Neutralizing Antibodies against Lassa Virus Lineage I

ABSTRACT Lassa virus (LASV) is the causative agent of the deadly Lassa fever (LF). Seven distinct LASV lineages circulate through western Africa, among which lineage I (LI), the first to be identified, is particularly resistant to antibody neutralization. Lineage I LASV evades neutralization by half of known antibodies in the GPC-A antibody competition group and all but one of the antibodies in the GPC-B competition group. Here, we solve two cryo-electron microscopy (cryo-EM) structures of LI GP in complex with a GPC-A and a GPC-B antibody. We used complementary structural and biochemical techniques to identify single-amino-acid substitutions in LI that are responsible for immune evasion by each antibody group. Further, we show that LI infection is more dependent on the endosomal receptor lysosome-associated membrane protein 1 (LAMP1) for viral entry relative to LIV. In the absence of LAMP1, LI requires a more acidic fusion pH to initiate membrane fusion with the host cell relative to LIV.

structure of LI-pfGP in complex with GPC-A antibody 25.10C and a 3.6-Å cryo-EM structure of LI-pfGP in complex with GPC-B antibody 18.5C-M30 ( Fig. 1; see also Fig. S1 to S3 and Tables S1 and S2 in the supplemental material).
Overall, both LI GP structures share the same architecture as LIV GP (PDB accession number 7S8H), with an average root mean square deviation (RMSD) of 1.2 Å when aligned to a single GP monomer within the trimer (Fig. 1A). Three copies of both 25.10C and 18.5C-M30 Fab fragments bind a single LI GP trimer. As found with our recent structure of LIV bound to 25.10C (27), the heavy chain of 25.10C primarily contacts residues required for LAMP1 binding in GP1 (loop 225 to 235), while the light-chain anchors to the GP2 fusion loop (Fig. 1B). The GPC-B antibody 18.5C-M30, which is derived from GPC-B MAb 18.5C (28,29), binds to a quaternary epitope spanning two adjacent monomers and bridging their GP1 to GP2 subunits (Fig. 1C). LI differs from other lineages at several key sites that could affect antibody activity, including position 95 in the GPC-A epitope and residues 62, 198, and 397, which are located in or near the GPC-B epitope. Here, we use the structures of LI-pfGP in complex with 25.10C and 18.5C-M30 to understand how these amino acid differences allow LI to evade neutralization by most known anti-LASV antibodies.
LI Arg 95 decreases the efficacy of GPC-A antibody 36.1F. Of the two known anti-LASV GPC-A MAbs, only 25.10C is effective against LI. The other, 36.1F, is LIV-specific and does not neutralize LI (26). The lineage-specific GPC-A MAb 36.1F binds an epitope that is highly conserved between LI and LIV (27). Indeed, position 95 is the only nonconserved residue at the 36.1F binding site between the two lineages (see Fig. S4 in the supplemental material). Moreover, the loop containing R95 exhibits a similar conformation between LI and LIV GP ( Fig. 2A). Thus, a Met or Arg residue at position 95, rather than any conformational alteration, likely determines antibody reactivity at that site. Modeling suggests that the LI M95R substitution would produce a steric clash with L104 of the 36.1F CDR H3 ( Fig. 2A). In contrast, the pan-LASV 25.10C has a shorter CDR H3 that does not contact residue 95 and instead interacts with E75 in GP1.  To better understand how R95 affects antibody efficacy, we used biolayer interferometry (BLI) to characterize the binding kinetics of GPC-A MAbs to LI-pfGP and LI-pfGP bearing an R95M substitution. Monomeric GP was used for these studies to enable use of the 1:1 fit model during analysis. Pan-LASV 25.10C binds wild-type LI, LI-R95M, and LIV with high affinity ( Fig. 2B; see also Table S3 in the supplemental material). Lineage IV-specific 36.1F does not bind wild-type LI but does bind LI, bearing an R95M mutation with kinetics and affinity comparable to those of LIV ( Fig. 2B; see also Table S3). Furthermore, while 36.1F does not neutralize pseudovirus particles bearing LI (ppVSV-LI), it robustly neutralizes ppVSV-LI-R95M (Fig. 2C). These results demonstrate that an Arg residue at position 95, which occurs in the GPC-A site of lineages I, V, VI, and VII, directly inhibits MAb 36.1F binding and neutralization.
