Vaccines at Retrovirus 2004

By Richard Jefferys*

The 11th Conference on Retroviruses and Opportunistic Infections took place from 8-11 February, drawing over 3,500 researchers and activists from around the world into the cavernous halls of San Francisco’s Moscone Convention Center. The main focus of the annual meeting is basic science and HIV treatment, but this year the completion of the AIDSVAX efficacy trials and the controversies surrounding the Phase III Thai prime-boost study ensured that vaccines were a hot topic, both in conference sessions and during the informal hallway tête �  têtes that are the signature of scientific meetings.

Caught in the act: The potential of neutralizing antibodies targeting fusion

Antibody expert Dennis Burton (Scripps Institute, La Jolla) reviewed the state of scientific knowledge regarding neutralizing antibody (NAb) responses to HIV, focusing on the potential of strategies to block HIV as it fuses with target cells (viral fusion). Burton began with the role of NAbs in protection against viral diseases for which licensed vaccines exist. Resolution of an acute viral infection is thought to be mediated primarily by innate and cell-mediated (T cell) immune responses, and NAbs arise after the symptoms of acute infection are resolved. Once present, NAbs typically persist and are available to immediately bind and help block the virus if a secondary exposure occurs (memory B cells can also replenish supplies of NAbs if necessary). However, Burton believes it’s unlikely that NAbs mediate complete protection – some virus probably leaks through and is dealt with by memory T cells that were also generated during the initial acute infection. Licensed vaccines essentially attempt to mimic the effects of primary infection via the generation of NAbs and memory T cell and B cell responses. Burton stressed that these approaches typically work well if natural infection leads to the generation of NAbs. But HIV infection rarely induces such responses, and even when present they almost never demonstrate broad activity against a diverse range of primary viral isolates. Therefore Burton thinks that the use of simple mimicry as an HIV vaccine strategy is not likely to produce a protective NAb response.

Burton went on to discuss the reasons why HIV infection generally fails to induce NAbs. Chief among them is the poor accessibility of conserved regions of the viral envelope, which are protected from antibody recognition by a highly variable and glycosylated gp120 “glycan shield.” Nevertheless, Burton emphasized that several monoclonal antibodies have been isolated that do possess broad neutralizing activity. An unpublished study by Burton’s colleague James Binley, in collaboration with Chris Petropoulos from Virologic Inc., tested the activity of a panel of these monoclonal antibodies against close to 100 HIV primary isolates (a mixture of molecular clones and quasispecies) from multiple different clades using Virologic’s luciferase-based neutralization assay. The results indicated an antibody known as 4E10 possesses activity against all viruses tested, although Burton described the potency as “moderate.” The next best antibody in terms of breadth of neutralization was 2F5. Burton pointed out that these two monoclonal antibodies both target gp41, a part of the viral envelope that is submerged beneath the glycan shield and only briefly exposed during the process of fusion between HIV and a CD4+ T cell. The relative conservation of gp41 epitopes that are exposed during fusion makes them an attractive target for antibody-based vaccine strategies, Burton suggested.

He then emphasized the importance of ongoing research in two key areas. Firstly, it will be important to delineate the levels of NAbs required to mediate protection against HIV infection in vivo, and understand whether NAbs and cell-mediated immune responses can work together to enhance protective efficacy. Secondly, the optimal strategy for designing immunogens to induce NAb responses in humans remains to be discovered. Burton listed the four leading approaches that are currently being pursued:

  • Production of recombinant molecules that mimic the mature gp120 trimer on HIV’s surface.
  • Production of envelope molecules that may normally express neutralizing epitopes weakly but engineer them to better express these epitopes (by either subtracting elements from the envelope to try and reveal epitopes, or conversely by adding elements to mask regions that induce ineffective responses and thus appear to interfere with the generation of NAbs).
  • Generation of stable intermediates of the fusion process to expose conserved epitopes to which antibody could gain access during viral entry into a target cell.
  • Production of epitope mimics of the broadly neutralizing monoclonal antibodies determined from structural studies of the antibody-antigen complexes.

Burton is particularly excited about the third approach, because the broad neutralizing activity of 4E10 and 2F5 demonstrates that it is feasible; it had been feared that the process of viral fusion might be too brief to allow antibody-mediated inhibition. The brevity does seem, however, to generally prevent the generation of fusion-directed NAbs in the setting of natural HIV infection. Thus, the challenge for vaccine designers is to present antigens that elicit 4E10 and 2F5-like antibodies in humans, and Burton feels that if this challenge can be overcome “we might be able to attack a real weakness of the virus because the fusion machinery is really very well conserved.”

