he HIV/AIDS pandemic is now more than 2 decades old. It is incontestably one of the most devastating infectious disease scourges in the history of humankind. Yet surprisingly, in areas outside sub-Saharan Africa there is little public understanding that we are living during a period that will be remembered by historians as characterized, in part, by one of the most destructive epidemics ever experienced by humankind, with human, political, and economic ramifications that will be felt throughout the current century, and, if we cannot devise strategies to stop it, perhaps beyond. It is also likely to be remembered as a prototype of ferocious emerging infection epidemics, either naturally occurring or deliberately induced, that will likely plague humankind unless we address the societal variables (led by population growth and global climate change) that provide the milieu in which these epidemics will arise and thrive (1).

It is now abundantly clear that no pharmacologic agent, no educational efforts directed to safe sex (regardless of how vigorously implemented), no nutritional modification will stop this epidemic. Halting the spread of HIV requires an effective vaccine. Currently, an estimated 60 million to 70 million people are either living with HIV or have died from it. Yet, that enormous number is considerably less than the toll forecast a decade or more ago (2), illustrating the inherent inaccuracies of prediction science when a comprehensive set of variables is unavailable. One of many possible explanations for the disparity between past estimates of HIV prevalence and mortality and current experience is the failure to account for natural immunity.

All past and current vaccine efforts are predicated on the assumption that a protective immune response to HIV can in fact be generated. No one contests the need for preventive and, more recently, therapeutic vaccines. Enormous intellectual, bench research, and clinical trials investment by industry and academia have been expended, with future acceleration both planned and demanded. Despite this, we currently have nothing available for general preventive use, and the HIV vaccine research community exhibits cautious optimism at best. On the positive side, this quest has expanded our understanding of human immunology by orders of magnitude, and the impediments to success are becoming clearer. Central challenges for our current vaccine development program include the facts that: 1) HIV targets CD4 lymphocytes, a central and coordinating element of the immune response; 2) HIV exists in a wide array of antigenically distinct subtypes, making the vaccine target highly diverse, with the capability to evolve rapidly through mutation and selection in response to environmental pressure; and 3) viral elements with important replication-cycle functions are frequently shielded from immune surveillance by means of secondary and tertiary structure and by posttranslational modifications. Traditional vaccine approaches--live attenuated, killed, and subunit vaccines--have all proven less than efficacious in both the nonhuman and human setting to date, although some efforts in these areas continue (3). Although the investigation of clinically relevant HIV-specific immune responses has been a constant focus of multiple laboratory investigators over time, it has never been clearer that a rational approach to vaccine design will require better understanding of protective, functional immunity.

What is the evidence for a protective immune response to HIV that supports the quest for a prophylactic vaccine? During the early years of the epidemic, there was little acceptance of the notion of natural resistance and scant evidence that the kind of resistance that would prevent infection could be generated. More attention should have been focused on the early studies of sexual partners of hemophiliacs, most of whom remained uninfected despite repeated exposure. The research community focused (quite appropriately) on virus strain, genital viral load, infectivity per contact, and ability of the virus to propagate in mucosal cells, but we should have suspected then that some of the noninfectivity related to host resistance. We will refer to this as "natural immunity or resistance" primarily to distinguish it from immunity induced by a vaccine.

Natural resistance can be thought of as falling into two general categories. The first is resistance of target cells to HIV infection due to target cell coreceptor mutations (primarily CCR5, but also CXCR4 and chemokine receptor ligand mutations). The initially described CCR5D32 mutation (32-base-pair deletion in the gene coding region) is protective in homozygous individuals but occurs infrequently and is limited primarily to Caucasians, and currently there is no evidence that this resistance can be manipulated to provide a vaccine that would protect target cells from infection (4, 5). Other mutations, particularly in the CCR5 regulatory, noncoding region, continue to be identified with variable prevalence across populations and complex, interactive effects on both transmission and disease progression (6-10).

