A systems approach to HIV-1 vaccines

Understanding the full breadth of immune responses that protect against HIV-1 may speed vaccine development.

More than three decades after the first reports of AIDS, we still lack an effective HIV-1 vaccine. Recent clinical trial results have hinted at the complexity of the immune responses that could protect against HIV-1. Not only broadly neutralizing antibodies but also immune functions mediated by the antibody Fc (fragment-crystallizable) region seem to be protective. A comprehensive description of anti-HIV-1 immunity could aid the design and testing of improved vaccines. In a recent paper in Cell, Chung et al.¹ studied the antibody profiles of participants in four vaccine clinical trials in unprecedented detail and used a systems analysis to better define immune 'correlates of protection' (Fig. 1 and Table 1).

Figure 1: An integrated serological profiling approach reported by Chung et al.¹ for identification of unique humoral Fc-related vaccine signatures (left) compared to other approaches (right).

Katie Vicari/Nature Publishing Group

Chung et al.¹ purified IgG from participants in four different HIV-1 vaccine trials (RV144, VAX003, HVTN204 and IPCAVD001). Six Fc effector functions and 58 biophysical characteristics were evaluated, including binding to Fcγ receptors and relative abundances of different antigen-specific antibodies in 120 samples. All 64 parameters were subjected to statistical analyses, including hierarchical clustering, multidimensional combined feature selection method, partial least-squares discriminant analysis and correlation network analysis, to identify unique signatures and differences that may be related to protection. Previous methods typically included many of these parameters but were rarely based on extensive evaluation of coordinated Fc functions.

Table 1 The four trials analyzed by Chung et al.¹

Vaccine developers have traditionally focused on discovering agents that elicit high titers of neutralizing antibodies against the pathogen. For HIV-1, considerable evidence suggests that neutralizing antibodies are indeed protective. In nonhuman primate models of HIV-1 infection, passive administration of neutralizing antibodies has achieved complete protection². A recent first-in-man trial showed that a broadly neutralizing antibody—an antibody capable of neutralizing most viral strains—can significantly reduce viral load in HIV-1-infected individuals³. However, monoclonal antibodies are expensive, and they are typically administered therapeutically, after infection, rather than prophylactically. Therefore, the quest for an efficacious AIDS vaccine continues.

One promising strategy, initiated by our studies of germline precursors of broadly neutralizing antibodies⁴, is to develop immunogens that stimulate maturation of antibody precursors⁵. Substantial effort along these lines is ongoing⁶, but the complexity of antibody maturation pathways remains a challenge. Another approach is to look beyond neutralizing antibody activity to other immune correlates of protection. Both neutralizing and non-neutralizing antibodies against HIV-1 could act through antibody Fc–related functions, such as antibody-dependent cellular cytotoxicity (ADCC), antibody-dependent cellular phagocytosis (ADCP) and antibody-dependent complement deposition (ADCD).

Recent analyses of HIV-1 vaccine trials seem to support a role of Fc-dependent processes. Of particular interest is the RV144 trial, a randomized, double-blind efficacy trial of a prime-boost HIV-1 vaccine that showed a modest (31.2%) reduction in the risk of infection at 42 months⁷. Interestingly, broadly neutralizing antibodies were not detected in any participant analyzed. Subsequent studies of samples from this trial found that antibodies (IgG) specific to a particular region (V1V2) of the HIV-1 envelope glycoprotein (Env) correlated inversely, and other antibodies (IgA) directly, with infection rates among those who were vaccinated⁸, and that IgG3 was important for protection⁹. In vaccinees with low IgA levels, ADCC, IgG avidity and neutralizing activity were inversely correlated with infection⁷, indicating that ADCC could contribute to protection in some vaccinees⁸.

In their 2014 study, Chung et al.¹⁰ sought to further dissect the role of Fc-mediated responses in the RV144 trial. They compared antibody profiles in plasma samples from 20 placebo recipients and 80 vaccinees with those from 10 placebo recipients and 50 vaccinees in the ineffective VAX003 trial. They showed that in the RV144 trial, highly functional IgG3 was elicited, whereas in the VAX003 trial the IgG3 response was negatively affected by IgG4, and also that IgG1 and IgG3 antibodies targeting the crown of the HIV-1 Env V2 were elicited only in the RV144 trial. These findings are important because the differences between IgG3 and IgG4, which are associated with coordinated responses targeting V2 epitopes, may underlie the observed modest efficacy in the RV144 trial. Interestingly, depletion of IgG3 significantly reduced both ADCP and ADCC in vitro, indicating that Fc effector functions could be important in vivo.

