Main

Anti-HIV drugs have extended the lives of millions of infected people, but the epidemic continues its course, with high rates of new infections in regions least able to cope medically, socially and financially. For example, in some particularly hard hit areas infection rates in young women soar from less than 1% at age 15 to in excess of 50% by their mid-twenties. A vaccine will be required to ultimately conquer AIDS, but many scientists believe that this is probably decades away, if even possible. Thus far, HIV vaccine trials in humans have resulted in either no or relatively unimpressive protection despite measureable immunogenicity of administered HIV antigens1,2,3,4. Indeed, natural immunity itself is far from protective, inducing both humoral and cellular immune responses but not leading to spontaneous clearance or protecting against HIV superinfection5. Therefore, simply generating an immune response similar to what is generated in natural infection is unlikely to immunize effectively against HIV. This conclusion supports the view that our fundamental approach to HIV vaccination needs to be re-examined.

We believe that the development of an effective HIV vaccine is an achievable goal, and that there are many ‘knowable unknowns’ that can be addressed immediately to advance this effort. The goal of this Perspective is to outline progress from studies of HIV and the closely related AIDS-inducing monkey retrovirus simian immunodeficiency virus (SIV), together with insights derived from studies of other viruses, showing that some level of control of lentiviruses can be achieved. This suggests that there are immune mechanisms that, if induced by a vaccine, may prevent HIV infection or at least limit the level of viraemia in infected people. In addition to the critical value in continuing and expanding current efforts to test vaccine candidates in humans, we emphasize the importance of a long-term plan to enhance understanding of basic immunological mechanisms that may be able to prevent amplification and systemic spread from the limited initial nidus of HIV infection (see the accompanying paper6). We need better integration of novel concepts and information on new immune system genes and mechanisms derived from work in model systems into vaccine science. These advances need to be rapidly translated into the HIV field to optimize chances for effective vaccination against AIDS. Achieving this goal will be greatly facilitated by engagement of scientists from diverse but currently isolated disciplines, who, by providing new insights, can accelerate progress towards an effective vaccine.

Gaps in knowledge

Given the failure of empirical HIV vaccine approaches, understanding the nature of the immune response needed for protection is the essential missing ingredient. We still lack fundamental knowledge regarding the nature, quality and quantity of immune responses that should be induced, the ideal antigens to include, how to overcome the tremendous sequence variability engendered by the error-prone HIV reverse transcriptase, or even whether preventive vaccine strategies should focus on protection from infection or protection from disease progression.

Classically, immunization has depended on programming memory T and B cells to remember encounter with antigen. The current vaccine failures indicate that this is not sufficient, and that immunization strategies need now to be engineered with close attention to the type and differentiation state of the immune response generated, the efficacy of vaccine-induced effector mechanisms against specific vulnerable steps in the pathogenesis of HIV infection, the capacity of the induced response to maintain activity at relevant body surfaces, the influence of the virome and microbiome on vaccine responses, the fitness cost of mutations forced on HIV by specific immune responses, the capacity of vaccine-generated immune responses to avoid exhaustion, and the capacity of the different parts of the immune response generated by vaccine to synergize with one another at the relevant sites. Because HIV initially establishes infection in a very small number of cells6, appropriate pre-existing immunity has the potential to prevent infection, clear infection, or limit replication so that the set point of viraemia is low enough during chronic infection to prevent progression to AIDS and inhibit spread from one person to another. A better understanding of the molecular mechanisms responsible for immune control of acute and chronic viral infections, and immune interactions with HIV at various stages of disease (Fig. 1 and ref. 6) will be needed to direct effective vaccine strategies to meet this desperate global need.

Figure 1: HIV pathogenesis and outstanding questions related to vaccination.
figure 1

Shown are the steps in HIV pathogenesis and ‘knowable unknowns’ that, when defined, may alter how we approach vaccination against HIV. APC, antigen-presenting cell.