Consensus env gene sequences encode functional & immunogenic proteins

One attractive strategy for reducing the genetic divergence between HIV vaccines and circulating viruses is to build vaccines that are based on consensus or ancestral viral sequences (see Bette Korber’s presentation covered in IAVI Report, Oct-Nov 2001). A key question related to this approach is whether artificially constructed gene sequences can encode functional and structurally intact viral proteins. Several studies presented at the conference addressed this issue. Feng Gao (Duke University, Durham) has constructed an env gene (christened CON6) based on the consensus sequence for the entire group M of HIV. Gao confirmed the structural integrity of the encoded Env protein by demonstrating that it binds monoclonal antibodies known to recognize linear, conformational and glycan-dependent epitopes in a broad array of primary HIV isolates. He then inserted the CON6 gene into DNA and recombinant vaccinia virus vaccines for preliminary immunogenicity studies in mice. Compared to env genes from clade B and C isolates, CON6 induced T cell responses (primarily CD4+) that showed substantially broader cross-clade recognition of HIV epitopes. However, neutralizing antibody responses were poor.

Denise Kothe (University of Alabama at Birmingham) has taken a similar approach utilizing consensus sequences for clade B and C env genes (obtained from Bette Korber) as opposed to the entire group M. Kothe first verified that these genes could mediate infectivity if inserted into an env-deleted HIV backbone. These “pseudovirions” were also sensitive to neutralization by monoclonal antibodies that have activity against primary HIV isolates. In ongoing work, Kothe is attempting to improve the ability of the consensus proteins to induce effective antibodies by making modifications that may unmask neutralizing epitopes.

CTL control of SIVmac239 challenge after DNA/Sendai immunization

Tetsuro Matano (University of Tokyo) presented results from a preclinical study that may offer solace to researchers developing CTL-based HIV vaccines. Matano and colleagues immunized eight non-Indian macaques with a multi-gene DNA construct (at week 0) followed by a booster using a Sendai virus vector encoding the SIV Gag protein (at week 6). Four macaques received sham vaccinations and served as controls. Three months after the boost all animals were challenged intravenously with SIVmac239.

Surprisingly, 5/8 immunized macaques demonstrated robust control of SIVmac239 viremia, and Matano reported that this has persisted out to a year of follow-up. Three of the eight (and all controls) displayed high viral loads that are more typical of SIVmac239 infection. In an attempt to understand the outcome, Matano compared the CTL responses and viral sequences in the five macaques that controlled viremia to the three that did not. The numbers of Gag-specific CTL (as measured by IFN-γ production) were not significantly different between the two groups. However, when the genetic sequence of the SIV gag gene was analyzed, it transpired that all five controllers had consistent amino acid changes in their virus suggestive of CTL escape. In three animals known to be descended from the same parent (and therefore presumed to share a class I MHC haplotype), the escape mutation was pinned down to a lysine to serine change at position 216 of the Gag 207-216 epitope. No such changes were seen in the remaining three immunized macaques.

Searching for an explanation, Matano’s group constructed a SIVmac239 variant containing the escape mutation and found that it had reduced in vitro replicative capacity compared to wild-type virus. Two unrelated macaques were then challenged with a mixture of the wild-type and mutated SIVmac239 and while both variants could be detected one week after challenge, the escape mutation had disappeared a week later, further confirming its replicative disadvantage. Matano’s study echoes recently published work by David Watkins (see Research Briefs, page TK) suggesting that CTL escape can, under some circumstances, inflict a severe fitness cost on the virus. In response to a question from Jeff Lifson, Matano noted that the macaques used in this experiment were from Myanmar, and that SIVmac239 may be slightly less virulent in this subspecies compared to the Indian macaques that are typically used by American researchers. The shared parentage of several of the animals that controlled viremia also suggests that, while promising, these results need to be confirmed in a larger and more genetically diverse group of macaques.

Another presentation that offered an optimistic perspective on the potential benefits of CTL-based vaccines was a poster by Zoé Coutsinos (Institut Cochin, Paris). Coutsinos looked at the long term outcome after challenge with the highly pathogenic SIVmac251, based on the presence or absence of an SIV-specific CTL response. Out of 17 animals that developed CTL in response to immunization with a lipopeptide construct, 16 are free of signs of simian AIDS after two years of follow up. In contrast, out of seven macaques that did not develop CTL responses after vaccination, only one was asymptomatic after two years, a highly statistically significant difference.