In the Cell paper just published, Chung et al.¹ expanded their previous study¹⁰ to include two additional trials that failed to show protection (HVTN204 and IPCAVD001) (Table 1). They measured multiple parameters using blood samples from all four trials: Fc-related functions (ADCC, ADCP, ADCD and three antibody-dependent natural killer (NK) cell functions—secretion of IFNγ and of MIP1β and degranulation) together with biophysical parameters (58 relative concentrations, including bulk IgG, IgG1, IgG2, IgG3 and IgG4; binding to gp140, gp70, p24, gp120 and gp41 from clades B, AE and A; ten relative concentrations of IgA included only in the analysis of the RV144 trial; and binding to FcγR2a, FcγR2b and FcγR23b). The analysis of such a large number of parameters from four different clinical trials was enabled by the use of much more sophisticated statistical analyses than the simple correlation statistics used in previous studies. This integrative approach should help to identify new parameters, and relationships between them, that could not only identify correlates of protection but also be useful in elucidating mechanisms of antibody elicitation by vaccine immunogens.

The authors initially used unsupervised hierarchical clustering to group vaccine regimens by the type of immunogen involved. Although this method revealed the dominance of immunogen type in eliciting distinct profiles, it was not possible to dissect specific features responsible for the separation of clusters. To obtain enhanced resolution of vaccine signatures, the authors used classification and discrimination methods as well as a correlation network analysis. The RV144 and VAX003 profiles were clearly separated through the use of as few as seven features. The analysis confirmed the previous finding of elevated IgG3 in the RV144 trial and dominance of IgG4 in the VAX003 trial. It also identified features that might be related to the lack of protection in VAX003, including elevated gp140-specific antibodies, NK cell degranulation and chemokine secretion, indicating that induction of those responses might not be important for an efficacious HIV-1 vaccine.

The authors also found that clear separation of the antibody signatures of the four vaccine trials could be achieved by analysis of just 15 of the 64 measured parameters. One interesting result was that the antigen itself, rather than the vector or immunization regimen, significantly affected the type of antibody profile elicited. The correlation networks for the four trials proved to be very different; this part of the analysis provides unique information about relationships between features that contribute to differences among the trials. Identification of networks with specific feature-function relationships could help to elucidate mechanisms underlying protection. For example, in the RV144 trial, the total IgG binding to V1V2 regions from the Env clades A and E was directly tethered to both ADCP and ADCC, suggesting that this specific response to V1V2 could drive the killing of infected cells through ADCP and ADCC.

The 'systems serology' analysis presented by Chung et al.¹ suggests that IgG3 in the RV144 could either be a surrogate of effective responses or contribute to an effective response through combination with multiple other antibody features, particularly IgG1. Interestingly, they also found that IgA was an antagonist of all of the IgGs specific to V1V2 except IgG3, and therefore IgA could be a marker of less effective immune responses. Finally, their results also indicated that ADCP could be crucial in combination with Fc-related functions for protection from the transmission of HIV-1 through mucosal barriers.

Although the approach of Chung et al.¹ promises to enable exquisitely detailed analyses of Fc-related vaccine signatures, there are practical and methodological concerns that will have to be taken into consideration when undertaking such analyses. For example, the assays used for the Fc effector functions are not easy to standardize across laboratories. Future studies would benefit from the use of blinded samples. It would also be of interest to test samples from lymphoid tissues, the major site of HIV-1 replication, in addition to blood samples. The results of systems analyses depend on the breadth and depth of the training data set and often cannot be generalized. Thus, the authors' conclusions may not hold for other immunogens or viruses. Finally, in the RV144 trial, only very limited protection was achieved and less than 1% of all participants were infected, so the significance of any identified correlate of protection remains to be determined.

Nevertheless, a systems serology approach could be readily applied to other trials and expanded to include additional parameters, such as sequence analysis of antibodyomes⁴ and molecular signatures of vaccines. Such studies might also be designed to address the important question of how to predict the efficacy of a novel vaccine based on in vitro studies or animal models. Finding the immunological correlates of vaccine protection for HIV-1 infection by considering qualitative and quantitative immune responses could help guide us to the next generation of vaccines¹¹.