PowerPoint slide

Full size image

Signals that vaccination may work

There are, finally, weak signals from a human vaccine trial, and more robust findings from efforts to vaccinate against SIV, indicating that a measure of protective immunity to HIV or SIV can be induced. A recent phase III trial combining an HIV envelope protein immunogen with a recombinant canarypox-HIV vector suggested an encouraging (albeit insufficient) 30% reduction in acquisition1. In monkeys, induction of virus-specific CD8 T-cell responses does not prevent infection, but if strong and broad enough, can curtail SIV replication after homologous or heterologous virus challenge7,8. Use of simian cytomegalovirus (CMV) as a vector to continuously deliver SIV antigens limits systemic infection in some animals after intrarectal challenge with SIV9. These and similar studies show that exposure to lentivirus antigens can lead to resistance to or control of infection and provide controlled models to address substantial gaps in our knowledge of what immune mechanisms are responsible for the observed effects.

Encouragement afforded by these vaccine results is enhanced by studies of natural transmission that indicate that induction of even modest protective immunity at mucosal surfaces may tip the balance towards less efficient spread of HIV. Only one-third of infants born to infected mothers acquire infection; sexual transmission occurs after between 1 out of 100 and 1 out of 1,000 exposures (although the risk may be far higher when the transmitting partner is acutely infected)10; exposure to infected blood products does not consistently lead to infection11; commercial sex workers may remain uninfected despite extensive exposure12; and most infections are initiated by a single virus13. These data indicate that the induction of even modest protective immunity at mucosal surfaces before infection may be effective at inhibiting spread of HIV. Indeed, this may be the mechanism of apparent protection in the recent Thai vaccine trial, which limited HIV acquisition but had no effect on set-point viral load once systemic infection was established. Detailed studies of the earliest events in transmission in animal models and humans are needed to define the mechanisms that explain resistance to infection in the vast majority of HIV exposures.

Détente teaches us how to go to war

Not all HIV infections lead to AIDS, indicating that it is possible for immunity to establish a very long-term détente with lentiviruses; the mechanisms involved might provide insight into the goals for vaccination (Fig. 2). For example, about one in 300 people maintain extremely low HIV loads without treatment, some for more than 30 years14, and HIV-2 pathogenesis is similar to HIV-1 pathogenesis and yet causes disease far less frequently15. Similarly, SIV can cause high-level viraemia, infect and deplete CD4 cells, and yet not cause AIDS16. Together these data indicate that there are as-yet-unidentified viral or host factors that attenuate lentivirus pathogenesis. Identification of these may define the type of immune responses needed to reach the goal of vaccinating to prevent disease progression and spread within the population (Fig. 2). A key step in this direction would be identification of novel molecules involved in control of chronic virus infection in model systems, and rapid translation of such new findings for detailed analysis during HIV and SIV infection.

Figure 2: Approach to vaccination and special challenges of HIV vaccination.
figure 2

Shown are the steps in development of a vaccine strategy for HIV. In blue are the key issues corresponding to each step that influence our capacity to immunize against HIV.

PowerPoint slide

Full size image

Fight using all the tools immunity offers

Current evidence overwhelmingly argues for vaccines that simultaneously elicit effective CD4, CD8 and B-cell responses; however, an integrated understanding of the relationships between innate and adaptive immune responses to HIV antigens, and how this might lead to enhanced immunization, is yet to be achieved (Fig. 3). Furthermore, although it is clear that innate immune mechanisms contribute to control of HIV17, it is not clear whether it is possible to harness this response via vaccination. Indeed, the HIV vaccine field has been largely divided into B-cell approaches and CD8 T-cell approaches. There is little evidence that any vaccine achieves protection through antibody alone, or ever prevents initial infection entirely; the same holds for T-cell responses. Antibody functions optimally in the setting of simultaneous CD4 T-cell responses that provide help; CD4 and CD8 responses are probably needed to limit or eliminate HIV replication at the initial site of infection through innate and adaptive mechanisms. Similarly, T-cell vaccines largely focused on CD8 T-cell responses7,8 are unlikely to succeed without antibody and HIV-specific CD4 T-cell responses9,18. Even if we knew what we wanted to induce in a vaccine, the science of engineering specific immune responses to exceed the capacity of the natural immune response is in its infancy. A key step in this important direction will be to define relationships and potential synergies between CD4 T-cell, CD8 T-cell, antibody and B-cell responses during vaccination and in situations in which long-term viral control is achieved in infected people.

