|Home | About | Journals | Submit | Contact Us | Français|
The objective of modern vaccine development is the safe generation of protective long-term immune memory, both prophylactic and therapeutic. Live attenuated vaccines generate potent cellular and humoral immunity [1–3], but numerous problems exist with these vaccines, ranging from production and storage issues to adverse reactions and reversion to virulence. Subunit vaccines are safer, more stable, and more amenable to mass production. However the protection they produce is frequently inferior to live attenuated vaccines and is typically confined to humoral, and not cellular immunity. Unfortunately, there are presently no subunit vaccines available clinically that are effective at eliciting cellular responses let alone cellular memory . This article will provide and overview of areas of investigation that we see as important for the development of vaccines with the capacity to induce robust and enduring cellular immune responses.
The importance of T cell immunity to vaccination is at least 2 fold. First, though neutralizing antibody is admittedly the reason for the success of the majority of vaccines thus far, this parameter is far less relevant (if not irrelevant) when considering therapeutic vaccination against established diseases such as HIV, Hepatitis C, Mycobacterium tuberculosis (mTB), or cancer. In these situations of chronic infection/disease, it is clear that potent cellular immunity is a prerequisite for vaccine efficacy . Second, even in situations where neutralizing antibody is protective, the biology of the infectious agent may make the generation of cellular immunity far more desirable than just antibody. A good example is influenza where a cellular response against antigens conserved across multiple strains would be expected to protect the host more broadly that the typical yearly vaccine-elicited neutralizing antibody responses against the highly variable HA and NA proteins. Therefore, there is a significant need for the development and implementation of new vaccine paradigms capable of priming robust CD4 and CD8+ T cell immunity.
The production of such a vaccine will require the identification not only of the appropriate antigens to serve as immunologic targets, but also the discovery of robust adjuvants, capable of performing the tasks listed above [6–9]. Following the conceptual leadership of Charlie Janeway , all those interested in the generation of adaptive immune responses from vaccination have spent the last 3 decades searching for entities that elicit innate immunity for use as vaccine adjuvants. Numerous advances have been made in the last 10 years regarding the identification of new innate signaling pathways as well as the identification of the agonists/ligands that instigate their stimulation [11–13]. These discoveries have presented the vaccine world with a veritable plethora of agonists and ligands to incorporate into rationally designed vaccine adjuants in the hopes that better days are ahead for the generation of enduring, vaccine elicited cellular responses.
Unfortunately, these advances in innate immunity have not necessarily translated into significant clinical advances in vaccines and/or vaccine adjuvants. Years of exploration have revealed that the mere generation of innate immunity does not guarantee that adaptive immunity will follow. That is, there is a distinction between innate immunity that simply elicits inflammation and innate immunity that effectively transitions into robust adaptive immunity, specifically T cell immunity. The failure to produce advances in vaccine biology commensurate with the explosion of information on innate activating pathways is due, at least in part, to a failure in addressing a number of “known unknowns” (with apologies to Donald Rumsfeld) in vaccine biology. In calling them as such, we recognize that not all of these known unknowns are universally agreed upon as known, unknown or even as unknowns worthy of knowing. Be that as it may, this review will present a short list of vaccinology considerations and the rationale behind why these are important parameters to explore for the purposes of ongoing and future vaccine design and development.
It has long been appreciated that additional signals outside of the trimolecular complex (MHC-peptide-antigen) are required for fulminate T cell activation . In the mid 1980’s, CD28 was identified, and has since been well characterized, as a required costimulatory molecule for functional T cell responses [15, 16]. Through associated kinase activity, ligation of CD28 with either CD80 or CD86 (B71/2) on antigen presenting cells results in upregulated and stabilized IL-2 mRNA expression, IL-2 production, mTOR activation, and subsequent upregulation of the pro-survival factor Bcl-xL. [17–19]. Collectively CD28 provided the necessary signal(s) along side of TCR activation for robust T cell proliferation and survival, the so-called 2 signal hypothesis of T cell activation .
