key: cord-0265221-ievabjox authors: Crooks, Emma T; Almanza, Francisco; D’addabbo, Alessio; Duggan, Erika; Zhang, Jinsong; Wagh, Kshitij; Mou, Huihui; Allen, Joel D; Thomas, Alyssa; Osawa, Keiko; Korber, Bette T; Tsybovsky, Yaroslav; Cale, Evan; Nolan, John; Crispin, Max; Verkoczy, Laurent K; Binley, James M title: Engineering well-expressed, V2-immunofocusing HIV-1 envelope glycoprotein membrane trimers for use in heterologous prime-boost vaccine regimens date: 2021-07-20 journal: bioRxiv DOI: 10.1101/2021.07.20.453076 sha: c7bb9458de0362c4a4132ac14d5f62d933fde12f doc_id: 265221 cord_uid: ievabjox HIV-1 vaccine immunofocusing strategies have the potential to induce broadly reactive nAbs. Here, we engineered a panel of diverse, membrane-resident native HIV-1 trimers vulnerable to two broad targets of neutralizing antibodies (NAbs), the V2 apex and fusion peptide (FP). Selection criteria included i) high expression and ii) infectious function, so that trimer neutralization sensitivity can be profiled in pseudovirus assays. Initially, we boosted the expression of 17 candidate trimers by truncating gp41 and introducing a gp120-gp41 SOS disulfide to prevent gp120 shedding. “Repairs” were made to fill glycan holes and other strain-specific aberrations. A new neutralization assay allowed PV infection when our standard assay was insufficient. Trimers with exposed V3 loops, a target of non-neutralizing antibodies, were discarded. To try to increase V2-sensitivity, we removed clashing glycans and modified the V2 loop’s C-strand. Notably, a 167N mutation improved V2-sensitivity. Glycopeptide analysis of JR-FL trimers revealed near complete sequon occupation and that filling the N197 glycan hole was well-tolerated. In contrast, sequon optimization and inserting/removing other glycans in some cases had local and global “ripple” effects on glycan maturation and sequon occupation in the gp120 outer domain and gp41. V2 mAb CH01 selectively bound trimers with small high mannose glycans near the base of the V1 loop, thereby avoiding clashes. Knocking in a N49 glycan perturbs gp41 glycans via a distal glycan network effect, increasing FP NAb sensitivity - and sometimes improving expression. Finally, a biophysical analysis of VLPs revealed that i) ∼25% of particles bear Env spikes, ii) spontaneous particle budding is high and only increases 4-fold upon Gag transfection, and iii) Env+ particles express ∼30-40 spikes. Overall, we identified 7 diverse trimers with a range of sensitivities to two targets that should enable rigorous testing of immunofocusing vaccine concepts. Author Summary Despite almost 40 years of innovation, an HIV vaccine to induce antibodies that block virus infection remains elusive. Challenges include the unparalleled sequence diversity of HIV’s surface spikes and its dense sugar coat that limits antibody access. However, a growing number of monoclonal antibodies from HIV infected donors provide vaccine blueprints. To date, these kinds of antibodies have been difficult to induce by vaccination. However, two antibody targets, one at the spike apex and another at the side of the spikes are more forgiving in their ‘demands’ for unusual antibodies. Here, we made a diverse panel of HIV spikes vulnerable at these two sites for later use as vaccines to try to focus antibodies on these targets. Our selection criteria for these spikes were: i) that the spikes, when expressed on particles, are infectious, allowing us to appraise our vaccine designs in an ideal manner; ii) that spikes are easy to produce by cells in quantities sufficient for vaccine use. Ultimately, we selected 7 trimers that will allow us to explore concepts that could bring us closer to an HIV vaccine. 243 soluble CD4 (sCD4) and measured the ability of V3 mAbs (39F and 14e) to inhibit infection of 244 CCR5-expressing cells (102). In this format, V3 mAbs neutralized 11 of our strains (S2 Fig). 245 CE217 was later verified as 14e-sensitive in mutation analysis (see below). 14e-sensitivity of the 246 other 5 strains of our panel could not be confirmed, as their infectivities were too low. For CH01 and its UCA, we also measured IC50s against PV produced in GnT1-cells. We 248 previously showed that CH01 saturation improves against GnT1-PV, presumably as clashes with 249 larger glycan head groups are eliminated (13). Similarly, PG9 neutralization is more effective 250 against B4GalT1+ST6Gal1-(abbreviated "B4G+ST6") modified PV that increases hybrid glycans 251 and terminal -2,6 glycan sialylation (13). Neither of these modifications overtly increase V3 252 sensitivity, suggesting that trimer folding is not impacted. Q23 was the most sensitive strain to MAbs CH01, VRC38 and their UCAs ( Fig. 2A; (23)) 254 and was also highly PG9-sensitive. Although Q23's tier 1B classification and moderate PGT145-255 sensitivity reflect a somewhat less compact V2 apex compared to other strains ( Fig. 1) , it is 256 nevertheless 14e-resistant, and may therefore be useful to prime V2 NAbs in a vaccine regimen. 257 Surprisingly, WITO, 001428, and X2278 were not neutralized by CH01 10g/ml ( Fig. 2A) , 258 contrasting previous data (97). This may be due to CH01's characteristic "sub-saturating" 259 neutralization and/or that our 'workhorse' CF2 assay is slightly less sensitive than the commonly 260 used TZM-bl assay (13). Indeed, CH01 exhibited IC30s of 0.04g/ml and 6g/ml, respectively 261 against WITO and X2278 ( In contrast, 13 of 15 GnT1-PVs were CH01-sensitive (Fig. 2B , top right; BB201 and 264 CNE58 GnT1-PVs did not infect sufficiently). For 10 of the CH01-sensitive GnT1-PVs, maximum 265 CH01 saturation was close to 100% (Fig. 2B , bottom right) and its IC30s were also lower (Fig. 266 2B, compare upper panels). VRC38, PG9 and PGT145 neutralized all of the group 1 strains, 267 except that PGT145 did not neutralize BB201. In several cases, they also neutralized group 2 268 strains ( Fig. 2A) . Only 6101 was resistant to all V2 NAbs, probably due to its missing N160 glycan 269 (Fig. 1, S1 Fig). Overall, Q23's exquisite sensitivity to CH01, its UCA and other V2 NAbs support its use 271 as a V2 priming immunogen. Several other CH01-sensitive strains could be useful in boosting. 272 However, as outlined below, selected strains may benefit from engineering to increase V2 273 sensitivity and/or trimer expression. 275 In situ membrane expression of candidate trimers. Trimer expression on VLP surfaces is improved by truncating gp160 at position 708, 277 leaving a 3 amino acid gp41 tail (gp160CT; Fig. S1 , Fig. 3A (compare lanes 1 and 2) ). Based on 278 our previous work, high VLP trimer expression is achieved by co-transfecting Env plasmids with 279 MuLV Gag (Fig. S1 in (23)) and Rev plasmids (used when Env is not codon-optimized). The SOS 280 mutation (501C+605C) further improves JR-FL trimer expression (Fig. 3A , compare lanes 2 and 281 3). E168K and E168K+N189A variants of JR-FL gp160CT SOS were also well-expressed (Fig. 282 3A, lanes 4 and 5). Gp160CT SOS mutants of strains Q23.17 and CNE58 were also better 283 expressed than their gp160 WT and gp160CT WT counterparts (Fig. 3A, lanes 6-10) , although 284 neither expressed as well as JR-FL. One explanation for the high SOS trimer expression is that 285 gp120 shedding is eliminated, evidenced by the lack of gp41 stumps (Fig. 3A , compare lanes 2 286 and 3). SOS and gp160CT mutations were made in Env plasmids previously used to make PV 288 of the 17 candidate strains (Fig. 1) , along with BG505 as a reference. SOS gp160CT Env In contrast to the mixed expression of group 1 strains, all 5 group 2 SOS gp160CT 302 mutants had high levels of trimer (Fig. 3B, lanes 14-18) . Gp160 and gp120 expression was high 303 in all cases (magenta and red dots in Fig. 3C ). Corresponding gp41 bands were also observed in 304 all but AC10. Overall, these blots reveal vastly different Env expression of different strains that 305 could greatly impact their utility in vaccines (92). The group 2 strains and the 4 group 1 strains 306 that express high levels of functional authentic gp120/gp41 trimers (Q23, T250, c1080 and WITO) 307 are of particular interest for follow up. Below, we used various strategies to modify the most 308 promising strains to knock in V2 sensitivity and/or to improve Env expression. We first modified our prototype vaccine strain, JR-FL, to try to improve its V2 and FP 312 sensitivity. The E168K+N189A mutant was sensitive to VRC38, PG9 and PGT145 (Fig. 1 ) and 313 partially CH01-sensitive (Fig. 2B ). V2 sensitivity might be improved by removing V1V2 clashing 314 glycans and/or by increasing strand C's overall basic charge. To maintain high expression, 315 modifications were made in the JR-FL SOS gp160CT background. We initially compared WT 316 and SOS PV NAb sensitivities. As shown previously, SOS PV infection can proceed after receptor 317 engagement by adding a low molarity reducing agent to break the gp120-gp41 disulfide (103). 318 WT and SOS PV neutralization profiles were broadly similar (S3A Fig). However, CH01 saturation 319 of SOS mutant PV was greater and an IC50 was measurable. Thus, the SOS mutant improves 320 trimer expression, retains V3 resistance and slightly improves CH01 sensitivity. Since the JR-FL N188 and N189 sequons overlap, only one site can be occupied by 322 glycan. We therefore investigated the effects of knocking out each sequon alone or together. expression. S158T infectivity was also unchanged, but S364T infectivity was reduced to ~15%. 480 Glycopeptide LC-MS revealed that S158T mutant glycosylation was broadly similar to the parent, 481 with score changes <+/-5 (Fig. 5C , S1 Data and analysis, S8 Fig) . Differences in both directions 482 were observed, e.g., at positions N156 and N301. Maturation states of the complex gp41 glycans 483 also differed, but these are prone to vary, as mentioned above (S1 Data and analysis). Elevated 484 sequon skipping/core glycans at N160 could be a direct consequence of the rare S158T mutation 485 adjacent to this sequon. Significant skipping also occurred at position N339 (Fig. 5D , S1 Data and 486 analysis). Overall, S158T caused some unwanted changes in glycan maturation and skipping. S364T dramatically increased glycan differentiation at N332 of the gp120 outer domain 488 (Fig. 5C, S8 Fig) , as did N301 and N386, to a lesser extent. N262 glycan data was not obtained. 489 Significant sequon skipping occurred at N339 and N637. The N362 glycan is not proximal to N295 490 and N332, suggesting allosteric effects of this mutant, as for S158T. Both S158 and S364 are 491 well-conserved across strains (S1 Fig) , suggesting that they are structurally important. We 492 therefore suggest that sequon optimization be considered only when it does not disturb conserved 493 residue(s). D197N successfully and completely knocked in the N197 glycan (S1 Data and analysis). 