|Home | About | Journals | Submit | Contact Us | Français|
The membrane-proximal external region (MPER) of the human immunodeficiency virus (HIV) envelope glycoprotein (gp41) is critical for viral fusion and infectivity and is the target of three of the five known broadly neutralizing HIV type 1 (HIV-1) antibodies, 2F5, Z13, and 4E10. Here, we report the crystal structure of the Fab fragment of Z13e1, an affinity-enhanced variant of monoclonal antibody Z13, in complex with a 12-residue peptide corresponding to the core epitope (W670NWFDITN677) at 1.8-Å resolution. The bound peptide adopts an S-shaped conformation composed of two tandem, perpendicular helical turns. This conformation differs strikingly from the α-helical structure adopted by an overlapping MPER peptide bound to 4E10. Z13e1 binds to an elbow in the MPER at the membrane interface, making relatively few interactions with conserved aromatics (Trp672 and Phe673) that are critical for 4E10 recognition. The comparison of the Z13e1 and 4E10 epitope structures reveals a conformational switch such that neutralization can occur by the recognition of the different conformations and faces of the largely amphipathic MPER. The Z13e1 structure provides significant new insights into the dynamic nature of the MPER, which likely is critical for membrane fusion, and it has significant implications for mechanisms of HIV-1 neutralization by MPER antibodies and for the design of HIV-1 immunogens.
The continued spread of human immunodeficiency virus (HIV) worldwide and, in particular, in sub-Saharan Africa, where an estimated 22 million people currently are living with HIV/AIDS, underscores the urgent need for a preventative vaccine. However, despite nearly 25 years of intense international research, a vaccine is not yet available. Passive immunization with broadly neutralizing antibodies can confer sterilizing protection against infection in animal models (4, 12, 39-41, 51, 64), providing encouragement for the development of an antibody-inducing component of an HIV type 1 (HIV-1) vaccine. Such a vaccine should elicit neutralizing antibodies with activity against the broadest range of primary circulating isolates. However, a lack of understanding of how to raise potent, cross-reactive antibodies by immunization, the so-called neutralizing antibody problem, is a major hurdle in this effort (6, 24, 72). Thus, an understanding of the structure and presentation of neutralizing epitopes on the virus and the antibodies that recognize them is vital for vaccine development.
The targets of antibody neutralization are the surface envelope (Env) glycoprotein trimers (gp120/gp41) that mediate the fusion of the viral membrane with that of the host. The majority of antibodies elicited during natural infection or immunization show limited or no cross-reactivity against diverse isolates. However, a few rare, broadly neutralizing, monoclonal antibodies have been isolated from HIV-1-infected individuals and exhibit activity against a wide range of isolates by binding to functionally conserved epitopes exposed on native gp120/gp41 trimers. These epitopes include the CD4 binding site, recognized by antibody b12, and a relatively well-conserved cluster of N-linked glycans, located on the outer domain of gp120, that is recognized by antibody 2G12 (12, 13, 71, 76). V3-directed antibodies, which are common in natural infection, also are able to sporadically neutralize across clades, as exemplified by 447-52D and F425-B4e8 (7, 16, 49, 66). The identification of three broadly neutralizing antibodies, 2F5, Z13, and 4E10, that target the conserved tryptophan-rich membrane-proximal external region (MPER) of gp41 has implicated this region as a highly promising vaccine target and has, therefore, spurred interest in its structural characterization (15, 35, 45, 47, 48, 50, 80).
The MPER plays a critical, but not fully understood, role in membrane fusion and is situated between the C-terminal heptad repeat (CHR) and the transmembrane domain (TM) of gp41 (Fig. (Fig.1).1). Following the binding of gp120 to the cell surface receptors CD4 and CXCR4/CCR5, the gp41 glycoprotein undergoes a series of conformational changes that trigger the membrane fusion activity. Notably, a relatively long-lived prehairpin intermediate of gp41 is formed, in which the coiled-coil of the N-terminal heptad repeats (NHR) extends so as to enable the fusion peptides to embed into the target membrane. In the postfusion or fusogenic state, the CHR and NHR reassemble into an antiparallel 6-helix bundle in a process that drives membrane fusion (18). The MPER contains several functionally conserved tryptophan residues that are critical for membrane fusion and viral entry, although the structural basis for their specific role has not been firmly established (22, 44, 58). Their mutation to alanine leads to the attenuation of viral infectivity, which is most pronounced for Trp666 and Trp672 (numbered according to the HXB2 isolate) (46, 58, 78). In addition, peptides based on the MPER can induce membrane leakage (68). Such membrane-disrupting properties of the MPER have been suggested to be functionally important in the expansion of the fusion pore created after receptor engagement (42, 44, 58, 68, 77).
