Yang, Otto O

Introduction

Historically, vaccine development has been a largely empiric process of mimicking natural responses to diseases where survival is usually associated with subsequent protective immunity. However, such empiric approaches have been unsuccessful (e.g., killed virus or recombinant envelope) or not feasible (e.g., live attenuated virus) in humans, forcing a rethinking of vaccinology. Given the poor success of natural immunity, it seems likely that a successful vaccine will need to move beyond mimicking nature and address the mechanisms of failure in the immunopathogenesis of infection. Classically, antibody responses have been excellent predictors of protective immunity for most vaccines. Unfortunately, effective humoral immunity against HIV-1 appears to be uncommon, and eliciting antibodies with antiviral activity has proved exceedingly difficult. By comparison, HIV-1-specific CD8 T lymphocytes (CTL) appear to exert a protective effect,and and are able – at least partially – to contain viral replication resulting in a delay to disease development in most infected persons. Thus, much attention has turned to producing disease-attenuating CTL responses by means of a vaccine.

Because of the requirement for antigen delivery to the HLA class I pathway to elicit CTL responses, a variety of novel vaccine approaches have been developed. Most of these have shown limited immunogenicity in humans, with the exception of recombinant Adenovirus serotype 5 (rAd5) with or without naked DNA priming. This approach has shown apparent immunogenicity by gamma interferon ELISpot assays, and has thus has been advanced to three large phase II and III trials in high-risk human populations. The highly anticipated results will be available over the next 1–3 years. If these studies do not show efficacy – such as reducing the rate of infection or reducing chronic viremia after infection – should the CTL-based approach be abandoned?

If the current rAd5 trials fail to demonstrate clinical impact, there are at least two caveats to be considered and addressed before abandoning a CTL-based vaccine. First, of necessity, the ELISpot has been the only standardized assay to assess vaccine immunogenicity due to its technical simplicity and remarkable sensitivity and precision. However, this assay has not been shown to reliably indicate immune control during chronic infection; it quantitates HIV-1-specific CTL but cannot predict the ability of those CTL to recognize HIV-1-infected cells ^[1,2]. It is a real possibility that a vaccine could elicit ELISpot-detectable CTL against HIV-1 sequences, which lack sufficient avidity to recognize HIV-1-infected cells ^[3] (especially given the emphasis on maximizing protein expression via codon optimization and other strategies). Second, the optimal targeting of CTL responses for effective antiviral immunity is unknown. The HIV-1 sequences to include in a vaccine have been the topic of considerable debate, with extensive discussions centering on the inclusion/exclusion of viral proteins such as Gag and Env. This review will focus on this second point, and propose an approach to the problem.

HIV-1 pathogenesis and Ockham's Razor

The ‘CTL paradox’ has been that HIV-1-specific CD8 CTL are found in chronically infected persons, yet infection is not cleared or sufficiently suppressed to prevent disease progression in the vast majority of such persons. Why does this response fail, when CTL can be so effective at clearing or containing other viruses? There is strong evidence for several failure mechanisms, including the following observations:(i) CD4 T lymphocyte help is lacking ^[4,5]; (ii) CTL are not properly activated/differentiated ^[6,7]; (iii) CTL have dysfunctional profiles of cytokine/enzyme production and/or proliferation ^[8–10]; (iv) CTL appear anergic and/or senescent ^[11] or prone to death ^[12]; (v) levels of virus-specific CTL wane in late disease ^[13,14]; (vi) epitope mutations can arise and allow escape from CTL recognition (reviewed in ^[15]). While this last mechanism would seem to provide a direct explanation, escape appears to occur very slowly ^[16,17] or not at all in epitope sequences during chronic infection ^[18]. This finding suggests that this escape may not be the ultimate explanation for failure of the CTL response during chronic infection. Failure thus appears to be multifactorial.

