Skip to content

Harvard Vaccine Team Exposes Potential Dark Side of Reliance on Cellular Immune Protection

The Fate of Monkey #798

By Richard Jeffreys

January 2002 proved to be a month of mixed blessings for the AIDS vaccine field. A slew of new papers in the prestigious journal

Nature publicly highlighted both the promise and potential pitfalls of new immunization strategies, raising the volume of scientific debates that have been quietly preoccupying researchers for some time. At the center of it all were two back-to-back articles released on January 17: one from a team of Merck researchers led by John Shiver, publicly debuting encouraging data from studies comparing multiple HIV vaccine constructs (including Merck’s proprietary adenovirus-based vaccine vector) in rhesus macaques; the second from Dan Barouch and Norman Letvin’s group at Harvard, presenting a cautionary tale of viral escape from vaccine-induced T-cell responses in the same animal model system. TAG’s new Project Director for Basic Science, Richard Jeffreys, sifted through the data and commentary to prepare this report.

The debate sparked by these data sets revolves around the level of protection that might be afforded by the induction of T-cell immunity against HIV. The immunization strategies employed in both studies successfully induced virus-specific CD4+ helper and CD8+ CTL (cytotoxic T-lymphocytes) responses, but neither afforded full protection from infection. Instead, the success of the vaccines was measured by their ability to stimulate the immune system to control viral replication and thus preserve CD4+ T-cell counts and prevent clinical disease, at least in the short term.

This type of outcome contrasts with the Holy grail of vaccinology, “sterilizing immunity,” wherein infection is entirely prevented or rapidly cleared, leaving no detectable trace (except for, sometimes, long-lasting immunity).

The conventional wisdom is that sterilizing immunity can only be achieved with the aid of neutralizing antibodies, and HIV has thus far proven resolutely resistant to this type of immune response (although experiments using high levels of infused lab-created antibodies, “passive immunization,” have prevented infection in an SIV model). The pursuit of partial protection has thus been promoted as something of a stop-gap measure while researchers continue to try to solve the antibody challenge.

Public dissent regarding this two-tiered approach has been muted — until now. It is the Harvard data that has finally drawn several partial protection pessimists into the open because it raises a chilling possibility: that a vaccine which offers only partial protection could end up leading to a worse outcome than no vaccination at all.

In the study, Barouch and his Harvard team found that a single viral mutation led to viral rebound, CD4 cell decline, symptomatic disease and ultimately death in one of eight vaccinated macaques. Up until that point, the monkey in question had been clinically and immunologically healthy for six months after an intravenous challenge (with the pathogenic SIV/HIV hybrid SHIV89.6P, either six or twelve weeks after the final immunization; see footnote). The mutation was apparently selected for by the vaccine-induced virus-specific CTL response.

Interviewed in a Mark Schoofs Wall Street Journal piece, primate researcher David Watkins raises the specter of such escape mutations occurring in vaccinated humans and being transmitted onwards, potentially leading to the emergence of — yes, that media favorite — a “supervirus.” While this appears to echo some of the extremely speculative arguments against global implementation of HAART, a recent modeling experiment by Andrew Read and colleagues from Edinburgh actually offers some basis for Watkins’ concerns. Read modeled the potential effects of vaccines that ameliorate disease but do not prevent infection and found that under some circumstances they could potentially select for pathogens with increased virulence. Importantly, however, this result becomes less likely if the vaccine also reduces onward transmission of the infection. The potential for enhanced virulence would also be reduced if the vaccine were able to fully protect some proportion of immunized individuals.

The views of Watkins illustrate the theoretical basis for an increasing bifurcation of opinion among HIV vaccine researchers. On one side, there is a cadre displaying considerable enthusiasm and optimism about prospects for T-cell based vaccines, including Norm Letvin and the U.K.’s Andrew McMichael. On the other, an increasingly vocal group — including Watkins but perhaps most often associated with Harvard primatologist Ron Desrosiers — argues for caution, even going so far as to characterize the current mood of optimism over new vaccines as “irresponsible.” Somewhere in the middle, stoic realists such as antibody expert John Moore from Cornell acknowledge that T-cell-based vaccines are well worth testing, but expect that the addition of an effective antibody-based approach will be required to achieve truly protective immune responses.

While they have served to highlight these outstanding questions pertaining to T-cell-based HIV vaccines, neither the Merck nor Harvard paper claims to provide data that can resolve them. And there may be a danger of the data’s being over-interpreted. The initial goal of both groups was to consistently raise CTL responses, a not-insignificant challenge as is evidenced by the decade-long travails of the ALVAC canarypox vector (see “A Tale of Two Trials” below). Also, in keeping with the preliminary nature of these experiments, only a limited number of viral antigens were employed: env and gag in the Harvard study and gag alone in Merck’s.

