Skip Navigation
Leading research to understand, treat, and prevent infectious, immunologic, and allergic diseases
Skip Content Marketing
  • Share this:
  • submit to facebook
  • Tweet it
  • submit to reddit
  • submit to StumbleUpon
  • submit to Google +

AIDS Vaccine Research Subcommittee Meeting—January 27-28, 2009

The AIDS Vaccine Research Subcommittee (AVRS) met in public session on January 27–28, 2009, in the Natcher Conference Center on the campus of the National Institutes of Health (NIH) in Bethesda, MD.

AVRS members present: Eric Hunter (Chair), James Bradac (Executive Secretary), Jay Berzofsky (ex officio), Deborah Birx, Dennis Burton, Kevin Fisher, Paul Johnson,
Jeffrey Lifson, Margaret Liu, Bonnie Mathieson (ex officio), Juliana McElrath, Nelson Michael (ex officio), Gary Nabel (ex officio), Louis Picker, Bali Pulendran, Nina Russell, Jerald Sadoff, Bruce Walker.

Other participating NIH personnel:

  • Carl Dieffenbach, Director, Division of AIDS (DAIDS), NIAID
  • Peggy Johnston, Director, Vaccine Research Program (VRP), DAIDS, NIAID
  • Patricia D’Souza, Vaccine Clinical Research Branch, VRP, DAIDS, NIAID
  • Michael Pensiero, Preclinical Research and Development Branch, VRP, DAIDS, NIAID
  • Adam Sherwat, Vaccine Clinical Research Branch, VRP, DAIDS, NIAID


  • Dan H. Barouch, Beth Israel Deaconess Medical Center
  • Michael Betts, University of Pennsylvania School of Medicine
  • Mark Connors, Laboratory of Immunoregulation, National Institute of Allergy and Infectious Diseases (NIAID)
  • Nicole Frahm, Fred Hutchinson Cancer Research Center
  • Ronald Germain, Laboratory of Immunoregulation, NIAID
  • David Heckerman, Microsoft Research
  • Robert Palermo, University of Washington
  • Rafick-Pierre Sékaly, Oregon Health Science University and Université de Montréal
  • Wendy Tan, Emory Vaccine Center
  • Daniel Zak, Institute for Systems Biology

Call to Order

Dr. Hunter called the meeting to order and asked the committee members and observers to introduce themselves. He also recognized the service and contributions of three AVRS members who are rotating off the subcommittee: Paul Johnson, Margaret Liu, and Julie McElrath.

STEP Analysis Update

Julie McElrath identified the two major scientific questions emerging from the STEP trial; namely, (1) What are the reasons for the lack of vaccine efficacy, and (2) What biological mechanisms explain the increased HIV-1 acquisition in the vaccine group? To date, 22 proposals have been submitted and 10 approved for studies using select vials of the 247,000 specimens that have been shipped and accessioned to date. In some cases, the experiments have already been conducted and data are under analysis.

  • In the area of comprehensive immunogenicity, for example, HIV-specific T cells were detected in 89 percent of vaccine recipients, but only 31 percent mounted both CD4+ and CD8+ responses after three doses, and the median percentage of antigen-specific CD8+ T cells was 43 percent lower than in HIV-infected long-term nonprogressors (LTNPs)
  • In the area of cofactors and covariates, other studies have shown that inherited HLA types are associated with HIV disease control (i.e., set point viral load [sVL]), with some alleles being protective (HLA-B57/B58/B27) and others being unfavorable (HLA-*3502/*3503/*3504/B*3501). sVL is not affected by most baseline covariates (treatment arm, race, age, region, circumcision). However, results did show that high Ad5 titers (>200) blunted T cell responses in vaccinees, particularly Gag responses. HIV infection in cases is not attributable to this reduce breadth of responses, but decreased targeting of Gag epitopes is associated with higher sVL
  • In the area of viral sequencing, investigators have already sequenced 318 full genomes from 55 cases, 23 percent of which have multiple variants, but there is no evidence of phylogenetic clustering. Final results are due in February 2009.
  • In the area of adenovirus-specific immunity, analysis of 82 cases in the STEP study indicated a possible increased infection rate in vaccine recipients with baseline Ad5+ nAb titers greater than 18. However, nested case-control studies of two other cohorts (MACS and HPTN 039) with a total of 368 cases showed no association between Ad5+ titer and acquisition of HIV

