Immunotherapy is currently transforming cancer treatment. Notably, immune checkpoint blockers (ICB) have shown unprecedented therapeutic successes in numerous tumor types, including cancers that were traditionally considered as nonimmunogenic. However, a significant proportion of patients do not respond to these therapies. Thus, early selection of the most sensitive patients is key, and the development of predictive companion biomarkers constitutes one of the biggest challenges of ICB development. Recent publications have suggested that the tumor genomic landscape, mutational load, and tumor-specific neoantigens are potential determinants of the response to ICB and can influence patients' outcomes upon immunotherapy. Furthermore, defects in the DNA repair machinery have consistently been associated with improved survival and durable clinical benefit from ICB. Thus, closely reflecting the DNA damage repair capacity of tumor cells and their intrinsic genomic instability, the mutational load and its associated tumor-specific neoantigens appear as key predictive paths to anticipate potential clinical benefits of ICB. In the era of next-generation sequencing, while more and more patients are getting the full molecular portrait of their tumor, it is crucial to optimally exploit sequencing data for the benefit of patients. Therefore, sequencing technologies, analytic tools, and relevant criteria for mutational load and neoantigens prediction should be homogenized and combined in more integrative pipelines to fully optimize the measurement of such parameters, so that these biomarkers can ultimately reach the analytic validity and reproducibility required for a clinical implementation. Clin Cancer Res; 22(17); 4309–21. ©2016 AACR.

Since their first introduction into the clinic and first approval in 2011 (1), immune checkpoint blockers (ICB) have transformed cancer treatment and allowed unprecedented improvements in overall survival (OS), progression-free survival (PFS), or overall response rates (ORR) in many aggressive diseases (2, 3). Most importantly, benefits of ICB have not been limited to the “traditional” immunogenic cancers, malignant melanoma and renal cell carcinoma (RCC), but have also been extended to other histologies classically described as “nonimmunogenic,” such as non–small cell lung cancer (NSCLC) or mismatch-repair–deficient colorectal cancer (MMR-deficient colorectal cancer; ref. 3). Despite these clear clinical advances, the biological mechanisms that underlie antitumor immunity and determine sensitivity to these agents, notably anti-programmed death receptor-1/-ligand 1 [anti–PD-(L)1] are still poorly understood. Moreover, a statistically significant proportion of patients, approximately 80%, among all tumor types included, still do not respond to these drugs, highlighting the urge for developing robust predictive biomarkers that would guide appropriate selection of patients. Recently, the tumor cell mutational burden has been correlated with clinical benefits of anti–PD-1 and anti-CTLA-4 therapy in various tumor types, including malignant myeloma (4, 5), NSCLC (6), and several DNA repair–deficient tumors (7–9). Predicted neoantigen load has also emerged as an interesting selection biomarker for predicting clinical benefit of these agents. Overall, a direct link between DNA repair deficiency, mutational landscape, predicted neoantigen load, and clinical activity of ICB is suggested.

In this review, we discuss the significance and the relevance of this correlation in solid tumors. We also provide critical insight into the methods and techniques that have been used for performing analyses of tumor mutational burden, predicted neoantigen load, and neopeptide formation. We further propose a comprehensive approach that would allow encompassing other potential predictive biomarkers for response to anti–PD-(L)1 inhibitors.

The original concept of immune surveillance, hypothesized in 1957 (10), and formally established in 1970 (11), postulated that the immune system alone could eliminate tumor cells in the early stages of carcinogenesis. Since then, this theory has been further enriched by the “immunoediting” notion (12), which describes how both innate and adaptive immunity contribute to carcinogenesis, notably by exerting a Darwinian selection pressure. Immunoediting classically consists of three distinct steps: (i) elimination: the innate and adaptive compartments coordinately drive immune rejection; (ii) equilibrium: through a clonal selection process, the dynamic balance between tumor and immune cells results in the emergence of specific tumor cell variants with increased resistance, which take advantage of acquired mutations; (iii) escape: the immune-resistant clones freely expand, circumventing both innate and adaptive immune responses.

A variety of mechanisms can facilitate tumor immune escape (Fig. 1). Among them, deregulation of immune checkpoint signaling has been observed in multiple malignancies (13–21). Immune checkpoints involve the interaction between a receptor expressed on T cells and its ligand located at the surface of antigen-presenting cells. This generates a costimulatory signal, which triggers either the activation or inhibition of T cells. Two major checkpoints regulate T-cell activation: (i) the CD28/CTLA-4 axis, which activates T cells upon engagement of CD28 with CD80 and CD86, and conversely inhibits T cells when CTLA-4 is engaged; and (ii) the PD-1 axis, which provides a strong inhibitory signal following binding of PD-L1 or PD-L2 to the PD-1 receptor (22). Contrary to CTLA-4, PD-1 is thought to act predominantly in the tumor microenvironment, where PD-L1 is overexpressed by multiple cell types, including dendritic cells, M2 macrophages, and tumor-associated fibroblasts (23).

Figure 1.

Mechanisms of immune escape in the tumor microenvironment. Several mechanisms, involving multiple immune components, contribute to tumor immune escape. (1) Immune recognition can be impaired following reduced expression of MHC class I molecules in malignant cells, resulting in decreased antigen presentation and consequently reduced detection by cytotoxic CD8+ T lymphocytes. (2) Cancer cells can activate immunosuppressive mechanisms by inducing immune cells' apoptosis through the expression of death signals (including FAS- and TRAIL-ligands). (3) Tumor cells release in the microenvironment a variety of immune-modulatory molecules that inhibit the immune system, such as IL6 and IL10, by inducing immunosuppressive Treg cells and MDSC, whereas the activity of cytotoxic CD8+ T cells and NK cells is inhibited. (4) This cytokine imbalance, combined with the secretion of TGFβ, COX-2, and PGE2, inhibits dendritic cell differentiation and maturation, thereby affecting antigen presentation and recognition by T cells. The release of additional immune modulators or metabolic regulators, such as IDO and arginase, also favors the establishment of an immunosuppressive tumor microenvironment. (5) Disrupted expression of immune checkpoint ligands by cancer cells provides coinhibitory signals to CD4+ and CD8+ T lymphocytes, preventing them from building a specific antitumor immune response. CCL, chemokine ligand; COX-2, cyclooxygenase-2; CXCL, chemokine (C-X-C motif) ligand; FAS-L, FAS-ligand; GM-CSF, granulocyte macrophage colony-stimulating factor; iDC, immature dentritic cell; IDO, indoleamine-2,3-deoxygenase; mDC, mature dentritic cell; MDSC, myeloid-derived suppressor cell; PD-1, programmed cell death 1; PD-L, programmed cell death ligand; PGE2, prostaglandin E2; TAN, tumor-associated neutrophil; TCR, T-cell receptor; Treg, regulatory T cells.

Figure 1.

Mechanisms of immune escape in the tumor microenvironment. Several mechanisms, involving multiple immune components, contribute to tumor immune escape. (1) Immune recognition can be impaired following reduced expression of MHC class I molecules in malignant cells, resulting in decreased antigen presentation and consequently reduced detection by cytotoxic CD8+ T lymphocytes. (2) Cancer cells can activate immunosuppressive mechanisms by inducing immune cells' apoptosis through the expression of death signals (including FAS- and TRAIL-ligands). (3) Tumor cells release in the microenvironment a variety of immune-modulatory molecules that inhibit the immune system, such as IL6 and IL10, by inducing immunosuppressive Treg cells and MDSC, whereas the activity of cytotoxic CD8+ T cells and NK cells is inhibited. (4) This cytokine imbalance, combined with the secretion of TGFβ, COX-2, and PGE2, inhibits dendritic cell differentiation and maturation, thereby affecting antigen presentation and recognition by T cells. The release of additional immune modulators or metabolic regulators, such as IDO and arginase, also favors the establishment of an immunosuppressive tumor microenvironment. (5) Disrupted expression of immune checkpoint ligands by cancer cells provides coinhibitory signals to CD4+ and CD8+ T lymphocytes, preventing them from building a specific antitumor immune response. CCL, chemokine ligand; COX-2, cyclooxygenase-2; CXCL, chemokine (C-X-C motif) ligand; FAS-L, FAS-ligand; GM-CSF, granulocyte macrophage colony-stimulating factor; iDC, immature dentritic cell; IDO, indoleamine-2,3-deoxygenase; mDC, mature dentritic cell; MDSC, myeloid-derived suppressor cell; PD-1, programmed cell death 1; PD-L, programmed cell death ligand; PGE2, prostaglandin E2; TAN, tumor-associated neutrophil; TCR, T-cell receptor; Treg, regulatory T cells.

Close modal

As opposed to historical immune-based approaches that were developed in traditionally immunogenic cancers, ICBs have allowed significant therapeutic successes in many solid tumors and hematologic malignancies. The anti-CTLA-4 ipilimumab (Yervoy, Bristol-Myers Squibb) was the first ICB to improve OS in malignant melanoma patients (1). In 2012, anti–PD-(L)1 therapies including the anti–PD-1 pembrolizumab (Keytruda, Merck), and the anti-PD-L1 atezolizumab (MPDL-3280A, Genentech/Roche), durvalumab (MEDI-4736, Astra Zeneca/MedImmune), and avelumab (MSB0010718C, Pfizer) entered clinical development. Very promising ORR in relapsing/refractory malignant melanoma, RCC, and NSCLC (3), associated with prolonged PFS and OS, led to their accelerated approval in 2014–2015, and the outstanding activity observed in several histologies (Supplementary Table S1) awarded them “drugs of the year” in 2013 (24). Since then, an exponential number of monotherapy or combination trials have been launched in multiple cancer types.

Contrary to immune escape, DNA repair deficiency has been successfully exploited as a therapeutic opportunity for more than 50 years with the use of traditional cytotoxic chemotherapies. If these DNA-damaging agents have initially been developed in a “one-size-fits-all” approach, DNA repair deficiencies are now being exploited in a much more targeted fashion, notably using targeted mechanism-based approaches, such as synthetic lethality (25–28).

