A Bangkok Surprise

Results of the immune correlates analysis of RV144 and advances in broadly neutralizing antibodies topped the developments reported at the annual AIDS vaccine conference

By Kristen Jill Kresge

If you polled researchers in the HIV vaccine field three years ago and asked them what they thought would come out of the RV144 trial in Thailand, the most popular answer would likely have been nothing. The prime-boost combination of Sanofi Pasteur’s canarypox vector-based candidate ALVAC-HIV (vCP1521) with AIDSVAX B/E, a genetically engineered version of HIV’s gp120 surface protein, was considered unlikely to work and many scientists thought the field would reap greater benefits by developing better candidates than by testing this regimen in the largest trial to date, involving more than 16,000 Thai volunteers (1).

Even after the results of RV144 were released two years ago showing the prime-boost regimen provided 31.2% protection against HIV infection—the first evidence of vaccine-induced protection—several researchers were still skeptical. The modest efficacy inspired many to raise questions about whether the protective effect was real or just a statistical fluke.

But now there may be fewer skeptics. After a two-year effort to elucidate the possible immune responses that correlated with the protection seen in RV144, the trial has once again yielded surprising findings. At the AIDS Vaccine 2011 conference that took place Sep. 12-15 in Bangkok, Thailand, Barton Haynes, who led the scientific steering committee that oversaw the collaborative and thorough RV144 correlates search, reported that two antibody responses were found to be significantly correlated with the risk of HIV infection among vaccine recipients in RV144. This finding generated several hypotheses and helped dispel doubts about whether the modest efficacy exhibited by the vaccine regimen was real. "The findings lend credence to the vaccine efficacy seen in the RV144 trial," said Haynes.

The first surprise, even for Haynes, was that any correlates at all were identified. Given the lack of support for RV144 when the trial began, investigators scaled back sample collection in the trial. This made the analysis somewhat akin to searching for a needle in a haystack. The second surprise was that while one antibody response was inversely correlated with HIV infection risk, the other was directly correlated with infection risk, suggesting this antibody response reduced the protective effect of the vaccine candidates.

While the correlates results are intriguing, it is still unclear exactly whether the antibody responses were directly responsible for the modest protection. Researchers are now aggressively investigating this.

Meanwhile, the blizzard of more potent and broadly neutralizing antibodies against HIV and elucidation of their targets on the virus has propelled research into the design of immunogens capable of eliciting such broadly neutralizing antibodies. Some of the advances in determining the structure of the targets of the broadly neutralizing antibodies (bNAbs) and the first generation of such immunogens were also presented in Bangkok.

The hunt for correlates 

The RV144 correlates presented at AIDS Vaccine 2011 were the result of a collaborative process by a team of researchers established soon after the results were reported in 2009 at the AIDS Vaccine meeting in Paris (seeRaft of Results Energizes ResearchersIAVI Report, Sep.-Oct. 2009). Since that time, under Haynes’s supervision, the team conducted a series of pilot studies, eventually settling on six primary and approximately 30 secondary assays that were used in case-controlled studies to try to determine what immunological measurements predicted HIV infection risk among RV144 volunteers over a three-year period.

The six primary assays that were selected measured the following immune responses: binding immunoglobulin (Ig)A antibodies in plasma; IgG antibody avidity to the A244 gp120, the antigen used in the vaccine candidates; antibody-dependent cellular cytotoxicity (ADCC); neutralizing antibodies against a six-tier panel of HIV isolates; binding IgG antibodies to the first and second variable loops of HIV Envelope, known as V1 and V2, scaffolded onto a gp70 from murine leukemia virus; and CD4+ T-cell responses as measured by secretion of several cytokines/chemokines, including interferon-γ, interleukin-2, and tumor necrosis factor α, among others.

The statistical analysis plan was developed and carried out by Peter Gilbert and colleagues at the Statistical Center for HIV/AIDS Research and Prevention (SCHARP), based in Seattle. The analysis had 80% power to detect an approximately 50% reduction in HIV infection rate and controlled for variables including gender and baseline behavioral risk in the multi-variate analysis. An independent team of statisticians validated the statistical analyses by Gilbert and colleagues.

