Could vaccination with AIDSVAX immunogens have resulted in antibody-dependent enhancement of HIV infection in human subjects?

Abstract

The immune-correlate analysis of the RV144 clinical trial revealed that human plasma IgA immune responses elicited by the RV144 vaccine correlated positively with a risk for HIV acquisition. This result once again emphasized that HIV vaccines can potentially have adverse effects leading to enhancement of infection. Here, we discuss previously reported evidence of antibody-dependent enhancement of HIV infection. We also describe how a structure-based epitope-specific sieve-analysis can be employed to mine the molecular mechanism underlying this phenomenon.

Introduction

Significant evidence suggests that the elicitation of a protective antibody (Ab) immune response by a vaccine could prevent acquisition of human immunodeficiency virus (HIV).^1,2 Indeed, the recent RV144 HIV vaccine clinical trial could be viewed as a proof-of-principle that a vaccine can prevent HIV infection in humans.³ However, in addition to a protective effect, the RV144 vaccine appears to have elicited plasma immunoglobulin A (IgA) responses in vaccinated human subjects that correlated positively with an increased risk of infection.⁴ Precisely understanding the mechanisms behind this potentially adverse immune response to RV144 vaccination may be extremely important to the design of a safe and effective anti-HIV vaccine.

A number of previous studies postulated that vaccination could potentially be not only ineffective, but actually harmful, by rendering vaccine recipients more susceptible to infection rather than protecting them. Such a phenomenon, called a vaccine-induced enhancement, is known for infections by various viral pathogens⁵ including members of lentivirus family (e.g., feline immunodeficiency virus^6-8). The first-ever clear evidence of vaccine-induced enhancement of HIV in clinical studies was also recently reported when an increased HIV acquisition risk was detected in selected subgroups of vaccinated subjects enrolled in the STEP study.^9,10

Antibody-Dependent Enhancement of Viral Infection

Antibody-dependent enhancement (ADE) is the molecular mechanism of enhancement of viral infection which has been previously documented for various viral pathogens. The most studied example of ADE is dengue virus, where documented increases in pathogenicity in humans were associated with prior heterotypic Abs.^11,12 Additionally, ADE has also been reported for infection with Murray Valley encephalitis, respiratory syncytial, ebola, and measles.^13-16 Even though the first evidence of ADE in HIV was discovered in vitro in late 1980s,¹⁷ the idea of ADE has been largely ignored by the HIV research community for many years until the recently completed immune-correlate analysis of RV144 clinical trial showed that the binding of non-neutralizing plasma IgA Abs to env correlated directly with the rate of infection in vaccine recipients.⁴ ADE, however, was likely not responsible for the infection enhancement observed in the STEP study mentioned in the Introduction section above as the STEP vaccine was not designed to elicit Abs.

ADE in HIV has historically been suspected to be associated with non-neutralizing functions of Abs: complement activation^18,19 and Fc receptor biding.²⁰ Such mechanisms increase local concentrations of fully functional virus in the proximity of target cell surface, substantially increasing the probability of binding and fusion of a virus to a target cell. Alternatively, both neutralizing and non-neutralizing anti-HIV Abs may also enhance the infection by inducing conformational changes and “locking” HIV's envelope gp41-gp120 spikes in conformations facilitating host receptor/co-receptor recognition by the virus.²¹

ADE of HIV Infection in Clinical Studies

Despite the fact that the ADE of HIV infection in vitro was discovered decades ago, no indication of ADE in clinical HIV trials had been observed until 2005, when statistical analysis of AIDSVAX clinical trials^22,23 revealed that VAX004 vaccinees with low rgp120 Ab responses had a rate of HIV infection higher than that of the ‘placebo’ cohort.²⁴ That observation was recently revived by the results of the RV144 immune-correlate analysis mentioned above.²³

Enhancement of HIV Infection in AIDSVAX Vaccine Trials

Sieve-analysis is a powerful bioinformatics approach which allows mining and understanding immune responses elicited by vaccination.²⁵ Previously, we reported an epitope-specific sieve-analysis of the VAX003 and VAX004 clinical trials of AIDSVAX vaccine.²⁶ Specifically, we profiled the distributions of Ab-targeted epitopes of several anti-V3 neutralizing monoclonal Abs (mAbs) among the sequences of the breakthrough HIV viruses isolated from the volunteers infected during the AIDSVAX trials. We compared the epitope-specific infection counts between vaccinated and placebo cohorts statistically to see if any epitope-specific narrow protective immune responses could have been elicited by AIDSVAX vaccination. In this review, the same epitope distribution data is, however, used to test the reverse hypothesis: to see if a significant increase in the rates of epitope occurrence could be observed in vaccinees comparing to placebo recipients. In other words, we test the null hypothesis of no enhancement of the HIV infection in the ‘vaccine’ group by using the left-tailed Fisher Exact Test instead of the right-tailed test applied in our previous published analysis.