DNA repair deficiency is one of the main drivers of genomic instability, a key hallmark of cancer (ref. 29; Table 1). It favors the accumulation of DNA lesions that can arise from two distinct processes: (i) exogenous lesions, resulting from exposure to mutagenic agents and carcinogens and (ii) endogenous defects, which arise as a consequence of cell metabolism and the inherent instability of DNA (30). Interestingly, some peculiar types of exogenous DNA damage are associated with specific patterns of mutations, also called mutational signatures. For example, the predominance of C-to-A transitions, due to the effect of the polycyclic hydrocarbons of tobacco smoke, is characteristically found in NSCLC (31). In melanoma, UV radiation creates pyrimidine dimers, which result in a high prevalence of C-to-T transitions on the untranscribed strand (32). Specific mutational signatures have also been reported in cancers with endogenous DNA damage repair defects, for example, BRCA1- or BRCA2-deficient high-grade serous ovarian and triple-negative breast cancers, which harbor frequent loss of heterozygosity (28, 33); MMR-deficient colorectal cancer (34), associated with a microsatellite instable phenotype and high mutational burden; and POLE-deficient endometrial cancers, which exhibit an ultra-mutated phenotype (7).

Table 1.

Type and frequency of DNA repair alterations in solid tumors

Alterations
Cancer typeGeneTypeFrequencyReferences
Non–small cell lung cancer BRCA1 Reduced mRNA and protein expression 44% (68) 
 FANCF Promoter methylation 14% (69) 
 ATM Somatic mutations 6% (69) 
 MSH2 Reduced protein expression 18%–38% (69) 
 ERCC1 Reduced protein expression 22%–66% (69) 
 RRM1 Loss of heterozygosity 65% (69) 
Small-cell lung cancer POLD4Reduced mRNA expression N.R. (70) 
Clear-cell renal cell carcinoma ATM Somatic mutations 3% (71) 
 NSB1 Somatic mutations 0.5%  
 MLH1 Homozygous deletion 3%–5% (72) 
 MSH2 Promoter hypermethylation N.R. (73) 
Urothelial carcinoma BRCA1 Somatic mutations 14% (74–76) 
 BRCA2 Somatic mutations 14%  
 PALB2 Somatic mutations 14%  
 ATM Somatic mutations 29%  
 MSH2 Loss of protein expression 3% (77) 
 ERCC2 Somatic mutations 12% (78) 
Head and neck cancer FANCBPromoter methylation 31% (79) 
 FANCFPromoter methylation 15%  
 FANCJ Reduced protein expression (IHC) N.R.  
 FANCM Reduced protein expression (IHC) N.R.  
 BRCA1 Reduced protein expression (IHC) N.R.  
 BRCA2 Reduced protein expression (IHC) N.R.  
 FANCD2 Reduced protein expression (IHC) N.R.  
Ovarian cancer BRCA1/BRCA2 Germline mutations 15% (80, 81) 
  Somatic mutations 35%  
  Promoter methylation 11%–35%  
 FANCF Promoter methylation N.R.  
 FANCD2 Reduced protein expression N.R.  
 BARD1 Germline mutations 6% (82) 
 BRIP1 Germline mutations 6%  
 PALB2 Germline mutations 6%  
 MRE11 Germline mutations 6%  
 RAD50 Germline mutations 6%  
 RAD51C Germline mutations 6%  
 NSB1 Germline mutations 6%  
 MSH6 Inactivating mutations 6% (82) 
Triple-negative breast cancer BRCA1 Germline mutations 5%–10% (80, 83) 
 BRCA2 Somatic mutations 10%  
Gastric cancer MLH1 Loss of protein expression (IHC) 18% (84) 
  Promoter hypermethylation 15%  
 MSH2 Loss of protein expression (IHC) 3%  
MMR-deficient colorectal cancer MRE11 Somatic mutations 75% (34, 85, 86) 
 RAD50 Somatic mutations 21%–46%  
 BRCA2 Somatic mutations 2%  
 MSH3 Somatic mutations 22%–51% (34, 85, 86) 
 MSH6 Somatic mutations 9%–38%  
 MLH3 Somatic mutations 9%–28%  
 POLD3 Somatic mutations 37% (34, 85, 86) 
Hepatocellular carcinoma NSB1 Somatic mutations 10% (87) 
 MSH2 Promoter hypermethylation 25% (88, 89) 
  Reduced protein expression 18%  
 PMS2 Promoter hypermethylation 15%  
 MLH1 Promoter hypermethylation 8%  
  Reduced protein expression 38%  
Biliary tract cancer MSH2 Loss of protein expression (IHC) 7% (90, 91) 
 MSH6 Loss of protein expression (IHC) 7%  
 MLH1 Loss of protein expression (IHC) 1.5%  
 PMS2 Loss of protein expression (IHC) 1.5%  
Prostate cancer BRCA2 Homozygous deletion/heterozygous deletion/frameshift mutation 14% (92, 93) 
 ATM Frameshift mutation 12%  
 PALB2 Frameshift mutation 4%  
 CHK2 Homozygous deletion 4%  
 FANCA Homozygous deletion 6%  
 BRCA1 Homozygous deletion 2%  
 MRE11 Frameshift mutation 2%  
 NSB1 Frameshift mutation 2%  
 MLH3 Frameshift mutation 4% (92, 93) 
Endometrial cancer MLH1 Promoter hypermethylation 30% (7, 94) 
 POLE Somatic mutations 10% (7, 94) 
Pancreatic cancer BRCA2 Germline mutations 1.5% (68, 95) 
 MSH2 Loss of protein expression (IHC) 15% (96) 
 MSH6 Loss of protein expression (IHC) 15%  
 MLH1 Loss of protein expression (IHC) 15%  
 PMS2 Loss of protein expression (IHC) 15%  
Alterations
Cancer typeGeneTypeFrequencyReferences
Non–small cell lung cancer BRCA1 Reduced mRNA and protein expression 44% (68) 
 FANCF Promoter methylation 14% (69) 
 ATM Somatic mutations 6% (69) 
 MSH2 Reduced protein expression 18%–38% (69) 
 ERCC1 Reduced protein expression 22%–66% (69) 
 RRM1 Loss of heterozygosity 65% (69) 
Small-cell lung cancer POLD4Reduced mRNA expression N.R. (70) 
Clear-cell renal cell carcinoma ATM Somatic mutations 3% (71) 
 NSB1 Somatic mutations 0.5%  
 MLH1 Homozygous deletion 3%–5% (72) 
 MSH2 Promoter hypermethylation N.R. (73) 
Urothelial carcinoma BRCA1 Somatic mutations 14% (74–76) 
 BRCA2 Somatic mutations 14%  
 PALB2 Somatic mutations 14%  
 ATM Somatic mutations 29%  
 MSH2 Loss of protein expression 3% (77) 
 ERCC2 Somatic mutations 12% (78) 
Head and neck cancer FANCBPromoter methylation 31% (79) 
 FANCFPromoter methylation 15%  
 FANCJ Reduced protein expression (IHC) N.R.  
 FANCM Reduced protein expression (IHC) N.R.  
 BRCA1 Reduced protein expression (IHC) N.R.  
 BRCA2 Reduced protein expression (IHC) N.R.  
 FANCD2 Reduced protein expression (IHC) N.R.  
Ovarian cancer BRCA1/BRCA2 Germline mutations 15% (80, 81) 
  Somatic mutations 35%  
  Promoter methylation 11%–35%  
 FANCF Promoter methylation N.R.  
 FANCD2 Reduced protein expression N.R.  
 BARD1 Germline mutations 6% (82) 
 BRIP1 Germline mutations 6%  
 PALB2 Germline mutations 6%  
 MRE11 Germline mutations 6%  
 RAD50 Germline mutations 6%  
 RAD51C Germline mutations 6%  
 NSB1 Germline mutations 6%  
 MSH6 Inactivating mutations 6% (82) 
Triple-negative breast cancer BRCA1 Germline mutations 5%–10% (80, 83) 
 BRCA2 Somatic mutations 10%  
Gastric cancer MLH1 Loss of protein expression (IHC) 18% (84) 
  Promoter hypermethylation 15%  
 MSH2 Loss of protein expression (IHC) 3%  
MMR-deficient colorectal cancer MRE11 Somatic mutations 75% (34, 85, 86) 
 RAD50 Somatic mutations 21%–46%  
 BRCA2 Somatic mutations 2%  
 MSH3 Somatic mutations 22%–51% (34, 85, 86) 
 MSH6 Somatic mutations 9%–38%  
 MLH3 Somatic mutations 9%–28%  
 POLD3 Somatic mutations 37% (34, 85, 86) 
Hepatocellular carcinoma NSB1 Somatic mutations 10% (87) 
 MSH2 Promoter hypermethylation 25% (88, 89) 
  Reduced protein expression 18%  
 PMS2 Promoter hypermethylation 15%  
 MLH1 Promoter hypermethylation 8%  
  Reduced protein expression 38%  
Biliary tract cancer MSH2 Loss of protein expression (IHC) 7% (90, 91) 
 MSH6 Loss of protein expression (IHC) 7%  
 MLH1 Loss of protein expression (IHC) 1.5%  
 PMS2 Loss of protein expression (IHC) 1.5%  
Prostate cancer BRCA2 Homozygous deletion/heterozygous deletion/frameshift mutation 14% (92, 93) 
 ATM Frameshift mutation 12%  
 PALB2 Frameshift mutation 4%  
 CHK2 Homozygous deletion 4%  
 FANCA Homozygous deletion 6%  
 BRCA1 Homozygous deletion 2%  
 MRE11 Frameshift mutation 2%  
 NSB1 Frameshift mutation 2%  
 MLH3 Frameshift mutation 4% (92, 93) 
Endometrial cancer MLH1 Promoter hypermethylation 30% (7, 94) 
 POLE Somatic mutations 10% (7, 94) 
Pancreatic cancer BRCA2 Germline mutations 1.5% (68, 95) 
 MSH2 Loss of protein expression (IHC) 15% (96) 
 MSH6 Loss of protein expression (IHC) 15%  
 MLH1 Loss of protein expression (IHC) 15%  
 PMS2 Loss of protein expression (IHC) 15%  

NOTE: Genes in blue are related to DSBR, in green to MMR, in red to NER, in orange to nucleotide synthesis, and in gray to DNA replication. Genes marked with an asterisk refer to data reported in cell lines only. Mutations or alterations in genes related to cell cycle are described in Supplementary Table S2.

