Anti-V2 antibodies virus vulnerability revealed by envelope V1 deletion in HIV vaccine candidates.

The efficacy of ALVAC-based HIV and SIV vaccines in humans and macaques correlates with antibodies to envelope variable region 2 (V2). We show here that vaccine-induced antibodies to SIV variable region 1 (V1) inhibit anti-V2 antibody-mediated cytotoxicity and reverse their ability to block V2 peptide interaction with the α4β7 integrin. SIV vaccines engineered to delete V1 and favor an α helix, rather than a β sheet V2 conformation, induced V2-specific ADCC correlating with decreased risk of SIV acquisition. Removal of V1 from the HIV-1 clade A/E A244 envelope resulted in decreased binding to antibodies recognizing V2 in the β sheet conformation. Thus, deletion of V1 in HIV envelope immunogens may improve antibody responses to V2 virus vulnerability sites and increase the efficacy of HIV vaccine candidates.

Introduction

The HIV recombinant canarypox-derived vector (ALVAC) and gp120-envelope proteins formulated in alum vaccine platform tested in the RV144 HIV vaccine trial was the first to reduce the risk of HIV acquisition in humans (31.2%) (Rerks-Ngarm et al., 2009). Serum IgG to the gp70-V1/V2 scaffold (Haynes et al., 2012) and to linear V2 peptides (Gottardo et al., 2013; Zolla-Pazner et al., 2014) have been identified as correlates of reduced risk of HIV acquisition. Sieve analysis further demonstrated genetic markers of immunologic pressure at positions 169 and 181 (Rolland et al., 2012) of V2, a region that binds to the α4β7 integrin (Lertjuthaporn et al., 2018). V2 is structurally polymorphic and can adopt β strand or α-helical conformations. However, V2 interaction with the α4β7 integrin is inhibited preferentially by antibodies recognizing its α-helical conformation (Lertjuthaporn et al., 2018).

The SIVmac251 macaque model, in which vaccinated animals are mucosally exposed to the highly pathogenic SIVmac251 at a dosage far in excess of HIV transmission in humans, recapitulated the modest vaccine efficacy observed in RV144 and identified antibodies to V2 as a correlate of reduced risk of SIV acquisition (Pegu et al., 2013; Vaccari et al., 2016). Furthermore, substitution of the alum adjuvant with MF59 (Vaccari et al., 2016) in the same animal model abolished the vaccine protection afforded by the ALVAC/gp120 vaccine platform, thereby predicting the recently announced lack of efficacy in the HVTN-702 HIV trial that used the MF59 adjuvant (Cohen, 2020).

The need to develop immunogens able to increase the efficacy of HIV vaccine candidates remains urgent. We found that antibodies to V1 interfere with the cytocidal function of antibodies to V2 and demonstrate here that the removal of V1, engineered to favor an α-helical conformation of V2, increases V2-specific cytocidal antibodies appearing to increase vaccine efficacy.

Results

Antibodies to a V1 region adjacent to V2 (V1a) are associated with increased SIVmac251 acquisition

We investigated the serum antibody responses to V1 and V2 using linear peptide arrays in a cohort of 78 macaques immunized with four different vaccine regimens and exposed by the same route to the same dose of an identical SIVmac251 stock (Figure S1A). The gp120 protein bivalent boost was adjuvanted in alum in three regimens (ALVAC-SIV/gp120 + alum, DNA-SIV/ALVAC-SIV/gp120 + alum, and Ad26-SIV/ALVAC-SIV/gp120 + alum) and in MF59 in the fourth (ALVAC-SIV/gp120 + MF59) (Vaccari et al., 2016, 2018). The efficacy of these regimens was evaluated as the average per-challenge risk of SIVmac251 acquisition compared with unvaccinated controls following intrarectal exposure to repeated, low doses of the virus. For simplicity, we hereafter refer to ALVAC-SIV/gp120 + alum and the DNA-SIV/ALVAC-SIV/gp120 + alum (with respective vaccine efficacies of 44% and 52%; p < 0.05) as the protective regimens (Figures S1B and S1C), and to ALVAC-SIV/gp120 + MF59 and Ad26-SIV/ALVAC-SIV/gp120 + alum (vaccine efficacies of 9% and 13%; p > 0.05) as non-protective regimens (Figures S1D and S1E) (Vaccari et al., 2016, 2018).

The levels of sera antibody reactivity to overlapping linear V1 (Starcich et al., 1986) peptides 15–24 (Figure S1F) did not differ between protective and non-protective vaccines (Figure S1G). However, when all vaccinated macaques were analyzed, animals with above-median antibody levels to V1 had a trend toward increased risk of SIVmac251 acquisition (p = 0.0658; Figure S1H). Analysis of antibody responses to V1 peptides in protective and non-protective vaccine subgroups showed that anti-V1 antibodies were associated with an increased risk of SIVmac251 acquisition only in the non-protective group (Figures 1A and S1I). Antibody responses to V1 peptides 23 and 24, encompassing the amino acid segment NETSSCIAQNNCTGLEQEQMISCKF, revealed higher reactivity in the non-protective vaccine subgroup when compared with the protective group (Figure 1B). The region designated here as V1a (peptide 23 to 24) lies directly N-terminal to a cryptic α4β7 integrin-binding site (Tassaneetrithep et al., 2014) in V2 (V2b), with both V1a and V2b being part of a continuous, exposed peptide segment at the extreme apex of the envelope trimer (Gorman et al., 2016; Julien et al., 2013; Liu et al., 2008; Pancera et al., 2014) (Figures 1C and 1D). V1a is also in tertiary contact with the V2 region (V2c) that contains the canonical tripeptide shown to bind to the α4β7 integrin (Arthos et al., 2008; Nakamura et al., 2012), as depicted in Figure 1D in a 3D homology model of the SIVmac251 trimer based on the HIV BG505 cryo-EM structure. The V1/V2 domain of a related cryo-EM structure of SIVcpzPtt is nearly identical (Andrabi et al., 2019). Based on these structural relationships, we hypothesized that antibodies to V1a may influence vaccine efficacy by interfering with antibody binding to V2 and tested different assays of anti-V1 and anti-V2-specific monoclonal antibodies (mAbs) cloned from vaccinated protected or vaccinated SIV-infected animals (Mason et al., 2016).