Figure 3: Immune responses to HIV and outstanding questions related to vaccination.
figure 3

Shown are the steps in the immune response to HIV. For each component of the immune response, colour-matched ‘knowable unknowns’ are specified which, when defined, may alter how we approach vaccination against HIV.

PowerPoint slide

Full size image

The case for re-embracing CD4 T cells

There is much controversy concerning the role of HIV-specific CD4 T cells in control versus promotion of HIV replication. HIV replicates best in vitro in activated CD4 T cells, there is evidence for selective infection of HIV-specific CD4 T cells by HIV19, and a correlation between SIV recrudescence and CD4 T-cell activation after depletion of CD8 T cells has been reported20. Such observations support concerns that induction of a large population of HIV-immune CD4 T cells might enhance rather than control HIV infection. This view is countered by overwhelming evidence that CD4 T cells are essential for antiviral responses in vivo including CD8 T-cell responses and effective B-cell responses in every system in which it has been studied (for example, see refs 21–23 and references therein). For example, CD8 T cells do not efficiently access mucosal sites to combat viral infection without CD4 T-cell help18, and the outcome of retroviral infection depends critically on speed and extent of virus-specific CD4 T-cell responses24. The basic rules of immunology therefore indicate that it will be impossible to generate long-lived, stable, high-level vaccine-induced immunity to HIV without inducing a protective CD4 T-cell response, but most HIV vaccine strategies have not specifically focused on inducing these responses.

Current data strongly suggest that HIV-specific CD4 T cells are beneficial rather than detrimental. The initial burst of HIV and SIV replication during acute infection does not occur in HIV- or SIV-specific CD4 T cells6, and whereas infection of these cells could conceivably add fuel to this early fire, a strong effector-memory CD4 T-cell response is associated with diminished, not enhanced, SIV replication after intrarectal challenge9. Moreover, although HIV preferentially infects HIV-specific CD4 cells, the overall magnitude of this increase is small, and the great majority of HIV-specific CD4 T cells are not infected in vivo, even in the presence of high-level viraemia19. Blockade of CD4 T-cell activation after CD8 T-cell depletion using anti-interleukin (IL)-15 antibody in SIV-infected macaques has no effect on SIV replication25, indicating that previous associations between CD4 T-cell activation and increased viraemia are correlative only20. Furthermore, CD4 T cells can provide direct antiviral effects24, and the presence of strong virus-specific CD4 T-cell responses distinguishes non-pathogenic HIV-2 infection from HIV-1 infection15. Possible support for a role for immune CD4 T cells in vaccination may be found in the Thai HIV vaccine trial results, in which decreased HIV acquisition was observed with a vaccine that induces HIV-specific CD4 T cells1. A much greater emphasis on understanding the antiviral effector properties of CD4 T cells is needed, and the role of these cells in control of HIV at mucosal sites and vaccination must be one of the highest priorities for future studies.

B cells and antibody come back into focus

Animal challenge studies indicate that a pre-existing neutralizing antibody response can prevent AIDS virus infection. Notably, a significant minority of HIV infections generate heterogeneous mixtures of polyclonal high-affinity broadly neutralizing antibodies to multiple different envelope epitopes26, giving hope that properly formulated vaccines might generate protective responses. However, it is unknown whether vaccine-induced elicitation of such antibodies can counter enormous and growing HIV quasispecies diversity and prevent initial infection on exposure. So far, vaccine-elicited antibody responses have not been protective and the form of the virus (free versus intracellular) that spreads is unknown (Fig. 1) so that epitopes exposed during natural transmission are undefined. Moreover, the envelope glycoprotein is very difficult to produce in its folded trimeric form; this may be required to provide an effective immunogen. Furthermore, because HIV spread between cells may occur by means of multiple mechanisms, including virological synapse-mediated cell adhesion coupled to viral endocytosis, it is not clear that antibody will be able to inhibit all forms of viral spread in vivo27.