From initial explorations using solution nuclear magnetic resonance, the structure of a 19-residue MPER peptide (residues 665 to 683) was found to be helical in dodecylphosphocholine micelles, with the hydrophobic and hydrophilic residues distributed evenly around the helix axis (62). Another study found that an MPER peptide comprising residues 659 to 671 adopts a 310-helix in water (10). More recently, the structure of an MPER peptide (residues 662 to 683) in liposomes was elucidated by a combination of nuclear magnetic resonance and spin-label electron paramagnetic resonance (69), and it was found to adopt a kinked, amphipathic structure composed of two helices connected by a short hinge (Phe673 and Asn674). Crystal structures of Fab 2F5 in complex with a 7-mer (E662LDKWAS668) and 17-mer encompassing residues 654 to 670 previously had revealed a mostly extended conformation characterized by a central β-turn involving Asp664, Lys665, and Trp666 (47, 48). This motif is the key recognition determinant for 2F5 and becomes deeply buried in the antibody combining site, suggesting that it is exposed at some stage in viral entry (45, 47, 78). The crystal structure of Fab 4E10 in complex with peptide-spanning residues W670NWFDITNW678 revealed an amphipathic α-helical structure with a narrow hydrophilic face (15). The N terminus of the 4E10 epitope forms a 310-helix that transitions into a regular α-helix at residue Asp674 and continues to Lys683, which constitutes the end of the gp41 ectodomain (14). Thus, while the structure of the MPER within functional, membrane-embedded Env trimers is not known, the observation that unconstrained peptides are able to adopt more than one defined structure suggests an inherent degree of flexibility.
Like 4E10, Z13 was identified from an HIV-1-infected individual, the former being isolated from an immortalized B-cell line and the latter from a bone marrow RNA phage display library (80). The epitope of MAb Z13 spans residues S668LWNWFDITN677, as determined by peptide mapping, scanning mutagenesis, and antibody competition studies (46, 80). This region lies between the 2F5 and 4E10 epitopes but overlaps more closely with 4E10 (Fig. (Fig.1).1). 4E10 and Z13 are both able to neutralize primary as well as laboratory-adapted isolates; nevertheless, Z13 is not as broadly neutralizing as 4E10, which has the greatest breadth of any HIV-1 antibody described to date (9). Z13e1 is an affinity-enhanced variant of Z13 and was evolved by randomizing the complementarity determining region (CDR) L3 loop sequence to identify tighter-binding mutants using phage display (46). Z13e1 displays higher affinity for both peptide and recombinant gp41 substrates, as well as increased neutralization potency, suggesting that the L3 mutations optimize binding to the linear MPER epitope. The neutralization breadth of Z13e1 is limited by the requirement for Asn671 and Asp674 in the MPER, which are approximately 71 and 58% conserved, respectively, among sequences in the Los Alamos HIV sequence database (80). Based on the clear relationship between Env trimer binding and neutralization, the neutralizing activity of Z13e1 derives from binding to a functional trimer (8, 20, 25, 43, 52, 55, 60, 73, 74). While Z13e1 and 4E10 have identical affinities for optimized linear peptides, Z13e1 is still about an order of magnitude less potent than 4E10 against a variety of primary isolates. Although the occlusion of the Z13e1 epitope on virion-associated trimers is thought to be the major limitation (46), the structural basis for the lower potency of Z13e1 relative to those of 2F5 and 4E10 is unclear.
Whereas neutralization by 4E10 depends critically on Trp672 and Phe673, Z13e1 instead requires the flanking Asn671 and Asp674 residues (46). Based on a helical model of the MPER, it was predicted that Z13e1 binds the narrow hydrophilic face that displays Asn671, Asp674, and Asn677 that is opposite that recognized by 4E10. As Z13e1 and 4E10 bind to functional trimers, both epitopes must be exposed at some stage before membrane fusion (20). To examine how Z13e1 recognizes its MPER epitope, we determined the crystal structure of Fab Z13e1 in complex with a 12-residue peptide corresponding to the core epitope with C-terminal flanking lysines to aid peptide solubility (W670NWFDITN677KKKK). The crystal structure at 1.8-Å resolution uncovers a conformation of the MPER that is distinct from that visualized in complex with 4E10. Our findings show that Z13e1 and 4E10 recognize different conformers of the MPER and reveal a novel conformational switch that is relevant for HIV-1 neutralization and membrane fusion.
Recombinant immunoglobulin G1(κ) [IgG1(κ)] for Z13e1 was overexpressed in CHO-K1 cells and purified by protein A affinity chromatography (GE Healthcare) as previously described (46). Fab Z13e1 was generated by the digestion of IgG with endoproteinase Lys-C by following a modification of the protocol for IgG 2F5 (47). In brief, IgG Z13e1 (15 mg/ml) was dialyzed against 50 mM Tris-Cl, pH 8.0, 350 mM NaCl, and treated with 100 mM dithiothreitol (1 h at 37°C). Following reduction, Z13e1 was dialyzed against 25 mM Tris-Cl, pH 8.0, followed by dialysis against 25 mM Tris-Cl, pH 8.0, 2 mM iodoacetamide (48 h at 4°C). Alkylated Z13e1 was dialyzed against 25 mM Tris-Cl, 1 mM EDTA, pH 8.5, and digested with 0.005 μg/μl sequencing-grade Lys-C (Roche Applied Sciences) for 4 h at 37°C. Digestion was stopped by the addition of 1 mM TLCK (Nα-p-tosyl-l-lysine chloromethyl ketone) and 0.4 mM leupeptin (Roche Applied Sciences). The mixture was loaded on HiTrap Protein A Fast Flow (GE Healthcare). Flowthrough fractions containing Fab were pooled and loaded on MonoS 10/30 GL (GE Healthcare) and eluted with a 0 to 0.5 M NaCl gradient in 50 mM sodium acetate, pH 5.0. Fab fractions were concentrated and further purified by size exclusion on a Superdex 200 10/300 GL (GE Healthcare) in 20 mM Tris-Cl, 200 mM NaCl, pH 8.0. The Fab peak was pooled and concentrated in size-exclusion buffer to 30 mg/ml, as determined with the Pierce bicinchoninic acid assay.