For the case-controlled studies, the correlates team analyzed samples from 41 HIV-infected vaccine recipients, 205 uninfected vaccine recipients, and 40 placebo recipients. The samples used in the analysis were collected at week 26 of the trial—two weeks after all six vaccinations (four ALVAC primes and two AIDSVAX boosts) were administered—when the immunogenicity peaked.

The results of the assays identified two so-called correlates of risk. The first statistically significant correlate was IgG antibodies that bind to the V1/V2 loops of HIV Env. The presence of these antibodies correlated with a 43% reduction in HIV infection rate. For volunteers with high levels of IgG binding antibodies to V1/V2 compared to those with medium- or low-levels of these antibodies, there was a 75% reduction in HIV infection rate. Volunteers with high levels of V1/V2 antibodies appeared to be protected, while those with lower levels received little or no protection from the vaccine regimen, Haynes said.

The second immune response identified as a statistically significant correlate of risk was plasma IgA antibodies that bind HIV Env. These IgA antibody responses were directly correlated with a 54% increase in HIV infection rate among vaccinated volunteers, suggesting these antibody responses reduced the protective effect of the vaccine regimen. There was, however, no evidence that these IgA responses were associated with an enhanced risk of HIV infection. When researchers compared HIV-infected vaccinees to placebo recipients, they found that the HIV infection rates among these groups were the same.

To investigate this correlate further, researchers did epitope mapping of the IgA responses to gp120. They identified the C1 peptide on gp120 as the binding site of the antibodies, which has been shown to be the target epitope of ADCC, according to Haynes. This led him to suggest that IgA antibodies that bind to C1 may block ADCC, a mechanism by which antibodies facilitate the destruction of HIV-infected cells (see Antibodies: Beyond NeutralizationIAVI Report, Jan.-Feb. 2010). This phenomenon has been observed in cancer, according to Haynes, who says that IgA antibodies can block ADCC responses against tumors.

Further exploration of this mechanism is now underway. Researchers are conducting an exploratory analysis to see whether low or high levels of IgA antibodies in plasma had an effect on ADCC responses, which were also measured in one of the six primary assays. Haynes also reports that additional studies will be conducted to see if plasma IgA interferes with any other immune responses.

Trial investigators did not collect any mucosal samples in RV144, so only plasma IgA antibodies can be analyzed. Secretory IgA at the mucosa, where it predominates, is in the form of a dimer, while only four percent of plasma IgA is dimeric. Haynes says plasma IgA also has a different potency than its mucosal counterpart. "It’s an open question about what implications this finding may have for mucosal immunity," said Haynes, adding that collection of mucosal secretions will definitely be incorporated into the RV144 follow-up studies.

Researchers also have several other studies planned or already underway to further investigate the correlates. "They [the correlates] give us an important lead on improving on these responses," said Haynes. "Now we have informed hypotheses and directions that come from a trial."

Haynes’s lab will be conducting passive administration studies of V2 monoclonal antibodies identified as a correlate in RV144 in nonhuman primate (NHP) studies to see if they are protective following challenge with a simian immunodeficiency virus (SIV)/HIV hybrid known as SHIV. Additional in vivo studies will also be used to investigate if IgA antibodies are capable of blocking ADCC.

There are also plans to re-evaluate the samples collected in the VAX003 and VAX004 trials—two Phase III trials conducted with only AIDSVAX. Both of these trials showed that AIDSVAX alone provided no protection in either men who have sex with men (MSM), injection drug users (IDUs), or high-risk women, but once again the samples from these trials may provide useful clues. The correlates findings from RV144 have sparked interest in whether similar immune responses were induced in VAX003 and 004 vaccine recipients but were perhaps overshadowed by the viral diversity or the quantity of virus the high-risk volunteers in these two trials were exposed to. "Challenge dose may overwhelm immunity," said veteran vaccinologist Stanley Plotkin, citing the Polio vaccine as an example.