Taken together, these findings suggest that T cell vaccines need to elicit a stronger and broader response to achieve a threshold of protection. However, which specific epitopes are recognized may be more important than the number. Ad5 vectors, as currently constructed, may not provide the “correct” memory response and may in fact dampen the response to HIV-1 transgenes, but there is no evidence that pre-existing Ad5 immunity increases susceptibility to HIV-1. In response to questions, McElrath added that investigators are still trying to determine if the epitopes recognized by vaccinees are those that confer protection. In addition, favorable HLA type doesn’t always convey protection. There will be further information and specific cases to present to AVRS in May.

Larry Corey presented findings on the impact of herpes simplex virus type 2 (HSV-2). In the STEP cohort, about 30 percent of all participants are HSV-2 positive, and it increases their likelihood of HIV infection independently of other risk factors, although there also appears to be some interaction between HSV-2 status, Ad5 serostatus and circumcision. Mechanistically, investigators theorize that HSV-2 infection induces a chronic inflammatory response in the mixed dendritic cells at the dermal-subdermal boundary near nerve endings, even when there is no apparent lesion, and that abrasion during coitus exposes these target cells to HIV. Consequently, HSV-2 infection may pose a barrier to HIV vaccines, pointing out the urgent need for an HSV-2 vaccine as well.

Correlates of Immunologic Control of HIV

Patricia D’Souza suggested that current assays may not be adequate to identify the correlates of protection. Clearly not all cytotoxic T lymphocytes (CTLs) are created equal—CD8+ cells show great promise, but several factors can contribute to the antiviral efficacy of CD8+ T cells: polyfunctionality, functional avidity, killing efficiency, evolutionary constraints on the epitope sequences, the kinetics of antigen presentation, and proliferative ability. In addition, the effectiveness of an immune response to a T cell vaccine may vary from person to person just as the immune response to HIV infection does, and this variation may be strongly related to HLA allele. Thus a T cell vaccine may augment the body’s genetically determined natural ability to respond to HIV, resulting in varying levels of control that depend on the person’s HLA allele.

Bruce Walker presented data on CTL function in the 500 LTNPs in the International HIV Controllers Cohort, some of whom have normal CD4+ counts and extremely low viral loads 30 years after seroconversion. Their ability to contain the HIV infection is related to differences in viral fitness that are mediated by the immune system. For example, the protective HLA B*57 allele is overrepresented in elite controllers and seems to be an important link between the MHC and viral fitness. In some way, CTL-expressed peptides (such as TW10, which inhibits p24 synthesis) exert selective pressure on the viral population, controlling viral load by limiting the virus’ ability to mutate and thus escape. It would appear that it is the HIV epitopes that are targeted by CTLs, rather than the HLA alleles, that are protective. The task now is to identify which HIV epitopes predict control and progression, and to develop a vaccine that elicits a response to those specific epitopes.

David Heckerman cited similar evidence to explain why the Merck vaccine failed to reduce viral load in STEP cases despite the induction of HIV-specific CD8+ T cells. Following vaccination, investigators assayed the response to 76 peptide “minipools” spanning Gag, Pol and Nef. Each response was attributed to a known epitope within each minipool and a matching HLA restriction in the individual, and the predicted responses were correlated with early sVL at weeks 8 and 12 after infection. The results showed that control of set point VL is correlated not with the total number of responses, but rather by the number of responses to specific, protective epitopes. These “good” epitopes are not limited to Gag, nor are they limited to those presented by protective alleles such as B*57 and B*27. This suggests that the design of a successful vaccine immunogen may hinge on including good epitopes and excluding others that distract the immune system from targeting regions associated with control of viral replication.

