Breakthrough Virus Neutralization Resistance as a Correlate of Protection in a Nonhuman Primate Heterologous Simian Immunodeficiency Virus Vaccine Challenge Study

ABSTRACT

Comprehensive assessments of immune correlates of protection in human immunodeficiency virus (HIV) vaccine trials are essential to vaccine design. Neutralization sieve analysis compares the neutralization sensitivity of the breakthrough transmitted/founder (TF) viruses from vaccinated and control animals to infer the molecular mechanisms of vaccine protection. Here, we report a robust neutralization sieve effect in a nonhuman primate simian immunodeficiency virus (SIV) vaccine trial (DNA prime/recombinant adenovirus type 5 [rAd5] boost) (VRC-10-332) that demonstrated substantial protective efficacy and revealed a genetic signature of neutralization resistance in the C1 region of env. We found significant enrichment for neutralization resistance in the vaccine compared to control breakthrough TF viruses when tested with plasma from vaccinated study animals, plasma from chronically SIV-infected animals, and a panel of SIV-specific monoclonal antibodies targeting six discrete Env epitopes (P < 0.008 for all comparisons). Neutralization resistance was significantly associated with the previously identified genetic signature of resistance (P < 0.0001), and together, the results identify virus neutralization as a correlate of protection. These findings further demonstrate the in vivo relevance of our previous in vitro analyses of the SIVsmE660 challenge stock, which revealed a broad range of neutralization sensitivities of its component viruses. In sum, this report demonstrates proof-of-concept that phenotypic sieve analyses can elucidate mechanistic correlates of immune protection following vaccination and raises a cautionary note for SIV and SHIV (simian-human immunodeficiency virus) vaccine studies that employ challenge strains with envelope glycoproteins that fail to exhibit neutralization resistance profiles typical of TF viruses.

IMPORTANCE With more than 2 million new infections annually, the development of an effective vaccine against HIV-1 is a global health priority. Understanding immunologic correlates of protection generated in vaccine trials is critical to advance vaccine development. Here, we assessed the role of vaccine-elicited neutralizing antibodies in a recent nonhuman primate study of a vaccine that showed significant protection against simian immunodeficiency virus (SIV) challenge and suggested a genetic signature of neutralization sensitivity. We found that breakthrough viruses able to establish infection in vaccinated animals were substantially more resistant to antibody-mediated neutralization than were viruses from controls. These findings suggest that vaccine-elicited neutralizing antibodies selectively blocked the transmission of more sensitive challenge viruses. Sieve analysis also corroborated a genetic signature of neutralization sensitivity and highlighted the impact of challenge swarm diversity. Our findings suggest an important role for neutralization sieve analyses as an informative component of comprehensive immune-correlates analyses.

INTRODUCTION

Substantial efforts toward the development of a preventive human immunodeficiency virus type 1 (HIV-1) vaccine have yet to produce a promising vaccine candidate. In the five large human HIV-1 vaccine efficacy (VE) trials conducted to date, only the RV144 trial demonstrated modest VE (1). Despite these disappointing clinical outcomes, intensive posttrial analyses have yielded insights into immune correlates, elucidating potential mechanisms of protection and generating testable hypotheses for future research. A key component of these posttrial assessments is sieve analysis (2–6). By comparing features of the viruses establishing infection in vaccine recipients to those in placebo recipients, sieve analyses have identified properties of the specific viruses blocked (or “sieved”) by vaccine-induced immune pressure. Examples include CD8 T cell epitopes in conserved viral structural proteins (6) and antibody targets in the envelope (5). Genetic sieve analyses employ statistical approaches to find sequence evidence of vaccine pressure on the large and divergent population of HIV-1 viruses to which trial participants are exposed (7, 8). For the RV144 and STEP trials, such genetic sieve studies helped define the mechanism of action of each vaccine strategy (2, 5, 6). The value of genetic sieve analysis is complementary to that of phenotypic analyses. In the RV144 trial, sieve analyses focused on the V1V2 region of envelope were used to test the hypothesis generated in correlates-of-risk studies that antibodies against the V2 region of envelope mediated protection. The genetic sieve effect found in RV144 was then further corroborated by analyses of immune responses of subjects in the RV144 study, forming a cohesive picture of potential immune mechanisms of protection in RV144 (9–13). A more broadly applied genetic sieve analysis identified other genetic determinants associated with vaccine-mediated responses, thereby raising other testable hypotheses of the RV144 mechanism of action (3). Thus, genetic sieve analyses are an integral part of posttrial analyses, which are able to both generate hypotheses and confirm and extend the findings of complementary phenotypic correlates studies.

Neutralization sieve analyses, wherein the neutralization sensitivities of breakthrough viruses to vaccine-induced antibodies between study arms are compared, have been proposed as potentially important components of HIV vaccine correlates studies (14, 15). In vaccine trials designed to elicit humoral responses, neutralization analyses may be particularly relevant. Of the five large human efficacy trials, only VAX004 has been tested for a neutralization sieve effect. VAX004 did not demonstrate overall VE, but it showed a trend toward protection in high-risk subjects (16), and in vitro assays identified specific immune responses that correlated with lower infection rates (17, 18). A neutralization sieve analysis of VAX004 tested the ability of antibodies in plasma from vaccinated subjects to neutralize breakthrough viruses in both study arms. That study found a statistically significant sieve effect, with the vaccine breakthrough viruses demonstrating greater resistance to neutralization by immune plasma samples (14). Due to limitations inherent to large human trials, such as the relatively small number of seroconversions, delays in diagnosis, and limited biospecimen availability, such phenotypic sieve analyses have not been implemented widely and have not been reported in a human trial demonstrating overall efficacy.