The details of Barouch’s work provide additional reasons for caution. The data derives from a study that was widely publicized when first published in Science in the fall of 2000. A DNA vaccine construct encoding SIV gag and HIV env was administered four times to rhesus macaques. Four animals received the DNA vaccine alone, while two additional groups of four animals each received a low dose of an IL-2 fusion molecule (IL-2/Ig) in either protein or DNA plasmid form at the time of the first two immunizations. Six weeks after the final booster, all macaques were intravenously challenged with SHIV89.6P. All animals became infected, but at the time of the publication of the Science paper, recipients of the vaccine plus IL-2 had controlled viremia and preserved their CD4 counts over 140 days of follow-up.

By contrast, four of eight controls had died and only two displayed some degree of immunologic control of the challenge virus. But subsequent to this initial report, one animal that received the vaccine plus IL-2 in protein form — monkey #798 — began to lose control of viremia at around week 24 post-challenge. This was followed by a loss of CD4 T cells (week 36), symptomatic clinical disease (week 44), and death from simian AIDS (week 52).

It is the sobering tale of this macaque that forms the basis of the Harvard group’s Nature paper. In collaboration with Northwestern University virologist Steve Wolinsky, the researchers went over the data to look for explanations for the apparent vaccine failure. Genetic sequencing of the virus revealed that between weeks 14 and 20, immediately prior to the viral breakthrough, a mutation occurred in a region of the gag protein targeted by the vaccine-induced CTLs. The mutation involved a single amino acid change (from threonine to isoleucine) which was absent from 8/8 viral isolates sampled at week 14, but present in 10/10 isolates sampled at week 20. Upon further analysis, CTLs targeting the original epitope were found to be 1,000-fold less efficient at recognizing the mutant virus than the original strain. Barouch concluded that it was this single point mutation which ultimately triggered the cascade of events leading to the death of monkey 798.

The data raise the question of whether such escape tactics will prove to be the Achilles heel of all T-cell based vaccine strategies. If such vaccines cannot prevent infection, will eventual immune escape and disease progression be inevitable? Could such escape variants be transmitted, and thus further diminish vaccine efficacy at the population level? The Harvard team notes that the best strategy for preventing escape may be broadening the vaccine-induced immune response (e.g., by including antigens other than just gag and env) and attempting to drive viral replication to the lowest level possible post-challenge.

In an interview with National Public Radio after the study was announced, Norman Letvin noted that, prior to the emergence of the CTL escape mutant, monkey 798 appeared to have slightly higher levels of viral replication than the other immunized animals. He also reported that these remaining seven macaques have continued to control viremia for more than 600 days of follow-up. Taken together, these observations suggest that while it is probably premature to conclude that all CTL-based vaccines are doomed to failure, the unavoidable implication is that increasing CTL selection pressure by vaccination could have unpredictable effects on the evolution of HIV. Careful long-term monitoring and follow-up will be critical in both animal and human studies of these approaches.


A Tale of Two Trials: Vaccine Researcher Questions the Need for Two Massive ALVAC Studies
The week following the publication of the Merck studies, John Moore became the first researcher to speak up about another muffled controversy: the plans for two separate, massive phase III trials of Aventis Pasteur’s canarypox-based HIV vaccine (ALVAC). Moore’s commentary, also published in Nature, suggests that an excess of competitiveness between the two trial sponsors — the Department of Defense (DOD) and the National Institutes of Health (NIH) — is leading to duplicative trials that waste both human and financial resources.

The DOD’s trial is to be conducted in Thailand, and plans to enroll 15,800 heterosexuals (the DOD’s regulations don’t allow them to work with gay men or intravenous drug users) into a trial of an ALVAC vector (vCP1521) encoding gp120 from a subtype E isolate, along with gp41 and gag/pro from subtype B. All volunteers will also receive two boosts with Vaxgen’s bivalent subtype B/E recombinant gp120 vaccine. The estimated cost is $35-40 million and the current proposed start date is mid-2002. The NIH trial is being planned for the United States, the Caribbean and South America through the HIV Vaccine Trials Network (HVTN) and is slated to involve 11,080 volunteers comprising both gay men and high-risk heterosexuals. The protocol uses a slightly different ALVAC vector (vCP1452) encoding env, gag/pro and regions of pol and nef that are rich in CTL epitopes. One arm of the study includes a boost with Vaxgen’s clade B gp120 vaccine. Estimated cost for the HVTN study is $60-80 million. (Plans for both studies are discussed in detail in the new IAVI Report, online at