495 (Fig. 5C, S8 Fig) . N301 maturation modestly increased. Gp41 glycans varied in complexity (S1 496 Data and analysis). Unlike the S158T and S364T mutants, N160 and N637 sequons were fully 497 occupied. However, like these other mutants, some skipping occurred at position N339 (Fig. 5D, 498 S1 Data and analysis). Overall, the effects of D197N on other glycan sites were milder than those 499 of S158T and S364T (Fig. 5C, S8 Fig) . Modeling suggests that the enhanced N301 glycan 500 maturation is a localized effect (Fig. 5A, S7 Fig, S8 Fig) , so it appears that overall trimer 501 conformation is not perturbed. Sequon-optimized D197N+S199T was inferior to D197N. It only filled the N197 site with 503 glycan to ~90% and caused dramatic glycan holes elsewhere, most notably at N463 that was 504~91% skipped (S1 Data and analysis). N262 partly toggled to complex. The N301 glycan became 505 immature. These differences help in comparison of the D197N+S199T versus D197N mutant (S1 506 Data and analysis, S8 Fig) . As for the S158T and S364T mutants, distal glycans were affected, 514 that would have given more complete insights into how the N49 glycan improves VRC34 515 sensitivity. While N611 is not close to the N49 glycan (Fig. 5A) , it is possible that smaller glycans 516 at the other gp41 sites provide space for the N611 glycan to move aside for VRC34 binding. 517 Glycan maturation at N301 was also modestly impacted. Our model suggests that some effects 518 are localized (Fig. 5A ), but others (e.g., N188) are distal, suggesting a global conformational 519 change, consistent with partial V3 non-nAb sensitivity (Fig. 4, S5C Fig) . N611Q led to increased N262 glycan maturation, decreased N463 glycan maturation and 521 partial skipping at N188 and N339 (Fig. 5D ). T49N+N611Q caused reduced N301 maturation (Fig. 522 5C, S8 Fig) and skipping at positions N339 and N463 (Fig. 5D) Analysis of N138A+N141A revealed that the N135 glycan is complex (at least in the 530 absence of these neighboring glycans). N135 was not detected on the parent, probably due to its 531 proximity to N138. Significant skipping at N262, N295 was observed, and, to a lesser extent, at 532 N339 (Fig. 5D ). The small amount of glycan detected at position N262 was far more mature than 533 on the parent (Fig. 5C ). Although this mutant was infectious (Fig. 4A, lane 12) , since this glycan 534 is structurally important, its absence could cause some misfolding (100). Glycan maturation 535 differences were also observed at positions N156, N188, N616 and N637. N138A+N141A Env complexed with CH01 exhibited radical glycan changes at some 537 positions: a shift to high mannose glycans at positions N135 and N188 is consistent with a 538 preference for high mannose glycans to minimize clashes at the binding site (S7 Fig, S8 Fig, Fig. 539 5A) (13). However, N356 and N463 glycans also increased in high mannose, despite being distal 540 from the CH01 epitope. Intriguingly, glycan N262 became less mature, while glycan N295 was 541 largely complex in CH01-bound sample, despite both being skipped in the unbound sample. 542 Conversely, the N332 glycan exhibited more skipping and glycan N616 was more complex in the 543 CH01-bound sample. These notable findings reveal the presence of glycan species in CH01 544 complexes that were not detected in the reference sample and vice versa. Thus, some glycans 545 exhibit more extensive variability than expected. The differences may reflect the idea that CH01 546 neutralizes the N138A+N141A mutant to a maximum of only ~75%, suggesting that it binds only 547 a fraction of trimers where N135 and N188 glycan clashes are minimal, that this fraction of trimers 548 carries other glycan variants that may further improve CH01 binding or that are inextricably linked 549 with the presence of small high mannose glycans at position N135. Overall, these mutants reveal that outer domain glycans (N156-N339) are prone to 551 maturation changes, while inner domain glycans N88, N356-N448 are largely unchangeable. 552 Sequon skipping was also more common at some outer domain glycan sites, particularly N339. 559 mutation improved VRC34 sensitivity as expected, while V2 NAbs were largely unaffected except 560 for a modest loss of PG9 sensitivity (Fig. 4B, lane 32) . Previous studies suggested that modifying 561 V3 sequence (106) and an S365V mutant (107) may improve V2 NAb sensitivity. However, trimers 562 mutated with a global V3 consensus sequence (lanl.gov) did not express efficiently, and S365V 563 had little effect (Fig. 4, lanes 33 and 34) . This suggests that cognate V2-V3 sequences are 564 important for folding and that the effect of S365V is context-dependent. Highly basic C-strands may initiate V2 NAb lineages via electrostatic interactions (36, 69, 566 108). However, a V169R mutant to render the JR-FL C-strand more like many V2-sensitive group 567 1 strains (Fig. 1 ) -was misfolded (Fig. 4, lane 17) , as described above. A D167N mutant provide 568 another way to increase strand C charge (Fig. 4B, lane 35) , as found in some V2-initiating 569 sequences (69). This further increased CH01 sensitivity (Fig. 4B, During V2 NAb ontogeny in natural infection, the C-strand may become more neutral, as 579 the virus attempts to escape NAbs. In turn, V2 NAbs evolve to be less dependent on electrostatic 580 charges and depend more on V2 "anchor" residues (36, 54, 69, 108). To mimic the "escape" 581 phenotype of such "late" viruses, we made an R166K+V169E variant. We also added back the 582 V1 and N611 glycans and modified the FP sequence to the second most common variant (8). 583 Including these changes in boosts could help V2 and FP NAbs evolve to tolerate sequence 584 variations and navigate glycans. However, none of the V2 NAbs neutralized this variant, most 585 likely because V169E eliminates V2 binding completely. However, the array of other mutants in 586 Fig. 4B provide a variety of options to increase V2 stringency in boosts without eliminating V2 587 sensitivity altogether. Conversely, VRC34 neutralized I515L comparably to other mutants that 588 retain the N611 glycan, suggesting that it tolerates this FP sequence variation. Finally, we investigated approaches to improve JR-FL Env processing at the lysine rich 590 gp120/gp41 junction (see S1 Text). While we were unable to improve cleavage efficiency by 591 mutation or furin co-transfection, data suggest that a lysine or arginine at position 500 (S1 Fig) 592 should be used as a repair mutation in strains where other residues are present. 594 An alternative PV neutralization assay for poorly infectious clones. Our standard neutralization assay uses pNL-LucR-E-and an Env plasmid to make PV for 596 infection of CF2.CD4.CCR5 cells (NL-Luc assay). In this assay, Q23 SOS gp160CT PV infection 597 was low and close to our arbitrary cutoff of 50,000 relative light units (RLUs), where neutralization 598 becomes difficult to distinguish from background. The "gold standard" TZM-bl protocol cannot be 599 used for SOS PV infection, as it involves overlaying cells on virus-antibody mixtures, which is 600 incompatible with our requirement to wash cells after briefly exposing them to 5mM DTT to break 601 the SOS bond of spikes attached to cellular receptors, allowing infection to proceed (103). We 602 therefore sought a different protocol that uses pre-attached cells. We adapted a PV assay 603 previously reported for coronaviruses, in which viral budding is driven by MuLV GagPol and 604 luciferase is carried by plasmid pQC-Fluc (109). PVs made this way mediated elevated infection 605 versus the NL-Luc assay for poorly infectious Q23, WITO and T250 SOS PV (Fig. 6A ). However, 606 JR-FL SOS PV infection (already high in the NL-Luc assay), was slightly lower in the pQC-Fluc 607 assay. To check if neutralization sensitivity was impacted by assay differences, we compared 608 PG9 neutralization of the four viruses in both assays. The NL-Luc assay resulted in high error 609 bars compared to the pQC-Fluc assay, most notably for Q23 and WITO (Fig. 6B) . Nevertheless, 610 PV PG9 sensitivities were comparable, suggesting that the pQC-Fluc assay is a reasonable 611 substitute whenever infectivity is too low in the NL-Luc assay. In addition to the engineering approaches above, we evaluated many others, exemplified 808 in S2 Text. We previously found that codon-optimized MuLV Gag drives higher yields of Env trimer 812 on VLPs, compared to pNL-LucR-E-(23). Electron microscopy showed that some particles bear 813 surface spikes (77, 112), although "bald" particles with no spikes are also common. We were 814 concerned that MuLV Gag-induced budding might outpace surface Env expression, decreasing 815 the proportion of "Env+" VLPs and/or spike density. To investigate the Gag-dependency of particle trimer expression, we co-transfected a 851 perspective on the increased Env output when transfecting with high amounts of Gag versus no 852 Gag (Fig. 10A, compare lanes 1 and 5) . Essentially, transfecting with a high dose Gag increases 853 particle numbers by only ~4-fold with a concomitant increase in Env, as particle production is 854 already high even with no Gag transfection. Overall, this data agrees quite well with the EM data 855 in that particles are 13-37% Env+ and the proportion of Env+ particles does not change much 856 with Gag co-transfection, which is perhaps not surprising, if Gag only raises above spontaneous 857 particle production levels by about 4-fold. 865 Subtracting autofluorescence from the ~125 PGT121 mAbs/particle lowers the estimate to ~109 866 PGT121 molecules per particle. If 3 PGT121 molecules can bind per trimer at saturation, this 867 suggests a spike density of ~30-40 trimers per particle. This aligns quite well with our earlier 868 estimate of 27 spikes per particle for our early JR-FL VLPs (77). There was not a clear difference 869 in the extent of particle PGT121 binding between samples (Fig. 10C) , suggesting that spike 870 density does not vary, consistent with the finding above that the proportion of Env+ particles also 871 does not change drastically with Gag co-transfection. Finally, single vesicle flow cytometry also 872 revealed that most particles were between 70-180nm in diameter, centered around 120nm, but 873 some particles appeared to be much larger, up to ~300nm, in agreement with EM data (Fig. 10B ). 