Abbreviations: DSBR, double-strand break repair; NER, nucleotide-excision repair; N.R., not reported.

It is somehow intuitive that the presence of high tumor mutational burden can increase the likelihood of neoantigens formation, and that the most mutated tumors may also be the most immunogenic ones (35). However, if high mutational burden has repeatedly been associated with response and improved outcome on ICB therapy, it would be naïve to conclude on that basis that there is a general correlation between DNA repair deficiency and sensitivity to anti–PD-(L)1 (Fig. 2), the reality being much more complex.

Figure 2.

DNA repair defects and their association with anti–PD-(L)1 efficacy in solid tumors. A, representation, per tumor type, of the median frequency of DNA repair deficiency (yellow pie charts) and the median efficacy of anti-PD-(L)1 (blue pie charts). For each histology, the median rate of DNA repair defects was calculated on the basis of literature data (see Table 1 for raw data). When DNA repair defects in distinct pathways were mutually exclusive, the sum of their frequency was taken; when overlaps were observed between several DNA repair defects, the median of all DNA repair defects was chosen. The frequency of additional defects in other genes relevant for DNA repair (i.e., genes involved in cell cycle regulation or DNA replication) were also evaluated and are depicted on the side of the pie chart graphs. Tumor types resulting from exposure to a mutagenic agent are highlighted by a skull. ORR reported in phase I, II, or III trials performed in the corresponding histologies were taken for estimating the efficacy of anti–PD-(L)1 inhibitors (see Supplementary Table S1 for raw data). The data cut-off for collecting anti–PD-(L)1 efficacy was January 2016. B, scatter plot illustrating the lack of statistically significant correlation between DNA repair mutation frequency and response to anti–PD-(L)1 therapies, highlighting the need to take into account additional parameters for predicting response to these drugs. DSBR, double-strand break repair; HPV, human papillomavirus; NER, nucleotide-excision repair; N.R., not reported; TNBC, triple-negative breast cancer.

Figure 2.

DNA repair defects and their association with anti–PD-(L)1 efficacy in solid tumors. A, representation, per tumor type, of the median frequency of DNA repair deficiency (yellow pie charts) and the median efficacy of anti-PD-(L)1 (blue pie charts). For each histology, the median rate of DNA repair defects was calculated on the basis of literature data (see Table 1 for raw data). When DNA repair defects in distinct pathways were mutually exclusive, the sum of their frequency was taken; when overlaps were observed between several DNA repair defects, the median of all DNA repair defects was chosen. The frequency of additional defects in other genes relevant for DNA repair (i.e., genes involved in cell cycle regulation or DNA replication) were also evaluated and are depicted on the side of the pie chart graphs. Tumor types resulting from exposure to a mutagenic agent are highlighted by a skull. ORR reported in phase I, II, or III trials performed in the corresponding histologies were taken for estimating the efficacy of anti–PD-(L)1 inhibitors (see Supplementary Table S1 for raw data). The data cut-off for collecting anti–PD-(L)1 efficacy was January 2016. B, scatter plot illustrating the lack of statistically significant correlation between DNA repair mutation frequency and response to anti–PD-(L)1 therapies, highlighting the need to take into account additional parameters for predicting response to these drugs. DSBR, double-strand break repair; HPV, human papillomavirus; NER, nucleotide-excision repair; N.R., not reported; TNBC, triple-negative breast cancer.

Close modal

The description of a correlation between mutational load and response to ICB was allowed by recent advances in next-generation sequencing (NGS) technologies, notably whole-exome sequencing (WES) and RNA-sequencing (RNA-seq). High mutational load, defined as >100 nonsynonymous single-nucleotide variants (nsSNV) per exome, was first associated with clinical benefit in melanoma patients treated with anti–CTLA-4 therapy (4, 5). Subsequently, Rizvi and colleagues correlated high mutational load (defined as >178 nsSNVs per exome) and durable clinical benefit in two partially independent cohorts of NSCLC patients receiving pembrolizumab (6). Of note, the study reported a significantly increased ORR in tumors exhibiting a smoking molecular signature. Moreover, in responders showing the highest mutational burden, specific mutations were identified in DNA repair genes, including POLD1, POLE, MSH2, BRCA2, RAD51C, and RAD17, thus supporting that DNA repair defects can increase tumor immunogenicity by favoring somatic mutations. Consistently, later findings showed higher response rates to anti–PD-1 therapy in MMR-deficient tumors (9, 36), and in BRCA2-mutated melanoma (37). Interestingly, in the latter study, mutational load did not correlate with tumor response but was associated with improved patient survival only, highlighting the role of additional factors influencing early tumor response and long-term OS.

Now, the major challenges that remain to be addressed to improve robustness of mutational burden include the definition of optimal tumor purity and sequencing depth, as well as the threshold for defining “high” and “low” mutational burden. Indeed, there is a significant overlap in mutation range between responders and nonresponders (4, 5): some patients still benefit from ICB despite very low mutation rates, and conversely, high mutational load does not always correlate with response. This is best illustrated by Hodgkin lymphoma, which is highly sensitive to PD-1 blockade (38) despite carrying virtually no mutation. Mutational signatures, that are functional readouts of the past and current disease biology in terms of DNA damage and DNA repair, could represent an additional genomic determinant of response to ICB (35, 39). Their use, combined with evaluation of mutational load and detection of mutations in DNA repair genes, may therefore allow better stratification of patients and identify ICB-sensitive tumors.

Importantly, the above-described analyses of the mutational landscape only provide an “instantaneous and descriptive” picture of a tumor genome. Even mutational signatures, in some cases, might exclusively reflect previous DNA repair deficiencies and may not be relevant markers of the actual DNA repair status of the tumor. It is therefore crucial to assess the potential for these mutations to functionally enhance antitumor immune responses by creating immunogenic neoantigens.

Two main classes of tumor antigens are classically described: (i) tumor-associated antigens (TAA), which are nonmutated self-antigens that are aberrantly expressed by cancer cells following genetic and epigenetic alterations, and (ii) tumor-specific antigens, which are neoantigens that form as a result of nonsynonymous mutations and are generally unique to a tumor. Among these, the latter only have been consistently associated with antitumor T-cell reactivity and clinical efficacy of ICB (40).

Although we can anticipate that highly mutated tumors are more prone to form neoantigens, the stochastic nature of neoantigen generation calls for a functional validation, as all formed neoantigens may not be immunologically relevant. If it is obvious that nsSNVs represent a mine of immunogenic mutations, frameshift, splice site mutations, and intragenic fusions are also prone to generate neoepitopes when nonfunctional proteins are directed to the proteasome (41). The correlation between mutational burden and predicted neoantigen load (as defined by the number of neoantigens potentially presented by the MHC class I) has been achieved by creating bioinformatics analysis pipelines that model the key steps of the antigen presentation process (Fig. 3 and Table 2): (i) expression of mutated proteins that are processed by the proteasome, and produce neopeptides; (ii) translocation of the neopeptides through the endoplasmic reticulum and binding to the MHC class I molecule with a sufficient affinity to enable T-cell presentation; and (iii) recognition of the presented neoantigen by a T-cell clone able to detect it.

Figure 3.

Pipeline for the identification of immune-relevant neoantigens. The typical pipeline consists of six main steps: (1) Tumor mutational load and specific mutations are identified using WES or WGS. Additional techniques such as CGH or MSI-profiling might be of interest to evaluate genomic instability but have not been validated yet in this indication. Moreover, WES is always a required starting point as DNA sequence information is required for subsequent prediction tools. (2) Using RNA-seq, previously generated sequencing data are filtered for gene expression to restrict neoantigen prediction to the set of translated mutations (“expressed nsSNV”). Subsequently, predictions for (3) proteasomal processing and (4) TAP-mediated transport of peptides are completed using dedicated algorithms. (5) To predict binding of peptides on MHC class I molecules, the previously selected peptides are implemented in a dedicated software that infers binding affinity to HLA molecules according to the HLA type of the patient. (6) Eventually, the predicted peptides may be synthesized to test for T-cell reactivity in vitro using the MHC multimer technology. Key technologies most often used in the literature are highlighted in bold. Techniques exclusively used to measure genomic instability are presented in dotted rectangles. ARB, average relative binding; CGH, comparative genomic hybridization; SMM, stabilized matrix method; WGS, whole-genome sequencing.

Figure 3.

Pipeline for the identification of immune-relevant neoantigens. The typical pipeline consists of six main steps: (1) Tumor mutational load and specific mutations are identified using WES or WGS. Additional techniques such as CGH or MSI-profiling might be of interest to evaluate genomic instability but have not been validated yet in this indication. Moreover, WES is always a required starting point as DNA sequence information is required for subsequent prediction tools. (2) Using RNA-seq, previously generated sequencing data are filtered for gene expression to restrict neoantigen prediction to the set of translated mutations (“expressed nsSNV”). Subsequently, predictions for (3) proteasomal processing and (4) TAP-mediated transport of peptides are completed using dedicated algorithms. (5) To predict binding of peptides on MHC class I molecules, the previously selected peptides are implemented in a dedicated software that infers binding affinity to HLA molecules according to the HLA type of the patient. (6) Eventually, the predicted peptides may be synthesized to test for T-cell reactivity in vitro using the MHC multimer technology. Key technologies most often used in the literature are highlighted in bold. Techniques exclusively used to measure genomic instability are presented in dotted rectangles. ARB, average relative binding; CGH, comparative genomic hybridization; SMM, stabilized matrix method; WGS, whole-genome sequencing.