Peptide 178-1 was synthesized as previously described (11) and dissolved in 10 mM Tris-Cl, pH 8.0, to a concentration of 50 mg/ml. Crystals of Fab Z13e1 in complex with peptide were obtained by cocrystallization. Z13e1 complex with peptide 178-1 was formed by the addition of peptide at a final concentration of 5 mg/ml to 27 mg/ml Fab prior to crystallization. Initial and optimized crystallization conditions were identified by a nanoliter screening format using the IAVI/JCSG/TSRI CrystalMation robotic system. Diffraction-quality crystals were obtained by equilibrating drops of protein mixed 1:1 with 20 to 22% polyethylene glycol 3350 (PEG 3350) and 200 mM potassium fluoride (KF) in sitting-drop vapor diffusion wells at 22°C for 3 days, followed by streak seeding and 2 days of growth. Crystals were harvested from mother liquor and transferred to a holding solution consisting of 20% PEG 3350, 150 mM KF, 100 mM NaCl, and 12.5 mM Tris-Cl, pH 8.0, and then to a cryoprotective solution consisting of 27.5% PEG 3350, 150 mM KF, 100 mM NaCl, 12.5 mM Tris-Cl, pH 8.0, and 12% glycerol and plunged into liquid nitrogen. Data were collected at beamline 11-1 at the Stanford Synchrotron Radiation Lightsource (SSRL) with crystals maintained at 100 K using a liquid nitrogen vapor stream. Diffraction data were processed with XDS (32) and CCP4 (17).
The Fab Z13e1 peptide 178-1 complex was determined by molecular replacement using the program EPMR (34), with Fab 17b (Protein Data Bank code 2NY1) as the search model (76). The constant regions of Fab 17b were placed first, followed by the variable domains with truncated CDR loops. Rigid body refinement was carried out by treating each domain (VL, CL, VH, and CH1) as a rigid body, and the resulting model subsequently was refined by one round of torsion angle simulated annealing by slow cooling from 5,000 K in phenix.refine (1), which was used in all subsequent refinement steps. The model then was developed by alternating rounds of manual rebuilding with the program COOT (23) and cycles of positional and temperature factor adjustment. Bulk solvent and anisotropic temperature factor correction were carried out, as was TLS (T, translation; L, libration; S, screw-motion) refinement in later rounds. Following simulated annealing refinement, clearly interpretable density was observed for the CDR loops and bound peptide, which were built manually in COOT. Water molecules were added manually by inspection of difference Fourier maps and refined. The model quality was assessed using the JCSG quality control server, which includes MolProbity for assessing model geometry (21). The analysis of the ionization state of HisH50 and the hydrogen bonding network involving TrpH47, TyrH100G, HisH50, and AspP674 was undertaken with Protonate3D (36) implemented in the Molecular Operating Environment suite (Chemical Computing Group). Final refinement statistics are summarized in Table Table1.1. Crystals of Fab′ derived from pepsin digestion could be grown by cross-seeding and diffracted to approximately the same resolution. As this structure shows no apparent differences from the Fab obtained by Lys-C digestion, only the latter structure is reported here.
Z13e1 Fab mutants were constructed in phagemid cloning vector pComb3X, which encodes the wild-type Z13e1 light and heavy chains, by use of the QuikChange XL mutagenesis kit (Stratagene). All mutants were verified by sequencing. The preparation of crude Fab supernatants was carried out as previously described (79), with the following modifications. The mutant clones and wild-type Z13e1 Fab were transformed separately into Escherichia coli XL1-Blue cells (Stratagene), and single colonies were used to inoculate 10 ml super broth medium containing 50 μg/ml of carbenicillin and 10 μg/ml of tetracycline. Cultures were incubated overnight at 25°C and harvested by centrifugation the following day. Cell pellets were lysed by being resuspended in 1 ml CelLytic B buffer (Sigma) and incubated for 20 min at room temperature. Lysates were clarified by centrifugation at 14,000 rpm and stored at −80°C prior to use. Triplicate crude Fab supernatants were prepared to lessen the effect of batch variation and used directly for enzyme-linked immunosorbent assay (ELISA), as described below.
Ninety-six-well plates (Costar) were coated with 2 μg/ml goat anti-human F(ab′)2 (Pierce), 2 μg/ml of gp41-MBP, and 2 μg/ml of peptide 178-1 in phosphate-buffered saline (PBS), and the plates were incubated overnight at 4°C. Wells then were washed with TPBS (PBS containing 0.05% Tween 20) and blocked with 4% nonfat dry milk at room temperature for 1 h. Subsequently, the wells were washed and 50 μl of crude Fab supernatant was serially diluted in TPBS containing 1% nonfat dry milk. After 1 h, the plates were washed again and further incubated with peroxidase-conjugated goat anti-human Fab (Pierce) diluted 1:1,000 in TPBS containing 1% nonfat dry milk. Finally, the wells were developed with tetramethylbenzidine substrate (Pierce). Optical densities (at 450 nm) were determined with a microplate reader (Molecular Devices). The concentration of Fab was determined by anti-Fab ELISA (full curve, threefold dilution series) using a standard curve generated with parental Fab Z13e1. Apparent affinities were determined from the antibody concentration at half-maximal binding. All samples were tested at least twice, and the mean was taken as the final reported value.
