Every year, Keystone Symposia sponsor two parallel conferences on HIV pathogenesis and vaccines. The meetings take place at the Keystone Resort high in the Rocky Mountains, allowing researchers to mix the latest data with a daytime trip to the ski slopes. This year's event offered no earth-shattering new insights, but provided some glimpses of future directions in HIV research.

Mike McCune from the Gladstone Institute in San Francisco reviewed data that challenges traditional definitions of "treatment failure" (McCune 2002). In collaboration with Steve Deeks from San Francisco General Hospital, McCune has shown that many individuals who appear to be failing protease inhibitor (PI)-based antiretroviral drug combinations—based on detectable viral load and the presence of drug-resistance mutations—continue to do well both clinically and immunologically. Deeks and the Gladstone group began with a series of hypotheses that might explain this phenomenon:

• Protease inhibitor-resistant virus may have reduced replicative capacity compared to wild-type virus.
• Reduced replicative capacity may be associated with reduced T cell activation and turnover.
• Reduced replicative capacity may spare T cell production by the thymus.
• Reduced replicative capacity may facilitate the generation of stronger HIV-specific CD4 and CD8 T cell responses.

The first hypothesis, suggested initially by test-tube data, was confirmed by Steve Deeks in a treatment interruption trial (Deeks 2001). In this study, individuals with resistant virus who were stable on a PI-containing combination stopped treatment for 12 weeks. Over the first few weeks after interruption resistant virus remained detectable and CD4 T cell counts stayed stable, but then wild-type (non-resistant) HIV rapidly emerged and entirely replaced the resistant virus population. This emergence of wild-type virus was associated with a significant increase in viral load and a decline in CD4 T cell counts.

Moving onto the second hypothesis, McCune measured markers of T cell activation and turnover in three groups of individuals (see also Deeks 2002): 17 untreated, 36 with virological failure (defined as a viral load >2,500 copies for >24 weeks despite being on PI-containing treatment for at least 18 months) and 18 with virological success (defined as a viral load of <50 copies for >24 weeks after receiving PI-containing treatment for at least 18 months). Comparing expression of the activation markers CD38 and HLA-DR on CD4 T cells between groups, McCune reported that they were significantly reduced (with a p value of 0.002) in the virological failure group compared to the untreated controls. Activation markers were also significantly lower in the virological success group compared to the virological failure group. A similar picture emerged when the same activation markers on CD8 T cells were measured. The proliferation marker Ki67 exhibited a somewhat different pattern: in CD4 T cells, expression was significantly higher in untreated individuals compared to those with virological failure or success, but there was no significant difference between the latter two groups.

McCune also measured T cell turnover in 17 individuals with virological failure using the deuterated glucose technique pioneered by Marc Hellerstein from the University of California at Berkeley. In this case, data were compared to historical controls (both untreated and those with virological success) from a previous study using the same technique. CD4 T cell turnover was significantly reduced in the virological failure cohort compared to untreated individuals, but, as seen with Ki67, the difference between the virological failure and success groups was not significant. The median CD4+ T cell half-lives for the three groups were 22 days (untreated), 68 days (virologic failure) and 82 days (virologic success). McCune emphasized that in untreated HIV infection, viral load correlates with the degree of T cell activation and turnover, whereas in individuals with PI-resistant virus this correlation is lost. He speculated that this situation may be analogous to that of sooty mangabey monkeys, which sustain high SIV viral loads but manifest little evidence of T cell activation and no disease progression (Kaur, 2002).

Addressing the third hypothesis, McCune noted that direct viral effects on thymic production are tough to prove in people. The thymus is not easily sampled, and direct measures of T cell output from this organ are still lacking. The strongest support for the idea that PI-resistant virus may be easier on the thymus comes from a study using a culture system (Stoddart 2001). Both PI-resistant and wild-type HIV isolates were taken from participants in Steve Deeks's treatment interruption study (described above). The ability of these isolates to replicate in thymic organ culture was then compared and, as predicted, PI-resistant viruses reproduced poorly compared to their wild-type counterparts.