As only a tiny subset of an infected person’s HIV quasispecies transmits6, the relationship between the diversity of sequences of the surface glycoprotein gp120 worldwide and appropriate vaccine epitopes is unknown, but it is possible that much of this diversity is irrelevant to immunogen design28,29. Recent results from large population screening studies have revealed the presence of novel broadly neutralizing antibodies in some infected people, and provide important new avenues for antigen design through targeting these epitopes that are subdominant or entirely unrecognized in the majority of infected persons (for example, see refs 28, 30). Reasonable levels of one of the few available broadly neutralizing monoclonal antibodies blocks low-dose intravaginal challenge with SIV using the HIV gp120 (SHIV)31,32, and encouraging recent studies show that complete protection from SIV challenge has been achieved by adeno-associated virus (AAV)-mediated continuous endogenous expression of immunoadhesins derived from neutralizing antibodies33.

Even if protective B-cell epitopes are found, the mechanisms responsible for antibody action in vivo need to be much better understood to foster induction of a polyfunctional humoral response (Fig. 3)30,34,35. Meeting this need is complicated by the lack of surrogate in vitro assays, other than virus neutralization, for protection. We may be missing epitopes that generate antibodies that protect through antibody-dependent cell-mediated cytotoxicity, complement fixation, or the capacity to block infection without neutralizing, as observed for other viruses (see refs 36–39 and references therein). Antibodies may also be important through interaction with Fc receptors on granulocytes, macrophages, dendritic cells, or B cells to alter antigen presentation, tropism, or replication in vivo36. Furthermore, B cells are more than simply factories for making antibody, as B cells unable to make antiviral antibody exert T-cell-dependent control of chronic virus infection in mice40. Efficient antibody-mediated control of Friend leukaemia retrovirus requires CD4 and CD8 T-cell responses and is regulated by specific major histocompatibility complex (MHC) alleles37,41. This may be particularly relevant to AIDS as specific MHC alleles are associated with more effective immune control of HIV infection. Unexplored interactions between B cells, antibody responses and T cells are knowable unknowns that are probably critical to HIV vaccine strategies (Fig. 3).

A major future challenge will be converting information on protective epitopes and antibody-mediated mechanisms of protection into protective vaccine antigens. Antibodies to HIV often target accessible gp120 epitopes that can mutate without fitness cost rather than subdominant epitopes that may be more conserved and thus more targetable. A fundamental question is how to generate CD4 T-helper cell activity for vaccine antibody responses to potentially protective subdominant epitopes. One approach to generating protective responses to subdominant epitopes is inoculation with viral vectors that express blocking antibody-like molecules33. Alternatively, the HIV glycoprotein may have to be engineered to optimize responses to cross-protective subdominant epitopes.

CD8 T cells are important yet insufficient

The failure of the first T-cell HIV vaccine in humans2,3 indicates that more needs to be understood about the breadth, specificity and quality of protective CD8 responses to HIV to optimize vaccine design (Fig. 3). Targeting of Gag epitopes is associated with lower viral loads than targeting Env, a finding that needs to be evaluated in SIV models to assess whether this should inform vaccine design42. Some SIV- and HIV-specific T cells are surprisingly ineffective at killing virus-infected targets43,44, and the relevance of killing to the protective activity of CD8 T cells in vivo remains unknown. Immunoregulatory networks that exhaust adaptive CD4 and CD8 responses during chronic infection probably contribute to CD8 ineffectiveness45,46,47,48. Studies using assays that more closely replicate the in vivo situation in which viral antigens are processed and presented by primary cells rather than being provided as synthetic peptides at supra-physiological doses are urgently needed, along with high-throughput single-cell analysis of T-cell phenotypes and gene expression, to define differences between effective and ineffective T-cell responses49.

Tuning the immune system to fight AIDS

To vaccinate against HIV we need to induce B and T cells that have and retain very specific differentiated properties allowing them to control replication of established infection and to function during the first critical hours of infection in the intestine or genitourinary tract (Fig. 1 and ref. 6). A major opportunity to engineer vaccine responses comes from studies in mice and non-human primates showing that the immune system is dampened during persistent virus infection, generating ‘exhausted’ T cells to prevent tissue damage (reviewed in ref. 22). The key observation for vaccine design is that manipulation of the mechanisms responsible for exhaustion at the time of first antigen exposure can fundamentally alter vaccine responses.

