The successes of antitumor immuno-based therapies and the application of next-generation sequencing to mutation profiling have produced insights into the specific targets of antitumor T cells. Mutated proteins have tremendous potential as targets for interventions using autologous T cells or engineered cell therapies and may serve as important correlates of efficacy for immunoregulatory interventions including immune checkpoint blockade. As mutated self, tumors present an exceptional case for host immunity, which has primarily evolved in response to foreign pathogens. Tumor Ags’ resemblance to self may limit immune recognition, but key features appear to be the same between antipathogen and antitumor responses. Determining which targets will make efficacious Ags and which responses might be elicited therapeutically are key questions for the field. Here we discuss current knowledge on antitumor specificity, the mutations that provide immunogenic targets, and how cross-reactivity and immunodominance may contribute to variation in immune responses among tumor types.

CD8 and CD4 ɑβ T cells recognize peptides in the context of class I and class II MHC, respectively. Viral immunologists have defined peptide MHC (pMHC) epitopes from numerous pathogens, prompting the emergence of a set of central principles that span infectious models. First, self-epitopes are distinguished from pathogenic epitopes at least in part by the mechanism of central tolerance, where TCRs strongly reactive to self-epitopes are deleted in the thymus by the process of negative selection (although self-reactive T cells do emerge in the periphery). Second, despite a potentially vast number of peptides that could be recognized by the T cell repertoire of the host, only a small portion of the overall viral proteome is sampled to produce the targets of host immunity. Third, individuals that share the same MHC will reliably produce similar response magnitudes to the same epitopes from the same pathogens. These consistent, structured profiles of immune response, termed immunodominance hierarchies, are regulated by numerous factors, including the extent of peptide presentation in the MHC, the kinetics of pMHC decay, and the nature of the corresponding TCR repertoire.

Because of central tolerance, only limited features of the tumor proteome should be accessible for immune recognition. When approaching the synapse between the human immune system and cancer, the importance of specificity can be viewed from both an immune- and tumor-centric perspective. Immunological specificity has been shaped evolutionarily by foreign pathogens resulting in a system normally capable of discerning self from nonself with a great deal of precision. In the context of cancer, however, the immune system is presented with an evolutionary conundrum: although safeguards such as immunological tolerance help prevent aberrant responses to self-antigens by T cells, they may also limit the diversity, repertoire, and function of tumor-reactive immune cells. This realization has resulted in an ongoing search for the holy grail of cancer: an antigenic target that is simultaneously abundant in cancerous cells and absent in normal tissues. The following sections and Table I describe the three broad categories of tumor Ags—tumor-associated Ags (TAAs), cancer-germline or cancer/testis Ags (CTAs), and tumor-specific Ags (TSAs)—and include a discussion on the specificity of these Ags and the extent to which each is shared among patients and/or specific cancer types.

Table I.
Types of tumor Ags and their advantages and disadvantages as therapeutic targets
Class of Tumor AgDescription of AgExamples of Ag TypeAdvantages of TargetingDisadvantages of Targeting
Ags with low specificity     
 Differentiation Ag Associated with proteins displaying a cell lineage-specific pattern of expression or present during specific developmental stages. CD19, MART-1, gp100, TRP2, CEA Antigenic targets shared between patients Antigenic expression on normal healthy tissues 
Development of off-the-shelf treatments Potential for on-target, off-tumor toxicity 
 Overexpressed Ag Normal cellular proteins expressed in greater abundance in cancerous cells. WT1, ERBB2, PRAME, RAGE-1, mesothelin Antigenic targets shared between patients Difficult to determine relative abundance on tumor compared with normal cells 
Development of off-the-shelf treatments Potential for on-target, off-tumor toxicity 
     
Ags with high specificity     
 Mutated Ag Gene mutations resulting in the expression of new peptides (from point mutations, altering the phase of a gene's reading frame, or chromosomal translocations). CDK4, KRAS, BRCA1/2, p53, TGF-βRII Decreased likelihood of on-target, off-tumor toxicity Antigenic targets are not shared between patients 
Potential sharing of driver/fusion mutations between patients with the same cancer type Requirement for patient-specific treatments 
 Oncogenic viral Ag Abnormal proteins expressed by cells infected with oncoviruses that can be at the origin of several types of cancers. HPV E6/E7, EBV EBNA1/LMP1/LMP2 Viral Ags shared between patients with cancers with viral etiology Relatively low frequency of cancer types with known viral etiology 
Development of off-the-shelf treatments 
 Cancer/testis (germline) Ag Expressed in testes, fetal ovaries, or trophoblasts, but absent in healthy somatic cells. Selectively expressed by specific types of cancer. MAGE, BAGE, GAGE, NY-ESO-1 Antigenic targets shared between patients with the same cancer type Are not ubiquitously expressed across cancer types 
Development of off-the-shelf treatments Potential for unanticipated on-target, off-tumor toxicity 
Class of Tumor AgDescription of AgExamples of Ag TypeAdvantages of TargetingDisadvantages of Targeting
Ags with low specificity     
 Differentiation Ag Associated with proteins displaying a cell lineage-specific pattern of expression or present during specific developmental stages. CD19, MART-1, gp100, TRP2, CEA Antigenic targets shared between patients Antigenic expression on normal healthy tissues 
Development of off-the-shelf treatments Potential for on-target, off-tumor toxicity 
 Overexpressed Ag Normal cellular proteins expressed in greater abundance in cancerous cells. WT1, ERBB2, PRAME, RAGE-1, mesothelin Antigenic targets shared between patients Difficult to determine relative abundance on tumor compared with normal cells 
Development of off-the-shelf treatments Potential for on-target, off-tumor toxicity 
     
Ags with high specificity     
 Mutated Ag Gene mutations resulting in the expression of new peptides (from point mutations, altering the phase of a gene's reading frame, or chromosomal translocations). CDK4, KRAS, BRCA1/2, p53, TGF-βRII Decreased likelihood of on-target, off-tumor toxicity Antigenic targets are not shared between patients 
Potential sharing of driver/fusion mutations between patients with the same cancer type Requirement for patient-specific treatments 
 Oncogenic viral Ag Abnormal proteins expressed by cells infected with oncoviruses that can be at the origin of several types of cancers. HPV E6/E7, EBV EBNA1/LMP1/LMP2 Viral Ags shared between patients with cancers with viral etiology Relatively low frequency of cancer types with known viral etiology 
Development of off-the-shelf treatments 
 Cancer/testis (germline) Ag Expressed in testes, fetal ovaries, or trophoblasts, but absent in healthy somatic cells. Selectively expressed by specific types of cancer. MAGE, BAGE, GAGE, NY-ESO-1 Antigenic targets shared between patients with the same cancer type Are not ubiquitously expressed across cancer types 
Development of off-the-shelf treatments Potential for unanticipated on-target, off-tumor toxicity 

The spectrum of tumor Ags, including those with low and high specificity for the tumor, is described, with specific examples of each type and the features that contribute to their therapeutic efficacy.

HPV, human papillomavirus.

Tumor-associated Ags.

Of all the tumor Ags studied to date, among the best characterized and earliest targets for cancer vaccine approaches are the TAAs. TAAs are normal host proteins that demonstrate distinct expression profiles between host and tumor cells. In general, the dysregulation of gene pathways as a result of mutations within the tumor cells results in the atypical expression of unmutated proteins that would otherwise be expressed at relatively lower levels, or not at all, in normal cells of that tissue type in its current developmental state. Specifically, TAAs are composed of what have been termed differentiation Ags, which are proteins that are shared between the tumor and the normal tissue of origin but distinct from other tissues, and overexpressed Ags, which are aberrantly expressed normal proteins that provide a growth and/or survival advantage to the tumors.

Differentiation Ags.