However, T cells do more than just proliferate, being called upon to express a variety of effector functions ranging from the production of pro- and anti-inflammatory cytokines to lysis of antigen bearing target cells [21, 22]. In addition, after pathogen/target clearance, T cells must contract to a memory pool that will remain poised for rapid helper/effector functions upon reintroduction of cognate antigen . The two signal model of T cell activation is not sufficient for providing information that can direct either the effector or memory arms of the response, and thus a minimal “3 signal” model was proposed. By way of analogy, in the dance that is a developing T cell response, signal 1 (MHC-peptide-TCR) serves to promote selection of a capable dance partner, signal 2 (CD28) enhances this recognition and promotes a more firm embrace between the dancers, while signal 3 serves as the music telling the dancers whether they are in a tango, foxtrot, or waltz.
Immunologically, the concept of signal 3 was coined and formalized for CD8+ T cells when Mescher and colleagues proposed that cytokines of the immunological milieu played a deterministic role in the fate of CD8 T cell responses to in vitro activation cultures and adjuvant antigen priming . They and others would go on to show that type I IFN and IL-12 acting through their receptors on the surface of T cells could program both the acute T cell response as well as the kind and quality of the T cell memory response. [24–30]. In the realm of CD4 helper T cell biology this paradigm fits with the ever expanding world of helper T cell differentiation in which specific cytokines direct CD4+ T cell fate into various cytokine producing Th1, Th2, Th17, etc. T cell subsets . Molecularly speaking, the factors capable of providing “signal 3” support were universally inflammatory cytokines, typically involving the induction of STAT1 or STAT4 signaling into the responding T cells . Activation of these signaling modalities eventually culminates in regulating the expression of transcription factors (Tbet, Eomes, RORγT, GATA3) that ultimately shape the fate of the cell [25–27, 29, 30]. This has been confirmed in numerous model systems where T cells deficient in various cytokine receptors show impairments in primary and secondary responses to certain pathogen challenge or vaccination [23, 32].
While the capacity for IL-12 or type 1 I IFN to serve as signal 3 mediators is undisputed, a growing number of observations indicate that these signaling pathways can at times be a liability for the generation of immune memory . Recent studies in both mouse and human indicate a central role for STAT3-inducing cytokines in the development of T cell memory, primarily due to the capacity of STAT3, and subsequent induction of SOCS3, to limit STAT4 activation and downstream Tbet expression [33, 34]. Thus, T cell memory is achieved only through the limitation of STAT4/Tbet. Indeed, unrestricted Tbet expression, which is downstream of both STAT1 and STAT4 “signal 3” mediators, has gained a reputation for being too much of a good thing, serving to severely blunt the development of long term memory [35, 36]. As immune memory is the object of vaccination, these data question what the optimal signal 3 mediators are in a vaccine setting.
In the search for non-cytokine signaling pathways, various TNF Receptor family members appear to support signal 3 functions in responding T cells , specifically under conditions of vaccination in which the generation of memory T cells is preferred over the generation of primary effectors. Recently we published that the generation of immune memory in response to vaccination using a combined TLR/CD40 adjuvant [38, 39] is independent of all the usual cytokine signal 3 suspects, instead relying upon the TNFR family members CD27 and OX40 . In these experiments both the primary and memory CD8+ response is critically dependent upon signals through CD27 principally and residually through OX40. In contrast, the CD4 response to this vaccination is more dependent upon OX40 with a residual dependence on CD70 . TNFR signaling promotes T cell survival through NFkB activation of the pro-survival member Bcl-xL [37, 42–45]. Besides contributing to T cell survival and memory formation, van Lier and colleagues showed that CD27 was specifically required for the survival of lower affinity T cells into the memory pool . When present, these lower affinity cells resulted in T cell pool with greater cross reactivity and better immune protection to challenge with a disparate, but related pathogen. Thus TNF receptors, with their capacity to promote the survival of immunity against a more broad diversity of pathogens, may represent a form of “signal 3” that might be specifically preferred during vaccination.
