Do clades matter for HIV vaccines?

By Patricia Kahn, Ph.D.

As thousands of people prepare to gather in Nairobi and New York for September’s AIDS meetings, a key question for vaccine developers is how to contend with the huge diversity of HIV strains circulating worldwide

 HIV is famously the most genetically diverse viral pathogen known—nowhere more so than in Africa—as well as one of the most rapidly mutating. That, plus the uneven global distribution of its nine genetic subtypes, or clades, poses one of the biggest scientific unknowns facing AIDS vaccine developers: is a single, “universal” vaccine against all strains possible? Or will it be necessary to make a slew of different vaccine formulations, each tailored to the most common strains in a given region? Even worse, could it mean that new formulations might be needed regularly, as with flu vaccines?

The answer will be key to how quickly, and at what cost, an AIDS vaccine can be widely distributed around the globe once a successful candidate is identified. Manufacturing even a single formulation and getting it out quickly to adults and adolescents soon after it is licensed will be far more complicated and costly than anything the public health field has ever attempted. Doing it with many different formulations, or with repeated updates, would be even more challenging.

Nailing down the impact of HIV diversity on vaccine responses is difficult, for several reasons. One is that the current system for classifying HIV diversity is based on genetic sequence, not immune properties, and hasn’t been translated into distinct “immunotypes”—which is what really matters for vaccines. While that task is slowly being tackled for epitopes targeting cellular immunity, it may be impossible for neutralizing antibodies (NAbs), where clades don’t seem to correlate with immune recognition.

Another complication is that the clade issue has become highly politicized. Until recently, vaccine development has had a lopsided focus on clade B strains, which dominate the epidemic in industrialized countries but cause only about 12% of infections globally. That disconnect helped mobilize developing countries to get involved with HIV vaccine testing, and spurred development of non-clade B candidates—which now greatly outnumber new clade B-based ones. But it also helped create a political logjam: Fears that testing vaccines based on “unmatched” strains exploits trial volunteers in developing countries sometimes engendered resistance even to early-stage trials of non-local clades, and raised pressures to tailor vaccine candidates to ever-finer, single-country levels.

Yet only by comparing vaccine efficacy in matched settings to partially or completely unmatched ones can the impact of HIV diversity ultimately be resolved. Moreover, “a country-by-country approach to vaccine development would be crippling,” says Francine McCutchan (Henry M. Jackson Foundation, Rockville), who leads the US military’s HIV global surveillance program. “It would make it very, very slow to get vaccines suitable for some hard-hit regions, especially in Africa.”

But despite these challenges, there are promising developments, along with some sobering ones. A growing body of data shows that immune responses to T-cell-based HIV vaccines, and to natural infection, often recognize HIV proteins from different clades—fueling optimism that, at least for vaccines targeting cellular immunity, some cross-clade protection can be achieved. And studies in infected people are revealing that a few antibodies which neutralize primary HIV strains also work against other clades, findings that are renewing hopes of designing immunogens able to elicit NAbs, a task that has long seemed intractable.

Things are also moving on the political front. “I see a major shift,” says Jose Esparza, coordinator of the WHO-UNAIDS HIV Vaccine Initiative, pointing as an example to a consensus document on clades released by the African AIDS Vaccine Programme at its June meeting (see Vaccine Briefs, p. 20). “There’s now widespread recognition across Africa that the question of diversity needs to be rigorously tested through well-designed clinical trials—including some in unmatched settings.” While the clade issue is most relevant for efficacy trials, at this point there’s progress at the Phase I level: South Africa—with an overwhelmingly clade C epidemic—has just approved a Phase I trial of a vaccine based on clade A, shortly after giving the green light to its first HIV vaccine study, which uses a clade C-based candidate (see Vaccine Briefs, p. 20). (Africa’s first HIV vaccine trial, a Phase I study in Uganda in 1998-2000, also used a vaccine from an unmatched clade—in this case, clade B, which is all that was available at the time.)

But at the same time, the picture of global diversity—and the task of tracking it—are getting more complicated. That’s partly because new HIV variants and recombinants are continuously generated, and established ones move into new geographic areas. But it also reflects the growing use of more sophisticated technologies to characterize HIV strains, which are uncovering some unexpected layers of complexity.

Describing global diversity

At first glance, it might seem that classifying and tracking HIV variation around the world is a straightforward, albeit mammoth, undertaking. But that hasn’t turned out to be so.