The ideal setting in which neutralization sieve analyses could contribute to vaccine trial assessment is a trial with demonstrated VE, a proposed neutralizing antibody (NAb)-based mechanism of action, and early diagnoses of infection with ample available clinical specimens (15). While this scenario has been challenging to achieve in human trials, it can be more readily accomplished in the nonhuman primate (NHP) model. Furthermore, the NHP model simplifies sieve analyses, as animals are exposed to a controlled challenge rather than a highly divergent population of circulating HIV-1 strains (19). Designed to enable comprehensive correlates analysis the NIH Vaccine Research Center (VRC) recently conducted a large preclinical NHP vaccine protection study (4). The VRC trial tested 80 animals with a DNA prime/recombinant adenovirus type 5 (rAd5) boost platform with three candidate vaccine inserts, including a T cell-optimized Gag mosaic immunogen, an Env mosaic immunogen, an SIVmac239 wild-type Env immunogen, and placebo (4). Following immunization, the animals were challenged by 12 repeat intrarectal inoculations with the heterologous virus swarm, SIVsmE660. The SIVsmE660 challenge stock used had ∼1.8% within-swarm diversity and ∼14% genetic distance in the env sequence from the vaccine insert virus, SIVmac239 (20). This VRC trial demonstrated that SIVmac239 Env-based immunization conferred 69% per-challenge VE (4). Extensive correlates analysis revealed that VE was associated with several antibody responses, and a genetic sieve analysis identified a 2-amino-acid signature (at amino acid positions 45 and 47) in the C1 region of envelope that was enriched in breakthrough infections of vaccinated animals (4). This C1 genetic signature, when analyzed in the context of well-characterized simian immunodeficiency virus (SIV) Env backbones, was associated with neutralization resistance in in vitro tests (4). However, this C1 genetic signature was not evaluated in the context of the homologous Env gp160 molecules of the actual breakthrough transmitted/founder (TF) viruses.

Here, we performed a neutralization sieve analysis of the breakthrough TF viruses from this DNA prime/rAd5 boost vaccine trial. This follow-up study had several objectives. First, we aimed to test the concept of vaccine-mediated sieving of neutralization-sensitive viruses and demonstrate a proof-of-concept that these types of studies can be a valuable component of correlates analysis. Second, we aimed to test the predictive capacity of the genetic signature of neutralization resistance found in VRC-10-332 (4), as it has been challenged by posttrial assessments of other SIVsmE660 vaccine challenge studies (21, 58). Third, we sought to assess the impact of the unique pattern of diversity in neutralization sensitivity of the SIVsmE660 challenge stock (20) on the outcome and interpretation of data from a preclinical trial of a vaccine that generated a substantial antibody-mediated immune response (4).

MATERIALS AND METHODS

VRC-10-332 prime/boost vaccine with SIVsmE660 challenge study.

The study protocol and results from this 80-animal trial were reported previously (4). Briefly, 80 rhesus macaques were divided into 4 groups. Each 20-animal group received vaccination with a DNA prime/rAd5 boost platform with one of three vaccine inserts: (i) a T cell-optimized Gag mosaic immunogen, (ii) an Env mosaic immunogen, (iii) an SIVmac239 wild-type Env immunogen, or (iv) placebo. Following immunization, the animals were given 12 intrarectal challenges with the heterologous virus swarm SIVsmE660. The SIVsmE660 challenge stock used had ∼1.8% within-swarm diversity and ∼14% genetic distance in the env sequence from the vaccine insert virus strain SIVmac239 (20). Only breakthrough viruses from wild-type SIVmac239-vaccinated animals and controls were assessed in this study.

Sequence alignments and diversity estimates.

All sequences were aligned manually or by using CLUSTALW (22) and inspected with MacClade 4.08 (23) to optimize alignments. The breakthrough TF env sequences were generated by single-genome sequencing (SGS) of plasma samples from peak viremia from animals in NHP trial VRC-10-332, as previously reported (20, 24). Briefly, plasma SIV RNA from ramp-up viremia (7 days postchallenge) was extracted, cDNA was synthesized, and SGS of full-length gp160 env was performed, as previously described (24, 25). A mean of 21.2 sequences (range, 10 to 38) were obtained for each animal (4). TF viruses were identified as the consensus sequence of each low-diversity lineage that conformed to conditions of a model of neutral virus evolution (24). In our study, the actual TF virus sequences were then used for SIV Env cloning. Sequences from the primary trial were uploaded to GenBank (accession numbers KF602252 to KF603880) (4). The VRC-10-332 trial was conducted with approval from the Vaccine Research Center Animal Care and Use Committee. The maximum likelihood phylogenetic tree was generated by using PhyML version 3.0 (26) with a TIM1+I+G model of substitution, including previously characterized reference SIVsmE660 TF env genes (20, 24). Pairwise diversity estimates were determined by using DIVEIN (27). Assessment for clustering of vaccine versus control breakthrough TF env genes and the C1 genetic signature was performed with a Slatkin-Maddison test (28, 29), and Hudson's nearest-neighbor (Snn) analyses were performed in HyPhy (29), using 10,000 permutations on data sets that excluded the nucleotides encoding amino acids at position 45 and 47.

env gene cloning.