In addition to pointing out that the slender differences between these trials render them duplicative, John Moore also raises several scientific questions relating to the study designs. Chief among these is the notoriously poor immunogenicity of the ALVAC vector. In what must surely be a record for any experimental medical intervention, ALVAC’s tortuous history has included over 40 phase I and II trials involving around 1900 volunteers. The vaccine has shown an ability to induce low-level HIV-specific CTL responses in about a third of participants at best, and these responses are rarely directed at more than one epitope. Moore also questions the scientific rationale for including a gp120 boosting component. No animal model study has shown an advantage to this approach, and recently presented data from Harriet Robinson’s group at Emory reported that adding a gp120 booster to a DNA/MVA vaccine actually reduced efficacy rather than improving it. Moore points out that efficacy data from Vaxgen’s phase III trial of gp120 will be available later this year, which will surely shed more light on this question.

To avoid duplication and wasted resources (and the potential of two huge failed trials further denting public confidence in the vaccine effort), Moore suggests that the DOD ALVAC trial go forward (perhaps with a more rational boost, for example, gag instead of gp120), while the HVTN concentrate on newer and more promising third-generation vaccine constructs such as Merck’s adenovirus and the DNA/MVA prime-boost regimen being developed by the International AIDS Vaccine Initiative (IAVI). While a counter-argument is that these newer agents are not yet ready for phase III evaluation, it is entirely conceivable that they will reach this milestone within the next two years. And even if both proposed ALVAC trials begin as scheduled, enrollment is likely to be a lengthy process. Alternatively, Moore offers that the HVTN could assume responsibility for conducting a single ALVAC trial while DOD concentrates on bioterrorism defense. Although John Moore’s advice is clearly not intended to be proscriptive, the overarching and long-overdue message of his commentary is clear: it’s vital that the plans for both ALVAC trials are subjected to open public discussion before they are finalized.


Dept. of Defense NIH/HVTN
Region Thailand Americas
Size 15,800 11,080
Immunogen ALVAC1 ALVAC2
Start Date mid-2002 ???
Cost $35-40M $60-80M
1. canarypox 1521 (clade B) with Vaxgen gp120 (clade B/E) boost

2. canarypox 1452 with Vaxgen gp120 (clade B) boost in one subgroup


The Importance of Letting T-Cells Take a Rest
One interesting aspect of Merck’s vaccine study appears to have escaped comment. The data comprised two sets of studies, one set evaluating the vaccine vectors given singly while the second set combined the DNA vector with either Ad5 or MVA in a prime-boost regimen. In the former studies, macaques were challenged with SHIV89.6P twelve weeks after the last immunization, while in the prime-boost the challenge was administered just six weeks after the booster shot. The data clearly demonstrates that the prime-boost approach induced larger T-cell responses than either the Ad5 or MVA vector given alone, but what about the post-challenge outcomes? Looking at the graphs, it appears that control of viral load and preservation of CD4+ T-cell counts was more consistent in the animals that received Ad5 and MVA alone compared to those that received prime-boost. So what’s going on here?

The explanation may relate to a fundamental tenet of T-cell immunology. CTL maven Rafi Ahmed has long noted that vaccine-induced T-cell responses need to reach a “resting memory” state in order to respond optimally to a subsequent boost or challenge. The canonical T-cell response to a vaccine involves a peak of proliferation, followed by a “death phase” and ending with a stable but lower-level population of resting memory cells. It can take several weeks for this process to play out in mice, and how long it takes in higher primates is currently unclear. It is possible that for the macaques in the Merck study that received Ad5 or MVA alone, the additional six weeks of rest between the final immunization and challenge may explain the otherwise counterintuitive results. This will be a key question to explore in future animal studies, and, according to Emilio Emini, data is forthcoming that will address the question more directly.

Footnote: SHIV89.6P is a hybrid SIV/HIV construct which contains the envelope of SIV and the core of HIV. More specifically, it was constructed by combining the genes tat, vpu, rev and env from the Dutch HIV-1 isolate 89.6 with the remaining genome of SIVmac239. The gag proteins of challenge virus (SIVmac239) and vaccine are therefore precisely matched, or “homologous.” SHIV89.6P is noted for its ability to cause an unusually rapid and typically irreversible CD4 T-cell loss, accompanied by the swift onset of simian AIDS and death. While use of this virus allows for a rapid analysis of vaccine-mediated protection from clinical disease, many researchers raise the point that SHIV89.6P does not reproduce the more prolonged course of HIV infection observed in humans — and therefore may not be truly representative of the human in vivo situation. The Merck investigators themselves concede that “the relevance of the SHIV 89.6P monkey challenge . . . has not been firmly established.”


Back To Top