Much of the interest in differentiation Ags as potential targets of the immune system and therapeutic approaches stems from studies on melanoma, where researchers have documented spontaneous T cell responses against peptides derived from GP100 (14), Melan-A/MART-1 (5, 6), and tyrosinase (7, 8). Although these responses have been observed in patients with melanoma due to the Ag expression patterns on melanoma cells, the potential to treat a wide range of patients continues to drive new research focused on specifically targeting cancerous cells that aberrantly express differentiation Ags. CD19 is another such differentiation Ag, in this case found on normal and malignant B cells, that can be targeted in patients with acute lymphoblastic leukemia and other B cell tumors (reviewed in Ref. 9). Recently, adoptive transfer of anti-CD19 chimeric Ag receptor (CAR) T cells in patients with relapsed and refractory acute lymphoblastic leukemia has shown high rates (up to 90%) of complete remission in part due to the strength and antitumor activity of this immunotherapy (9). Despite the demonstrable sensitivity of anti-CD19 therapies in clearing B cell malignancies, the specificity of such approaches is expected to include noncancerous cells as well due to the near ubiquity of CD19 expression among B cells throughout their development. Many promising differentiation Ags may therefore exhibit off-target antigenicity, the consequences of which can range from temporary loss of inessential cells to the permanent destruction of vital tissues that may require secondary treatment or ongoing supplementation. Therefore, the most ideal differentiation Ags are those derived from proteins that would normally only be expressed during early ontogeny.

Overexpressed Ags.

Overexpressed Ags are another class of TAAs that have been shown to play a role in driving the malignant phenotype of many tumors. In leukemic cells, Wilms tumor 1 is commonly overexpressed and helps drive the oncogenic process (10, 11). In several epithelial tumors, such as breast and ovarian cancer, overexpression of ERBB2 (HER2/NEU) is typically associated with poor prognoses, but this protein may also serve as a potential immunotherapy target because of its increased expression on the surface of cancerous cells that exhibit heightened proliferation (12, 13).

Much of the early interest in targeting TAAs is owed to the sharing of Ags among patients and across various types of cancer, which allows for broadly administrable, off-the-shelf treatments. However, a major limitation in targeting TAAs results from their similarity to self-peptides, which greatly limits endogenous T cell responses due to central tolerance or lower TCR–pMHC binding affinity compared with those associated with recognition of foreign Ags (14). Findings from several recent studies also caution against using adoptive transfer or receptor engineering approaches against TAAs, as previous efforts have resulted in unexpected on-target, off-tumor toxicities, as described in the unfortunate case of a patient with metastatic colon cancer who was treated with CAR T cells against ERBB2 and died as a result of off-tumor effects acting on lung epithelial cells that expressed low levels of ERBB2 (15).

Cancer/testis Ags.

Van der Bruggen et al. (16) identified the first human tumor Ag, MAGE-1, a CTA recognized by endogenous T cells. CTAs are expressed in testes, fetal ovaries, or trophoblasts, but are otherwise absent in healthy somatic cells. Since this study in the early 1990s (16), an increased focus on specificity, and by extension the realization that a more in-depth characterization of Ag localization would be useful, resulted in increased efforts to target CTAs. The attractiveness in targeting CTAs can, in part, be attributed to: 1) their disrupted gene regulation in various tumor types, 2) their limited expression in normal tissues, 3) their lack of presentation in germline and trophoblastic cells, which do not display MHC class I molecules on their surface, and 4) their immunogenic potential. Although vaccine and adoptive cell therapies targeting CTAs have shown promise in some clinical trials with NY-ESO-1 (17) and MAGE (18), a recent attempt to use adoptive cell therapies to target MAGE-A3 led to severe toxicities and even death as a result of unanticipated expression of this gene in the brain (18).

Tumor-specific Ags.

TSAs generally arise from tumor-specific mutations, which result in the exclusive expression of neoantigens in tumors and, by definition, their absence in normal cells. However, another source of TSAs that have been shown to elicit tumor recognition by T cells are the viral proteins expressed by cells infected with oncoviruses such as human papillomavirus (19, 20) and EBV (21). In either case, the cancer-restricted expression intrinsic to TSAs theoretically plays into the immune system’s strength of distinguishing self from nonself by bypassing the mechanisms that would otherwise eliminate the tumor-reactive T cells that bind self-pMHC with high affinity. Support for the notion that neoantigens are sufficiently dissimilar to self to be targetable by the immune system was shown in two seminal human studies in 2005 (22, 23). In a study by Robbins and colleagues (23), ex vivo-expanded tumor infiltrating lymphocytes (TILs) were adoptively transferred into a patient with melanoma. In addition to the complete regression of the tumor, the patient also exhibited persistent T cell populations that recognized mutated GAS7 and GAPDH. Wölfel and colleagues (22), on the other hand, used cDNA library screens to show that autologous T cells (using clonal T cells and mixed lymphocyte tumor cell cultures) had greater responses against neoantigens than TAAs in human melanoma. This early work, in addition to the recent advancements in next-generation sequencing applications and epitope prediction algorithms, has resulted in a major shift toward the utilization of personalized immunotherapies to selectively target TSAs.

The efforts to exploit TSAs as therapeutic targets and to characterize responses in patients undergoing immune checkpoint blockade (ICB) therapy has led to in-depth mapping of TSA-specific T cells in a number of studies. One finding that has consistently emerged, particularly from studies of melanoma, is that relatively few putative neoepitopes generate a detectable T cell response (24). “Hit rates” of finding a tetramer positive or peptide-reactive T cell population in patients have been as low as 0.5–2% of screened Ags. For example, in one study a patient had 249 nonsynonymous mutations in a melanoma, of which 126 were predicted to bind to HLA-A*0201. Screening these neoantigens against the patient’s own TILs resulted in a response to 2 of the 126 putative neoantigens. However, T cell responses were generated to a much larger proportion of these peptides in in vitro cultures of PBMCs from healthy donors (25). In another study on melanoma, researchers generated 75 tetramers corresponding to potential epitopes mapped to the patient’s tumor, yet only one T cell response was detected (26).

There are multiple potential explanations as to why so few of the potential neoepitopes in a tumor elicit a T cell response, including that the predicted neoepitopes may not actually be processed and presented on the tumor. Some studies have addressed this by experimentally validating that the predicted peptides can bind the predicted HLA molecule (27). In a study screening a patient with chronic lymphocytic leukemia with this method, where 18 candidate neoepitope peptides were experimentally confirmed to bind HLA, only one neoepitope elicited a detectable response. Beyond an “overestimate” of available neoepitopes, it is also possible that nonresponsiveness to particular neoepitopes may result from constraints on the available repertoire that arise due to similarity to self (although, as discussed below, this seems likely to have a modest effect on limiting responses). An implication of this low response rate, though, is the concern that low mutation burden tumors, including many pediatric tumors, may not generate enough TSA and TAA to be effectively targeted by endogenous T cell responses.

The generation of tumor neoantigens is a direct result of the genomic instability that gives rise to cancer cells, where the accumulation of mutations and genomic rearrangements within a cell can disrupt important gene pathways (e.g., those that prevent cell death, limit cellular division, or cause further genomic instability) by interfering with the normal expression or functionality of genes integral to such processes (28). Of particular relevance to tumor-specific antigenicity are the subset of these somatic mutations and rearrangements that ultimately result in the synthesis of mutated and/or chimeric amino acid chains, parts of which can subsequently be processed for presentation in the MHC. Importantly, the genomic changes underlying tumor neoantigens can be true drivers of the cancer phenotype or simply tumor-specific genetic hitchhikers of those drivers, with Ags in the latter category potentially more susceptible to immune escape.