The breakthrough TF Envs were molecularly cloned for the production of pseudovirus and phenotypic analyses. We used the sequences generated in the primary challenge study that represented each of the TF viruses. Of the 60 TF viruses identified (41 in controls and 19 in vaccinated animals) in the primary trial, 7 TF sequences (from 6 controls and 1 vaccinated animal) were unable to be cloned because the exact TF virus amplicon or sequence was not available from the primary trial. To reduce the probability of generating molecular env clones with Taq polymerase errors, we reamplified the clones from the first-round PCR product under the same nested PCR conditions but used only 25 cycles. Correctly sized amplicons identified by gel electrophoresis were gel purified by using the QIAquick gel purification kit according to the recommendations of the manufacturer (Qiagen), ligated into the pcDNA3.1 Directional Topo vector (Invitrogen Life Technologies), and transformed into TOP10 competent bacteria. Bacteria were plated onto LB agar plates supplemented with 100 μg/ml ampicillin and cultured at room temperature for 3 days. Single colonies were selected and grown overnight in liquid LB broth at 30°C with shaking at 225 rpm, followed by plasmid isolation. Finally, each molecular clone was sequence confirmed to be identical to the previously determined env sequence of the TF env amplicon.

Pseudovirus preparation and titration.

Pseudovirus was prepared by transfecting 3 × 10⁶ 293T cells cultured overnight in 10-cm² tissue culture dishes with 4 μg of the rev-env expression plasmid and 4 μg of the HIV-2 backbone construct pJK7312AΔEnv by using Fugene 6 (Roche Applied Science, Indianapolis, IN). Pseudovirus-containing culture supernatants were harvested 2 days after transfection, cleared of cellular debris by low-speed centrifugation, and stored in aliquots at −80°C. Viruses were titrated on TZM-bl reporter cells (8129; NIH AIDS Research and Reference Reagent Program), which contain Tat-inducible luciferase and a β-galactosidase gene expression cassette. Infectious titers on 24-well plates were measured based on β-galactosidase production, representing the number of infection events (infectious units [IU]) per microliter of virus stock, as described previously (30, 31).

Antibodies and sera.

Monoclonal antibodies (MAbs) 3.11H, 6.10F, 1.10A, 1.7A, 6.10B, and 1.4H were provided by J. Robinson (Tulane University Medical Center, New Orleans, LA), and their neutralization properties and target epitopes were reported previously (32–35). MAbs ITS01, ITS06.02, ITS08, and ITS09.03 were isolated from SIV-infected macaques. Their epitope specificity and cross-reactivity were characterized by using an indirect enzyme-linked immunosorbent assay (ELISA) for binding to the SIV gp120 monomer, gp140 foldon trimer, and overlapping 15-mer SIV Env peptides as well as by using a competition ELISA with CD4-IgG (R. Mason and M. Roederer, submitted for publication). These MAbs were further characterized for virus neutralization potency and cross-reactivity by using the TZM-bl pseudovirus neutralization assay (36). Plasma samples from vaccinated animals were obtained at peak immunogenicity and characterized previously (4). SIV-infected macaque plasma samples were provided by N. Letvin (Harvard University, Boston, MA) and were obtained from chronically infected rhesus macaques that were intrarectally inoculated and productively infected with SIVsmE660 for at least 1 year.

Neutralization assays.

Virus neutralization by plasma and MAbs was assessed on TZM-bl cells as described previously (30, 31). TZM-bl cells were seeded at 1 × 10⁴ cells per well and cultured in 96-well plates overnight. Virus stock dilutions were made to final concentrations in Dulbecco's modified Eagle's medium (DMEM) containing 6% fetal bovine serum (FBS) and 40 μg/ml DEAE-dextran (Sigma-Aldrich, St. Louis, MO) to achieve 2,000 IU/well. Viruses with low numbers of IU per microliter were added at no less than 1,500 IU/well. Equal-volume virus dilutions and 5-fold serially diluted sera or MAbs were mixed and incubated at 37°C for 1 h. Supernatants were then removed, and 80 μl of these mixtures was added. Medium-only and virus-only control wells were included as background and 100% infectivity, respectively. Luciferase activity was measured after 48 h of incubation at 37°C by using Bright-Glo according to the manufacturer's instructions (Promega). All assays were done in triplicate in each of at least two independent experiments. The previously described neutralization-sensitive TF virus CP3C-P-A8 was used as a standard neutralization-sensitive virus (4, 20, 37, 38). CP3C-P-A8 is a TF virus that resulted from intrarectal inoculation with low-dose SIVsmE660, as described previously (37). To calculate the concentration of antibody that neutralized 50% of virus infection (50% inhibitory concentration [IC₅₀]), the antibody dose-response curves were fit with a four-parameter logistic equation by using Prism 5.0 (GraphPad Software, Inc., San Diego, CA). When 50% neutralization was not achieved at the highest concentration of plasma or antibody used, the IC₅₀ was recorded as being higher than the highest concentration used.