The envelope mutations were introduced into the pSVIIIexE7pA-JR2 template (78) by a QuikChange XL mutagenesis kit. After sequence verification, the pseudotyped HIV-1JR2 mutants, competent for a single round of infection, were generated in HEK 293T cells with the luciferase reporter plasmid pSG3Δenv and env complementation, as described previously (78). The pseudotyped virus then was assayed for neutralization using TZM-bl cells as target cells (46). Serially diluted antibody was mixed 1:1 with the pseudoviruses and incubated for 1 h at 37°C prior to addition to target cells. After 48 h of incubation at 37°C, the luminescence in relative light units was measured using an Orion microplate luminometer (Berthold Detection Systems). The extent of virus neutralization was determined as the percent reduction of viral infectivity compared to that of an antibody-free control. All experiments were performed in triplicate and repeated at least twice.
Coordinates and structure factors for Fab Z13e1 in complex with peptide 178-1 have been deposited in the Research Collaboratory for Structural Bioinformatics' Protein Data Bank (http://www.pdb.org) under accession code 3FN0.
To examine the recognition of the MPER by Z13e1 and gain insight into the differences between Z13e1 and 4E10, the crystal structure of Fab Z13e1 was determined in complex with an MPER peptide to 1.8-Å resolution. The 12-residue peptide includes 8 native gp41 residues corresponding to the core epitope of Z13 (W670NWFDITN677) followed by a polylysine solubility tag. In a previous study, this peptide was named 178-1 and was a member of a series designed to assess the boundaries of Z13e1 and 4E10 recognition (11, 46). The peptide is numbered according to the HXB2 isolate (TrpP670, AsnP671, TrpP672, PheP673, AspP674, IleP675, ThrP676, and AsnP677) and contains a P chain identifier.
IgG1(κ) was secreted from a stable CHO-K1 cell line transfected with the vector pIgG-Z13e1 (46). Fab Z13e1 was prepared by endoproteinase Lys-C digestion by following the reductive alkylation protocol previously described for 2F5 (47). Crystals of the complex were grown by the incubation of Fab (27 mg/ml) with peptide 178-1 (5 mg/ml), followed by sitting-drop vapor diffusion. Crystals normally grew after several days at 22°C, but only in a subset of equivalent drops. Diffraction-quality crystals were obtained by streak seeding and belong to tetragonal space group P41212. The structure was solved by molecular replacement and refined to a final Rcryst of 21.0% (Rfree = 25.2%) at 1.8-Å resolution. The final model contains Fab residues L3 to L211 (light chain) and H1 to H227 (heavy chain), numbered according to the nomenclature of Kabat et al. (31), and peptide residues P670 to 678. Only backbone atoms are visible for LysP678, which corresponds to a Trp in the native sequence, and the rest of the polylysine tag is disordered. All CDR loops are ordered except for the apex of H3, which largely is disordered. The final refined model has excellent geometry for the Fab and the peptide with all residues, except for AlaL51 (65), in the most favored regions (Table (Table11).
The crystal structure of Fab Z13e1 reveals the canonical immunoglobulin ß-sandwich fold with an elbow angle of 168°. CDR L1 adopts canonical structure 2, L2 canonical structure 1, and H2 canonical structure 2. H1 adopts canonical structure 1, but the peptide bond between IleH30 and AsnH31 is flipped to allow interaction of the backbone amide of AsnH31 with the MPER peptide. Thus, the hydrogen bond normally present between the carbonyl of residue 29 and amide of residue 31 is absent. The conformation and location of the CDR loops and bound peptide are depicted in Fig. Fig.2A2A.
In Z13e1, L3 contains the affinity-enhancing mutations that were selected by phage display and adopts a novel conformation. The L3 sequence AL90RLLLPQL96 (underlined residues are mutated from the wild-type Z13 κ light chain sequence QL90RSDWPRL96) is more hydrophobic, particularly in the stretch of tandem Leu residues that replace SerL92 and AspL93. As the canonical structures for L3 are dependent on the hydrogen bonding capability of the GlnL90 side chain (80% conserved in κ sequences) (2, 70), mutation to Ala allows the loop to adopt a noncanonical conformation. The most striking feature is that residues L95 to L97 now loop out from the combining site and are stabilized by a hydrogen bond between the main chain carbonyl of LeuL93 and the side chain hydroxyl of ThrL97 (Fig. (Fig.2B2B).
Z13 can dimerize promiscuously with three lambda chain variants in addition to the κ light chain, which gives the highest activity. However, the lambda chains have different canonical L3 structures and do not contain GlnL90, suggesting that the wild-type κ L3 sequence does not contribute directly to binding or is not optimally packed against the heavy chain (80). The best L3 sequence match among antibodies of known structure is Fab 80R (Protein Data Bank entry 2GHW), which differs by only two residues from Z13wt L3: AspL93→Asn and ArgL96→Pro. The L3 of Fab80R adopts a distorted canonical 1 conformation, which suggests that Z13wt also is likely to adopt a canonical 1 conformation, consistently with expectations based on the observed frequency (2). However, modeling of the Z13wt L3 loop based on the 80R conformation results in numerous close contacts, particularly with ArgL91, TrpL94, and ArgL96, implying that some local adjustment must take place in the ligand-bound structure.
