Immunogenicity of somatic mutations in human gastrointestinal cancers

Abstract

It is unknown whether the human immune system frequently mounts a T cell response against mutations expressed by common epithelial cancers. Using a next-generation sequencing approach combined with high-throughput immunologic screening, we demonstrated that tumor-infiltrating lymphocytes (TILs) from 9 out of 10 patients with metastatic gastrointestinal cancers contained CD4⁺ and/or CD8⁺ Tcells that recognized one to three neo-epitopes derived from somatic mutations expressed by the patient’s own tumor. There were no immunogenic epitopes shared between these patients. However, we identified in one patient a human leukocyte antigen–C*08:02–restricted T cell receptor from CD8⁺ TILs that targeted the KRAS^G12D hotspot driver mutation found in many human cancers. Thus, a high frequency of patients with common gastrointestinal cancers harbor immunogenic mutations that can potentially be exploited for the development of highly personalized immunotherapies.

Genetic aberrations underpin all cancers. These genetic alterations are specific to cancers and are not present in normal tissues; thus, treatments that specifically target the protein product of these genetic alterations may provide clinical benefit in the absence of normal tissue toxicities. Cancer immunotherapies such as adoptive cell therapy with tumor-infiltrating lymphocytes (TILs) or immune checkpoint inhibitors have demonstrated clinical activity in patients with metastatic melanoma, smoking-induced lung cancer, renal cell carcinomas, and cancers with DNA mismatch-repair deficiency (1–7). Increasing correlative evidence suggests that some of these clinical responses are likely mediated by T cells that target somatic mutations expressed by the patients’ tumors (7–13). In a direct demonstration of the therapeutic potential of immune targeting of cancer mutations, we recently identified mutation-specific CD4⁺ T cells in a patient with metastatic epithelial cancer originating from the bile ducts and observed objective regression of lung and liver metastases ongoing now at 20 months after treating the patient with a highly enriched population of these mutation-specific T cells (14). This patient’s tumors contained only 26 mutations, which indicates that the immune system can mount a clinically relevant antimutation T cell response against cancers with a low mutation load (14). However, immune checkpoint blockade therapies are ineffective against the majority of metastatic gastrointestinal (GI) cancers (2, 7), which on average have a lower number of mutations as compared with those of melanoma, smoking-induced lung cancers, and cancers with DNA mismatch-repair deficiencies (7, 15, 16). This suggests that mutation-reactive T cells maybe rare or absent in the majority of these patients. If so, this would likely pose major challenges for the development of immunotherapies that target mutations in many patients with GI cancers. Thus, we aimed to determine whether immunogenic mutations are common in patients with metastatic GI cancers.

To this end, we used whole-exome or whole-genome sequencing to identify somatic mutations present in the metastatic tumors derived from nine additional patients with cancers originating from the colon, rectum, esophagus, bile ducts, or pancreas (Table 1 and table S1). The number of mutations ranged from 10 to 155 (Table 1 and tables S2 to S9) when using previous methods to call mutations (17). However, to evaluate any low-coverage and low-confidence mutations, we relaxed the mutation call criteria for most samples (17) and thus evaluated between 38 and 264 putative mutations (Table 1). In parallel, we generated multiple TIL cultures from the metastatic lesions of each patient. To test whether any of the TIL cultures from each patient recognized their own tumor mutations, we used a tandem minigene (TMG) approach as previously described (8, 14, 17). Briefly, these TMGs comprise a string of minigenes, which are genetic constructs that encode an identified mutation flanked on each side by the 12 wild-type amino acids from the parent protein, except in the case of frameshift mutations, in which the cDNA was translated until the next stop codon. After in vitro transcription, the TMG RNAs are then individually transfected into autologous antigen-presenting cells (APCs), allowing for the potential processing and presentation of all mutated epitopes by each of the patient’s major histocompatibility complex (MHC) class I and class II molecules, followed by a coculture with the different TIL cultures. Because we and others cannot consistently grow tumor cell lines from metastatic gastrointestinal cancers, this approach enables the reliable screening of tumor-specific antigens (the mutanome) without the requirement of a tumor cell line.

Table 1.

Mutation-reactive T cells in metastatic GI cancers. NE, not evaluated.

