Skip directly to content

Adenovirus Vaccine Vectors and HIV Acquisition Risk: The View from the Summit

By Richard Jefferys, Project Director

On September 19th, 2013, the National Institute of Allergy and Infectious Diseases (NIAID) sponsored a daylong mini-summit to discuss the safety of adenovirus vaccine vectors in populations at risk for HIV infection. The meeting was prompted by an accumulation of data indicating that receipt of an adenovirus serotype 5 (Ad5)-based HIV vaccine candidate significantly enhanced, rather than reduced, the risk of acquiring infection in two efficacy trials: STEP, which primarily recruited men who have sex with men in the Americas, and Phambili, which recruited heterosexual men and women in South Africa (these trials, particularly STEP, have been covered extensively on the blog in the past – see the adenovirus vectors category).

In April 2014, a perspective article was published in the journal Science offering a brief summary of the proceedings. The authors are Anthony Fauci, Director of NIAID, Mary Maravorich, Director of the Vaccine Research Program in the Division of AIDS at NIAID, Carl Dieffenbach, Director of the Division of AIDS at NIAID, and two researchers who served as the summit co-chairs: Eric Hunter and Susan Buchbinder. One unequivocal recommendation that emerged from the meeting is highlighted: further HIV vaccine studies using Ad5 vectors are “not appropriate.” Development of Ad5 vectors for other diseases also appears unlikely, but whether the Ad vectors based on alternative serotypes that are currently being studied might increase the risk of acquiring HIV infection remains unclear.

To briefly revisit the history: Ad vectors have proven vastly superior to previous vectors (such as those based on poxviruses) for inducing CD8 T cell immune responses. Because evidence indicates that CD8 T cells play an important role in controlling several pathogens that have proven difficult to design vaccines against, including HIV, TB, malaria, and hepatitis C, there is understandably a great deal of interest in the potential to develop vaccines for these infections using the adenovirus vector platform. Three efficacy trials involving Ad5-based HIV vaccine candidates have been conducted: STEP, Phambili and HVTN 505.

STEP and Phambili were designed to administer three shots of a Merck Ad5 vaccine encoding Gag, Pol and Nef proteins. Vaccinations in STEP were stopped in September 2007 due to lack of efficacy, but evidence quickly emerged of an enhanced risk of HIV infection in a subgroup of participants (uncircumcised men with high titers of preexisting antibodies to Ad5). Over extended follow-up, the increased risk of HIV among vaccine recipients became statistically significant when the entire trial population was analyzed.

The STEP results led to the early cessation of immunizations in the Phambili trial (most participants only received one or two of the planned series of three shots); initial results from Phambili did not show enhancement but, as principal investigator Glenda Gray explained at the summit, a statistically significant difference in HIV infections between the vaccine and placebo groups has emerged over time (these data have since been published in Lancet Infectious Diseases).

HVTN 505 is a more recent trial conducted in men who have sex with men (MSM) and transgender women in the Americas that was designed to avoid risk of enhanced HIV acquisition by limiting enrollment to circumcised individuals with no detectable preexisting antibodies to Ad5. The vaccine regimen comprised three DNA immunizations and just a single shot of a different Ad5 vector to that used in STEP and Phambili, encoding multiple HIV Env proteins from different virus subtypes in addition to Gag, Pol and Nef. The trial did not show efficacy and, as described by Scott Hammer at the summit, follow up to date has not revealed evidence of an increased HIV acquisition risk among vaccine participants.

Peter Gilbert presented a meta-analysis of the three trials, which confirmed a significant 32-33% increase in the risk of HIV acquisition among vaccine recipients that was essentially entirely driven by the STEP and Phambili results (which showed a 41% increase in risk when analyzed separately from HVTN 505). Gilbert also delved into subgroup data, noting that the elevation in HIV risk associated with vaccination was greatest in uncircumcised participants with antibodies to Ad5 in the period up to 18 months of follow up, but then appeared to shift to circumcised individuals who lacked antibodies to Ad5 at baseline. However, he noted that results from earlier timepoints were likely more robust because more of the data accrued during the period before the studies were unblinded.