Exhausted cells express downregulatory receptors, lack multiple effector functions, and show weak proliferative responses. Exhaustion is best studied in T cells, but may also affect B cells50,51. For T cells, exhaustion is mediated by factors including FOXP3-positive regulatory T (Treg) cells, inhibitory receptors such as PD1 and CTLA4 and others22,48, and cytokines such as IL-10 (refs 52, 53) or transforming growth factor-β54. In mice infected with lymphocytic choriomeningitis virus (LCMV) recent data show that IL-21, perhaps made by CD4 T cells, prevents exhaustion55,56,57. Blockade of IL-10 or PD1 increases control of chronic LCMV and or SIV infection52,58,59. This is relevant to vaccine strategies as control of chronic LCMV infection is enhanced by either blockade of IL-10 during therapeutic DNA vaccination52 or blockade of PD1 during vaccinia vector vaccination60. These data support linking studies of immunoregulatory mechanisms operating during chronic infection to vaccination efforts. IL-10 and PD1 inhibit CD8 responses by independent mechanisms; simultaneous blockade of both may be advantageous53. New data show that administration of the ‘immunosuppressive’ drug rapamycin counterintuitively enhances memory CD8 T-cell differentiation61, indicating that pharmacological manipulations at the time of vaccination may enhance differentiation of induced T or B cells.

Administration of antigen in specific contexts such as via dendritic cells may induce immune responses that can enhance chronic immune control of HIV62. This striking finding indicates that similar approaches to vaccination might induce immune responses that either eliminate the initial nidus of HIV infection or limit replication and thus control set point viraemia. Similarly, one might argue to block induction of Treg cells during vaccination. However, counterintuitively, elimination of Treg cells can blunt the mucosal immune response and control of vaginal infection with herpes simplex virus, a finding potentially related to altering the balance between mucosal and lymphatic responses63. This provides a cautionary note as one may not be able to translate all mechanisms of chronic control of HIV to the vaccination effort. Another major issue is how to use adjuvants, which may, in addition to enhancing the overall immunogenicity of an antigen, be used to direct and optimize the differentiation of cells responding to vaccine antigens. Here again are knowable unknowns: it is important to determine how to engineer augmented vaccine responses via manipulation of immunoregulatory pathways during vaccination (Fig. 3).

Battling HIV in mucosal tissues

HIV delivers a near-lethal hit to the host immune system by killing a huge proportion of T cells in the gastrointestinal tract (ref. 6 and Fig. 1), and persistent viraemia may feed off of systemic immune activation associated with mucosal damage during HIV infection. Mucosal events may contribute to HIV and SIV progression, variation in vaccine responses, and variations in measurements of surrogate markers of protection. Continuous activation of mucosal HIV-immune cells may be required for protective vaccine responses given the speed of mucosal HIV pathogenesis6. The immune effector mechanisms that target HIV in the mucosa are unknown, and relationships between mucosal and systemic immunity are poorly understood, leading to uncertainty over the benefits of mucosal versus systemic vaccination. Tissue-specific immune responses have a key role in control of viral infections22,64,65, and thus differences among the vagina, penis, intestine and oropharynx might determine vaccine efficacy. Furthermore, emerging evidence from humans and mice suggests that specific mechanisms maintain local immunity in skin and genital tissues64,65. As studies of HIV or SIV immunity typically analyse peripheral blood cells, we may be working from an inaccurate view of immunity to HIV and SIV. An increased focus on the role of mucosal events in HIV vaccination is therefore needed.

Although appropriate emphasis has been placed on local responses, there is nevertheless convincing evidence of effective induction of mucosal responses to enteric murine norovirus infection and human papilloma virus after subcutaneous vaccination66,67, for mucosal protection against SIV after subcutaneous vaccination with a CMV-based vector9, and for generation of substantial mucosal responses after systemic immunization in mice and macaques68. Thus, systemic vaccination can elicit effective mucosal immunity, and systemic immunoglobulin-G (IgG) responses, rather than mucosal IgA responses, may be a reasonable vaccine goal. If lymphocytes with specific differentiated properties are essential for mucosal immunity to HIV they must be better defined to optimize systemic induction of mucosal immunity. The mucosal immune responses associated with continuous antigen and potential adjuvant effects69 of live vaccine vectors such as cytomegalovirus9 or live attenuated SIV70, and with the immunization scheme used in the Thai vaccine trial, need to be defined.