Macaque low dose mucosal challenge models

A longstanding concern regarding the use of macaque models in AIDS vaccine research is that animals are challenged (usually intravenously) with doses of virus far higher than is typical in a real-world (typically mucosal) exposure to HIV. High doses are used to ensure that all unvaccinated control animals become infected, since failure to infect controls often prevents a clean analysis of any vaccine effects in the immunized animals. However, this equates to a per-exposure risk of infection of 100%, whereas the per-exposure risk of HIV infection in humans is thought to be in the range of 0.01% to as high as 10% in male-male transmission (Nat. Rev. Microbiol. 1:25; 2003).

Two studies presented at the conference—one theoretical and the other experimental—investigated novel study designs based on repeated, low-dose viral challenges. A poster by Roland Regoes and Silvija Staprans (Emory University, Atlanta) asserted that low-dose macaque challenge experiments involving repeated exposures may be a feasible methodology for assessing candidate vaccines, based on the results of a mathematical model. Their results suggested that a vaccine with 80% protective efficacy could be evaluated using as few as 5-7 animals per group with a statistical power of >95%. In contrast, a single-dose challenge experiment of a comparable vaccine would require as many as 20 animals per group to achieve the same statistical power.

Ron Otten (Centers for Disease Control, Atlanta) described the development of a low-dose vaginal challenge model in pig-tailed macaques. To assess feasibility, three groups of two animals were given weekly atraumatic vaginal challenges with either 10, 2 or 0.2 TCID50  (50% tissue culture infectious doses) of the R5-tropic SHIV162p3 virus. Both animals given the highest dose became systemically infected after three exposures, while those given 2 TCID50 became infected after 4-8 exposures. Neither animal that received the lowest dose became infected after twelve exposures.

Otten then utilized the 10 TCID50 dose to evaluate the potential protective efficacy of the microbicide candidate, cellulose acetate phthalate (CAP). Three of four CAP recipients remained uninfected after a total of 12 exposures, while three controls and one CAP-treated macaque became infected after the first three challenges. Despite the small numbers of animals used in the experiment, Otten was able to obtain a statistically significant p value of 0.015 when the rate of infection was compared between the controls and the CAP group. Otten noted that an attractive feature of this model is that the significance of the p value increases as a function of the number of exposures given, suggesting that the approach could also be applied to the evaluation of potential vaccine candidates. A study published after the conference by David Watkins also concludes that low dose mucosal challenges can be used to evaluate vaccine efficacy (J. Virol. 78:3140; 2004).

Disconnect between interferon-g production and cytotoxicity among HIV-specific CD8+ T cells

The daunting complexity of human immunology presents a severe challenge for researchers attempting to delineate the optimal technologies for measuring vaccine- or infection-induced immune responses. The challenge is well-illustrated by the fact that no definitive correlates of protection against HIV infection have yet emerged, and the immunological mechanisms by which rare HIV-infected long term non-progressors (LTNP) control viremia and maintain health remain relatively obscure. Despite these problems, researchers continue to make incremental progress, and a number of studies addressed some of the outstanding questions in this area.

Mathias Lichterfeld (Partners AIDS Research Center, Boston) presented new data on the measurement of HIV-specific CD8+ T cell responses in infected individuals. The widely used ELISPOT assay quantitates HIV-specific CD8+ T cells based on their ability to produce the cytokine interferon (IFN)-g, but this approach may not always capture cells capable of exerting cytotoxic (cell-killing) activity. Lichterfeld compared results obtained using ELISPOT with a new flow cytometry-based assay that assesses the cytotoxicity of CD8+ T cells by looking for the induction of caspase-3 substrates (a marker for apoptosis or cell death) in target cells. One advantage of the caspase-3 assay compared to traditional tests of cytotoxic activity is that it does not require any culturing of the CD8+ T cells in the laboratory. Lichterfeld found that the IFN-g ELISPOT results correlated poorly with cytotoxic activity. However, a superior correlation emerged when the ELISPOT was modified to capture CD8+ T cells producing the cytokine TNF-α in addition to IFN-g. Lichterfeld pointed out that these results are consistent with murine models of CD8+ T cell exhaustion, which have demonstrated that the capacity for IFN-g production can be maintained long after other important functions—such as cytotoxicity and the production of TNF-α—are lost.