In response to questions, Heckerman said that investigators are currently taking a closer look at what defines a “good” epitope, but that we may need a higher bar for “good.” For example, some of the same mutations are seen in both LTNPs and progressors, and more data will be needed to tease out the exact parameters of “protection.” Those additional data will soon become available, allowing investigators to also look at post-infection events and to answer questions about how immune response prevents compensatory “escape” mutations in HIV.

Mark Connors presented data suggesting the importance of T cell function in immune control. Analysis of LTNPs reveals that neither HLA profile in general, nor B*57 in particular, is necessary or sufficient to suppress viral load. However, LTNPs do have a higher proportion, and/or more active population, of early CD8+ effector T cells that secrete interferon-gamma (IFN-γ), granzyme-B and perforin. Investigators tested the role of these effector molecules in vitro killing assay, first mixing effector cells with labeled target cells for an hour and then measuring for granzyme-B. They observed a large-scale killing that correlated with the delivery of granzyme-B to labeled cells, suggesting a qualitative increase in the ability of HIV+ T cells to kill infected cells. Assays of PBMCs from HVTN 071, which involved the Merck trivalent vaccine, revealed no such pattern; the vaccine produces a response profile that is more similar to progressors than to LTNPs. These results suggest that the proliferation and loading of CD8+ effector T cells play a key role in controlling virus, but they also suggest the need to induce a slower, long-term proliferation and loading. They also reveal the need for a better assay of effector molecules, something similar to the in vitro killing assay.

In response to questions, Connors repeated that killing is a quality of T cells, regardless of peptide or profile, but that finding the optimal peptide will result in a better killing. The goal is an assay that distinguishes between phenotypes with no overlap. These results were obtained at 36 hours after infection, but it would be useful to repeat the experiment at day 6 to find out exactly what’s being expanded. Similarly, results from PBMCs don’t tell us what’s going on in tissues; it would be useful to repeat the experiment with mucosal samples, especially cells from the site of infection. Investigators have not yet conducted tetramer inhibition of killing to measure its specificity, for example, if Env-specific T cells are more potent than Gag-specific T cells. There don’t appear to be differences among patient groups, but these results (using SF162 virus, a clade B isolate) have not been repeated in other clades.

Analysis of Ad5-Specific Responses in Vaccines

Michael Pensiero noted that the STEP findings with regard to the rAd5 vector have stimulated renewed interest in the study of wild type adenovirus infection. Pertinent research questions include the correlation between Ad5 immunity and HIV infection, expansion of T cell populations in Ad5+ vaccinees, response to challenge, impairment of immune response to HIV transgenes, and implications for vaccine design.

Nicole Frahm suggested that a more comprehensive understanding of vector-specific cellular immune responses will be necessary before vectors are used in future vaccine trials. Her lab examined PBMCs from 413 STEP trial participants and measured their responses to the Ad5 vector (without inserts) and 1,773 Ad5 15-mer peptides overlapping by 11 amino acids. Results showed that Ad5-specific T cell responses are detectable in the majority of individuals from the STEP trial and are influenced by Ad5 serostatus and vaccination. Responses to empty vector are predominately CD4+ T cell-mediated; in placebo recipients, responses are more frequent in subjects with nAb titers >18, but magnitudes are similar. In vaccine recipients, response rates and magnitudes are significantly dampened by pre-existing nAb titers >18. Responses to Ad5 peptides lead to significant increases in CD8+ T cell response rates in all vaccinees, regardless of Ad5 serostatus, and magnitudes are significantly increased in Ad5 seronegative subjects; CD-4+ response rates increase only in seronegative subjects. Epitope mapping reveals that CD4+ responses are predominately directed at hexon and core proteins, while CD8+ responses are directed at an additional target, E3/E4, with only minor differences in specificity between vaccinees and placebo recipients. Ninety-two percent of the targeted peptides are completely conserved across group C adenoviruses (Ad1, 2, 5, and 6), which may explain the presence of Ad-specific T cell responses in the absence of Ad5 nAbs.