Applying Ockham's Razor to CTL failure

While CTL ineffectiveness may be multifactorial, it seems unlikely that there is a confluence of multiple independent events that coincidentally contribute to this phenomenon. Is there a single underlying process that unifies these multiple avenues of CTL dysfunction? The course of disease appears to be set during primary infection because the ensuing ‘set-point’ of viremia is highly predictive of disease progression rate; thus, an early event likely predetermines a path for long-term failure. Here I make the case that immunodominance and CTL mis-targeting during early infection sets forth a chain of events that dooms the CTL response in the long term (Fig. 1). (i) Immunodominance is determined by factors such as epitope affinity for the HLA molecule, CTL avidity for the epitope-HLA complex, and naive T cell receptor repertoire (reviewed in ^[19]). In contrast, the sequence conservation (constraint) of a given epitope is dependent on virologic factors such as protein structure or enzyme function, which are unrelated to the above host immune factors. The majority of the HIV-1 proteome is highly variable (Fig. 2), indicating considerable plasticity to tolerate sequence changes, and thus immunodominant epitopes initially targeted by CTL are likely to fall in such variable regions by simple chance. (ii) HIV-1 has high rates of replication and mutation, generating tremendous sequence variation such that every possible combination of one or two mutations is produced daily in an infected individual ^[20]. Epitope mutations therefore occur constantly in vivo, and those mutants that are unrecognized by CTL yet viable allow HIV-1 to escape. Thus the early immunodominant CTL responses are prone to rapid escape, because they target variable sequences. This scenario is supported by the strikingly common observation of rapid epitope escape mutation in acute infection patients ^[21–26]. (iii) ‘Original antigenic sin,’ a phenomenon originally described as perseveration of strain-specific antibody responses against influenza virus with inability to generate antibodies against subsequent strains ^[27,28], tends to constrain as well the re-targeting of CTL responses against a pathogen when epitopes change ^[29,30]. After escape of early immunodominant responses, those CTL decay and subdominant responses can arise ^[24–26,31], but these may be limited by ‘original antigenic sin’. This phenomenon is illustrated by reported cases of HIV-1-infected individuals who were superinfected with second strains of HIV-1; CTL targeting of epitope variants in the second strain was narrower ^[32,33]. Also consistent with this process is the observation that while CTL responses can occasionally be raised against epitope escape variants ^[31,34], this event appears to be the rare exception. By the time of chronic infection, repeated cycles of escape/decay of CTL responses and replacement by subdominant responses have led to selection of CTL recognizing relatively conserved epitopes ^[18,35], which are likely constrained from escape. (iv) During very early infection, however, tissue CD4 T lymphocytes (the major reservoir) sustain massive and un-reversed depletion within days to weeks after infection ^[36–38]. Moreover, HIV-1-specific helper cells are particularly vulnerable to direct infection and death ^[39], and are disproportionately depleted. Given early immunodominance of CTL responses against variable epitopes, ensuing escape mutation, and limited CTL retargeting due to ‘original antigenic sin’, HIV-1 evades the antiviral control of CTL during the early phase of infection and replicates at high levels. This process results in a loss of help for HIV-1-specific CTL, contributing to hypofunction ^[5]. Murine data in CD4-knockout mice have demonstrated that the genesis of the antiviral CTL response does not require helper responses, but that memory differentiation and long term antiviral function are defective when help is unavailable ^[40]. These factors in concert probably contribute to CTL dysfunction, and inability to shut down HIV-1 replication. (v) Chronic immune activation of the CD8 T lymphocyte compartment drives disease progression. Activation, as reflected by the marker CD38 on these cells, strongly predicts disease progression ^[41,42]. Because CTL proliferation is antigen driven, persisting HIV-1 replication drives abnormal CTL activation and differentiation ^[6,7]. Continued proliferation over years further induces replicative senescence of CTL, which leads to poor proliferative and functional status ^[11,43]. In turn, these processes favor viral replication and loss of help, feeding back in a vicious cycle to inexorably reduce the efficiency and number of CTL over time.

Fig. 1

Fig. 2

Thus, immunodominance against poorly conserved epitopes is a single flaw that could trigger a cascade of events leading to the multiple mechanisms for CTL failure. Indirect evidence for this hypothesis is the observation that HLA B*57 is strongly associated with better immune control of HIV-1 during early and chronic infection ^[44]. CTL responses against highly conserved B*57 epitopes are immunodominant even in early infection ^[45] as opposed to the shifting pattern of epitope recognition observed for other HLA types such as A*02; earlier targeting of conserved epitopes may be a key factor in the protective influence of B*57. Also suggesting the importance of early HIV-1 control, antiretroviral treatment even relatively late in acute infection can help preserve HIV-1-specific CD4 T helper responses and at least temporarily boost immune control of viremia ^[46,47].

