Barcelona introduced a new word—and a new worry—into the common AIDS parlance: superinfection. In one of the conference’s most widely-discussed presentations, Bruce Walker (Massachusetts General Hospital, Cambridge) described the case of an HIV-positive man who became infected with a second, closely related strain—despite having strong cellular immune responses that were controlling the first virus without drugs. The finding set off alarms, in the press and the corridors, that it could portend dim prospects for developing an HIV vaccine (at least one based on cellular immunity) that protects against even minimally divergent strains.

Walker’s wasn’t the only report of superinfection, although the others went relatively unnoticed. In one [Link], Stephanie Jost from the University of Geneva described a man who became superinfected with a clade B virus about two years after initial infection with an A/E strain. (This finding, first reported at the Retrovirus Conference in Feb. 2002, was since published in the New Eng. J. Med. 347:731; 2002.) And a Thai-CDC collaboration found two potential cases of superinfection in IDUs, each involving a second subtype ([Link]; J. Virol. 76:7444; 2002 [PubMed]), although the methodology used could not completely rule out simultaneous transmission (or two separate transmissions occurring close in time). But none of these three patients was found to have immune responses that recognized the new strain at the time of superinfection—in contrast to Walker’s patient, where cross-recognition was clearly present.

From a public health perspective, the take-home message is straightforward: Superinfection can occur, and can lead to worsening of disease—making safe sex precautions essential for HIV-positive individuals.

But for vaccines, the lessons are more ambiguous. Alongside the question of failed immune protection, it’s unclear whether this reflects what would happen with a vaccine. Nor is it known whether superinfection (especially with a virus of the same clade) is a rare event, or far more common than people have realized. While the first question is largely unanswerable in the short-term—and Walker spelled out some caveats about extrapolating to vaccines—data on the frequency of superinfection should emerge from an ongoing study in Tanzania [Link].

Superinfection with a same-clade virus

According to Walker’s talk and information from Marcus Altfeld, who carried out this work with Todd Allen in Walker’s lab, the superinfection case involved a male patient who was participating in an ongoing trial of structured treatment interruption. He had begun HAART shortly after becoming infected and interrupted for the first time about 18 months later; virus rebounded immediately, and he resumed treatment. During a second interruption he controlled virus much better, maintaining low loads (~1,000) for about 6 months and CD4 counts in the 700-900 range.

But then virus suddenly spiked again, and the patient developed symptoms typical of acute HIV infection, including fever, lymphoadenopathy and fatigue. After another few months of treatment, a third interruption led to immediate viral rebound.

Then came an unexpected finding. Analysis of virus from these two spikes revealed what looked like a different strain than the original one—also clade B, but diverging by about 12% at the protein sequence level. Even with sensitive (PCR-based) techniques, the new strain couldn’t be detected in blood samples taken before the spike that occurred 6 months into the second interruption. Consistent with the patient’s report of an un-protected sexual exposure around that time, the researchers concluded that this represents a superinfection.

The immunological findings were also striking: At the time of superinfection, the patient had strong, broad T-cell responses to HIV—over 25,000 spot-forming cells per million PBMC, measured by Elispot analysis for interferon-gamma-producing cells. What’s more, these were directed against more than two dozen different HIV epitopes spread across the viral genome. After superinfection, about half of these responses persisted, and they turned out to recognize epitopes present in both viruses; there were even three new CD8 responses detected, all specific to the second virus.

Yet this wasn’t enough to prevent the second infection. "I anticipated, as did others," says Walker, "that this level of immunity would be cross-protective."

Unraveling protection and CD8 responses

The finding that it wasn’t resonated strongly with vaccine developers, who routinely evaluate CTL-based vaccines by the same T-cell responses that Walker’s group measured. "These data challenge our notions that magnitude and breadth translate into protection," says Kent Weinhold of Duke University (Durham), who heads the central immunology laboratory of the HIV Vaccine Trials Network. "They tell us that we need to look more deeply at parameters of cellular immunity. This really opens our eyes that protection is much more complicated."

That view is also beginning to emerge from studies on what constitutes an effective (viremia-suppressing) response in infected people. For example, Mark Connors of NIH has compared CD8 responses in a group of long-term non-progressors (LTNP) who meet stringent criteria (infected for 13 years or more; viral loads below 50; normal CD4 counts) to those in progressors, all without HAART. His findings: even patients with high viremia and progressive disease have high numbers of HIV-specific CD8 cells, while LTNP have similar levels but narrower, more focused responses (PNAS 97:2709; 2000 [PubMed]). Other researchers, including Mike Betts and Louis Picker (J. Virol. 75:11983; 2001 [PubMed]), Marylyn Addo in Walker’s group and Merck’s John Shiver also have recent data showing a lack of correlation in infected patients between levels of HIV-specific CD8 cells (measured by interferon-gamma production) and the ability to control viremia.