The typical mutation load underlying cancer cells varies widely, from as few as 0.8 to over 47 coding mutations per megabase (estimated median), with some tumor mutation burdens projected at over 1200 per megabase (29). These variations in mutation load have been well studied in relation to patient age (e.g., (29)) and cancer/tissue type, with adult skin and lung cancers among those typically characterized by the most mutations and leukemias among those characterized by the fewest (2931). The number of novel peptides resulting from tumor mutations likewise varies in proportion to total mutation burden, and the vast majority of studies aimed at identifying immune response to cancer-derived neoepitopes have as a result focused on those cancers that average among the highest of mutation burdens. From those studies, on average fewer than 2% of the mutations investigated in these highly mutated tumors have been shown to elicit endogenous T cell responses (25, 26, 32), such that cancer neoantigens are often thought to result from a probabilistic process in which a greater number of mutations result in a greater likelihood of generating an immunogenic neoantigen. The general acceptance of this hypothesis within the field of immuno-oncology has proven particularly discouraging for the prospects of using immunotherapies broadly in cancer treatment, as many adult tumors and most pediatric cancers exhibit relatively few somatic mutations. A recent study provides further insight into the tumor mutational landscape by identifying HLA alleles predicted to be good presenters of particular tumor driver mutations. A subsequent statistical analysis indicated those tumor drivers were unlikely to occur within the context of those particular HLA haplotypes (33). This analysis and another report found that when such tumors did emerge in an individual carrying an HLA allele that could present the corresponding driver mutation, the tumors were often associated with allelic loss of the relevant HLA gene (34). Regardless, the discovery of endogenous T cell responses to tumor-specific neoepitopes provides both a targeted list of Ags for patient-specific vaccination and the opportunity to identify receptors that may prove integral to the success of T cell engineering therapeutics.

Whereas vaccination and TCR engineering approaches require the direct identification of the targeted TSAs, a separate class of anticancer immunotherapy, ICB, is not bound by these same constraints. Remarkable clinical activity against a variety of tumor types has been attributed to the ability of several ICB therapies (mostly targeting CTLA-4 and PD-1) to enhance T cell activity against cancer neoantigens (3537). Although immunotherapeutic approaches using ICB have not deliberately focused on the antigenic targets recognized by a patient’s T cells, data from clinical trials on patients with melanoma (38, 39) and non-small cell lung carcinoma (4042) suggest that mutational load and neoantigen abundance is positively correlated with objective response rates in patients treated with ICB (43). Although these findings provide indisputable evidence that the immune system recognizes and targets neoantigens in cancers with high rates of somatic mutation, this does not preclude the possibility that similar (or perhaps even more favorable) responses could be obtained in tumors with lower mutational burdens. In a recent study, Munson et al. (44) found that the extent of CD8 T cell infiltration into tumors and TCR sharing across patients with breast cancer correlated with improved survival. Additionally, TCR sharing was readily detected between patients with breast cancer (in both TILs and blood), whereas the same TCRs were only found sparingly in control blood from cancer-free donors. These findings suggest that the shared TCRs in cancer-bearing patients are tumor-specific and may recognize shared Ags among patients with breast cancer. It is important to note that the authors did not conclusively rule out the possibility that the shared TCRs could also be recognizing viral or environmental Ags shared among patients; however, the idea that these shared TCRs were specific for tumor Ags was further supported by data showing that, in comparison with T cells from tumor-free lymph nodes, T cells from tumor-involved lymph nodes were 4-fold more likely to contain TCRs also present in the primary tumor TILs. Further studies using reconstructed TCRs would be required to demonstrate specificity of the TCRs for tumor Ags, but regardless, these important findings will likely reinvigorate efforts to identify and prioritize therapeutic treatments that can be used to exploit the immune system and selectively target tumor Ags across various types of cancer.

As described previously, cancer arises due to an accumulation of genetic alterations, which leads to the production and processing of mutant proteins (neoantigens) that are otherwise absent from host cells (28, 30). From a purely theoretical perspective, the upper limit of expressed neoantigens could be approximated from the number of somatic mutations and genome rearrangements present in a specific tumor type. However, it is well known from viral models that this number would grossly overestimate the actual number of expressed and presented Ags, as the processing pathways involved with MHC class I Ag presentation greatly diminishes the number of potential antigenic peptides (45). Although peptide processing contributes to the narrowing pool of presented peptides, the greatest factor affecting the abundance of pMHC complexes and influencing the immunogenicity of the resultant epitopes comes from insufficient peptide-MHC binding affinities, which results in unstable pMHC complexes and significantly limits the expression of the corresponding peptides on the cell surface. A recent paper by Abelin et al. (46) used a monoallelic cell expression system and mass spectrometry of eluted peptides in an attempt to more accurately profile the HLA peptidome while accounting for the role of expression and peptide-MHC binding in peptide presentation. Although this approach accurately reflects the peptides presented by single HLA allele-transduced B cells, future studies and analyses would be necessary to determine whether these findings are generalizable to primary tumor cells that express multiple HLA alleles. Furthermore, in order for these peptides to induce an immune response, T cells must contain a cognate TCR capable of binding the pMHC complex with sufficient avidity to induce an effective T cell response, which results in a polyclonal pool of CD8+ T cells recognizing immunodominant and subdominant Ags. This phenomenon is known as immunodominance (45), and is likely another mechanism limiting the breadth of targets that can be recognized by the immune system. Immunodominance is the structured response hierarchy among epitope-specific T cell populations targeting subsets of epitopes from an antigenic mix. This has been studied frequently in the context of pathogenic infections, where a small subset of potential epitopes generates cognate T cell responses and those responses have reliable magnitude hierarchies. Thus, infections with the same pathogen in animals sharing the same MHC haplotype will generally produce the same response profile.

The mechanisms of immunodominance are not yet fully elucidated. Several factors have been proposed to contribute, including the number of pMHC complexes presented on the cell surface, the precursor frequency of the responding T cell population, the binding affinity and off-rate of peptide for MHC, and the avidity of the pMHC–TCR interaction (4749). However, some of these associations have not held up with additional data, such as the contribution of precursor frequency, which was found not to be correlated with response size in the influenza model system (50).

Immunodominance could function in the tumor environment by focusing the response on a small subset of presented neoepitopes or TAA-derived epitopes. A study of hepatocellular carcinoma TAA responses found clear evidence of immunodominance across multiple patients, with a small subset of expressed TAA being targeted by each patient, although there did not seem to be consistency in which Ags were preferentially targeted among patients (51). As noted, several factors could ultimately influence the presentation of a particular epitope in a given patient, including protein expression variation, processing efficiency, and HLA polymorphisms. The breadth of the response was associated with improved progression-free survival and proved somewhat analogous to observations in hepatitis C virus where CD8 T cell response breadth has been correlated with improved viral control (52). A strong hierarchy has also been observed in at least one melanoma patient’s responses to the NY-ESO-1 TAA (53). Although these findings show clear instances of immunodominance with regard to the antitumor response by CD8 T cells, the underlying mechanisms driving these Ag hierarchies are unclear. Further studies are needed to address whether specific factors associated with immunodominance disproportionately contribute to shaping the immune repertoire as it is currently unclear whether the relative level of expression between tumor Ags, the rapidity by which competing tumor Ags induce an immune response, or some other factor plays the greatest role in shaping the repertoire.

The potential for immunodominance to regulate antitumor responses has been appreciated for some time and raises important considerations for immune therapies and cancer vaccination (54, 55). Indeed, variation in immunodominance among individuals with distinct HLA haplotypes contributed to differential outcomes after NY-ESO-1 vaccination (56). As efforts continue to determine the key predictors of immunodominant responses, these parameters could be incorporated into the tumor Ag computational pipelines to improve their accuracy and utility (Fig. 1). If the apparent low response rate to the TSA and TAA landscape in tumors is in part due to immunodominant focusing, one consequence of this might be that even tumors with low mutation burdens will elicit useful responses. That is, a larger proportion of the available neoepitopes will be targeted in these tumors, because the limitation on TSA- and TAA-driven responses is driven not by a lack of antigenicity but rather by the immune system’s immunodominant focusing. Future studies exploring responses in low mutation burden tumors should address this hypothesis (57).