Close modal
Table 2.

Advantages and drawbacks of the available techniques to identify immunogenic mutations/neoantigens

Technique or softwarePlatformStrengthsWeaknessesRelevance for antigenome predictionReferences
WGS Mutational profiling Both coding and noncoding DNA sequences are analysed. Sequencing depth is usually low, which prevents detection of some subclonal mutations. ++ (97) 
WES Mutational profiling Provides high sequence coverage across exome, increasing reliability and ability to detect subclonal mutations. 
  • (i) Only covers the ≈1% coding regions of the genome.

  • (ii) Some mutations may be missed due to uneven capture efficiency across exons.

 
+++ (97) 
MSI profiling Microsatellite instability Several methods all well-validated. Only provides information on the microsatellite instability. (9) 
CGH Genomic instability Global picture of the overall genomic instability. 
  • (i) Only provides copy-number variations and translocations of large portions of the genome.

  • (ii) No access to the DNA sequence.

 
 
RNA-seq Expression profiling and coding mutation analysis 
  • (i) Focuses on translated mutations only, that are the most likely to have functional consequences.

  • (ii) Analysis not restricted to known genes: potential for discovering novel transcripts, splice variants or fusions.

  • (iii) Possibility to correlate mutational data with gene expression.

 
  • (i) Access to matched normal is key but cannot be achieved in many cases: hard to distinguish tumor-specific mutations from polymorphisms.

  • (ii) Limited calling of mutations within RNA species due to their low levels, either because of low level gene expression or because of mRNA stability.

 
+++ (97) 
NetCHOP 20SPCM (WAPP package)FragPredict (MAPPP package) Proteasomal processing prediction trained on in vitro data N.R. Predictions from in vitro data do not capture the full complexity of proteasomal processing. (98) 
PCleavage      
NetCHOPCterm Proteasomal processing prediction trained on in vivo data 
  • (i) In vivo data provide accurate prediction as predictions are made on the entire processing machinery (action of several proteasomes, cytosolic proteases…)

  • (ii) May also capture transport efficiency.

 
N.R. +++ (98, 99) 
PredTAP TAP transport prediction No comparative study available. No comparative study available. N.R. (98, 100) 
SVMTAP (WAPP package)      
SMM (stabilized matrix method) Allele-specific HLA binding affinity prediction N.R. (ii) Does not account for non-linearities and interdependencies between amino acids. (101–103) 
ARB average relative binding (matrix-based methods)      
NetMHC [artificial neural networks (ANN)-based method] Allele-specific HLA binding affinity prediction Nonlinear model. Does not allow prediction for all known HLA alleles. ++ (103, 104) 
NetMHCpan [Pan-specific artificial neural networks (ANN)-based method] Pan-specific HLA binding affinity prediction 
  • (i) Allows predictions to be made for all known HLA Class I alleles, including alleles for which no prediction is available with NetMHC.

  • (ii) NetMHCpan is the best-performing method for allele-specific HLA binding affinity prediction.

 
N.R. +++ (105) 
Athlates HLA typing N.R. 
  • (i) Early tool with lower accuracy than that of up-to-date tools.Restricted to the use of WES data

 
(98) 
Polysolver HLA typing 
  • (i) Provides improved retrieval and alignment of HLA reads.Polysolver infers HLA-type information with 97% sensitivity and 98% precision from exome-capture sequencing data.

  • (ii) Allows identification of patient-specific mutations in HLA alleles.

 
(ii)Restricted to the use of WES data +++ (106) 
OptiType HLA typing 
  • (i) Performs fully automated HLA typing with four-digit resolution on NGS data from RNA-Seq, WES and WGS technologies.

  • (ii) OptiType showed an accuracy of 99.3% on two-digit-level and of 97.1% on four-digit-level typing using datasets of RNA-Seq, WES and WGS technologies.

 
  • (i) Zygosity detection occasionally fails in cases where alleles with high sequence similarity constitute a heterozygous locus.

  • (ii) Not able to resolve all ambiguities for every genotype.

 
+++ (107) 
MHC multimer technology T-cell reactivity analysis 
  • (i) Gold-standard assay to identify immunogenic peptides. Can be used to detect even low frequencies of antigen-specific T cells on small amounts of clinical material.

  • (ii) “Peptide exchange technology” allows the production of large collections containing a lot of different peptide–MHC complexes for T-cell staining.

 
N.R. +++ (44, 97) 
Technique or softwarePlatformStrengthsWeaknessesRelevance for antigenome predictionReferences
WGS Mutational profiling Both coding and noncoding DNA sequences are analysed. Sequencing depth is usually low, which prevents detection of some subclonal mutations. ++ (97) 
WES Mutational profiling Provides high sequence coverage across exome, increasing reliability and ability to detect subclonal mutations. 
  • (i) Only covers the ≈1% coding regions of the genome.

  • (ii) Some mutations may be missed due to uneven capture efficiency across exons.

 
+++ (97) 
MSI profiling Microsatellite instability Several methods all well-validated. Only provides information on the microsatellite instability. (9) 
CGH Genomic instability Global picture of the overall genomic instability. 
  • (i) Only provides copy-number variations and translocations of large portions of the genome.

  • (ii) No access to the DNA sequence.

 
 
RNA-seq Expression profiling and coding mutation analysis 
  • (i) Focuses on translated mutations only, that are the most likely to have functional consequences.

  • (ii) Analysis not restricted to known genes: potential for discovering novel transcripts, splice variants or fusions.

  • (iii) Possibility to correlate mutational data with gene expression.

 
  • (i) Access to matched normal is key but cannot be achieved in many cases: hard to distinguish tumor-specific mutations from polymorphisms.

  • (ii) Limited calling of mutations within RNA species due to their low levels, either because of low level gene expression or because of mRNA stability.

 
+++ (97) 
NetCHOP 20SPCM (WAPP package)FragPredict (MAPPP package) Proteasomal processing prediction trained on in vitro data N.R. Predictions from in vitro data do not capture the full complexity of proteasomal processing. (98) 
PCleavage      
NetCHOPCterm Proteasomal processing prediction trained on in vivo data 
  • (i) In vivo data provide accurate prediction as predictions are made on the entire processing machinery (action of several proteasomes, cytosolic proteases…)

  • (ii) May also capture transport efficiency.

 
N.R. +++ (98, 99) 
PredTAP TAP transport prediction No comparative study available. No comparative study available. N.R. (98, 100) 
SVMTAP (WAPP package)      
SMM (stabilized matrix method) Allele-specific HLA binding affinity prediction N.R. (ii) Does not account for non-linearities and interdependencies between amino acids. (101–103) 
ARB average relative binding (matrix-based methods)      
NetMHC [artificial neural networks (ANN)-based method] Allele-specific HLA binding affinity prediction Nonlinear model. Does not allow prediction for all known HLA alleles. ++ (103, 104) 
NetMHCpan [Pan-specific artificial neural networks (ANN)-based method] Pan-specific HLA binding affinity prediction 
  • (i) Allows predictions to be made for all known HLA Class I alleles, including alleles for which no prediction is available with NetMHC.

  • (ii) NetMHCpan is the best-performing method for allele-specific HLA binding affinity prediction.

 
N.R. +++ (105) 
Athlates HLA typing N.R. 
  • (i) Early tool with lower accuracy than that of up-to-date tools.Restricted to the use of WES data

 
(98) 
Polysolver HLA typing 
  • (i) Provides improved retrieval and alignment of HLA reads.Polysolver infers HLA-type information with 97% sensitivity and 98% precision from exome-capture sequencing data.

  • (ii) Allows identification of patient-specific mutations in HLA alleles.

 
(ii)Restricted to the use of WES data +++ (106) 
OptiType HLA typing 
  • (i) Performs fully automated HLA typing with four-digit resolution on NGS data from RNA-Seq, WES and WGS technologies.

  • (ii) OptiType showed an accuracy of 99.3% on two-digit-level and of 97.1% on four-digit-level typing using datasets of RNA-Seq, WES and WGS technologies.

 
  • (i) Zygosity detection occasionally fails in cases where alleles with high sequence similarity constitute a heterozygous locus.

  • (ii) Not able to resolve all ambiguities for every genotype.

 
+++ (107) 
MHC multimer technology T-cell reactivity analysis 
  • (i) Gold-standard assay to identify immunogenic peptides. Can be used to detect even low frequencies of antigen-specific T cells on small amounts of clinical material.

  • (ii) “Peptide exchange technology” allows the production of large collections containing a lot of different peptide–MHC complexes for T-cell staining.

 
N.R. +++ (44, 97) 

NOTE: Multiple NGS technologies, bioinformatics tools, and pipelines are available to analyze tumor samples and predict immunogenic mutations/potential neoantigens in patients (see corresponding steps in Fig. 3). Primarily, genomic data are generated using various NGS technologies, most frequently including WES and RNA-seq to integrate both nsSNVs and expressed nsSNVs. These data are then analyzed using dedicated prediction algorithms corresponding to each step of the neoantigen generation biological process. These filtering tools guide the selection of immunogenic neoantigens among the bulk of candidate neoantigens. Although official guidelines are currently lacking on which tool should preferably be used, most often used algorithms include NetCHOPCterm for proteasomal processing prediction and NetMHC/NetMHCpan for HLA binding prediction. Eventually, a functional validation may be performed using an in vitro T-cell reactivity assay to validate the immunogenicity of the predicted neoantigens.

Abbreviations: CGH, comparative genomic hybridization; MSI, microsatellite instability; N.R., not reported; WGS, whole-genome sequencing.