L3 is not involved in any direct contacts with the gp41 peptide, for which all Fab interactions are with the heavy chain. The noncanonical conformation allows LeuL92 to drop down in the bottom of the combining site and pack closely between the aromatic side chains of TyrH100G and TrpH47, optimizing the interface between VL and VH (Fig. (Fig.2).2). LeuL92 does provide a somewhat long van der Waals contact to ThrP676, but this interaction is unlikely to contribute significantly to binding affinity (Tables (Tables22 and and3).3). To assess the importance of LeuL92, this residue was mutated to Ala, and the resulting affinity for peptide was determined (Table (Table4).4). In addition, a double reversion to the wild-type κ sequence was constructed by the mutations LeuL92→Ser and LeuL93→Asp. The LeuL92→Ala single mutant and LeuL92→Ser, LeuL93→Asp double mutant showed 2.3 and 1.9% binding relative to that of Z13e1 in peptide ELISA, respectively. Although the reversions may not be compatible with the Z13e1 L3 loop structure, these results illustrate the importance of LeuL92 for optimal binding.
Z13e1 has a relatively long 17-residue H3 that encodes mainly hydrophobic, aromatic, and flexible residues (VH95AIGVSGFLNYYYYMDVH102) (80). H3 displays the common ß-bulge torso, stabilized by a salt bridge between ArgH94 and AspH101 and a hydrogen bond between the side chain of TrpH103 and carbonyl oxygen of MetH100I (2). The apex of H3 is composed of GlyH100A, PheH100B, LeuH100C, and AsnH100D and has the highest B values of the model. This degree of disorder is consistent with the expected flexibility and observed lack of interactions with peptide 178-1, although some of the long CDR H3s in other human antibodies are remarkably well ordered (15, 47, 59, 66). As in 4E10, the tip of Z13e1 H3 bends away from the binding site and does not interact with bound peptide. Only residues at the H3 base contact peptide, including IleH97, GlyH98, ValH99, TyrH100E, TyrH100F, and TyrH100G (Table (Table2).2). Small and flexible residues (GlyH98, ValH99, SerH100, and GlyH100A) precede the two hydrophobic residues, PheH100B and LeuH100C, at the tip of the loop and may provide sufficient flexibility to adapt to the membrane-bound epitope.
Fab Z13e1 was cocrystallized with peptide 178-1 (W670NWFDITN677KKKK), which includes most of the residues in the Z13e1 epitope but lacks Ser668 and Leu669 (11). Peptide 178-1 (Kd [dissociation constant], ~40 μM) was critical for obtaining crystals, as the free N-terminal amine on TrpP670 is required for crystal packing. Peptides extended by the addition of N-terminal Ser668 and Leu669 or C-terminal Trp678 and Leu679 did not yield crystals with the Fab.
The peptide is bound in a meandering S-shape conformation, with residues 670 to 671 displaying approximately helical backbone angles and residues 674 to 677 adopting a standard type I β-turn or α-helical conformation (Table (Table5).5). The N-terminal turn orients the side chains of residues TrpP670, AsnP671, TrpP672, and PheP673, along with their corresponding backbone amides, in the same direction, suggesting a helical turn conformation. The C-terminal portion of the peptide adopts a type I β-turn that also resembles a single helical turn, and it is stabilized by a hydrogen bond between the carbonyl of AspP674 and the amide of AsnP677 and via a side chain-mediated hydrogen bond between the AspP674 carboxyl and ThrP676 hydroxyl. The polar AspP674, ThrP676, and AsnP677 residues are clustered together (Fig. (Fig.3),3), forming a prominent hydrophilic surface. The S-shaped conformation allows the close approach of the aliphatic side chain of IleP675 to the indole of TrpP670, with their respective Cα atoms separated by only 5.3 Å, defining a sharp bend in the peptide.
The combining site of Z13e1 buries 429 Å2 of surface area on the peptide and a corresponding 473 Å2 on the antibody, which are within the range expected for Fab-peptide complexes (400 to 700 Å2) (67), and this compares favorably with the 451-Å2 buried surface area for the equivalent peptide epitope of 4E10 (15). Since peptide 178-1 lacks Ser668 and Leu669, which contribute significantly to binding affinity and may have specific interactions with the antibody, the buried surface area for the full epitope is expected to be slightly greater. As in 4E10, some residual available surface area in the combining site, particularly for H3, L2, and L3, may indicate that residues outside the core epitope are involved in binding.
Z13e1 uses all three heavy-chain CDR loops to bind peptide, with the greatest contribution from H2 (42% of total buried surface area on antibody) followed by H3 (32%) and H1 (23%). L3 makes one minor van der Waals interaction that accounts for only 3% of the total contacts. In total, 80 van der Waals contacts and 8 hydrogen bonds are made between Z13e1 and peptide (Table (Table2).2). IleH97 (H3), GlyH98 (H3), AsnH31 (H1), and IleP675 (peptide) contact one face of the TrpP670 indole. The indole nitrogen of TrpP670 donates a hydrogen bond to the side chain carbonyl of AsnH31. AsnP671 is a highly contacted residue, forming hydrogen bonds with the side chain and backbone amide of AsnH31, a water-mediated hydrogen bond with the backbone amide of TyrH53, and a hydrogen bond between its backbone carbonyl and the amide of GlyH54. In contrast, TrpP672 and PheP673 make relatively few side chain contacts with Z13e1 (Fig. (Fig.3B).3B). The TrpP672 indole makes van der Waals contacts with TyrH53 and GlyH54 in H2, whereas PheP673 is completely solvent exposed, although its backbone carbonyl makes two water-mediated hydrogen bonds to ThrH56 (Fig. (Fig.3B3B).