Statistical analyses.

Generalized estimating equations (GEEs) (39) were used to compare control to vaccinated animals, adjusting for the potential dependence caused by multiple viruses from the same animal. Models used either a linear or log link, as appropriate for the outcome under consideration, with robust, exchangeable correlation structures. Assays with results above the detection limit were analyzed by using random-effect tobit models (40) to account for censoring and within-animal dependence. Analyses considering both the study arm and C1 genetic signature (e.g., see Fig. 5B and C) were adjusted for multiple comparisons via Holm-Sidak adjustment. Analyses were performed by using StataMP 14.0 (Stata statistical software, release 14, 2015; StataCorp, College Station, TX).

FIG 5.

The C1 genetic signature predicts neutralization resistance. (A) The neutralization curve for a representative immune plasma from animal ZG12 segregated by C1 genetic signature (TR, sensitive; A/K, resistant) and study arm demonstrates clear separation between all TR viruses and all A/K viruses and modest separation between vaccine TR and control TR viruses. (B and C) IC₅₀s to SIVsmE660-infected macaque plasma samples (B) and vaccinated-animal plasma samples (C) of vaccine breakthrough and control TF Envs segregated by C1 genotype. (D to F) Maximal neutralization of TF Envs by plasma samples from vaccinated animals (D) and SIV-infected animals (E) and MAbs (F) correlates strongly with C1 genotype, regardless of study arm. In panels B and C, IC₅₀s of >1:500 are shown as red symbols.

RESULTS

Breakthrough TF Envs from vaccinated and control animals.

The 80-animal VRC vaccine challenge study tested a DNA prime/rAd5 boost strategy and found 69% per-challenge VE in SIVmac239 Env-immunized animals (4). To test for a neutralization sieve effect, we first generated a set of breakthrough TF Env clones from the SIVmac239 Env-vaccinated and control arms. The TF env sequences were derived by single-genome sequencing (SGS) of peak plasma viral RNA (vRNA) obtained at 7 days postchallenge from breakthrough infections of the SIVmac239 wild-type Env-immunized animals (n = 18 TF env sequences from 15 infected animals) and controls (n = 35 TF env sequences from 20 infected animals). The SIVmac239-vaccinated and control env sequences had mean and maximum diversities of 0.72% and 1.53%, respectively. As shown in Fig. 1, these TF env sequences are genetically similar to previously characterized SIVsmE660 TF viruses arising from low-dose rectal challenge from different but related SIVsmE660 stocks, which were dispersed throughout the phylogeny (20, 37). The breakthrough TF env sequences clustered within the phylogeny by both study arm (vaccine versus control) and C1 genetic signature (A/K versus TR) (P = 0.0001 and P < 0.0001, respectively, by a Slatkin-Maddison test and P < 0.0001 for both comparisons by Hudson's nearest-neighbor test). The breakthrough TF Envs were cloned, pseudotyped with an Env-minus backbone, and tested for entry into TZM-bl target cells. As expected for TF viruses, all 53 Envs conferred efficient virus entry.

FIG 1.

Phylogeny of SIVsmE660 breakthrough TF Envs. Shown is a maximum likelihood phylogenetic tree of gp160 env sequences from the 53 breakthrough TF viruses in SIVmac239 wild-type Env-vaccinated and control animals from the VRC DNA prime/Ad5 boost study with 11 reference SIVsmE660 TF env sequences (20). Breakthrough TF env sequences from the control arm are shown in red with the suffix .con, those from vaccine arm are in shown in green with the suffix .vac, and previously reported SIVsmE660 TF env sequences are in shown in black. The C1 genetic signature of neutralization resistance, A/K, is indicated with yellow stars and the suffix .AK, and the TR signature is indicated with the suffix .TR. There was statistically significant phylogenetic clustering of both the vaccine breakthrough TF env sequences and the A/K genetic signature.

Neutralization sensitivity of TF breakthrough Envs to SIVmac239-vaccinated macaque plasma samples.

To determine if vaccine-elicited NAbs selectively blocked transmission of the more neutralization-sensitive TF viruses, we tested the neutralization sensitivity of the 53 breakthrough Envs to prechallenge plasma samples from four vaccinated animals. The immune plasma samples came from two animals that were protected throughout the course of the 12-challenge study and from two animals that became infected after the second and seventh challenges, respectively (4). As a positive control, these vaccine plasma samples were first tested for neutralizing activity against a well-characterized neutralization-sensitive TF virus, CP3C-P-A8 (4, 20, 38). Despite heterogeneity in SIV acquisition outcomes, the plasma samples from these four animals exhibited similar neutralization potencies, with IC₅₀s ranging between 2.2 × 10⁻⁴ and 7.5 × 10⁻⁴ plasma dilution. The breakthrough TF Envs from vaccinated and control animals demonstrated a large range in neutralization sensitivity to the immune plasma samples, with IC₅₀s ranging from 3.05 × 10⁻⁶ to >0.002. Examples are shown for the immune plasma sample from animal 08D038 in Fig. 2A and B. The TF Envs exhibited one of two neutralization patterns: highly sensitive, with typical sigmoidal neutralization curves and maximal neutralization of ∼60 to >95%, or highly resistant, with neutralization curves showing a flattened sigmoid pattern that generally did not even reach an IC₅₀. The more resistant phenotype was highly enriched in the vaccinated than in control animals: a mean of 13.3 of the 18 (74%) vaccine breakthrough TF Envs were highly neutralization resistant (i.e., they did not reach 50% neutralization at the highest plasma concentration), compared with a mean of 6.3 of the 35 (17%) control Envs (P < 0.0001 for all comparisons). Similarly, the breakthrough TF Envs from vaccinated animals were significantly more neutralization resistant to the immune plasma samples, as measured by both IC₅₀ (Fig. 2C) and maximal neutralization (Fig. 2D) (P < 0.0005 for all comparisons). Notably, the two breakthrough TF Envs from animals 08D038 (Env clone PRBO13) (Fig. 2B) and 8-82 (Env clone P3D7) (data not shown) were highly resistant to their autologous plasma samples.

