AIDS Vaccines, From Monkeys to People

An Interview with John Shiver 

For many years, John Shiver, Merck's Executive Director of Viral Vaccine Research, has been a regular speaker at major conferences on HIV vaccines. With 11 clinical trials in progress and an extensive pre-clinical program for testing candidate vaccines in monkeys, Merck's program ranks among the major HIV vaccine development efforts.

Shiver, who earned a PhD in chemistry and then trained in immunology at the National Institutes of Health, has been at Merck since 1991. He now oversees not only the company's HIV Vaccine research activities but also its work on vaccines against other viral pathogens.

The previous IAVI Report (Feb-Apr 2003) reviewed Merck's clinical program on HIV-1 vaccines. Here, Shiver speaks with IAVI Report editor Patricia Kahn about the company's newly-launched international trial and on lessons learned along the way about using the rhesus macaque model for AIDS vaccine development.

Can you tell us about the international trial just launched with one of your vaccines?

The trial will enroll 435 volunteers at sites in the US, Puerto Rico, Brazil, Peru, Haiti, Thailand, South Africa and Malawi, which are part of the HIV Vaccine Trials Network. We’re testing a “proof-of-concept” vaccine containing the HIV-1 gag gene in an adenovirus vector. The vaccine is already being studied in several hundred North American volunteers, so we know it looks safe and can induce good immune responses. This trial will gather more safety and immunogenicity data in people of diverse genetic backgrounds. What new information do you expect to get by scaling up the numbers and diversity of volunteers?

The key question is whether people of diverse genetic backgrounds respond equally well to the vaccine—even though they may target different epitopes. We’ll also analyze how well the vaccine-induced responses in these populations recognize HIV-1 antigens based on different clades, as we’re doing in the North American volunteers. And we’ll continue exploring the effect of pre-existing immunity to the adenovirus vector. 

Your newer studies all involve adeno-based vaccines. Are these now your main focus? 

We already know a lot about our adeno-based candidates, and see many advantages: immunological, convenience of construction, and so on. But we’re taking a broad approach to viral vectors, exemplified by two studies we presented at the recent Keystone meeting. One is the testing of three different poxviruses in prime-boost combinations with adeno. The other is our work on different adenovirus serotypes, which we’ll also test in various prime-boost combinations. 

Our goal is to collect a set of vaccine tools and analyze each one alone and in combination. Right now, pre-clinical data suggest that prime-boost combinations of adenoviruses and poxviruses, or of different adenovirus serotypes, are our most promising vaccine tools for inducing cellular immunity. The next step—finding the best combinations—is unpredictable: you still have to rely basically on intuition and the empiricism of seeing what happens. 

So it will take a while to figure out what our best options are. But we’ll make these decisions based on a substantial body of clinical data.

You mentioned genetic diversity of human populations as one rationale for the international trial. At Keystone you presented some monkey data relevant to genetic differences and vaccine protection. Can you describe these findings? 

We’re trying to develop a broad perspective about our challenge studies, by looking at choices like which monkeys or challenge viruses to use, and then seeing what differences result from this choice. 

Our earlier studies on adeno-based vaccines used animals harboring the MamuA*01 MHC class I allele [a genetic marker associated with greater natural ability to control SIV and SHIV replication] and challenged with SHIV89.6P. These animals have controlled viremia to undetectable levels, with detectability being about 50 virus copies per ml of plasma, and stayed there so far for 2-3 years. When we induced stronger immune responses with a really good DNA prime plus adjuvant, and an adenovirus boost, it didn’t improve on this—we couldn’t crimp the primary viremia further, and chronic viral load can’t go below undetectable. All animals showed a strongly dominant response to the Gag CM9 epitope, although they also responded to other epitopes. 

The new experiments tested different adeno-based vaccines against the same challenge strain, in monkeys with genetic backgrounds that don’t give immunodominant responses to Gag [MamuA*01-negative]. We found that vaccination lowered viral set point 100-fold or more compared with unvaccinated animals. That’s a significant drop. But the load was still a few thousand copies, which is a lot more than 50. 

This tells us that strong immunodominance for one Gag epitope confers a big advantage against SHIV challenge. But even in animals without this advantage, two or three orders of magnitude is still a lot of control—actually, an outcome many people would probably be satisfied to see in a human study. Think about it: if we start with a setpoint of 30,000, which is what’s found in untreated people in the MACS cohort [a long-term, prospective US-based cohort of HIV-positive men] and drop it by two logs, we’re in the hundreds. That should mean a much better clinical prognosis. These results also tell us that SHIV is a more stringent challenge in MamuA*01-negative monkeys than in positive animals—another indication that genetic background is important. 

What else have you gleaned about genetic background and control of viremia?

We’ve seen that even without a vaccine, MamuA*01-negative animals tend to have a worse clinical outcome after challenge than the A*01-positives. They advance to AIDS much more rapidly. 