During the discussion following Gilbert’s talk, Glenda Gray cautioned against reading too much into the subgroup analysis, noting that the selection of 18 months as the cut off was arbitrary. Several participants pondered whether the results might pin the blame for enhancing HIV risk solely on the Merck Ad5 vector, but in response it was pointed out that HVTN 505 excluded the population that appeared to be at highest risk in STEP. Furthermore, follow up from HVTN 505 remains comparatively limited.

The results of the meta-analysis were subsequently presented at the AIDS Vaccines 2013 conference in Barcelona and the reactions of several HIV vaccine researchers were covered in an article for Nature News by Sara Reardon. The article highlights a key human element of the debate around Ad vector safety that the Science perspective is not in a position to address explicitly: there are scientists who have made an extensive, good-faith investment in the development of Ad vectors as vaccines for infections that pose a major threat to health globally, and some of these individuals are understandably skeptical of the notion that their candidates might increase HIV risk.

One researcher who has conducted pioneering studies of alternative Ad serotypes as vectors, Dan Barouch, expresses doubt that the enhancement effect documented in Phambili was attributable to the vaccine. A commentary that accompanies the publication of the Phambili results in Lancet Infectious Diseases, authored by Merlin Robb and Nelson Michael from the US Military HIV Research Program (which is studying alternate Ad serotype vaccine vectors), calls the findings “mechanistically confusing” and argues that “one possible explanation for this curious clinical outcome is that it is the result of unknown biases introduced by unblinding.”

The Phambili investigators themselves, who have looked thoroughly for alternative explanations for their data, state: “the consistency between the STEP and Phambili studies point to a potential biological explanation for our findings.” To a large extent these may just reflect typical scientific differences of opinion, but the possibility of some intellectual conflict-of-interest among researchers developing Ad-based vaccines, whether conscious or unconscious, should be borne in mind. There was at least one mention of this concern at the summit: a participant pointed out that two members of the panel charged with discussing the implications of STEP and Phambili for other Ad vectors—Hildegund Ertl and Dan Barouch—lead efforts to develop candidates based on alternative serotypes. 

The continuing uncertainty around the safety of Ad vectors highlights the need to understand the potential mechanism by which HIV risk was increased in STEP and Phambili. One important difference between Ad vectors and those tested previously that was emphasized by immunologist Rafi Ahmed is that they are derived from a pathogen that is common in nature, which people are frequently exposed to (adenoviruses cause severe colds). Recipients of poxvirus vectors are fortunately not likely to encounter smallpox, whereas exposure to multiple Ads is normal. James Wilson, an expert in Ad biology, noted that Ad infections can also persist and that they may be considered part of a “virome” in the human body, akin to the way commensal bacteria constitute a microbiome.

The issue of natural exposure to Ads in the environment is central to one of the suggested explanations for why HIV acquisition risk was increased in the Ad5 trials (a possibility articulated in the discussion section of the paper describing the Phambili results): the vector induced Ad-specific CD4 T cell responses, which were drawn to mucosal sites of HIV exposure by natural Ad infection, because Ads persist at these locations in the body. Larry Corey, head of the HIV Vaccine Trials Network (which conducted both the STEP and Phambili trials), drew a parallel to HSV-2 infection: studies from Corey’s laboratory suggest that the consistent 2-3 fold elevation in HIV acquisition risk associated with HSV-2 is caused by the presence of HSV-2-specific CD4 T cells at mucosal sites (where they are engaged in immune surveillance of HSV-2). These HSV-2-specific CD4 T cells express CCR5 and represent targets for HIV.

Likewise, Ad-specific CD4 T cells engaged in immune surveillance of natural Ad infection at mucosal sites could also potentially influence susceptibility to HIV, and Ad vectors would be likely to increase the numbers of these cells. Furthermore, risk could be influenced by variables unmeasured in STEP and Phambili, including magnitude of mucosal Ad-specific CD4 T cells at baseline and over time, as well as the timing and intensity of exposures to natural Ads with the potential to modulate these CD4 T cell responses (the complex nature of these interacting factors could conceivably be a reason why the enhancement results appear “mechanistically confusing”).