Genoveffa Franchini, chief of the animal models and retroviral vaccine sections at the National Cancer Institute, has seen evidence for this in NHPs. Using a low-dose challenge model, Franchini can replicate the RV144 results in a monkey model with about 30% of macaques protected against SIVmac251 challenge following vaccination with a similar regimen based on SIV. However, if the dose is increased, the protective effect is lost. "If you’re using too much virus you can’t see vaccine efficacy," said Franchini. There are also similarities in the antibody responses elicited in NHPs. The animals protected against low-dose challenge appear to have higher anti-gp120 antibody levels, and the antibody avidity to the V2 loop appears to be important, though Franchini says more experiments are necessary to show whether this is the mechanism of protection in the macaque studies.

In addition to challenge dose, the mode of transmission may also be important. One of the unique aspects of RV144 was the trial population—volunteers were largely heterosexuals at low risk of acquiring HIV. Studies of the earliest stages of HIV infection have shown that heterosexual transmission is predominantly (about 80% of the time) the result of a single transmitted or founder virus that establishes infection (see HIV Transmission: The Genetic BottleneckIAVI Report, Nov.-Dec. 2008). Whereas, in MSM and IDUs, researchers have reported that on average a much higher number of founder viruses are transmitted. Katharine Bar of the University of Alabama at Birmingham summarized the research findings in MSM and IDUs. She said that two studies have shown that in approximately 40% of MSM, multiple founder viruses established infection, while two studies in IDUs have provided conflicting results. In one study, approximately 60% of volunteers in a small cohort of IDUs were infected by multiple founder viruses (as many as 16 variants), while in another small study the percentage of IDUs infected with multiple transmitted founder viruses was the same as MSM.

Bar said that in VAX003, 44% of HIV-infected volunteers were infected with multiple variants, which she said sets a higher bar for vaccine protection than in heterosexual cohorts, such as the population enrolled in RV144. "AIDSVAX may have had a modest vaccine effect that did not rise to the level of overt vaccine protection," she said. In contrast to the VAX003 population, Morgane Rolland of the University of Washington reported in Bangkok that 75% of HIV infections in RV144, among both vaccine and placebo recipients, were the result of a single transmitted founder virus.


What next? 

No one is quite sure exactly what the correlates findings mean for the development of an HIV vaccine. Giuseppe Pantaleo, chief of the division of immunology and allergy at the Centre Hospitalier Universitaire Vaudois in Lausanne, Switzerland, is confident that the RV144 trial will continue to inform the field. "The RV144 correlates work is clearly going to guide us on the future of HIV vaccine development," he said. Jerome Kim, deputy director of science at the US Military HIV Research Program, a key collaborator on RV144, was more cautious. "Any results may be unique to this vaccine," he said. "We have to bear that in mind as we look to the next step in HIV vaccine development."

Other researchers think that a more effective vaccine will have to induce bNAbs against the virus. "Even though we’re getting advances toward non-neutralizing antibodies, there’s still a big gap in potency between these and broadly neutralizing antibodies," said Robin Shattock, professor of mucosal infection and immunity at Imperial College London.

Meanwhile, other researchers see the correlates findings as a way to improve upon the 31% efficacy seen in RV144, perhaps even increasing the efficacy to a high enough level that it might lead to a first generation AIDS vaccine. "Even a partially effective vaccine, if it reduced transmission, would have an overall effectiveness that would be quite high," said Plotkin. And this pathway to a partially effective vaccine means finding one that works through non-neutralizing activities. "I believe in neutralizing antibodies, but there’s more than one way to skin a cat," said Robert Gallo, founder and director of the Institute of Human Virology (IHV) in Maryland.

The first step in improving on the RV144 results is extending the duration of the immune responses. In RV144, efficacy after one year (six months after the full vaccine regimen was administered) was as high as 60%. Although measuring efficacy at this time point was not part of the pre-specified trial analysis, it has been intriguing to many researchers and suggests that improving the durability of the immune responses induced by this vaccine regimen might dramatically increase the efficacy.

