The role of neoantigens in response to immune checkpoint blockade

Abstract

Immune checkpoint blockade has demonstrated substantial promise for the treatment of several advanced malignancies. These agents activate the immune system to attack tumor cells. For example, agents targeting CTLA4 and programmed cell death 1 (PD-1) have resulted in impressive response rates and, in some cases, durable remissions. Neoantigens are mutations that encode immunologically active proteins that can cause the immune system to recognize the affected cell as foreign. Recent data have made it clear that these mutations are, in large part, the functional targets of immune checkpoint blockade. This review summarizes the key discoveries leading up to this important conclusion and discusses possible applications of neoantigens in cancer therapy.

Introduction

The immune system has evolved to detect and destroy foreign pathogens. The notion that such a mechanism might be relevant to eliminating cancer cells originated in 1893, when William Coley in New York observed an extraordinary tumor regression after bacterial infection. In 1909, Paul Ehrlich hypothesized that ongoing immune surveillance repressed the growth of cancers that would otherwise develop more frequently (1). This process has subsequently been recognized to involve a complex relationship between tumor cells and the immune system, summarized by the cancer immunoediting hypothesis by Burnet, Thomas and, most recently, Schreiber (2). According to this theory, the emergence of tumors is, at least in part, the result of loss of equilibrium between tumor cells and immune system-mediated elimination, resulting in immune escape and tumor growth.

One mechanism of immune escape occurs when tumor cells co-opt inherent immunoregulatory mechanisms that normally exist to prevent host autoimmunity and restrain excessive immune reactivity. In the absence of these immune checkpoints, T cells are activated when a peptide is presented through an HLA molecule to the TCR (this triggers ‘signal one’ in the T cell) together with the simultaneous interaction of a co-stimulatory molecule, such as CD80 or CD86, with their receptor on T cells, CD28 (‘signal two’) (3). Professional antigen-presenting cells are able to help provide both signals, leading to priming and expansion of naive T cells. Non-hematopoietic cells in the body are able to present endogenous peptides on MHC class I molecules, a process that is involved in immunosurveillance for viral infections. These processes have already been elegantly reviewed extensively over the years and will not be revisited here (4–6).

A number of processes have been identified that modulate TCR signaling. CTLA4 was first cloned in 1987, a molecule homologous to CD28, and is able to bind to CD80/CD86 with strong affinity (7). Its expression is up-regulated in response to T-cell activation (8–10) and induces an inhibitory signal in the T cell.

In 1996, it was first observed that inhibiting CTLA4 in mice led to tumor regression (11). In 2010, a phase 3 trial investigated the efficacy of ipilimumab, a humanized mAb targeting CTLA4, for the treatment of metastatic melanoma. This seminal trial demonstrated, for the first time, an improvement in overall survival among patients with metastatic melanoma who were treated with immunotherapy, achieving responses in 11% of patients (12). This ultimately led to the US Food and Drug Administration (FDA) approval of ipilimumab for metastatic melanoma in 2011. Interestingly, long-term follow-up of over 1800 melanoma patients treated with ipilimumab showed that approximately 20% of patients experienced durable remission of their tumor that may or may not have been reflected in upfront objective response rate measurements (13). Ipilimumab was approved in 2015 for use in the adjuvant setting for stage 3 melanoma on the basis of positive phase 3 trial results (14).

In parallel, the receptor called programmed cell death 1 (PD-1) was identified in 1992. It is expressed on activated T cells and binds to ligands (PDL-1 or PDL-2) that are commonly expressed on tumor cells (15, 16). This binding induces a T-cell inhibitory signaling cascade (17).

In 2014, the FDA approved antibodies targeting PD-1—pembrolizumab and nivolumab—for the treatment of metastatic melanoma. Both are fully humanized IgG4 antibodies targeting PD-1 that selectively block the interaction of the receptor with PD-L1 and PD-L2. Nivolumab achieved response rates of 32% in patients with metastatic melanoma whose tumors progressed after anti-CTLA4 therapy (18) and 40% in previously untreated patients who had metastatic melanoma with wild-type BRAF (19). Similar results (46% overall response rate) were seen with pembrolizumab in ipilimumab-refractory melanoma (20). In 2015, both agents were FDA approved for use in non-small cell lung cancer (NSCLC), after similar response rates of 18–19% were observed (21, 22). Nivolumab has also recently received an FDA-approved indication for renal cell carcinoma, on the basis of a 25% response rate in a phase 3 trial (23). Anti-PD-1 therapy is currently under clinical investigation in many additional cancer types, with early phase trial data showing similar or superior response rates in bladder, esophageal, gastric, head and neck and Merkel cell cancers and in lymphomas.