Another observation is that unvaccinated animals occasionally control virus spontaneously. I’ve seen this with SIV, and even more with SHIV—we see about 1 in 5 or 6 animals which do that, regardless of their MHC background. I’m not sure why. But it shows that there’s a lot going on in individuals that determines disease course, whether they’re monkeys or people. 

That’s why human studies need to look at a lot of diversity in volunteers, in terms of their genetic backgrounds, to really get a sense of how effective a vaccine will be. And it’s why international studies are so crucial—to start getting a handle on this. 

You’ve also looked at different challenge viruses. What have you learned?

At Keystone, we showed data on SIVmac239, which is more pathogenic than SHIV, in both MamuA*01-positive and -negative animals challenged intrarectally. The question was whether vaccines that controlled SHIV89.6P would give at least partial control against this challenge. 

Only one group did—the one given a DNA-prime, adeno boost, which also gave the best T-cell responses. The relative control of viremia isn’t necessarily due to these specific vaccines, but might happen with any vaccines that get T-cell responses to this higher level. In the MamuA*01-positive animals, DNA/adeno reduced SIV loads 10-30-fold below the control levels, at least out to 260 days. We still need to follow this longer—it’s a young study. 

That’s more than what many studies find with SIV, although still far from full protection. But is a challenge virus like SIVmac239 perhaps too stringent?

I don’t know. Well, yes, I know one thing. Both SHIV and SIV give extremely high viremia in macaques—between 100,000 and over one million viral copies of virus per cc of plasma within a few weeks of challenge. They maintain this level for months, until AIDS-associated pathogenesis sets in. Clearly, that’s much more viremia, and faster pathogenesis, than HIV-1 causes in people, where median viral setpoint is about 30,000. 

One consequence is that, so far, it takes two different types of vaccines to control SIV viremia in monkeys. But weare seeing successes in monkeys, so there’s a strong rationale for moving these approaches into people.

To me the value of these challenge systems—even if the disease they cause is exaggerated compared to humans—is that they help us discriminate among candidates. If no vaccines impacted viremia or disease at all, we couldn’t learn anything about what to test in humans. But we do have vaccines that control SHIV or SIV. So our foot is in the door. The task is to use animals for defining which combinations of vectors and antigens open the door further, and then to translate this experience into human trials. 

But I wouldn’t say that a particular vaccine approach is the one thing to move into people because it’s so great in monkeys. Instead, you use the animals to make a short list of what to take forward into people. 

Given all these variables, what challenge model will you use from now on?

I’ll use both SIV and SHIV, and probably stick to MamuA*01-negative animals for SHIV. I don’t think there’s any further room to mine for control of SHIV with these vaccines in MamuA*01-positives. 

You mentioned that your successfully vaccinated monkeys are still protected 2-3 years after challenge, with no sign of viral escape. That’s especially surprising since your vaccines carry only one HIV gene. Why does virus escape in some studies but not others? 

This is an important point. Our approach, which is based on our experience with anti-retroviral drugs, is that you need strong immune pressure against the virus. You need it as quickly possible, and it needs to be fairly diverse. Even though our adenovirus carries only gag, it’s very potent in monkeys—the magnitude of T-cell responses is strong prior to challenge, and it tends to recognize multiple Gag epitopes. 

I collaborated on the DNA vaccine studies presented at Keystone by Dan Barouch and Norman Letvin [reporting viral escape for several animals in two challenge studies ]. Those are early vaccines developed in my group, but that elicit relatively weak immune responses. And where responses are weak, they also tend to be narrow. Less immune pressure on the virus means greater potential for escape. 

That doesn’t mean escape can’t happen where responses are stronger and broader. But it stands to reason—and our data bear it out—that better containment of HIV is much more likely when there’s a stronger immune response from the start.

By now Merck has tested certain approaches in both monkeys and humans. How well did monkeys predict the human results? 

With adenovirus, many of our findings are very similar. You get the same magnitude of responses in people and in monkeys, the same proportions of CD8 and CD4 contributions, the same durability of responses and diversity of epitopes. In terms of dose-response, responses to adenoviruses in people exceed what we saw in monkeys, at least in people without pre-existing antibodies to Ad5. 

With DNA, immunogenicity decreased as we moved from mice into bigger rodents like rats or guinea pigs, and then into monkeys. It took more immunizations and more DNA, and the responses became lower. But we still came up with a dose and regimen in monkeys that consistently primed T-cell responses. Even with DNA in saline, nearly every monkey makes a sustained T-cell response, and the CRL1005 adjuvant increased responses by about 5-fold. But in people, that trend falls off: our best DNA approach gives responses in only 40% of the volunteers; they’re not very durable, and don’t synergize well with adenovirus vaccine. The synergy between vaccines was there, though, from the animal work.

These results also suggest another conclusion: cytokines and adjuvants generally don’t translate well from monkeys into humans. 

So would you say that non-human primate models can help identify the best approaches, but not the best dosage or immunization regimes? In general, yes, but may depend on the animal species and the specific vaccine. For example, we’ve done some testing of DNA vaccines in baboons, and found that their dose response, and the impact of adjuvants, were close to what we saw in people. 