So it’s back to the "holy grail" question: what defines a protective response—or, as Walker rework-ed the question in his talk, what correlates with loss of viral control? Part of the answer may involve having responses to the "right" epitopes, which several labs are working to identify. Connors is looking for qualitative or functional differences between protective and ineffective responses, focusing on differences in the ability of CD8 cells from progressors versus LTNP to divide (Nature Immunol., in press). Others are looking at markers that might further define protective cells among interferon-gamma-producing CD8 populations [Link], and at CD4 help. "The fact that most of the vaccines being tested rely on CD8 responses make this qualitative difference very important to understand," says Connors.

Do these findings translate to vaccines?

All this still leaves the underlying questions of why Walker’s patient could not stave off a second infection, even across a relatively small strain difference, and whether this predicts that the same might happen with vaccines. Walker raised several caveats, and other researchers offered some of their own. One is that "we are dealing with an HIV-infected person who most likely has immune deficiencies," says Altfeld, and even subtle damage could impair his ability to make fully functional responses. Looking specifically at T-cell help, there were signs of a downward trend in the patient’s CD4 levels (from 1,000-1,300 during treatment to 700-900 before superinfection), although Altfeld said this did not reach statistical significance.

Another key point is that immunity established by vaccination before HIV exposure is very different from that induced by active infection—which could mean differences in qualitative properties and/or responses to specific epitopes. "If the same degree of broad immunity were induced by a vaccine in an uninfected person, it might be much more effective," said Walker.

Primatologist Mark Lewis (Southern Research Institute, Frederick) raises a very different point, asking whether protection has truly failed. "We don’t expect CTL-based vaccines to block infection," he says, and therefore the superinfection "wasn’t at all surprising. But we don’t know what the outcome of this patient will be. Is his immune response going to change his ultimate disease course?" [As of mid-September 2002, Altfeld said the patient is off treatment after a fourth round and controlling viral load—but with an elevated setpoint of 10,000-40,000.] Going a step further, "Isn’t this what we’d expect without neutralizing antibodies in the mix?" asked another researcher.

How common is superinfection?

Another crucial question, from both vaccine and public health perspectives, is how often superinfection occurs. "A single case tells us that it’s possible to become infected despite broad CTL responses," says Walker. "It doesn’t tell us how likely this is." To date there have been only sporadic reports of double infections, although the huge number of HIV recombinant strains—both those in circulation and the much larger group of strains unique to one or a few people—suggest that this may not be rare at all. But there has been no systematic surveillance up to now, and no clarity on whether these double infections reflect transmission of two strains at one time (or shortly after one another) or true superinfections, with a second infection occurring after the first one is well-established.

That’s exactly what Francine McCutchan (Henry M. Jackson Foundation, Rockville), Michael Hoelscher (University of Munich) and collaborators are now studying in Tanzania, a country with three major clades in circulation. They first carried out a pilot study involving 100 infected, high-risk women ("bar girls"), says McCutchan, and found that an astonishing 45% harbored unique inter-clade recombinants—each of which must have arisen from a double infection, either in the women themselves or in a proximal sexual contact.

To pin down the frequency of inter-clade double- and superinfections, they also established a cohort of 600 bar girls in the context of the HSIS Bar Workers Health Project, (presented by Oliver Hoffman [Link]. Recruited without regard to HIV status, initial HIV prevalence was 68%, with an incidence after one year of 14.1 infections/100 person-years. Now two years into the three-year study, the women continue to be monitored every three months for HIV infection, and blood samples collected for later analysis. Key questions: are double infections found, and if so, when do second infections tend to occur relative to first ones? And can people who become doubly infected be distinguished immunologically from those who don’t?

Looking ahead

Studies like these should help resolve whether superinfection represents a common event or a fascinating but rare curiosity—although they don’t yet tackle the question of superinfection within clades, which Altfeld hopes researchers will start looking for now in the wake of the attention this issue has received. But vaccine developers don’t all seem as fazed as initial reports implied. "As nicely worked up as this is, it’s a case report in one individual," says primate researcher Jeff Lifson (National Cancer Institute, Frederick). "I wouldn’t discount something as important as preventive vaccines because of one patient. It’s certainly sobering, but by itself not enough to make me get my nihilist hat back on."

------------
^*Patricia Kahn is editor of the IAVI Report