Patient ID	Age/sex	Tumor type	Number of mutations†	Number of mutations assessed‡	Number of TIL cultures assessed	Number of TIL cultures with mutation reactivity§	Mutated protein recognized	Amino acid change	T cell type	Frequency of mutation-reactive TCR in tumor (%)	Rank of mutation-reactive TCR in tumor
3737*	45/F	Bile duct	26	25	5	5	ERBB2IP	E805G	CD4	0.009	2718
								E805G	CD4	0.375	10
3812	44/M	Bile duct	48	179	5	0	–	–	–	–	–
3942	46/F	Rectal	155	144	6	2	NUP98	A359D	CD8	0.67	5
						4	KARS	D356	CD8	0.020	1143
						3	GPD2	E426K	CD4	0.037	862
3948	48/M	Esophageal	84	211	5	2	PLEC	E1179K	CD4	NE	NE
						2	XPO7	P274S	CD4	NE	NE
						2	AKAP2	Q418K	CD4	NE	NE
3971	49/M	Colon	118	118	23	11	CASP8	F67V	CD8	1.25	3
3978	46/F	Bile duct	39	38	9∥	1¶	ITGB4	S1002I	CD4	NE	NE
3995	50/M	Colon	58	154	19∥	2	TUBGCP2	P293L	CD8	0.023	1056
						15	RNF213	N1702S	CD8	0.60	15
						2	KRAS	G12D	CD8	0.055	527
4007	52/M	Colon	134	264	23	4	SKIV2L	R653H	CD8	0.090	121
								R653H	CD8	0.014	887
						5	H3F3B	A48T	CD8	1.19	4
4032	46/M	Colon	101	222	24	12	API5	R243Q	CD8	0.083	126
								R243Q	CD8	0.059	187
						1	RNF10	E572K	CD8	0.030	423
						7	PHLPP1	G566E	CD8	0.081	129
4069	57/M	Pancreatic	10	97	15	1	ZFYVE27	R6H	CD8	0.088	278

^*Patient 3737 was previously reported in (14).

^†As determined by Personal Genome Diagnostics (PGDx) from whole-exome sequencing (17).

^‡determined when mutation call criteria was relaxed (17). The mutation call criteria was not relaxed for patients 3737, 3942, and 3978.

^§Reactivity was determined with IFN-γ ELISPOT or flow cytometry for 4-1BB or OX40 up-regulation upon coculture with mutated TMGs or peptides.

^∥The indicated number of TIL cultures includes T cell populations that were sorted and expanded from a tumor digest based on PD-1 expression as described in the supplementary materials.

^¶Identified from a FACS-purified TIL population as described in the supplementary text.

In a representative patient with metastatic colon cancer (patient 4007), 23 TIL cultures derived from individual tumor fragments of a single metastatic liver lesion were initially screened for reactivity against 17 TMG constructs, which together encoded 264 minigenes (table S10). We evaluated both the secretion of the effector cytokine interferon-γ (IFN-γ) using the enzyme-linked immunospot (ELISPOT) assay and up-regulation of the T cell activation marker 4-1BB using flow cytometry because these approaches can provide complementary and nonredundant information about antigen-specific T cell responses. Five of 23 and 4 of 23 TIL cultures contained variable levels of reactivity against TMG-7 (TIL cultures 4, 6, 16, 22, and 23) and TMG-14 (TIL cultures 7, 8, 16, and 18), respectively, as determined with IFN-γ ELISPOT (Fig. 1A, top). Flow cytometric analysis of the T cell activation marker 4-1BB suggested that the reactivities against TMG-7 and TMG-14 were mediated by CD8⁺ T cells (Fig. 1A, bottom). To identify which mutated antigens were being recognized in TMG-7 and TMG-14, we first enriched the TMG-7– and TMG-14–reactive cells by sorting and expanding the CD8⁺ T cells from TIL cultures 8, 16, and 23 that up-regulated 4-1BB upon stimulation with TMG-7 or TMG-14 (Fig. 1B). These enriched TMG-7– and TMG-14–reactive cells were then cocultured with autologous dendritic cells (DCs) that were individually pulsed with the mutated peptides encoded by TMG-7 or TMG-14, respectively (table S10). TMG-14–reactive CD8⁺ T cells from both TIL cultures 8 and 16 recognized the SKIV2L^R653H mutation, whereas the TMG-7–reactive CD8⁺ T cells from TIL culture 23 recognized the H3F3B^A48T mutation (Fig. 1C). We sequenced the T cell receptors (TCRs) of the two SKIV2L^R653H-mutation-reactive T cell populations and found that they did not share the same TCR sequences (table S19). The T cell responses were mutation-specific because genetic engineering of autologous open repertoire peripheral blood T cells with the TMG-7– and TMG-14–reactive TCRs (table S20) conferred reactivity to the mutated H3F3B^A48T and SKIV2L^R653H peptides, respectively, with no reactivity observed against the wild-type peptides (Fig. 1D). Thus, our screening approach enabled the identification of patient-specific mutation-reactive CD8⁺ T cells.