The N- and C-terminal halves of the peptide are bridged by the hydroxyl of TyrH33, which hydrogen bonds to the carbonyl of TrpP670 and amide of IleP675. The hydrophilic cluster of AspP674, ThrP676, and AsnP677 inserts into the combining site, resulting in the burial of the AspP674 and ThrP676 side chains (Fig. (Fig.3C).3C). AspP674 accepts hydrogen bonds from HisH50 N2 and a water molecule coordinated by the backbone carbonyl of ThrH57. IleP675 is packed tightly against IleH97, GlyH98, TyrH33, and TyrH100E. TrpP670 and ValH99, whose side chain is largely disordered, also contact IleP675. The ThrP676 side chain, which packs against TyrH100E, involves LysH58 in hydrogen bonding interactions. AsnP677 is the last ordered residue in the peptide and interacts only with the -amine of LysH58.
Selected positions in Z13e1 and the MPER were mutated to assess the effect on binding and neutralization. Z13e1 mutants were tested for binding to peptide 178-1 and recombinant gp41 (a maltose binding protein fusion) in an ELISA format (Table (Table4),4), whereas mutations on the HIV-1 envelope protein itself were incorporated into a pseudotyped HIV-1JR2 virus (Table (Table6)6) (46). TyrH33 hydrogen bonds with the backbone atoms of TrpP670 and IleP675, both of which are invariant residues. The mutation of TyrH33→Ala or TyrH33→Phe, which preserves hydrophobic packing in the combining site but removes the hydrogen bonding interactions to the side chain, knock out binding (<0.1%). IleH52 packs against TyrH33 and provides numerous van der Waals contacts to backbone atoms of the peptide. The mutation of IleH52 to Ala results in <1% binding. We tested additional mutations in the extended H3 tip. Although PheH100B does not contact peptide, the mutation of this residue or the deletion of the entire apex of H3 (data not shown) resulted in a 99% reduction in binding, a phenotype reminiscent of an analogous mutation in 2F5 H3 (79). Overall, the mutagenesis studies strongly support the key roles of the antibody-antigen contacts established by the crystal structure.
Neutralization by Z13e1 is critically dependent on Asp674 (46, 80). HIV-1 isolates with Asp (58% conserved) or Glu (1.1%) at 674 are neutralized by Z13e1, whereas those with Asn (12.6%), Ser (22.7%), or Thr (1.6%) are not (46). These results imply that a carboxylate at 674 is necessary for recognition. The structure reveals a complex hydrogen bonding network between the AspP674 carboxylate and the antibody (Fig. (Fig.2B2B and and3C).3C). In the unbound state, HisH50 Nδ1 would be expected to accept a hydrogen bond from the TyrH100G hydroxyl, as TyrH100G must accept a hydrogen bond from the indole NH of TrpH47. However, upon binding to the AspP674 carboxylate, HisH50 Nδ1 could deprotonate TyrH100G to form a three-centered, delocalized salt bridge. The equilibrium between the neutral and charged states of HisH50 could serve to delocalize the charge of AspP674 by a proton relay mechanism. However, in the major protonation state, AspP674 and HisH50 likely interact through a hydrogen bond. The negative charge on AspP674 is presumed to be delocalized between Oδ1, which accepts two hydrogen bonds from ThrP676, and Oδ2, which in turn hydrogen bonds with HisH50. An asparagine at position 674 would require a hydrogen bond between its side chain NH2 and the protonated N2 of HisH50, which would destabilize the complex. In isolates bearing a Ser at 674, no interaction with HisH50 is possible. Thus, the structure shows that only a carboxylate can favorably interact with HisH50.
The neutralization of HIV-1JR2 bearing the most commonly observed substitutions of Asn671 also was examined; Ser671 and Asp671 variants were neutralized as effectively as wild-type JR2, whereas the Thr671 variant was resistant (Table (Table6).6). Together, Asn671, Ser671, and Asp671 are observed in approximately 94% of isolates, with Thr671 accounting for fewer than 6% of sequenced envelopes. Thr671 may not adopt the appropriate rotamer to interact with AsnH31 and, thus, would confer neutralization resistance. The Ala substitution for 671 eliminates all three hydrogen bonds observed in the crystal structure. The mutation of AsnH31 to Ala resulted in 8.9% binding relative to that of Z13e1 (Table (Table4).4). Thus, the structure clearly shows why Asn671→Ala and Asp674→Ala substitutions result in neutralization resistance to Z13e1 and helps explain the neutralization profiles of common variants (46, 80). These results extend the previous Ala-scanning studies and indicate that the requirement for a carboxylate at 674 is the limiting factor in Z13e1 neutralization breadth.
The crystal structure of Z13e1 unexpectedly has revealed a strikingly different conformation of the gp41 MPER than that observed previously and gives insights into why this antibody differs substantially from the other MPER broadly neutralizing antibodies, 2F5 and 4E10, despite the recognition of an overlapping epitope. Peptide Ala scanning previously had assigned the Z13e1 core epitope as S668LWNWFDITN677, which overlaps with those of 2F5 (L661ELDKWASL669) and 4E10 (N671WFDITNWLW680). The Z13e1 epitope consists of two helical turns that are perpendicular to one another, and it differs from the more extended α-helical conformation of the slightly downstream 4E10 epitope.