Importantly, in the phase 3 studies of melanoma and of lung and renal cell cancers, the response rates were associated with significant improvements in overall survival and, in some cases, long-term remissions. Still, the majority of patients do not achieve responses to these therapies, highlighting the need for improvement, either with combination therapies or novel agents targeting alternate immunoregulatory checkpoints.

For example, CTLA4 and PD-1 checkpoints are complementary and nonredundant, and pre-clinical models demonstrated synergy with dual blockade. These findings led to a phase 3 clinical trial investigating combination therapy with ipilimumab and nivolumab that achieved an unprecedented 58% objective response rate in metastatic melanoma with combination therapy, including a 12% complete response rate (24, 25). In lung cancer, combination treatments have resulted in promising response rates but at the cost of substantial toxicities (26).

There is currently significant interest in targeting other immunoregulatory molecules. PD-L1 and PD-L2 are found on tumors and antigen-presenting cells. Several trials are underway examining these agents alone or in combination with anti-PD-1 agents (27–30). There are additional receptors such as TIM-3 (limits CD4⁺ and CD8⁺ T-cell responses and activates T_reg cells (26, 27)) and LAG-3 (binds to MHC class II molecules to decrease T-cell activation and also enhances T_reg cells (31)) that are currently being investigated in early phase clinical trials as therapeutic targets (32, 33). New targets include, but are not limited to, IDO1, 4-1BB, TIGIT, VISTA, GITR, KIR, OX40, ICOS and others.

Neoantigens can mediate recognition of a cancer as ‘foreign’ by the immune system

The removal of the ‘brakes’ on the immune system with immune checkpoint blockade has allowed for the remarkable improvements in clinical outcomes described above. However, even with the brakes off, the adaptive immune system still must recognize some portion of cancer as foreign in order to facilitate selective elimination of a cancer. Evidence that tumor-specific antigens facilitate immune elimination of a cancer was described over 50 years ago with the use of chemically or UV-induced tumors in murine models (34–36). Prehn and Main demonstrated that chemically induced fibrosarcomas would be rejected by syngeneic mice that were immunized prior to transplantation, and this immunization was incomplete if a different cancer was used for transplantation or non-existent if normal tissue was utilized for immunization (34).

Theoretically, these tumor-specific antigens could consist of cancer–testis antigens, tissue differentiation genes, over-expressed genes (amplified oncogenes) in the tumor or tumor-specific neoantigens (37). Cancer–testis antigens consist of genes that are normally only expressed in germ cells but are often aberrantly expressed in some cancers because of epigenetic dysregulation (38). As germ cells do not express MHC class I proteins, these could serve as tumor-specific antigens. Alternatively, lineage-specific genes such as melan-A or GP100 have been shown to induce T-cell responses from human patients with tumors, although why tolerance to these antigens is incomplete is unclear (37, 39–41).

In contrast to other types of tumor-specific antigens, neoantigens are novel peptides that are not normally found in the host and are unique to a particular cancer (42). These peptides are not subject to central tolerance. Neoantigen can arise from somatic mutations (or other genetic alterations) that result in the production of a novel peptide, or from viral peptides in virally induced cancers (e.g. human papilloma virus-related or EBV-related tumors).

Mandelboim et al. provided initial pre-clinical evidence that neoantigens could facilitate recognition of a cancer by identifying a mutation in connexin 37 that induced a CTL response in a murine lung cancer model (43). They then demonstrated that peptide vaccination with this neoantigen could protect mice from metastasis and reduce the metastatic load in mice with preexisting metastasis (44). Subsequently, multiple groups identified CTLs against mutated genes, such as MUM1, CDK4 and CTNNB1 (β-catenin) in human melanomas (45–47). Other groups have identified unique antigens secondary to point mutations that can induce CTLs in human tumors (reviewed by Sensi and Anichini) (36).

Lennerz et al. comprehensively analyzed a single patient’s T-cell repertoire for reactivity against neoantigens, differentiation antigens and cancer–testis antigens (48). They identified T cells reactive against five neoantigens and, using peripheral blood samples collected over time, demonstrated that immunoreactivity against neoantigens was the dominant form of response by the immune system in response to tumors. Similarly, Robbins and Rosenberg identified T cells that responded to mutated forms of GAS7 and GAPDH in a patient who achieved a complete response to adoptive immunotherapy with autologous tumor-infiltrating lymphocytes (49).