Collectively, the data suggest that the notion of a “signal 3” may have a good deal of flexibility, and it is even likely that signals beyond cytokines or TNF receptors may be competent for supporting complete T cell activation and programming of memory. A true reductionist approach suggests perhaps a 4 signal model consisting of peptide/MHC and CD28 to support initial activation (signals 1 and 2), cytokine mediated expansion and differentiation matched to the inflammatory milieu (Signal 3a), and TNFR-mediated survival (signal 3b) of both effectors and long lived memory. A major area of focus for the future should be the exploration and identification of the spectrum of pathways capable of contributing to signals 3a/b and how to best target these for the purposes of vaccine development.
Given that the Dendritic Cell (DC) is saddled with the responsibility of providing the majority, if not all, of signals 1–3b described above, it is somewhat surprising that we have relatively little quantitative understanding of what the optimal antigen-presenting DC looks like. Since its discovery in the mid to late 1970s by Steinmann and Cohen [47–49], numerous studies have confirmed the critical function of DC in bridging the innate and adaptive immune system. Various subsets of DCs have been identified in both rodents and humans, each with common and unique T cell stimulatory capacities. The central position of DCs in eliciting T cell responses made it obvious that the creation of effective vaccines required harnessing the capability of DCs to prime an effective adaptive immune response. Therefore, in order to develop vaccines that produce T cell responses, it became imperative that an adjuvant must induce the maturation and activation of the appropriate DC subset.
The question remains as to which DC subset is best suited to prime a given T cell response . Numerous lines of evidence indicate that specific DC subsets in both mouse and human specialize in cross presenting exogenously provided antigens in the class I pathway [50, 52–58]. Targeting these DC subsets, either by the form of antigen or nature of the adjuvant, could optimize a vaccine toward eliciting particular kinds of T cell expansion. For example, we recently demonstrated that Langerhans cells (LC) in the skin can be induced to cross-present antigen to CD8 T cells using a TLR7 agonist-conjugate vaccination platform . In contrast, Heath and colleagues have shown that dermal CD103+ DCs, not LCs, are the dominant cross presenting DC subset during HSV challenge [60–62]. In further contrast, both CD103+ and CD103− DCs in the lung have been variably implicated as the major cross presenting DC in models as diverse as influenza challenge and apoptotic cell encounter [52, 63–66]. Therefore, the data suggest that the apparent specialization of specific DC subsets for different modes of antigen presentation may have more to do with the nature of the challenge and/or form of the antigen/adjuvant than the functional capacity of the DC.
In addition, we have little to no quantitative appreciation as to what DC “maturation” and “activation” mean at the molecular level. In the tradition of the model described above, the simplest definition of DC “activation” would be the production by the DC of all the necessary factors (antigen/MHC, CD80/86 expression, IL-12/IFN production, CD70/OX40L expression) that can deliver all 4 signals. This however does not answer the question of how much of each of these signals can be considered enough. Is a 2-fold elevation of CD86 expression above background sufficient? Or does it need to be 5 fold? Can 4 fold more signal 3 compensate for 2 fold less signal 2? And how much of, and which, signal 3 is preferred? We, and many others, have identified the expression of the TNF ligands CD70 and OX40L on the DCs as potent initiators of CD8 and CD4 T cell expansion, respectively [39–41, 52, 58, 67–76]. Given the potency of the receptors for these ligands to serve as signal 3a/b mediators in supporting T cell expansion, they are attractive targets for priming an effective adaptive immune response following vaccination. Indeed, the targeting of CD27 in the context of infection or dendritic cell immunization can even promote CD8 T cell memory in the absence of CD4 T cells . This CD8 memory phenotype/functionality would be preferred in individuals with diminished CD4 T cell counts, such as HIV+ or HepC+ individuals. While these data seem to suggest that CD27 stimulation is always a boon to the development of T cell memory, we (RMK, unpublished data) and others [46, 78–82] have observed a deleterious effect of over exuberant CD27 stimulation on the development of competent CD8 T cell memory. Substantial quantitative analysis will need to be done on the effect of specific DC activation/maturation profiles on subsequent development of immunity before we will have a clear picture of what, when, and for how long which activation markers/costimulatory molecules are required to prime protective CD4 and CD8 T cell memory.