Nicos Karasavvas of the Armed Forces Research Institute of Medical Sciences quantified the decline in IgG antibody responses that occurred in the RV144 trial based on the results of a peptide microarray evaluation. This evaluation showed that IgG antibody responses to cyclic V2 peptides dropped significantly by 28 weeks after the last injection. "They declined very rapidly with time," said Karasavvas. A 10-fold decrease in antibody responses occurred between two weeks after the final injection and 28 weeks after the final vaccination was administered.

To improve upon the durability of the immune responses, researchers are planning follow-up studies to RV144 with an additional AIDSVAX boost. The hope is this will extend the efficacy seen in RV144 beyond one year.

Gallo has had similar problems with duration of immune responses in his work at IHV. "Antibodies to Envelope don’t last," he said. In experiments with colleagues at IHV, Gallo says they have seen sterilizing protection against a repeat, low-dose SHIV challenge in NHPs vaccinated with their full-length single chain immunogen without induction of conventional bNAbs, but the protection is lost in about four months (seeVaccine BriefsIAVI Report, May-June 2011).

Another approach to improving on the RV144 efficacy is using a different canarypox vector. One of the alternative vectors being considered for RV144 follow-up trials is NYVAC, a canarypox vector developed by the EuroVac consortium. Pantaleo presented data in Bangkok on the immune responses induced by the second-generation NYVAC vector expressing trimeric gp140. This vector has been evaluated in combination with an HIV Env boost, and this induced better immune responses when an additional boost was administered at 12 months than a DNA prime and NYVAC boost, according to Pantaleo. He also presented on a second-generation, replication competent NYVAC vector that he says induced even better antibody and ADCC responses than the replication deficient form.

Yet another approach to improving on the RV144 results is changing the immunogen. According to Haynes, the correlates results are already leading to the design of new immunogens. Part of this work involves gaining a better understanding of the gp120 antigen AE.A244 that was tested in RV144. This was a relatively unique immunogen, according to Kim, and part of this uniqueness stems from a gD protein from herpes simplex virus that is tagged on to part of the A244 antigen. This antigen was originally developed at the biotechnology company Genentech and the gD protein was included because of its ability to pull antibodies out of serum.

Inclusion of this protein in the A244 gp120 antigen results in a 10-fold increase in binding affinity with CH01 and PG9, two of the recently identified bNAbs that target the V2 and V3 loops of HIV Env. While binding to these quaternary antibodies was increased, the gD molecule on A244 has a marginal effect on gp120 binding to linear epitopes. The A244 antigen also binds to and recognizes the germline sequences of these antibodies, according to Kim. "Any vaccine antigen has to bind to and recognize germline sequence and A244 does this," he says. But just what role the A244 antigen played in the protection afforded by the RV144 vaccine regimen is unknown. Haynes’s lab is currently exploring this.

Nabbing more bNAbs 

The flurry of discoveries of new bNAbs has become more like a blizzard. Among the latest additions to the antibody armamentarium are a collection of 17 antibodies isolated from four individuals from IAVI’s cohort of chronically HIV-infected individuals (2)—the same cohort that first led to the isolation of PG9 and PG16, two of the more potent bNAbs identified in the past few years (3).

The target of many of the 17 new antibodies is different, however, from that of PG9 and PG16. "All of them seem to be glycan dependent," said Ian Wilson, the Hansen professor of structural biology at The Scripps Research Institute (TSRI) in La Jolla, California. Another antibody, 2G12, one of the original handful of bNAbs that researchers had to work with, is also glycan dependent, but according to Wilson these new antibodies are more complex than 2G12 and also are "extremely potent." He says that several antibodies in this new family are 10 times more potent than recently isolated bNAbs, and 100 times more potent than other antibodies that were described much earlier. The potency and neutralization breadth of the PGT antibody family were assessed using a panel of 162 HIV pseudoviruses, representing all HIV subtypes currently in circulation. While a few of the PGT antibodies are not as broadly neutralizing, many of them are quite potent. For example, the PGT128 antibody neutralizes about 70% of HIV isolates, as compared to approximately 80% of isolates for PG9 and PG16, but PGT128 can neutralize 50% of isolates at a concentration of 0.1 μg/ml, illustrating its potency.