Figure 1

V1 Antibody levels are associated with increased risk of SIVmac251 acquisition

(A) Time of acquisition in macaques immunized with non-protective vaccines and mean O.D. sum of serum responses to V1 peptides 15–24 at week 27: high (n = 19) and low (n = 20) anti-V1 values.

(B) Serum antibody against peptides 23 and 24 (V1a) in animals vaccinated with protective (diamonds, n = 39) and non-protective (inverted triangles, n = 39) vaccines 3 weeks after the last immunization and 1 week before challenge (week 27), data shown as mean with SD.

(C) Amino acid sequence of V1 and V2 (SIVmac251- K6W). Sequences are represented as follows: V1 (black); V1a (black dotted line); V2 (red); V2b (red dotted line; recognized by mAb NCI09); V2c (green dotted line; recognized by mAb NCI05). MAb ITS41 recognizes the V1a epitope, and NCI04 and NCI06 recognize amino acids in the N-terminal region of V1.

(D) Spatial relationship of V1 (olive), V2b (red), V2c (green), and the canonical V2 tripeptide, DLV, that binds the α4β7 integrin (purple) in the gp120 trimer.

(E) NCI05 and NCI09 binding to SIVmac239-infected A66 p24 Gag-positive cells in a representative experiment of staining of SIVmac251-infected cells.

(F) The average percentage of Gag and NCI05 positive (green) or NCI09 positive (red, n = 2), data shown as mean with SD.

(G) SIVmac251 virion capture assay (n = 7 experiments, black dots): virion input (gray) or virion captured by beads coated with NCI05 (green), NCI09 (red), or mouse IgG (negative control) (Rhesus IgG isotype was also used as a negative control, with n = 3, mean = 5.09 SIV RNA in transformed log copies/mL and SD = 0.08. Both NCI05 and NCI09 mAb were statistically higher than Rhesus IG control, data not shown). Data in the graph are shown as mean with SD. Statistical analyses comparing two groups was done using Mann-Whitney test; when comparing three groups or more Kruskal-Wallis test with Dunn’s multiple comparison test was used, and the infection curves were analyzed using Log Rank (Mantel-Cox test).

See Figure S1.

Antibody to V1a decreases anti-V2 antibody cytotoxicity and ability to inhibit gp120 binding to α4β7

The mAbs (NCI09 and NCI05) cloned from the vaccinated and protected animal P770 (Vaccari et al., 2016) were cross-reactive with SIVmac251 and SIVsmE543 gp120, V1/V2 scaffolds, and cyclic V2 peptides (Figures S2A–S2C). Linear peptide mapping (Figure S2D), peptide competition (Figures S2E and S2F), and crystallography (Figures S3 and S4; PDB: 6VRY) demonstrated that NCI09 recognized the TGLKRDKTKEY epitope in V2b. NCI05 did not bind to linear SIVmac251-K6W peptides (Figure S2C), but its binding to cyclic V2 was competed by peptides encompassing the SIVmac239TGLKRDKKKEYNETWYSAD amino acid sequence (Figure S2F). From the same animal, we also obtained two V1-specific mAbs, NCI04 and NCI06, recognizing the CNKSETDRWGLTK epitope located N-terminal to V1a (Figures 1C and S2C). None of these mAbs neutralized tier 2 SIVmac251 or SIVSME660. NCI05 neutralized tier 1 SIVSME660 but not tier 1 SIVmac251. NCI06 and NCI09 had low neutralizing activity against tier 1 SIVmac251 (Figure S2G). Both NCI05 and NCI09 bound to gp120 on the surface of Gag-positive SIVmac239-infected cells (Figures 1E, 1F, S5A, and S5B) and to SIVmac251 virions (Figure 1G). Functionally, NCI05 and NCI09 mAbs inhibited SIV gp120 binding to the α4β7 integrin in a cell adhesion assay (Lertjuthaporn et al., 2018; Wibmer et al., 2018) (Figure 2A) and mediated antibody-dependent cell-mediated cytotoxicity (ADCC; Figure 2B). Of interest, the mAb ITS41 recognizing V1a (Mason et al., 2016) inhibited binding of NCI09 to the gp120 on the surface of SIVmac251-infected cells (Figure 2C), as well as the binding of NCI05 and NCI09 mAbs to gp120 SIVmac251-M766 (Lertjuthaporn et al., 2018; Wibmer et al., 2018) (Figures 2D and 2E). Prebound NCI05 and NCI09 were not affected by ITS41, demonstrating asymmetric competition (Figures S5C and S5D). The NCI06 mAb, which recognizes a peptide distal to V1a not in contact with V2b or V2c in the 3D envelope structure (Figures 1C and 1D), did not interfere with NCI09 binding to gp120 (Figures S5E and S5F). In addition, increasing amounts of ITS41 reversed NCI09 inhibition of gp120 binding to the α4β7 integrin (Figures 2F and S5G) and inhibited NCI05 and NCI09-mediated ADCC (Figures 2G and 2H). NCI04 did not affect NCI05 or NCI09-mediated ADCC (Figures 2I and 2J). Of interest, both ITS41 and NCI04 mediate ADCC to a much lower extent than NCI05 and NCI09 despite having the identical Fc region since all antibodies were cloned in an expression vector that joined variable regions with the same Fc scaffold. This result highlights the importance of Fab properties and epitope accessibility for ADCC.