In 2012, researchers from the groups of Schreiber (50) and of Jacks (51) demonstrated the importance of neoantigens in the process of immunoediting in pre-clinical murine models. Immunoediting, a revision of the immune-surveillance hypothesis, posits that the immune system serves as an extrinsic tumor suppressor. Immunoediting consists of three sequential phases: elimination, equilibrium and escape (1, 52). In the elimination phase, the adaptive and innate immune system are able to destroy a cancer before it becomes clinically apparent. Tumors that survive the elimination phase move into equilibrium, where the adaptive immune system prevents rapid growth of the tumor but simultaneously shapes the immunogenicity of the tumor. In the escape phase, tumors have acquired the ability to bypass the immune system and become clinically apparent.

To investigate the role of neoantigens in immunoediting, Schreiber’s group created a chemically induced sarcoma in an immune-deficient murine host, resulting in an ‘unedited’ tumor (50). This unedited tumor was typically rejected when orthotopically transplanted into wild-type mice but grew in other immune-deficient mice. They also identified less immunogenic sub-clones of this tumor that survived transplantation into wild-type mice. Sequencing these clones and performing bioinformatics analysis (see section below for details) led to the identification of a mutation in SPTBN2 (spectrin-β2) as responsible for the rejection. Forced expression of this antigen in tumors transplanted into wild-type mice led to rejection. These results along with the observations from Jacks’ group (51) suggested that, in mice, neoantigens play a key role in inhibiting tumor development.

Cancer genetics influences the neoantigen landscape in tumors

The majority of neoantigens identified that lead to an immune response in both human tumors and murine models have not arisen from traditional oncogenes or tumor-suppressor genes but rather from passenger genes, that is, genes unlikely to directly contribute to oncogenesis (53, 54). The number of passenger mutations for a particular tumor depends on the mutational load and this, in turn, indicates the chance that an immunogenically relevant neoantigen will be present (55, 56).

Over the past decade, comprehensive genomic analysis from The Cancer Genome Atlas (TCGA) of a wide variety of solid and liquid tumors has characterized the average mutational load for different types of malignancies. These studies have revealed a broad range of mutational loads from less than 0.1 mutation per megabase pair (Mbp) for pilocytic astrocytoma to over 10 mutations per Mbp for melanoma (57, 58). Not surprisingly, the tumor historically thought to be the most immunogenic (melanoma) also has the highest mutational load. Further, in the TCGA data, an increased mutation load is correlated with increased cytolytic activity (RNAseq-derived measurement of T-cell and NK cell infiltration) (59, 60). This further supports the notion that tumors with an increased mutation load have increased immune activity.

Two critical determinants of mutational load are carcinogen exposure and defects in DNA-repair pathways. Exposure to UV light is the major carcinogen in melanoma and results in the majority of mutations in these tumors (61). Melanomas without evidence of UV exposure typically have a lower mutational load and presumably fewer immunogenically relevant neoantigens (61). Similarly, in NSCLC, exposure to tobacco increases the mutational load of tumors and, more specifically, mutations that result in transversions (62, 63).

Defects in DNA-repair pathways could also lead to further increases in the number of mutations and subsequent neoantigens through a hypermutator phenotype (56, 64). One such example we identified involves mutations in the polymerase proofreading domain of POLE and POLD1 that disturb the fidelity of the enzyme. These mutations can result in 100 or more mutations per Mbp (65, 66). Similarly, defects in mismatch repair (e.g. Lynch syndrome) in colon cancers can result in increased mutational loads and an increase in the number of frameshift mutations (67). Unlike point mutations, frameshifts can result in completely novel peptides that may be especially immunogenic (68, 69). Neoantigens from missense mutations fall into at least two classes: type I neoantigens that change an amino acid in regions that contact the TCR CDR3 but do not alter anchor residues and type II neoantigens that change anchor residues and alter binding affinities to MHC molecules (Fig. 1).

Fig. 1.

Example of the types of neoantigens. Missense mutations can alter amino acids in peptides that are normally present in regions that make contact with the TCR. These mutations do not alter the binding affinity of the peptide to MHC molecules but may make the peptide immunogenic (top). Mutations can also occur on locations that convert a previously unpresented peptide into one that now binds to MHC molecules. These mutations can generate a new anchor residue that promotes the binding of the mutated peptide onto MHC complexes (bottom).