IFN-g ELISPOT assays are also commonly used to quantify HIV-specific CD4+ T cell responses, but two studies also raised concerns as to whether this approach is capturing a fully functional cell population. Alexandre Harari (Laboratory of AIDS Immunopathogenesis, Lausanne) showed that in controlled CMV infection, virus-specific CD4+ T cells comprise a relatively balanced mix of populations secreting the cytokine IL-2 alone, IL-2+ IFN-g and IFN-g alone. HIV-specific CD4+ T cells in LTNPs displayed a similar cytokine profile. In stark contrast, HIV-specific CD4+ T cells in progressive infection were skewed toward a population only capable of making IFN-g. The quantity of HIV-specific CD4+ T cells as measured based on IFN-g production alone was not correlated with control of viral load, but there was an inverse correlation between the numbers of IL-2 and IL-2+ IFN-g producing cells and the level of HIV viremia. A poster from Harriet Robinson’s group at Emory University described a similar finding in HIV-infected children.

Stephen De Rosa (Vaccine Research Center, Bethesda) investigated the cytokine profiles of T cell responses to DNA vaccines encoding HIV gagpol and env in HIV-negative volunteers and compared them to those seen in infected individuals. While vaccination induced T cells that produced a somewhat similar mix of cytokines as those seen by Alexandre Harari in CMV infection and LTNPs (with some cells producing either IL-2 or IFN-g alone and others producing both), the proportion of cells producing the various cytokines differed in infected individuals, with a tendency to skew toward IFN-g production. These data suggest that that quantifying HIV-specific T cells based on IFN-g alone may not reveal the full spectrum of immune responses induced by vaccination, and that immunization might induce T cells with broader functional capabilities than those that typically arise in the setting of natural infection (although De Rosa acknowledged that the relationship between these various cell populations and the ability of a vaccine to protect against or control HIV infection remains unknown).

Beyond help: Direct effector functions of HIV-1-specific CD4+ T cells

The role of CD4+ T cells in the immune response is generally considered to involve the provision of help to CD8+ T cell and B cells. In the absence of such help, CD8+ T cell and B cell responses generally function poorly or not at all. Occasionally, however, it appears that CD4+ T cells can exert direct cytotoxic activity themselves. Phillip Norris (Partners AIDS Research Center, Boston) has been investigating this phenomenon in the setting of HIV infection and in a poster documented cytotoxic activity by HIV-specific CD4+ T cells in four out of ten individuals studied. There has been controversy as to whether such activity may emerge as a result of long term culture in the laboratory, but in one individual the cells could be identified directly ex vivo. This particular study participant has maintained a viral load of <50 copies for over 20 years without treatment, and his HIV-specific CD4+ T cells could mediate >3 log suppression of HIV replication in vitro. A poster by John Zaunders also reported the identification of a cytotoxic HIV-specific CD4 T cell population in an infected LTNP and identified a similar CMV-specific population in healthy CMV-positive adults. It appears that these cells may represent an underappreciated component of antiviral immunity in humans.

VaxGen denouement: No efficacy in racial subgroups, no efficacy in Thai trial

Dean Follman (National Institutes of Health, Bethesda) addressed the one lingering question regarding VAX 004, the recently completed efficacy trial of VaxGen’s recombinant gp120 AIDSVAX vaccine candidate. The question relates to VaxGen’s controversial claim that AIDSVAX showed statistically significant protective efficacy among black study participants and a combination of racial subgroups that the company characterized as “Blacks, Asians and Others.” Follman led a committee appointed to investigate this claim, made up of representatives from NIH, Centers for Disease Control (CDC), the SCHARP Statistical Center at the University of Washington, and VaxGen.

One of the basic tenets of statistical analyses is that conducting multiple comparisons of different subgroups in a trial greatly increases the risk of finding an erroneously significant result. There are several standard statistical tools that can be applied in order to reduce the risk of error, but when VaxGen first presented their trial results, these tools had not been employed (although the company initially claimed otherwise). After reanalyzing the data, Follman’s team found that a significant result could be obtained by chance about 22 – 24% of the time when data from 15 subgroups were evaluated, leading them to conclude that the unadjusted subgroup data originally presented by VaxGen were spurious.

In the same conference session, Punnee Pitisutithum (Mahidol University, Bangkok) gave the first public presentation of results from the second AIDSVAX efficacy trial, which recruited 2,546 Thai intravenous drug users at high risk for HIV infection. As reported last November in a press release from VaxGen, the vaccine again failed to show efficacy. Pitisutithum fleshed out the details, showing that 106/1161 vaccine recipients became infected vs. 105/1155 that received placebo. Pitisutithum also stated that there were no significant differences in viral load, CD4+ T cell counts or time to initiation of antiretroviral therapy (ART) between vaccinees and placebo recipients. Taken together, Pitisutithum’s and Follman’s data sets should finally lay to rest any outstanding questions regarding the efficacy of AIDSVAX.

*Richard Jefferys is Basic Science Project Director at the Treatment Action Group, a New York-based organization advocating for HIV research