McCune was able to produce a larger body of data in support of his fourth hypothesis. The Gladstone team employed intracellular cytokine staining to measure the magnitude of the HIV-specific CD4 and CD8 T cell response (this technique captures HIV-specific T cells based on their ability to produce the cytokine interferon-gamma when stimulated with viral antigens). This analysis included 28 untreated individuals, 66 experiencing virological failure on PI-based treatment and 87 categorized as a virological success on PI-based treatment. McCune reported that the total gag-specific CD4 T cell response was significantly higher in the virological failure group than either the untreated (p value of 0.007 for the comparison) or virological success (p=0.001) groups. There was a significant correlation between gag-specific CD4 and CD8 T cell responses in all groups. McCune highlighted the fact that the magnitude of the HIV-specific CD8 T cell response correlated with maintenance of a stable viral load over four months of follow-up. Looking at the magnitude of the virus-specific T cell response and viral load from a single timepoint, McCune found that the data fitted a "bell-shaped curve": individuals with viral loads less than 1,500 copies showed a positive correlation between the size of the gag-specific T cell response and viral load, whereas those above the 1,500 copy cut-off exhibited a negative correlation. McCune's results echo other recent studies suggesting that there may be qualitative differences in virus-specific T cell responses that are not a captured by measuring interferon-gamma alone.

McCune concluded his presentation with the provocative "Goldilocks Hypothesis"—that in some cases, the persistent presence of PI-resistant virus may be beneficial in terms of preventing immune activation, sparing thymic output and fostering functional anti-HIV CD4 and CD8 responses. This appears to account for the "disconnect" seen in the Deeks cohort and other virologic "failures" who maintain or increase CD4 cell counts and do not progress clinically despite continually measurable viral load. Understanding the underlying cause of this change in viral pathogenicity could potentially shed light on the long-standing mysteries of the benign SIV infections seen in sooty mangabey and African green monkeys and thus open up new avenues of HIV research.

Squeezing Into Phe43
Richard Wyatt from the National Institutes of Health presented an update on the search for potential neutralizing antibody targets on HIV's notoriously evasive envelope glycoproteins. Probing the structure of gp120 (in collaboration with many other investigators, including Joe Sodroski from Harvard), Wyatt has homed in on a cavity known as Phe43, in the portion of the protein that binds to CD4 (the cavity is named after the amino acid from CD4 that fits into it: phenylalanine 43). Phe43 is relatively well conserved in diverse HIV and even SIV isolates, making it an attractive target for both drug discovery and antibody design. The problem, however, is that unless the virus is in the process of binding to CD4 prior to entry into a cell, this region of gp120 is relatively inaccessible. In Wyatt's vivid description, the V1 and V2 loops of gp120 act as "a pair of calipers" that close off access to Phe43.

Wyatt believes it may be possible to create a gp120 protein that is locked into the CD4 binding conformation, in other words a gp120 where the V1 and V2 calipers are open and the Phe43 cavity is visible to the immune system. Using such a protein as an immunogen might induce antibodies capable of blocking the gp120/CD4 binding process, although the brevity of this interaction might make it hard for such an antibody to be in the right place at the right time. Wyatt feels it's a hypothesis worth testing, and is busily creating mutant viruses in the hope of finding one that produces gp120 in, or at least close to, the desired confirmation. In terms of drug discovery, Wyatt believes that Bristol-Myers Squibb's new experimental entry inhibitor may interact in some way with Phe43 and he has been attempting to access the compound in order to further his studies, but he ruefully reported that so far the effort has stalled due to intellectual property issues.

Tales of Memory T cells
The opening lecture of the Keystone meeting was given by cellular immunologist Rafi Ahmed from the Emory University Vaccine Center, whose work in murine models has provided critical insights into how T cells respond to chronic viral infections and vaccination (Ahmed 2002). Ahmed's talk focused on the generation, maintenance and function of "memory" T cells, which are known to be key components of the protective immune response. Ahmed was among the first to show that memory T cells can persist in the body in the absence of their specific antigen. Based on these seminal results, Ahmed's group has since conducted an intensive search for the factors that govern memory CD8 T cell survival. At Keystone he reported, with some relief, that they had finally found one—the cytokine interleukin 15 (IL-15). Ahmed found that mice lacking genes for IL-15 (so-called IL-15 "knockout" mice) experience a gradual attrition of their memory CD8 T cell pools. In normal mice, memory CD8 T cells undergo a process known as "homeostatic proliferation" whereby each memory T cell creates a fresh copy of itself every 24-48 hours (also known as "self-renewal"). In IL-15 knockout mice, memory CD8 T cells can not proliferate in this way, leading to a reduction in numbers over time. Ahmed's results hint that IL-15 might be used therapeutically in order to boost memory CD8 T cell responses. Currently, the factors governing the survival of memory CD4 T cells remain relatively obscure since the actions of IL-15 appear specific to memory CD8 T cells.