The original definition of clades was based on short sequences, mostly within the HIV envelope, and has expanded to include nine clades (designated A through K, with no E or I). As the amount of sequence data grew over time—all captured in the NIH-supported database now run by Bette Korber at the Los Alamos National Laboratory (LANL) in New Mexico ( became clear that each HIV gene shows a different, and characteristic, degree of variation. Env heads the list, with up to 35% sequence diversity between clades, 20% within one clade and even 10% in a single infected person. At the other end of the spectrum, the Gag and Pol proteins show only 10-15% sequence divergence across clades.

But over the past few years, as more HIV labs have established high-throughput sequencing technologies, the field has accumulated hundreds of full-length HIV sequences from around the world—allowing large-scale comparison of whole genomes rather than just selected regions. The result: “a very rapid change in the epidemiological picture,” says McCutchan, whose updated map of global HIV diversity (included in this issue as a special poster and available at www.iavi.org/iavireport) illustrates the 10 major epidemiological patterns.

One big change is a growing awareness of the role recombinants play in the epidemic, especially in regions where several clades co-circulate. Some of them, like the B/F recombinant found in parts of South America, were previously thought to be pure subtypes. (A few were already known to be important circulating strains, especially the A/G recombinant called CRF02,_AG, for Circulating Recombinant Form, common in west and central Africa, and CRF01_AE, originally called clade E, in Southeast Asia.) Even more surprising, says McCutchan, is the high number of unique recombinants (those found so far only in a single person). For example, in Tanzania, where McCutchan and collaborator Michael Hoelscher (University of Munich and director of Tanzania’s Mbeya Medical Research Program, MMRP) have dozens of full-length sequences from low-risk populations, at least 40% are turning out to be unique recombinants (see figure 3). Similarly, full-length sequencing of HIV from low-risk adults in Uganda and Thailand, countries that each have two important co-circulating strains, found about 30% and 13% recombinants, respectively (see figs. 1 and 2).

Since recombination can only occur if two viral strains have co-infected a single cell, these findings hinted at yet another unexpected layer of diversity: that double infections may be far more common than previously recognized. Consistent with this notion, a prospective study of 600 female bar workers in southwestern Tanzania suggests so far that, astonishingly, a substantial proportion of these high-risk women show preliminary evidence of double infection with HIV from two different clades (see article, "Studying HIV Diversity in Multi-Clade Regions", p.17).

But characterizing these dual infections and pinpointing when each one occurred is proving to be technically daunting, even with the project’s large collection of blood samples from individuals at different time points. The difficulty is that the relative proportions of the two strains fluctuate widely over time, according to Hoelscher—one strain may remain barely detectable for months, then suddenly emerge to dominate the viral population in the blood. With superinfection now a hot issue in the HIV field (see IAVI Report, Jul-Sep 2002), resolving whether the infections happened more or less simultaneously, or if one occurred after the other was well-established (true superinfection), is a high priority for these researchers.

The issue has enormous implications for mapping diversity, in terms of identifying the proportions of different strains circulating in a population, and for genotyping HIV in samples used for studies of cross-clade immunity. And even though dual infections may prove less common outside highly-exposed groups like the Tanzanian cohort, HIV varies up to 10% even in singly-infected people. “I think we’ll find that sampling from a single time point, and sequencing only one full clone, has pretty poor power to reveal what’s going on,” says McCutchan. “We’ve uncovered a gigantic can of worms. It will take new technologies to sort this out.”

Yet there’s a bright spot in the data on HIV variation, she adds. Looking at the global numbers, it emerges that four clades (A through D) plus two CRFs (01 and 02, both of which are about 70% clade A) account for over 90% of all infections worldwide. From this perspective, diversity can be boiled down to 4 key clades, plus small contributions from the non-A segments of these two CRFs—a more manageable focus for vaccine developers.

But the missing link remains to connect this increasingly detailed understanding of genetic diversity with a picture of immune properties. The usual way of classifying viruses immunologically is based on serotype—a group of antigenically-related strains recognized by a specific reference antibody, usually a neutralizing antibody. By this criterion, “there’s absolutely no sense of serotypes for HIV,” says antibody expert David Montefiori (Duke University, Durham). “Although antibody binding tracks somewhat based on clade, we don’t know if neutralizing serotypes even exist and can be defined.” He also points out that it’s extremely difficult to pinpoint NAb epitopes, which are often based on 3D shape and map to discontinuous sequences.