Besides mutational load, another key genetic contributor to immune response and recognition may be the underlying genetic heterogeneity of a particular tumor. In the past decade, it has become increasingly clear that tumors are genetically heterogeneous, and multiple sub-clonal populations with distinct mutational profiles can exist in any one patient’s tumor (70, 71). Our group and others have identified that tumor clonality is prognostic in many cancer types regardless of the underlying treatment, with tumors with increasing intratumoral genetic heterogeneity having worse outcomes (71–73). In terms of immune recognition of a tumor, theoretically, if a neoantigen is not present in all sub-clones of a tumor, the immune system ultimately will not be able to recognize the entire cancer as foreign, and a sub-clone may survive to propagate the tumor. Recently, our group collaborated with the Swanton and Quezada laboratories to show that clonal neoantigens—neoantigens that are present in most if not all clones—tend to be more immunogenic (74).

As an extension of our hypotheses on the effects of clonality on immunotherapy efficacy, the extent of metastatic tumoral heterogeneity should have strong effects on how immune checkpoint agents work. If a patient has a highly heterogeneous tumor burden among metastatic tumors, then it stands to reason that some metastatic clones may not possess the relevant neoantigens needed for immune clearance. Alternatively, other patients may have relatively homogenous metastases derived from tumors with less heterogeneity. These latter patients would be predicted to respond better to immune checkpoint therapy (Fig. 2).

Fig. 2.

The effects of clonal heterogeneity on immunotherapy efficacy. Shown in the figure are two patients with tumors each harboring the same six neoantigens. Two neoantigens, denoted by red shapes, need to be present for immune-mediated clearance following immune checkpoint therapy, while blue shapes represent neoantigens not recognized by the immune system. If a tumor possesses high heterogeneity, some clones and/or metastatic lesions may possess the needed neoantigens while others may not. As such, immunotherapy would not clear all metastases and the cancer will progress (top). If the tumor has few sub-clones and all clones possess the two neoantigens necessary for response, then immunotherapy will clear all metastases and result in disease control (bottom).

Some immunologists have suggested that only one dominant neoantigen per tumor drives the immunotherapy response. This is highly unlikely in light of the genetic complexity seen between different metastatic deposits from the same patient and even within the same tumor. If it was really one dominant neoantigen that drove response in human tumors, one would anticipate that the tumor would easily adapt around this mutation (like with targeted therapies). The durability of response to immunotherapies is inconsistent with this hypothesis as is the genetic diversity of metastases. Hence, concepts of immunodominant antigens derived from pre-clinical models may not accurately reflect the genetic complexities of human tumors (50, 75).

Neoantigen predictions from cancer sequencing data

Segal et al. undertook one of the first attempts to use exome sequencing data to characterize neoantigens for a series of breast and colon cancer patients. This group performed an exercise where they predicted mutated peptides that would bind to HLA-A*02:01 (76). Although these results are likely to be largely incorrect, given that patient-specific HLA alleles were not used, the study did illustrate the potential usefulness of the general approach.

Since then, multiple groups have expanded on this method of using sequencing data to generate putative neoantigens that we previously reviewed in detail (77). Briefly, this process entails first sequencing the whole exome of a tumor to identify somatic point mutations and frameshift mutations. Although somatic mutation calling is commonplace (determining which base pairs in the tumor genome are different from the host genome), there remain significant differences in both somatic mutation calls and insertion/deletion (indel) calls between different experienced genome centers (78, 79). For example, the International Cancer Genome Consortium (ICGC) provided 16 well-regarded genome analysis centers the same FASTQ files (i.e. the format for sharing the sequences and the per-base quality of data from high-throughput sequencers) and discovered significant discrepancies in mutation calls between the centers (78).

Centers that had better accuracy typically relied on multiple mutation callers in their pipelines, and in our own neoantigen pipeline, we combine the results of four different callers and merge information to enhance both sensitivity and specificity (56). It is therefore not surprising that neoantigen calls differ significantly between work by different research groups. It is of paramount importance to use highly accurate mutation-calling pipelines—that frequently use more than one algorithm—in neoantigen prediction.

The second step entails having mutations filtered by looking only at the genes that are expressed utilizing RNAseq data from the same tumor. Although this step intuitively makes sense and has worked in pre-clinical settings, results in clinical settings have been mixed (80). Since clonality is critical and metastases often develop from small sub-clonal populations, it is important to not use expression data as a blunt filter to discount neoantigens that are not apparently highly expressed based on sequencing from limited tissue sources such as biopsies.