LI evades neutralization by most GPC-B antibodies. The GPC-B site is a major neutralization determinant; more than half the anti-LASV antibodies target the GPC-B epitope (26). Notably, nearly all GPC-B MAbs are derived from the same heavy-chain germ line (IGHV3-21) and thus exhibit nearly identical heavy chains. Likewise, these antibodies poorly neutralize lineage I (see Fig. S5 in the supplemental material). We previously demonstrated that antibodies 18.5C, 25.6A, and 37.7H incompletely neutralize LASV-LI ppVSV pseudovirions and fail to neutralize authentic LI LASV (28,29). BLI analysis further reveals that wild-type GPC-B antibodies exhibit more rapid dissociation from monomeric LI relative to LIV, suggesting that a high antibody off-rate may decrease antibody efficacy (see Fig. S6 and Table S3 in the supplemental material).
The GPC-B antibody 18.5C can be engineered to completely neutralize LI ( Fig. 3A) (28,29). This enhanced MAb, 18.5C-M30, introduces two Arg residues into the heavy chain via the insertion of one Arg into CDR H2 at position 54 and an Arg substitution into CDR H3 residue 100 (L100R). Here, we determined the structure of 18.5C-M30 in complex with LI-pfGP. This structure shows that the two engineered Arg residues form salt bridges to Asp400 and Asp407 of LI-pfGP (Fig. 3B). These Asp residues are conserved across all LASV lineages. Binding kinetics show that 18.5C-M30 has an increased affinity for the LI GP monomer compared to parental 18.5C, primarily due to a slower off-rate (see Fig. S6 and Table S3). Hence, neutralization of LI achieved by engineering 18.5C-M30 was likely associated with enhanced affinity to conserved parts of LASV GP.
Our LI-18.5C-M30 structure pinpoints the following three LI mutations at or near the GPC-B site that decrease antibody efficacy and, thus, are responsible for the immune evasion   Neutralizing Antibodies against Lassa Virus mBio LI also contains an Arg substitution at position 198 instead of the Ser found in all other lineages. This substitution is located in a helix spanning residues 196 to 207 that lies just above the GPC-B epitope and is adjacent to a complex glycan at N389 that sterically occludes binding by GPC-B antibodies (Fig. 4C) (28,29). In X-ray crystal structures of LIV GP, the 196 to 207 helix is ordered, and the Ser 198 side chain faces the interior of GP (28,29). In both LI cryo-EM structures presented here, this loop is disordered (residues 196 to 207). Modeling suggests that the R198 substitution may shift this helix outward to create space for the bulky Arg sidechain, which in turn may shift the N389 glycan even further over the GPC-B epitope, increasing the steric hindrance in this region (Fig. 4C). Indeed, removal of this glycan via an N389D mutation results in complete neutralization by 18.5C, 25.6A, and 37.7H (Fig. 4D). Furthermore, the R198S mutation improved neutralization of LI pseudovirions by 18.5C, and particularly improved neutralization by the two other GPC-B antibodies 25.6A and 37.7H. Kinetic analysis of GPC-B MAb binding to monomeric LI GP bearing the R198S mutation shows little difference in on-or off-rate compared to wild-type LI GP ( Fig. S6 and Table S3). Hence, the increased neutralization afforded by this mutation is likely due to increased access to the GPC-B epitope rather than through an increase in antibody affinity for GP.
Together, these three LI mutations at positions 62, 198, and 397 contribute to the immune evasion of LI from GPC-B antibodies. The impact of positions 62 and 198 varies by antibody, but Gln 397 markedly impacts every GPC-B MAb tested here by decreasing both affinity and neutralization potency. Thus, Gln 397 is likely the most impactful substitution related to antibody-mediated neutralization of LI at the GPC-B site.
LI LASV exhibits a greater dependency on LAMP1 for viral entry. Thus far, interactions of Lassa virus with the LAMP1 receptor have been characterized for LIV-GP only (20,22,23). Our cryo-EM structures of LI GP revealed alternate conformations of rotamers in regions of GP implicated in LAMP1 binding, particularly the pH-sensing residues H91 and H229 and adjacent residues Y93 and R95 (Fig. 5A). In light of the importance of this region in LAMP1 binding, we analyzed the ability of pseudovirus particles bearing LI-or LIV-GP (ppVSV-LI or -LIV) to infect haploid cells in the absence of LAMP1 (HAP1/LAMP1 2 ) or the presence of native endosomal expression of LAMP1. When LAMP1 is present, LI and LIV pseudovirus display equal levels of infectivity. In contrast, when LAMP1 is knocked out, infection of ppVSV-LI decreases considerably relative to ppVSV-LIV, indicating that LI-GP is more dependent on LAMP1 for efficient viral entry than LASV-LIV (Fig. 5B).