Wilson and colleagues, in collaboration with Bill Schief, now a principal scientist at IAVI’s Neutralizing Antibody Center based at TSRI, then determined the crystal structure of the Fab (Y-shaped) portion of the PGT128 antibody in complex with an engineered glycosylated gp120 outer domain construct containing a truncated V3 loop, work that has since been published (Science 2011 doi:10.1126/science.1213256). This crystal structure revealed that PGT128 engages two glycans, as well as the terminal end of the V3 loop. "This is a pretty extensive epitope," said Wilson. Susan Zolla-Pazner, a professor of pathology at New York University, said that while the PGT family of antibodies is certainly more potent and broadly neutralizing than other antibodies, the phenomenon of V3-targeting antibodies also binding to glycans had been documented previously (4).

Wilson said that when the PGT128 antibody was modeled onto the HIV Env trimer using electron microscopy, it showed that "this epitope is pretty accessible." This, along with the fact that PGT128 is so potent, suggests that this epitope may be a good target on which to design vaccine immunogens.

But do they protect? 

The next frontier in developing bNAb-based vaccine candidates is designing vaccine immunogens capable of inducing these bNAbs in HIV-uninfected individuals. At the VRC, the focus is primarily on designing immunogens based on the CD4-binding site, which is the target of bNAbs such as VRC01. "We want the most minimal immunogen that has the CD4 site but nothing else," said Jeffrey Boyington, a staff scientist at the VRC. To accomplish this, Boyington and colleagues are adding their CD4-binding site immunogen into the E2 domain of a Chikungunya virus-like particle (VLP).

When this fusion construct was tested in rabbits, researchers found the sera could neutralize some tier 1 HIV isolates after two or three injections. In experiments in rhesus macaques, researchers found injecting the animals with gp140 trimers followed by a boost with the modified Chikungunya VLP construct resulted in elicitation of neutralizing antibodies with greater specificity to the CD4 binding site, according to Boyington. Although these antibodies can only neutralize tier 1 isolates, it provides a proof of principle, and now there are many more tools and immunogens in the pipeline, Boyington said. "We have a long way to go but I think there’s a path ahead of us," said Barney Graham, chief of the clinical trials core laboratory at the VRC.

While immunogen design work continues, Graham and colleagues are preparing for trials to test passive immunization of the bNAb VRC01 to answer the question of whether this antibody is capable of blocking HIV infection. Studies in NHPs have shown passive immunization of VRC01 is able to block SHIV infection. At a dose of 20 mg/kg, VRC01 prevents infection following a high-dose rectal SHIV challenge in all of the monkeys studied, while at a lower dose of 5 mg/kg of VRC01, only half (two of four) of the monkeys were protected, suggesting dose of antibody is critical to the level of protection.

The VRC plans to conduct a series of Phase I and II trials of passive immunization of VRC01 to establish the proof of concept that this antibody is capable of blocking HIV infection. The goal is also to define the specificity, potency, and function of VRC01 to provide targets for vaccine-induced protection.

The trials include a Phase I study in HIV-infected and uninfected adults, a Phase I safety and pharmacokinetic study in infants in the US, a Phase I safety study in infants in developing countries, and a Phase IIb study to see if VRC01 can prevent HIV infection in infants. The first Phase I study in adults is expected to start in late December, and if all goes as planned, the Phase IIb study in infants would start in mid-2013.

Originally, the VRC also planned to conduct a Phase IIb trial in adults; however, this study is on hold for now because it would require producing 30kg of VRC01, as compared to 3kg for all of the other Phase I and IIb studies combined. "The problem is in the manufacturing," said Graham, who acknowledged that this trial would be the most informative. Researchers are now looking at several different ways to engineer the antibody so less would need to be delivered. They are also trying to improve the manufacturing capacity.

1. Science 303, 316, 2004
2. Nature 477, 466, 2011
3. Science 326, 285, 2009
4. J. Gen.Virol. 73, 3099, 1992; J. Virol. 76, 9035, 2002