976 hinge on the use of rigidifying disulfides like DS-SOSIP that prevent misfolding. Conversely, 977 functional trimers are flexible, increasing the potential for engraftments to perturb trimer folding, 978 for example, due to incompatibilities between a new V1V2 loop with the V3 loop of the scaffold. Our efforts to improve gp120/gp41 processing were also not successful. Non-basic 980 residues at position 500 may reduce processing (S1 Text), but reverting H500R (in combination 981 with other mutations) did not improve T250 infectivity or expression. Although SOSIP processing 982 benefit from the "R6" mutation (122) and furin co-expression (123), both of these approaches led 983 to reduced membrane trimer expression and infectivity. This may be because the furin site is 984 largely inaccessible on membrane trimers, regardless of enzyme or substrate sensitivity. Indeed, 985 JR-FL membrane trimers were among the most effectively processed, while those of many other 986 strains remain predominantly uncleaved. The scarcity of well-expressed membrane trimer strains is perhaps analogous to the 989 problem of identifying soluble trimers, except that the nature of the challenge differs. Our efforts 990 revealed that mutants with valuable effects soluble trimers were, aside from the SOS mutant, 991 unhelpful in the context of membrane trimers. This underlines the extensive differences between 992 these two forms of Env that present different challenges. Indeed, different strains make better 993 prototypes in the two formats, e.g., JR-FL for membrane trimers and BG505 for SOSIP. Thus, 994 while expression is a problem for membrane Envs, it is not a problem for soluble trimers. 995 Conversely, while a key challenge for SOSIP is rendering them to be closed trimers, this is 996 generally not a problem for membrane trimers, as their membrane context prevents them from 997 naturally adopting a triggered conformation but requires I559P and other mutations in soluble 998 format. Another challenge was that we insisted that selected membrane trimers be functional, so 1180 concepts for immunofocusing V2 and FP NAbs. Ultimately, success in multiple polyclonal outbred 1181 models would provide strong support for clinical translation. A V169R mutation would increase strand C positive charge, possibly improving 345 V2 sensitivity. However, neither the N135A+V169R double mutant nor the V169R alone were 346 infectious or expressed trimer (Fig. 4A, lanes 5 and 17). Y173 and Y177 of the V2 loop Since the 351 Y173H mutant alone did not affect CH01 sensitivity (S5A Fig), we infer that the increased CH01 352 sensitivity of N135A+Y173H is due to N135A. N135A+V181I also improved infectivity, albeit 353 insufficiently to measure PV sensitivity (Fig. 4A Compared to N135A+Y173H, these mutants were slightly more VRC38-sensitive, but 357 slightly less CH01-sensitive (Fig. 4B compare lanes 6 4A, compare lanes 3 and 10) 360 and modestly improved CH01 sensitivity and saturation, but not to the extent of N135 glycan 361 knock out mutants (Fig. 4B, compare lanes 3, 6, 8, 9 and 10; S5A Fig). VRC38 and PGT145 362 sensitivity was also higher, but PG9 sensitivity was unchanged (Fig. 4B, lanes 3 and 10) However, if 366 the N141 sequon of JR-FL membrane trimers is only 50% occupied by glycan, as reported 367 previously (79), the negligible effect of N141A is perhaps not surprising. Incomplete N141 368 occupation may be due to spatial competition with the N138 sequon. If so, N141 occupation may 369 increase when N138 is absent. Accordingly, a N138A+N141A double mutant exhibited improved 370 CH01, VRC38 and PG9 sensitivity compared to N138A alone To try to further improve CH01 sensitivity, we removed all 3 V1 glycans together, first as V2 loop, this glycan could impact V2 NAb binding (Fig. 5A). The 377 N135A+N138A+N141A+D197N+S199T mutant was well-expressed and infectious. Unlike the 378 single N135A mutant, an additional Y173H mutation was not needed Since the N197 glycan lies at the edge of the V2 apex (Fig. 5A), it is possible 383 that the D197N+S199T mutant used in combination with the triple V1 glycan mutant may directly 384 impact V2 NAb sensitivity. Compared to the parent, the D197N mutant alone exhibited higher 385 PGT145 sensitivity but weaker or no PG9 and CH01 sensitivity In this context, N135A showed high VRC38 393 and PGT145 sensitivity (Fig. 4B, compare lanes 6 and 21). However, these mutants were neither 394 CH01-nor PG9-sensitive. Therefore, the reduced PG9 saturation of the E168K+N188A+N189A 395 mutant noted above We also knocked in the rare N49 glycan (modeled in Fig. 5A) 401 that is carried by well-expressed strains WITO and AC10 (S1 Fig), hoping to boost JR-FL trimer 402 expression We therefore 405 toggled the N49 and N611 glycans, with or without D197N. T49N and N611Q glycan mutants did 406 not appreciably impact PV infectivity or trimer expression (Fig. 4A, lanes 25-31). N611Q improved 407 VRC34 sensitivity, while T49N did so to a lesser extent (S5B Fig). T49N+N611Q was only 408 marginally more VRC34-sensitive than N611Q alone (Fig. 4B, lanes 3, 25, 27 and 30). T49N 409 mutants were all modestly 39F-sensitive (Fig. 4B, lanes 25 Notably, N49 mutants caused a slight gp41 mass decrease, coupled with the expected 419 slight gp120 mass increase (S6A Fig, compare lanes 1 and 2). The gp41 mass decrease of the 420 T49N mutant was smaller than that of the N611Q mutant that knocks out a gp41 glycan (S6A Fig, 421 lanes 1, 3, 5 and 6) Conversely, T49N+N611Q did not reduce gp41 mass further than N611Q alone 425 (S6A Fig, lanes 5 and 6). Thus, the effect of T49N on gp41 is eliminated when combined with (S6A Fig, lanes 7-10). However, gp41 mass did not change, as 428 expected. Expression of these V1 mutants was somewhat weaker than the parent As reported previously (13), parent JR-FL gp41 was endo H-432 resistant, consistent with complex, fucosylated glycans. However, N49 mutant gp41 exhibited a 433 ladder of endo H-sensitive species, consistent with the idea that the N49 glycan limits gp41 glycan 434 maturation (S6B Fig, lanes 2 and 4) To gain further insights into the effects of the various mutations above that add remove or 439 modify particular sequons, we assessed glycan occupation and maturation by glycopeptide in-440 line liquid chromatography mass spectrometry (LC-MS) (67, 81) Thus, the untrimmed high mannose glycan, M9Glc, has a score of 1, while the most highly 443 branched and fucosylated complex glycan HexNAc(6+)(F)(at each site are summarized in Fig. 5B. The nature of glycans at each site generally 446 match a previous report that categorized JR-FL PV Env glycans by another method However, the N160 and N386 We next evaluated glycan score differences at each site in pairs of samples. Score 450 changes were recorded in a dot plot (Fig. 5C) for sites that were >10% occupied by glycans 451 (excluding core glycans) in both samples Sequon skipping and core glycans are shown in Fig We first compared two preparations of JR-FL SOS E168K+N189A VLP trimers This revealed minor differences in gp120 glycans, with a modest difference high mannose 456 trimming at position N156 (Fig. 5C, S8 Fig). Gp41 glycans were all heavy and complex (S1 Data 457 and analysis, S7 Fig). Sequon skipping was rare and varied between samples Glycan core was found occasionally 460 (~5% or less) at 4 sites in one sample, but not at all in the other (S1 Data and analysis; average 461 % core shown in Fig. 5D). Several sequons could not be assigned a glycan A 465 comparison of 'parent' trimers and monomeric JR-FL gp120 revealed that glycan types (i.e., high 466 mannose or complex) were similar at many positions (S1 Data and analysis, S8 Fig). However, 467 gp120 monomer glycans were more differentiated at positions N88, N156, N160, N241, perhaps 468 reflecting the greater access to glycan processing enzymes (Fig. 5C). Conversely, glycans N295 469 and N301 were less mature, which, as for SOSIP Neither mutant affected 614 Q23 is highly CH01 UCA-sensitive (Fig. 2) and expresses well (Fig. 3B-D), so may be 615 ideal for V2 NAb priming. Q23 SOS is also highly V2-sensitive (Fig. 7B, lane 1, S9A Fig). To try 616 to further increase Q23's V2 sensitivity, we removed the two V1 glycans at positions N133 and 617 N138 alone and together. These mutants reduced infectivity and expression (Fig. 7A, lanes 1-4), 618 but had little effect on V2 NAb or CH01 UCA sensitivities To try to increase FP NAb sensitivity, we tested the effect of N611A alone and together 621 with N88A. As for JR-FL, N611A improved VRC34 sensitivity. Vaccine-elicited FP NAb vFP16.02 622 also neutralized this mutant (Fig. 7B, lane 5). When N88A was overlaid, vFP16.02 was still able 623 to neutralize, but VRC34 sensitivity was lost Together, there was 629 no net effect on expression (Fig. 7A, lane 9). Notably, D49N knocked in vFP16.02 sensitivity and 630 increased VRC34 sensitivity (Fig. 7B, compare lanes 1 and 7), as for JR-FL (Fig. 4). PGT145 and 631 CH01 sensitivities were also slightly higher CH01 sensitivity was moderately higher than the parent (S9A Fig). 14e saturation also 634 increased, although it did not achieve an IC50, suggesting partial exposure (S9D Fig). CH01 UCA 635 sensitivity was unaffected (Fig. 7B, lane 10, S9B Fig). Overall, this further suggests that the N49 636 glycan opens the trimer slightly to expose V2 and V3 targets. To try to further increase V2-637 sensitivity, we overlaid the D167N mutant. However, this showed loss of PGT145, CH01 and 638 CH01 UCA sensitivity, overt V3 sensitivity and poor expression (Fig. 7A and B, lane 11, S9A, B 639 and D Fig). Since N49 and D167N may both modestly increase V3 sensitivity, together they may 640 lead to overt V3 sensitivity we wondered 643 if essentially the reverse mutation, i.e., knocking out Q23's basic residue by a R169I mutation 644 might improve its expression. However, this was not the case, and sensitivity to V2 and FP NAbs 645 was reduced or eliminated Given Q23's CH01 UCA sensitivity, we tested if it could also stimulate CH01 UCA double knock in' mice, i.e. expressing both heavy and light chain 649 rearrangements (110) were effectively labeled by WITO SOSIP (S10A Fig). As expected Q23 SOS D49N+N611A VLPs also stimulated ex vivo CH01 UCA 651 dKI+ splenic B cells effectively and this result titrated (S10B Fig). GnT1-VLPs induced more 652 robust stimulation, whereas bald VLPs did not stimulate cells (S10B Fig). Thus, Q23 VLPs may 653 be highly effective at priming CH01-like specificities in a vaccine regimen We next attempted to improve WITO, another well-expressed (Fig. 3B) We next checked the effects of removing the 3 V1 glycans at positions 133, 140 and 663 145 alone and together. Expression of N133A was lower, but N140A and N145A expressed like 664 the parent (Fig. 7C, S11B Fig). Sensitivities to multiple V2 NAbs were slightly higher (Fig. 7D, 665 lanes 2-4). However, removing all 3 glycans together reduced CH01 sensitivity and also caused 666 some V3 sensitivity To accelerate screening, we combined this 671 double mutant with TL514-515GI (FPvar1) and N611A to knock in FP sensitivity. However, this 672 mutant was overtly V3-sensitive and CH01-resistant (Fig. 