Tumor neoantigens: from basic research to clinical applications

Tumor neoantigen, or tumor-specific antigen (TSA), is the repertoire of peptides that displays on the tumor cell surface and could be specifically recognized by neoantigen-specific T cell receptors (TCRs) in the context of major histocompatibility complexes (MHCs) molecules [1‐5]. From an immunological perspective, tumor neoantigen is the truly foreign protein and entirely absent from normal human organs/tissues. For most human tumors without a viral etiology, tumor neoantigens could derive from a variety of nonsynonymous genetic alterations including single-nucleotide variants (SNVs), insertions and deletions (indel), gene fusions, frameshift mutations, and structural variants (SVs) [2, 5]. For virally associated tumors, such as human papillomavirus (HPV)–related cervical or oropharyngeal cancer, Merkel cell polyomavirus (MCPyV)–related Merkel cell carcinoma (MCC) and Epstein-Barr virus (EBV)–related head and neck cancers, any epitopes derive from open reading frames (ORFs) in the viral genome also contribute to the potential source of neoantigens [6‐8]. In recent years, emerging evidence has suggested that neoantigens play a pivotal role in tumor-specific T cell-mediated antitumor immunity. As summarized in several elegant reviews [2, 4, 5, 9], this evidence includes, but is not limited to, (1) the occurrence of antitumor immune response via T cell recognition of neoantigen, (2) the relationship between tumor mutation/neoantigen burden and clinical outcomes to immune checkpoint blockade, and (3) the promising antitumor effects of therapeutic vaccines or adoptive T cell transfer based on neoantigen.

Unlike neoantigens, another two common types of tumor antigens, named tumor-associated antigens (TAAs) and cancer-germline antigens (CGAs), are not only expressed on the tumor cell surface, they would also be found on healthy or immune-privileged tissues (especially reproductive tissues including testes, fetal ovaries, and trophoblasts) with low levels of expression [3, 10, 11]. Therapeutic vaccines based on the TAAs or CGAs have obtained dismal results mainly due to the central and peripheral tolerance mechanisms [12]. Moreover, high-affinity TCRs for TAAs are preferentially depleted because of positive selection, and the affinities of the remaining TCRs for TAAs are lower than those for neoantigens and other foreign antigens [13]. In addition, since TAAs or CGAs could still have low levels of expression in normal tissues, targeting them may result in severe autoimmune toxicities related to immune activation in non-target tissues, such as severe hepatitis, colitis, rapid respiratory failure, renal impairment, and even treatment-related death [14].

Theoretically, neoantigen is an ideal immunotherapy target because they are distinguished from germline and could be recognized as non-self by the host immune system [5]. Neoantigen-specific immune reactions are not easily subject to complex immune tolerance mechanisms. Additionally, it may be less likely to trigger autoimmunity because they do not express on normal cells. Neoantigen-based therapeutic personalized vaccines and adoptive T cell transfer have shown promising preliminary results. Furthermore, recent studies suggested the significant role of neoantigen in immune escape, immunoediting, and sensitivity to immune checkpoint inhibitors.

In this review, we systematically summarize the recent advances of understanding and identification of tumor-specific neoantigens and its role on current cancer immunotherapies. We also discuss the ongoing development of strategies based on neoantigens and its future clinical applications. We do hope that this review could help us better understand the ongoing development of strategies based on neoantigens and its future clinical applications.