Fig. 1. Identification of personalized mutation-specific T cells in a patient with metastatic colon cancer (4007).

(A) Twenty-three different TIL cultures were cocultured with autologous DCs transfected with an irrelevant TMG RNA, or the indicated TMG construct encoding the various putative mutations identified with whole-exomic sequencing. T cell responses were measured the next day by means of IFN-γ ELISPOT assay (top) and flow cytometric analysis for 4-1BB up-regulation on CD8⁺ T cells (bottom). (B) TIL cultures 8, 16, and 23 were cocultured with DCs transfected with an irrelevant TMG RNA, or TMG-7 or TMG-14 RNA, and CD8⁺ T cells that up-regulated 4-1BB were purified by means of FACS and expanded. (C) IFN-γ ELISPOT assay (left axis) and flow cytometric analysis of 4-1BB expression (right axis) on TMG-7– and TMG-14–reactive CD8⁺ T cells isolated in (B) after an overnight coculture with DCs pulsed with long peptides encoded by either TMG-7 or TMG-14, or transfected with TMG-7 or TMG-14 RNA. (D) Autologous open-repertoire peripheral blood T cells were genetically modified with the TCRs derived from the SKIV2L and H3F3B mutation-reactive T cells identified in (B) and (C) and then cocultured with DCs pulsed with the indicated wild-type (wt) and mutated (mut) long peptides. Flow cytometric analysis is gated on live CD8⁺ T cells, and TCR transduction efficiencies ranged between 60 and 80%. Plate-bound antibody to CD3 (OKT3) was used as a positive control in all coculture assays. “>” ELISPOT assay is not accurate above ~500 spots. Data from (B) to (D) are representative of at least two independent experiments.

For patient 3995 with metastatic colon cancer, in addition to the identification of patient-specific, mutation-reactive CD8⁺ T cells against RNF213^N1702S and TUBGCP2^P293L (fig. S1), we also detected a low-level CD8⁺ TIL reactivity against the KRAS^G12D hotspot mutation (Fig. 2A). We purified by means of fluorescence-activated cell sorting (FACS) and expanded the low frequency of KRAS^G12D-reactive CD8⁺ T cells (Fig. 2B) and confirmed that the enriched population indeed specifically recognized APCs when pulsed with the mutated KRAS^G12D peptide or when transfected with full-length KRAS^G12D RNA (Fig. 2C). Genetic engineering of autologous open-repertoire T cells with the TCR isolated from the KRAS^G12D-specific T cells (Fig. 2C and table S20) redirected human leukocyte antigen (HLA)–C*08:02–restricted reactivity to COS-7 cells transfected with KRAS^G12D (Fig. 2D) as well as KRAS^G12D-expressing pancreatic cancer cell lines (Fig. 2E). In addition to the secretion of IFN-γ, the KRAS-mutation reactive TILs specifically produced tumor necrosis factor (TNF) and displayed cytolytic potential against pancreatic cancer cell lines expressing HLA-C*08:02 and KRAS^G12D (Fig. 2F). The KRAS^G12D mutation did not appear to be immunogenic in all patients for we did not detect KRAS-mutation reactive TILs in the two patients, 4032 and 4069, whose tumors harbored KRAS^G12D (tables S8 and S9). These two patients did not express the HLA-C*08:02 allele (table S23), although they did express other MHC-I alleles predicted to bind the KRAS^G12D mutation with moderate affinity (table S24). Thus, in addition to distinct patient-specific immunogenic mutations, we also identified an immunogenic mutation in the shared KRAS^G12D driver mutation.