This would be the worst-case scenario, because Ad-specific CD4 T cells are highly cross-reactive and respond to multiple Ad serotypes (including variants from chimpanzees). Vaccines based on non-Ad5 serotypes remain in development for multiple infections including HIV, TB, malaria and hepatitis C; if the proposed mechanism involving Ad-specific CD4 T cells accounted for the STEP and Phambili results, all of these vaccine candidates might have the potential to significantly increase susceptibility to HIV. Not only that, but the enhanced risk may be lasting, because Ad vectors induce long-lived memory Ad-specific CD4 T cell responses.

The Science perspective does cover this possibility; it includes a figure that cites vector-specific CD4 T cell responses and environmental reexposure to vector antigens (i.e. exposure to natural Ad infection) as factors that may enhance HIV acquisition risk. It also states: “Although there is no current evidence of increased risk of HIV acquisition for other Ad vectors, the presence of rAd5-activated T cells in tissue and the degree of shared epitopes between Ad serotypes that are recognized by CD4+ T cells should be considered as a possible area of concern.”

Unfortunately, as James Wilson described, it is challenging to try and model these phenomena in macaques because they are not susceptible to human adenoviruses. In his words: “Is there a good animal model to simulate a human Ad infection? The answer is no.” Although some macaque studies have used a version of human Ad5 that is described as species-adapted, Wilson bluntly stated: “that is also not a good animal model.”

Even with Ad infections that are naturally harbored by macaques, Wilson stressed that the immune response profile is very different to that seen in humans, with Ad-specific T cell responses higher in the gut-associated lymphoid tissue than the blood (the opposite pattern is seen in humans). Macaque Ad infections also appear more persistently active. Despite these differences, Ads that naturally infect macaques may still be a better model of human Ad infection than other available options.

For this reason, Mike Betts is assessing whether enhancement of SIV acquisition occurs in macaques after vaccination with a vector based on a macaque Ad, simian adenovirus-7 (SAdV7). Betts and colleagues previously presented data at the AIDS Vaccine 2012 conference showing that the vector increases Ad-specific CD4 T cell responses and provokes increased mucosal CD4 T cell activation (this study was eventually published online in the Journal of Virology). At the summit Betts was only able to show early, inconclusive results from the ongoing attempt to recapitulate the STEP results (so far, during a series of low-dose SIV challenges, 3/10 vaccinated macaques have become infected versus 1/10 controls).

The summit also included presentation of evidence that argues against the idea that Ad-specific CD4 T cells might be responsible (or solely responsible) for the enhanced HIV risk in STEP and Phambili. These presentations are mentioned briefly in the Science perspective, although there are some caveats associated with them that are not described.

Chris Miller discussed the macaque study his group published in the Journal of Virology in February 2012 (the paper is available open access). A species-adapted Ad5 was used to model prior exposure to the virus. Animals infected with this Ad5 were divided into two groups, which received immunizations with either the Merck Ad5 vector encoding SIVmac239 gag, pol and nef or the vector alone. For comparison, two groups of macaques not exposed to the species-adapted Ad5 received the same vaccination regimens, and a fifth group naïve to both Ad5 and vaccination served as controls. A week after the last immunization, all groups were challenged weekly with SIVmac251 via penile exposure, initially at a very low dose followed by higher doses (22 challenges were administered in total).

Two macaques in the Ad5-exposed, SIV vaccine group became infected at the lowest SIV challenge dose (during his presentation at the summit, Miller misspoke and said three animals) compared to none of the animals in any other group. In two of the reported statistical analyses this result represented a trend, and in one statistical model it emerged as statistically significant. There were no significant differences in SIV acquisition among the groups at higher doses. In fact four out of nine animals in the Ad5-exposed, SIV vaccine group remained uninfected after all 22 challenges, the highest proportion of any group, which seems to run somewhat counter to the evidence of enhancement at the low challenge dose.