7D, lane 7, S11A Fig) -essentially the 673 reverse of the desired effect. The same mutant lacking the G300N+R305K (Fig. 7D, lane 9) was 674 not overtly V3-sensitive, suggesting that G300N+R305K causes misfolding. TL514-515GI slightly 675 improved VRC34 sensitivity (Fig. 7D lane 8), and, as expected We next tested the effects of removing the unusual N49 and N674 glycans that also exist 678 in the well-expressed AC10 strain (Fig. 1, Fig. 3, S1 Fig). Both mutations reduced Env expression 679 (Fig. 7C, lanes 10 and 11, S11B Fig) AE and virtually absent elsewhere (S11C Fig). Above, we 23 682 saw that knocking in the N49 glycan had no effect on JR-FL expression (S6 Fig) and improved 683 Q23 expression slightly (Fig. 7A), suggesting that N49 impacts expression in some, but not all 684 scenarios. The N674 glycan is slightly more prevalent (13% conserved) and is present in 5 of our 685 17 strains (Fig. 1, S1 Fig). However, none of these other N674 glycan-containing strains were 686 well-expressed Finally, we combined the FP-immunofocusing mutant TL514-515GI+N611A with D167N This improved WITO sensitivity 689 to multiple V2 NAbs, albeit with a moderate increase in V3 sensitivity (Fig. 7D, lane 12, S11A Fig), 690 similar to the JR-FL D167N mutant T250 is a well-expressed group 1 strain. We initially repaired the gp160CT SOS parent Poorly 698 saturating 14e neutralization suggested partially open trimers, so it was unsurprising that the 699 D49N mutant led to overt V3 sensitivity (Fig. 8B, lane 3, S12C Fig). Removing one or both V1 700 glycans led to a modest increase in CH01 sensitivity In contrast, D167N lost V2 sensitivity, similar to the D49N mutant, 702 and became overtly V3-sensitive CE217 is among the most V2-sensitive group 1 strains, but is modestly expressed (Fig Expression was not further 710 increased by K49N (Fig. 8C, lanes 2 and 3, S13A Fig). D386N to fill in a glycan hole also had little 711 effect, aside from slightly decreased PGT145 and PG9 sensitivity (Fig. 8D, lane 4). G300N (104) 712 further decreased PGT145 neutralization activity (Fig. 8D, lane 5), but had little effect on CH01 713 and its UCA (S13 Fig B, C) Removing V1 and V2 clashing glycans generally improved V2 sensitivity (Fig. 8D 717 E Fig). N189A and N195A mutants were also measurably susceptible to the CH01 UCA (Fig. 8D, 718 lanes 8 and 9, S13C Fig). Removing both N189A and N195 glycans, however, did not further 719 increase CH01 or VRC38 sensitivity (Fig. 8D, lane 10). Indeed, N137A+N189A+N195A eliminated 720 CH01 and VRC38 sensitivities altogether (Fig. 8D, lane 11). Adding D167N to N195A improved 721 PG9, but not CH01 sensitivity (Fig. 8D, compare lanes 9 and 12). Furthermore, VRC38 and 722 VRC34 sensitivities were reduced and lost Since K49N and D167N mutations may both induce partial V3 sensitivity that become overt 724 when in combination, we reverted N49K. This modestly improved V2 NAb sensitivity and slightly 725 decreased V3 sensitivity that nevertheless remained overt K170Q mutants were made to try to reduce V2 sensitivity, for possible 727 late boosting. However, these mutants almost completely eliminated V2 sensitivity, reduced trimer 728 expression and were overtly V3-sensitive (Fig. 8C and D, lanes 14 and 15). Finally, an S399G cause significant changes in mAb sensitivities but reduced expression and infectivity Gp160 truncation (gp160CT WT) led to a loss 736 of PG9 sensitivity, which was partly restored by the SOS mutant, albeit with some V3-sensitivity 737 (S14A Fig). D49N markedly improved expression (S14B Fig) but led to overt 14e-sensitivity. This 738 contrasted the modest effect of N49 on the V3 sensitivities of JR-FL and Q23, probably because 739 the SOS parent is already partially V3-sensitive, like T250 SOS parent (S12 Fig) These latter mutants 745 are more V3-sensitive than they are to V2 NAbs, suggesting a significant loss of trimer 746 compactness that is incompatible with our goal to immunofocus on V2. Therefore, any further 747 mutants should not be combined with either D49N or H375S. Since the parent virus is partially 748 V3-sensitive, a strategy akin to T250 may be effective (avoiding D167N but removing clashing 749 glycans). However, we did not pursue c1080 further at this point Given our success with JR-FL (Fig. 4), we took a similar strategy with other group 2 strains The AC10 parent is already PG9-and PGT145-sensitive and lacks a clashing N130 glycan (Fig so we removed these from AC10 first, alone and together with innermost glycans. V1 glycan 759 mutants N137A and N137A+N142A led to modest changes in PGT145 and PG9 sensitivity, but 760 remained CH01-and VRC38-resistant (Fig. 9B, S15A Fig). V2 glycan mutants N185A and 761 N185A+N184A either eliminate an overlapping sequon or both V2 sequons (Fig. 1, S1 Fig), 762 leading to PG9 resistance, but retained PGT145-sensitivity and resistance to 14e, CH01 and 763 VRC38 (Figs. 9B and S15A). Finally, D167N mutant caused increased PG9 sensitivity, detectable 764 CH01 sensitivity, but also partial V3-sensitivity. Although PGT145-sensitivity was intact, it was the best expressed clone in our panel (Fig. 3B-D) and is also PGT145-sensitive 769 (Figs 1 and 2A) Removal of V1 and V2 glycans significantly increased sensitivity to multiple V2 mAbs (Fig. 9D, 775 lanes 4-7 and S15B Fig). Removal of V2 glycans N183 and N187 knocked in VRC38 sensitivity Accordingly, we made several initial 781 mutants in combination: E49N to try to maximize expression, N130H to eliminate a V2 clashing 782 glycan and E305K to try to improve V1V2 packing (104) PG9 sensitivity may be due to N130H 785 mutation and/or E305K. VRC34 and 14e sensitivities were likely a result of E49N. N611Q 786 improved VRC34 sensitivity, as expected (Fig. 9F, lane 3). Finally, we removed potentially 787 clashing V1 and V2 glycans, starting with those closest to the base of each loop (i.e. N136A and 788 N193A), then double mutants. Both single mutants (N136A and N193A) resulted in detectable 789 CH01 IC50s, albeit sub-saturating (Fig. 9F, lanes 4 and 7) unlike that in JR-FL (Fig. 4) Of the group 2 strains, 6101 is the poorest expressing (Fig. 3B-D) and also lacks V2-817 fixed amount of WITO SOS gp160CT with 10-fold decreasing Gag doses Another 819 sample was generated by transfecting a high dose of Gag alone. Supernatants were filtered and 820 1000x concentrated. As expected, higher doses of MuLV Gag drove production of more 821 particulate Env (Fig. 10A). However, Env was detected even when 1,000-fold less Gag or no Gag 822 was co-transfected (Fig. 10A, compare lanes 1, 4 and 5) Gag only "bald" VLPs (Fig. 10B, top row) provided 825 a reference to help identify Env spikes on other samples 2-4) revealed particles with surface structures not detected on bald VLPs that we infer 827 as Env spikes. These putative spikes do not adopt clear propellor-like structures, perhaps 828 because bald particles like those in the 831 top row were also prevalent in all the other samples. Counting Env+ and Env-particles in each 832 sample revealed that approximately a quarter to a third were Env+. There was a trend towards a 833 greater proportion of Env+ particles in samples made with little or no Gag. Thus, our concern that 834 efficient Gag-induced particle production might lead to an overwhelming proportion of bald VLPs 835 appears to be unfounded rows 2-4, that may be detached spikes, that may dissociate due to VLPs collapsing 837 during the process of negative staining. Finally, the size of particles varied. Many particles were 838 approximately 100nm diameter, although some were much larger, up to ~300nm in diameter We next analyzed the same VLP samples by single vesicle flow cytometry (vFC™ which uses a fluorogenic membrane probe to detect and size vesicles, 842 and fluorescent mAbs to measure vesicle surface cargo by immunofluorescence Using this method with Alexa647-labeled PGT121 revealed a modest Transfecting sample (Fig. 10D). Co-848 transfecting Env reduced the particle count by ~60% to 1.5 x 10 8 /µl (Fig. 10D, left) For new strains to be adopted for vaccine use, 877 we insisted they each satisfy 4 key criteria. First, that they are well-expressed, like our JR-FL 878 prototype. Second that they are high quality, i.e., with sequons fully occupied, gp120/gp41 well-879 processed and not overtly V3-sensitive (partial V3 sensitivity is acceptable, as long as it is not 880 greater than V2 sensitivity). Third, that, collectively, they cover a range of sensitivities to desired 881 target(s), ranging from acute, UCA-triggering to that of typical transmitted isolates. Fourth Our exhaustive engineering efforts are summarized in S16 Fig. At completion, we 885 identified 7 trimers from 17 initial strains that each satisfy our criteria (S16 Fig Modified Q23, T250 and CE217 trimers could be used in priming, boosting with modified WITO Given our stiff selection criteria, the remaining clones did 888 not 'make the cut' for a variety of reasons: poor expression (CAP45, KER2018, CNE58, BB201, 889 001428, X2278), poor infectivity (CM244, 6101), V3 sensitivity (c1080) and insufficient V2-890 sensitivity (AC10) The 892 complexities that govern trimer phenotypes in our criteria defy prediction by Env sequence 893 alignments. Nevertheless, to accelerate the analysis of clones in the latter part of this project, we 894 began by initially filling in glycan holes and "repairing" insertions, deletions and overlapping 895 sequons -thus correcting "errors of nature Although CT 900 removes retention signals (115), our observation that C-terminal tags reduce trimer expression 901 (S2 Text), suggests that the weaker expression of Envs with cytoplasmic tails is sequence-902 independent. Our default CT mutant leaves a 3AA cytoplasmic tail (S1 Fig); further truncation 903 did not improve trimer expression. Also like our JR-FL prototype, SOS mutants of various strains 904 improved trimer expression Native trimers are flexible: 906 apex folding is influenced by the V1V2 loops that sit atop and mask the V3 loop on the resting 907 spike (95) its effect on our trimers was modest, as was SOS (102). Thus, SOS gp160dCT 910 trimers exhibited relevant Our findings are in line with the idea that the trimer apex exists in a variety of states, 912 ranging from fully closed and V2-sensitive to fully open, overtly V3-sensitive In 916 contrast, partly V3-sensitive T250 was prone to become overtly V3-sensitive and therefore had to 917 be dealt with more carefully (S12 Fig). Perhaps the most striking example of how trimer 918 conformations change with mutation was for CE217, where PGT145-sensitivity was lost in the 919 most CH01-sensitive mutants that also gained some V3-sensitivity, then CH01 sensitivity was 920 then lost as V3 sensitivity became overt with more mutations (Fig. 8D). Thus trimers sample a 921 