Figure 2

Functional activity of V1 and V2 mAbs

(A) Adhesion of gp120 SIVmac251-M766 to the α4β7 integrin only (dark gray) in the presence of vedolizumab, a mAb anti-α4β7 integrin used as a positive control at a concentration of 0.5 μg/mL (light gray), or in the presence of NCI05 (green) and NCI09 (0.25 μg/mL; red). Data shown as mean with SEM.

(B) ADCC mediated by NCI05 and NCI09 in the CEM-based assay (n = 3), data shown as mean with SD.

(C) Competition of NCI09 binding to SIVmac251-infected CD4+ T cells by increasing amounts of ITS41 (0, 2.5, and 5 ng/mL). The dot plot displayed is of one representative experiment (n = 2).

(D and E) Inhibition of gp120 SIVmac251-M766 binding to mAbs (D) NCI05 or (E) NCI09 by ITS41, or albumin and mAb ITS01 with CD4 binding specificity as controls.

(F) Adhesion of SIVmac251-M766 to α4β7 integrin in the absence or presence of NCI09 (1.25μg/mL), or with increasing concentrations of prebound ITS41, followed by 1.25 μg/mL of NCI09. As the concentration of ITS41 is increased, the inhibitory activity of NCI09 is lost. Data shown as mean with SD.

(G) Inhibition of ADCC mediated by mAbs NCI05 and (H) NCI09 by increasing amounts of ITS41 (n = 3).

(I and J) NCI04 does not compete with (I) NCI05 or (J) NCI09-mediated ADCC (n = 3). Data shown as mean with SD. Statistical analyses were performed using two-way ANOVA test with Dunnett’s multiple comparison test.

See Figures S2–S5.

V1-deleted immunogens designed to favor V2 α-helical conformation reduce the risk of SIVmac251 acquisition

The functional interference of ITS41 with NCI05 and NCI09 raised the hypothesis that deletion of V1 in SIV/HIV envelope immunogens could increase V2 accessibility, enhance the level of V2 functional antibodies, and increase vaccine efficacy. To test this, we designed V1-deleted gp120 proteins (gp120ΔV1) by symmetrically truncating V1 at its stem, since its origin and insertion (stem) to the V1/V2 domain connect the A and B β strands (McLellan et al., 2011). The V1/V2 domain remaining after deletion of the gp120 V1 (gp120ΔV1) was energy minimized as previously described (Abagyan and Totrov, 1994; Cardozo et al., 1995). The search predicted a stable, low-energy, partially α-helical V2 conformation in gp120 (Figure 3A and Table S1). As control, we designed another V1-deleted gp120 (gp120ΔV1gpg) by inserting the Gly-Pro-Gly β turn at the excision point with the purpose of minimizing disruption to the crystallographically visualized V1/V2 Greek key β sheet fold (Figures 3B and Table S2). We then expressed M766-based gp120ΔV1 and gp120ΔV1gpg proteins in Chinese Hamster Ovary (CHO) cells together with wild-type gp120 (gp120WT; Table S3). The purified monomeric gp120ΔV1 and gp120ΔV1gpg proteins were stable and unrecognized by the anti-V1 NCI06 and ITS41 mAbs (Figure S5H), bound to NCI05 and NCI09 by ELISA (Figures 3C–3E), immune precipitation, and western blot better than gp120WT (Figures S5H and S5I). Of interest, gp120ΔV1 and gp120ΔV1gpg also bound better to simian soluble CD4 than the gp120WT (Figure 3F). These data are consistent with increased exposure of V2 epitopes and the CD4-binding site in the gp120ΔV1 and gp120ΔV1gpg antigens (Ching and Stamatatos, 2010).

Figure 3

Vaccine design and virological outcome

(A) Model of the gp120ΔV1V2 α helix (green). The predicted α-helical structure of the V2c peptide was imposed on the V2c segment in gp120 and the local backbone energy minimized.

(B) Model of the gp120ΔV1gpg β strand.

(C–E) ELISA binding of the purified (C) gp120WT, (D), gp120ΔV1, and (E) gp120ΔV1gpg with α-V2 mAbs NCI05 and NCI09 and the α-V1 mAb NCI06.

(F) Binding of simian CD4-Ig to gp120WT, gp120ΔV1, and gp120ΔV1gpg.

(G) Schematic representation of the study design. Each vaccinated group included 14 young macaques, and the control group consisted of 18 naive young macaques. All animals were simultaneously exposed to weekly low doses of SIVmac251 by the intrarectal route beginning at week 17.