Assessing the impact of CTL escape mutations: acute versus chronic infection

This unifying hypothesis for CTL failure also provides explanations for seemingly inconsistent data regarding epitope escape mutation. Human studies of epitope escape mutation have demonstrated highly variable results; this variability can be explained by the timing of studies in different stages of infection. CTL targeting differs substantially between early (pre-establishment of set-point viremia) and chronic infection (quasi-steady state viremia during the asymptomatic phase) ^[48–50]. This finding is consistent with early immunodominant CTL targeting of unconstrained viral epitopes followed by decay of those CTL after epitope mutation ^[17,24,26]. Subsequently during chronic infection, replacement by subdominant CTL responses against more conserved epitopes takes place to reach relative stability during chronic infection, reflected by set-point viremia ^[18,35]. Accordingly, escape mutations occur in many epitopes concurrently during acute infection, whereas, during chronic infection, mutations occur in single epitopes infrequently if at all (see below).

Another paradox has been that the development of specific epitope escape mutations usually has not been observed to correlate to obvious changes in viremia. This lack of correlation would be consistent with escape mutation occurring en masse during acute infection (observed as concurrent escape within several epitopes ^[21–26]). Moreover, during chronic infection, an isolated escape mutation in a single epitope can take place (among multiple non-mutating epitopes), but this event occurs infrequently due to CTL targeting of epitopes in conserved regions of the viral proteome ^{[16–18,35,51]}. Thus it would be expected that an escape mutation in any single epitope during either phase of infection would not correlate to major changes in immune control.

While some reports have suggested that escape mutation in single epitopes causes increased set-point viremia during chronic infection ^[16,52,53], such examples are rare and other studies have shown no such association ^[17], or disease progression without epitope mutation ^[54,55]. A disproportionate contribution of different epitopes to control of viremia is a possible explanation for these inconsistent findings. However, an alternative reason is that increased viral replication resulting from global CTL dysfunction (see above, and Fig. 1) allows development of escape mutations that are difficult to generate because they require multiple (compensatory) mutations ^[51,56].

In summary, widespread escape mutation in multiple epitopes simultaneously during early infection is a key process setting the course for subsequent hypofunction of CTL through early loss of HIV-1-specific helper CD4 T lymphocytes (and original antigenic sin). In contrast, although CTL have re-targeted to conserved epitopes during chronic infection, CTL dysfunction plays the dominant role for failure of immune containment; viral epitope escape mutation would have only a minor role during this phase.

Interrupting the pathogenic process

The scenario outlined in Fig. 1 suggests an early immune misstep, early immunodominance of poorly conserved epitopes, which offers a point for intervention. Subverting this pattern of immunodominance by directing initial CTL responses to target highly conserved regions of HIV-1 could interrupt the cascade of events causing CTL failure by: avoiding early escape, avoiding original antigenic sin, increasing early antiviral efficacy of CTL, preventing or reducing the early loss of helper cells, optimizing long term CTL function, and suppressing HIV-1 replication to low levels that prevent immune activation and disease (Fig. 3).

Fig. 3

What should go into an HIV-1 vaccine?

Assuming that an effective vaccine delivery system is developed to target CTL against HIV-1, what viral sequences should be delivered? There has been vigorous debate on this topic recently, almost all of which has centered on the question of which protein genes should be included in a vaccine. It has been argued that early expressed proteins (Tat, Rev, and Nef) should be included because early epitope expression allows better antiviral activity of CTL in vitro ^[57,58]. Others have argued that Env should be included for its potential to concurrently elicit neutralizing antibody responses. Still others have argued that large studies of patients show statistically significant negative associations of Gag-specific CTL responses and positive associations of Env- and Nef-specific CTL responses with viremia ^[59,60], and thus a vaccine should deliver Gag but not Env or Nef. Finally, it has been proposed that a vaccine should be inclusive of multiple proteins, to maximize breadth of CTL epitope targeting both to minimize the chance for escape as well as to maximize the chance of generating effective CTL in the absence of a clear understanding of the determinants of CTL efficacy.