In response to questions, Frahm added that the multiple natural preinfection with group C adenoviruses may have a huge priming effect, leading the immune system to respond strongly to the vector while distracting it from the HIV inserts. This might be avoided by using a less common vector; much will be learned from ongoing trials of using modified vaccinia Ankara and canarypox vectors. It might be possible to somehow knock out the early (vector) response and focus the response on the transgenes.

Michael Betts presented data from analysis of samples from a Phase 1 trial of the Merck vaccine that show no difference in the expression or functionality of CD4+ T cells in Ad5 seropositive and seronegative vaccinees, and no difference in the induction of immunity. Vaccination increases the number of CD4+ cells early, especially in seronegative vaccinees, but by week 20, the responses are similar. For this reason, it is unlikely that Ad5 serostatus can account for the higher rate of infection in vaccinees.

In response to questions, Betts said that expression of CD67 (a possible activation marker) does not change at any time point. Experimental design forced investigators to infer what was happening in tissue from what could be measured in PBMCs; trafficking markers are not yet available and must come from a larger specimen base.

Dan Barouch addressed the same underlying question, whether Ad5-specific immunity acts to increase the number of targets for HIV, also using data from a Phase 1 trial of the Merck vaccine. He found no correlation between Ad5 serostatus and Ad5-specific T cell responses. Seronegative vaccinees became seropositive for Ad5, but they showed no detectable responses to Ad26, Ad35, or Ad42. Vaccination did not lead to changes in CD4+ subsets or sustained activation of CD4+ T cells. Taken together, these data reduce the plausibility of the hypothesis that Ad5 nAb titer >18 contributes to HIV acquisition in STEP vaccinees. Barouch’s lab is currently looking at seroprevalence of alternative Ad vectors (5, 26, 35, and 48) in the STEP study to determine if baseline nAbs to other Ad serotypes correlate with enhanced acquisition.

In response to questions, Barouch said that Ad5 immunity remains a risk factor in uncircumcised men. There is little cross-reactivity among the 51 human adenovirus serotypes, except within the 6 subgroups, and there is much to be learned from rare and weakly infective varieties such as Ad26, even if they are less-than-perfect vectors. More data are needed on mucosal responses, which might be investigated in nonhuman primate (NHP) studies, for example, nasal challenge with replication-competent Ad5 followed by mucosal biopsies, repeated with other Ad types. Heterologous prime and boost is another strategy that deserves further investigation. Historical data suggest that NHPs are not a perfect model of human response to Ad5.

Daniel Zak presented a systems-level analysis of the effects of Ad5 immunity on vaccine responses, using data from the HVTN 071 trial. Using microarrays to determine which genes and pathways have been activated at intervals of 6 to 168 hours after vaccination, investigators found robust, consistent responses across 10 subjects. Maximum response comes at 24 hours, with 583 genes consistently upregulated and 482 downregulated, most of them having functions related to immune response, antigen processing and presentation, and lymphocyte proliferation. Ad5-specific immunity attenuated the response of 148 of these genes (84 up, 64 down), possibly because these subjects have baseline differences in the expression of adjuvancy-relevant genes induced by the vaccine. Investigators also identified several novel correlates for the magnitude of CD8+ responses.

In response to questions, Zak added that Ad5 immunity attenuates a number of responses, but it doesn’t appear to amplify any responses. It’s not clear whether this happens because the stimulus is reduced or the response is reduced; this could be tested through dose reduction. It would be desirable to have data from lower doses (of vaccine) and at earlier and later timepoints. At present, no samples are available beyond a week, and it would be desirable to repeat the analysis at 20 days. Yellow fever and other attenuated virus vaccines have slower responses than Ad5 vaccines.

Wendy Tan presented data from mouse studies showing that different vectors induce memory T cells of different qualities. Specifically, a comparison of Ad5 with lymphocytic choriomeningitis virus (LCMV) and Listeria monocytogenes (LM) showed that Ad5 compares favorably on some measures (e.g., absolute number of CD8+ T cells), but relatively low on others (e.g., polyfunctionality and proliferation capacity). In other words, the Ad5 vector induces a response that is high in magnitude but restricted in breadth. DNA priming can significantly increase the magnitude of Ad5-induced CD8+ T cell response, but it has minimal effect on its breadth or functionality.