(H) Risk of SIVmac251 acquisition in animals immunized with ΔV1 envelope immunogens.

(I) Binding of week 17 (minus the baseline) serum to V2cE543 peptide in immunized animals (WT, n = 14; ΔV1, n = 14; ΔV1gpg, n = 13), data shown as mean with SD.

(J) Correlation between serum antibodies to V2cE543 and number of intrarectal challenges (WT, n = 14; ΔV1, n = 14; ΔV1gpg, n = 13).

(K) Correlation of serum inhibition of the α4β7 integrin (expressed on RPMI8866 cells) to V2cE543 and intrarectal challenges (WT, n = 2; ΔV1, n = 2; ΔV1gpg, n = 4). The correlation was performed only with data from animals that had inhibition above the assay cutoff. The infection curves were analyzed using Log Rank (Mantel-Cox test); data comparison between the three vaccinated groups was done with non-parametrical Kruskal-Wallis test with Dunn’s multiple comparison test. The correlation analyses were performed using the non-parametric Spearman rank correlation method with the exact permutation two-tailed p-values calculated.

See Tables S1–S6 and Figures S5–S7.

We tested the efficacy of the V1-deleted immunogens using a DNA-SIV-prime/ALVAC-SIV/with gp120 protein + alum monovalent boost regimen followed by low-dose intrarectal exposures to SIVmac251. We designed a modified vaccine regimen aimed at magnifying a possible difference in the efficacy of the wild type and ΔV1 immunogens. Here, we halved the amount of SIV Gag DNA in the prime and performed a single protein boost (rather than two) with ALVAC-SIV using the SIVmac251 gp120M766 alone, omitting the two SIVSME543 gp120GC7V protein boosts (Vaccari et al., 2016, 2018). In previous studies, the association between antibodies to V2 and a decreased risk of SIVmac251 acquisition were notably revealed by SIVSME543 antigens but, curiously, not by SIVmac251 antigens.

We vaccinated three groups of 14 macaques each with two inoculations (weeks 0 and 4) of plasmid DNAs expressing SIV gp160WT (group 1; Table S4), SIV gp160ΔV1 (group 2; Table S5), or SIV gp160ΔV1gpg (group 3; Table S6) together with SIV p57 Gag. All groups received one boost at week 8 with ALVAC-SIV expressing gp120WT, and each group was administered a final boost at week 12 consisting of the same ALVAC-SIV together with the SIVmac251-M766 gp120WT (group 1), gp120ΔV1 (group 2), or gp120ΔV1gpg (group 3) protein adjuvanted with alum Alhydrogel (Figure 3G). Alongside a simultaneous control group of 18 naive macaques, the vaccinated macaques were exposed weekly to a total of 11 low doses of SIVmac251 by the intrarectal route, beginning at 5 weeks from the last immunization (week 17). A significant decrease in the risk of SIVmac251 acquisition was observed following immunization with the gp160ΔV1 DNA and gp120ΔV1 protein immunogens engineered predominantly to favor the α-helical V2 conformation (vaccine efficacy 57%; p = 0.04; Figure 3H) but not following vaccination with wild-type envelope immunogens in group 1 or ΔV1gpg envelope immunogens in group 3 (Figures S6A and S6B). We observed no sustained, significant difference in the level of plasma viral RNA in animals that became infected in each vaccinated group compared with controls (Figures S6C–S6I) and only a transient trend of decreased SIV DNA levels in the rectal mucosa in group 2 (Figure S6J). The lack of vaccine efficacy in group 1 was not entirely unexpected given both the decreased amount of DNA used in the prime and, perhaps more importantly, the omission of the two SIVSME660 gp120GC7V boosts, as the antibody level to cyclic V2E543 was a main correlate of reduced risk in two independent studies (Vaccari et al., 2016, 2018). The different outcomes in groups 2 and 3 compared with controls suggested that the inferred differences in the V2 conformation of the ΔV1 and ΔV1gpg immunogens might have quantitatively or qualitatively affected the antibody response to V2.

Antibodies inhibiting V2-α4β7 interaction are associated with a decreased risk of SIVmac251 acquisition

Protection in RV144 correlated with a non-glycosylated V2 peptide that adopts an α-helical 3D conformation (Aiyegbo et al., 2017) and encompasses the canonical α4β7 integrin-binding site in V2. In previous macaque studies, Ab binding to the V2 SIVSME543 peptide (but not SIVmac251) was associated with a decreased risk of SIVmac251 acquisition (Vaccari et al., 2016, 2018). We therefore engineered an isolated V2 peptide (V2cE543) corresponding to the V2c region of SIVSME543 (a clone of SIVSME660) by identifying the beginning and ending amino acids between V2 positions 165 and 181 that adopt an α-helical conformation. Serum reactivity to the V2cE543 peptide (DKKIEYNETWYSRD) was higher (trend) in animals immunized with the ΔV1 immunogens (Figure 3I) and correlated with a decreased risk of SIVmac251 acquisition (R = 0.36, p = 0.02; Figure 3J). Furthermore, the level of inhibition of V2cE543 binding to the α4β7 integrin in the eight animals that had values above the cutoff of the assay correlated with a decreased risk of SIVmac251 acquisition (R = 0.73, p = 0.046; Figures 3K and S7A). Reactivity to an equivalent SIVmac251 peptide did not reveal any association with risk of acquisition (data not shown). These animals were all immunized with SIVmac251-based immunogens, suggesting that the V2cE543 conformation is better able to capture antibodies associated with a decreased risk of SIVmac251 acquisition, in agreement with prior observations (Vaccari et al., 2016, 2018). Serum recognition of SIVmac251 V2 linear (Figures S7B–S7D) or cyclic peptide (Figure S7E), of the entire gp120 peptide array (sorted as responses to V4, C3, and C5, with responses to C3 and C5 being highest in animals immunized with the ΔV1 immunogens; Figures S7F–S7H), had no apparent association with a decreased risk of SIVmac251 acquisition.