number of conformations, including intermediates that are partially V3-sensitive but retain some 922 V1V2 NAb sensitivity (107) Thus, the effect of SOS on NAb sensitivity depends on the apex 926 conformation of parent trimers. Notably, PGT145 sensitivity of gp160CT SOS trimers was 927 somewhat reduced in all cases, as neutralization depends on a tightly folded apex that is slightly 928 perturbed by both the SOS and CT mutations. Overall, our findings are consistent with a slightly 929 activated SOS trimer state (116) Given the frequency of these misfolded mutants, it is not surprising that they are well-934 documented in other studies (3, 34, 104, 117-120)trimer interactions, this is not necessarily the case. Instead, it may simply be that the 937 mutant disturbs trimer quaternary interactions. This latter appears to be the case for a quartet of 938 V3 and V2 mutations (104) that we hoped might tighten the apex Thus, while mutations throughout the trimer can adversely impact proper folding, attempts to 941 "close" partially open trimers face a complex and nuanced challenge to identify the residues that 31 rendered the trimer open in the first place. For example, in the case of T250, the V1 glycans 943 clearly limit V2 apex folding/V2 NAb sensitivity We 951 also made a large number of point mutants, including many reported previously (33, 88, 95, 106, 952 107). Overwhelmingly, these repairs did not markedly improve trimer expression, infection or NAb 953 sensitivity, suggesting that none of them address the underlying reasons for poor membrane 954 trimer expression. Thus, while these repair strategies may be effective in some formats In some cases, adding an 959 N49 glycan improved trimer expression (WITO, c1080) but not in others (JR-FL etc). How might 960 the N49 glycan impact trimer expression? Other ineffective attempts to improve expression included modified signal peptides and 964 codon optimization. Different expression plasmids or lentiviral vectors were also unable to 965 improve expression consistently. This suggests that transcription is not a major bottleneck for 966 membrane Env expression. Moreover, making lentiviral cell lines is quite labor-intensive and 967 eliminates the flexibility afforded by switching Env plasmids when making VLPs by transfection. 968 Domain swaps offer the potential benefit of a well-expressed Env as a scaffold for 969 immunogenic domains However, in our hands, they had little effect on membrane trimer expression (see example in S2 The discrepancy may be due to the fact that In our hands, V1V2 and gp120 chimeras were overtly V3-sensitive or non-functional This is a potential advantage over soluble trimers, where 1002 appraisal is limited to binding assays that may not track perfectly with neutralization. We were 1003 excited to observe that SOS PVs were functional for several other strains aside from our JR-FL 1004 prototype. However, we were not surprised that infectious counts were lower This presented a trade-off between deciding either select WT trimers because they are more 1006 functional or selecting SOS trimers because they express better. However, the pQC-Fluc assay 1007 dramatically improved infection sensitivity so that all strains were useable. While we do not know 1008 why the pQC-Fluc assay improves infectious counts Overall, membrane trimer expression appears to depend on many factors, rendering it 1015 difficult to identify and resolve expression bottlenecks. Thus, it is important to begin with clones 1016 that at least express modest levels of Env. One strategy we did not attempt but that may be helpful 1017 would be to screen for highly expressed clonal relatives of V2-sensitive However even if successful, such clones may not satisfy our other criteria and 1019 may still require rounds of mutation and screening Sequon skipping and optimization 1021 Glycopeptide analysis revealed that sequon skipping is usually limited for membrane The resulting 1023 epigenetic glycan holes are problematic because they induce non-NAbs that could distract from 1024 immunofocusing strategies However, we 1026 only observed N160 skipping in one parent sample and the S158T mutant. The latter is consistent 1027 with a previous study in which S158T caused N160 skipping in BG505 SOSIP (84). Thus, it follows 1028 that when two sequons are closely juxtaposed (a 1 amino acid gap in this case), reducing the 1029 efficiency of the first site can increase occupancy of the second. This may explain why S158 1030 heavily predominates over T158 in natural isolates. Similarly, sequon optimization mutants S364T 1031 and D197N+S199T were unhelpful. Moreover, in both cases, skipping increased at various distal 1032 sites. Overall, sequon optimization has little benefit for This was manifested in the T49N mutant's dramatic changes in glycan scores at various 1037 positions, as well as sequon skipping at N195 and in the core glycan at N625. While it might be 1038 expected that knock in of the N49 glycan would decrease the maturation of neighboring glycans, 1039 e.g. N276 and N637, it affected distal glycans. Furthermore, paradoxically, some glycans became 1040 more differentiated. This suggests that N49 knock in influences the network of closely spaced 1041 glycans on the trimer surface with variable and sometimes far reaching effects that are not easily 1042 understood by rigid models of trimer structure The effects of glycan knock in can differ between SOSIP and 1047 membrane trimers. Thus, for SOSIP, the presence of N197 glycan decreased N156 and N160 1048 glycan trimming, but not for JR-FL membrane trimers Removing N611, a well-conserved glycan caused similar ripples in glycan maturation and 1051 sequon skipping like N49 knock in, consistent with a role in maintaining trimer architecture Notably, some distal glycans became less 1053 mature, despite the reduced overall glycan count, again suggesting perturbation of the trimer 1054 glycan network. The N138A+N141A exhibited a similar phenotype Considering sum of the effects of mutants in Fig. 5C, sequon skipping in JR-FL membrane 1057 trimers is concentrated in the outer domain of gp120, between positions 156 and 339 and rarely 1058 occurs elsewhere We speculated that high glycan numbers and/or lack of glycan holes (gaps in structural 1063 glycans) might drive folding and associate with high expression. However, glycan toggling in 1064 either direction did not consistently impact expression. That said, as covered above, adding the 1065 N49 glycan did improve expression for some strains. Furthermore, glycan removal in several 1066 cases reduced expression It may be that glycan content is a delicate balance of sufficient glycans to drive 1068 folding, but not so many that glycan overcrowding occurs, which could result in high mannose 1069 glycan bias (13), sequon skipping or a reduction in folding kinetics Overall, the varied effects of glycan toggling on proximal and distal glycans provides 1071 reason for caution. For example, a glycan hole that increases V2 apex sensitivity will only be 1072 effective if it does not open up other unwanted glycan holes. On the other hand, if "off-target" 1073 holes differ between successive vaccine shots Our mutant screening efforts paradoxically suggest that V1V2 glycans further from the V2 Eliminating the outermost glycans also had more benefit for CE217. The same 1079 may be true for T250 and KNH1144, although in these cases, to accelerate the discovery efforts, 1080 we combined V1 and V2 glycan knock outs, precluding a clear comparison of single glycan 1081 knockouts. Furthermore, the N130 glycan is absent in V2-sensitive strains and its removal from 1082 group 2 strains KNH1144 and sc422 improved their V2 sensitivities. However, as mentioned 1083 above knocking the N197 glycan did not appreciably impact either expression or V2 1085 sensitivity, although toggling this glycan can impact V2 sensitivity of other strains Although this could suggest that CH01 recovered an 1090 earlier "high mannose" trimer glycoform, not all sequons were affected. For example, the N616 1091 glycan was more mature. It could be that glycan maturation at different sites is co-dependent, via 1092 glycan network effects. If so, however, how can we reconcile the rarity and indeed total absence 1093 of glycoforms at several sites in the corresponding uncomplexed parent? Considering that CH01 1094 substantially neutralizes the N138A+N141A mutant, we would expect the trimer glycoform it binds 1095 to register prominently amid the trimer glycovariants that constitute the parent sample. However, 1096 this assumes that all glycoforms are equally infectious, which may not in fact be the case. Indeed, 1097 a significant fraction of JR-FL trimers remains uncleaved and therefore non-functional V2 bNAbs typically bind one-per-spike (62, 95). C strand charge is at a premium during 1102 the early ontogeny of V2 NAbs. Therefore, we engineered trimers to increase C-strand's charge 1103 (Fig. 1), taking care to use substitutions that are acceptable based on sequence alignments Of these, 166 and 171 were present in all but 6101. Our attempts to fix the C-strand and 1106 other repairs in this strain, however, did not result in sufficiently functional trimers. K/R168 was 1107 present in all but JR-FL, where E168K was an effective knock in. D167N, as found in V2 NAb 1108 initiating clones consistently improved V2 sensitivity This was surprising, considering that V169R mutation 1112 of ADA improves its V2 sensitivity and decreased V3 sensitivity (95). Thus, the benefits of 1113 knocking in a basic residue at position 169 is context-specific. The reverse mutation K169I/E 1114 reduced expression and increased V3 sensitivity in CE217 Biophysical analysis of VLPs 1118 We investigated the basis for the dominant proportion of bald particles over Env+ particles 1119 in the hope that new information might illuminate ways to improve VLP quality and possibly spike 1120 density, which may assist NAb priming. In the past, flow cytometric measurement of individual 1121 particles has been challenging. However, new state-of-the-art methods supported by a set of 1122 accepted guidelines are now enabling quantitative and reproducible measurement of individual 1123 extracellular vesicles and their molecular cargo (129) 114) provide a background of 1126 particles, among which Gag-or Env-bearing particles are also released. Some particle production 1127 may be driven by Env alone, given that Env's heavy glycosylation promotes ER stress. This may 1128 explain why VLPs invariably contain uncleaved gp160, as Env released under stress bypasses 1129 furin processing. Some particles may arise from spontaneous budding of endogenous Gag in the 1130 293T cell genome, e.g. HERV. Finally, some particles may derive