In response to questions, Tan acknowledged that 1011 viral particles is a high dose, but when the experiment was repeated at 107 particles, the quality of the response was irregular. It’s not clear whether the differences are due to the vector or the dose; investigators haven’t repeated every vector at every dose. Proliferation data are slow to collect and are not yet available. Investigators are still trying to determine the target cell for Ad5-specific CD8+ T cells, but they assume they are dendritic cells. There are significant differences between human and mouse immune systems, but the mouse appears to be a reliable model for Ad5 response and memory. Investigators have not looked at data from earlier Ad5 vaccines at the NIH Vaccine Research Center (VRC).

Program Updates

Adam Sherwat presented plans for HVTN 505, a proof-of-concept trial of the DNA prime + rAd5 vaccine developed by VRC. The protocol calls for enrolling roughly 1,300 HIV-negative men having sex with men (MSM) in the United States. Enrollment is restricted to men who are circumcised and Ad5-seronegative; enrolling the 1,300 participants is expected to take about 36 months, and it is anticipated that 3.6 percent of participants will acquire HIV infection each year. The trial will be monitored for safety, futility, and endpoints; a meeting with FDA was held in January 2009.

In response to questions, Sherwat said that participants would also be screened for MHC/HLA phenotype, but not until after they are enrolled. Earlier stratification might yield stronger results, but screening is already onerous, and investigators don’t want to extend the process. Safety monitoring will include a crossline cutoff, but weekly review of viral load and other endpoints will not be possible in this trial, primarily because finger-prick techniques are not ready. The trial should have 80 percent power to detect a 57 percent difference in protection against infection.

James Bradac presented an update on preclinical activities in the DAIDS Vaccine Research Program (VRP). In response to the vaccine summit, VRP is putting greater emphasis on vaccine discovery research, with new money for basic research and efforts to make better use of NHP resources. NIAID published two RFAs in August 2008—the Highly Innovative Technologies to Interrupt Transmission of HIV (HIT-IT) Program, for which applications were received in November 2008, and the Basic HIV Vaccine Discovery Program, for which applications were received in January 2009. First awards for both programs are scheduled for July 2009. In addition, DAIDS has sponsored workshops on supply, demand, and use of NHP resources and on B-cell vaccines; an RFA should be ready soon for a B-Cell Biology Network for AIDS Vaccine Development program.

Previously existing programs will continue. Six new Phased Innovation Awards were made in 2008, and some of the earliest awardees are ready for the transition from R21 phase (proof of concept) to R33 phase (expanded development) without having to submit new applications. In response to questions, Bradac explained that the R21 milestones in the Phased Innovation Award program are self-imposed and serve as part of the 18-month progress report and transition review.


In the discussion that followed, Eric Hunter indicated that the job of the AVRS is to transform its deliberations into recommendations to DAIDS, including suggestions for priorities and directions and (if possible) a series of action items. A summary of the present meeting, with PowerPoint presentations, will be made available to members within a month, and AVRS will reexamine these questions in its May meeting.

Other participants suggested that topics such as insert design and vector effects are so complex that they may be too much for any one group to handle. Similarly, what funding mechanism is most appropriate? NIAID has more control over a contract than it does over grants, but clearly a consortium approach will be needed. Funding should be made available for a consensus conference and a mechanism found to support joint efforts. Other participants indicated that we need a better understanding of the basics—for example, today’s presentations suggest that it is more important to target the right epitopes than it is to elicit a massive but untargeted response. Perhaps a “speed dating” approach would help to answer the basic questions; set up and attack a few straw men, then let the decisions on program directions be outcome-driven.