The appropriate dose of antigen in a vaccine study is a crucial determinant of vaccine efficacy. Ours and others [38, 83–87] data reveal that small changes in the dose of antigen produces a corresponding increase in the magnitude of the T cell response as well as the level of protective immunity generated. Some older data might seem to suggest that high doses of antigen can result in the phenomenon known as high zone tolerance. However, these studies were largely focused on the effect of high levels of antigen stimulation on either B cells  or on T cell clones in vitro [89–93], and no in vivo equivalent for T cells has been observed with the exception of clonal exhaustion which is discussed in further detail below. Simply put, all recent in vivo data indicate that T cell responses are highly dose responsive… the more antigen you put in, the more T cells you get out (Figure 1). With this in mind, it comes as considerable surprise that the mechanism underlying this direct relationship between antigen dose and T cell magnitude is unclear. Indeed, if anything, the literature is somewhat contradictory to these empirical findings. CD8+ T cells, upon encounter with sufficient antigen and costimulatory signals, undergo preprogrammed rounds of proliferation regardless of the ongoing presence or absence of the antigen [84, 94–97]. If the presence or absence of antigen does not effect the extent of proliferation for an antigen specific T cell, this questions how the amount of antigen can affect the magnitude of the response? One answer to this question is that antigen dose must affect the degree to which T cells are recruited into the response. However recent bar coding studies [98, 99] have confirmed what most of us suspected from older experiments using adoptive transfer or more recent experiments using tetramer enrichment of endogenous repertoires; namely that recruitment of antigen specific T cells into most responses is close to 100%. While the solution to this apparent conundrum almost certainly lies in how increasing antigen produces subtle, but critical, variations in T cell survival, recruitment and/or proliferation, the increase in magnitude of a T cell response to larger antigen doses is still an unspoken mystery of T cell biology.
Be that as it may, the data in experimental models regarding the direct relationship between antigen dose and T cell magnitude is unyielding. Despite this, the antigen doses used in clinical vaccines are surprisingly low. The antigen doses used in most animal mouse vaccine studies are generally in the range of 1–10mg/kg. In a 20g mouse, this comes to somewhere between 20–200μg of antigen. Curiously, the same μg amounts of antigen are used in the majority of clinical vaccines (e.g. HepB vaccines use 5–20ug HepBsAg), despite the fact that humans represent a 3000 fold increase in mass and at least a 200 fold increase in surface area!  Given the degree to which we experimentally appreciate the importance of antigen dose to successful T cell responses, this raises the question as to why the antigen doses are not scaled up in the clinic to meet the size of a larger host? As far back as the late 1800s, it was observed that various metabolic parameters were similar between the same or even different species when using surface area, not mass, for scaling purposes . Therefore drugs are often dosed as a function of mg/m2 rather than mg/kg. A mathematical correction is made for the fact that these parameters do not change in straight-line proportion to surface area, but in the end, using this type of calculation for vaccine purposes serves to effectively reduce the recommended antigen dosing for higher animals. For example, the recommended calculation for an antigen with an optimal dose of a 4 mg/kg (12.06 mg/m2) in a mouse produces a scaled and corrected dose in the human of 0.32mg/kg.