The second category is innate and acquired immunity, with by far the most investigative emphasis being focused on the latter. [Innate immunity—first-line defense mechanisms which do not require priming with immunogens (e.g., interferon and natural killer cell responses)—will not be addressed; those interested in this growing field are referred to a recent review (11).] Some of the earliest studies evaluated the generation of humoral immunity in the form of broadly neutralizing circulating or mucosal antibody. Its potential usefulness is based, in part, on vertical and horizontal transmission studies in which neutralizing antibody in the donor (positive partner or mother) was associated with reduced transmission. Antibody responses observed to date in the natural setting or in laboratory or vaccine models have generally been strain-specific or nonprotective. However, these responses may not be comparable or relevant to neutralizing antibody in the uninfected partner. In animal models, neutralizing antibody can prevent a simian virus-HIV chimeric infection, but the amount of neutralizing antibody required was far in excess of that found in humans (12). On the basis of current evidence, it is difficult to argue for a primary role of neutralizing antibody in natural resistance to HIV (13). In addition, the passive administration of antibody is not feasible in the general, at-risk population. Maneuvers to expose epitopes protected by secondary/tertiary structure or posttranslational modifications in vaccine preparations hold some promise, but doubts regarding clinical utility remain.

The acquired immune response that fuels the majority of current HIV vaccine efforts is the CD8 cytotoxic T lymphocyte response (CTL). Studies of acute infection in animals and humans have associated upswings in HIV-specific CTL activity with relative control (or at least down modulation) of viral replication (14-16). Despite this association, it has been quite difficult to prove cause (CTL effector function) and effect (replication control). An interesting experiment in support of effective CTL function is the loss of viral replication control in animal models where CD8 lymphocytes are depleted by means of specific antibody infusions (17, 18).

A particularly powerful (and difficult) approach to better defining protective immunity to HIV is the study of natural resistance in the human epidemic. The response of the scientific community to elucidating mechanisms of HIV transmission and disease progression in human cohorts has undergone three overlapping transformations. First, there was a focus on progression of the disease and the cofactors that enhance transmission, especially the presence of certain other sexually transmitted infections. This was followed by assessments of the factors that lead to long-term nonprogression, with particular interest in viral strain differences and in host immune responses, especially those that are CD8 lymphocyte based. The third phase has been the scrutiny of those who, despite conventional or intensive exposure, remain uninfected. Multiple studies have examined discordant couples, paying special attention to sexual behavior patterns, characteristics of the HIV-positive partner (viral load, immune parameters), and the presence of sexually transmitted diseases. Increasingly, however, studies on nontransmission have focused on extraordinarily heavily exposed persons (particularly women) who have, by all currently available tests, remained HIV uninfected.

Our own studies, started in 1991, initially involved discordant couples in stable heterosexual relationships. The protocol, developed after lengthy discussions, mistakenly limited immunologic studies to the HIV-positive partner, again discounting the possibility of resistance to infection. A total of 212 discordant and, belatedly, 68 concordant couples were enrolled. This large cohort provided some expected findings (anal sex increased risk as did herpes infection in either partner) (19), some new information (Mycoplasma genitalium infection in either partner served as a promoting cofactor) (20), and one particularly interesting yet simple observation: The nontransmitters had unusually robust CD8+ responses (measured as numbers of CD8 lymphocytes (21); no functional studies were initially performed). In a later subset analysis, these CD8+ cells were found in all cases to be functionally active, able to reduce HIV multiplication in a CD8+ cell noncytolytic assay (22). This suggested the possibility that increased numbers of functional CD8+ cells reduced circulating viral load, potentially also in semen, thereby minimizing partners' exposure and risk. An inverse relationship was demonstrated between CD8+ levels and virus load, but in multiple logistic regression, the magnitude of the CD8+ count was an independent variable in likelihood of transmission. (That finding will be referred to later in regard to vaccines.) A similar study conducted by Nelson and colleagues of Thai concordant and discordant couples did not find an increase in CD8+ T lymphocytes in nontransmitters (no functional assays were performed) (23).

Our study would have ended at this point had it not been for a telephone call from a woman who heard about the study and related that she was repeatedly HIV negative by PCR despite frequent unprotected sex with a partner who had AIDS, the sexual activity continuing when he had penile ulcers and she had vaginal herpes. She felt we were obligated to explain why she was ostensibly uninfected. This, of course, made her situation similar to that of the uninfected Gambian and Nairobi sex workers (24, 25). Like those groups, we then gathered 18 highly exposed, persistently HIV-negative women and undertook a comprehensive analysis of their immune defenses. One advantage we had was the availability of 12 HIV-positive partners.