Another option would be to organize consortia around specific questions, or teams, to generate and test specific hypotheses (e.g., can the choice of vector control the effector response in mucosa?). A specific example would be a Viral Vector Safety Group that could apply standard measures across multiple vectors, thereby addressing cross-cutting questions in vaccine design. This would be a good topic for an upcoming Global HIV Vaccine Enterprise meeting. There seemed to be agreement that vectors and inserts should be addressed separately, but with attention to synergies and a focus on outcome, what we want the vector and insert to achieve.

The meeting adjourned at 5:10 p.m. and reconvened the following morning at 8:30 a.m.

Applying Systems Biology Approaches to the Study of Vaccine Efficacy

James Bradac opened the session by reporting that the NIAID Division of Microbiology and Infectious Diseases (DMID) has awarded four grants in FY 2009 for Systems Biology for Infections Diseases:

  • Battelle-Pacific Northwest National Laboratory (Salmonella enterica and Yersinia pestis)
  • Institute for Systems Biology (influenza virus and Staphylococcus aureus)
  • Stanford University (Mycobacterium tuberculosis)
  • University of Washington (influenza virus and SARS-associated coronavirus [SARS-CoV])

In addition, planning is underway for a FY 2010 initiative from DAIDS using the P01 or P50 mechanism for Dissecting HIV Immune Response: A Systems Biology Approach. The purpose of the following presentations is to help AVRS make recommendations for FY 2011 on what additional initiatives should be undertaken, using what funding mechanisms. Two specific topics of interest are: (1) What questions should be addressed using systems biology techniques and available human and NHP samples and datasets, and (2) What systems biology technologies should be applied to vaccine research samples on a routine basis?

Ron Germain introduced systems biology as a combination of the principles of engineering, mathematics, physics, and computer science with extensive experimental data, at many different scales, that can be used to develop a quantitative and conceptual understanding of biological systems and phenomena, permitting prediction and accurate simulation of complex biological events and behaviors. Inputs include genomic and epigenomic data, large-scale expression profiling, high-throughput proteomic analysis, four-dimensional imaging datasets, multiparameter nanoscale analysis, and global RNAi and chemical library screening. Outputs include engineering diagrams, linkage maps and process cartoons, as well as quantitative simulation models of the network organization of cell components, differentiated cells, diseases, etc. These models allow investigators to predict the effect of changing one or several components. Conventional approaches to these questions and datasets are ponderous and difficult to update, but the Simulation Immunology (Simmune) computer program allows biologists to construct and run complex models with the necessary mathematics “in the background.” Simmune supports computational models at all scales, from molecular interactions to multicellular systems, allowing investigators to understand, predict, and validate complex interactions.

This and other tools are being developed and disseminated by NIAID’s Program in Systems Immunology and Infectious Disease Modeling (PSIIDM) and the new trans-NIH Center for Human Immunology (CHI), but the present state of the art is limited; tools are flat and static, and models don’t mine the data for new and unexpected results. A major goal of CHI is to understand how the human immune system (not that of a model organism) operates in healthy individuals and in the course of disease. This will require the development of new analytic tools, the collection of more data into larger datasets, and far more sophisticated, iterative exploitation of those data to improve our understanding of the immune system. One important endpoint will be “high-density datasets,” and another will be a massively parallel laboratory-on-a-chip that combines biochemistry, genomics, proteomics, and disease fingerprinting to produce a differential diagnosis and prognosis for each patient as a whole organism.

In response to questions, Germain said that models can be tested both by iteration (making and testing multiple predictions) and by sensitivity testing (comparing predictions with real biology). Building the model inescapably adds to the researcher’s understanding of the system, which will proceed from simple to complex. Microarray analysis currently represents a kind of “low-hanging fruit” that is ripe and ready for analysis; perturbation of systems will lead to more sophisticated insights. The current plan is to use a “Wikipedia” approach, creating a public database of structural biology models that can by copied and tested, but not modified, by outside users. In each case, the model (and any subsequent changes by the moderators) will be linked to experimental data. In this way, the field, not CHI, curates the models.

