews, B. W. (1968). Solvent content of protein crystals. J. Mol. Biol. 33, 491–497. 33. Tsodikov, O. V., Record, M. T., Jr & Sergeev, Y. V. (2002). A novel computer program for fast exact

Crystal Structure of Ab2/3H6 Fab-2F5 FabVComplex

34. 35.

36.

37.

38.

39.

calculation of accessible and molecular surface areas and average surface curvature. J. Comput. Chem. 23, 600–609. Martin, A. C. R. (1996). Accessing the Kabat antibody sequence database by computer. Proteins: Struct. Funct. Genet. 25, 130–133. Zwick, M. B., Komori, H. K., Stanfield, R. L., Church, S., Wang, M., Parren, P. W. et al. (2004). The long third complementarity-determining region of the heavy chain is important in the activity of the broadly neutralizing anti-human immunodeficiency virus type 1 antibody 2F5. J. Virol. 78, 3155–3161. Pai, E. F., Klein, M. H., Chong, P. & Pedyczak, A. (2000). Fab′–epitope complex from the HIV-1 crossneutralizing monoclonal antibody 2F5. World Intellectual Property Organization patent WO-00/61618. Ofek, G., Tang, M., Sambor, A., Katinger, H., Mascola, J. R., Wyatt, R. & Kwong, P. D. (2004). Structure and mechanistic analysis of the anti-human immunodeficiency virus type 1 antibody 2F5 in complex with its gp41 epitope. J. Virol. 78, 10724–10737. Julien, J.-P., Bryson, S., Nieva, J. L. & Pai, E. F. (2008). Structural details of HIV-1 recognition by the broadly neutralizing monoclonal antibody 2F5: epitope conformation, 2F5 CDR H3 loop mobility and anion-binding site. J. Mol. Biol. Submitted for publication. Tian, Y., Ramesh, C. V., Ma, X., Naqvi, S., Patel, T., Cenizal, T. et al. (2002). Structure–affinity relationships in the gp41 ELDKWA epitope for the HIV-1 neutralizing monoclonal antibody 2F5: effects of side-chain and backbone modifications and conformational constraints. J. Pept. Res. 59, 264–276.

919 40. Menendez, A., Chow, K. C., Pan, O. C. & Scott, J. K. (2004). Human immunodeficiency virus type 1-neutralizing monoclonal antibody 2F5 is multispecific for sequences flanking the DKW core epitope. J. Mol. Biol. 338, 311–327. 41. Bryson, S., Cunningham, A., Ho, J., Hynes, R. C., Isenman, D., Barber, B. et al. (2001). Cross-neutralizing monoclonal anti-HIV-1 antibody 2F5: preparation and crystallographic analysis of free and epitope-complexed forms of Fab′. Protein Pept. Lett. 8, 413–418. 42. Bradford, M. M. (1976). A rapid and sensitive method for the quantitation of microgram quantities of protein utilizing the principle of protein–dye binding. Anal. Biochem. 72, 248–254. 43. Otwinowski, Z. & Minor, M. (1997). Processing of X-ray diffraction data collected in oscillation mode. In Methods in Enzymology, Vol. 276: Macromolecular Crystallography, Part A (Carter, C. W., Jr & Sweet, R. M., eds), pp. 307–326, Academic Press, New York. 44. Kabsch, W. (1993). Automatic processing of rotation diffraction data from crystals of initially unknown symmetry and cell constants. J. Appl. Crystallogr. 26, 795–800. 45. Brünger, A. T., Adams, P. D., Clore, G. M., DeLano, W. L., Gros, P., Grosse-Kunstleve, R. W. et al. (1998). Crystallography and NMR system (CNS): a new software system for macromolecular structure determination. Acta Crystallogr., Sect. D: Biol. Crystallogr. 54, 905–921. 46. Emsley, P. & Cowtan, K. (2004). Coot: model-building tools for molecular graphics. Acta Crystallogr., Sect. D: Biol. Crystallogr. 60, 2126–2132. 47. DeLano, W. L. (2002). The PyMOL Molecular Graphics System. DeLano Scientific, Palo Alto, CA.