On a more promising note, mapping T-cell epitopes (which are based strictly on sequence) and defining their patterns of cross-clade recognition is easier, and work in this area is ongoing

Vaccines and cross-clade responses

If there’s been one set of findings to cheer in the recent past, it’s the emerging clinical data suggesting that immune responses to T-cell-based vaccines frequently recognize at least some HIV strains of other clades—albeit possibly fewer epitopes and/or at a lower magnitude.

For example, trials of Merck’s adenovirus-based candidate, which carries a clade B-derived gag gene, found that 10 out of 13 volunteers who responded to vaccine also recognized peptides from clades A and C Gag. Consistent with these findings, an international study of people infected with HIV of different clades detected similar frequencies of cross-clade T-cell responses to Gag and to the relatively conserved proteins Pol and Nef, but far fewer responders to the more diverse Rev and Tat proteins. Cross-clade CTL responses are also seen with canarypox-based vaccines, as first shown in studies of Ugandan volunteers given clade B-based vaccines and who often responded to some A and D peptides (J Infect Dis 182:1350;2000). Overall, says Larry Corey, who directs the NIH-sponsored HIV Vaccine Trials Network, “clades aren’t making the kind of difference people thought they would.” He also predicts that “cross-clade responses will be more the norm than the exception.”

But the case is far from proven. It’s unknown whether cross-reactivity will translate into cross-protection, says Jaap Goudsmit, chief scientific officer at Crucell, a Dutch biotechnology company; without more detailed data on which epitopes are and aren’t recognized across clades, and which ones matter for protection, he’s reserving judgment. Lower levels of responses across clade are another potential factor—although new data showing that a candidate Ebola vaccine protects monkeys even after one dose, which induces a weaker response than a full prime-boost regimen (Nature 424:681;2003), indicate that less-than-optimal responses can suffice for a successful vaccine. And 90% sequence conservation between, say, the vaccine and an infecting Gag protein still corresponds to an average of one amino acid difference per epitope, write Tomas Hanke and Andrew McMichael—which, despite some “wobble” (tolerance for mismatch), could abrogate a significant number of responses across clades (Vaccine 20:1918;2002). Some loss was seen in the Ugandan canarypox vaccine study, where cross-clade responses were sometimes as strong as those to vaccine strain (as measured by peptide titrations in Elispot assays), and sometimes weaker.

In any case, studies of cross-clade recognition—including some with new non-clade B candidates—are yielding a growing body of valuable information. And as Peggy Johnston (Assistant Director for AIDS Vaccines, National Institutes of Allergy and Infectious Diseases) points out, “this will allow the field to make decisions based on data, not conjecture,” she says—decisions such as what vaccine strains are most promising for testing in regions with particular non-matching clades in circulation.

Merck’s John Shiver, who heads the company’s vaccine research program (see article), points to another potentially important (and under-recognized) component of cross-clade protection: immune responses to HIV antigens not present in the vaccine. This notion emerged from Merck’s studies of its gag-only candidates in monkeys, which show “terrific” responses to Nef after challenge with either SIV or SHIV—responses rarely seen post-challenge in unvaccinated animals. “This suggests that vaccines may only need to generate enough of an immune response so that after infection, you can make lots of natural responses to the infecting virus, whatever clade it is,” he says. “The vaccine may not have to generate all the heterologous [cross-clade] coverage.”

For vaccines that target the antibody-producing B-cells, the picture remains much bleaker. So far, no vaccine tested in humans has generated neutralizing antibodies (NAb) to anything beyond the vaccine strain and a few closely related isolates, and there are few ongoing clinical trials involving candidates that even target this arm of the immune system.

Yet data on HIV-infected people show that cross-neutralizing antibodies do exist—findings that have re-kindled efforts to find strategies for generating them via vaccination.

Designing Vaccines for Breadth

All this leaves vaccine developers still operating largely in the dark in terms of how to design for maximal breadth.

For T-cell-based vaccines, a common starting point is to use the most conserved regions of HIV—first and foremost the gag gene (or protein), followed by pol and sometimes nef. These are usually derived from a primary HIV isolate, sometimes selected for a particular biological property such as use of the CCR5 receptor, and/or origin in the geographic region where the vaccine will be used. In a variation on this theme, the San Diego-based company Epimmune developed a candidate containing highly conserved epitopes (rather than whole genes) from across the genome, selecting further for those recognized by the most common HLA genotypes. This candidate recently entered Phase I clinical trials in the US and Botswana.