Three points are noteworthy. First, the calculation above is used predominantly for more traditional drugs and therapeutics and is actually not followed at all when identifying a dose of antigen appropriate for a vaccine. If it were, then the 4mg/kg dose mentioned above would translate into injecting over 19mgs of protein into a 60kg human patient! Clearly this is ridiculously high for the purposes of generating either antibody or T cell responses. However, it may well be closer to appropriate than the ~20ug used in many current vaccines in the clinic, a dose that translates to 0.0003 mg/kg (< 0.01mg/m2) which is well below anything even remotely successful for generating T cell responses in mouse or man. Second, since the biology of secondary lymphoid tissue into which the vaccine drains is arguably more related to surface area than mass, there is a reasonable argument that antigen doses should indeed be scaled to surface area. However, not only is the average volume of an individual human lymph node larger than that of a mouse, but the local draining lymphoid tissue in a human is far more diffuse than in the mouse. Any 1cm of mouse surface area might have access to one or two lymph nodes. In contrast, the same 1cm in the human might drain to 5–20 lymph nodes, depending on the precise location. To make matters even worse for T cell vaccine enthusiasts, antigen processing and presentation are woefully inefficient in response to non-live (subunit) vectors, far less efficient than the direct recognition of antigen by specific B cells. Collectively, this means that an antigen will be heavily diluted by the larger volume of human draining lymphoid tissue as compared to the mouse, resulting in the greater likelihood that the local concentration of protein might slip below the threshold necessary for effective presentation to T cells but not below the threshold for the induction of antibody responses. Third, the majority of vaccines use seroconversion (antibody) as a primary readout of vaccine efficacy. As antibody responses are far less responsive to, and dependent upon, increases in antigen dose, the lowest dose of antigen capable of inducing antibody is utilized. Thus, while it is little wonder that smaller amounts of antigen have been deemed acceptable in the clinic, it is less well appreciated that this dose is chosen to the complete detriment of T cell responses.
Ultimately this suggest that any attempts to determine the appropriate antigen dose for eliciting T cell responses in humans should take these factors into account. While the 19mgs/dose in the example above is surely overkill, it may be an order of magnitude closer to the right dose compared to the many-orders-of-magnitude-too-low that is currently in clinical practice. If preclinical animal models have any significance, the vast majority of our clinical vaccines may be starving the potential for antigen-hungry T cell responses right out of existence.
While antigen dose has both an intuitive and historical connection to vaccine efficacy, the influence of the adjuvant dose is far less appreciated. Innate cytokines have a substantial impact on the development of adaptive immunity, in particular with respect to the differentiation of responding cells into effectors. Indeed, the 3-signal model described above incorporates the necessary function of these cytokines, originally IL-1, IL-12 and type I IFN [21, 22, 102–104]. However, an excess of a cytokine such as IL-12 has the capacity to drive responding cells into terminal differentiation, providing a robust primary effector response but a dramatically reduced magnitude of long lived immune memory . Regulators of metabolism such as mTOR control not only the source of energy within the responding T cells but are now seen to regulate the formation of immune memory in a fashion analogous to excess IL-12 . Indeed, the deleterious effects of both mTOR and IL-12 on immune memory have a common root as both promote over expression of the differentiation factor Tbet which, when left unchecked, enacts a transcriptional program that suppresses the ability of another transcription factor, Eomesodermin, to promote long lived memory [35, 36].
The upshot is that these cytokines and transcription factors operate on a goldilocks principle in which the contribution of each must be “just right” in order for the appropriate development of a long lived memory pool. As it is the responsibility of the adjuvant to induce the production and release of these inflammatory cytokines and factors, this implies that the dose of adjuvant could have important consequences on the induction, or lack thereof, of immune memory. This prediction bears out as we have observed the dose of adjuvant to have an impact on the phenotype, though not the magnitude, of the antigen specific T cell response, with the cells becoming more effector-like (hi KLRG1 expression ) as the dose of adjuvant increases (Figure 2). This suggests that the wrong dose of adjuvant might create levels of inflammation that begin to impinge on the goldilocks threshold. Thus there is reason to be mindful of the dose of adjuvant in our vaccines to be sure they are optimal for the generation of cellular immune memory. While more may be better for antigen doses (see above) the same is unlikely to be true for the adjuvant.