The set of mutations can then be virtually translated into predicted proteins and neoepitopes that result from a non-synonymous mutation. These candidate neoepitopes can then be assessed for their ability to bind HLA molecules and transverse the antigen-processing machinery by determining proteasome cleavage sites and TAP transport efficiency (81, 82). The tools for the latter two factors are, arguably, still in need of much optimization.

As mentioned, using one of several computational algorithms, binding of mutated peptides to patient-specific HLA alleles can be performed for HLA class I or class II molecules to identify neoantigens (83–85). Class I prediction has generally been more robust and used more routinely in oncology studies.

These tools consist of two classes of prediction approaches—allele-specific and pan-specific tools. In general, these tools use techniques from machine learning (such as neural networks) to train a predictor using a large database of known peptide-MHC pairs such as from the Immune Epitope Data Base (IEDB). As many rarer HLA alleles do not have experimentally measured data, the pan-specific tools make further extrapolations using similar HLA alleles (86). Allele-specific MHC class I prediction works reasonably well with an area under the curve of around 0.75 for most tools (84). Although the pan-specific tools, such as NetMHCpan, predict binding for a larger swath of HLA alleles, they may not produce as reliable results for rarer alleles and hence could make the ultimate interpretation of predictions more difficult (80, 87).

In vitro or ex vivo testing is frequently used to confirm which neoepitopes can elicit an immune response in a particular patient (88). These include in vitro T-cell stimulation assays, tetramer analysis and so on. Computational algorithms to determine the immunogenicity of neoepitopes have been developed, but their reliability in practice remains uncertain (89).

Historically, MHC class II prediction has proven less accurate than class I prediction, which may in part be due to the nature of the binding pocket of MHC class II molecules (90). Poor predictive results have led to it not being used as routinely in oncology applications, although improvements in performance have been recently obtained (91). Recent pre-clinical work with three different murine models demonstrated that a large fraction of mutations can serve as MHC class II epitopes and that these can be used in a vaccination strategy to lead to tumor rejection (92). CD4⁺ T cells that recognize neoantigens (i.e. MHC class II epitopes) have been found in patients with melanoma and cholangiocarcinoma (93, 94). Although MHC class I-restricted CD8⁺ T cells are required for the efficacy of immune checkpoint blockers (75), the importance of class I versus class II epitopes in immune recognition and response to immune manipulations remains to be determined.

The role of neoantigens in mediating response to immune checkpoint blockade

In 2013, Schumacher’s group presented a case report and some preliminary data in which they demonstrated the existence of a neoantigen-specific T-cell population in a partial responder to anti-CTLA4 therapy (95). Rosenberg’s group similarly demonstrated that neoantigens mediated responses in three patients who received adoptive T-cell therapies, suggesting that neoantigens can lead to responses to a broad class of immunotherapies (96). Schreiber’s group demonstrated in murine model systems that neoantigens mediated response to both anti-CTLA4 and anti-PD-1 therapies (75). Moreover, vaccination against these epitopes produced results similar to checkpoint blockade in mice.

Our group made the first direct link between mutation load and outcome to immune checkpoint blockade in 2014 by examining a series of outlier responses to anti-CTLA4 therapy in patients with melanoma (55). We determined that the mutation load was associated with outcome in both a discovery set of patients and a subsequent validation set. Shortly thereafter, we reported that the mutation load also predicted the outcome of patients treated with anti-PD-1 therapy in NSCLC (56). After we first showed that the mutational load predicts clinical benefit from immune checkpoint blockade, at least three other studies subsequently confirmed our findings (64, 80, 87).

In the NSCLC group of patients from our study, those with evidence of smoking-related changes in their genome appeared to have improved outcomes as did those with alterations in the DNA damage response. Subsequently, a group from the Johns Hopkins University presented an elegant study linking patients who had mismatch-repair defects with response to anti-PD-1 therapy in colon and uterine cancers (64). As discussed above, patients with mismatch-repair defects have an increased mutational load. Shortly thereafter, another group independently verified that mutational load was associated with response to anti-CTLA4 therapy (80). Together, data from our group and others—in melanoma and in NSCLC, colon and uterine cancer—show that an increased mutation load is correlated with response to immune checkpoint blockade therapy. Moreover, these data indicate the importance of environmental exposure (UV irradiation and smoking) and DNA-repair defects such as microsatellite instability on the production of neoantigens.