FIG 2.

Neutralization sensitivity of vaccine breakthrough and control TF Envs to prechallenge immune plasma from vaccinated animals. The 53 breakthrough TF Envs (n = 35 from control animals, and n = 18 from Env-vaccinated animals) were tested for neutralization sensitivity to immune plasma samples from 4 Env-vaccinated study animals (animals ZG05, ZG12, 8-82, and 08D038). (A and B) Infectivity curves for the control (A) and Env-vaccinated (B) animals are shown for a representative immune plasma sample from animal 08D038, where solid lines indicate the C1 signature “TR” and dashed lines indicate the C1 signature “A/K.” The autologous TF Env from animal 08D038, Env PRBO13, is shown as a bold black dashed line (B). (C and D) The TF Envs from Env-vaccinated animals were significantly more resistant than the TF Envs from control animals, measured by IC₅₀ (C) and maximal neutralization (D). In panel C, IC₅₀s of >1:500 are shown as red circles.

Neutralization sensitivity of TF breakthrough Envs to SIV-infected macaque plasma samples.

To determine whether the neutralization sensitivity differences between study arms were specific to this vaccine's antibody response or were more generalizable, we assessed the sensitivity of the vaccine breakthrough and control TF Envs to SIVsmE660-infected macaque plasma samples. Compared to the prechallenge immune plasma samples described above, the two representative SIV-infected macaque plasma samples had modestly increased neutralization potencies against a neutralization-sensitive TF Env (Env clone P5B23) and were similarly nonreactive against a more resistant TF Env (Env clone PRBO13), as shown in Fig. 3A and B. When the panel of 53 breakthrough TF Envs was tested against the SIV-infected plasma samples, the Envs from vaccinated animals were significantly more resistant, as measured by the IC₅₀ (Fig. 3C) and maximal neutralization (Fig. 3D) (P < 0.0001 for all comparisons). Notably, plasma from one of the two SIV-infected animals, A5V045, was capable of substantially more potent neutralization of the majority of the control TF Envs, as reflected by a median IC₅₀ that was 2 to 3 logs lower than that for the plasma samples from vaccinated animals (Fig. 2C and 3C). Due to the numbers of neutralization-resistant TF Envs, however, the mean IC₅₀s for all six plasma samples were within 0.3 logs of each other.

FIG 3.

Neutralization sensitivity to plasma samples from SIV-infected macaques. (A and B) The neutralization potencies of plasma samples from two chronically SIVsmE660-infected macaques (animals A5V026 and A5V045) and four Env-vaccinated macaques (animals ZG05, ZG12, 08D038, and 8-82) were tested against the neutralization-sensitive TF Env clone P5B23 (A) and the resistant TF Env clone PRBO13 (B). (C and D) Vaccine breakthrough TF Envs were significantly more resistant to plasma samples from SIV-infected macaque than were control TF Envs as assessed by IC₅₀ (C) and maximal neutralization (D). In panel C, IC₅₀s of >1:500 are shown as red circles.

Epitope specificity of neutralization sensitivity of TF breakthrough Envs.

To characterize the specificity of the neutralization sieve effect seen in this trial, we tested the vaccine breakthrough and control TF Envs for sensitivity to a comprehensive panel of anti-SIV monoclonal antibodies (MAbs) targeting V1, V2, V3, V4, CD4 binding site, and CD4-induced epitopes. The SIVsmE660 isolate was shown previously to be highly sensitive to neutralization at multiple epitopes, including V3, with IC₅₀s in the nanogram/milliliter range (20). Figures 4A and B show infectivity curves for one of the two V3-targeting MAbs tested, 6.10F, which displays neutralization potency against the majority of control TF Envs comparable to that against the previously described SIVsmE660 isolate stock (20). When the breakthrough vaccine and control TF Envs were compared, the vaccine TF Envs again show markedly greater resistance to 6.10F. Indeed, when tested with 10 MAbs targeting six distinct gp120 epitopes, the breakthrough TF Envs from vaccinated animals were globally more neutralization resistant than control Envs, as measured by IC₅₀ (P < 0.005 for all comparisons) (data not shown) or maximal neutralization (P < 0.008 for all comparisons) (Fig. 4C and D). The most potent neutralization, by both measures, was seen with the antibodies targeting V3 and the CD4 binding site but was statistically higher in the control Envs for all epitopes, reflecting a generally open conformation, higher intrinsic envelope reactivity (41), or global sensitivity of the neutralization-sensitive SIVsmE660 breakthrough TF viruses.