This modeling pipeline has been overall successful in correlating mutational load with predicted neoantigen load. First, mutational and predicted neoantigen loads were significantly correlated with clinical benefit in melanoma patients treated with ipilimumab (5). Consistently, it was suggested that tumors displaying >10 nsSNVs/Mb may produce sufficient neoantigens to generate ant-tumor immunogenicity, whereas tumors with <1 nsSNV/Mb may not (41). Further consistent observations were made in DNA repair–deficient tumors, including MSI-high tumors (7, 36), BRCA-mutated ovarian cancer (8), and melanoma (37). However, genomic instability and accumulation of mutations is a double-edged sword process, which both favors the generation of immunogenic neopeptides, but also allows emergence of less immunogenic new clones that escape immune surveillance, thereby favoring primary or acquired resistance. High intratumor heterogeneity (ITH) has indeed been correlated with poorer outcome, whereas sensitivity to ICB is associated with low ITH and high clonal neoantigens (42). This underlines the paradoxical role of DNA repair defects in dictating response to ICB. Although DNA repair–deficient tumors exhibit high genomic instability and high mutational/neoantigen burdens, they are also the most likely to display high ITH due to their propensity to provoke random mutations (43). If this observation represents a strong biological argument for treating patients with ICB early in the course of the disease, when genomic instability is high, and ITH low, we can hypothesize that each clone within the tumor will retain some degree of intrinsic genomic instability, and participate in the generation of immunogenic neoantigens.

Therefore, a key issue is the determination of which epitopes will actually prime T-cell responses, among the bulk of released epitopes. The study of a melanoma patient who experienced complete response after 3 months of ipilimumab treatment revealed that, out of 1,657 nsSNVs, the tumor only displayed 448 immunologically relevant epitopes, and no more than two of them were identified as able to trigger a patient-specific antitumor T-cell response (44). In a similar analysis, Rizvi and colleagues demonstrated that response to pembrolizumab in a NSCLC patient was associated with the T-cell response against a single neoantigen resulting from a nsSNV in HERC1 (6). In another study evaluating response to anti-CTLA-4 in malignant myeloma, Snyder and colleagues identified a set of consensus tetrapeptide sequences exclusively shared by patients exhibiting long-term clinical benefit (4) and being necessary and sufficient for the activation of an antitumor T-cell response; these results were unfortunately not confirmed in two later studies (5, 37).

Mutational burden and predicted neoantigen load also shape the nature and functional properties of antitumor immune infiltrates. The presence of tumor-infiltrating cytotoxic T-lymphocytes (CTL) has been correlated to higher immunogenic mutation rate, using RNA-seq data (45). Rooney and colleagues further described that predicted neoantigen load correlated with the cytolytic activity of intratumoral CTLs and natural killer (NK) cells (46), but that a given mutation rate was associated with distinct cytolytic activities across different histologies. For instance, cervical cancers exhibit higher cytolytic activity than melanoma, although this cancer type is not as sensitive to ICB. This suggests that both tissue-specific and tumor-specific factors contribute to immune escape regulation. Interestingly, this work also proposed a model for correlating the subclonal evolution of tumor genetics with the cytolytic activity of surrounding CTLs and NK cells, thereby reinforcing the link between continuous tumor genetics drift and immune escape.

Overall, the data presented above support that high mutational burden associates with increased neoantigens formation and tumor immunogenicity. However, the very high attrition rate, from a high mutational burden to the very few neoepitopes that will eventually produce an antitumor immune response, illustrates well the complexity of predicting tumor immunogenicity using genomic data alone. Furthermore, other mechanisms, including oncogenic stress (47–49), secretion of immunosuppressive cytokines (e.g., IL10; ref. 50), or downregulation of MHC class I (51), also modulate tumor immunogenicity, and mutational burden is only one component of the determinants of ICB sensitivity.

“Tumor-related” biomarkers

Beyond tumor “antigenome,” several biomarkers are being developed to predict response to anti–PD-(L)1 therapies (52, 53). The most promising and best validated one is probably PD-L1 expression assessment by immunohistochemical staining on tumor and/or tumor-infiltrating immune cells (3, 54–57). However, this biomarker currently lacks sensitivity—some PD-L1-negative patients consistently experience clinical benefit (58, 59), and specificity— not all PD-L1–positive tumors benefit from anti–PD-(L)1 therapy (2, 60). Furthermore, the parameters of PD-L1 staining scoring are highly variable, notably the anti-PD-L1 antibody (clone SP142 and clone SP2063, Ventana; and clone 28-8 and clone 22C3, Dako), the platform (PD-L1 IHC pharDx, Dako; OptiView DAB IHC Detection Kit, Ventana), the cells of interest (cancer cells, stromal cells, immune tumor-infiltrating cells), the positivity threshold (1%, 5%, 10%, or 50%), as well as the tumor material used for analysis (fresh versus archived material, and primary versus metastatic tumor; ref. 61). Moreover, PD-L1 expression can be constitutive or inducible (e.g., INFγ-mediated induction; ref. 62). Together, these elements represent significant hurdles for reaching the reproducibility and analytic validity that is required for any companion biomarker development and clinical implementation.

“Immune-related” biomarkers

Beyond tumor-related biomarkers, the exploration of immune infiltrate characteristics may also provide interesting biomarkers. Analysis of pretreatment samples from melanoma and NSCLC patients responding to pembrolizumab revealed higher CD8+ T-cell levels at the tumor-invasive margin, as compared with nonresponders (63, 64). Some more complex immune signatures have also been explored: for example, Ribas and colleagues described an immune gene expression signature associated with gain in both ORR and PFS in melanoma patients treated with pembrolizumab (65), which is being explored in other histologies. More recently, an eight-gene signature reflecting preexisting immunity, the “T-effector/IFNγ signature,” was explored in the phase II POPLAR trial. High signature expression levels appeared to predict OS (but not PFS or ORR) benefit in atezolizumab-treated patients (66).

Immunomonitoring strategies, that is, repeated assessment of dynamic circulating biomarkers involved in immune response, have also been proposed. These dynamic biomarkers, which include notably cytokines and inflammatory mediators (Supplementary Table S3; ref. 67), can be monitored at several timepoints on trial using a simple blood test. If these circulating biomarkers have not been robust enough so far to predict responders to ICB (52), they clearly represent a powerful and practical tool for monitoring patient response, and deserve as such active investigation.

Anticipating primary treatment resistance

Finally, as is the case for any targeted therapy, and especially considering the cost of ICB and their associated biomarkers, early prediction of resistance is key. The very recent work by Hugo and colleagues in melanoma (37) identified a transcriptional signature associated with resistance to anti–PD-1 therapy. Exclusively found in the pretreatment tumors of nonresponding patients, this “innate anti–PD-1 resistance” (IPRES) is characterized by the upregulation of genes involved in the regulation of epithelial–mesenchymal transition (EMT), cell adhesion, extracellular matrix remodeling, angiogenesis, and wound healing. Very interestingly, this signature was not predictive of resistance to anti–CTLA-4 therapy, but found at variable frequencies across most common cancers, suggesting that some mechanisms of ICB resistance might be shared by different histologies.

In the aggregate, these data highlight that a comprehensive and integrated approach, which would encompass tumor genetics, immune checkpoint expression, microenvironmental, and immune-monitoring data, is highly needed to best select patients.

How could we improve and expand the use of DNA repair deficiency, mutational burden, and predicted neoantigen load for selecting patients that are the most likely to benefit from anti–PD-(L)1 therapy? Targeted sequencing of hotspot mutations in DNA repair gene panels provides useful but limited information, as it misses nongenetic forms of DNA repair defects (e.g., secondary to epigenetic alterations), and, most importantly, does not functionally evaluate the tumor DNA repair capacity. The decreasing costs and expanding availability of NGS technologies open interesting perspectives for their broader use in clinical routine, and mutational load is a simple parameter that is easily calculated and technically reproducible, allowing the comparison and/or merging of various patient series. We can therefore reasonably hope that, with increasing numbers and open data sharing, relevant mutational thresholds for predicting sensitivity to anti–PD-(L)1 therapy, as well as tumor purity and sequencing depth that are required, will be soon better defined in a histotype-specific fashion. The pipeline optimization and establishment of reference guidelines for predicting neoantigen load will also accelerate the clinical implementation of the latter work. Together, especially if integrated with PD-L1 IHC scoring and signatures of primary resistance, these data might rapidly become robust enough to be clinically implemented.

However, several challenges will still need to be addressed: (i) tumor material is not always available, and efforts should be made to develop equivalent assays on circulating biomarkers, such as cell-free tumor DNA; (ii) tumor heterogeneity needs to be anticipated (42); (iii) the immunogenic potential of mutations other than nsSNVs (including fusion transcripts and aberrantly expressed splice variants) requires further exploration; (iv) developing an integrated approach, that would also encompass tumor microenvironment and immune infiltrates characteristics, as well as immunomonitoring data, warrants further investigation; and (v) last but not least, cost-efficacy and health economics studies will be needed to determine which approach will eventually be the most relevant and sustainable.

Together, these challenges open very stimulating perspectives and one can be certain that several exciting revolutions are still to come soon in immuno-oncology.

A. Marabelle is a consultant/advisory board member for Amgen, Biothera Pharmaceuticals, GlaxoSmithKline, Lytix Biopharma, Nektar, Novartis, Pfizer, Roche/Genentech, and Seattle Genetics. J.-C. Soria is a scientific cofounder of Gritstone Oncology and is a consultant/advisory board member for AstraZeneca, MSD, Pfizer, and Roche. No potential conflicts of interest were disclosed by the other authors.

R.M. Chabanon was supported by the Fondation Philanthropia.