Ahmed also provided an early look at data obtained using microarray technology to examine gene expression in CD8 T cells. Microarrays can provide a readout of the genetic activity occurring within a cell by comparing DNA extracted from the cell with DNA containing vast numbers of known genes. Fluorescent tags illuminate the genes that are active; the brighter the signal, the more active the gene. Ahmed's group was interested in the changes in gene expression that accompany the transition from naive T cell to memory T cell, a transition that occurs in response to a new vaccination or infection. It has been appreciated for some time that memory T cells possess enhanced functional capabilities that play a key role in vaccine-induced protection from infection or disease, but exactly how these skills develop is still largely unknown. Ahmed looked at 8,700 genes and discovered a complex suite of alterations that occurred during the naïve to memory transition. Notably, the gene expression profile of memory T cells was not set into a stable pattern until several weeks after immunization. This time lag in the development (or differentiation) of fully-fledged memory T cells may be important for scheduling vaccinations that aim to prime and then boost a memory T cell response.

To address this issue further, Ahmed focused on one apparent marker of memory T cell differentiation. The gene expression data suggested that CD8 T cells lose, but then reacquire, a surface marker known as L-selectin or CD62L as they transition from naïve to memory. Ahmed showed that the re-expression of CD62L on memory CD8 T cells correlated with the development of a robust capacity to proliferate in the face of viral challenge, and that this capacity was associated with enhanced protection against the murine virus LCMV. Linking these data to prime-boost protocols, Ahmed went on to demonstrate that the dose of antigen influenced the length of time it took for memory CD8 T cells to reacquire CD62L: the higher the dose, the longer it took for the CD8+CD62L+ phenotype (christened "central memory" T cells) to develop. Ahmed suggested that a strong priming immunization would likely mimic this effect, making it prudent to wait longer prior to attempting a booster vaccine. Conversely, a weak prime would lead to a quicker conversion from naïve to central memory T cells, suggesting that a booster could be administered within a shorter time period. These observations may have additional implications for the timing of viral challenges in animal model experiments: for vaccines based on T cell immune responses, a longer wait between boost and challenge might improve rather than diminish protection. Similarly, in light of Ahmed's data, studies of therapeutic immunization followed by antiretroviral treatment interruption might need to carefully evaluate the optimal interval between immunization and interruption. The same considerations could come into play in trials of repeated structured treatment interruptions where the aim is an "autoimmunization" effect: if an interruption succeeded in inducing new memory T cell responses, they may need to time to mature into a central memory phenotype in order to respond optimally during a subsequent treatment interruption.

Quality vs. Quantity
An aspect of Ahmed's discussion that resurfaced throughout the remainder of the meeting involved assessing not just quantitative but qualitative aspects of T cell responses. Commenting on the use of gamma-interferon production to measure the presence of antigen-specific T cells, Ahmed pointed out that in functionally compromised or "exhausted" CD8 T cells, the capacity to produce this cytokine is retained long after other functions (such as proliferation, IL-2 production and TNF-alpha production) are lost. Ahmed's comment echoes data from Norm Letvin's laboratory showing that CD8 T cells from long-term non-progressor and vaccine-protected macaques retain the ability to produce IL-2 and TNF-alpha when compared to CD8 T cells from animals with progressing disease (McKay, 2002).

Among the other presentations at Keystone that investigated qualitative aspects of the virus-specific T cell response was a poster from Mark Boaz of Kings College School of Medicine in London (Boaz, 2002). The study compared HIV-specific CD4 T cell responses in 16 long-term non-progressors (LTNPs) and 15 patients with progressive disease. Measuring CD4 T cell responses to HIV's p24 protein based on gamma-interferon production revealed no differences between the two groups, confirming results from other recently published studies (Betts, 2001). But when Boaz analyzed p24-specific CD4 T cells from the two groups for co-expression of gamma-interferon and IL-2, a statistically significant difference emerged: LTNP had higher frequencies of these dual cytokine-producing cells than progressors (p value of 0.0015).