One of the key sources of data for immune correlates of HIV immunity that might be converted into vaccine strategies is the nature of immunity to HIV during chronic infection (Fig. 2). Mucosal-damage-associated translocation of intestinal contents, or immune-suppression-associated intestinal infection, may lead to systemic immune activation that fosters HIV disease progression16,71. Furthermore, SIV-infected non-progressors show fewer signs of intestinal barrier dysfunction than progressors16. Studies in mice deficient for innate immune genes also show that alterations in mucosal barrier function result in systemic activation of adaptive immunity72. Mucosal damage is therefore a new factor to be considered with regards to immune effects of HIV infection, and might well alter measurements of HIV-immune cells or antibody and thus contribute to false conclusions about the effects of HIV on immunity, and vice versa, during chronic infection.

The antigens, pathogens, or microbial products that induce systemic immune activation, and may therefore complicate assessment of correlates of HIV immunity, are yet to be defined. Translocated intestinal products may not be the entire cause of systemic immune activation. It has been presumed that standard diagnostic tests rule out a role for pathogens in HIV- and SIV-related enteropathy, but this is not tenable as metagenomic analyses reveal many novel enteric viruses in humans and primates (for example, see refs 73, 74) whose role in HIV/AIDS has not been assessed. Furthermore, monkeys harbour widely variant populations of intestinal bacteria that may influence progression to AIDS and systemic vaccine responses75. The virome and microbiome of the host have profound effects on the nature of the mucosal and systemic immune response to new pathogens, including the basal level of activation of innate immune cells and the development of T-cell subtypes in the intestinal mucosa22,39,76. Future studies of immune cell function in HIV-infected individuals and of HIV vaccine responses need to be designed while bearing in mind the potential role of the microbiome and virome in determining vaccine responses.

Exploiting evolution in a vaccine trap

HIV sequences have tremendous diversity; circulating HIV isolates exhibit up to 35% sequence variation in gp120 (ref. 77) and tremendous variability is generated in an individual after HIV passes through the transmission bottleneck6,13,78. The HIV quasispecies probably has pathogenetic properties independent of the capacity to mutate to evade immunity. For example, poliovirus studies show that a quasispecies can confer pathogenetic properties such as neurotropism and virulence during viral pathogenesis79.

The tremendous variability of HIV also exposes some weaknesses. In simultaneously infected adult identical twins, the earliest CD8 epitope escape mutations arose at the same nucleotide in both individuals, indicating that there are constraints on HIV evolution, and that some escape pathways are detrimental to viral fitness80. Population studies show predictable pathways to escape within epitopes presented by specific HLA alleles81,82. These mutations, particularly in Gag, can impair viral fitness83,84. It is perhaps not surprising that CD8 T cells targeting epitopes in Gag are associated with lower viral load than CD8 T cells targeting the more mutable gp120 (ref. 42). Part of the role of HLA B*57 in people able to control HIV replication to undetectable levels without medication (‘elite controllers’) seems to be generation of fitness-impairing epitope mutations that cannot easily be compensated83,85. Such CD8 T-cell-induced mutations associated with lower viral load can revert after transmission, providing in vivo evidence of impaired fitness86,87. The ability to define the fitness landscape of HIV now exists, and this information could have a significant impact on vaccine development.

Careful observation of HIV evolution under selective pressure in vivo also has the potential to provide key new insights into the nature of potent immune responses. Clear signatures of selection have been identified that are not accounted for by known CD4 or CD8 T-cell, or B-cell, epitopes82; known epitopes do not explain all CD4 and CD8 T-cell responses observed88. CD8 T-cell control of chronic infection with herpesviruses and polyomaviruses can involve non-classical MHC molecules, and similar responses have been reported in humans22. Thus, there are additional epitopes, and also potentially other mechanisms of innate or adaptive immune control, that await discovery.