1.
Hodi
FS
,
O'Day
SJ
,
McDermott
DF
,
Weber
RW
,
Sosman
JA
,
Haanen
JB
, et al
Improved survival with ipilimumab in patients with metastatic melanoma
.
N Engl J Med
2010
;
363
:
711
23
.
2.
Robert
C
,
Schachter
J
,
Long
GV
,
Arance
A
,
Grob
JJ
,
Mortier
L
, et al
Pembrolizumab versus ipilimumab in advanced melanoma
.
N Engl J Med
2015
;
372
:
2521
32
.
3.
Topalian
SL
,
Hodi
FS
,
Brahmer
JR
,
Gettinger
SN
,
Smith
DC
,
McDermott
DF
, et al
Safety, activity, and immune correlates of anti-PD-1 antibody in cancer
.
N Engl J Med
2012
;
366
:
2443
54
.
4.
Snyder
A
,
Makarov
V
,
Merghoub
T
,
Yuan
J
,
Zaretsky
JM
,
Desrichard
A
, et al
Genetic basis for clinical response to CTLA-4 blockade in melanoma
.
N Engl J Med
2014
;
371
:
2189
99
.
5.
Van Allen
EM
,
Miao
D
,
Schilling
B
,
Shukla
SA
,
Blank
C
,
Zimmer
L
, et al
Genomic correlates of response to CTLA-4 blockade in metastatic melanoma
.
Science
2015
;
350
:
207
11
.
6.
Rizvi
NA
,
Hellmann
MD
,
Snyder
A
,
Kvistborg
P
,
Makarov
V
,
Havel
JJ
, et al
Mutational landscape determines sensitivity to PD-1 blockade in non-small cell lung cancer
.
Science
2015
;
348
:
124
8
.
7.
Howitt
BE
,
Shukla
SA
,
Sholl
LM
,
Ritterhouse
LL
,
Watkins
JC
,
Rodig
S
, et al
Association of polymerase e-Mutated and microsatellite-instable endometrial cancers with neoantigen load, number of tumor-infiltrating lymphocytes, and expression of PD-1 and PD-L1
.
JAMA Oncol
2015
;
1
:
1319
23
.
8.
Strickland
KC
,
Howitt
BE
,
Shukla
SA
,
Rodig
S
,
Ritterhouse
L
,
Liu
JF
, et al
Association and prognostic significance of BRCA1/2-mutation status with neoantigen load, number of tumor-infiltrating lymphocytes and expression of PD-1/PD-L1 in high grade serous ovarian cancer
.
Oncotarget
2016
;
7
:
13587
98
.
9.
Le
DT
,
Uram
JN
,
Wang
H
,
Bartlett
BR
,
Kemberling
H
,
Eyring
AD
, et al
PD-1 blockade in tumors with mismatch-repair deficiency
.
N Engl J Med
2015
;
372
:
2509
20
.
10.
Burnet
M
. 
Cancer; a biological approach. I. The processes of control
.
Br Med J
1957
;
1
:
779
86
.
11.
Burnet
FM
. 
The concept of immunological surveillance
.
Prog Exp Tumor Res
1970
;
13
:
1
27
.
12.
Dunn
GP
,
Old
LJ
,
Schreiber
RD
. 
The three Es of cancer immunoediting
.
Annu Rev Immunol
2004
;
22
:
329
60
.
13.
Hino
R
,
Kabashima
K
,
Kato
Y
,
Yagi
H
,
Nakamura
M
,
Honjo
T
, et al
Tumor cell expression of programmed cell death-1 ligand 1 is a prognostic factor for malignant melanoma
.
Cancer
2010
;
116
:
1757
66
.
14.
Taube
JM
,
Anders
RA
,
Young
GD
,
Xu
H
,
Sharma
R
,
McMiller
TL
, et al
Colocalization of inflammatory response with B7-h1 expression in human melanocytic lesions supports an adaptive resistance mechanism of immune escape
.
Sci Transl Med
2012
;
4
:
127ra37
.
15.
Shi
S-J
,
Wang
L-J
,
Wang
G-D
,
Guo
Z-Y
,
Wei
M
,
Meng
Y-L
, et al
B7-H1 expression is associated with poor prognosis in colorectal carcinoma and regulates the proliferation and invasion of HCT116 colorectal cancer cells
.
PLoS One
2013
;
8
:
e76012
.
16.
Mu
CY
,
Huang
JA
,
Chen
Y
,
Chen
C
,
Zhang
XG
. 
High expression of PD-L1 in lung cancer may contribute to poor prognosis and tumor cells immune escape through suppressing tumor infiltrating dendritic cells maturation
.
Med Oncol
2011
;
28
:
682
8
.
17.
Yang
CY
,
Lin
MW
,
Chang
YL
,
Wu
CT
,
Yang
PC
. 
Programmed cell death-ligand 1 expression in surgically resected stage i pulmonary adenocarcinoma and its correlation with driver mutations and clinical outcomes
.
Eur J Cancer
2014
;
50
:
1361
9
.
18.
Gao
Q
,
Wang
XY
,
Qiu
SJ
,
Yamato
I
,
Sho
M
,
Nakajima
Y
, et al
Overexpression of PD-L1 significantly associates with tumor aggressiveness and postoperative recurrence in human hepatocellular carcinoma
.
Clin Cancer Res
2009
;
15
:
971
9
.
19.
Kuang
D-M
,
Zhao
Q
,
Peng
C
,
Xu
J
,
Zhang
J-P
,
Wu
C
, et al
Activated monocytes in peritumoral stroma of hepatocellular carcinoma foster immune privilege and disease progression through PD-L1
.
J Exp Med
2009
;
206
:
1327
37
.
20.
Ke
W
,
Kryczek
I
,
Chen
L
,
Zou
W
,
Welling
TH
. 
Kupffer cell suppression of CD8+ T cells in human hepatocellular carcinoma is mediated by B7-H1/programmed death-1 interactions
.
Cancer Res
2009
;
69
:
8067
75
.
21.
Konishi
J
,
Yamazaki
K
,
Azuma
M
,
Kinoshita
I
,
Dosaka-Akita
H
,
Nishimura
M
. 
B7-H1 expression on non-small cell lung cancer cells and its relationship with tumor-infiltrating lymphocytes and their PD-1 expression
.
Clin Cancer Res
2004
;
10
:
5094
100
.
22.
Topalian
SL
,
Drake
CG
,
Pardoll
DM
. 
Immune checkpoint blockade: a common denominator approach to cancer therapy
.
Cancer Cell
2015
;
27
:
450
61
.
23.
Chen
DS
,
Irving
BA
,
Hodi
FS
. 
Molecular pathways: next-generation immunotherapy-inhibiting programmed death-ligand 1 and programmed death-1
.
Clin Cancer Res
2012
;
18
:
6580
7
.
24.
Robert
C
,
Soria
JC
,
Eggermont
AMM
. 
Drug of the year: programmed Death-1 receptor/Programmed Death-1 Ligand-1 receptor monoclonal antibodies
.
Eur J Cancer
2013
;
49
:
2968
71
.
25.
Farmer
H
,
McCabe
N
,
Lord
CJ
,
Tutt
ANJ
,
Johnson
DA
,
Richardson
TB
, et al
Targeting the DNA repair defect in BRCA mutant cells as a therapeutic strategy
.
Nature
2005
;
434
:
917
21
.
26.
Fong
PC
,
Boss
DS
,
Yap
TA
,
Tutt
A
,
Wu
P
,
Mergui-Roelvink
M
, et al
Inhibition of poly(ADP-ribose) polymerase in tumors from BRCA mutation carriers
.
N Engl J Med
2009
;
361
:
123
34
.
27.
Ledermann
J
,
Harter
P
,
Gourley
C
,
Friedlander
M
,
Vergote
I
,
Rustin
G
, et al
Olaparib maintenance therapy in platinum-sensitive relapsed ovarian cancer
.
N Engl J Med
2012
;
366
:
1382
92
.
28.
Lord
CJ
,
Ashworth
A
. 
The DNA damage response and cancer therapy
.
Nature
2012
;
481
:
287
94
.
29.
Hanahan
D
,
Weinberg
RA
. 
Hallmarks of cancer: the next generation
.
Cell
2011
;
144
:
646
74
.
30.
Lindahl
T
. 
Instability and decay of the primary structure of DNA
.
Nature
1993
;
362
:
709
15
.
31.
Pleasance
ED
,
Stephens
PJ
,
O'Meara
S
,
McBride
DJ
,
Meynert
A
,
Jones
D
, et al
A small-cell lung cancer genome with complex signatures of tobacco exposure
.
Nature
2010
;
463
:
184
90
.
32.
Pfeifer
GP
,
You
Y-H
,
Besaratinia
A
. 
Mutations induced by ultraviolet light
.
Mutat Res
2005
;
571
:
19
31
.
33.
Timms
KM
,
Abkevich
V
,
Hughes
E
,
Neff
C
,
Reid
J
,
Morris
B
, et al
Association of BRCA1/2 defects with genomic scores predictive of DNA damage repair deficiency among breast cancer subtypes
.
Breast Cancer Res
2014
;
16
:
1
9
.
34.
Hewish
M
,
Lord
CJ
,
Martin
SA
,
Cunningham
D
,
Ashworth
A
. 
Mismatch repair deficient colorectal cancer in the era of personalized treatment
.
Nat Rev Clin Oncol
2010
;
7
:
197
208
.
35.
Alexandrov
LB
,
Nik-Zainal
S
,
Wedge
DC
,
Aparicio
SA
 Jr
,
Behjati
S
,
Biankin
A V
, et al
Signatures of mutational processes in human cancer
.
Nature
2013
;
500
:
415
21
.
36.
Dudley
JC
,
Lin
M-T
,
Le
DT
,
Eshleman
JR
. 
Microsatellite instability as a biomarker for PD-1 blockade
.
Clin Cancer Res
2016
;
22
:
813
20
.
37.
Hugo
W
,
Zaretsky
JM
,
Sun
L
,
Song
C
,
Moreno Homet
B
,
Hu-Lieskovan
S
, et al
Genomic and transcriptomic features of response to anti-PD-1 therapy in metastatic melanoma