FIG 4.

Neutralization sensitivity to a panel of 10 MAbs. The vaccine breakthrough TF Envs were more resistant than control TF Envs to anti-SIV Env MAbs targeting epitopes throughout gp120. (A and B) Neutralization curves of representative anti-V3 MAb 6.10F for control (A) and vaccine breakthrough (B) TF Envs, where solid lines indicate the C1 signature TR and dashed lines indicate the C1 signature A/K. (C and D) Maximal neutralization by all 10 MAbs at high concentrations was significantly greater in control than in vaccine breakthrough TF Envs.

The C1 genetic signature predicts neutralization sensitivity.

Finally, we tested whether the 2-amino-acid signature identified by Roederer and colleagues conferred neutralization resistance to the TF breakthrough Envs. Analyses of data from the primary trial found that the resistant C1 genotype (either alanine at position 45 or lysine at position 47 [A/K]) was enriched in the breakthrough TF Envs of the SIVmac239 vaccine recipients compared to the SIVsmE660 challenge stock and non-Env-immunized TF viruses (73.6% versus 21.4% and 22.4%, respectively) (4).

Roederer and colleagues then showed that changing these two amino acids (from TR to AK) could convert the neutralization-sensitive virus strain CP3C to a resistant phenotype; the converse change (from AK to TR) changed the highly neutralization-resistant virus strain CR54 to a sensitive phenotype (4). To test whether this C1 signature conferred neutralization resistance across the spectrum of breakthrough TF Envs in which it naturally occurred, we compared the phenotypes of the breakthrough TF Envs by C1 signature. In support of the findings of the primary trial, we found that the A/K signature was strongly associated with neutralization resistance, as shown in the representative neutralization curve for the immune plasma sample from animal ZG12 (Fig. 5A). When segregated by both study arm and genetic signature, the A/K TF Envs within each study arm were significantly more neutralization resistant than the TR Envs, as measured by both maximum neutralization (data not shown) and IC₅₀ (Fig. 5B and C). Comparison of just TR Envs revealed a consistent, but largely statistically nonsignificant, trend toward greater neutralization resistance in breakthrough vaccine TR Envs than in control TR Envs across plasma samples and MAbs (Fig. 5B and C). Comparison of all breakthrough Envs regardless of study arm also demonstrated significantly greater neutralization resistance in A/K Envs by IC₅₀ (data not shown) and maximal neutralization (Fig. 5D to F). Indeed, after adjusting for the C1 signature (A/K versus TR), a regression analysis showed that the difference in neutralization sensitivity between vaccine and placebo TF Envs remained statistically significant (P < 0.05) for 4 of the 10 plasma samples or antibodies (animals ZG05, AV5045, ITS08, and ITS06.02) (data not shown). Thus, the C1 genotype accounted for much of the variability in neutralization sensitivity.

DISCUSSION

Comprehensive analyses of immune correlates in HIV vaccine trials are essential to iterative vaccine design. Assessments of NAb responses are particularly valuable, as NAbs are likely to be a central component of a highly effective HIV vaccine (42). Here, we report a robust neutralization sieve effect in an NHP vaccine trial that demonstrated substantial protective efficacy. Together with correlates-of-risk and genetic sieve analyses, the selection for neutralization-resistant breakthrough TF viruses in vaccinated animals corroborates NAb as a mechanistic correlate of protection (4, 43). The sieve analyses also corroborate the C1 genetic signature of neutralization resistance identified in the primary trial (4), which was highly predictive of phenotypic resistance. Furthermore, our results demonstrate the in vivo relevance of the previously described diversity in neutralization sensitivity of the SIVsmE660 challenge stock (20) and its consequent strengths and limitations as a vaccine challenge.

Sieve analyses of vaccine trials compare breakthrough infections in the vaccine and placebo groups to determine the characteristics of the pathogens that were blocked, or “sieved,” by vaccine-induced immune responses. In well-conducted, placebo-controlled trials with similar challenges (using either effective randomization in human clinical trials or identical challenge viruses in NHP trials), differences between study arms can be attributed to vaccination (2, 7). Thus, the conceptual framework behind sieve analysis has long been applicable to the clinical assessment of vaccines, especially vaccines against antigenically diverse pathogens. The identification of differences in relevant characteristics (e.g., serotypes or phenotypic qualities) of breakthrough infections allows estimation of differential efficacies and testing of specific hypotheses (7). Genetic sieve analyses of data from HIV vaccine trials have been highly instructive to vaccine development by identifying vaccine-mediated immune pressures and their specific viral targets (2, 3, 5, 6). The neutralization sieve analysis performed here was based on the genetic sieve framework but was specifically designed to test the NAb mechanism of action proposed by the correlates-of-risk and genetic sieve analyses of data from the primary trial (4). In the context of mutually reinforcing orthogonal correlates studies, the selection for resistant viruses in vaccinated animals seen consistently throughout our analyses confirms and extends our understanding of the NAb mechanism of protection.