ADCC to gp120-coated cells or SIV-infected cells correlates with a decreased risk of SIVmac251 acquisition

ADCC was a secondary correlate of reduced risk in individuals with low IgA levels in RV144 (Tomaras et al., 2013). ADCC activity mediated by the plasma from animals in the WT, ΔV1, and ΔV1gpg groups was performed using the target EGFP-CEM-NKr-CCR5-SNAP cells (Orlandi et al., 2016) (T lymphoblastoid cell line CEM-based assay) coated with purified gp120WT, gp120ΔV1. Analysis of the coated cells demonstrated that NCI05 or NCI09 bound less well to CEM cells coated with gp120WT than those coated with gp120ΔV1 and gp120ΔV1gpg (Figures S8A–S8D). However, ADCC measured with cells coated with the gp120WT did not differ among the animal groups (Figures 4A–4C), suggesting that V1 is not a major target of ADCC. Animals vaccinated with the ΔV1 immunogen (group 2) mounted significantly higher ADCC titers directed to ΔV1 than to the ΔV1gpg antigen (Figure 4B). Animals immunized with the ΔV1gpg immunogen (group 3) mounted lower ADCC titers directed to ΔV1gpg than to ΔV1 antigens (Figures 4C and Table S7). ADCC was also performed using target cells infected with SIVmac251 (Lewis et al., 2019), and no differences were observed among the animal groups in this assay (Figure S8E). These data demonstrate that the two V1-deleted immunogens differ both in their ability to induce and to reveal cytotoxicity activity when used in the CEM-based ADCC assay.

Figure 4

ADCC directed to ΔV1 gp120 associated with decreased risk of SIVmac251 acquisition

(A–C) CEM-based ADCC titers in animals immunized with (A) WT, (B) ΔV1, or (C) ΔV1gpg envelope immunogens on target cells coated with gp120WT, gp120ΔV1, or gp120ΔV1gpg (WT, n = 14; ΔV1, n = 14; ΔV1gpg, n = 13 animals). Data shown as mean with SD.

(D–F) Correlation of ADCC titers directed to the ΔV1 antigen in animals immunized with (D) WT, (E) ΔV1, or (F) ΔV1gpg envelope immunogens and time of SIVmac251 acquisition (WT, n = 14; ΔV1, n = 14; ΔV1gpg, n = 13 animals).

(G and H) Correlation of ADCC titers in animals immunized with (G) ΔV1gpg or (H) WT envelope immunogens on target cells coated with gp120WT and time of SIVmac251 acquisition (WT, n = 14; ΔV1gpg, n = 13 animals).

(I) Correlation of percentage of specific ADCC killing of SIVmac251-infected cells in animals immunized with ΔV1 envelope immunogens and time of SIVmac251 acquisition.

(J) Correlation of ADCC titers on the CEM-based assay gp120ΔV1 and ADCC measured on SIVmac251-infected cells in animals immunized with ΔV1 immunogens.

(K) Correlation of ADCC titers (week 17) on target CEM cells coated with gp120ΔV1 in animals immunized with ΔV1 immunogens and CCR5+α4β7– Th2 CD4 T cells (from PBMC collected at week 13; ΔV1, n = 14). Data comparison between the three vaccinated groups was done with non-parametrical Kruskal-Wallis test with Dunn’s multiple comparison test, and the correlation analyses were performed using the non-parametric Spearman rank correlation method with exact permutation two-tailed p-values calculated.

See Table S7 and Figures S8.

We performed correlation analyses using the non-parametric Spearman test to assess the relationship of ADCC titers to the three antigens and the risk of SIVmac251 acquisition and found that ADCC directed to gp120ΔV1 protein is significantly correlated with a reduced acquisition risk in the WT and ΔV1-vaccinated groups, and a correlation (trend) was observed in the ΔV1gpg group (WT, R = 0.59, p = 0.03; ΔV1, R = 0.76, p = 0.003; ΔV1gpg, R = 0. 51, p = 0.08; Figures 4D–4F and Table S7), indicative of the ΔV1 antigen's ability to capture protective antibodies. In group 3, a correlation with decreased acquisition was also observed with ADCC directed to WT gp120 (R = 0. 61, p = 0.03; Figures 4G and Table S7). Strikingly, there was a correlation with ADCC directed to WT and an increased risk of SIVmac251 acquisition in animals immunized with the WT immunogens (R = −0.55; p = 0.04; Figures 4H and Table S7), suggesting that ADCC leading to a decreased risk of viral infection in all groups is better elicited and exhibited in the CEM-based ADCC assay by the ΔV1 protein.