It has been said that there are only two things one needs to know about physics; F=ma and don’t spit into the wind. Examining the last 100 years of vaccine development reveals an equivalent two-rule mantra for vaccinology; always use an adjuvant, and always have the adjuvant form an antigen depot. While the need to incorporate the adjuvant is indisputable, the requirement for an antigen depot is highly suspect. The history behind the use of an antigen depot, originally through the use of precipitates and later though the use of various kinds of emulsions, can be traced back to the simple fact that they worked. In the early 1900’s, Alexander Glenny postulated that his precipitation of antigen using aluminum potassium sulfate resulted in better immune protection due to the precipitate forming a slowly dissipating antigen depot, thereby allowing the host an extended period of time for antigen recognition and response [107–110]. This hypothesis regarding the importance of the antigen depot largely persists to this day despite evidence to the contrary. For example, while the antigen in the precipitate does indeed persist for many weeks in vivo , the removal of antigen precipitate depots/nodules two weeks after initial vaccination has no impact on the magnitude of the antibody response[112, 113]. These and other data have served to fuel the understanding that the affect of the precipitate/emulsion on the efficacy of the vaccine has far more to do with its activation of innate immunity than any requirement for the formation of an antigen depot. However, this knowledge has not necessarily translated upwards and changed the formulation of adjuvants in the clinic. This is again mostly due to that fact that an antigen depot does not seem to have any deleterious effect on the magnitude of antibody elicited. Since the majority of vaccines in the clinic are geared toward the production of humoral immunity, this has led to the perception that, though an antigen depot may not be necessary for vaccine efficacy, it probably can’t hurt.
Recent data indicates that at least for the purposes of cellular immunity, the last point above is likely to be wrong. While T cell responses are certainly observable following vaccination with emulsion-based adjuvants [114–116], recent data indicate that the magnitude of the T cell response generated is a fraction of what it could be in the absence of the emulsion. Estaban Celis at the Lee Moffit Cancer Center and Willem Overwijk at MD Anderson have shown convincingly that the success of a soluble adjuvant formulation in promoting robust CD8+ T cell expansion and immune protection against cancer growth can be substantially reversed by incorporating the use of an emulsion ( and WO personal communication). Similar data in our hands is consistent with their observations (Figure 3). The emulsion appears to compromise the T cell response as a result of the very antigen depot that was ironically once viewed as so critical to the vaccine success. T cells generated by the vaccine recirculate and become trapped at the site of the antigen depot . In the situation of a tumor bearing host, the injection site can become the chief competitor to the tumor site in attracting tumor specific T cells, distracting the cells away from the intended target and effectively sabotaging the entire purpose of the vaccination (WO, personal communication). The solution to this problem (pun intended) may be as simple as just removing the emulsion. For saponin/MF59/QS21-based adjuvants this will be problematic as the emulsion is the adjuvant. However, TLR-based adjuvants (Oncovir, MPL, CpG) can easily be formulated without the use of an emulsion.
It is worth clarifying at this point that the deleterious effects of antigen trapped in a depot are almost certainly due to the form and site in which the antigen persists and not due to the simple persistence of antigen alone. Antigen persists long after the clearance of a number of infectious agents [119–125] and this persistence can be beneficial to the immune protection of the host. A good example is influenza where flu antigen persists in the lung draining lymphoid tissue well beyond the sterile clearance of the virus [119–121, 123]. In these circumstances, stimulation of flu-specific memory T cells by the persisting antigen within the lymphoid tissue can effect their phenotype, function [119, 123]. The subsequent trafficking of these cells into the lung parenchyma provides an enhanced degree of immune protection against reinfection, an enhancement that dissipates along with the loss of the persisting antigen over time. This type of antigen persistence must also be distinguished from the excessive antigen load observed after challenge with certain chronic viral infections such as LCMV clone 13, HepC or HIV [126–131]. It is probably not coincidence that the immune exhaustion seen in the host after these chronic infections bears striking resemblance to the loss of T cell function in a host containing an emulsion-based vaccine adjuvant (Figure 3). Collectively the data suggest the tentative conclusion that long term, lymphoid based antigen persistence has the potential for substantial immune benefit while chronic antigen depots within peripheral compartments (from either chronic viral replication or emulsion based adjuvants) is highly detrimental.