The significant enrichment of vaccine breakthrough TF Envs that were highly resistant to neutralization by immune plasma samples strongly suggests that vaccine-induced NAbs were capable of blocking infection by neutralization-sensitive SIVsmE660 challenge viruses and mediated protection in this trial. The selection for resistant viruses was consistent across the four tested plasma samples from vaccinated animals, which showed comparable neutralization profiles and similarly increased proportions of neutralization-resistant vaccine breakthrough TF Envs (Fig. 2C and D). Thus, despite variability in clinical outcomes, the induced humoral response was uniform across tested animals. The sieve effect was also consistent and statistically robust as assessed by several neutralization metrics, including IC₅₀, maximal neutralization, and proportion of resistant Envs (P < 0.008, for all comparisons). Employed to translate with well-validated studies of HIV-1 neutralization (36) and to account for the atypical neutralization pattern seen with SIVsmE660 (i.e., incomplete neutralization across antibodies) (4, 20, 21), these different metrics were mutually reinforcing.

Robust selection for resistant viruses was also seen with infected plasma samples and a broad panel of SIV-specific MAbs; these analyses enabled further insights into the induced NAb response in this vaccine trial. Differences in neutralization titers between immune and infected plasma samples allowed a measure of the relative potency of the induced humoral response. Indeed, while TF Envs that were resistant to immune plasma samples were largely resistant to infected plasma samples, the sensitive Envs were neutralized at significantly lower concentrations of infected than immune plasma samples (median IC₅₀s of 3.98 ×10⁻⁶ and 4.46 ×10⁻⁵ plasma dilution, respectively [P < 0.0001]). This difference in median IC₅₀s between immunized and infected plasma samples (Fig. 3A) suggests a moderately weaker humoral immune response in vaccinated than in infected animals. Interestingly, the sieve effect also remained robust for each of the 10 MAbs targeting six discrete gp120 epitopes (Fig. 4). This finding reinforces the global sensitivity to neutralization of SIVsmE660 (20, 38) and suggests that vaccine-induced antibodies targeting any of the epitopes tested here, including several that are largely occluded on primary HIV-1 viruses (44), had the potential to confer protection against SIVsmE660 challenge. Indeed, inference of the characteristics of the viruses sieved by the vaccine suggests that the blocked viruses possessed an extreme and global neutralization sensitivity, akin to a tier 1a classification (45) or high intrinsic Env reactivity (41), which is quite different from the majority of TF HIV-1 strains against which an HIV-1 vaccine must ultimately protect (24). In total, the selection for resistant viruses shown consistently across immunized and infected polyclonal plasma samples and MAbs targeting multiple gp120 epitopes provides strong evidence of a vaccine-mediated NAb response that protected against the acquisition of the highly neutralization-sensitive viruses comprising much of the SIVsmE660 swarm.

In the primary trial, Roederer and colleagues identified a 2-amino-acid signature in C1 that was highly overrepresented in vaccine breakthrough env sequences and conferred neutralization resistance when tested in standard SIVsmE660 Env backbones (4). This finding has been questioned in recent analyses of different SIVsmE660 vaccine challenge experiments (46–48) that showed neither enrichment for the C1 genetic signature nor a neutralization sieve effect (21, 58). While the characterization of breakthrough Envs by Burton et al. (58) was methodologically similar to our study, the SIV vaccine trials from which the breakthrough TF Envs were identified were substantially different (46–48). Notably, the parent trials used SIVsmE660 stocks with less genotypic and phenotypic diversity (i.e., Burton et al. found both a significantly lower frequency of A/K viruses in their SIVsmE660 challenge stock than in the challenge stock used in the VRC trial and fewer neutralization-resistant breakthrough Envs than in the present study). Furthermore, the three SIV vaccine challenge trials studied by Burton et al. did not induce protective NAb responses (i.e., VE correlated with higher-avidity binding antibodies [49] and not NAbs) (46–48). Thus, the significant differences in challenge stocks and mechanisms of induced immune responses of the primary trials may explain the seemingly contradictory findings and emphasize the importance of challenge stock and correlates analysis in interpretation of data from vaccine trials. Indeed, our analyses found that the C1 signature was highly predictive of phenotypic neutralization resistance in breakthrough TF Envs for both vaccine-specific reagents (immune plasma samples from vaccinated animals) and more general reagents (plasma samples from SIV-infected animals and a broad panel of MAbs). The strength and generalizability of the association suggest its utility as a screening tool. For example, in a future trial, the C1 regions of SIVsmE660 challenge stocks could be sequenced in a high-throughput manner to determine the relative proportions of neutralization-sensitive and -resistant viruses before they are employed experimentally. While strongly associated with phenotypic resistance, the C1 signature did not account for all of the differences in neutralization sensitivity within the challenge stock or the complete effect of the phenotypic sieve. As shown in Fig. 5, there was overlap between the neutralization profiles of the A/K and TR viruses across assays. Furthermore, in 4 of the 10 plasma samples tested, selection of resistant viruses persisted after controlling for the C1 genetic signature. Of note, Roederer and colleagues reported a genetic sieve effect in TR viruses within the mosaic breakthrough virus sequences at position 162 in the V1 region of env in Env-vaccinated plasma, but this signature was not predictive in either the sequence analysis of SIVmac239 wild-type Env-vaccinated animals in the primary study (4) or our phenotypic assays (data not shown). Thus, our analyses strongly support the predictive capacity of the C1 genetic signature, show that it accounts for the majority of the diversity in neutralization sensitivity in the SIVsmE660 swarm, and suggest its utility as a screening tool to measure SIVsmE660 neutralization diversity.