Fig. 2. Identification of KRAS^G12D-mutation-specific T cells in a patient with colorectal cancer (3995).

(A) IFN-γ ELISPOT assay of P3W5+6 TILs cocultured overnight with DCs transfected with irrelevant (Irrel.) TMG or TMG-3 RNA (which encodes the KRAS^G12D minigene), or DCs pulsed with wt or KRAS^G12D 24–amino acid (AA)–long peptides. Numbers are spots per 3 × 10⁴ TILs. (B) P3W5+6 TILs were stimulated with DCs pulsed with wt or KRAS^G12D 24-AA-long peptides, and 4-1BB⁺ CD8⁺ T cells were purified by means of FACS and expanded. (C) IFN-γ ELISPOT assay and flow cytometric analysis of 4-1BB expression on the KRAS^G12D-enriched CD8⁺ T cells from (B) cocultured overnight with DCs pulsed with wt or KRAS^G12D 24-AA-long peptides, or DCs transfected with full-length wt or KRAS^G12D RNA. (D) IFN-γ enzyme-linked immunosorbent assay (ELISA) of T cells genetically modified with the KRAS^G12D-reactive TCR cocultured with COS-7 cells cotransfected with nothing (Mock) or the indicated KRAS gene along with nothing (No HLA), or the HLA-B and -C alleles expressed by the patient. (E) IFN-γ ELISA of T cells genetically modified with the KRAS^G12D-reactive TCR cocultured with various pancreatic cancer cell lines transduced with nothing (Mock) or the HLA-C*08:02 allele. The presence or absence of endogenously expressed KRAS^G12D is shown. AS, ASPC-1; MD, MDA-Panc48; PK, PK-45p; FA, FA6-2; HP, HPAC; Bx, BxPC-3 (KRAS-wt); A8, A818.8 (KRAS^G12R); SK, SK-PC3 (KRAS^G12V); MI, MIA PaCa-2 (KRAS^G12C). (F) KRAS^G12D-reactive TILs enriched from (B) were cocultured for 6 hours with pancreatic cancer cell lines transduced with nothing (Mock) or the HLA-C*08:02 allele, and flow cytometry was used to assess CD107a expression and TNF production by means of intracellular cytokine staining. Autologous APCs (peripheral blood mononuclear cells) pulsed overnight with wt or KRAS^G12D 24-AA-long peptides were used as control target cells. Data are gated on CD8⁺ T cells expressing the KRAS^G12D-reactive TCR Vβ5.2. “>” ELISPOT assay is not accurate above ~500 spots. Error bars are ±SD. All data are representative of at least two independent experiments.

We used similar approaches to test whether mutation-reactive TILs could be detected in the seven remaining patients with metastatic colon, rectum, esophagus, bile duct, and pancreatic cancer and found mutation-reactive T cells in six patients (Table 1, figs. S2 to S7, and supplementary text). Our strategy of lowering the mutation-call threshold led to the identification of one immunogenic mutation that would not have been identified if we had used previous, more stringent methods to call mutations (the ZFYVE27 mutation in patient 4069) (Table 1 and tables S9 and S18). Both the number of mutation-reactive TIL cultures and the frequency of mutation-reactive T cells were often highly variable (Table 1 and figs. S1 to S7). Some of the immunogenic mutations identified in this report were in genes with known biological relevance to tumorigenesis, such as KRAS, PHLPP1, and API5 (Table 1) (18–20). AH of the identified mutations recognized by CD8⁺ T cells encompassed minimal T cell epitopes that were predicted to rank among the top 2% of peptides that bound to one or more of the patients’ MHC-I molecules (tables S21 and S23), and some of these epitopes were tested and found to be recognized by the appropriate T cells (figs. S4F; S6, C, E, J, and K; and S7C). Most of the mutations recognized by CD4⁺ T cells were encompassed within epitopes that were also predicted to bind with moderate affinity to at least one of the patients’ own HLA-II alleles (tables S22 and S23). Whole-transcriptome analysis revealed that the immunogenic mutations were in genes demonstrating a wide range of expression levels [2.9 to 185.4 fragments per kilobase of transcript per million mapped reads (FPKM)] within the metastatic lesions (tables S21 and S22).