Currently, it is well-known that the immune system possesses an extraordinary ability to distinguish self from non-self, and recognize and target non-self antigens on abnormal cells [4]. However, our understanding of tumor neoantigens and its important role in antitumor immune response is a long and tortuous process (Fig. 1). We can go back to the late nineteenth century when William Coley, father of cancer immunotherapy, firstly attempted to leverage the patient’s immune system to treat cancer [15, 16]. In spite of the tremendous outcomes in individual cases, his findings were abandoned due to the huge success of chemotherapy and radiotherapy in control of various cancers. The investigation of the immune system in carcinogenesis and control of tumor growth during progression has recurred in the early part of the twentieth century [17]. In 1943, Gross et al. firstly showed that mice could be protected against subsequent re-exposure from tumor cells after surgical removal of the same tumors [18]. They also found that analogous protection against tumor cells could be induced by first exposing mice to lethally irradiated tumor cells. Their findings revealed that the immune system can recognize and eliminate malignant cells. Ten years later, another group further found that mice were immune against a second challenge with the same tumor cells after resection of carcinogen-induced tumors, supporting the idea of the existence of antitumor immunity [19]. Nevertheless, the nature of antigens that could substantially trigger antitumor immune response was unclear during this period. Several decades later, De Plaen and colleagues reported a significant finding that antitumor T cells could recognize aberrant peptides derived from tumor-specific mutations in a methylcholanthrene (MCA)-induced mouse tumor model [20]. They further identified the first T cell–recognized neoantigen by using a cDNA library screen. After that, a series of neoantigens derived from somatic mutations were identified in various human tumors including melanoma and renal cell carcinoma [21, 22]. Another significant advance in our understanding of neoantigens occurred in the early twenty-first century, when Huang et al. found nearly complete regression in a melanoma patient after infusion of a cell product with a high proportion of neoantigen-reactive T cells [23] and Lennerz et al. reported that the T cells of the patient with melanoma were reactive against five mutated peptides resulting from somatic point mutations and T cells against mutated epitopes was clearly predominated over the response to TAAs [24]. Similarly, the Rosenberg group found that neoantigen-specific T cells could persist at high levels in both the tumor and peripheral blood 1 month after adoptive transfer in a patient with melanoma that experienced a complete regression following adoptive transfer of ex vivo–expanded autologous tumor-reactive tumor-infiltrating lymphocytes (TILs)[25]. Collectively, these studies provided initial evidence that neoantigens play a vital role in the naturally occurring antitumor T cell response. With the advent and wide application of next-generation sequencing (NGS) technology, we have obtained a more and more profound understanding on tumor neoantigens. In 2012, two research groups independently and firstly applied NGS technology to identify immunogenic neoantigens in mouse tumor models [26, 27] and reported the protective effects of neoantigen vaccines in B16 tumor model [27]. Soon after, NGS technology was widely applied to explore T cell recognition of neoantigens in cancer patients [28]. Subsequent studies indicated that T cell responses against mutated antigens are frequently observed within TIL products in types of cancers including melanoma, lung cancer, colorectal cancer, cholangiocarcinoma, and squamous cell carcinoma of the head and neck [29‐33]. Nonetheless, whether the fraction of patients with detectable neoantigen-specific T cell responses is comparable across these tumor types remains largely unknown at this moment. Recent advance in our understanding of neoantigens is derived from the research of immune checkpoint inhibitors targeting cytotoxic T lymphocyte-associated protein 4 (CTLA-4) and programmed cell death protein 1 (PD-1) on T cells. The cancer-immunity cycle indicates that T cells recognized neoantigens displayed by MHCs on tumor cells is the first step to eliminate the established tumor via using immune checkpoint inhibitors [34, 35]. Mechanically speaking, there should be a significant correlation between tumor neoantigen load and clinical efficacy of immune checkpoint inhibitors. This relationship was also demonstrated by several elegant studies [33, 36‐38]. However, some cancers with high level of neoantigen load are not as sensitive to immune checkpoint inhibitors as expected. It is likely to be associated with the neoantigen clonality [39] but true mechanism of this phenomenon is not fully understood, suggesting that we still need to pay more effort to deepen our understanding of tumor neoantigens and its role in antitumor immune response.

The genetic changes harbored by a neoantigen could transform a self-protein into a functionally non-self-protein. The neoepitopes being presented by MHC-I could then drive immunogenic responses through several potential mechanisms. First, neoepitopes derived from mutations could improve MHC-I binding affinity and then result in the presentation of MHC-I ligands that would ordinarily not be presented during T cell selection [40]. Second, neoepitopes may increase the stability of the TCR–MHC-I interaction even with similar binding affinities as wild-epitopes and can induce different immune responses [41]. Third, mutation-induced changes to flanking amino acid residues significantly interfere the presentation of MHC-I viral epitopes. Theoretically, a mutation harbored within a neoantigen could drive the presentation of an adjacent unmutated MHC-I ligand that has escaped immune tolerance due to poor processing [42, 43].