Our studies help explain and also extend the previous mapping and mutagenesis aimed at defining the neutralization breadth of Z13e1 (46, 80). The mutagenesis of pseudotyped HIV-1JR2 in single-round infectivity assays revealed that Asn671 and Asp674 are critical for Z13e1 neutralization. Clearly, Z13e1 is limited by its dependence on Asp674, which is deeply buried and interacts specifically with a histidine on the framework region of the antibody. Accordingly, the mutation of HisH50 is one of the most destructive, in terms of binding affinity, of those analyzed in this study (Table (Table4).4). Z13e1 also makes numerous interactions with the peptide backbone, which are insensitive to sequence variation, and recognizes the side chains of invariant residues Trp670, Trp672, and Ile675, although these can be individually mutated to Ala without disrupting neutralization (46). Surprisingly, Z13e1 is tolerant of native sequence variation at Asn671 and insensitive to the common Ser substitution of Thr676 (Table (Table6).6). Thus, the structure reveals both the basis of specificity in this broadly neutralizing antibody and the limitations of its breadth of recognition for other clades and strains.
To gain insight into how Z13e1 interacts with the MPER as a whole, we modeled the natural N- and C-terminal extensions of the core epitope on the Z13e1 core epitope structure (Fig. (Fig.4A).4A). We drew upon the reported observation (30) that crystal packing influences the conformation of Trp670 in the 2F5 structure (30, 47). New crystal forms of Fab 2F5 with different peptides show that the C terminus (W666ASLW670) of the 2F5 epitope can adopt a helical turn (30). As W670NWF673 in the Z13e1 structure is compatible with a helical turn, the C terminus of the 2F5 peptide epitope structure can be connected with the N terminus of the Z13e1 peptide epitope structure to create a short helix composed of residues 666 to 672 (Fig. (Fig.4A).4A). Water molecules are found near the modeled main-chain carbonyl groups of Ser668 and Leu669, indicating that the hydrogen bonding requirements of peptide 178-1 are not fully met and, hence, are consistent with a truncated helix. The resulting model suggests that N-terminally extended peptides that include residues 666 to 669 have increased affinity for Z13e1 due to the propagation of this helix and perhaps also mediate interactions with the apex of H3 (Fig. (Fig.4A).4A). Similarly, the C terminus of the epitope may be modeled by appending residues 678 to 683, as defined in the crystal structure of 4E10 (14).
The resulting composite MPER structure consists of two helices bent by ~90° at Phe673 and reveals that Z13e1 binds in the elbow region of the MPER. The composite model is largely, although not completely, amphipathic and suggests an orientation for the MPER with respect to the membrane, as depicted in Fig. Fig.4B.4B. The majority of hydrophobic and mainly aromatic residues would insert into, or interact closely with, the membrane, including Trp666, Leu669, Trp672, Phe673, Trp678, Tyr681, and Ile682. The hydrophilic residues are localized on the opposite face; importantly, Asp664 and Asp674 both point away from the membrane. Trp670, Ile675, and Leu679 constitute hydrophobic residues that are not predicted to insert deeply into the membrane but form a hydrophobic cluster on the interior of the elbow. Although Trp670 and Ile675 are accessible to Z13e1, they could interact with phospholipid or other residues in gp41 in the native trimer, in which case Z13e1 may displace these endogenous contacts. Such features are reminiscent of membrane-interacting structures (26), particularly the membrane binding domain of PGHS-1 cyclooxygenase (53, 54).
A comparison of the 4E10 and Z13e1 structures yields significant insights into structural flexibility of the MPER. Trp672 and Phe673 are the most highly contacted residues in the 4E10 epitope, with buried surface areas of 123 and 108 Å2, respectively. 4E10 completely buries these side chains, which also contact each other through a T-stacking (edge-to-face) arrangement of the aromatic rings. In contrast, Phe673 is the least contacted residue in the Z13e1 complex, with only water-mediated interactions with the peptide backbone and a buried surface area of just 9 Å2. Trp672 has more extensive interactions with Z13e1 than Phe673, primarily via backbone hydrogen bonding and van der Waals contacts with the accessible face of the indole ring, burying 54 Å2 of total surface area in the Z13e1 combining site. In support of the different conformations of the MPER in both crystal structures, Z13e1 effectively neutralizes a Phe673→Leu variant, a known 4E10 escape mutation (27, 78), whereas neutralization of this mutant by 4E10 is compromised considerably (Table (Table6).6). Thus, a large discrepancy is found in the inferred degree of exposure of these key residues on native trimers.
The composite model suggests an explanation and a mechanism for how the Z13e1 and 4E10 epitope conformations can be readily interconverted to one another. In the Z13e1 conformation, the C-terminal helical turn is capped by Asp674 (Fig. (Fig.4C4C and Table Table5),5), which is 58% conserved and can be replaced by serine (23%) or asparagine (13%), both of which are common helix-capping residues (56, 57). In contrast, when 4E10 binds to the MPER, the C-terminal helix is extended and, instead, is capped by Asn671 (Table (Table5),5), which is 71% conserved and also can be replaced only by the helix-capping residues serine (22%) and threonine (6%). Thus, these residues appear to act as a switch that enables the extension and stabilization of a longer C-helix in which Trp672 and Phe673 rotate out of the membrane. Therefore, Z13e1 binds a Trp672/Phe673-occluded structure, while 4E10 ultimately recognizes a Trp672/Phe673-accessible conformation, although it is not clear whether these residues would be accessible in the initial encounter complex (69).