Multiple laboratories around the United States joined the effort that shall hereafter be designated The New Jersey Medical School Led Consortium (NJMSLC). The goal was to elucidate a common defense mechanism that could form the underpinning of vaccine development and testing. As it turned out, the premise of a single natural defense pattern proved incorrect; four different defense mechanisms and multiple individual patterns were demonstrated (22, 26). This should not have been surprising given the strenuous debate at conferences and in the medical literature regarding the relative value of segments of the immune system in HIV defense. Investigators tend to study areas of interest--innate immunity, a variety of CD8 lymphocyte effector functions, CD4 lymphocyte helper activity, neutralizing humoral immunity, as well as human genetic polymorphisms that have an impact on host-virus interplay--but rarely in a comprehensive fashion. All can and have made cases for their individual areas of expertise, and all have fallen short to date. The complex regulatory network underpinning the human immune response, with CD4 T helper cell function serving as both coordinator and primary HIV target, suggests that the identification of a single, critical protective effector function is unlikely. This has emerged as the single most pressing problem in HIV vaccine design and evaluation and supports the continued study of natural resistance.

Thirteen highly exposed, uninfected women demonstrated one or more of the following immune defense mechanisms (Table 1): CD8 noncytotoxic HIV-suppressive activity, CD4 proliferation to HIV antigens, CD8 lymphocyte production of macrophage inflammatory protein-1b, and CD8 HIV-specific interferon-g secretion (ELISPOT). Although these women reported frequent, repeated sexual contact with their infected partner, only five had ongoing exposures and the remainder had multiple-year interludes between exposure period and testing. Additional assays of immune defense which were generally negative, suggesting no contribution to seronegativity, included mucosal and circulating neutralizing antibody, single-cell cytokine expression, and CD4 lymphocyte susceptibility to infection. As in the larger cohort study, the HIV-infected male partners who had not yet progressed to AIDS exhibited markedly increased CD8 lymphocyte counts as a group, and all seven tested demonstrated suppression of HIV replication by CD8 lymphocytes (Table 2). Their median CD8+ counts ranged from 1446 to 3199 cells/mm³ (median of laboratory normals = 484/mm³, 99th percentile = 1075/mm³).

A review of 26 studies on naturally resistant populations revealed that only nine analyzed more than one protective immune defense, and only five could be considered reasonably comprehensive (22, 24, 26-53). (Where there were multiple reports on a single population, they were considered together as one study.) Most focused on CTL (determined by bulk or limiting dilution chromium-release assays and, more recently, by flow cytometry of tetramer-stained lymphocytes and ELISPOT). In two of the studies, CTL was evaluated by classic chromium-release assays in combination with ELISPOT; in both, the latter appeared more sensitive (26, 53).

Although it is difficult to compare studies on natural resistance because most have measured only one or two defenses and techniques vary substantially, the most frequently studied natural defenses implicated in effective responses against HIV are:

• CTL effector response
• CD4 T lymphocyte helper responses (determined by proliferative assays, interleukin-2 (IL-2) release, and more recently by ELISPOT)
• Noncytolytic CD8+ HIV suppression
• CD8+ beta chemokine production
• HIV-specific mucosal neutralizing immunoglobulin A (IgA)

There have been surprisingly few studies on circulating neutralizing IgG or IgA antibody.

In almost all the studies showing one or more specific defenses, a significant percentage of those studied fail to show that defense, and every defense thus far described has not been found in at least one competently carried out study when specifically targeted. The NJMSLC is representative. As shown in Table 1, no more than 50% showed any of the documented protective defenses, and we did not find CTL by chromium-release assay, mucosal antibody, or circulating neutralizing antibody. Additionally, 20% of the exposed unfected exhibited none of the defenses we investigated. Thus, there is not a monolithic pattern of putative defense, usually a significant percentage of any cohort do not demonstrate the defense mechanisms being studied, and there are significant problems with designating any given defense as dominant.