FIGURE 1.

Features that determine antitumor T cell reactivity. Two major types of Ags, tumor-associated and tumor-specific, can be recognized by endogenous T cell responses. The ability for epitopes derived from these Ags to be detected by responding T cells is modulated by the host’s HLA type and the epitope’s processing and presentation efficiency. Intratumoral heterogeneity may also allow individual tumor cells to escape recognition. On the T cell side, immunodominance hierarchies can be generated leading to an individual Ag being the major target of the response. Additionally, holes in the TCR repertoire and T cell tolerization and exhaustion can limit response efficacy.

FIGURE 1.

Features that determine antitumor T cell reactivity. Two major types of Ags, tumor-associated and tumor-specific, can be recognized by endogenous T cell responses. The ability for epitopes derived from these Ags to be detected by responding T cells is modulated by the host’s HLA type and the epitope’s processing and presentation efficiency. Intratumoral heterogeneity may also allow individual tumor cells to escape recognition. On the T cell side, immunodominance hierarchies can be generated leading to an individual Ag being the major target of the response. Additionally, holes in the TCR repertoire and T cell tolerization and exhaustion can limit response efficacy.

Close modal

As discussed earlier, TSAs may provide the most promising targets for tumor clearance while limiting undesirable effects on normal tissue. TSA-focused strategies involve the elicitation or in vitro generation of pools of T cells that have been tuned against these defined TSAs by vaccination or TCR transduction. Notably, these Ags are generally extremely similar to self, with a single mutation underlying the majority of characterized TSAs. The exceptions are the rarer mutations generated by insertion, deletion, or fusion, which have the potential to create epitopes that differ from the “parent” self-peptide(s) by multiple consecutive amino acids.

Due to the special nature of the relationship between TSA-derived epitopes and self, there are two concerns that arise when considering them as targets: repertoire limitations and cross-reactivity. Because of negative selection, TCRs directly responsive to self should be underrepresented or entirely missing from the peripheral repertoire. Although TSAs are by definition not identical to self, their proximity may constrain the potential responding repertoire if most TSA-reactive TCRs would also have been reactive against the parent self-epitope. However, a series of recent studies have suggested that the restrictions on the repertoire by negative selection may be somewhat limited. One study focusing on human T cell responses found that T cells against known self-antigens, such as SMCY in males, could still be found in the periphery. Although these cells were present at lower levels than in females and in a tolerogenic state in males, they could be reactivated with higher doses of Ag (58). In the same study, a single epitope was mutated at each residue that would permit change without preventing peptide-MHC binding, and each mutation generated an epitope that reliably bound T cells. The interpretation of these results is that there were few absolute holes in the repertoire as conventionally understood and that clonal deletion only “pruned” self-reactive specificities. Thus, even if an epitope is proximal to self, there will likely still be T cells capable of recognizing it.

TSA-derived epitopes generated by single nucleotide variants are analogous to the situation that arises during an infection when an epitope is mutated at a single residue. This has been well studied in the context of viral infections, where natural selection can favor single nonsynonymous mutations in the viral genome that facilitate immune escape. Frequently, these mutations can significantly limit or even fully evade the immune repertoires generated against the original epitope, for instance allowing pathogen persistence in the case of HIV and hepatitis C virus (59). A TSA generated by such a variant may similarly not share significant overlap in reactive immune repertoires with T cells that would target the unmutated self-peptide. Yet, various neoepitope prediction pipelines have incorporated distance-to-self as a factor in predicting the potential immunogenicity of a TSA (60, 61). At least two factors can be considered for this assessment: 1) whether the mutation creates a variation that will generate a potential cleavage site for peptide processing, and 2) whether the potential epitope is sufficiently distinct biochemically that the responding TCRs are unlikely to cross-react with self. The latter factor is currently somewhat difficult to model, given our limited ability to predict peptide–MHC–TCR interactions at a structural level, but highly divergent amino acid residues are generally more likely to create distinct epitope surfaces.

Further demonstration of the repertoire’s ability to mount anti-TSA responses comes from vaccination studies utilizing TSA targets in melanoma. In one study, healthy HLA-matched donors were used to elicit responses against tumor Ags from melanoma patients (25). Many of these were private mutations, and the healthy donors could elicit robust responses to these mutated epitopes, although it should be noted that the method used to elicit the healthy donor responses (long-term in vitro stimulation) was not applied to PBMCs from the patients. More recently, two groups have successfully immunized melanoma patients with targeted arrays of TSAs and adjuvant, demonstrating that even autologous responses can be boosted in patients that have presumably had significant exposure to these Ags (62, 63).

Although there is good evidence that repertoire limitations can be overcome, there is still the potential that the targeted approaches of TSAs may generate cross-reactive responses to the parent self-epitope, in effect breaking tolerance. This concern is particularly significant for adoptively transferred TCR approaches that do not rely on the elicitation of autologous responses. The goal of these therapies would be analogous to CAR TCR therapies, where TCRs specific for the TSA-associated epitope are engineered into autologous T cells that are then expanded and infused back into the patient. However, if the TCRs exhibit significant cross-reactivity to the self-epitope, a large infusion of cells bearing these receptors could target the unmutated parent self-epitope. Beyond off-target affinity for the parent epitope, unexpected reactivities could also occur against other endogenous peptides, as was reported for a trial of TCR-based therapy targeted at MAGE-A3, where the used TCR also targeted a self-epitope from the striated muscle protein titin (64). Notably, this TCR was derived from an endogenous response but was affinity-enhanced via phage display selection. The wild-type TCR did not respond to the titin-derived epitope except at very high concentrations, suggesting that it would not have mediated the same reactivity. This result serves as an important cautionary note for engineered TCR approaches and suggests that endogenous peripheral tolerance mechanisms are able to shape antitumor responses away from significant normal self-reactivity.

Several approaches have been suggested to limit self-reactivity by engineered TCRs. As noted, relying solely on endogenously generated TCRs may significantly limit potential self-reactivities due to editing by central and peripheral tolerance mechanisms. In contrast, TCRs selected in vitro or TCRs from humanized mouse models may have undetected reactivities and have the potential to be antigenic themselves (65). Additionally, some concerns have been raised about promiscuous pairing of introduced TCR chains with the endogenous chains present in a T cell, leading to hybrid receptors that may have markedly varied targets (66). To avoid this possibility, altering the TCR constant regions to pair preferentially with the cognate engineered chain has been achieved by the introduction of a second cysteine or by swapping human constant regions for murine homologs; both strategies resulted in higher expression of TCR on the cell surface (67, 68).

The potential for off-target or undetected specificity of an introduced TCR raises the issue of the cross-reactivity of the TCR repertoire, which has been estimated by various methods to be quite vast (exceeding 1 million peptides per TCR in one experimental report (69)). The argument for the inherent cross-reactivity of the TCR repertoire can be inferred directly from the basic principles of adaptive immunity: given the broad reactivity of the immune system to novel Ags, the potential landscape of MHC-binding peptides, and the size of the TCR repertoire in any individual, the size of the potential pMHC target pool is many orders of magnitude greater than the size of an individual’s repertoire. Yet although it is clear that a TCR should be highly cross-reactive, it is less clear in which “neighborhood” that cross-reactivity will occur (70).

One landmark study addressing this question identified the targets of individual TCRs by screening yeast display libraries, where researchers demonstrated that an individual TCR targeted relatively conserved features of a given peptide surface but was agnostic to even dramatic changes at other residues in the peptide (71). Although the potential peptide targets contained a strikingly high diversity of peptides, there were consistent commonalities among the peptide groups that even allowed prediction of reactive peptides not yet observed. Furthermore, even peptides that were reactive that bore no similarity could be “connected” by a series of intermediate reactive peptides that were a single mutation away from another reactive peptide.