These findings define a weakness in HIV’s most potent defence, the capacity to mutate, and indicate the potential for effective immunization via an ‘evolutionary trap’, driving HIV into a fitness dead end by targeting epitopes in fitness-critical regions of viral proteins. Experiments in both HIV and in animal models in which immune pressure on viruses can be easily controlled will be needed to develop the methods for evolutionary trap vaccination. This will require integration of computational biology, evolutionary biology and high-throughput sequencing to determine the sequence space within which a ‘fit’ virus must reside. Full-length genomes will have to be analysed to define escape mutations and linkages between these mutations and intragenic and extragenic suppressor mutations that restore fitness. Studies have already shown that immune-induced mutations in gp120 are more frequent yet less likely to confer a fitness cost to HIV84 than those in Gag that impair fitness83,84. However, even then compensatory mutations can sometimes rapidly restore viral fitness89. Combining an evolutionary trap with mosaic vaccines based on computationally derived sequences representing maximal coverage of HIV diversity is an attractive approach90.

Weapons deployed by vaccination against HIV

In the end, inducing immune responses to HIV will be effective only if potent effector mechanisms are deployed against the virus. Because viral integration and establishment of a pool of latently infected cells occurs very rapidly after transmission (Fig. 1 and ref. 6), these responses will have to be present continually or arise extremely rapidly to prevent establishment of chronic infection. Unfortunately, we still do not know what an effective immune response to HIV includes, and therefore are compromised in our quest for a vaccine (Figs 1, 2, 3). Recent focus has been on inducing polyfunctional T-cell responses as a surrogate for effective responses, but which functions are truly relevant remains unknown. Significant effort needs to be directed to defining vaccine-relevant effector mechanisms that can eliminate the initial nidus of HIV infection, prevent systemic spread from an initial localized mucosal infection, or block the establishment of latency. The prevention of latency might enhance vaccination by preventing the virus from accessing a quiescent state of infection invisible to the immune system (see below). In this area, particular emphasis needs to be placed on defining novel mechanisms as currently known effector responses do not consistently correlate with control of HIV.

Clues to potential novel immune effector mechanisms come from analysing the complex cellular factors that promote or restrict HIV replication91,92. HIV has evolved mechanisms to thwart many host proteins that limit viral replication. The importance of defining the relationship between immune evasion strategies, host restriction and immunity is shown by the recent demonstration that Rfv3, which determines the effectiveness of antibody against Friend retrovirus in mice, is encoded by Apobec3, which encodes a protein that is a homologue of human APOBEC3G and APOBEC3F proteins targeted by HIV vif93,94,95. The task we face is defining mechanisms that are not effectively blocked by HIV and which can be deployed by vaccine-generated immune cells. Cytotoxic killing of an infected cell requires recognition that occurs optimally only after HIV is integrated into the genome and proteins are expressed. Breaking the cycle of replication and latency may therefore require different or additional mechanisms, in particular those that block early events in HIV pathogenesis (Fig. 1 and ref. 6).

One potential weapon that might be used against HIV would be prevention or regulation of latency. It is possible that the capacity of HIV to enter an immunologically silent latent state in which the provirus is present but antigenic proteins are not expressed early after infection of the mucosal surface could be responsible for vaccine failures or inefficient vaccination. Thus, immune responses that prevent the establishment of latency might enhance vaccine effectiveness. Latent reservoirs of HIV and SIV are incompletely defined96, and this knowledge is critical to the vaccine effort. Reactivation of HIV from latency is regulated by multiple overlapping factors including chromatinization of the provirus, methylation, T-cell activation, NFκB and cytokines including tumour-necrosis factor-α97,98,99. Notably, the immune system can regulate the nature of latency in cells infected with herpesviruses100, and thus the nature of latency is not necessarily cell intrinsic—a potentially important clue that needs to be exploited. Latency in macrophages or other cells that may have a role in persistent infection is less well understood than in T cells. Therefore, it is not known whether vaccine-induced effector mechanisms should target HIV in macrophages or dendritic cells, in addition to T cells. A further profound advantage of defining immune and cell-intrinsic control of latency in more detail could be the development of additional animal models in which mice are adapted to HIV98, or HIV is adapted to additional primates99.