LAMP1 is not absolutely essential for cell entry; LIV LASV infects at 15 to 30% of wild-type levels in LAMP1-knockout cells (Fig. 5B) (20,23). LAMP1 does increase efficiency of infection, however, by increasing the pH threshold at which membrane fusion occurs to a pH of #5.5 (20,22,23). To investigate why LI is more dependent on LAMP1 for cell entry, we examined the fusogenic profile of LI versus LIV in the absence of LAMP1, using acid-bypass assays that force viral membrane fusion to occur at the cell surface rather than the endosome. Here, LIV requires a pH of #4.5, while LI requires a pH of #4.0 to enable optimal cell entry (Fig. 5C). Hence, the increased dependency on LAMP1 displayed by LI relative to LIV is likely linked to the more acidic pH required for fusogenic activity. LI may require the more hydrolytic low-pH of late endosomes to fuse in a LAMP1-independent manner and, therefore, have a far lower infection efficiency.
The mechanism behind the more acidic fusion pH required for LI is currently understudied. Residue R95 is positioned near residues H91 and H229 (H92 and H230 in LIV), which belong to the LASV His triad and mediate the pre-to postfusion conformational changes required for viral entry (22). Hence, the R95 substitution may modulate the ability of these residues to sense pH and, therefore, alter interactions with LAMP1 in the endosome. Similarly, R198 is located in a loop implicated in LAMP1 binding, and mutation of residues adjacent to this position reduces interaction with recombinant LAMP1 (32). To understand if either of these variations in LI GP are responsible for its greater dependence on LAMP1, we analyzed whether R95M and R198S pseudovirions displayed differential fusogenic profiles compared to wild-type LI. Additionally, we examined pseudovirions including K55R, Y62ins, and Q397H mutations, since LI exhibits unique mutations at these sites. However, each mutation showed a similar profile to that of wild-type LI pseudovirions (see Fig. S7 in the supplemental material), suggesting the molecular basis for the more acidic fusion pH required by LI lies beyond these single point mutations. Neutralizing Antibodies against Lassa Virus mBio

DISCUSSION
The genetic diversity exhibited between LASV lineages and the propensity for these lineages to evolve are major hurdles toward the design of broadly reactive therapeutics and vaccines (31). LI is the most genetically divergent lineage of LASV relative to the prototypical LIV (33). This work emphasizes the impact of single residue substitutions on antibody efficacy and further explains why the majority of neutralizing antibodies against the immunodominant GPC-A and GPC-B epitopes on LASV GP are ineffective against LI.
At the GPC-A site, the LI R95 substitution alone prevents neutralization by the antibody 36.1F. R95 is also observed in lineages V, VI, and VII and likely inhibits 36.1F-mediated neutralization of these lineages as well. Likewise, LASV lineages II and III contain nearby mutations at GP1 residues 73 and 75. These lineages, like LI, are also refractory to neutralization by 36.1F. Hence, this region may be a hot spot for viral evolution and highly vulnerable to escape by antibodies that bind in a 36.1F-like manner. Inclusion of R95 in future vaccine candidates may help elicit antibodies that neutralize in a 25.10C-like manner.
At the GPC-B site, neutralization is dictated by a network of contacts between antibody and GP as well as access to the site itself. For 18.5C, engineering novel GP contacts through an R54 insertion and an L100R mutation in the heavy chain can rescue its ability to completely neutralize LI. Curiously, GPC-B MAbs 25.6A and 37.7H have naturally occurring arginine residues in their heavy chain CDRs (R31 and R100 in 25.6A and R55 and R100 in 37.7H), yet neither of these antibodies can neutralize wild-type LI. We also find that each of the GPC-B antibodies examined here show improved neutralization of LI GP bearing an R198S mutation, which we hypothesize increases access to the GPC-B epitope. Similarly, removal of the N389 glycan that occludes the GPC-B site also enables robust neutralization by GPC-B antibodies. Hence, site access is likely an additional determinant for GPC-B efficacy against the divergent LASV LI.
Wild-type 18.5C poorly neutralizes ppVSV-LI but completely neutralizes LI with point mutations that allow for new hydrogen bonds via the insertion of Y62 in GP1 or the Q397H substitution in GP2. Moreover, the substitution of LI Q397 to the His found in all other lineages also enables neutralization by 25.6A and 37.7H-making it the only GPC-B mutation to equally impact all three GPC-B MAbs examined in this study. This residue is centrally located in the GPC-B site, and the identity of this residue appears to be a critical determinant for this class of antibodies. At the GPC-B site, specifically including Q397 in future vaccine candidates may help elicit a broadly reactive immune response by forcing novel GPC-B MAbs to abandon the pi-stacking interaction with LIV H398 and generate 18.5C-M30-like antibodies.