The predictive aspects of this work suggest that it might be possible to determine if a TCR is likely to react with both a TSA-derived epitope and the parent peptide, although this capability is beyond current in silico approaches without sufficient training data sets.

T cell-mediated therapies for tumors can be broadly classified in three categories: 1) adoptive cell therapies, 2) vaccines, and 3) immune modulating therapies including ICB and cytokine therapy. Characterizing specificity is required for the first two types. Although ICB can be used without any knowledge of the TSA or TAA that T cells are targeting, several studies have found associations between tumor mutation burden (and in particular putative neoepitope burden) and improved outcomes (29, 40, 72). The reason for this association is not entirely clear; it may correlate with overall antitumor response magnitude, or certain high-quality responses may be more likely to be generated if a large number of TSAs are available for targeting.

For therapies in development, including vaccines and adoptive cell therapies, specific targets that might be shared across individuals are particularly attractive for generating off-the-shelf therapies that can be quickly applied. CD19 CAR T cells, NY-ESO-1 vaccines, and adoptive TCR therapies targeting common fusions and driver mutations are all broadly applicable across an entire tumor type or even multiple tumor types. One concern with these approaches is that they focus the response on a single target (though it may include multiple epitopes). Targeting multiple mutations is more likely to prevent tumor escape and account for the highly heterogeneous mutation landscapes characteristic of most tumors. However, finding multiple targets in a single tumor will likely require using private antigenic targets, (i.e., derived from mutations that are patient-specific). This introduces complexities in therapeutic development and delivery, but the tremendous promise of these approaches justifies the investment in developing workflows that will make these therapies widely tractable.

Isolating and cloning tumor-specific TCRs can now be done relatively rapidly. The actual determination of specificity remains as the final bottleneck, although determining specificity may not be required in every case. Further, recent advances include the development of algorithms to predict TCR specificity based on training data, and future progress in such endeavors likely stands to at least increase the efficiency of matching TCR sequences with epitopes (73, 74). Cloning and expressing TCRs offer the advantage that the cells in which they are transduced could be engineered to have increased functional activity, reduced likelihood of developing exhaustion phenotypes, and safety features such as kill switches or multiple Ag specificity requirements (75, 76). As they would generate high levels of tumor specificity with reduced off-target effects, these features seem to offer the best hope for safe and efficacious therapies even if many technological challenges remain.

This work was supported by Cancer Center Support Grants P30CA021765 and R01AI107625 and by the American Lebanese Syrian Associated Charities.

Abbreviations used in this article:

     
  • CAR

    chimeric Ag receptor

  •  
  • CTA

    cancer/testis Ag

  •  
  • ICB

    immune checkpoint blockade

  •  
  • pMHC

    peptide MHC

  •  
  • TAA

    tumor-associated Ag

  •  
  • TIL

    tumor-infiltrating lymphocyte

  •  
  • TSA

    tumor-specific Ag.