.
Cell
2016
;
165
:
35
44
.
38.
Ansell
SM
,
Lesokhin
AM
,
Borrello
I
,
Halwani
A
,
Scott
EC
,
Gutierrez
M
, et al
PD-1 Blockade with nivolumab in relapsed or refractory Hodgkin's lymphoma
.
N Engl J Med
2015
;
372
:
311
9
.
39.
Helleday
T
,
Eshtad
S
,
Nik-Zainal
S
. 
Mechanisms underlying mutational signatures in human cancers
.
Nat Rev Genet
2014
;
15
:
585
98
.
40.
Lennerz
V
,
Fatho
M
,
Gentilini
C
,
Frye
RA
,
Lifke
A
,
Ferel
D
, et al
The response of autologous T cells to a human melanoma is dominated by mutated neoantigens
.
Proc Natl Acad Sci U S A
2005
;
102
:
16013
8
.
41.
Schumacher
TN
,
Schreiber
RD
. 
Neoantigens in cancer immunotherapy
.
Science
2015
;
348
:
69
74
.
42.
McGranahan
N
,
Furness
AJ
,
Rosenthal
R
,
Ramskov
S
,
Lyngaa
R
,
Saini
SK
, et al
Clonal neoantigens elicit T cell immunoreactivity and sensitivity to immune checkpoint blockade
.
Science
2016
;
351
:
1463
9
.
43.
Burrell
RA
,
McGranahan
N
,
Bartek
J
,
Swanton
C
. 
The causes and consequences of genetic heterogeneity in cancer evolution
.
Nature
2013
;
501
:
338
45
.
44.
Van Rooij
N
,
Van Buuren
MM
,
Philips
D
,
Velds
A
,
Toebes
M
,
Heemskerk
B
, et al
Tumor exome analysis reveals neoantigen-specific T-cell reactivity in an ipilimumab-responsive melanoma
.
J Clin Oncol
2013
;
31
:
439
42
.
45.
Brown
SD
,
Warren
RL
,
Gibb
EA
,
Martin
SD
,
Spinelli
JJ
,
Nelson
BH
, et al
Neo-antigens predicted by tumor genome meta-analysis correlate with increased patient survival
.
Genome Res
2014
;
24
:
743
50
.
46.
Rooney
MS
,
Shukla
SA
,
Wu
CJ
,
Getz
G
,
Hacohen
N
. 
Molecular and genetic properties of tumors associated with local immune cytolytic activity
.
Cell
2015
;
160
:
48
61
.
47.
Akbay
EA
,
Koyama
S
,
Carretero
J
,
Altabef
A
,
Tchaicha
JH
,
Christensen
CL
, et al
Activation of the PD-1 pathway contributes to immune escape in EGFR-driven lung tumors
.
Cancer Discov
2013
;
3
:
1355
63
.
48.
Azuma
K
,
Ota
K
,
Kawahara
A
,
Hattori
S
,
Iwama
E
,
Harada
T
, et al
Association of PD-L1 overexpression with activating EGFR mutations in surgically resected nonsmall-cell lung cancer
.
Ann Oncol
2014
;
25
:
1935
40
.
49.
Green
MR
,
Monti
S
,
Rodig
SJ
,
Juszczynski
P
,
Currie
T
,
O'Donnell
E
, et al
Integrative analysis reveals selective 9p24.1 amplification, increased PD-1 ligand expression, and further induction via JAK2 in nodular sclerosing Hodgkin lymphoma and primary mediastinal large B-cell lymphoma
.
Blood
2010
;
116
:
3268
77
.
50.
Berger
S
,
Siegert
A
,
Denkert
C
,
Köbel
M
,
Hauptmann
S
. 
Interleukin-10 in serous ovarian carcinoma cell lines
.
Cancer Immunol Immunother
2001
;
50
:
328
33
.
51.
Reichel
J
,
Chadburn
A
,
Rubinstein
PG
,
Giulino-Roth
L
,
Tam
W
,
Liu
Y
, et al
Flow-sorting and exome sequencing reveals the oncogenome of primary Hodgkin and Reed-Sternberg cells
.
Blood
2015
;
12
:
1061
72
.
52.
Herbst
RS
,
Soria
J-C
,
Kowanetz
M
,
Fine
GD
,
Hamid
O
,
Gordon
MS
, et al
Predictive correlates of response to the anti-PD-L1 antibody MPDL3280A in cancer patients
.
Nature
2014
;
515
:
563
7
.
53.
Yearley
JH
,
Gibson
C
,
Yu
N
. 
PD-L2 expression in human tumors: relevance to anti-PD1 therapy in cancer [abstract]
.
In: Proceedings of the European Cancer Congress 2015; 2015 Sep 25–29; Vienna, Austria
.
Brussels (Belgium)
:
European Cancer Organisation
; 
2015
.
Abstract nr 18LBA
.
54.
Garon
EB
,
Rizvi
NA
,
Hui
R
,
Leighl
N
,
Balmanoukian
AS
,
Eder
JP
, et al
Pembrolizumab for the treatment of non-small-cell lung cancer
.
N Engl J Med
2015
;
372
:
2018
28
.
55.
Brahmer
JR
,
Rizvi
NA
,
Lutzky
J
,
Khleif
S
,
Blake-Haskins
A
,
Li
X
, et al
Clinical activity and biomarkers of MEDI4736, an anti-PD-L1 antibody, in patients with NSCLC
.
J Clin Oncol
32
:
5s
, 
2014
(
suppl; abstr 8021⁁
).
56.
Soria
JC
,
Cruz
C
,
Bahleda
R
,
Delord
JP
,
Horn
L
,
Herbst
RS
, et al
Clinical activity, safety, and biomarkers of PD-L1 blockade in non-small cell lung cancer (NSCLC): additional analyses from a clinical study of the engineered antibody MPDL3280A (anti-PDL1) [abstract]
.
In: Proceedings of the European Cancer Congress 2013; 2013 Sep 27–Oct 1; Amsterdam, the Netherlands
.
Brussels (Belgium)
:
European Cancer Organisation
; 
2013
.
Abstract nr 3408
.
57.
Taube
JM
,
Klein
A
,
Brahmer
JR
,
Xu
H
,
Pan
X
,
Kim
JH
, et al
Association of PD-1, PD-1 ligands, and other features of the tumor immune microenvironment with response to anti-PD-1 therapy
.
Clin Cancer Res
2014
;
20
:
5064
74
.
58.
Patel
SP
,
Kurzrock
R
. 
PD-L1 expression as a predictive biomarker in cancer immunotherapy
.
Mol Cancer Ther
2015
;
14
:
847
56
.
59.
Kerr
KM
,
Tsao
M-S
,
Nicholson
AG
,
Yatabe
Y
,
Wistuba
II
,
Hirsch
FR
. 
Programmed death-ligand 1 immunohistochemistry in lung cancer: in what state is this art?
J Thorac Oncol
2015
;
10
:
985
9
.
60.
Robert
C
,
Long
GV
,
Brady
B
,
Dutriaux
C
,
Maio
M
,
Mortier
L
, et al
Nivolumab in previously untreated melanoma without BRAF mutation
.
N Engl J Med
2015
;
372
:
320
30
.
61.
Madore
J
,
Vilain
RE
,
Menzies
AM
,
Kakavand
H
,
Wilmott
JS
,
Hyman
J
, et al
PD-L1 expression in melanoma shows marked heterogeneity within and between patients: Implications for anti-PD-1/PD-L1 clinical trials
.
Pigment Cell Melanoma Res
2015
;
28
:
245
53
.
62.
MacMicking
JD
. 
Interferon-inducible effector mechanisms in cell-autonomous immunity
.
Nat Rev Immunol
2012
;
12
:
367
82
.
63.
Tumeh
PC
,
Harview
CL
,
Yearley
JH
,
Shintaku
IP
,
Taylor
EJM
,
Robert
L
, et al
PD-1 blockade induces responses by inhibiting adaptive immune resistance
.
Nature
2015
;
515
:
568
71
.
64.
Hu-Lieskovan
S
,
Goldman
J
,
Han
M
,
Zaretsky
J
,
Shintaku
I
,
Wolf
B
, et al
High intratumoral T-cell infiltration correlated with mutational load and response to pembrolizumab in non-small cell lung cancer [abstract]
.
In: Proceedings of the 16th World Conference on Lung Cancer; 2015 Sep 6–9; Denver, CO
.
Aurora (CO)
:
International Association for the Study of Lung Cancer
; 
2015
.
Abstract nr ORAL3105
.
65.
Ribas
A
,
Robert
C
,
Hodi
FS
,
Wolchok
JD
,
Joshua
AM
,
Hwu
W-J
, et al
Association of response to programmed death receptor 1 (PD-1) blockade with pembrolizumab (MK-3475) with an interferon-inflammatory immune gene signature
.
J Clin Oncol
33
, 
2015
(
suppl; abstr 3001
).
66.
Fehrenbacher
L
,
Spira
A
,
Ballinger
M
,
Kowanetz
M
,
Vansteenkiste
J
,
Mazieres
J
, et al
Atezolizumab versus docetaxel for patients with previously treated non-small-cell lung cancer (POPLAR): a multicentre, open-label, phase 2 randomised controlled trial
.
Lancet
2016
;
387
:
1837
46
.
67.
Manson
G
,
Norwood
J
,
Marabelle
A
,
Kohrt
H
,
Houot
R
. 
Biomarkers associated with checkpoint inhibitors
.
Ann Oncol
2016
;
27
:
1199
206
.
68.
O'Sullivan
CC
,
Moon
DH
,
Kohn
EC
,
Lee
J
,
Sullivan
CCO
,
Moon
DH
, et al
Beyond breast and ovarian cancers: PARP inhibitors for BRCA mutation-associated and BRCA-like solid tumors
.
Front Oncol
2014
;
4
:
42
.
69.
Postel-Vinay
S
,
Vanhecke
E
,
Olaussen
KA
,
Lord
CJ
,
Ashworth
A
,
Soria
J-C
. 
The potential of exploiting DNA-repair defects for optimizing lung cancer treatment
.
Nat Rev Clin Oncol