from FBS that carries ubiquitous 1131 vesicles. After transfection, cells are washed in PBS and replaced with 1% FBS medium a day 1132 later This 1134 explanation is weakly supported by the observation that co-transfecting high doses of Gag+Env 1135 leads to higher total particle production than Env alone and a higher total number and fraction of 1136 Env+ particles (Fig. 10D, right lanes 2 and 4) Is there a way to purify Env+ particles? Immunocapture might work but eluting them may be 1140 problematic. If Env-EVs are higher in some other marker (i.e. CD81) because they bud differently 1141 or originate from FBS well-expressed, V2-sensitive and functional 1144 membrane trimers, how will we use them in vaccine studies? A first step will be to fix the FP 1145 sequence of all clones to fusion peptide variant 1 (FP8var1) so that we can begin investigating 1146 the immunofocusing on this site at the same time as the V2. The most CH01 UCA-sensitive Q23 1147 mutant would be the ideal prime. T250 and CE217 that are also slightly CH01 UCA-sensitive 1148 group 1 strains would also be useful in early shots. WITO and the 3 group 2 strains could be used 1149 as boosts. It may be useful to prime with D167N+N611Q mutant trimers and/or where clashing 1150 glycans are removed, and then gradually reverse these modifications in boosts Thus, by decreasing CH01-1154 sensitivity in boosts, NAbs may gradually develop an ability to navigate glycans and sequence 1155 diversity, while gaining N88 contacts for FP NAbs and N160 contacts for V2 NAbs with increasing 1156 dependency on conserved anchor residues within strand C. The V3 occlusion and high sequon 1157 occupation of most trimers should help avoid inducing non-NAbs to the V3 loop and glycan holes As a result, successive shots may induce new 1163 strain-specific Abs, rather than building sufficiently on lineages initiated by preceding shots. In the 1164 other study (54), it is unclear why bNAbs did not develop. Possible explanations might be the 1165 repertoire limitations of rabbits for making V2 NAbs or insufficient shared memory T cell help 1166 between shots Superinfection with a new and diverse strain, is perhaps the closest natural 1170 infection scenario to our proposed SHPB regimens. Evidence suggests that superinfection 1171 doesn't promote NAb breadth (130). However, an important difference in SHPB is that trimers are 1172 modified to immunofocus on NAbs, whereas superinfecting trimers are unlikely to share 1173 vulnerable sites. Depending on preliminary tests, to keep NAbs "on track" between shots, we may 1174 need to reduce or eliminate strain diversity. The variety of mutants of each strains will provide 1175 ample ways to reduce strain diversity in our regimens, as needed HIV-1 Env plasmids. Abbreviated names of Env strains are given first, with full names and 1186 GenBank references in parentheses, as follows: Q23 (Q23.17; AF004885.1), WITO (WITO.33, 1187 AY835451.1) 1191 (X2278.c2.B6, FJ817366.1), JR-FL (JR-FL, AY669728.1), AC10 (AC10.29, AY835446.1) Full-length Env clones of the above strains, commonly used to make PVs for neutralization 1195 assays, were obtained from the NIH AIDS Reagent Repository, the Vaccine Research Center and 1196 The Scripps Research Institute. In many cases, these Env plasmids used expression plasmids 1197 such as pCI ii) Gag and Rev plasmids. A plasmid expressing murine leukemia virus (MuLV) Gag (23) Whenever Env plasmids used native codons, we co-transfected pMV-Rev 0932 that expresses 1201 codon optimized HIV-1 Rev to maximize Env expression Glycosyltransferase plasmids Glycosyltransferase plasmids pEE6.4_B4GalT1 (expressing 1203 -1,4 galactosyltransferase and pEE14.4_ST6Gal1 (expressing -galactoside -2,6-1204 sialyltransferase were co-transfected at a ratio of 1% and 2.5% total plasmid DNA MAb plasmids were obtained from their producers and the NIH AIDS Reagent 1206 Repository For VLP production, Env plasmids were co-transfected in Human Embryonic Kidney 293T along with the MuLV 1214 Gag plasmid (23) and pMV-Rev 0932, as needed. 48 hours later, supernatants were collected, 1215 precleared, filtered, and pelleted at 50,000g in a Sorvall SS34 rotor. To remove residual medium, 1216 pellets were washed in 1ml of PBS, recentrifuged in a microcentrifuge at 15,000 rpm Luc.R-E and an Env plasmid using PEI Max. Briefly, PV was incubated with 1224 graded dilutions of mAbs for 1 hour at 37 o C, then added to CF2Th.CD4.CCR5 cells, plates were 1225 spinoculated, and incubated at 37 o C (13). For wild-type (WT) PV, plates were incubated for 3 1226 days, after which luciferase was measured. For SOS PV, following a 2-hour incubation, 5mM DTT 1227 was added for 15 minutes to activate infection. The mAb/virus mixture was replaced by fresh 1228 media ii) pQC-Fluc assay. PV were produced by co-transfecting Env plasmids with pMLV GagPol and 1230 pQC-Fluc-dIRES (abbreviated as pQC-Fluc) (109). The resulting PV were used in neutralization 1231 assays with CF2Th Post-CD4 assay. PV were mixed with sCD4 with or without V3 mAbs 14e or 39F Blue Native PAGE-Western Blot 1236 VLPs were solubilized in 0.12% Triton X-100 in 1mM EDTA. An equal volume of 2x sample 1237 buffer (100mM morpholinepropanesulfonic acid (MOPS), 100mM Tris-HCl, pH 7.7, 40% glycerol, 1238 and 0.1% Coomassie blue) was added. Samples were spun to remove any debris and loaded 1239 onto a 4-12% Bis-Tris NuPAGE gel (Thermo Fisher) and separated for 3 hours at 4C at 100V Proteins were then transferred to polyvinylidene difluoride (PVDF) membrane, de-stained, and 1241 blocked in 4% non-fat milk in PBST. Membranes were probed with a cocktail of mAbs 39F, 2F5, 1242 b12, 4E10, 14e, and PGT121, followed by alkaline phosphatase labeled anti-human Fc conjugate 1243 (Accurate Chemicals) and were developed using VLPs were denatured by heating with 2-mercaptoethanol for 10 minutes at 90 o C, then 1247 mixed with Laemmli buffer, then loaded onto 4-12% Bis-Tris NuPAGE gel (Invitrogen) Biolabs) was added to the samples after reduction and denaturation, followed by incubation for 41 Proteins were transferred onto PVDF membrane, de-stained, and blocked in 4% non-1251 fat milk in PBST. Membranes were probed as for BN-PAGE blots above Next, Env proteins were reduced and alkylated with 20mM 1257 iodoacetamide (IAA) for 1h in the dark, followed by a 1h incubation with 20mM DTT to eliminate 1258 residual IAA. Alkylated Env proteins were buffer exchanged into 50mM Tris/HCl, pH 8.0 using 1259 Vivaspin columns (3 kDa). Aliquots were digested separately overnight using trypsin and 1260 chymotrypsin (Mass Spectrometry Grade, Promega) VLPs were processed in the same way, except that were initially buffer exchanged 1263 into 50mM Tris HCl 0.1% Triton X-100 (w/w) to disperse lipids. To identify the glycome of the 1264 trimers that complexed with mAb CH01, VLPs were mixed with excess CH01 and incubated for 1265 1h at 37ºC. Screw cap spin columns were incubated with protein A-agarose for 10 minutes to 1266 allow for spin column resin equilibration before washing with gentle Ag-Ab binding buffer VLP-CH01 complexes were then applied to the spin columns and left to 1268 incubate for 30 minutes. Columns were washed twice with gentle Ag-Ab binding buffer prior to 1269 elution in 100-200 μL gentle Ag-Ab elution buffer (Thermo Fisher Scientific). Eluted VLP-CH01 1270 mixtures were then buffer exchanged into 100μL 50mM Tris/HCl pH 8.0 for subsequent reduction 1271 and alkylation Liquid chromatography-mass spectrometry (LC-MS) glycopeptide analysis 1274 Peptides were dried again, re-suspended in 0.1% formic acid and analyzed by nanoLC ESI MS with an Ultimate 3000 HPLC (Thermo Fisher Scientific) system coupled to an Orbitrap 1276 Eclipse mass spectrometer (Thermo Fisher Scientific) using stepped higher energy collision-1277 induced dissociation (HCD) fragmentation. Peptides were separated using an EasySpray PepMap 100 C18 3μM 75μM x 1279 2cm) was used in line with the LC prior to separation with the analytical column. LC conditions 1280 were as follows: 280 minute linear gradient consisting of 4-32% acetonitrile in 0.1% formic acid 1281 over 260 minutes, followed by 20 minutes of alternating 76% acetonitrile in 0.1% formic acid and 1282 4% acetonitrile in 0.1% formic acid, to ensure all the sample elutes from the column. The flow rate 1283 was set to 300nL/min. The spray voltage was set to 2.7 kV and the temperature of the heated 42 1284 capillary was set to 40°C. The ion transfer tube temperature was set to 275°C. The scan range 1285 was 375−1500 m/z. Stepped HCD collision energy was set to 15% Orbitrap at a resolution MS1=120,000, MS2=30,000. The AGC target for MS1 was set to standard 1288 and injection time set to auto which involves the system setting the two parameters to maximize Glycopeptide fragmentation data were extracted from the raw file using Byos Data were evaluated manually for each glycopeptide; the peptide was 1294 scored as true-positive when the correct b and y fragment ions were observed, along with oxonium 1295 ions corresponding to the glycan identified. The MS data was searched using the Protein Metrics 1296 305 N-glycan library with sulfated glycans added manually. The relative amounts of each glycan 1297 at each site as well as the unoccupied proportion were determined by comparing the extracted 1298 chromatographic areas for different glycotypes with an identical peptide sequence. All charge 1299 states for a single glycopeptide were summed. The precursor mass tolerance was set at 4 ppm 1300 and 10 ppm for fragments. A 1% false discovery rate (FDR) was applied. The relative amounts of 1301 each glycan at each site as well as the unoccupied proportion were determined by comparing the 1302 extracted ion chromatographic areas for different glycopeptides Hex(10+) was defined as M9Glc, HexNAc(2)Hex(9−5) was classified as M9 to M3. Any of these structures containing a fucose were categorized as FM Hex(, compositional isomers are grouped, so, for example, a triantennary glycan contains Core glycans refer to truncated 1311 structures smaller than M3. M9glc-M4 were classified as oligomannose-type glycans. Glycans 1312 containing at least one sialic acid were categorized as NeuAc and at least one fucose residue in 1313 the Glycans were categorized into I.D.s ranging from 1 (M9Glc) to 19 (HexNAc(6+)(F)(x)) These values were multiplied by the percentage of the corresponding glycan divided by the total 1316 glycan percentage excluding unoccupied and core glycans to give a score that pertains to the 43 1317 most prevalent glycan at a given site. Arithmetic score changes were then calculated from the 1318 subtraction of these scores from one sample against others as specified The model representation of the JR-FL SOS E168K+N189A trimer was constructed using