T cell analysis in vaccination

Abstract

A major problem in the analysis of vaccine candidates is the lack of any agreed upon surrogates of efficacy, which means that for diseases that depend on a strong T cell response (HIV, TB especially) the only option is to perform an efficacy trial, involving thousands of subjects, enormous costs, and years before the results are known [1]. We also know that T cell responses are an important part of most pathogen responses, and so identifying key T cell response metrics in early vaccine trials would be generally useful. Given our ignorance of what the most important variables are, what would we like to measure and how can this be accomplished, especially given the explosion of new technologies that are available? What follows is a consideration of what should be measured, with the caveat that some of these will be more important than others.

Magnitude of the T cell response to a vaccine

There are a number of both classical and emerging ways to assess the magnitude of a T cell response. This can range from stimulating PBMCs from vaccinated subjects with whole pathogens/lysates/infected cells to antigen-specific assays such as ELISPOTs, peptide pools or tetramers. In the case of the later, highly multiplexed methods like those of Newell, where up to 00 different specificities can be assessed [2,3,4•] simultaneously with excellent sensitivity (0.01–0.001%). ELISPOT’S can also be used, and the Sette group has used them very effectively [5] but the assay is not very sensitive (>0.1% threshold) and so, may miss many responses.

Oligo-tagged tetramers/multimers or nano particles have also been used [6,4•], however since each multimers needs to be at a high enough concentration to bind cognate T cells effectively, there will be a point at which a staining solution will saturate and the proteins will precipitate. That threshold will be on the order of ~10 mgs/mL or less. Still, multiple aliquots of different tetramer collections could extend this method into the thousands, if the input samples are large enough.

Functionality and phenotype of responding T cells

It is reasonable to assume that the functionality of the T cell response has an important bearing on the degree of protection, but what is the evidence that the particular phenotype a T cell has, is effective? In Mtb infections, where the Th1* predominates, at least in the blood [7], it is worth noting that ~90% of latent carriers are protected for life. Thus it could be proposed that this is a fairly effective control mechanism, although it’s perfectly likely that other T cell types, while less numerous, might also be making key contributions. Similarly, if a robust antibody response is critical for the effective elimination of a pathogen before it can infect cells, then a Tfh bias would be preferred [8]. Here labeled tetramers or other multiplex methods give the most flexibility since they can identify pathogen-specific T cells regardless of function. Non-specific stimulation and then intracellular cytokine staining can then reveal the particular type or types of T cell with those characteristics or even that they are non-functional due to exhaustion or senescence. There is also evidence that T cells which make multiple cytokines indicate a more robust response [9,10]. Interestingly, Huang et al. [11•] found that highly expanded T cells-specific for Mtb antigens or CMV showed a senescent phenotype. Thus clonal expansion alone is no guarantee of protection. More recently, high throughput single cell methods using Next gen sequencing for the analysis of T cell responses have become increasingly common. The first of these was developed by Han et al. [12], which focused on single cell amplification and sequencing of both TCR chains and a set of cytokines and transcription factors that could designate the type of T cell that was responding. This typically yielded data on 500–1000 cells at a time and has been used extensively by my group, and also included primers that measure general transcriptional characteristics or epigenetic characteristics. Very recently, 10X Genomics has developed a single cell transcriptional system and more recently added a TCR sequencing component. These methods can generate both chains of an αβ TCR (or γδ TCRs as well) and identify on the order of 2000 genes expressed in that cell. This falls short of the 5–10k genes that are expressed in most T cells, due to the fact that they tend to have much less RNA than most cell types. This means that the method will favor the most abundant genes and then give a fairly random sample of the many that are very rare (which is most). Thus only a fraction of CD4⁺ T cells (in our experience) will ‘call out’ the CD4 transcript. This seriously hinders the efficiency of the method and so a very useful adjunct to this method is to use oligo-tagged antibodies (to CD4, among many others available) that are used to stain the cells first and then identified along with the RNAs in that cell during the amplification and sequencing steps [13,14]. In this way one can at least know what the range of gene expression might be in a given major cell type. Data of this type has also been very useful in identifying novel subtypes within the ‘classical’ T cell types and this increased cellular resolution will undoubtedly be helpful in analyzing vaccine and pathogen responses.

Specific targets

It is also reasonable to think that some potential targets of T cells might be more effective than others in controlling a pathogen. It has long been known for example, that many specific antibodies are not effective in controlling a pathogen, and may even represent a smokescreen the pathogen might use to divert the immune system. What these targets might be in terms of T cell responses has not been easy to discern, but recent progress in HIV analysis has yielded a new and potentially enabling strategy. This relies on earlier work of Chakraborty and colleagues who developed computational methods to define-specific points of vulnerability, specific sequence variants that could not coexist in HIV, presumably because the reduced viral fitness [15], this was later expanded to develop ‘fitness landscapes’ for HIV and other viruses [16]. But the recent paper of Gaiha et al. [17••] takes this further by refining this analysis and showing that such vulnerability regions are in fact targets of T cell responses in elite controllers-the small fraction of individuals infected with HIV that are able to control the virus indefinitely. Thus analyses of this type may be an excellent way to pinpoint pathogen epitopes that would make particularly effective vaccines.