More recently, researchers have begun looking at artificial sequences derived by computer analysis, rather than actual circulating viruses, as sources of vaccine strains. These are often consensus sequences, made by analyzing a set of sequences (say, primary isolates of clade C) and choosing the nucleotide found most commonly at each position. Or they may be ancestral sequences, which represent the most likely common ancestor to a group of isolates. The rationale for these approaches—articulated in depth by Bette Korber (Science 296: 2354; 2003)—is that they minimize the genetic distance between the vaccine and the pool of circulating strains; in contrast, a primary isolate might be an “outlier,” genetically speaking, relative to many other isolates of the same subtype.

These sequences have not yet been incorporated into vaccines, although similarity to a clade consensus sequence is sometimes used to help select primary isolates for vaccine strains. But several groups are developing Env immunogens from consensus or ancestral sequences, and two teams reported at the 2003 Retro-virus conference that they seem to fold and function like real Env proteins, and to show potential as cross-clade immunogens. Nancy Haigwood and Jim Mullins (University of Washington, Seattle) made two 'proof- of-concept’ ancestral clade B Env proteins, which so far (as a DNA vaccine in rabbits) induce “reproducible but fairly low titers" of NAbs that cross-neutralize a clade C isolate, says Haigwood (poster 409). And researchers from Duke University, LANL and the University of Alabama reported on an immunogen made from a consensus Env sequence of all major clades, and which was recognized by several clade B and C antisera (from infected patients)—unlike Env from either clade, which reacted best with same-clade sera (poster 410). Both posters are available at www.retroconference.org/2003.

Other strategies for generating broad NAbs are based on modifying the shape of the Env antigen, rather than focusing on sequence (IAVI Report, Dec 2002-Jan 2003, p.1). These include development of native, trimeric structures; modification of Env to remove the more variable regions (exemplified by Chiron’s gp140 immunogen, which should enter clinical trials this year); generating immunogens that bind to well-characterized, neutralizing broad Nabs, or which expose normally-hidden neutralizing epitopes (fusion intermediates).

And for both B- and T-cell-based vaccines, several groups are combining immunogens from different clades to create “cocktail” vaccines (for example, the A/B/C candidate described in the article on p.7). In the future, cocktails could also contain mixtures of primary isolates (or shape-based immunogens) plus consensus sequences.

Vaccines and cross-clade responses

Last but far from least of the hurdles in developing a broad HIV vaccine will be the challenge of determining just how broadly protective a vaccine actually is. Intended to help mobilize support for the steps this will require, the new African AIDS Vaccine Programme document emphasizes that clade-mismatched trials are a crucial part of the solution, and that politics must recognize this reality.

The document also proposes some guidelines for decision-makers. For Phase I/II testing, it advocates moving ahead with good candidates, regardless of the subtypes involved, for the sake of other benefits—such as building capacity for running trials, and for conducting scientific and ethics reviews; establishing dialogs among scientists, policy makers and communities; and gathering data on a vaccine’s ability to induce cross-clade responses in diverse populations. It also recommends that decisions about efficacy trials should be based on evidence of cross-reactivity between a candidate vaccine and the unmatched clades and/or CRFs circulating in the trial population, along with a good safety record from Phase I/II studies.

From a scientific perspective, this testing is likely to involve several different trial scenarios, says veteran vaccine developer Don Burke, who directs the Center for Immunization Research at Johns Hopkins University. One is to compare the efficacy of vaccines based on a single clade in matched versus unmatched settings—the strategy Merck is likely to pursue with its clade B-based candidates. Another is to ask whether clade-matched vaccines work better than unmatched ones—studies that could be done with two different vaccines in one setting, or by testing one vaccine in a region with multiple circulating clades, powering the trial to detect efficacy in at least one of them. In all cases, getting an answer will require careful analysis of breakthrough infections. Trial design, along with logistics, gets more complicated for multi-clade candidates and/or sites with multiple clades in circulation (See articles), since both these variables raise the number of volunteers needed to identify statistically significant trends in vaccine cross-protection. But with the politics gradually aligning more closely with the science, and a growing roster of potential trial sites in the picture, there should at least be a few less obstacles in the way.