Tough choices ahead

There is hope for HIV vaccination. There are many knowable unknowns (Figs 1 and 3) that can guide this effort, and there is much scientific knowledge yet to be applied. However, there are also major scientific challenges that must be met to optimize chances of generating an effective vaccine. Although a successful vaccine might be around the corner in the next human trial, instead the path to an effective vaccine might be many years long and require novel discoveries and insights into the functioning of the immune system. Thus, the scientific issues addressed in this Perspective are best considered in the broader context of how to structure the research effort to ensure success. Because many of the basic immunological issues that need to be solved to conquer AIDS are similar to those for hepatitis C, hepatitis B, malaria, tuberculosis and cancer, enhancing basic understanding of HIV immunity will have broad impacts. Calls for a greater focus on basic science to facilitate HIV vaccine research were expressed 16 years ago101. We, and many others102, make the same call today, as the kind of effort commensurate with the problem is far from being realized.

So what are the barriers, in addition to the complexity of the immunology, to rapid progress in basic science relevant to HIV vaccination? Most important among these barriers, and most controversial because it has important practical, philosophical, political, scientific and financial implications, is the separation of most HIV-related research from research in other areas of immunology and virology.

Considering the many issues that need to be solved to give us the best long-term chance of effective HIV vaccination (for example, Figs 1 and 3), it becomes clear that not all of the answers can efficiently come from studies of HIV. The history of research on HIV immunology is filled with paradigm shifts based on studies of viruses and species commonly defined as not AIDS related. Examples include T-cell exhaustion, Treg cells, PD1, CTLA4, inhibitory cytokines, and many others. Many were first defined in mice by studying LCMV, a virus unrelated to HIV, and many have taken years to translate from murine to human studies. Studies of other human pathogens are also highly relevant. For example, systems-based studies of human vaccines that do work, such as the yellow fever virus vaccine, provide new genes and pathways, some entirely unexpected, that may contribute to highly functional CD8 T-cell and B-cell vaccine responses49. Studies of cytomegalovirus led to the development of an HIV vaccine approach effective against mucosal challenge9. There is no debate that HIV vaccine research has benefitted significantly from research in other areas, and vice versa. The question is how best to move forward to realize the great potential offered by integrating currently isolated efforts.

We need, as a scientific community, to break down barriers between the AIDS effort and other relevant areas of science to access previously untapped human and technological resources for HIV vaccination. But this needs to be done in context, such that those with expertise from other fields are able to effectively tap into the wealth of current knowledge on HIV pathogenesis. It is not enough to simply funnel more funding to current AIDS research efforts. The key point here is that there is no lack of enthusiasm among scientists outside the HIV field to get involved in HIV/AIDS related research; rather there is merely a lack of a way in. The study section system and foundation efforts should be reformed with this principle in mind. Studies in multiple systems, especially those that define novel mechanisms and pathways related to chronic viral infection, should be linked to ongoing HIV efforts and pursued with urgency proportionate to the severity of the global AIDS crisis and under the AIDS umbrella.

We propose that AIDS-related research be redefined to promote the integrated study of all mechanisms that may facilitate understanding HIV pathogenesis and immunity. The HIV/AIDS problem is too severe to allow the relevant science to be restricted by policies that separate disciplines that could have a real impact on this deadly human pathogen. Although we must, of course, study vaccine candidates in humans, we must also effectively engage scientists who can bring new perspectives to combating HIV/AIDS. A long-term plan to enhance the basic science relevant to HIV pathogenesis and immunology, broadly defined, is as essential today as it was 16 years ago101. Targeted funding, even in small amounts, specifically designed to integrate currently isolated but potentially transforming efforts, may be the best possible investment that can be made at this juncture. We must finally apply the full power of modern science to the AIDS vaccine efforts, and by defining the knowable unknowns translate this knowledge to vaccine-mediated protection against a pathogen that has already caused over 30 million deaths and shows no sign of relenting.