The logical questions to follow such a conclusion are i) what is the half life of antigen following subunit vaccination, ii) what is the effect of a (non-emulsion) adjuvant on that persistence, and iii) would increasing the persistence of antigen within lymphoid tissue after vaccination have the same beneficial effect as does the persistence of antigen after infectious challenge? While it is generally assumed that soluble whole antigen has a relatively short half life within the host, recent data from our lab has shown a surprising longevity for whole antigen (3+ weeks) when in a soluble formulated adjuvant and at a dose that produces a robust primary T cell response (BAT and RMK, manuscript in preparation). Our data seem to suggest that long term antigen persistence is not only inevitable after an appropriately adjuvanted, formulated and dosed subunit vaccination, but is probably beneficial as well. While a more complete understanding of the impact/necessity of antigen persistence will be tied closely to the issues of both antigen dose and adjuvant:antigen ratio, the data indicate the need to more closely monitor this “unknown” as a possible predictor of overall vaccine efficacy with respect to the generation of T cell responses.
The major contributions of vaccination to global public health have taken place in the last 50 years including the eradication of Smallpox in 1980 and the near-elimination of Polio today. There are now estimated to be over 120 vaccines in routine use worldwide, which annually save about 2.5–3 million lives, and could, with optimal use, save almost twice as many [132–134]. It is not an overstatement to say that vaccination is the most successful medical intervention known to mankind. In light of these achievements, it is humbling to acknowledge that, though our capacity to quantitate an immune response has dramatically elevated in recent years, our ability to interpret these data in a way that predicts immune protection in humans is still lacking. For example, live attenuated infectious agents (ie. Vaccinia virus, MMR) produce both humoral and cellular immunity lasting for many decades [135–138]. However, we still know relatively little about the phenotypic correlates that characterize this T cell memory. To complicate matters further, the vast majority of basic immunology data is derived from animal models, and we have yet to fully ascertain how well vaccine principles from these models (in particular the ones described in detail above) are directly applicable to the human. A good case study on this point is TGN1412, an anti-CD28 superagonist antibody, which induced robust Treg responses in mice yet led to cytokine storm and shock in humans in Phase I trials. These results underscore the fact that we do not always know how well animal model data, even sometimes in higher primates, apply to the human response and immunologic protection.
In recent years much work has been done to describe phenotypic and functional differences between T cells that persist past the primary immune response [106, 140–143]. These different memory cell subsets have been characterized phenotypically by their cell surface expression of markers, such as KLRG1, CD127, CCR7 and CD62L, into either central or effector memory subsets. Due to the self-renewing nature of central memory T cells (Tcm) and their preferential localization to lymphoid organs, as compared to the shorter life span of effector memory T cells (Tem) that preferentially localize in the peripheral tissues, the field has largely considered the generation of Tcm’s as the primary goal of vaccination . But is this phenotype universally preferred or is it specific to the infectious challenge being protected against? Several studies have suggested a more critical role for Tem [121, 123, 145, 146], or the more recently identified tissue resident memory (Trm) subset [147–150], in T cell mediated immune protection, especially with regards to infections that are largely restricted to peripheral tissues. While Tcm CD8 T cells have proliferative capacity upon reinfection [144, 151], Tem and Trm cells have the capacity to immediately produce perforin and granzyme in a lytic response [148,152–154]. This functional distinction may be responsible for the recent observations that T cell mediated immune protection against influenza may be more dependent on T cell effector/effector memory cells than Tcm . In contrast, protection against a systemic infection, such as LCMV in the mouse or Listeria monocytogenes in mouse or human, may best be mediated by Tcm cells .