Bali Pulendran provided an example of how systems biology can provide insights into the mechanisms by which the empirical vaccines of the past mediate their efficacy. In this case, yellow fever vaccine 17D (YF-17D) is one of the most effective vaccines ever made, producing a multipronged immunity that persists for over 30 years after vaccination, yet we do not understand how it works. Previous research had established that YF-17D activates multiple dendritic cell subsets via Toll-like receptors (TLRs) 2, 7, 8, and 9 to stimulate polyvalent immunity. Further analysis led to the identification of innate immune “signatures” that can predict the quantity, quality, and persistence of antigen-specific CD8+ T-cell and B-cell responses to FY-17D. Not surprisingly, several of these genes have functions related to viral infection; one in particular, EIF2AK4, is linked to a conserved process by which the cell produces granules in response to several kinds of stress, including heat, oxidation, nutrient limitation, and UV radiation, as well as viral infection. The predictive value of this gene can then be tested in EIF2AK4-knockout mice, which indeed induce greatly diminished IFN-γ in T cells. But YF-17D induces whole families of genes; Pulendran suggested that the data on this and other genes in response to YF-17D could be mined for another 20 years.

Investigators are beginning to apply these same techniques to other vaccines. For example, an analysis of flu vaccines showed that there was very little overlap between the genes induced by trivalent inactivated influenza vaccine, which is injected, and live attenuated influenza vaccine, which is administered as a nasal mist. An analysis of acute infections identified 35 genes that best discriminated patients with influenza A virus from those with E. coli or S. pneumonia infection. In the longer term, investigators hope to identify the innate immune signatures of vaccine immunogenicity and efficacy for “good” and “suboptimal” vaccines, as well as the innate correlates of CTLs, nAbs, Th1, Th2, central memory, effector memory, etc., and incorporate them into a global access database that can be used to design a “vaccine chip” as well as vaccines themselves. It seems unlikely that they will discover any single, universal correlate of immunogenicity, although clusters of correlates are possible. The more important question is what new biological insights can be ascertained from these signatures.

In response to questions, Pulendran added that in some cases, this approach reveals the signature of the target cells rather than the vaccines. Investigators haven’t done SNP analysis of key genes, but they can make informed guesses. Further work is also needed on memory response—does the primary response predict long-term protection? This is an experiment that could be done in humans, but it might be useful to test it first in the NHP model of yellow fever.

Rafick-Pierre Sékaly described the capabilities of the Illumina chip, which uses systems biology techniques to process, analyze, and identify gene pathways elicited in response to protective vaccines, natural protection (e.g., LTNPs), and chronic infection in different types of cells. Experience has shown that a strong memory T cell response leads to protective immunity; analysis with the Illumina chip shows that memory CD4+ T cells in LTNPs show distinctive gene expression profiles. An example is FOX03a, which expresses a protein that targets genes with antiproliferative and/or apoptotic functions; silencing FOX03a rescues memory T cells in chronically infected HIV subjects from Fas-medicated apoptosis. Further analysis demonstrates that a mutation in the FOX3a N-terminus also increases the persistence of T cell memory. A similar phenomenon appears to occur in the memory B cells of LTNPs.

Sékaly said that investigators are now using the Illumina chip to repeat their yellow fever work in three separate labs (Atlanta, Montreal, and Lausanne) to identify the genes and networks induced by YF-17D. Preliminary results show that YF-17D induces a multicellular innate immune response involving genes associated with IFN and TLR activation, macrophages, natural killer cells, Th1 and Th2 cells, and activation of the complement system. YF-17D also induces the expression of genes associated with assembly of the inflammasome, secretion of IL-1α and ß and IL-7, and the modulation of transcription factors and antigen-processing genes—in short, a complex adaptive immune response that is also long-lasting. These gene expression signatures have been validated in the in vitro Mimic model and in the NHP (macaque) model. However, there doesn’t appear to be a single correlate of protection; the whole (the constellation of response) is definitely greater than the sum of its parts.