ADCC activity measured on SIVmac251-infected cells demonstrated no difference among the vaccinated groups (Figure S8E). However, the percentage of specific ADCC killing of SIVmac251-infected cells correlated significantly with a decreased risk of SIVmac251 acquisition in the non-parametric Spearman test only in the ΔV1-immunized group (R = 0.58, p = 0.03; Figure 4I). In this group, the level of ADCC directed to CEM cells coated with the ΔV1 protein correlated (trend) with the ADCC measured against infected cells (R = 0.43, p = 0.13; Figure 4J) and the frequency of vaccine-induced T helper (Th) 2 cells (R = 0.72, p = 0.007; Figures 4K, S8F, and S8G), suggesting that Th2 cells promote protective ADCC activity.

V2-specific ADCC, but not neutralizing antibody titers, correlates with a decreased risk of SIVmac251 acquisition

Next, we examined the contribution of anti-V2 antibodies to the ADCC measured with the CEM-based assay by using purified NCI05 and NCI09 F(ab’)2 as competitor, since both antibodies proved equally capable of mediating ADCC against gp120ΔV1-coated cells (Figures 2A and 2B). Both the NCI05 and NCI09 F(ab’)2 competed approximately 40% of serum ADCC directed to the gp120ΔV1 antigen in animals immunized with this immunogen (group 2; Figures 5A and 5B and Figures S8H–S8K). Of importance, the V2-specific serum ADCC activity inhibited by NCI05 or NCI09 F(ab’)2 in the ΔV1-immunized group (NCI05 mean delta = 15.12% +/− 2.96; NCI09 = 14.09% +/− 5.21) correlated with a decreased risk of SIVmac251 acquisition (NCI05, R = 0.67, p = 0.01; NCI09, R = 0.54, p = 0.05; Figures 5C and 5D), whereas the remaining non-V2-specific ADCC activity did not (data not shown).

Figure 5

V2-specific ADCC associated with decreased risk of SIVmac251 acquisition

(A and B) ADCC killing percentage in animals immunized with ΔV1 immunogens on gp120ΔV1-coated target cells following pre-incubation of target cells with (A) NCI05 F(ab’)2 or (B) NCI09 F(ab’) (ΔV1, n = 14).

(C and D) The V2-specific ADCC inhibited by (C) NCI05 or (D) NCI09 F(ab’)2 correlated with a decreased risk of SIVmac251 acquisition in the ΔV1 group (ΔV1, n = 14).

(E and F) ADCC killing percentage in animals immunized with ΔV1gpg immunogens on gp120ΔV1--coated target cells following pre-incubation of target cells with (E) NCI05 F(ab’)2 or (F) NCI09 F(ab’) (ΔV1gpg, n = 13).

(G and H) Correlation of V2-specific ADCC inhibited by (G) NCI05 or (H) NCI09 F(ab’)2 in the ΔV1gpg group with risk of SIVmac251 acquisition (ΔV1gpg, n = 14).

(I and J) Comparison of V2-specific ADCC revealed by (I) NCI05 F(ab’)2 or (J) NCI09 F(ab’) competition in the ΔV1 and ΔV1gpg immunized animals.

(K and L) Correlation of serum reactivity at the end of the immunization to linear V2 peptides (K) 27 and (L) 29 (Figure S7B) and level of SIV DNA in rectal mucosa at 2 weeks after infection in animals vaccinated with protective vaccine regimens (including ΔV1) that became infected (n = 46).

(M and N) Correlation of serum reactivity at the end of the immunization to V2 peptides (M) 27 and (N) 29 and level of SIV DNA in rectal mucosa at 2 weeks after infection in animals vaccinated with non-protective vaccines (including the WT and ΔV1gpg immunized animals in the current study) that became infected (n = 60). Data represented as mean with SD. Data comparisons between two paired or unpaired groups were done with Wilcoxon signed-rank test and Mann-Whitney test, respectively. The correlation analyses were performed using the non-parametric Spearman rank correlation method with exact permutation two-tailed p-values calculated.

See Table S8 and Figures S1A–S1D, S6A–S6B, S7, and S8.

An identical analysis of the sera of animals immunized with the ΔV1gpg immunogen demonstrated approximately 20% inhibition by both mAbs F(ab’) (NCI05 mean delta = 8.60% +/− 3.05; NCI09 = 8.74% +/− 3.18; Figures 5E, 5F, S8L, and S8M). We observed no correlation with V2-specific ADCC activity inhibited by NCI05 or NCI09 (delta) and the risk of SIVmac251 acquisition in this group (Figures 5G and 5H). The level of estimated V2-specific serum ADCC activity inhibited by either NCI05 or NCI09 F(ab’)2 was significantly higher in the ΔV1 than the ΔV1gpg group (NCI05: p = 0.0001; NCI09: p = 0.0040; Figures 5I and 5J). Extension of our analyses to antibody-dependent cellular phagocytosis (ADCP) and antibody-dependent neutrophil activation (ADNP) using beads coated with the gp120ΔV1 protein (Mahan et al., 2016) revealed that none of these responses correlated with SIVmac251 acquisition (Figures S8N and S8O and data not shown). Serum neutralizing antibody titers against the tier 2 SIVmac251-CS41 were highest in the ΔV1 group as expected (Ching and Stamatatos, 2010) (Figure S8P) but did not correlate with the risk of infection (data not shown). Neutralizing titers to tier 1B SIVmac251-M766 did not differ among the immunization groups (Figure S8Q), but strikingly, their levels correlate with an increased risk of SIVmac251 acquisition in the ΔV1 group (Table S8 and Figure S8R). These data support the idea that the ΔV1 immunogen engineered to favor a V2 α-helical conformation to elicit qualitatively different V2-specific ADCC titers providing a plausible explanation for the difference in vaccine efficacy observed in groups 2 and 3.