Neoepitopes can be identified in various ways. Prior to the advent of massively parallel NGS, labor-intensive individual cDNA library screening was used to screening T cell–reactive neoepitopes [24]. When NGS became a routine technique, the ability to identify tumor-specific genetic mutations altering the protein coding regions became rapid and high throughput, facilitating neoantigen prediction. This is accomplished by applying machine learning algorithms that model aspects of the MHC-I processing and presentation pathway to patient tumor exome data to predict potential neoepitope targets [44, 45]. Predicted neopeptides can next be synthesized and tested for reactivity by autologous T cells using various assays such as ELISPOT, fluorescently labeled HLA tetramers, or barcode-labeled HLA multimers [46]. A classic procedure for identification of neoepitopes was shown in Fig. 2.

Several machine learning–based epitope prediction tools are available, such as NetMHCpan [47], NetMHCIIpan [48], MHCflurry [49], ConvMHC [50], PLAtEAU [51], and NetCTLpan [52]. For a neopeptide to become a neoepitope, two properties must be fulfilled: the peptide must be processed and presented by HLA, and the presented peptide must be recognized by a suitable T cell. Therefore, although these approaches show immense potential, current neoepitope prediction methods based on sequencing and predictions of epitope processing and presentation result in a low rate of validation. Bjerregaard et al. summarized published data from 13 publications on human neopeptides originating from single amino acid substitutions for which T cell reactivity had been experimentally tested. Less than 3% were reported to elicit the T cell response [53]. One major reason may be that the machine learning algorithms are highly dependent on the datasets available for training and testing [54]. As a widely used resource, The Immune Epitope Database and Analysis Resource (IEDB) hosts a database of experimentally validated epitopes. But its datasets of validated T cell epitopes found in databases are almost entirely formed of epitopes from bacteria or viruses and were not obtained by standardized experimental methodologies in cancer models [55].

Recent developments in HLA peptidomics for class I and II HLA molecules have been relevant for the improvements in the available epitope datasets [56, 57]. Abelin et al. set up their own dataset (> 24,000 peptides) and identified thousands of peptides bound to 16 different HLA class I alleles to quantify the contribution of factors critical to epitope presentation, such as protein cleavage and gene expression [58]. A commercial platform, EDGE (Epitope Discovery in cancer GEnomes) was constructed on deep learning to a large (N = 74 patients) HLA peptide and genomic dataset from various human tumors and could increase the positive predictive value of HLA antigen prediction by up to nine-fold [59]. Another strategy to improve the ability to predict neoepitope was the integration of potential immunogenicity assessments to the prediction process. MuPeXI algorithm ranks predicted neoepitopes by a priority score that is based on inferred abundance, MHC binding affinity, and an immunogenicity score based on similarity to non-mutated wild-type peptide [60] EpitopeHunter algorithm, which integrates RNA expression with immunogenicity prediction algorithm based on the hydrophobicity of the TCR contact region [61, 62]. Neopepsee algorithm using a machine learning algorithm trained on epitope features, including antigen processing and presentation, amino acid characteristics, the binding difference between wild-type and mutant epitope, and similarity to known epitopes, to predict the immunogenicity and reduce the false-positive rate [63]. In conclusion, to maximizing the probability of identifying clinically relevant neoepitopes, multiple method epitope prediction and advanced neoepitope quality metrics are warranted.

Cancer immunoediting is a conceptual framework integrating the immune system’s dual host-protective and tumor-promoting roles. During cancer immunoediting, the host immune system shapes tumor fate in three phases: elimination, equilibrium, and escape [64]. Decades of researches have revealed the dual role of the immune system in tumorigenesis. Recent work on cellular or animal model and clinical study on one cancer patient case [26, 64‐67] have shown unequivocal evidence that the immune system can facilitate cellular transformation, prevent or control tumor outgrowth and shape the immunogenicity of tumors [68‐70]. Studies using tumor exome sequencing to predict candidate neoantigens provide insights into the contribution of mutated peptide antigens encoded by somatic mutations to tumor antigenicity in human cancers [28, 71].