To determine the endogenous frequency of the mutation-reactive T cells (table S19) infiltrating the metastatic lesions, we performed TCR-Vβ deep sequencing on the cryopreserved metastatic tumor lesions. As shown in Table 1 and fig. S8, the frequency of the identified mutation-reactive T cells infiltrating the metastatic lesions was variable, ranging from 0.009 to 1.25% of all T cells within a given tumor. Of the 17 identified mutation-reactive TCRs, four ranked within the top 10 most frequent TCRs within the tumor (rank range, 3 to 2718) (Table 1). The low ranking of some mutation-specific TCRs could suggest poor clonal expansion, survival, and/or infiltration of these T cells in the tumor environment. The antigens recognized by the other TCRs in the tumor are unknown but could be comprised of nonmutated tumor antigens, nontumor antigens, and/or other mutated antigens not identified by our assays. Often, only a minority of TIL cultures derived from the same metastatic lesion harbored detectable levels of IFN-γ-producing mutation-reactive T cells, and different TIL cultures were enriched for T cells reactive to different mutations (Fig. 1, Table 1, and figs. S1 to S7). This heterogeneity and relatively low frequency of neo-epitope T cell reactivity may be a function of the intratumoral genomic heterogeneity observed in human cancers (21, 22) and may partially explain the lack of efficacy observed in patients with metastatic GI cancers treated with immune-based therapies.

In vivo antitumor activity can theoretically be achieved by T cells targeting either passenger and/or driver mutations, so long as a sufficient number of tumor cells express the targeted mutation (or mutations). The immunological targeting of driver mutations (23–25) maybe desirable over passenger mutations, and it is thus pertinent that we identified a HLA-C*08:02–restricted TCR that recognizes the shared KRAS^G12D driver mutation. Because the adoptive transfer of autologous peripheral blood lymphocytes—genetically engineered to express tumor-reactive TCRs or chimeric antigen receptors—can elicit regression of widespread cancer in some patients with metastatic disease (1), the isolated TCR may extend TCR gene therapy to HLA-C*08:02 patients whose tumors express KRAS^G12D. The KRAS^G12D mutation is expressed in ~45% of pancreatic adenocarcinomas (26), ~13% of colorectal cancers (27), and at lower frequencies in other common solid cancers, and the HLA-C*08:02 allele is expressed in up to ~8 and ~11% of American Caucasoid and Black ethnicities, respectively. Thus, in the United States alone, thousands of patients with GI cancers each year would potentially be eligible for immunotherapy with this single KRAS^G12D-reactive TCR.

Together with our recent Report demonstrating the existence of mutation-reactive CD4⁺ T cells in a patient with metastatic cholangiocarcinoma (14), we have found that 9 out of 10 patients with metastatic GI cancers elicit T cell responses against at least one somatic mutation expressed by their tumors. The neo-epitope T cell responses identified in our patients were elicited against epithelial tumors with relatively low to moderate mutation burdens that are not susceptible to effective therapy with current checkpoint modulators (2, 7). Of our 10 patients described in (Table 1, four have been treated with enriched populations of T cells targeting predominantly one mutated antigen expressed by their autologous tumor. Patient 3737 (14) continues to experience tumor regression ongoing at 20 months posttreatment, whereas patient 4069 had a transient regression of multiple lung metastases, and 4007 and 4032 had no objective response. For patient 3737, over 23% of the circulating T cells at 1 month after treatment comprised the adoptively transferred mutation-specific T cells, whereas limited persistence (less than 1%, and even undetectable in some cases) of the mutation-reactive T cells was observed in the blood of the other three patients treated with an enriched population of mutation-specific T cells. These results suggest a need for enhancing the potency and persistence of adoptively transferred mutation-specific T cells through strategies such as the introduction of mutation-specific TCRs into naive or central memory T cells with high proliferative capacity, the targeting of driver mutations, the simultaneous targeting of multiple mutations, and/or combining adoptive cell therapy with other immunomodulators such as checkpoint inhibitors. Nonetheless, the observation that virtually all patients with metastatic GI cancers harbor tumor-mutation-specific T cells provides opportunities for the development of personalized vaccine and/or adoptive cell therapies that target immunogenic tumor mutations expressed by common epithelial cancers.