Antibodies to V2 correlate with SIV DNA in the mucosa

We tested whether the cytocidal function of V2-specific antibody limits the early seeding of the virus in vaccinated animals that become infected. First, we measured serum reactivity to linear V2 peptides 27 and 29 (corresponding to V2b and V2c) in animals immunized with protective vaccines, including group 2 described in the current study (Figures 3H, S1B, and S1C), or with non-protective vaccines, including groups 1 and 3 of the current study (Figures S6A, S6B, S1D, and S1E). An inverse correlation with serum reactivity and the level of SIV DNA in the rectal mucosa was found in protective (peptide 27: R = – 0.37, p = 0.01; peptide 29: R = – 0.38, p = 0.01; Figures 5K and 5L) as well as non-protective vaccines (a trend for peptide 27: R = – 0.22, p = 0.09; peptide 29: R = – 0.34, p = 0.007; Figures 5M and 5N). Collectively, these data suggest that anti-V2 antibodies may inhibit infection by more than one mechanism.

V1 deletion in HIV A244 gp120 decreases V2 β sheet conformation

To address the potential for differences in HIV and SIV envelope structures, we tested the relevance of our finding to HIV by generating two V1-deleted HIV clade A/E A244 gp120 proteins. The first, A244ΔV1a, was designed by deleting the LTNVNNRTNVSNIIGNITD peptide and leaving the natural nine-amino-acid loop in V1 (Table S10 and Figure S9A) with the intent of minimizing tension in the adjacent V2 loop. In the second construct, A244ΔV1b, the nine-amino-acid loop was replaced with the corresponding nine amino acids of the SIVmac251ΔV1 antigen already proven to favor an α-helical V2 structure eliciting cytocidal antibodies correlating with a decreased risk of SIV acquisition (Table S11 and Figure S9A). A244WT, A244ΔV1a, and A244ΔV1b (Tables S9–S11) gp120 antigens expressed in 293 cells (Figures S9B–S9D) were probed in ELISA with mAb PG9, which recognizes V2 in a β sheet conformation, or CH58 and CH59, which recognize V2c in an α-helical conformation (Bonsignori et al., 2012; Gorny et al., 1994; Liao et al., 2013). The PG9 mAb bound better to gp120 A244WT than to both the gp120 A244ΔV1a and gp120 A244ΔV1b proteins (Figures 6A and 6B). The CH58 and CH59 mAbs had similar reactivity to A244 WT and the A244ΔV1a and A244ΔV1b proteins (Figures 6C–6F), suggesting that V1 deletion in the HIV gp120 A244ΔV1a and gp120 A244ΔV1b proteins shifts the structural equilibrium of V2 and reduces the V2 β sheet conformation (recognized by PG9) without affecting the V2 α-helical conformation (recognized by CH58 and CH59). These ΔV1 HIV immunogens therefore represent reagents matched to the V1-deleted SIV immunogens (that reduced the risk of virus acquisition) suitable to test whether focusing the antibody response to a V2 α-helical conformation improves the efficacy of HIV vaccine candidates.

Figure 6

HIV A244ΔV1 gp120 immunogens are preferentially recognized by human antibodies binding to V2 in α-helix conformation

(A, C and E) ELISA kinetic of HIV gp120 A244 WT, gp120 ΔV1a, and gp120 ΔV1b reactivity to (A) PG9 (anti-V2 antibody that recognizes V2 β Barrel conformation), (C) CH58, and (E) CH59 (anti-V2 antibodies recognizing V2 α-helical conformation).

(B, D and F) Average antibody binding in two experiments at the highest concentration of (B) PG9, (D) CH58, and (F) CH59 antibodies.

(G) Pictorial representation of the hypothesized abundance of the α-helical and β sheet V2 conformations in the WT gp120 (left), ΔV1 (center), and ΔV1gpg (right) apex trimers (not to scale). Data represented as mean with SD.

See Tables S9–11 and Figure S9.