Interactions between the immune system and tumors have clear functional significance for tumor control, as the immune system exerts evolutionary pressure on highly immunogenic tumor clones through the process of immunoediting [26, 65, 72], and antitumor immune responses can be enhanced therapeutically by agents such as immune checkpoint inhibitors [73]. However, until recently, several multi-omics studies demonstrated the direct impact from immune pressure and immune editing on clonal evolution of tumor cells in metastases, benefit from the development of bioinformatic approaches on neoantigen prediction and on tumor evolutional and metastasis model based on somatic mutational landscape from lesions from different space and time [74‐79]. A total of 258 samples from different regions of 88 early stage, treatment-naïve NSCLC were examined in the Tracking Cancer Evolution through Therapy (TRACERx) NSCLC study [77]. Similar to genomic heterogeneity, the immunological landscapes of different regions of the same tumor can vary dramatically. An increase in the ratio of observed-to-expected neoantigens was noted from clonal to subclonal mutations among tumors with a low level of immune cell infiltration, which possibly reflects an ancestral immune-active microenvironment that has subsequently become cold. Tumors with high or heterogeneous levels of immune cell infiltration had significantly lower levels of expressed clonal neoantigens than those with limited levels of infiltration (median 29% and 35% versus 41%; P < 0.01). The pattern of neoantigen quantity and/or quality changes reflecting the immunoediting was also observed in other solid tumors including pancreatic cancer [79], colorectal cancer [74], melanoma [75], and glioblastomas [78].

Tumor cells could evolve several mechanisms to escape immune responses. On the one hand, cancers can hijack mechanisms developed to limit inflammatory and immune responses and escape the immuno-elimination. On the other hand, the metabolic or genetic alterations of tumor cells can render themselves invisible to the immune system or can favor the generation of an extracellular milieu preventing immune cell infiltration or cytotoxicity. And the tumor clone possessing neoantigens with potent immunogenicity tends to be eliminated during immunoediting. There are several ways through which immunogenic neoantigens could escape from immunological surveillance. At the DNA level, chromosomal instability–induced copy number alterations may drive the loss of neoantigens [77]. At the RNA level, the neoantigen expression could be decreased for the promoter hypermethylation, while epigenetics mechanism could not account for all the transcriptomic neoantigen depletion. Additional mechanisms of neoantigen transcription repression need elucidation [76, 77, 80]. At the protein level, the machinery to presenting antigen peptides could be disrupted by mutations affecting HLA heterozygosity, MHC stability, HLA enhanceosome, or neopeptide generation [77, 81]. All these different mechanisms through which the tumor hiding target to evade immune predation provide series potential clinical setting worth to be exploited [82].

Recent preclinical and clinical studies have shown that neoantigen-based approaches are able to induce robust antitumor immune responses in individual tumor microenvironment (TME). The two main approaches targeting tumor neoantigen that are well-established now include neoantigen-based cancer vaccines and neoantigen-based adoptive cell transfer (ACT) treatment. Also, combinational therapies employing both neoantigen-based approaches and immune checkpoint blockade (ICB) are underway to overcome ICB-induced immune resistance and maximize antitumor immune activity [83].

In conclusion, emerging evidence suggests that tumor neoantigen plays a pivotal role in immune escape, antitumor immune response and successful cancer immunotherapies. Both neoantigen-based vaccine and ACT treatments show very promising antitumor effect together with high specificity and safety in preliminary studies. Combinatorial approaches of neoantigen-based therapies with other immunotherapies (e.g., immune checkpoint inhibitors) and conventional treatments are ongoing, and the results are eagerly anticipated. With a better understanding of biological properties and role of neoantigen in antitumor immunity, there are abundant reasons to believe that neoantigen-based therapeutic strategies have a bright future in cancer immunotherapies.