Reversed hypothesis testing reveals that the infection counts for the 3 epitopes, those targeted by mAbs 268-D, 447–52D, and 537–10D, are significantly (p < 0.05) higher in the ‘vaccine’ VAX004 cohort comparing to the ‘placebo’ VAX004 cohort (Table 1). Notably, 2 of the epitopes, the ones targeted by mAbs 268-D and 447–52D, are also statistically significant after Bonferroni correction for multiple hypothesis testing. None of the comparisons for the presence of other epitopes in breakthrough viruses result in detection of a statistically significant difference, including the comparison for the epitope targeted by the mAb 3791, which was not present in the AIDSVAX immunogens (internal negative control, as described by Shmelkov, et al²⁶). These data suggest that the VAX004 vaccine resulted in the elicitation of Abs that increased the risk of infection of the vaccinees with viruses decorated with the epitopes targeted by mAbs 268-D, 447–52D, and 537–10D.

Table 1.

Mapping anti-V3 mAb epitopes in gp120 sequences of HIV breakthrough viruses infecting the AIDSVAX study population

AIDSVAX Trial ID:	VAX003				VAX004
Viral sequence	Placebo	Vaccine	Left-tailed p-value	Bonferroni correction	Placebo	Vaccine	Left-tailed p-value	Bonferroni correction
2219 containing	23	12	0.9867	1.0000	85	156	0.5126	1.0000
2557 containing	21	13	0.9525	1.0000	75	153	0.1003	0.7021
268-D containing	1	2	0.4962	1.0000	33	105	0.0002	0.0014
3074 containing	84	82	0.6633	1.0000	103	181	0.8207	1.0000
3791 containing	9	9	0.5884	1.0000	1	1	0.8752	1.0000
447–52D containing	16	18	0.4126	1.0000	74	168	0.0024	0.0168
537–10D containing	9	12	0.3137	1.0000	64	142	0.0241	0.1687
Unknown sequence	6	8	n/a	n/a	8	24	n/a	n/a
Total infected	105	106	n/a	n/a	127	241	n/a	n/a

Note: Numbers in ‘Placebo’ and ‘Vaccine’ columns are the numbers of human subjects infected with viruses bearing specified Ab-targeted epitopes. ‘Left-tailed p-value’ columns show p-values computed with the left-tailed Fisher Exact test. Bonferroni correction bounds were computed using a factor of 7 for each of the 2 trials separately.

One could hypothesize that if the increase in epitope-specific infection counts in the ‘vaccine’ cohort of the VAX004 trial as compared to the ‘placebo’ truly occurred as a result of the AIDSVAX vaccination, there should be no such increase in counts for the same Ab-targeted epitopes in the STEP cell-mediated vaccine study. Therefore, the STEP study can serve as an external negative control. We performed the analyses described above on the breakthrough sequences from the STEP clinical trial²⁷ and, as expected, no significant difference in distribution of Ab-targeted epitopes was observed between ‘vaccine’ and ‘placebo’ cohorts (data not shown).

Elicitation of mAb 268-D By Vaccine Components May Enhance HIV Infection

The most obvious explanation of the statistically significant increase in infection counts for the epitopes targeted by mAbs 268-D, 447–52D, and 537–10D in the vaccinated VAX004 cohort is that elicitation of Abs with the same specificity as one or several of these mAbs by the AIDSVAX B/B immunogen resulted in ADE of HIV infection. Indeed, Kliks, et al. previously demonstrated that the outcome of HIV interaction with some mAbs depends on the sequence of the V3 loop of viral env and specifically, mAb 268-D can induce both neutralization as well as enhancement of infection by various HIV strains.²⁸ This study suggests that the highly significant (p = 0.0002) increase in infection rates of the epitope targeted by mAb 268-D (Table 1) seen in VAX004 trial may relate to activation of ADE mechanisms in trial subjects as a result of AIDSVAX B/B vaccination.