Discussion

We have herein demonstrated that antibodies to the V1 region, exposed at the apex of the virion envelope trimer and adjacent to a conserved V2 region, have opposing effects on SIVmac251 acquisition in vaccinated macaques. Anti-V2 antibodies bound to infected cells and virions inhibited V2 binding to α4β7 and mediated ADCC, whose level and function correlated with a decreased risk of virus acquisition. In contrast, anti-V1 antibodies interfered with the ability of anti-V2 antibodies to bind gp120, mediate ADCC, and inhibit gp120 and α4β7 integrin interaction and furthermore correlated with an increased risk of virus acquisition. Whether antibodies to V1 affect V2 recognition by rendering specific V2 Ab-targeted epitopes less accessible by steric hindrance (competitive) or by stabilizing V1/V2 in a Greek-key β sheet fold through allosteric (non-competitive) inhibition remains to be determined. These mechanisms are not mutually exclusive in the context of polyclonal antibody responses. We demonstrate here that vaccination with the ΔV1 immunogens engineered to favor the V2 α-helical conformation (Figure 6G, center) elicited higher cytocidal V2 antibodies correlating with vaccine efficacy than immunogens with V2 in a β sheet conformation (Figure 6G, right) that afforded no vaccine efficacy. Overall, the results presented here are consistent with the finding of an alternative, unconstrained V2 α-helical conformation, distinct from that visualized in most envelope crystallographic structures, that was targeted by antibodies from the sera of volunteers in RV144 (Aiyegbo et al., 2017). This raises the hypothesis that V1 may not only decrease V2 accessibility to the immune system by direct masking but may also enforce a constrained V2 β strand conformation less accessible to the antibodies mediating ADCC via a non-competitive mechanism. Indeed, we demonstrate here that the SIV ΔV1 and ΔV1gpg antigens differed in their epitope accessibility. It is notable that the structure adopted by the isolated peptide V2c used in our studies differs significantly from the β strand form found in both the HIV-1 prefusion-closed trimer derived from stabilized gp120-gp41 linked by artificial disulfide bond (SOS) in combination with isoleucine-to-proline (IP) change in the gp41 (SOSIP) trimers (Medina-Ramírez et al., 2017) and the scaffolded V2 structures bound to HIV-1 bNAbs (Jiang et al., 2016). The structure inferred in our study is instead closer to that identified in co-crystals of HIV V2 peptides in complex with mAbs derived from an uninfected RV144 vaccinee (Liao et al., 2013), suggesting that the SIV V2 can adopt an α-helical structure analogous to the structure of V2 in HIV that has previously been linked to a reduced risk of HIV acquisition in humans (Lertjuthaporn et al., 2018).

The V2 envelope region is important in viral transmission and seeding gut inductive sites, as inferred by studies on transmitted HIV variants (Cavrois et al., 2014; Chohan et al., 2005; Jiang et al., 2016; Ritola et al., 2004; Rong et al., 2007; Sagar et al., 2006; Smith et al., 2016). Our findings here suggest that antibodies to V2 may interfere at different steps of viral transmission, from disrupting the interaction of the α-helical conformation of V2 with α4β7 during the establishment of infection to inhibiting virus spread by ADCC. Indeed, a recent study by Goes et al., 2020 demonstrated that V2 interaction with α4β7 provides a co-stimulatory signal that increases activation and proliferation of CD4+ cells and consequent HIV replication, suggesting that V2 may make gut resident T cells more receptive for viral infection. Furthermore, this phenotype was inhibited by antibodies recognizing the HIV α helix but not the β sheet conformation. Similar results were also obtained for SIV using the NCI09 mAb that was instrumental in our studies to reveal V2-specific ADCC responses correlating with a decreased risk of SIVmac251 acquisition (Figure 5D).

It is likely that the V1 of gp120 has evolved in SIV and HIV to counteract antibodies targeting the vulnerable V2 α-helical conformation. Antibody interference to HIV gp41 has been observed (Verrier et al., 2001), but interfering antibodies that target the apical gp120 domains and the V1/V2 have not been previously described. Interfering antibodies have been described in other viruses as well, including the Western equine encephalitis, polio, hepatitis C, and influenza viruses and the SARS-coronavirus (Dulbecco et al., 1956; Nicasio et al., 2012; Sautto et al., 2012; To et al., 2012; Tripp et al., 2005; Zhong et al., 2009).

The V1 deletion strategy employed here is relevant to the HIV vaccine design as removal of V1 from the A244 gp120 envelope decreases mAb PG9 binding to the protein, as we have demonstrated. Of particular significance is that this detrimental V1 element remains a component of most current Env-based vaccine candidates, suggesting that these candidates may exhibit improved efficacy with V1 deletion.

In summary, we have demonstrated that antibodies to V1 counteract functional antibody responses to viral vulnerability sites in V2. Minimizing the confounding role of V1, by its deletion or other means, presents a new opportunity to understand the biochemical basis of V2-associated viral vulnerability toward developing a fully efficacious vaccine for HIV.

Limitations of the study

The current study demonstrates that SIV envelope immunogens with the V1 region deleted (ΔV1) to favor the V2 α-helical conformation induced significant vaccine protection, whereas deletion of V1 to favor the V2 β sheet conformation, or V1 repleted (Wild Type) SIV immunogens, were not protective. Although the ΔV1 immunogens were engineered to assume these different conformations, we were unable to confirm their conformation experimentally. It was encouraging, however, that V1 deletion in the HIV A244 envelope, engineered to favor the V2 α-helical conformation, resulted in the loss of binding to monoclonal antibodies recognizing V2 in a β sheet conformation. In addition, the level and interference activity of V1-specific antibodies may vary depending on the HIV clade (Shen et al., 2015). Lastly, the predictive value of our preclinical study in Indian rhesus macaques for humans remains unclear. It is noteworthy, however, that the identical SIVmac251 macaque model recapitulated and predicted (Vaccari et al., 2016) the efficacy of the RV144 human trial (performed in more than 16,000 volunteers in Thailand) and the lack of efficacy of the HVTN702 trial (more than 5,000 volunteers in South Africa), respectively.

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by Dr. Genoveffa Franchini (franchig@mail.nih.gov).

Materials availability

All unique/stable reagents generated in this study are available from the Lead Contact with a completed Materials Transfer Agreement.

Data and code availability

The accession number for the structural analysis of the NCI09 antibody reported in this paper is PDB: 6VRY. All other data can be made available upon request.

Methods

All methods can be found in the accompanying Transparent methods supplemental file.