In addition to their distinct structures, 4E10 and Z13e1 also bind to opposite orientations of the MPER (Fig. (Fig.5A).5A). With Z13e1, the N terminus of the peptide contacts H3 and H1 and extends toward L3. In contrast, the peptide interaction with 4E10 begins at L3 and ends at H3. As a consequence, the light chains are oriented on opposite sides of the MPER when the peptide structures are superposed (Fig. (Fig.5B).5B). Both antibodies are oriented on one side of the putative membrane plane; however, the angle of the approach of the two antibodies is different. Since the MPER is thought to be difficult to access, it is surprising that residues 670 to 677 would be so well exposed as to allow two such widely different angles of approach. The epitope structure superposition may, therefore, indicate that these two MPER conformations are presented differently on the trimer and are associated with unique temporal and/or steric constraints. Alternatively, rearrangement of the MPER after initial antibody recognition could account for the apparent divergent approach and orientation of the two antibodies. Such a mechanism seems less likely for Z13e1, which binds less effectively to the hydrophobic residues presumed to be occluded. If 4E10 does extract Trp672 and Phe673 from the membrane, as recently suggested (69), it likely takes advantage of a readily accessible conformation of the MPER.
The strict conservation of helix-capping residues at 671 and 674 suggests that the two alternate conformations are important for viral fitness and likely represent a conformational switch that plays a role in membrane fusion. In the 4E10 helical conformation of the MPER epitope, the backbone is sequestered and the peptide presents an extensive hydrophobic face, with the conserved tryptophan residues distributed in a collar around the helical axis. As such, the helix may be able to interact with more than one hydrophobic surface, which could occur, for example, during the close apposition of viral and target membranes, as has been suggested (5). The elbow structure of the MPER bound by Z13e1 is more clearly amphipathic and would be expected to interact with only a single hydrophobic surface, presumably the viral membrane in the prefusion conformation. Interconversion between the two forms, perhaps upon activation by receptor binding, may modulate tertiary contacts with other regions in gp41, or even gp120, or interactions of the MPER with lipids. Moreover, the two conformations must differ in the elbow angle between the N- and C-helices of the MPER, which could affect the curvature of the viral membrane. Thus, the two MPER conformations may have distinct roles in membrane fusion, at different stages, due to differential interactions with the viral membrane or via the modulation of contacts made elsewhere in the trimer.
In summary, the crystal structure of Z13e1 with bound gp41 MPER peptide reveals a novel conformation of the MPER that is relevant for neutralization. Analysis of the complex also sheds light on the mechanism of neutralization and the structural basis of specificity by this broadly neutralizing human HIV-1 antibody and further suggests substantial structural plasticity in this conserved region of gp41. Neutralizing antibodies to the MPER are rare, as only three broadly neutralizing monoclonal antibodies have been described, and serum-mapping studies have suggested that the hydrophobic MPER is not very immunogenic (8, 27, 33, 37, 38, 75). It also has been suggested that 4E10 is polyspecific and reacts with cardiolipin, whereby B-cell regulation (tolerance control) may normally delete MPER-targeting antibodies. Although this view remains controversial (3, 28, 29, 37, 61), the recognition of hydrophobic sequences that would result in equally hydrophobic and sticky binding sites, such as those in 4E10, may be problematic for the immune system and for autoimmunity. On the other hand, it is encouraging that human antibodies can be elicited that are as disparate in their fine specificity for the MPER as 4E10 and Z13e1, which neutralize by binding to very different conformations of the MPER and to different subsets of residues. It also is possible that even more conformations relevant for neutralization exist but have not yet been structurally characterized. The structural characterization of novel MPER-targeting antibodies, particularly those elicited by non-B clades (8, 27), will be important for understanding the structural repertoire of the MPER. Immunogens based on known conformations and considerations of accessibility, including the masking of the hydrophobic surfaces, may be superior to grafts of unrestrained peptides. Definitive evidence about the full extent of the conformational transitions that occur during membrane fusion and how the MPER is presented in its native context on the membrane surface will require the structure of the prefusion gp120/gp41 trimer, as well as the complete postfusion structure of gp41, including the MPER, perhaps embedded in a lipid environment. However, as such structures already have proven difficult to obtain, the determination of MPER structures bound to neutralizing antibodies, such as described here, is extremely valuable and critical for the structure-assisted design of HIV vaccines and immunogens.
We thank the staff of the Stanford Synchrotron Radiation Lightsource (Beamline 11-1) for beamline support and assistance with data collection, A. Hessell for valuable technical assistance, D. Marciano and M. Elsliger of the JCSG for assistance with the IAVI/TSRI/JCSG CrystalMation robot, and R. Pantophlet and D. Leaman for useful discussions.
This work was supported by NIH grants GM46192 (I.A.W. and R.L.S.), AI69993 (M.B.Z.), and AI33292 (D.R.B.), NIH/NIAID NRSA fellowship AI74372 (R.P.), Austrian Science Fund J2845-B13 (J.S.G.), and American Foundation for AIDS Research fellowship 106427-34-RFHF (R.M.C.). The JCSG is supported by NIH NIGMS (U54 GM074898). The Scripps Research Institute thanks the International AIDS Vaccine Initiative for its scientific and development support and financial assistance.
This is manuscript no. 19745-MB from The Scripps Research Institute.
The authors have no conflicting financial interests.
Published ahead of print on 10 June 2009.