The argument that interruption of antigenic stimulus in commercial sex workers accounts for their late seroconversions is intriguing (54). In contrast to those findings, others, including our group, find that CTL responses persist at least several years after the last potential exposure. We found no relation between the ability to demonstrate potential defenses and time from last exposure; indeed, one woman showed an extraordinary CD4 lymphocyte proliferative response to multiple envelope antigens 13 years after her last potential exposure. That supports some hope that vaccines based on natural resistance will provide long-lasting protection.

A number of recent developments, although quite exciting, have further clarified the complex challenge of HIV prevention vaccine design. Aggressive therapy timed very early during acute infection appears to preserve critical CD4 helper lymphocyte function (in some cases), leading to a lower viral set point and ultimately to the possibility of "structured treatment interruptions" in which the host may control viral replication, at least in the short term (55-57). This demonstration of "assisted natural resistance" lends hope to the concept of control of viral replication as opposed to prevention, whether one is considering autoimmunization or therapeutic immunization.

Recent analyses of CTL specificity and avidity for antigen in both animal and human studies suggest that high avidity responses during acute infection encourage viral mutation and CTL escape (58-60). This demonstration of the plasticity of HIV in the face of traditional defense mechanisms is quite sobering. Further evaluation of the functional avidity of immune responses seems warranted with possible future incorporation into considerations for vaccine planning.

A sophisticated study of CTL specificity in the Kenyan sex worker cohort suggests that CTL-directed viral epitopes detected in heavily exposed, uninfected individuals are different from those detected once infection is established, even in those sampled before and after seroconversion (61). There may be substantial wisdom in further evaluating CTL response and the associated novel, protective viral epitopes in exposed, uninfected cohorts and in using differential epitope recognition for vaccine design.

A rapidly expanding body of knowledge is emerging regarding the heterogeneity of human HLA proteins and their role in establishing effective HIV immune responses. Antigen-presenting cells process viral proteins and present peptides, in the context of specific class I and II HLA antigens, to CD8 and CD4 T lymphocytes, respectively. Dominant CTL-directed viral epitopes in the context of specific HLA alleles have been demonstrated in individuals and can evolve over time. Individual HLA makeup varies considerably from person to person and in particular across ethnic backgrounds and geographic regions. A limited number of HLA alleles have been associated with decreased (or increased) risk for infection (or for disease progression once infection is established) (62, 63). Given the complexity of HLA background across individuals and ethnic groups, together with the geographic distribution of viral strains, the challenges are substantial for identifying important protective viral epitopes and translating that into a broadly effective vaccine approach.

Perhaps the most compelling evidence for the role of HLA genes in susceptibility/resistance to HIV-1 infection comes from the investigation of cohorts of high-risk individuals with chronic or repetitive exposure. Resistance to HIV infection in a cohort of highly exposed, persistently negative female sex workers in Thailand was associated with an increased frequency of HLA-B18 (27). In a cohort of female sex workers in Kenya, a decreased risk for HIV infection was associated with a cluster of related class I alleles—HLA-A2/6802 (64). Moreover, resistance was consistently demonstrated to be independently linked with HLA-DRB1^*01. Increased susceptibility to HIV was linked with HLA-A*2301 in the same cohort. The HLA-A2/6802 cluster/supertype was also associated with a decreased risk of perinatal transmission in mothers of the same ethnic group. These seronegative women have strong cytotoxic T cell responses to conserved HIV peptides presented by this HLA-A2/6802 supertype.

Studies with several viral systems, including HIV, suggest that there are families of closely related HLA alleles, like the cluster described above, that may present the same or highly similar peptide antigens. These have been termed HLA supertypes and the targeted viral epitopes, supertopes. Supertypes may be represented in as much as 40% to 50% of the general population and, therefore, may be desirable targets for broadly effective vaccines (65-68). Combinations will certainly be necessary, and there are many potential limitations. Population-based epitope/HLA studies have often been conducted in cohorts from developed countries with Caucasian majorities. Recent emphasis has been on generating similar comprehensive databases for communities and ethnic groups that vaccine efforts will target. Identification of HLA alleles linked with natural resistance to HIV infectivity is critical for the design of peptide-based vaccines.