Marie Claire Gauduin from the New England Primate Research Center also presented data on virus-specific CD4 T cell functionality, this time in macaques infected with pathogenic or attenuated strains of SIV (Gauduin, 2002). Gauduin assessed the ability of CD4 T cells specific for the gag p55 antigen to proliferate, using a new technique that combines staining for intracellular cytokine production with a marker of cell division called CFSE. In macaques infected with attenuated SIV, p55-specific CD4 T cells began to divide between 48-60 hours after in vitro antigen stimulation and by 96 hours almost all p55-specific cells had undergone division. In stark contrast, p55-specific CD4 T cells from animals infected with pathogenic SIV underwent little to no division in the same assay, suggesting that these responses had been functionally compromised by an as yet unknown mechanism.

Switching T Cells On and Off
If there were an award for best movie of the conference, it would undoubtedly go to James Allison from the University of California at Berkeley (Allison, 2002). Allison's group has been investigating CTLA-4 and CD28, two signaling molecules that T cells use when communicating with other immune system cells. Both CTLA-4 and CD28 belong to a family of "co-stimulatory" molecules that can play a complex variety of roles in T cell activation. T cells are typically activated (or switched on) by an encounter with an antigen-presenting cell (APC). APCs specialize in slicing proteins into small fragments called peptides and presenting them to T cells for inspection (the ultimate purpose is to trigger an immune response if the peptide is foreign, e.g. from a virus, but not if the peptide is from the self). If the peptide fits into the T cell's receptor—a sort of docking bay structure on the surface of the cell—the contact between T cell and APC will be prolonged and other molecules will move around the cells to gather at the site of T cell/APC contact (this site is called the "immunological synapse). These molecules form additional contacts between the T cell and the APC, allowing more signals to be sent back and forth. These signals serve to further refine how the T cell will respond to the peptide it has grasped in its receptor.

Allison managed to film examples of this process in action, focusing on the movements of CD28 and CTLA-4 as they shift position and congregate at the immunological synapse. The motivation for the experiment was that CD28 and CTLA-4 both form contacts with the same molecule on APC (B7), but send opposing signals to T cells: if CD28 molecules move to the immunogical synapse and engage B7 on the APC, the signal that gets sent to the T cell enhances activation. But if CTLA-4 molecules beat CD28 in the race to the synapse, they will bind B7 and the signal that gets sent to the T cell suppresses activation. What Allison's movies revealed is that CD28 generally has the advantage—these molecules are expressed on the surface of most T cells and move rapidly around to the site of APC contact, leading to T cell activation. In contrast, CTLA-4 molecules are retained inside the cell in the cytoplasm and cannot travel to the synapse so easily. Tagging the two molecules with green fluorescent protein allowed a dramatic visualization of these movements: as the two blobs representing the T cell and APC conjoined, brightly illuminated CD28 molecules could be seen slithering quickly around the perimeter of the T cell to gather at the point where the surfaces of the APC and T cell met. In the case of CTLA-4, the fluorescent molecules emerged from within the T cell on the opposite side from the APC and migrated to the immunological synapse at a slower pace.

Having identified these differences, Allison wanted to know if the interaction between T cell receptor and peptide had any influence on the recruitment of CD28 and CTLA-4 to the immunological synapse. His curiosity resulted in a striking finding: the avidity of the interaction between the T cell receptor and peptide was directly proportional to the amount of CTLA-4 that moved to the synapse. In other words, the more tightly the T cell receptor bound to the peptide, the more likely that CTLA-4 would be engaged and thus the T cell response suppressed rather than activated. Although it may seem arcane, this finding has important ramifications in terms of how the body launches a T cell immune response to a vaccine or infection. For example, many researchers have hypothesized that vaccines will work better if they induce T cells that bind to pathogen-derived peptides with high avidity. Allison's work offers the confounding suggestion that, while such high avidity T cell responses may look great according to in vitro assays that do not require APC/T cell interactions, in vitro they might be switched off by the rapid mobilization of CTLA-4 to the immunological synapse.