Furthermore, we found that LI has a greater dependence on LAMP1 for efficient entry than LIV. When LAMP1 is knocked out, ppVSV-LI infectivity is considerably less efficient than infection by ppVSV-LIV. The underlying molecular basis for this phenomenon lies beyond single substitutions in LI. Certainly, these results suggest variability in receptor dependence among LASV isolates and merit further exploration within and beyond LI.
Promising vaccine platforms that present full-length LASV GP have focused solely on LIV (34)(35)(36). Our data suggests that antibodies elicited from vaccination with LIV-GP may lack sufficient efficacy against LI and LI-like viruses, such as the newly emergent lineages LV to LVII, which each bear the same M95R mutation as LI. Mutations at the GPC-B site may be particularly important, as they render LI resistant to the most immunodominant group of anti-LASV neutralizing antibodies. Future vaccine candidates may, therefore, benefit from an immunogen that incorporates the LI-like R95 and/or Q397 to promote the formation of pan-LASV antibody interactions that may better accommodate variations at these positions. Moreover, to avoid the undesirable consequence of eliciting antibodies that do not target LIV M95/H397, a mixed immunogen or prime/boost strategy may provide an alternative approach to elicit a more broadly reactive response than solely including LI or LIV-like glycoproteins.
Overall, the studies here highlight key determinants on LASV-GP that will promote the development of affordable, broadly-reactive, and widely-accessible medical countermeasures against the multiple LASV lineages. Moreover, a comprehensive understanding of the molecular features on LASV-GP from different lineages will narrow the prioritization of therapeutics and vaccines to test in future studies.

MATERIALS AND METHODS
Cell lines. HEK293T (ATCC CRL-3216) and Vero (ATCC CCL-81) cells were cultured in high-glucose Dulbecco's modified Eagle's medium (DMEM) containing L-glutamine (Invitrogen, Carlsbad, CA) supplemented with 10% fetal bovine serum (FBS) (Omega Scientific, Tarzana, CA) and 1% penicillin-streptomycin solution (Penn/Strep). Cells were maintained at 37°C in a humidified atmosphere with 5% CO 2 . ExpiCHO cells were cultured in ExpiCHO expression medium and maintained at 37°C in a humidified atmosphere with 8% CO 2 . Drosophila S2 cells were cultured in Schneider's Drosophila medium at 27°C in stationary flasks. Stable cell lines were adapted to serum-free conditions and maintained with shaking at 27°C.
Expression of IgG. IgGs were expressed and purified according to references 27-29. Briefly, ExpiCHO-S cells were grown in shaker flasks in ExpiCHO expression medium in a humidified chamber at 37°C and 8% CO 2 . Cells were passaged every 3 to 4 days during early-log-phase growth at 4 Â 10 6 to 6 Â 10 6 cells/mL. One day before transfection, cells were diluted with prewarmed ExpiCHO expression medium to a final density of 3 Â 10 6 to 4 Â 10 6 cells/mL. On the day of transfection, the cell density was adjusted with ExpiCHO expression medium to 6 Â 10 6 cells/mL. Cells were transfected with a 1 to 1.15 ratio of heavy-chain to light-chain plasmid DNA using an ExpiFectamine CHO transfection kit. Cells were fed the next day according to the manufacturer's instructions. Antibodies were purified from clarified supernatants using protein A affinity chromatography via a HiTrap PrismA MAbSelect column. IgGs were eluted with 0.1 M citrate buffer, pH 3.4, and neutralized with a 1/ 10 volume of 1 M sodium phosphate, pH 8.0, before dialysis into phosphate-buffered saline (PBS).
Production and purification of Fab fragments from IgG. Purified IgG antibodies were digested by incubating with 5% papain (wt/wt) for 3 h at 37°C. The resulting Fab fragments were then purified using a Kappa select column (GE Healthcare) and further purified by size exclusion chromatography (SEC) using an S75 Increase 10/300 column (GE Healthcare).
Expression and purification of LI-pfGP and derivatives. Expression and purification of soluble LI-LASV-pfGP ectodomain monomers were performed as previously described (27)(28)(29). Briefly, the LI-LASV-GP soluble ectodomain monomer (residues 1 to 424) was modified to introduce the cysteine mutations K206C and G359C, a helix breaking E328P mutation, and the mutations L257R and L258R to alter the native S1P cleavage site to a furin protease cleavage site for production in Drosophila S2 cells (Invitrogen). This construct also contains an added LPETG amino acid sequence at the LI-GPCysR4 C terminus that allows ligation to the trimerization domain (PDB accession number 1NOG). S2 cells were grown to a density of 1 Â 10 7 cells/mL, and protein expression was induced using 500 mM CuSO 4 . Protein was purified from the supernatant by Strep-Tactin affinity chromatography and the StrepII tags removed by overnight incubation with EKMax (Thermo Fisher). The resulting protein was further purified by SEC using an S200 Increase column (GE Healthcare).