1
Bakker
,
A. B.
,
M. W.
Schreurs
,
G.
Tafazzul
,
A. J.
de Boer
,
Y.
Kawakami
,
G. J.
Adema
,
C. G.
Figdor
.
1995
.
Identification of a novel peptide derived from the melanocyte-specific gp100 antigen as the dominant epitope recognized by an HLA-A2.1-restricted anti-melanoma CTL line.
Int. J. Cancer
62
:
97
102
.
2
Cox
,
A. L.
,
J.
Skipper
,
Y.
Chen
,
R. A.
Henderson
,
T. L.
Darrow
,
J.
Shabanowitz
,
V. H.
Engelhard
,
D. F.
Hunt
,
C. L.
Slingluff
Jr
.
1994
.
Identification of a peptide recognized by five melanoma-specific human cytotoxic T cell lines.
Science
264
:
716
719
.
3
Kawakami
,
Y.
,
S.
Eliyahu
,
C. H.
Delgado
,
P. F.
Robbins
,
K.
Sakaguchi
,
E.
Appella
,
J. R.
Yannelli
,
G. J.
Adema
,
T.
Miki
,
S. A.
Rosenberg
.
1994
.
Identification of a human melanoma antigen recognized by tumor-infiltrating lymphocytes associated with in vivo tumor rejection.
Proc. Natl. Acad. Sci. USA
91
:
6458
6462
.
4
Kawakami
,
Y.
,
S.
Eliyahu
,
C.
Jennings
,
K.
Sakaguchi
,
X.
Kang
,
S.
Southwood
,
P. F.
Robbins
,
A.
Sette
,
E.
Appella
,
S. A.
Rosenberg
.
1995
.
Recognition of multiple epitopes in the human melanoma antigen gp100 by tumor-infiltrating T lymphocytes associated with in vivo tumor regression.
J. Immunol.
154
:
3961
3968
.
5
Coulie
,
P. G.
,
V.
Brichard
,
A.
Van Pel
,
T.
Wölfel
,
J.
Schneider
,
C.
Traversari
,
S.
Mattei
,
E.
De Plaen
,
C.
Lurquin
,
J. P.
Szikora
, et al
.
1994
.
A new gene coding for a differentiation antigen recognized by autologous cytolytic T lymphocytes on HLA-A2 melanomas.
J. Exp. Med.
180
:
35
42
.
6
Kawakami
,
Y.
,
S.
Eliyahu
,
K.
Sakaguchi
,
P. F.
Robbins
,
L.
Rivoltini
,
J. R.
Yannelli
,
E.
Appella
,
S. A.
Rosenberg
.
1994
.
Identification of the immunodominant peptides of the MART-1 human melanoma antigen recognized by the majority of HLA-A2-restricted tumor infiltrating lymphocytes.
J. Exp. Med.
180
:
347
352
.
7
Brichard
,
V.
,
A.
Van Pel
,
T.
Wölfel
,
C.
Wölfel
,
E.
De Plaen
,
B.
Lethé
,
P.
Coulie
,
T.
Boon
.
1993
.
The tyrosinase gene codes for an antigen recognized by autologous cytolytic T lymphocytes on HLA-A2 melanomas.
J. Exp. Med.
178
:
489
495
.
8
Wölfel
,
T.
,
A.
Van Pel
,
V.
Brichard
,
J.
Schneider
,
B.
Seliger
,
K. H.
Meyer zum Büschenfelde
,
T.
Boon
.
1994
.
Two tyrosinase nonapeptides recognized on HLA-A2 melanomas by autologous cytolytic T lymphocytes.
Eur. J. Immunol.
24
:
759
764
.
9
Maude
,
S. L.
,
D. T.
Teachey
,
D. L.
Porter
,
S. A.
Grupp
.
2015
.
CD19-targeted chimeric antigen receptor T-cell therapy for acute lymphoblastic leukemia.
Blood
125
:
4017
4023
.
10
Rodrigues
,
P. C.
,
S. N.
Oliveira
,
M. B.
Viana
,
E. I.
Matsuda
,
A. E.
Nowill
,
S. R.
Brandalise
,
J. A.
Yunes
.
2007
.
Prognostic significance of WT1 gene expression in pediatric acute myeloid leukemia.
Pediatr. Blood Cancer
49
:
133
138
.
11
Scharnhorst
,
V.
,
A. J.
van der Eb
,
A. G.
Jochemsen
.
2001
.
WT1 proteins: functions in growth and differentiation.
Gene
273
:
141
161
.
12
Disis
,
M. L.
,
D. R.
Wallace
,
T. A.
Gooley
,
Y.
Dang
,
M.
Slota
,
H.
Lu
,
A. L.
Coveler
,
J. S.
Childs
,
D. M.
Higgins
,
P. A.
Fintak
, et al
.
2009
.
Concurrent trastuzumab and HER2/neu-specific vaccination in patients with metastatic breast cancer.
J. Clin. Oncol.
27
:
4685
4692
.
13
Iqbal
,
N.
,
N.
Iqbal
.
2014
.
Human epidermal growth factor receptor 2 (HER2) in cancers: overexpression and therapeutic implications.
Mol. Biol. Int.
2014
:
852748
.
14
Stone
,
J. D.
,
D. T.
Harris
,
D. M.
Kranz
.
2015
.
TCR affinity for p/MHC formed by tumor antigens that are self-proteins: impact on efficacy and toxicity.
Curr. Opin. Immunol.
33
:
16
22
.
15
Morgan
,
R. A.
,
J. C.
Yang
,
M.
Kitano
,
M. E.
Dudley
,
C. M.
Laurencot
,
S. A.
Rosenberg
.
2010
.
Case report of a serious adverse event following the administration of T cells transduced with a chimeric antigen receptor recognizing ERBB2.
Mol. Ther.
18
:
843
851
.
16
van der Bruggen
,
P.
,
C.
Traversari
,
P.
Chomez
,
C.
Lurquin
,
E.
De Plaen
,
B.
Van den Eynde
,
A.
Knuth
,
T.
Boon
.
1991
.
A gene encoding an antigen recognized by cytolytic T lymphocytes on a human melanoma.
Science
254
:
1643
1647
.
17
Robbins
,
P. F.
,
S. H.
Kassim
,
T. L.
Tran
,
J. S.
Crystal
,
R. A.
Morgan
,
S. A.
Feldman
,
J. C.
Yang
,
M. E.
Dudley
,
J. R.
Wunderlich
,
R. M.
Sherry
, et al
.
2015
.
A pilot trial using lymphocytes genetically engineered with an NY-ESO-1-reactive T-cell receptor: long-term follow-up and correlates with response.
Clin. Cancer Res.
21
:
1019
1027
.
18
Morgan
,
R. A.
,
N.
Chinnasamy
,
D.
Abate-Daga
,
A.
Gros
,
P. F.
Robbins
,
Z.
Zheng
,
M. E.
Dudley
,
S. A.
Feldman
,
J. C.
Yang
,
R. M.
Sherry
, et al
.
2013
.
Cancer regression and neurological toxicity following anti-MAGE-A3 TCR gene therapy.
J. Immunother.
36
:
133
151
.
19
Stevanović
,
S.
,
A.
Pasetto
,
S. R.
Helman
,
J. J.
Gartner
,
T. D.
Prickett
,
B.
Howie
,
H. S.
Robins
,
P. F.
Robbins
,
C. A.
Klebanoff
,
S. A.
Rosenberg
,
C. S.
Hinrichs
.
2017
.
Landscape of immunogenic tumor antigens in successful immunotherapy of virally induced epithelial cancer.
Science
356
:
200
205
.
20
Draper
,
L. M.
,
M. L.
Kwong
,
A.
Gros
,
S.
Stevanović
,
E.
Tran
,
S.
Kerkar
,
M.
Raffeld
,
S. A.
Rosenberg
,
C. S.
Hinrichs
.
2015
.
Targeting of HPV-16+ epithelial cancer cells by TCR gene engineered T cells directed against E6.
Clin. Cancer Res.
21
:
4431
4439
.
21
Thompson
,
M. P.
,
R.
Kurzrock
.
2004
.
Epstein-Barr virus and cancer.
Clin. Cancer Res.
10
:
803
821
.
22
Lennerz
,
V.
,
M.
Fatho
,
C.
Gentilini
,
R. A.
Frye
,
A.
Lifke
,
D.
Ferel
,
C.
Wölfel
,
C.
Huber
,
T.
Wölfel
.
2005
.
The response of autologous T cells to a human melanoma is dominated by mutated neoantigens.
Proc. Natl. Acad. Sci. USA
102
:
16013
16018
.
23
Zhou
,
J.
,
M. E.
Dudley
,
S. A.
Rosenberg
,
P. F.
Robbins
.
2005
.
Persistence of multiple tumor-specific T-cell clones is associated with complete tumor regression in a melanoma patient receiving adoptive cell transfer therapy.
J. Immunother.
28
:
53
62
.
24
Pritchard
,
A. L.
,
J. G.
Burel
,
M. A.
Neller
,
N. K.
Hayward
,
J. A.
Lopez
,
M.
Fatho
,
V.
Lennerz
,
T.
Wölfel
,
C. W.
Schmidt
.
2015
.
Exome sequencing to predict neoantigens in melanoma.
Cancer Immunol. Res.
3
:
992
998
.
25
Strønen
,
E.
,
M.
Toebes
,
S.
Kelderman