SWISS-MODEL based on an existing structure of the 426c DS-SOSIP D3 trimer (pdb: 6MYY) Glycans were modelled on to this structure based on the most abundant glycoform identified from 1324 site-specific glycan analysis using WinCoot version 0.9.4.1 and PyMOL version 2.5.0. For sites 1325 which were not identified, a Man9GlcNAc2 glycan was modelled. Conditional color formatting was 1326 used to illustrate the predominant glycoforms of modeled glycans, as follows: green Phenotyping and calcium flux of CH01 UCA dKI-derived splenocytes splenocytes were phenotyped 1331 with 0.5μg/mL of anti-B220 BV650, anti-CD19 APC-R700 (both from Becton Dickinson) and 1332 WITO-SOSIP-BV421 HIV Env tetramers Thermo Fisher) for 30 min. To evaluate B-cell stimulation, splenocytes 1334 were stained with anti-B220 BV650 and anti-CD19 APC-R700 for 40 minutes HBSS, pre-stained cells were loaded with Fluo-4 via by mixing with equal volumes of After another HBSS wash, cells were resuspended in 1339 calcium-containing HBSS and incubated at room temperature for 5 minutes before activation by 1340 anti-IgM F(ab′)2 (Southern Biotech) or VLPs. Fluo-4 MFI data for total B-cells Negative-stain electron microscopy was applied to a freshly glow-discharged carbon-coated 1345 copper grid for 10-15 s and removed using blotting paper. The grid was washed with several 1346 drops of buffer containing 10 mM HEPES, pH 7.0, and 150 mM NaCl, followed by negative 1347 staining with 0.7% uranyl formate Representative images of VLPs were acquired 1349 with a Thermo Scientific Talos F200C transmission electron microscope operated at 200 kV and 44 1350 equipped with a Ceta CCD camera. The magnification was Flow cytometry analysis of particles. Particle concentration, size, Env+ fraction and spike 1354 density were determined by single vesicle flow cytometry (113, 114), using a commercial kit 1355 (vFC TM Assay kit Thermo Fisher) for 1h at RT 1358 and analyzed using membrane fluorescence to trigger detection. Data were analyzed using FCS 1359 Express (De Novo Software), and included calibration using a vesicle size and fluorescence 1360 intensity standards. The analysis included a pre-stain dilution series to determine the optimal 1361 initial sample dilution and multiple positive and negative controls, per guidelines of the 1362 International Society for Extracellular Vesicles (ISEV) (129). A detailed description of vFC TM 1363 methods and controls can be Figure 1. Key features of candidate Env strains. 17 strains were placed in two groups: i) group 1 strains are naturally sensitive to multiple V2 NAbs; ii) 5 group 2 strains exhibit high 1373 membrane trimer expression. Strain names are abbreviated (see Methods). An asterisk in total 1374 glycans per gp160 protomer indicates overlapping sequons in CE217 (N396 and N398) Glycan holes are listed whenever a 80% 1376 conserved glycan is absent. V2 sensitivity features are shown, including glycans involved in nAb 1377 binding or clashes, loop lengths. A double asterisk for the CNE58 V1 loop denotes a possible 1378 internal hairpin disulfide loop (Fig. S1). Rare glycans N49 and N674 are shown. Strand C 1379 sequence (AA166-171) is shown with basic residues in blue and acidic residues in red, along with 1380 overall strand C charge. The amino acid at position 500 may influence gp120/gp41 processing 1381 (gray highlights non-lysine or arginine residues) Total Env expression, as judged 1383 by SDS-PAGE-Western blot (see Fig. 3C and D) Figure 2. V2 NAb sensitivity of candidate strains. A) MAb IC50s against 17 candidate PVs 1386 bearing full-length wild-type (WT) gp160 spikes, except for WITO, AC10, 6101, KNH1144 and 1387 sc422, that were gp41 cytoplasmic tail-truncated (gp160CT). For CH01 and VRC38, UCA 1388 sensitivities are shown GnT1-PV data BB201 and CNE58 strains are not shown, as infection was insufficient. B) CH01 IC30s and % maximum CH01 neutralization saturation in unmodified (left) and GnT1-(right) Gp160CT and SOS mutations improve expression of candidate strains. A) VLP 1395 trimer expression with or without gp41 truncation (gp160CT) and SOS mutations, probed with 1396 anti-gp120 and anti-gp41 mAb cocktail. SOS gp160CT trimer expression of candidate strains 1397 visualized by B) BN-PAGE-Western blot and by SDS-PAGE-Western blot, probing with anti-1398 gp120 (C) or anti-gp41 (D) mAb cocktails. All Envs were expressed using robust expression 1399 plasmids (pVRC8400 or pCDNA3.1), except for Q23 in part A lanes 6-8 and BB201 in part B lane 1400 10, for which pCR3.1 was used Effect of mutations on JR-FL SOS gp160CT trimer expression, infectivity and 1404 NAb sensitivity. Effect of mutations on A) JR-FL gp160CT SOS trimer infectivity, and total Env 1405 expression (quantified by SDS-PAGE-Western blot), and B) mAb sensitivity. The mutant with the 1406 most desirable features is highlighted in red (lane 35). The V3 consensus Figure 5. Effects of mutants on JR-FL membrane trimer glycan maturation and occupation Glycans are colored in shades of green (high mannose) or magenta (complex), 1414 according to their score. Thus, untrimmed high mannose glycans are dark green and trimmed 1415 high mannose glycans are shown in lighter hues of green. Conversely, heavy complex glycans 1416 are shown in dark magenta, whereas smaller complex glycans are shown in lighter hues of 1417 magenta. Some glycans, rendered in gray, were not resolved in the JR-FL parent and therefore 1418 have no glycan score VLPs determined by LC-MS. Glycans were assigned scores by their degree of maturation C) Changes in glycan scores at each position between sample pairs. A negative score 1422 implies a shift to less mature glycan and vice versa. Data are only shown at positions where a Comparison of NL-Luc and pQC-Fluc assays for HIV pseudovirus infectivity and 1428 neutralization sensitivity. A) JR-FL E168K+N189A, Q23 D49N+N611A, WITO and T250 SOS 1429 gp160CT PVs, produced using NL-Luc or pQC-Fluc plasmid sets were compared for infection 1430 of CF2.CD4 CCR5 cells, assayed as RLUs. The dotted line marks an arbitrary cutoff for infection, 1431 below which data become unreliable. B) Comparative PG9 sensitivity of the same PV Effect of mutations on A) Q23 and C) WITO gp160CT SOS 1436 trimer infectivity (by the pQC-Fluc assay) and total Env expression Effect of mutations on A) T250 and C) CE217 gp160CT SOS 1441 trimer infectivity (by pQC-Fluc and NL-Luc assays, respectively) and expression (by SDS-PAGE-1442 Western blot). B) and D) mAb sensitivity of mutants Effects of Group 2 strains AC10, sc422 and KNH1144 SOS gp160CT mutations 1445 on trimer expression, infectivity and NAb sensitivity Fluc assay) and expression (by SDS-1447 PAGE-Western blot). B), D) and F) mAb sensitivity of mutants. 1448 1449 Figure 10. A quarter of particles from transfections with Env plasmid carry surface Env. 1450 293T cells were transfected with WITO SOS gp160CT and/or MuLV Gag Supernatants were precleared, filtered and 1,000-fold concentrated. Samples were probed by Upper panels show particle diameters and fluorescence intensities of 1455 samples stained with Alexa-647-labeled PGT121. In the lower panel, we show total particle counts 1456 versus Alexa-647 fluorescence. D) Total particle counts and Env+ particle counts per µl of the 1457 samples indicated (left) and % Env+ particles as a proportion of total particles (right) Tier 2 HIV-1 Neutralizing Antibodies Bind to Quaternary Epitopes Involving Glycan-Deficient 1463 Patches Proximal to the CD4 Binding Site Sequential and 1465 Simultaneous Immunization of Rabbits with HIV-1 Envelope Glycoprotein SOSIP.664 Trimers from 1466 Clades A, B and C Targeted N-1468 glycan deletion at the receptor-binding site retains HIV Env NFL trimer integrity and accelerates 1469 the elicited antibody response Broadly Neutralizing Antibodies to HIV and Their Role in Vaccine 1471 Design C3/465 glycan hole cluster in BG505 HIV-1 envelope is the major neutralizing target involved in 1474 preventing mucosal SHIV infection HIV-1 VACCINES. HIV-1476 1 neutralizing antibodies induced by native-like envelope trimers Vaccination with Glycan-Modified HIV NFL Envelope Trimer-Liposomes Elicits Broadly Neutralizing Antibodies to 1480 Multiple Sites of Vulnerability Epitope-based vaccine design 1482 yields fusion peptide-directed antibodies that neutralize diverse strains of HIV-1 Glycan Shield of the Native HIV Envelope Are a Target of Trimer-Elicited Neutralizing Antibodies Epitopes for 1488 neutralizing antibodies induced by HIV-1 envelope glycoprotein BG505 SOSIP trimers in rabbits 1489 and macaques Structure and immune 1491 recognition of trimeric pre-fusion HIV-1 Env Glycoengineering 1495 HIV-1 Env creates 'supercharged' and 'hybrid' glycans to increase neutralizing antibody potency, 1496 breadth and saturation Recent progress in broadly neutralizing antibodies to HIV Antibody Lineages with Vaccine-1500 Induced Antigen-Binding Hotspots Develop Broad HIV Neutralization Development of a 1503 3Mut-Apex-Stabilized Envelope Trimer That Expands HIV-1 Neutralization Breadth When Used 1504 To Boost Fusion Peptide-Directed Vaccine-Elicited Responses Humanized Immunoglobulin Mice: Models for HIV Vaccine Testing and 1506 Studying the Broadly Neutralizing Antibody Problem Human Ig knockin mice to study the development and 1508 regulation of HIV-1 broadly neutralizing antibodies HIV-1 Vaccines Based on Antibody Identification, B Cell Ontogeny, 1510 and Epitope Structure Targeting B-cell germlines and focusing 1512 affinity maturation: the next hurdles in HIV-1-vaccine development? Expert Rev Vaccines Identification of Common 1515 Features in Prototype Broadly Neutralizing Antibodies to HIV Envelope V2 Apex to Facilitate 1516 Vaccine Design Structures of Env V1V2 with broadly neutralizing antibodies reveal commonalities that enable vaccine 1519 design Virus-like Particles 1521 Identify an HIV V1V2 Apex-Binding Neutralizing Antibody that Lacks a Protruding Loop Developmental pathway for potent V1V2-directed HIV-neutralizing antibodies Analysis of a Clonal 1527 Lineage of HIV-1 Envelope V2/V3 Conformational Epitope-Specific Broadly Neutralizing 1528 Antibodies and Their Inferred Unmutated Common Ancestors Binding of 1530 inferred germline precursors of broadly neutralizing HIV-1 antibodies to native-like envelope 1531 trimers Antigen modification regulates competition of broad and narrow neutralizing HIV antibodies Rational HIV 1536 immunogen design to target specific germline B cell receptors Anti-HIV B Cell 1540 Lines as Candidate Vaccine Biosensors Vaccine Design to Target Germline Precursors of Glycan-Dependent Broadly Neutralizing HIV-1 VACCINES Priming a broadly neutralizing antibody response to HIV-1 using a germline-targeting immunogen Del Moral-Sanchez I, et 1548 al. Design and crystal structure of a