LI-pfGP trimerization and purification. LI-pfGP trimer was formed according to reference 27. Briefly, 40 mM LI-pfGP monomer was ligated to 14 mM 1NOG trimerization domain using 1.35 mM Sortase A enzyme in buffer containing 50 mM Tris, pH 7.5, 150 mM NaCl, and 2 mM CaCl 2 and incubating for 1 h at room temperature. The ligation reaction was quenched with 50 mM iodoacetamide. The resulting LI-pfGP trimer was purified using an S200 Increase column.
Antibody-LI-pfGP complex formation. Purified LI-pfGP trimer was incubated with excess Fab for at least 1 h at room temperature (RT). LI-pfGP trimer-Fab complexes were then purified by SEC using an S200 Increase column (GE Healthcare).
Cryo-EM sample preparation and data collection. Purified LI-LASV-pfGP-25.10C complexes were concentrated to 0.7 mg/mL, and a 3-mL aliquot was mixed with 1 mL 0.02 mM lauryl maltose neopentyl glycol (LMNG) detergent. No detergent was added to the LI-pfGP-18.5C-M30 complex. The samples (3 mL) were applied to C-flat 2/1 copper grids that had been plasma cleaned for 30 s in a NanoClean model 1070 (Fischione Instruments) using a mixture of 25% oxygen and 75% argon. Grids were blotted for 10 s to remove excess solution and then plunge-frozen in liquid ethane using an FEI Vitrobot (Thermo Fisher Scientific). Frozen grids containing LI-pfGP-25.10C were imaged using a Titan Krios (Thermo Fisher Scientific) equipped with a Gatan K2 detector, while grids containing LI-pfGP-18.5C-M30 were imaged using a Titan Krios equipped with a Gatan K3 detector. Movies for the LI-pfGP-25.10C complex were collected at a magnification of Â46,300 in super resolution mode and correspond to a calibrated pixel size of 0.548 Å/pixel. Movies were collected in a single session with a defocus range between 1.0 and 2.5 mm underfocus. Movies for the LI-pfGP-18.5C-M30 complex were collected at a magnification of Â75,750 in counting mode that corresponds to a calibrated pixel size of 0.6656. A full description of the cryo-EM data collection parameters is presented in Tables S1 and S2 in the supplemental material.
Cryo-EM data processing. All data processing was carried out using cryoSPARC v2.14.2 (37). Movies were motion-corrected using Patch motion correction and subsequently contrast transfer function (CTF) corrected using Patch CTF estimation.
For the LI-GP-25.10C complex, particle projections were picked from the micrographs using Topaz (38), downsampled by a factor of 2, and then subjected to reference-free two-dimensional (2D) class averaging to generate a particle stack of 152,874 particles that were used for ab initio reconstructions and subsequent three-dimensional (3D) and CTF refinements. The final C3 symmetrized reconstruction reported a resolution of 3.09 Å according to the gold standard Fourier shell correlation criterion (39).
For the LI-pfGP-18.5C-M30 sample, particle projections were also picked using Topaz (38) but were downsampled by a factor of 3 before reference-free 2D averaging, which generated a final particle stack of 380,210 particles. After generating an ab initio with subsequent 3D refinements, the micrographs were unbinned so that they were downsampled by a factor of 1.95 before undergoing a final series of 3D and CTF refinements. The final C3 symmetrized reconstruction reported a resolution of 3.59 Å according to the gold standard Fourier shell correlation criterion (39).

ACKNOWLEDGMENTS
We are grateful to Dewight Williams at Arizona State University (ASU) for help with cryo-EM data acquisition and acknowledge the use of facilities within the Eyring Materials Center at ASU. We further acknowledge Ruben Diaz Avalos at the La Jolla Institute for Immunology (LJI) for cryo-EM data acquisition at facilities at LJI. We are grateful to Sean Whelan, then at Harvard Medical School, for the acquisition of the (HAP1/LAMP1 2 ) cell line. We thank Heather M. Callaway, Sean Hui, Dawid Zyla, and Michael J. Norris for helpful discussions.
We gratefully acknowledge R21 AI137809, U19 AI142790, and R01AI141251 for financial support of this project. Adrian Enriquez holds a Postdoctoral Enrichment Program Award from the Burroughs Wellcome Fund. Conceptualization