NHP models are designed to answer questions that cannot be readily or feasibly addressed in human clinical trials. As with all model systems, SIV infection of macaques has limitations in accurately representing HIV-1 infection of humans (50). Adaptations to the model, including low-dose mucosal challenge and genetic matching of the animals' major histocompatibility complex class 1 alleles, have improved its biologic relevance (37, 51). Optimization of SIV challenge viruses to recapitulate circulating HIV-1 TF viruses, however, has remained a persistent challenge (24, 38). As demonstrated here, a strength of the SIVsmE660 isolate as a vaccine challenge virus is its heterogeneity. The diversity present in the SIVsmE660 swarm employed in this study (genetic diversities of 1.8% in the challenge stock [4] and >1.5% in T/F env sequences, with a range of >4 logs in IC₅₀s of TF Envs in plasma samples from infected macaques) allowed for the detection and delineation of vaccine-generated protection. Many correlates analyses with the NHP model, including genetic and phenotypic sieves, are dependent on sufficient diversity within the challenge stock to differentiate between breakthrough viruses. Thus, heterogeneous challenge stocks are essential to both recapitulate the diverse HIV-1 challenge faced by humans in communities throughout the world (19) and facilitate comprehensive correlates analysis. In trials of vaccines designed to induce antibody responses, the neutralization sensitivity of challenge stocks is of paramount importance. To address how accurately SIVsmE660 represents primary HIV-1 strains, we recently characterized the neutralization sensitivity of this commonly used challenge stock (4, 38, 46–48, 52). We found that the SIVsmE660 isolate was comprised of a majority of highly neutralization-sensitive viruses with a smaller fraction of more resistant viruses (20). Our findings prompted us to hypothesize that incomplete protection against low-dose mucosal SIVsmE660 challenge may result from protection against only the highly neutralization-sensitive variants. Thus, while the SIVsmE660 swarm possesses viruses with a range of neutralization sensitivities, substantial VE could be achieved with very modest antibody responses. Here, we used the neutralization sieve methodology to directly assess vaccine breakthrough viruses and indirectly infer the characteristics of the viruses that were prevented from establishing infection. The highly neutralization-sensitive viruses (median neutralization titers of >1:100,000 for infected plasma samples) blocked by vaccine-induced immune responses demonstrate the sensitivity of the SIVsmE660 challenge strain for the detection of modest vaccine-induced humoral responses. In this way, challenge with neutralization-sensitive SIVsmE660 can be used as a less stringent, heterogeneous vaccine challenge that is heterologous to SIVmac239-based immunogens. Importantly, the highly neutralization-sensitive viruses comprising the majority of SIVsmE660 strains differ substantially from the primary HIV-1 strains that are the target of an HIV vaccine (24, 53). The more resistant viruses comprising a minor component of the SIVsmE660 isolate differ from primary HIV-1 strains as well, demonstrating incomplete neutralization with a nonsigmoidal neutralization dose-response curve and “plateaus” of retained infectivity at high antibody concentrations. Such plateaus have been associated with antibodies targeting glycan-dependent and membrane-proximal external-region epitopes in HIV-1 (54–56) but are not associated with virus-specific neutralization patterns across epitopes in primary HIV-1 isolates. While this type of incomplete in vitro neutralization has unclear in vivo ramifications, it differentiates SIVsmE660 and the tier 2 and 3 TF HIV-1 strains against which a highly effective vaccine must protect and merits further study. As the vaccine field improves the breadth and potency of induced antibody responses, a more robust challenge swarm will be needed to perform a more rigorous check of vaccine protection. Thus, our findings highlight the in vivo importance of the high proportion of neutralization-sensitive viruses comprising the SIVsmE660 challenge stock (20) and the need for more authentic viral lineages to accurately recapitulate HIV-1 and confer a robust vaccine challenge. As the field adopts other SIV- and simian-human immunodeficiency virus (SHIV)-based challenge models, development of swarms with both heterogeneity and biologically relevant neutralization phenotypes should be pursued.

In conclusion, the development of an effective vaccine against HIV-1 acquisition is of paramount importance to global public health. Comprehensive, orthogonal, and mutually reinforcing correlates analyses are essential to fully understand the results of each of the time- and resource-intensive human and NHP vaccine trials completed to date (57). Our findings demonstrate proof-of-concept that phenotypic sieve analyses can confirm and extend proposed mechanisms of action and correlates of protection while simultaneously evaluating key components of model systems. This study was specifically designed to test a NAb correlate; however, a similar format of phenotypic testing of breakthrough TF viruses could be employed to test other immune mechanisms. For example, in human or NHP trials showing partial protection without a NAb correlate (1, 46–48), sensitivity to other antibody effector functions of breakthrough TF Envs could be assayed in a high-throughput manner to either test a proposed mechanism of protection or generate hypotheses in studies without a clear correlate. For these reasons, phenotypic sieve analyses of breakthrough viruses should be considered and incorporated in future trials.