native-like HIV-1 envelope trimer that engages multiple 1549 broadly neutralizing antibody precursors in vivo Quantification of the Impact of the HIV-1-Glycan Shield on Antibody Elicitation Effects of partially 1554 dismantling the CD4 binding site glycan fence of HIV-1 Envelope glycoprotein trimers on 1555 neutralizing antibody induction Glycoform Heterogeneity and Localized Diversity Govern the Initiation and Maturation of a V2 1558 Apex Broadly Neutralizing Antibody Lineage Neutralization-guided 1560 design of HIV-1 envelope trimers with high affinity for the unmutated common ancestor of CH235 1561 lineage CD4bs broadly neutralizing antibodies HIV-1 envelope 1563 glycan modifications that permit neutralization by germline-reverted VRC01-class broadly 1564 neutralizing antibodies Glycans Function as 1566 Anchors for Antibodies and Help Drive HIV Broadly Neutralizing Antibody Development Bacterially derived 1569 synthetic mimetics of mammalian oligomannose prime antibody responses that neutralize HIV 1570 infectivity Glycan Masking Focuses 1572 Immune Responses to the HIV-1 CD4-Binding Site and Enhances Elicitation of VRC01-Class 1573 Precursor Antibodies Consistent elicitation of 1575 cross-clade HIV-neutralizing responses achieved in guinea pigs after fusion peptide priming by 1576 repetitive envelope trimer boosting Virus-Like Particle Based Vaccines Elicit Neutralizing Antibodies against the HIV-1 Fusion Peptide. Vaccines (Basel) Functional 1583 Relevance of Improbable Antibody Mutations for HIV Broadly Neutralizing Antibody 1584 Development Development of broadly neutralizing antibodies from 1586 autologous neutralizing antibody responses in HIV infection Development of broadly neutralizing antibodies in HIV-1 infected 1589 elite neutralizers Ontogeny-based immunogens for the 1591 induction of V2-directed HIV broadly neutralizing antibodies Structural basis for broad and potent 1593 neutralization of HIV-1 by antibody VRC01 HIV-1-Neutralizing Antibodies Targeting the CD4 Supersite in 14 Donors Human Antibodies that Neutralize HIV-1: Identification, Structures, 1598 and B Cell Ontogenies Identification of a CD4-1600 Binding-Site Antibody to HIV that Evolved Near-Pan Neutralization Breadth Maturation Pathway from 1603 Germline to Broad HIV-1 Neutralizer of a CD4-Mimic Antibody Elicitation of Neutralizing Antibodies Targeting the V2 Apex of the HIV Envelope Trimer in a Wild-Type Animal 1606 Model Strategies for a multi-stage neutralizing antibody-based 1608 HIV vaccine Sequential 1610 Immunization Elicits Broadly Neutralizing Anti-HIV-1 Antibodies in Ig Knockin Mice Antibody Signatures and Application to Epitope-Targeted Vaccine Design The Chimpanzee SIV 1616 Envelope Trimer: Structure and Deployment as an HIV Vaccine Template Immunogenicity in Rabbits of HIV-1 SOSIP Trimers from Clades A, B, and C, Given Individually, 1620 Sequentially, or in Combination Rationally Designed 1622 Vaccines Targeting the V2 Region of HIV-1 gp120 Induce a Focused, Cross-Clade-Reactive, 1623 Biologically Functional Antibody Response A short 1625 segment of the HIV-1 gp120 V1/V2 region is a major determinant of resistance to V1/V2 1626 neutralizing antibodies gp120 V1/V2 domain with broadly neutralizing antibody PG9 Synthetic 1630 glycopeptides reveal the glycan specificity of HIV-neutralizing antibodies Structural basis for diverse N-glycan recognition by HIV-1-neutralizing V1-V2-directed antibody 1634 PG16 An 1638 HIV-1 antibody from an elite neutralizer implicates the fusion peptide as a site of vulnerability Site-Specific Glycosylation of Virion-Derived HIV-1 Env Is Mimicked by a Soluble Trimeric Immunogen Functional implications of the 1644 binding mode of a human conformation-dependent V2 monoclonal antibody against HIV Viral 1647 Escape from HIV-1 Neutralizing Antibodies Drives Increased Plasma Neutralization Breadth 1648 through Sequential Recognition of Multiple Epitopes and Immunotypes Stabilization of the V2 loop improves the presentation of V2 loop-associated broadly neutralizing 1652 antibody epitopes on HIV-1 envelope trimers Interdomain Stabilization 1654 Impairs CD4 Binding and Improves Immunogenicity of the HIV-1 Envelope Trimer. Cell Host 1655 Microbe A next-generation 1657 cleaved, soluble HIV-1 Env Trimer, BG505 SOSIP.664 gp140, expresses multiple epitopes for 1658 broadly neutralizing but not non-neutralizing antibodies Antigenic Map of a Cleaved Soluble HIV-1 Envelope Trimer Particulate Array 1662 of Well-Ordered HIV Clade C Env Trimers Elicits Neutralizing Antibodies that Display a Unique V2 1663 Cap Approach Murine Antibody 1665 Responses to Cleaved Soluble HIV-1 Envelope Trimers Are Highly Restricted in Specificity Multi-parameter 1668 exploration of HIV-1 virus-like particles as neutralizing antibody immunogens in guinea pigs, 1669 rabbits and macaques A comparative 1671 immunogenicity study of HIV-1 virus-like particles bearing various forms of envelope proteins, 1672 particles bearing no envelope and soluble monomeric gp120 Immunogenicity of Stabilized HIV-1 Envelope Trimers with Reduced Exposure of Non-1675 neutralizing Epitopes Differential 1677 processing of HIV envelope glycans on the virus and soluble recombinant trimer Global site-specific N-1680 glycosylation analysis of HIV envelope glycoprotein Composition 1682 and Antigenic Effects of Individual Glycan Sites of a Trimeric HIV-1 Envelope Glycoprotein Structure and Immune Recognition of the HIV Glycan 1685 Shield Microscopy-Based Epitope Mapping Defines Specificities of Polyclonal Antibodies Elicited during 1688 HIV-1 BG505 Envelope Trimer Immunization Enhancing glycan 1690 occupancy of soluble HIV-1 envelope trimers to mimic the native viral spike Structural basis for membrane 1693 anchoring of HIV-1 envelope spike The Cytoplasmic Tail Slows the Folding of Human Immunodeficiency Virus Type 1 Env from a Late Prebundle 1696 Configuration into the Six-Helix Bundle Guided Redesign Increases the Propensity of HIV Env To Generate Highly Stable Soluble Trimers Glycine 1701 Substitution at Helix-to-Coil Transitions Facilitates the Structural Determination HIV-1 Envelope Trimer Design and Immunization 1704 Strategies To Induce Broadly Neutralizing Antibodies Soluble Prefusion Closed 1706 DS-SOSIP.664-Env Trimers of Diverse HIV-1 Strains A Universal 1708 Approach to Optimize the Folding and Stability of Prefusion-Closed HIV-1 Envelope Trimers Relationship of HIV-1 and 1711 SIV envelope glycoprotein trimer occupation and neutralization Incorporation of high levels of 1713 chimeric human immunodeficiency virus envelope glycoproteins into virus-like particles Dense Array of Spikes on HIV Virion Particles Trimer Affects Accessibility to Broadly Neutralizing Antibodies at Its Apex Identification and 1720 characterization of a naturally occurring, efficiently cleaved, membrane-bound, clade A HIV-1 Env, 1721 suitable for immunogen design Variable loop glycan dependency of the broad and potent HIV-1-1726 neutralizing antibodies PG9 and PG16 Envelope Glycan Shield at Transmission Determines Neutralization Breadth Cryo-EM Structure of a 1731 Fully Glycosylated Soluble Cleaved HIV-1 Envelope Trimer Evolution of 1733 an HIV glycan-dependent broadly neutralizing antibody epitope through immune escape Characterizing 1736 anti-HIV monoclonal antibodies and immune sera by defining the mechanism of neutralization Redox-triggered 1739 infection by disulfide-shackled human immunodeficiency virus type 1 pseudovirions Trimer Apex Stabilize the HIV-1 Envelope in a Closed, Antibody-Protected Conformation Quantitative assessment of the preferences for the amino acid residues 1745 flanking archaeal N-linked glycosylation sites Characterization of structural features and diversity of variable-region determinants of related 1748 quaternary epitopes recognized by human and rhesus macaque monoclonal antibodies 1749 possessing unusually potent neutralizing activities Envelope CD4 Binding Loop Reveals Residues Controlling Distinct Trimer Conformations. PLoS 1752 Pathog Co-evolution 1754 of HIV Envelope and Apex-Targeting Neutralizing Antibody Lineage Provides Benchmarks for 1755 Vaccine Design SARS-CoV-2 spike-protein 1757 D614G mutation increases virion spike density and infectivity Initiation of HIV neutralizing 1759 B cell lineages with sequential envelope immunizations Envelope residue 375 substitutions 1761 in simian-human immunodeficiency viruses enhance CD4 binding and replication in rhesus 1762 macaques Nature of 1764 Nonfunctional Envelope Proteins on the Surface of Human Immunodeficiency Virus Type 1 Methamphetamine 1767 use alters human plasma extracellular vesicles and their microRNA cargo: An exploratory study Generation and Application 1770 of a Reporter Cell Line for the Quantitative Screen of Extracellular Vesicle Release. Front 1771 Pharmacol Systematic analysis of intracellular trafficking 1773 motifs located within the cytoplasmic domain of simian immunodeficiency virus glycoprotein 1774 gp41 Envelope Glycoproteins Adopt Downstream Conformations That Remain Responsive to 1777 Conformation-Preferring Ligands Changes in 1779 Structure and Antigenicity of HIV-1 Env Trimers Resulting from Removal of a Conserved CD4 1780 Binding Site-Proximal Glycan Enhancing exposure of HIV-1 neutralization epitopes 1782 through mutations in gp41 Development of an anti-HIV vaccine eliciting broadly neutralizing 1784 antibodies Trimer Apex Reveals Key Hydrophobic Constraints That Maintain the HIV-1 Envelope Spike in a 1787 Closed State. mBio Folding of the human immunodeficiency virus type 1 envelope 1789 glycoprotein in the endoplasmic reticulum Enhancing the 1791 proteolytic maturation of human immunodeficiency virus type 1 envelope glycoproteins A recombinant human 1794 immunodeficiency virus type 1 envelope glycoprotein complex stabilized by an intermolecular 1795 disulfide bond between the gp120 and gp41 subunits is an antigenic mimic of the trimeric virion-1796 associated structure Exploiting glycan 1798 topography for computational design of Env glycoprotein antigenicity Integrity of Glycosylation Processing of a Glycan-Depleted Trimeric HIV-1 Immunogen Targeting 1802 Key B-Cell Lineages HIV-1 Glycan 1804 Density Drives the Persistence of the Mannose Patch within an Infected Individual Networks of HIV-1 Envelope Glycans Maintain Antibody Epitopes in the Face of Glycan Additions 1808 and Deletions Conserved Role of an N-Linked Glycan on 1810 the Surface Antigen of Human Immunodeficiency Virus Type 1 Modulating Virus Sensitivity to 1811 Broadly Neutralizing Antibodies against the Receptor and Coreceptor Binding Sites MIFlowCyt-EV: a framework for standardized reporting of extracellular vesicle flow cytometry 1815 experiments HIV Superinfection 1817 Drives De Novo Antibody Responses and Not Neutralization Breadth