Robert Palermo described the use of structural biology techniques to study innate immunity and early host response as determinants of pathogenesis in RNA respiratory viruses and the potential for their use in NHP models of HIV. At present, investigators are conducting functional genomic studies in NHP models and in PBMCs from STEP trial participants, including Env subunit boosts, expression profiling, and comparisons of pathogenic and nonpathogenic outcomes. Results to date indicate that different vaccine regimes induce different patterns of gene expression; correlation with immunological and virologic data is ongoing, but preliminary analysis suggests a deeper and more complex response to a nonpathological infection than to a fatal infection. In a suitably coordinated endeavor, investigators believe that they can use similar techniques in NHP models to discover expression differences between responders and nonresponders, or between controllers and noncontrollers. Systems biology greatly increases the volume of data generated by such studies, but it also provides the computational tools for integrating those data in order make sense of them.

Daniel Zak presented data from an NHP study of innate immune response to 10 different vaccine-adjuvant regimes. Thirty-eight animals were vaccinated with different combinations of SIV-Gag with or without one or more TLR adjuvant(s), or with Ad5-Gag only, boosted at 10 and 20 weeks, and challenged at 50 weeks. Results showed that regimes including TLR-3 induce the strongest CD4+ responses, and that TLR-3 by itself induces the strongest CD8+ responses. However, TLR7 and 8 induce the strongest and most consistent expression responses after primary vaccination, especially from IFN-stimulated genes. Further microarray analysis revealed two novel predictors of immunity: stronger induction of CCR2B and stronger repression of KLF5 are associated with stronger CD4+ T cell responses. Mamu A*01 macaques were better protected than non-A*01 animals, and their lower sVL correlated with induction of STAT2 and IRF9 (IFN-stimulated transcription factors), NCF1/NOXO2 (proliferation of phagocytes), and TRAIL (apoptosis-inducing ligand). Novel genes correlated with lower sVL were E2F2 (represses cell cycle genes for quiescence) and PKD2L2 (cation channels). In general, the response on day 2 was predictive of long-term outcome.

In response to questions, Zak added that PBMCs had not been sorted for cell type, and investigators would need naïve cells to track changes in specific subpopulations. PBMCs are immediately preserved once drawn, and there’s little chance for changes in expression prior to analysis. Investigators collect both blood chemistry and microarray data to get parallel information, which can be deconvoluted later.


Participants agreed that these were exciting results and that it would be desirable to incorporate some structural biology techniques in ongoing HVTN trials. One problem is deciding what responses to look for, in what cells, at what intervals (from 1 hour to 1 week). In a human trial, however, this means managing a huge number of blood draws. Unfortunately, total PBMC data from acute infection will not correlate with disease outcome; T cells are needed. Technologies for blood draws and analysis will no doubt improve, thanks in part to improved microarray and chip design. For this reason, some money should go to improvements in high-throughput assays.
Participants acknowledge the capacity of structural biology to deal with large volumes of data, but the goal is to detect a signal—several participants suggested that it is better to overdesign a study than to have masses of useless data. On the other hand, basic biology often benefits from having as much data as possible. A continuing problem is the management of the resulting databases; who will curate them, and, more importantly, who will pay for them? What are the minimal standards for a useful database?

There are several ongoing vaccine trials (dengue, TB, and malaria, as well as HIV) whose results could be analyzed with these techniques. It would also be useful to employ these techniques for a pseudo-retrospective reanalysis, using data from past trials: what cells were present at different times, and does the pattern explain the outcome?

Gary Nabel said that VRC is starting to look at differences in gene expression in specific cell types, and with any luck some pattern might jump out from whole PBMCs. Technology may soon provide a way to separate PBMCs by cell type. He thinks that it would be best to focus on specific cell types. There have been no plans to use systems biology approaches in HVTN 505, but he’ll discuss various possibilities with the protocol committee.

Other participants pointed to a continuing cultural conflict between basic biology, which is curious about everything, and public health, which wants results that can be applied to human health. Consortium approaches will be the best way to get these two cultures to work together.

The meeting adjourned at 12:00 noon.

Last Updated June 27, 2011