Some of our best minds and consortia have understandably pushed forward with HIV vaccine design and testing efforts in the context of an absence of a correlate or correlates of protective immunity. We have been left with postulating a best guess or estimate of a desired vaccine immune response. The most recent setback was the HIV Vaccine Trials Network Study 203 (prime-boost vaccine strategy with attenuated canarypox vaccine followed by a recombinant HIV envelope protein boost) that was to serve as a steppingstone for a subsequent large efficacy trial: HVTN 501. This is a good example of a combination approach with the prime targeting CTL stimulation and the boost stimulating antibody production. HVTN 203 set an ELISPOT response threshold at a minimum of 30% of vaccinees that would be required as a correlate of immunity for the 501 Study. As this target ELISPOT response rate was not achieved in HVTN 203, the 501 Study was cancelled as currently designed. (Note that a very similar prime-boost strategy in a large efficacy trial will probably be conducted in Thailand without the correlate of immunity design.)

It seems that the best clues we have to anti-HIV protection are the natural defenses as evidenced by a broad array of nonhuman primate and human studies. The differences observed in type, quality, and quantity of different immune responses in resistant and acutely and chronically infected populations strongly suggest basing vaccine design decision-making on data from resistant and possibly acutely infected individuals, populations in which further studies and data are desperately needed.

Despite limitations with regard to strength of the evidence, we now have multiple studies that have defined a set of potentially effective defenses, and these may form the basis for designing effective vaccines. Given the uncertainty concerning the role of neutralizing antibody and striking differences in studies with regard to the presence and frequency of potential defenses, a limited or narrowly focused assessment of host immune responses is likely to result in invalid conclusions.

Because we are unable to assign accurate relative values to the catalog of HIV-specific human immune responses, the most prudent and comprehensive approach to vaccine development and testing would be to measure potential efficacy by testing the ability of the candidate vaccine to elicit each of the documented potential defenses, including but not limited to:

• CD8+ noncytotoxic HIV-1 suppression of both R5 and X4 strains
• CD4+ cell stimulation (measured by proliferative index and IL-2 and interferon-g release)
• CD8+ beta chemokine production
• CD8+ CTL by ELISPOT; tetramer staining; or flow-detected interferon-g production in response to specific stimulation
• Neutralizing antibody against envelope epitopes
• HIV-specific mucosal IgA

A candidate vaccine or, more desirably, combination vaccine design should be selected on the basis of the number of potential defenses elicited. A vaccine candidate targeting a broad array of immune defenses would be more attractive than a vaccine candidate that could elicit only one or two. The suggested broad testing would also permit combining two or even three vaccine vehicles to elicit as comprehensive an array of immune defenses as feasible. This approach also accommodates the rapid development of knowledge regarding relevant immune defenses and the parallel development of new assays.

Of course, there is no guarantee this approach will succeed, but, given the multiple different patterns of apparent defenses in those who are naturally resistant, a comprehensive standardized testing protocol is more likely to produce a vaccine that creates resistance in a larger percentage of vaccine recipients. If we do not do standardized comprehensive vaccine candidate testing, we run a huge risk, as more and more vaccine candidates are created, of having many competing vaccines, all capable of eliciting CTL and neutralizing antibody, all requiring extensive field testing, and each with the potential for not protecting at least one-third and perhaps more than two-thirds of the vaccine recipients.

Even with a comprehensive testing protocol, there will be many nagging questions, including:

• Is natural resistance based on the putative defenses demonstrated in those populations? Could it be that natural resistance is based on entirely different parameters?
• Will any single vaccine or combination protect in all geographic areas, against all strains and clades?
• Will potential protective effects be vitiated, at least in part, by the nutritional deficits that afflict at least one-third of the world's population?

Additionally, there is the issue of breakthroughs in apparent defenses, such as documented in the Kenyan sex worker population. Strength and duration of natural resistance is likely to depend on multiple variables, including genetic makeup, nutritional status, and nature of the immunogen.

What is abundantly clear is that we do not have the luxury of time. Only an effective vaccine will stop this expanding epidemic. We cannot afford multiple competing vaccines that may not address the antigenic needs of a large minority, even a large majority, of the target population. We believe any approach to developing and testing candidate vaccines other than by use of a standard, comprehensive protocol (to be modified as more studies are completed) based on natural resistance would be illogical and potentially disastrous.

References