TCR repertoire analysis

A new approach to the analysis of T cell responses is to take raw TCR sequence data, either bulk or single cell (with paired sequences) and to analyze it with algorithms that allow one to cluster sequences according to their peptide-MHC specificity. Early examples of this are the GLIPH algorithm, which looks for shared motifs within CDR3 sequences [18] or the TCRdist algorithm which estimates sequence similarities [19]. Using the GLIPH method on ~5700 TCR sequences enriched for Mtb specificities from 22 individuals latently infected with the bacterium identified five commonly shared sequence groups, where most of the individual sequences in a cluster shared an Mtb peptide-HLA class II specificity. This is a very active area and the promise of these small studies is that one can use this on much larger data sets that will identify hundreds or thousands of such specificity groups, to arrive at a ‘TCR-ome’ and that this may show differences in the elicited TCR repertoire between responders and non-responders to a given vaccine candidate, and allow the selection of the best candidates based on small clinical studies [1]. The promise of this approach has advanced further very recently with the development of a new version of the GLIPH algorithm (GLIPH2), capable of analyzing millions of TCR sequences at much faster speeds, and without some of the artifacts encountered earlier [20••]. This paper also demonstrates a whole genome screening method for CD4⁺ T cell epitopes for Mtb [20••]. Since this is one of the largest genomes known for a pathogen, at 4.4 MB, this could be a useful strategy for any others.

Prior exposures

Previous pathogen exposures to the other strains or even completely unrelated ones [21] can be very important in how a given individual or group of individuals may respond to a vaccine. In influenza work, this is now a major topic, as it has been evident for many years that prior flu strain exposure contributes significantly to how an individual responds to a given vaccine. This goes back to the ‘original antigenic sin’ observations of Thomas and others [22] and the recent work of Gostic [23] indicating that the predominant strain exposures in different eras could be critical to the chances of a fatal infection in later years. T cell responses are also notoriously cross reactive [21], and thus we must also consider the possibility that exposure to unrelated pathogens might play an important role in vaccine responses. Here, advances in technology provide at least a possible route to this, with the development of array-based methods to broadly screen for pathogen exposure [24,25].

Non classical T cells

A number of unique T cell types have been described that have a role in infectious disease responses, but generally have not been targeted in vaccine development, but probably should be. The best known of these non-classical TCRs are those of the γδ TCR type, which constitutes a minority of 3–20% of T cells in healthy humans, as where mice deficient in these cells are more sensitive to infectious diseases. Interesting, humans deficient in γδ T cells have yet to be described, and in evolution, organisms across a wide spectrum of vertebrate evolution that have been characterized this cell type along with αβ T cells and B cells [26]. Even in the most primitive vertebrates, such as Hagfish or Lampreys which use completely different genes for their antibody equivalents, there is a third cell type thought to be equivalent to γδ T cells [27]. Their TCRs recognize intact antigens directly, without a need for antigen processing and presentation, but their actual role in a pathogen response is generally not clear. An exception is in a murine form of Malaria (Falciparum chabaldi) where Mamedov et al. [28••] showed that a distinct species of γδ T cell arises late in the infection and blocks recurrent parasitemia through its secretion of MCSF, presumably through some action on monocytes or macrophages.

NKT cells utilize αβ TCRs, but typically have an invariant α TCR, and do not recognize their short chain fatty acid ligands in a way that makes use of TCR diversity the way that peptide-MHCs are recognized [29]. Their substrate are specific non-classical MHCs that can bind these unique ligands. They seem chiefly to recognize bacterial antigens, and so should be considered in developing those vaccines. Similarly, MAIT cells have been shown to use a similar strategy to recognize metabolites, particular a vitamin (so far) [30,31]. Single RNAseq analysis has shown that at least the NKT and MAIT cells that arise in the context of Mtb stimulation share virtually all of the same expressed genes [11•], suggesting that they are the same cell type that have been adapted to recognize different classes of ligands, and don’t require the diversity of classical αβ or γδ T cells.

Individual variation HLA and beyond

Another challenge of human vaccine development is remarkable diversity and variety of HLA alleles, estimated to be 20 000 or more [32]. This diversity is likely to be part of a pathogen escape plan where at least some individuals will be able to mount a survivable response based on having an allele or set of alleles unique suited for a particular pathogen. This is good in terms of species survival, but means that the overall response to a given vaccine may be widely variable, and so presents challenges. It is also clear that male and female responses differ, with women generally mounting more effective responses, at least in part likely due to testosterone acting to suppress inflammatory responses [33]. But human variation doesn’t end there, as there are likely many subclinical gene polymorphisms in immune genes, estimated to be 10–20% of the genome, that further broaden the response to a given pathogen/vaccine [34]. This means that even pilot studies of vaccine candidates should involve a significant number of individuals, to ensure that a vaccine candidate gives maximal protection to even those with weak immune systems. Elderly individuals are typically in this category, but some younger people might also. It is also not unreasonable to design different strength vaccines for different groups, as has already happened in influenza vaccines for example [35].