The mechanistic link between high mutation load and response to checkpoint blockade involves an increase in the number of neoantigens as mutation numbers rise. In our reports on anti-CTLA4 therapy in melanoma and anti-PD-1 therapy in NSCLC, we identified T-cell responses to a number of computationally predicted neoantigens (55, 56). Moreover, the level of neoantigen-responsive CD8⁺ T cells corresponded to the tumor burden (56). In the melanoma report, we subsequently identified a series of recurrent peptide features that occurred only in responders in both cohorts of patients we examined, some of which shared homology to micro-organisms responsible for specific infections (see below).

However, analysis by Van Allen et al. in a group of melanoma patients treated in Germany did not find the same set of recurrent tetramers in their long-term responding patients (80). This may have resulted from differences in clinical or biological features between the cohorts or may suggest that identifying broadly applicable neoantigen signatures will require larger number of patients, as neoantigen space (the number of possible mutated peptides of length 9, i.e. 20⁹) is quite large. Alternatively, Van Allen et al. used NetMHCpan, a method that relies on structural inference rather than only on experimental data to call neoantigens (80). It is interesting that they called more neoantigens than mutations in their analysis, which suggests the presence of noise in the data set, a hypothesis that is consistent with their observation that their results are not dependent on patient-specific HLA identity.

Reanalysis of recent pre-clinical evidence strongly suggests that some cross-reactivity between neoantigens and micro-organisms may play a critical role in response to therapy and suggests the possibility of shared features in antigens in patients who respond (see below) (97, 98).

Although neoantigens appear to play a critical role in mediating the response to checkpoint blockade, other types of tumor-specific antigens may still contribute in recognizing tumors as foreign, once the brakes are released (42). Antigens that T cells recognize in patients appear to include cancer–testis antigens, tissue differentiation antigens or over-expressed genes (37, 99). As discussed previously, work by Lennerz et al. suggests that although CTLs against neoantigens and other antigens may be present, serial sampling suggests the response to neoantigens is dominant (48). However, comprehensive analysis for all types of antigens has not been performed for patients receiving checkpoint blockade yet. This will be a key area to investigate as the field moves forward.

A relationship between neoantigens and epitopes from infectious agents?

Two recent pre-clinical studies in murine models suggest that the gut microbiome may play an important role in mediating the response to checkpoint blockade (97, 98). In the study from Zitvogel’s group, response to anti-CTLA4 therapy in a murine sarcoma model depended on the presence of a species of Bacteriodes and was abolished with antibiotic treatment or absent in germ-free mice (98). Moreover response to anti-CTLA4 therapy could be restored by gavage with Bacteriodies, immunization with Bacteriodies polysaccharides or transfer of Bacteriodes-specific T cells.

These data suggest that possible mimicry may occur between commensal/pathogenic micro-organisms and neoantigens and may underlie differences in response to therapy between patients. Importantly, the ability of adoptive transfer of bacteria-specific T cells to impart the clinical benefit of anti-CLTA4 implies some involvement of T-cell cross-reactivity that we had posited to be a possibility (55). Indeed, our prior work in melanoma suggested that epitopes found in responding patients were homologous to a number of microbes and viruses, including herpes family viruses and gut microbiota (55). Further, it has recently been reported that virus-specific CD8⁺ T cells infiltrate melanoma (100).

These data suggest that there may be cross-reactivity between T cells that recognize prior infectious epitopes and then, subsequently, recognize neoantigens in a malignancy. In fact, recent work by Garcia’s group using genetically encoded degenerate peptide libraries has demonstrated that TCRs with a known specificity for a self-derived peptide presented on MHC class II can cross-react with a number of related peptides from environmental entities (101). They noted that many cross-reacting peptides were homologous to microbial antigens and, further, that core sequences of tetramers or pentamers dictated T-cell recognition.

Concluding remarks

Here, we have summarized some recent impactful findings on cancer neoantigens. The rapid progress of this field has reinvigorated interest in the field of immunogenetics. Although immunology was originally a field with a rich history of genetic analysis (i.e. VDJ recombination, somatic hypermutation, etc.), it has, in the last decade, strayed away from genetics. The great interest and significant effort now being put forth to study neoantigens highlight the potential benefits that can be reaped from thoughtful, ‘out-of-the-box’ genetic approaches to immunologic problems.

Funding

T.A.C. is funded by the Pershing Square Sohn Cancer Alliance, the STARR Cancer Consortium, the Frederick Adler Chair and Stand Up to Cancer.