,
M. M.
van Buuren
,
W.
Yang
,
N.
van Rooij
,
M.
Donia
,
M.-L.
Böschen
,
F.
Lund-Johansen
,
J.
Olweus
,
T. N.
Schumacher
.
2016
.
Targeting of cancer neoantigens with donor-derived T cell receptor repertoires.
Science
352
:
1337
1341
.
26
Cohen
,
C. J.
,
J. J.
Gartner
,
M.
Horovitz-Fried
,
K.
Shamalov
,
K.
Trebska-McGowan
,
V. V.
Bliskovsky
,
M. R.
Parkhurst
,
C.
Ankri
,
T. D.
Prickett
,
J. S.
Crystal
, et al
.
2015
.
Isolation of neoantigen-specific T cells from tumor and peripheral lymphocytes.
J. Clin. Invest.
125
:
3981
3991
.
27
Rajasagi
,
M.
,
S. A.
Shukla
,
E. F.
Fritsch
,
D. B.
Keskin
,
D.
DeLuca
,
E.
Carmona
,
W.
Zhang
,
C.
Sougnez
,
K.
Cibulskis
,
J.
Sidney
, et al
.
2014
.
Systematic identification of personal tumor-specific neoantigens in chronic lymphocytic leukemia.
Blood
124
:
453
462
.
28
Hanahan
,
D.
,
R. A.
Weinberg
.
2000
.
The hallmarks of cancer.
Cell
100
:
57
70
.
29
Chalmers
,
Z. R.
,
C. F.
Connelly
,
D.
Fabrizio
,
L.
Gay
,
S. M.
Ali
,
R.
Ennis
,
A.
Schrock
,
B.
Campbell
,
A.
Shlien
,
J.
Chmielecki
, et al
.
2017
.
Analysis of 100,000 human cancer genomes reveals the landscape of tumor mutational burden.
Genome Med.
9
:
34
.
30
Vogelstein
,
B.
,
N.
Papadopoulos
,
V. E.
Velculescu
,
S.
Zhou
,
L. A.
Diaz
Jr.
,
K. W.
Kinzler
.
2013
.
Cancer genome landscapes.
Science
339
:
1546
1558
.
31
Lawrence
,
M. S.
,
P.
Stojanov
,
P.
Polak
,
G. V.
Kryukov
,
K.
Cibulskis
,
A.
Sivachenko
,
S. L.
Carter
,
C.
Stewart
,
C. H.
Mermel
,
S. A.
Roberts
, et al
.
2013
.
Mutational heterogeneity in cancer and the search for new cancer-associated genes.
Nature
499
:
214
218
.
32
Gros
,
A.
,
M. R.
Parkhurst
,
E.
Tran
,
A.
Pasetto
,
P. F.
Robbins
,
S.
Ilyas
,
T. D.
Prickett
,
J. J.
Gartner
,
J. S.
Crystal
,
I. M.
Roberts
, et al
.
2016
.
Prospective identification of neoantigen-specific lymphocytes in the peripheral blood of melanoma patients.
Nat. Med.
22
:
433
438
.
33
Marty
,
R.
,
S.
Kaabinejadian
,
D.
Rossell
,
M. J.
Slifker
,
J.
van de Haar
,
H. B.
Engin
,
N.
de Prisco
,
T.
Ideker
,
W. H.
Hildebrand
,
J.
Font-Burgada
,
H.
Carter
.
2017
.
MHC-I genotype restricts the oncogenic mutational landscape.
Cell
. DOI: 10.1016/j.cell.2017.09.050.
34
McGranahan
,
N.
,
R.
Rosenthal
,
C. T.
Hiley
,
A. J.
Rowan
,
T. B. K.
Watkins
,
G. A.
Wilson
,
N. J.
Birkbak
,
S.
Veeriah
,
P.
Van Loo
,
J.
Herrero
,
C.
Swanton
;
TRACERx Consortium
.
2017
.
Allele-specific HLA loss and immune escape in lung cancer evolution.
Cell
. DOI: 10.1016/j.cell.2017.10.001.
35
Schumacher
,
T. N.
,
R. D.
Schreiber
.
2015
.
Neoantigens in cancer immunotherapy.
Science
348
:
69
74
.
36
Gubin
,
M. M.
,
M. N.
Artyomov
,
E. R.
Mardis
,
R. D.
Schreiber
.
2015
.
Tumor neoantigens: building a framework for personalized cancer immunotherapy.
J. Clin. Invest.
125
:
3413
3421
.
37
Sharma
,
P.
,
J. P.
Allison
.
2015
.
The future of immune checkpoint therapy.
Science
348
:
56
61
.
38
Ribas
,
A.
,
L. H.
Camacho
,
G.
Lopez-Berestein
,
D.
Pavlov
,
C. A.
Bulanhagui
,
R.
Millham
,
B.
Comin-Anduix
,
J. M.
Reuben
,
E.
Seja
,
C. A.
Parker
, et al
.
2005
.
Antitumor activity in melanoma and anti-self responses in a phase I trial with the anti-cytotoxic T lymphocyte-associated antigen 4 monoclonal antibody CP-675,206.
J. Clin. Oncol.
23
:
8968
8977
.
39
Weber
,
J. S.
,
S.
O’Day
,
W.
Urba
,
J.
Powderly
,
G.
Nichol
,
M.
Yellin
,
J.
Snively
,
E.
Hersh
.
2008
.
Phase I/II study of ipilimumab for patients with metastatic melanoma.
J. Clin. Oncol.
26
:
5950
5956
.
40
Rizvi
,
N. A.
,
M. D.
Hellmann
,
A.
Snyder
,
P.
Kvistborg
,
V.
Makarov
,
J. J.
Havel
,
W.
Lee
,
J.
Yuan
,
P.
Wong
,
T. S.
Ho
, et al
.
2015
.
Cancer immunology. Mutational landscape determines sensitivity to PD-1 blockade in non-small cell lung cancer.
Science
348
:
124
128
.
41
Gettinger
,
S. N.
,
L.
Horn
,
L.
Gandhi
,
D. R.
Spigel
,
S. J.
Antonia
,
N. A.
Rizvi
,
J. D.
Powderly
,
R. S.
Heist
,
R. D.
Carvajal
,
D. M.
Jackman
, et al
.
2015
.
Overall survival and long-term safety of nivolumab (anti–programmed death 1 antibody, BMS-936558, ONO-4538) in patients with previously treated advanced non–small-cell lung cancer.
J. Clin. Oncol.
33
:
2004
2012
.
42
Rizvi
,
N. A.
,
J.
Mazières
,
D.
Planchard
,
T. E.
Stinchcombe
,
G. K.
Dy
,
S. J.
Antonia
,
L.
Horn
,
H.
Lena
,
E.
Minenza
,
B.
Mennecier
, et al
.
2015
.
Activity and safety of nivolumab, an anti-PD-1 immune checkpoint inhibitor, for patients with advanced, refractory squamous non-small-cell lung cancer (CheckMate 063): a phase 2, single-arm trial.
Lancet Oncol.
16
:
257
265
.
43
Yarchoan
,
M.
,
B. A.
Johnson
III
,
E. R.
Lutz
,
D. A.
Laheru
,
E. M.
Jaffee
.
2017
.
Targeting neoantigens to augment antitumour immunity.
Nat. Rev. Cancer
17
:
209
222
.
44
Munson
,
D. J.
,
C. A.
Egelston
,
K. E.
Chiotti
,
Z. E.
Parra
,
T. C.
Bruno
,
B. L.
Moore
,
T. A.
Nakano
,
D. L.
Simons
,
G.
Jimenez
,
J. H.
Yim
, et al
.
2016
.
Identification of shared TCR sequences from T cells in human breast cancer using emulsion RT-PCR.
Proc. Natl. Acad. Sci. USA
113
:
8272
8277
.
45
Yewdell
,
J. W.
2006
.
Confronting complexity: real-world immunodominance in antiviral CD8+ T cell responses.
Immunity
25
:
533
543
.
46
Abelin
,
J. G.
,
D. B.
Keskin
,
S.
Sarkizova
,
C. R.
Hartigan
,
W.
Zhang
,
J.
Sidney
,
J.
Stevens
,
W.
Lane
,
G. L.
Zhang
,
T. M.
Eisenhaure
, et al
.
2017
.
Mass spectrometry profiling of HLA-associated peptidomes in mono-allelic cells enables more accurate epitope prediction.
Immunity
46
:
315
326
.
47
Moon
,
J. J.
,
H. H.
Chu
,
M.
Pepper
,
S. J.
McSorley
,
S. C.
Jameson
,
R. M.
Kedl
,
M. K.
Jenkins
.
2007
.
Naive CD4(+) T cell frequency varies for different epitopes and predicts repertoire diversity and response magnitude.
Immunity
27
:
203
213
.
48
Obar
,
J. J.
,
K. M.
Khanna
,
L.
Lefrançois
.
2008
.
Endogenous naive CD8+ T cell precursor frequency regulates primary and memory responses to infection.
Immunity
28
:
859
869
.
49
Lazarski
,
C. A.
,
F. A.
Chaves
,
S. A.
Jenks
,
S.
Wu
,
K. A.
Richards
,
J. M.
Weaver
,
A. J.
Sant
.
2005
.
The kinetic stability of MHC class II:peptide complexes is a key parameter that dictates immunodominance.
Immunity
23
:
29
40
.
50
Cukalac
,
T.
,
J.
Chadderton
,
W.
Zeng
,
J. G.
Cullen
,
W. T.
Kan
,
P. C.
Doherty
,
D. C.
Jackson
,