Allison speculated that the phenomenon he has observed could serve at least two useful purposes: it could suppress high avidity T cell responses to self-derived peptides (potentially preventing autoimmunity), and it could broaden the immune response by preventing a small number of narrowly targeted, high avidity T cells from dominating (potentially reducing the risk of immune escape). To illustrate the latter point, Allison conjured up the image of T cells competing for access to peptides being presented by APC "like pigs at a trough." A few T cells with a high avidity for a few pathogen-derived peptides could easily dominate an immune response, and mutations in these peptides could lead to the pathogen escaping from immune surveillance.

In concluding the presentation, Allison raised the possibility that therapeutic blockade of CTLA-4 might enhance T cell responses in some disease states. He noted that anti-CTLA-4 antibody infusions are currently being studied in humans for the treatment of several cancers. Recent evidence that CTLA-4 is upregulated in HIV infection has also led Israeli researcher Zvi Bentwich to suggest that anti-CTLA-4 approaches might enhance T cell responses to HIV (Leng, 2002).

The Rare HLA Advantage
The closing talk at the Keystone meeting was given by gene genius Bette Korber from the Los Alamos National Laboratory. Korber presides over the voluminous HIV databases at Los Alamos, which contain data on viral genetic sequences, immunological epitopes, drug resistance-associated mutations, and vaccine trials. Recently, Korber has been investigating the effects of HLA variation on both viral load setpoints and disease progression using data from the Chicago site of the Multicenter AIDS Cohort Study (MACS).

For background, the HLA (Human Leukocyte Antigen) system comprises the molecules that process and present peptides to the immune system. There are two major classes of HLA molecules: Class I HLA molecules are found on almost every cell in the body and present peptides to CD8 T cells. Class II HLA molecules are found on more limited range of immune system cells (mainly dendritic cells, macrophages and B-cells) and present peptides to CD4 T cells. Both class I and II HLA molecules come in hundreds of different versions, dependent on the HLA genes inherited from our parents. The precise shape and size of an HLA molecule governs its ability to associate with a diverse array of peptides and present them to T cells. HLA molecules thus exert a profound influence on the body's ability to mount a broad and effective T cell response to any given pathogen. Each individual has over 40 different genes that encode HLA molecules. The class I genes are divided into different regions (or loci), with the most important being HLA-A, -B and -C. The major class II genes are HLA-DP, -DQ and -DR. There are many variations of these genes in the human population and thus many variant HLA molecules. The different versions of the genes are known as alleles, and a complex classification system is used to characterize the specific HLA alleles that an individual has inherited (this inheritance is known as the HLA type).

Korber analyzed the HLA types of 1,000 Chicago MACS participants, 564 of whom were HIV-positive and 436 HIV-negative. Viral load setpoints prior to treatment were available for 481 infected individuals. To refine her analyses, Korber grouped HLA types into "supertypes"—this groups together class I HLA genes based on structural similarities in the molecules they encode. The novel and important finding that emerged from Korber's work was that the frequency of class I HLA supertypes in the population was correlated with the viral load setpoint in people with that particular supertype. To put it another way: the less common the class I HLA supertype, the lower the viral load setpoint and vice versa. Korber stated that three-quarters of the MACS population could be grouped into one of six class I HLA supertypes, suggesting that such analyses could provide useful prognostic information for individuals.

Korber also reported that African American men in the Chicago MACS had lower viral load setpoints, on average, than their Caucasian counterparts. Although the numbers were relatively small (44 African American men were included in the analysis), this observation echoes data on African American women published by Kathy Anastos using data from the Women's Interagency HIV Study (Anastos 2000). Whether these results are linked to an increased prevalence of rare HLA supertypes among African Americans remains to be elucidated.

In terms of an explanation for the relationship between class I HLA supertype and viral load setpoint, Korber believes that it lies in the transmission of viruses with immune escape mutations—if transmission occurs between two people with the same or similar supertype, mutations that allowed escape from the CD8 T cell response of the transmitter will also allow the virus to evade the CD8 T cell response in the newly infected individual. Conversely, if transmission occurs between two people with very different HLA supertypes, escape mutations that allowed the virus to evade the CD8 T cell response in the transmitter will confer little advantage in the newly infected individual because their class I HLA molecules will present a different set of viral peptides to the immune system. Korber suspects that this phenomenon may also play a role in driving the evolution of HIV subtypes, which tend to cluster among genetically related populations.