Conclusions

Statistically significant differences between the occurrences of overlapping epitopes targeted by mAbs 447–52D, 537–10D, and 268-D in VAX004 ‘placebo’ and ‘vaccine’ cohorts could turn out to be an extremely important observation. The most obvious explanation of this difference is that a vaccine-induced enhancement (namely ADE) of infection has occurred as a result of AIDSVAX vaccination. A problem with this scenario is that in phase 1 and 2 of the AIDSVAX vaccine trials, there was no evidence of ADE.²⁹ On the other hand, there is experimental evidence that at least one of the 3 potentially enhancing in the VAX004 mAbs can induce ADE in vitro.²⁸

It is important to note that the signature motifs of the epitopes targeted by mAbs 268-D (10-[R,K]xx[H,R]xxPxR-18), 447–52D (16-PxR-18), and 537–10D (9-Rxxxx[I,M]xPxR-18) have the same sequence pattern 16-PxR-18.^30,31 Therefore, a set of sequences bearing the motif 268-D as well as a set of sequences bearing the motif 537–10D are subsets of a bigger set of sequences with the 447–52D motif. That characteristic makes the results of the analysis for each of the 3 epitopes potentially interdependent. Therefore, it is likely that the true ADE effects induced by one of these mAbs (or even some other mAb with a similar epitope) could be responsible for the increase in numbers of breakthrough HIV sequences containing epitopes of the other mAbs.

Interestingly, the increase in epitope-specific infection rates only occurred in VAX004 trial vaccinees but not in VAX003 trial vaccinees. In addition to being a convincing internal control, this observation raises the possibility that the enhancement of HIV infection was only induced by the AIDSVAX B/B immunogen used in VAX004. Indeed, epitopes targeted by the mAbs 268-D, 447–52D, and 537–10D were present in both the GNE8 and the MN strains used for the VAX004 trial vaccine but none of them were present in the A244 strain used together with the MN strain for the VAX003 trial vaccine.²⁶ It is also possible that an increase could have also occurred in VAX003 vaccinees but was not detected by the statistical approach employed due to low infection rates with viruses bearing the 268-D-targeted epitope (primarily subtype B) in the VAX003 Thai population where subtype E is prevalent (only 3 subjects were infected with 268-D epitope-decorated viruses in VAX003 trial, see Table 1). Finally, it is important to remember that the study population in the VAX004 trial (men-who-have-sex-with-men and heterosexual transmission) was different from the population of the VAX003 trial (IVDU and heterosexual transmission).

Interestingly, following the completion of the AIDSVAX trials, Gilbert, et al.²⁴ reported that VAX004 vaccinees with low rgp120 Ab responses had a rate of HIV infection higher than that of the ‘placebo’ cohort while the vaccinees with medium responses had a rate of infection comparable to that of the ‘placebo’ cohort and the vaccinees with high responses had a rate of infection lower than that of the ‘placebo’ cohort.²⁴ In their paper Gilbert, et al. discussed the possibility of infection enhancement in VAX004 vaccinees with low Ab responses but rejected the idea due to the lack of evidence. However, our analyses of these data raise the question of a possible AIDSVAX vaccine-induced enhancement of HIV infection yet again.

Intricate mechanisms, many of which have been poorly studied, could form the basis of the vaccine-induced enhancement of HIV infection.³² The detection of enhancement via the Ab-targeted epitopes, rather than any element of the Ab, revives the intriguing hypothesis that the observed enhancement is dependent on the tertiary structural presentation of epitopes on gp120²¹ rather than on the Ab Fc region as has been most frequently studied previously for HIV.^33,34 The study by Kliks, et al.²⁸ suggests that there may be a very delicate balance between the induction of neutralization and enhancement of infection by the same Ab species. The analysis discussed in the current review and the constantly evolving state-of-the-art epitope-delineation informatics approaches^30,31,35-38 may be important for mining diverse HIV-host data, such as clinical trial data, and understanding the molecular basis underlying the effects of ADE. Further experimental assessment of the serum from patients infected during the AIDSVAX and RV144 trials may also reveal important insights into the possibility of antibody-dependent enhancement of HIV infection upon vaccination.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Acknowledgments

We thank the NIAID-funded HIV Vaccine Trials Network and the STEP Study Team for providing sequences of breakthrough viruses from STEP clinical trial. We also thank Dr. Jennifer Fuller for providing comments and editing the manuscript. A GSID HIV Data Browser subscription was purchased for this work.

Funding

This work was supported by grants from the National Institutes of Health R01AI084119 and DP1DA036478 to TC.