Epitopes are proteins, but proteins are not epitopes

Selecting whole proteins for vaccines is a historical approach, reminiscent of strategies applied to Hepatitis B virus, Clostridium tetani, and other pathogens for which successful antibody-based vaccines have been developed. However, from the standpoint of CTL responses, this approach may not be relevant. For a given CTL, antiviral activity is determined by: (i) its efficiency in killing an infected cell ^[61,62]; (ii) the temporal relationship of this killing to viral replication ^[57,58]; and (iii) the propensity for the epitope to mutate and allow escape ^[63]. Beyond expression level, the protein context of the epitope is minimally relevant. For example, it is likely that Gag-targeting is not intrinsically more protective than Env-targeting, but that the observed statistical trends in vivo ^[59,60] are reflections that epitopes from within these proteins follow overall trends based on their parent proteins. Both Gag and Env are highly expressed proteins, but Gag is relatively conserved in sequence relative to Env. Accordingly, Gag epitopes will be more conserved than Env epitopes on average, and less prone to escape on average, explaining the association of Gag-specific CTL to reduced viremia across large numbers of infected persons. However, CTL targeting a highly conserved epitope in Env would be less prone to escape than CTL targeting a poorly conserved epitope in Gag, and therefore potentially more effective. Thus, it is important to consider CTL targeting in terms of epitope properties, rather than whole proteins, and vaccine design should not be constrained by thinking of viral proteins as antigenic units.

Subverting immunodominance

If early immunodominance against variable epitopes is the central process that leads to CTL failure, then avoiding this process with a vaccine could be the solution to promoting effective responses that delay or prevent disease. A vaccine administered before natural infection could have the opportunity to set CTL memory before exposure to whole HIV-1 and avoid natural immunodominance. Because memory-recall responses in adaptive immunity are rapid, vaccine-primed memory CTL against otherwise subdominant epitopes could respond quickly to HIV-1 challenge, thus having the edge over the usually immunodominant but yet naive CTL responses. By avoiding escape, these vaccine-primed CTL responses could exert earlier control of CD4 T lymphocyte loss, preserving the help that is required for CTL efficacy in chronic infection and avoiding the limitation of original antigenic sin (Fig. 3). Clinical evidence suggests that subdominant CTL responses can be effective in controlling infections ^[64].

How might this be implemented in vaccine design? Including whole proteins in a vaccine could recapitulate the problem of natural immunodominance, because variable regions of these proteins would also be available for CTL targeting. The solution may be to include only highly conserved regions of the virus, thus limiting the epitope choices to those falling in very constrained sequences and precluding potentially immunodominant responses against variable regions. Even relatively conserved proteins such as Gag contain variable regions and codons; omitting these from a vaccine could maximize the benefit of general Gag CTL targeting. Similarly, variable proteins such as Env contain conserved regions, offering potential targets for effective CTL responses.

Given that most of the HIV-1 proteome is variable, can this be accomplished? The few regions that are highly conserved may be too short to yield a breadth of CTL responses, and for any given HLA class I profile, there may be few epitopes. However, it is unclear that CTL breadth is crucial for immune control. For other viral infections where CTL responses are effective, these responses can be highly focused on one or two immundominant epitopes ^[65,66]. Breadth may be less important if CTL targeting is optimized, and having a few responses against conserved epitopes may suffice. Another caveat is that conserved sequences vary between HIV-1 clades. The design of a vaccine that will be immunogenic for CTL against selectively highly conserved epitopes across varying human populations in areas where HIV-1 clades differ will be a substantial challenge for implementing this strategy.

Conclusions

HIV-1 presents a challenge for vaccine development that has bested the historically successful empiric approach of mimicking natural infection. Creating an effective vaccine against HIV-1 will likely require strategies that address the pathophysiologic failure of immunity against the virus. The CTL response is an arm of immunity that appears partially effective during natural infection, and addressing the mechanism of CTL failure will be crucial for design of a vaccine intended to prevent or attenuate disease via CTL. The solution may be to avoid natural patterns of immunodominance that lead to CTL failure during acute infection, by excluding inappropriately variable sequences from the vaccine. Accomplishing this task will require novel approaches that account for the great variations of virus and host.

Note added in proof

While this article was in press, Merck and the HIV-1 Vaccine Trials Network announced the early termination of the STEP trial of a rAd5 HIV-1 vaccine, due to lack of protection measured by infection rate or setpoint viremia level. This serious setback in the quest for an effective HIV-1 vaccine underscores the importance of considering mechanisms of CTL failure in vaccine design.

Acknowledgements

The author has been supported by PHS grants AI043203 and AI051970, and would like to thank countless colleagues for sharing their ideas and discussion.

References