Miller’s conclusion is that the results implicate the SIV-specific CD4 T cell response as responsible for increasing SIV infection risk, and he cited several other studies that have reported that immune responses to retroviruses can be associated with enhanced susceptibility. However, as Rafi Ahmed noted, the effect was only seen in macaques pre-exposed to the species-adapted Ad5, suggesting an interaction between vector-specific immunity and SIV-specific immunity. In the J. Virology paper, Miller and colleagues write: “further studies are necessary to confirm the enhancement of virus acquisition” but as yet no attempts to replicate the experiment have been reported. Several summit participants’ cited Miller’s study as supporting the view that other Ad vectors will not increase HIV risk, but, absent replication, it seems like a slender thread of evidence on which to hang that hope.

A novel attempt to specifically address the whether vaccines based on other Ad serotypes might influence HIV susceptibility was described by Angelia Eick-Cost from the US Military. The analysis took advantage of the fact that military recruits recently began receiving a live oral vaccine incorporating Ad4 and Ad7 after a period during which the vaccine was not recommended. HIV prevalence was compared between vaccine and non-vaccine time periods, and no significant difference was seen. Over one year of follow up, there were 7 HIV infections among 99,990 recruits given the vaccine, versus 10 infections in 99,984 recruits who did not received the vaccine. Glenda Gray noted several limitations to the data, including the apparently very low exposure to HIV in the population and the lack of information on routes of transmission. Analyses of STEP have suggested that the enhancement in risk associated with Ad5 vaccination was skewed toward the penetrative partner.

In summarizing the outcomes of the summit for their Science article, Anthony Fauci and colleagues cite several recommendations beyond the top-line caution against further trials of Ad5 vectors:

  • For vaccines designed to induce T cell responses, the potential risk associated with activating CD4 T cells (whether HIV-specific or vaccine vector-specific) needs to be considered and analyzed.
  • Another important consideration is “the potential for re-exposure to vector-related antigens in the environment with subsequent restimulation of the vector response. This is particularly important when evaluating viral vectors, including alternative Ad vectors.”
  • “Future clinical testing of Ad-based vaccines should evaluate the levels and distribution of both vector and insert responses in target tissues where HIV acquisition is known to occur.”
  • Further non-human primate studies could potentially help distinguish the contribution of vector-specific and HIV-specific CD4 T cell responses to enhancing risk of infection by comparing vectors containing viral antigens to empty vectors or vectors containing other antigens.
  • Non-human primate studies may also have utility for defining biomarkers associated with increased risk of HIV acquisition.
  • Mucosal immunity needs to be better understood, including influences of the mucosal microbiome and virome on vaccination. Because adenoviruses have been found to persist at mucosal sites, the authors note that “for Ad-based vectors, understanding components of risk related to the level of Ad exposure and persistence will be essential.”
  • Trials of vaccine vectors designed to induce T cell immunity against diseases other than HIV may need to consider the possibility of enhanced HIV acquisition risk, particularly if trials are performed in areas with a high HIV incidence (e.g. there are Ad-based vaccine candidates being evaluated for TB prevention). HIV incidence should be monitored in these trials, and for a period afterward.


Subsequent to the summit, in September 2015, the meta-analysis presented by Peter Gilbert was published in PLOS One. At around the same time, the journal also published a paper by Glenda Gray’s research group describing extended follow up of a subset of participants from the Phambili trial. The latter study confirmed the increased risk of HIV acquisition associated with receipt of the Ad5 vaccine in men, but presented evidence that while there was no efficacy in women, there was also no significant enhancement of the risk of HIV infection.

The discussion sections of these two papers offer insight into the continuing differences of opinions among researchers regarding the mechanisms by which Ad5 immunization may have increased HIV acquisition in male recipients. Gilbert and colleagues state:

“We hypothesize that the HIV antigen design and the HIV-specific immune response patterns are more likely to influence susceptibility to HIV than the transient and non-specific immune stimulation of a vaccine vector.”

While the paper from Gray’s group offers the countervailing view that “vaccination may have led to an increased number of activated Ad5-specific T cells that then served as target cells for HIV-1.” 

Trials of alternative serotype Ad vectors (such as Ad26) as HIV vaccine platforms are ongoing, making it important for researchers to try and resolve the persisting uncertainties about what happened in the STEP and Phambili trials.