2012
;
9
:
144
55
.
70.
Huang
QM
,
Tomida
S
,
Masuda
Y
,
Arima
C
,
Cao
K
,
Kasahara
TA
, et al
Regulation of DNA polymerase POLD4 influences genomic instability in lung cancer
.
Cancer Res
2010
;
70
:
8407
16
.
71.
The Cancer Genome Atlas Network
. 
Comprehensive molecular characterization of clear cell renal cell carcinoma
.
Nature
2013
;
499
:
43
9
.
72.
Feng
C
,
Ding
G
,
Jiang
H
,
Ding
Q
,
Wen
H
. 
Loss of MLH1 confers resistance to PI3Kβ inhibitors in renal clear cell carcinoma with SETD2 mutation
.
Tumor Biol
2015
;
36
:
3457
64
.
73.
Yoo
KH
,
Won
KY
,
Lim
S-J
,
Park
Y-K
,
Chang
S-G
. 
Deficiency of MSH2 expression is associated with clear cell renal cell carcinoma
.
Oncol Lett
2014
;
8
:
2135
9
.
74.
Nickerson
ML
,
Dancik
GM
,
Im
KM
,
Edwards
MG
,
Turan
S
,
Brown
J
, et al
Concurrent alterations in TERT, KDM6A, and the BRCA pathway in bladder cancer
.
Clin Cancer Res
2014
;
20
:
4935
48
.
75.
Plimack
ER
,
Dunbrack
RL
,
Brennan
TA
,
Andrake
MD
,
Zhou
Y
,
Serebriiskii
IG
, et al
Defects in DNA repair genes predict response to neoadjuvant cisplatin-based chemotherapy in muscle-invasive bladder cancer
.
Eur Urol
2015
;
68
:
959
67
.
76.
Mullane
SA
,
Werner
L
,
Guancial
EA
,
Lis
RT
,
Stack
EC
,
Loda
M
, et al
Expression levels of DNA damage repair proteins are associated with overall survival in platinum-treated advanced urothelial carcinoma
.
Clin Genitourin Cancer
2016
;
14
:
352
9
.
77.
Bai
S
,
Nunez
AL
,
Wei
S
,
Ziober
A
,
Yao
Y
,
Tomaszewski
JE
, et al
Microsatellite instability and TARBP2 mutation study in upper urinary tract urothelial carcinoma
.
Am J Clin Pathol
2013
;
139
:
765
70
.
78.
Weinstein
JN
,
Akbani
R
,
Broom
BM
,
Wang
W
,
Verhaak
RGW
,
McConkey
D
, et al
Comprehensive molecular characterization of urothelial bladder carcinoma
.
Nature
2014
;
507
:
315
22
.
79.
Stoepker
C
,
Ameziane
N
,
van der Lelij
P
,
Kooi
IE
,
Oostra
AB
,
Rooimans
MA
, et al
Defects in the Fanconi anemia pathway and chromatid cohesion in head-and-neck cancer
.
Cancer Res
2015
;
75
:
3543
5
.
80.
Lee
JM
,
Ledermann
JA
,
Kohn
EC
. 
PARP inhibitors for BRCA1/2 mutation-associated and BRCA-like malignancies
.
Ann Oncol
2014
;
25
:
32
40
.
81.
Burgess
M
,
Puhalla
S
. 
BRCA 1/2-mutation related and sporadic breast and ovarian cancers: more alike than different
.
Front Oncol
2014
;
4
:
19
.
82.
Walsh
T
,
Casadei
S
,
Lee
MK
,
Pennil
CC
,
Nord
AS
,
Thornton
AM
, et al
Mutations in 12 genes for inherited ovarian, fallopian tube, and peritoneal carcinoma identified by massively parallel sequencing
.
Proc Natl Acad Sci U S A
2011
;
108
:
18032
7
.
83.
Bryant
C
,
Rawlinson
R
,
Massey
AJ
. 
Chk1 inhibition as a novel therapeutic strategy for treating triple-negative breast and ovarian cancers
.
BMC Cancer
2014
;
14
:
570
.
84.
Bass
AJ
,
Thorsson
V
,
Shmulevich
I
,
Reynolds
SM
,
Miller
M
,
Bernard
B
, et al
Comprehensive molecular characterization of gastric adenocarcinoma
.
Nature
2014
;
513
:
202
9
.
85.
Miquel
C
,
Jacob
S
,
Grandjouan
S
,
Aime
A
,
Viguier
J
,
Sabourin
JC
, et al
Frequent alteration of DNA damage signalling and repair pathways in human colorectal cancers with microsatellite instability
.
Oncogene
2007
;
26
:
5919
26
.
86.
The Cancer Genome Atlas Network
. 
Comprehensive molecular characterization of human colon and rectal cancer
.
Nature
2012
;
487
:
330
7
.
87.
Wang
Y
,
Hong
Y
,
Li
M
,
Long
J
,
Zhao
YP
,
Zhang
JX
, et al
Mutation inactivation of Nijmegen breakage syndrome gene (NBS1) in hepatocellular carcinoma and intrahepatic cholangiocarcinoma
.
PLoS One
2013
;
8
:
e82426
.
88.
Hinrichsen
I
,
Kemp
M
,
Peveling-Oberhag
J
,
Passmann
S
,
Plotz
G
,
Zeuzem
S
, et al
Promoter methylation of MLH1, PMS2, MSH2 and p16 is a phenomenon of advanced-stage HCCs
.
PLoS One
2014
;
9
:
e84453
.
89.
Wani
Y
,
Notohara
K
,
Tsukayama
C
,
Okada
S
. 
Reduced expression of hMLH1 and hMSH2 gene products in high-grade hepatocellular carcinoma
.
Acta Med Okayama
2001
;
55
:
65
71
.
90.
Moy
AP
,
Shahid
M
,
Ferrone
CR
,
Borger
DR
,
Zhu
AX
,
Ting
D
, et al
Microsatellite instability in gallbladder carcinoma
.
Virchows Arch
2015
;
466
:
393
402
.
91.
Abraham
SC
,
Lee
JH
,
Boitnott
JK
,
Argani
P
,
Furth
EE
,
Wu
TT
. 
Microsatellite instability in intraductal papillary neoplasms of the biliary tract
.
Mod Pathol
2002
;
15
:
1309
17
.
92.
Mateo
J
,
Carreira
S
,
Sandhu
S
,
Miranda
S
,
Mossop
H
,
Perez-Lopez
R
, et al
DNA-repair defects and olaparib in metastatic prostate cancer
.
N Engl J Med
2015
;
373
:
1697
708
.
93.
Abeshouse
A
,
Ahn
J
,
Akbani
R
,
Ally
A
,
Amin
S
,
Andry
CD
, et al
The molecular taxonomy of primary prostate cancer
.
Cell
2015
;
163
:
1011
25
.
94.
The Cancer Genome Atlas Research Network
Integrated genomic characterization of endometrial carcinoma
.
Nature
2013
;
497
:
67
73
.
95.
Rustgi
AK
. 
Familial pancreatic cancer: genetic advances
.
Genes Dev
2014
;
28
:
1
7
.
96.
Riazy
M
,
Kalloger
SE
,
Sheffield
BS
,
Peixoto
RD
,
Li-Chang
HH
,
Scudamore
CH
, et al
Mismatch repair status may predict response to adjuvant chemotherapy in resectable pancreatic ductal adenocarcinoma
.
Mod Pathol
2015
;
28
:
1383
9
.
97.
Heemskerk
B
,
Kvistborg
P
,
Schumacher
TNM
. 
The cancer antigenome
.
EMBO J
2013
;
32
:
194
203
.
98.
Backert
L
,
Kohlbacher
O
. 
Immunoinformatics and epitope prediction in the age of genomic medicine
.
Genome Med
2015
;
7
:
119
.
99.
Peters
B
,
Bulik
S
,
Tampe
R
,
Van Endert
PM
,
Holzhütter
H-G
. 
Identifying MHC class I epitopes by predicting the TAP transport efficiency of epitope precursors
.
J Immunol
2003
;
171
:
1741
9
.
100.
Zhang
GL
,
Petrovsky
N
,
Kwoh
CK
,
August
JT
,
Brusic
V
. 
PRED(TAP): a system for prediction of peptide binding to the human transporter associated with antigen processing
.
Immunome Res
2006
;
2
:
3
.
101.
Peters
B
,
Sette
A
. 
Generating quantitative models describing the sequence specificity of biological processes with the stabilized matrix method
.
BMC Bioinform
2005
;
6
:
132
.
102.
Bui
HH
,
Sidney
J
,
Peters
B
,
Sathiamurthy
M
,
Sinichi
A
,
Purton
KA
, et al
Automated generation and evaluation of specific MHC binding predictive tools: ARB matrix applications
.
Immunogenetics
2005
;
57
:
304
14
.
103.
Trolle
T
,
Metushi
IG
,
Greenbaum
JA
,
Kim
Y
,
Sidney
J
,
Lund
O
, et al
Automated benchmarking of peptide-MHC class I binding predictions
.
Bioinformatics
2015
;
31
:
2174
81
.
104.
Lundegaard
C
,
Lund
O
,
Nielsen
M
. 
Accurate approximation method for prediction of class I MHC affinities for peptides of length 8, 10 and 11 using prediction tools trained on 9mers
.
Bioinformatics
2008
;
24
:
1397
8
.
105.
Hoof
I
,
Peters
B
,
Sidney
J
,
Pedersen
LE
,
Sette
A
,
Lund
O
, et al
NetMHCpan, a method for MHC class i binding prediction beyond humans
.
Immunogenetics
2009
;
61
:
1
13
.
106.
Shukla
SA
,
Rooney
MS
,
Rajasagi
M
,
Tiao
G
,
Dixon
PM
,
Lawrence
MS
, et al
Comprehensive analysis of cancer-associated somatic mutations in class I HLA genes
.
Nat Biotechnol
2015
;
33
:
1152
8
.
107.
Kohlbacher
O
. 
OptiType: precision HLA typing from next-generation sequencing data
.
Bioinformatics
2014
;
30
:
3310
6
.