Even less is known about whether T cell memory can be generated that will preferentially result in different effector responses after rechallenge. In CD4 T cell biology, both Th1-like memory and Th0-like memory cells have been identified, the former giving rise to only Th1 effectors and the latter capable of giving rise to a diversity of T cell phenotypes and functionalities [156–158]. Just how broadly this applies to other subset phenotypes is yet unknown, as is any information on the potential for pre-committed CD8+ memory T cell populations. For example, what might a memory precursor CD8+ T cell that is pre-committed to IL-17 production (Tc17) look like and what benefits might it have? The tissue specific inflammatory properties of IL-17 may well produce an environment that is optimal for fighting diseases such as cancer or other chronic infections. Indeed, recent data has shown successful tumor regression following recruitment of CD8 T cells which produce IL-17[159, 160]. If this degree of diversity can be achieved in the pool of memory cells, then perhaps we may need to tailor our vaccines to generate different memory T cell subsets depending on the infection and/or diseases against which they might protect.
The situation is further complicated by the fact that the majority of observations regarding T cell memory phenotype and immunologic protection have been made in animal models, raising the additional question as to whether any of these correlates truly apply in a human clinical setting. It is noteworthy that the original distinction of memory T cell subsets into Tem and Tcm was made for human T cells [154, 161], indicating at least some degree of connection between mouse and human data. Further, there appears to be a good correlation between the phenotype of T cell exhaustion observed in animal models with that seen in the clinic in situations of chronic viral infection and cancer [162–165]. A few inroads have been made in attempting to prospectively define immunologic correlates of protection predicated on vaccines that are already known to work effectively in the clinical setting [162, 166–168]. One in particular found that robust humoral responses to the yellow fever vaccine correlated with expression of the B cell growth factor TNFRSF17, and robust cellular responses correlated with C1qB and eukaryotic translation initiation factor 2 alpha kinase 4 expression . While the importance of these genes is intuitive once identified, it is informative that no preceding mouse model or study predicted that these genes would be the major harbingers of immune protection. Correlates of protection even more unexpected than these may well be lurking just below the surface, and a comprehensive “systems biology” analysis of both live and attenuated vaccines will be necessary before these correlates are fully appreciated.
Historically, the vaccinologist can be compared to one who looses his keys in a large parking lot but spends his time only able to look in one small corner of it for the simple reason that that is the location of the lampost. This is not meant as a metaphor for an effort in futility, but as a reminder that we can only effectively study that which we can see. More broadly put, our conclusions are only as comprehensive as our tools of analysis are at distinguishing between the knowns, the unknowns and the known unknowns. Thankfully, a number of technological developments have given us the capacity to measure T cells and antibody producing cells down to vanishingly small numbers, as well as to measure many qualitative and quantitative aspects of innate immune activation. It remains to be determined just how much of the parking lot these tools will allow us to see, or how predictive those observations will be, but the unknowns described above are good places to start shining the light.
The authors thank Catherine Haluszczak for expert lab assistance. R.M.K. is a founder of ImmuRx Inc., a vaccine company whose intellectual property is based on the combined TLR agonist/anti-CD40 immunization platform. R.M.K. is an inventor on patent applications filed by the University of Colorado and licensed by ImmuRx Inc. MAB was supported Irvington Institute-Cancer Research Institute/Eugene V. Weissman Memorial Fellowship. Remaining authors were supported by grants from the NIH (AI06877, AI066121) and the Department of Defense (W81XWH-07-1-0550). DoD support was associated with funding for the Center for Respiratory Biodefense at National Jewish Health.
The remaining authors have no conflicts to report.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.