S. J.
Turner
,
N. L.
La Gruta
.
2014
.
The influenza virus-specific CTL immunodominance hierarchy in mice is determined by the relative frequency of high-avidity T cells.
J. Immunol.
192
:
4061
4068
.
51
Flecken
,
T.
,
N.
Schmidt
,
S.
Hild
,
E.
Gostick
,
O.
Drognitz
,
R.
Zeiser
,
P.
Schemmer
,
H.
Bruns
,
T.
Eiermann
,
D. A.
Price
, et al
.
2014
.
Immunodominance and functional alterations of tumor-associated antigen-specific CD8+ T-cell responses in hepatocellular carcinoma.
Hepatology
59
:
1415
1426
.
52
Sung
,
P. S.
,
V.
Racanelli
,
E.-C.
Shin
.
2014
.
CD8(+) T-cell responses in acute hepatitis C virus infection.
Front. Immunol.
5
:
266
.
53
Jackson
,
H.
,
N.
Dimopoulos
,
N. A.
Mifsud
,
T. Y.
Tai
,
Q.
Chen
,
S.
Svobodova
,
J.
Browning
,
I.
Luescher
,
L.
Stockert
,
L. J.
Old
, et al
.
2006
.
Striking immunodominance hierarchy of naturally occurring CD8+ and CD4+ T cell responses to tumor antigen NY-ESO-1.
J. Immunol.
176
:
5908
5917
.
54
Chen
,
W.
,
J.
McCluskey
.
2006
.
Immunodominance and immunodomination: critical factors in developing effective CD8+ T-cell-based cancer vaccines.
Adv. Cancer Res.
95
:
203
247
.
55
Schreiber
,
H.
,
T. H.
Wu
,
J.
Nachman
,
W. M.
Kast
.
2002
.
Immunodominance and tumor escape.
Semin. Cancer Biol.
12
:
25
31
.
56
Bioley
,
G.
,
P.
Guillaume
,
I.
Luescher
,
A.
Yeh
,
B.
Dupont
,
N.
Bhardwaj
,
G.
Mears
,
L. J.
Old
,
D.
Valmori
,
M.
Ayyoub
.
2009
.
HLA class I - associated immunodominance affects CTL responsiveness to an ESO recombinant protein tumor antigen vaccine.
Clin. Cancer Res.
15
:
299
306
.
57
Chang
,
T.-C.
,
R. A.
Carter
,
Y.
Li
,
Y.
Li
,
H.
Wang
,
M. N.
Edmonson
,
X.
Chen
,
P.
Arnold
,
T. L.
Geiger
,
G.
Wu
, et al
.
2017
.
The neoepitope landscape in pediatric cancers.
Genome Med.
9
:
78
.
58
Yu
,
W.
,
N.
Jiang
,
P. J. R.
Ebert
,
B. A.
Kidd
,
S.
Müller
,
P. J.
Lund
,
J.
Juang
,
K.
Adachi
,
T.
Tse
,
M. E.
Birnbaum
, et al
.
2015
.
Clonal deletion prunes but does not eliminate self-specific αβ CD8(+) T lymphocytes.
Immunity
42
:
929
941
.
59
Bowen
,
D. G.
,
C. M.
Walker
.
2005
.
Mutational escape from CD8+ T cell immunity: HCV evolution, from chimpanzees to man.
J. Exp. Med.
201
:
1709
1714
.
60
van Buuren
,
M. M.
,
J. J.
Calis
,
T. N.
Schumacher
.
2014
.
High sensitivity of cancer exome-based CD8 T cell neo-antigen identification.
OncoImmunology
3
:
e28836
.
61
Schubert
,
B.
,
L.
de la Garza
,
C.
Mohr
,
M.
Walzer
,
O.
Kohlbacher
.
2017
.
ImmunoNodes - graphical development of complex immunoinformatics workflows.
BMC Bioinformatics
18
:
242
.
62
Ott
,
P. A.
,
Z.
Hu
,
D. B.
Keskin
,
S. A.
Shukla
,
J.
Sun
,
D. J.
Bozym
,
W.
Zhang
,
A.
Luoma
,
A.
Giobbie-Hurder
,
L.
Peter
, et al
.
2017
.
An immunogenic personal neoantigen vaccine for patients with melanoma.
Nature
547
:
217
221
.
63
Sahin
,
U.
,
E.
Derhovanessian
,
M.
Miller
,
B.-P.
Kloke
,
P.
Simon
,
M.
Löwer
,
V.
Bukur
,
A. D.
Tadmor
,
U.
Luxemburger
,
B.
Schrörs
, et al
.
2017
.
Personalized RNA mutanome vaccines mobilize poly-specific therapeutic immunity against cancer.
Nature
547
:
222
226
.
64
Linette
,
G. P.
,
E. A.
Stadtmauer
,
M. V.
Maus
,
A. P.
Rapoport
,
B. L.
Levine
,
L.
Emery
,
L.
Litzky
,
A.
Bagg
,
B. M.
Carreno
,
P. J.
Cimino
, et al
.
2013
.
Cardiovascular toxicity and titin cross-reactivity of affinity-enhanced T cells in myeloma and melanoma.
Blood
122
:
863
871
.
65
Davis
,
J. L.
,
M. R.
Theoret
,
Z.
Zheng
,
C. H. J.
Lamers
,
S. A.
Rosenberg
,
R. A.
Morgan
.
2010
.
Development of human anti-murine T-cell receptor antibodies in both responding and nonresponding patients enrolled in TCR gene therapy trials.
Clin. Cancer Res.
16
:
5852
5861
.
66
Kuball
,
J.
,
M. L.
Dossett
,
M.
Wolfl
,
W. Y.
Ho
,
R.-H.
Voss
,
C.
Fowler
,
P. D.
Greenberg
.
2007
.
Facilitating matched pairing and expression of TCR chains introduced into human T cells.
Blood
109
:
2331
2338
.
67
Cohen
,
C. J.
,
Y. F.
Li
,
M.
El-Gamil
,
P. F.
Robbins
,
S. A.
Rosenberg
,
R. A.
Morgan
.
2007
.
Enhanced antitumor activity of T cells engineered to express T-cell receptors with a second disulfide bond.
Cancer Res.
67
:
3898
3903
.
68
Cohen
,
C. J.
,
Y.
Zhao
,
Z.
Zheng
,
S. A.
Rosenberg
,
R. A.
Morgan
.
2006
.
Enhanced antitumor activity of murine-human hybrid T-cell receptor (TCR) in human lymphocytes is associated with improved pairing and TCR/CD3 stability.
Cancer Res.
66
:
8878
8886
.
69
Wooldridge
,
L.
,
J.
Ekeruche-Makinde
,
H. A.
van den Berg
,
A.
Skowera
,
J. J.
Miles
,
M. P.
Tan
,
G.
Dolton
,
M.
Clement
,
S.
Llewellyn-Lacey
,
D. A.
Price
, et al
.
2012
.
A single autoimmune T cell receptor recognizes more than a million different peptides.
J. Biol. Chem.
287
:
1168
1177
.
70
Singh
,
N. K.
,
T. P.
Riley
,
S. C. B.
Baker
,
T.
Borrman
,
Z.
Weng
,
B. M.
Baker
.
2017
.
Emerging concepts in TCR specificity: rationalizing and (Maybe) predicting outcomes.
J. Immunol.
199
:
2203
2213
.
71
Birnbaum
,
M. E.
,
J. L.
Mendoza
,
D. K.
Sethi
,
S.
Dong
,
J.
Glanville
,
J.
Dobbins
,
E.
Ozkan
,
M. M.
Davis
,
K. W.
Wucherpfennig
,
K. C.
Garcia
.
2014
.
Deconstructing the peptide-MHC specificity of T cell recognition.
Cell
157
:
1073
1087
.
72
Yuan
,
J.
,
P. S.
Hegde
,
R.
Clynes
,
P. G.
Foukas
,
A.
Harari
,
T. O.
Kleen
,
P.
Kvistborg
,
C.
Maccalli
,
H. T.
Maecker
,
D. B.
Page
, et al
.
2016
.
Novel technologies and emerging biomarkers for personalized cancer immunotherapy.
J. Immunother. Cancer
4
:
3
.
73
Dash
,
P.
,
A. J.
Fiore-Gartland
,
T.
Hertz
,
G. C.
Wang
,
S.
Sharma
,
A.
Souquette
,
J. C.
Crawford
,
E. B.
Clemens
,
T. H. O.
Nguyen
,
K.
Kedzierska
, et al
.
2017
.
Quantifiable predictive features define epitope-specific T cell receptor repertoires.
Nature
547
:
89
93
.
74
Glanville
,
J.
,
H.
Huang
,
A.
Nau
,
O.
Hatton
,
L. E.
Wagar
,
F.
Rubelt
,
X.
Ji
,
A.
Han
,
S. M.
Krams
,
C.
Pettus
, et al
.
2017
.
Identifying specificity groups in the T cell receptor repertoire.
Nature
547
:
94
98
.
75
Sharpe
,
M.
,
N.
Mount
.
2015
.
Genetically modified T cells in cancer therapy: opportunities and challenges.
Dis. Model. Mech.
8
:
337
350
.
76
Ghoneim
,
H. E.
,
A. E.
Zamora
,
P. G.
Thomas
,
B. A.
Youngblood
.
2016
.
Cell-intrinsic barriers of T cell-based immunotherapy.
Trends Mol. Med.
22
:
1000
1011
.

The authors have no financial conflicts of interest.