So you could optimize your immunization regimens in baboons after macaque studies identify the best overall candidates? The thought frequently occurs to me that, yes, I’d like to see responses in baboons with a particular regimen. But these are large animals—30 or 40 kilos, not 3 or 4 kilos like macaques, so they’re much more difficult and expensive to work with. And our list of candidates is short enough to be manageable for clinical testing. Besides, you can only do so much, and clinical studies are our top priority right now. 

What are your criteria for deciding which approaches to move into Phase I testing? 

There are lots of things. We’re working to generate complete datasets, characterizing and comparing vaccines in terms of T-cell responses—overall magnitude, CD4 versus CD8 contributions, how many epitopes they recognize, and of course their ability to control viremia after challenge. When you do the science right, the answers are actually pretty clear as to which vaccines look better. 

Beyond how well the vaccine performs, a particular vaccine may be more attractive for development based on pragmatic things. Can we demonstrate preclinical safety and create the right type of information for FDA review? How well can we make the vaccine? 

In evaluating T-cell responses, most groups rely on assays that detect interferon-gamma. Are you satisfied that this detects the right cells?

I would not accept any single immunological parameter as the gatekeeper for clinical trials. We assess immune function based on many types of data. Besides interferon-gamma, we measure cytokines like IL-2 and TNF-alpha. We look at CD4 and CD8 contributions. We have a highly validated cytotoxicity assay, which correlates strongly with interferon-gamma responses for CD8 cells. Tetramer assays are great because they provide a reference point, an upper limit on the frequencies of antigen-specific cells. And we look at cellular phenotypes: Do the responding cells look more like activated or resting cells? Are they en route to a lymph node, or emerging from one?

It’s not clear what all these data mean. But they help you stay broad-minded about decisions, and to remember that you don’t really know what you’re doing. 

That said, when I look for a common thread in our challenge studies, I have to say that it’s the magnitude and diversity of CD8 T-cell responses.

We can’t define a cutoff where you know that above this level, every animal will control virus. The numbers are too limited. But at this point, it looks like the stronger the response and the more epitopes it targets, the more likely an animal will control virus, at least for some time. Maybe we can still improve our predictions by looking at subsets of these determinants. But I think we’re looking in the right direction. 

Why is it so hard to find definitive correlates of protection in monkeys? 

I think there are indications of correlates, in terms of trends. But it’s hard to nail down more precisely without lots of data, which in turn means lots of monkeys. And you’ll never have enough monkeys with the types of determinants we’re looking at. The correlate will probably be something like a probability of controlling viremia as a composite of immune response strength, diversity, CD4-CD8 contributions and ability to establish memory. Given the complexities of multi-component correlates, you won’t resolve this without a Phase III study in people. It’s not like hepatitis B vaccine, where if you get to a specific antibody titer, you’re considered protected. 

Is Merck working on neutralizing antibodies? 

We’ve had our share of negative data. But we kept an active effort in this area, and showed some results in posters at Keystone and in publications over the last year. These efforts are mostly directed at gp41 and gp120-CD4 complexes. So far we haven’t been able to define an interesting immunogen based on these proteins, but there are still some approaches to try.

How important are neutralizing antibodies for making a highly effective HIV vaccine? 

There’s no doubt in my mind that a good antibody-inducing component will ultimately yield a better vaccine. But I also think we can have a useful vaccine if a good antibody component isn’t found. 

By 'good antibodies,’ I mean antibodies that neutralize most wild-type primary isolates of HIV. Adding recombinant gp120 to a viral vector in one arm of a trial, or putting an envelope gene into a viral vector, is not a solution, because the responses they induce aren’t functional, in the sense that hey don’t meet the goal of what antibodies need to do.

Do you think sterilizing immunity is achievable? 

I think neutralizing antibodies might increase the probability of getting sterilizing immunity, of not having productive infections, and shifting average viremias lower. How much depends on how good the T-cell part of the vaccine is. 

Do you have a timetable for deciding which candidates to move into efficacy trials? 

Our goal now is to finish defining our basic vaccines, through the one we consider most promising—the trivalent vaccine with gag, pol and nef, which is now going into people. We’re also expanding into populations similar to those where efficacy studies will be done. You’ll see more and more sites opening up that are relevant to efficacy questions, whether it’s for studies with the HVTN or new sites we’re opening up with the trivalent vaccine. 

We’re planning very aggressively. I can’t give you a time frame because it’s not defined yet. But we’re planning for diverse efficacy studies with a clade B vaccine. While those efficacy studies are underway, we’ll continue developing our other potential vaccine components, so that we’ll have a short list of next best choices by the time the first efficacy data come in. 

Even a vaccine that works will probably have shortcomings. We’ll see what these are, and with an understanding of our next best tools, we’ll marry the two together and test efficacy again. That’s our basic strategy.