Eradication of Large Solid Tumors by Gene Therapy with a T-Cell Receptor Targeting a Single Cancer-Specific Point Mutation

Cancers are thought to evolve from normal cells to a malignant cell population that often is highly diversified genetically and epigenetically. Most of this evolution seems to occur even before the tumor reaches a size that can be detected clinically. Genetic diversity results from mutations in genes that make the incipient cancer prone to accumulate additional mutations (e.g., mutations in p53 or possibly in p68, a co-activator of p53; ref. 1) and/or drive malignant growth. The diversity provides cancers with variants that can escape specific treatments (2). Mutations in cancers regularly include nonsynonymous nucleotide substitutions that result in single amino acid substitutions (AAS). These somatic mutations are the basis of truly cancer-specific antigens (also referred to as neoantigens or neoepitopes) recognized by T cells (3).

This discovery two decades ago was soon confirmed in human cancer (4, 5) and followed by the suggestion that tumor-specific T-cell antigens recognized by T cells on a cancer may be predicted by analyzing its somatic mutations, an approach referred to as “reverse immunology” (6). However, wide interest in targeting tumor-specific AAS only developed more recently (7, 8) following the availability of high-throughput genomic analysis and computer algorithms predicting neoepitopes. Interest further intensified with the finding that the clinical efficacy of tumor-infiltrating T cells (TIL; ref. 9) or immune checkpoint inhibitors (10, 11) may correlate with endogenous T-cell responses to neoepitopes. Consequently, stimulation of T cells with mutant peptides is being combined with immune checkpoint inhibitors (12) that can activate and/or rescue exhausted endogenous cancer-specific T cells (13). The success of these approaches, however, may be limited because they rely on stimulating tolerant T cells, which revert to an ineffective state after transient activation (14).

To avoid these limitations, unbiased peripheral T cells of patients can be engineered to express antigen receptors of a chosen specificity. For example, T cells expressing anti-CD19 chimeric antibody receptors can eradicate advanced leukemia and lymphomas even though normal B cells are also ablated (15). Because targeting other non-mutant self-antigens on tumors by adoptive T-cell transfer caused severe toxicity and only little efficacy (16), we engineered T cells to express TCRs recognizing a shared (or trunk) mutation to determine the conditions by which targeting a single cancer-specific point mutation eradicates a progressively growing, genetically heterogeneous cancer. To our knowledge, this is the first study to show that under the right conditions, mutation-specific TCR gene therapy provides an effective, truly tumor-specific cancer treatment.

Genomic DNA and total RNA were extracted, whole-exome and RNAseq libraries were prepared, and sequenced by 150 bp paired-end reads on NextSeq 500 Desktop Sequencer or HiSeq2500 Sequencer (Illumina). Detailed information are provided in Supplementary Methods.

Low-quality reads (more than 80% of bases were base quality less than 20) were excluded using the FASTX Toolkit (Gregory J. Hannon, Cold Spring Harbor Laboratory, NY). For whole-exome sequencing, sequence reads were mapped to the mouse reference genome mm10 using Burrows-Wheeler Aligner (BWA v0.7.10; ref. 17). Possible PCR duplicated reads were removed (Picard v1.91; Broad Institute, Cambridge, MA), and read pairs with a mapping quality <30 and mismatches more than 5% of read length were excluded. For RNAseq, sequence reads were mapped to the mouse reference genome mm10 using STAR (18), and possible PCR duplicated reads and reads with a mapping quality <30 were excluded. Somatic variants were called using a Fisher exact test-based method (19) and annotated using ANNOVAR (20). Detailed information on variant calling are provided in Supplementary Methods.

Binding affinity to H-2K^b and H-2D^b was predicted for 8-, 9-, and 10-mer peptides containing nonsynonymous variants and found to be expressed in Bulk (variant reads ≥1 in RNAseq). Amino acid sequences were analyzed by the NetMHC 3.4 server (21) and binding affinity is expressed as IC₅₀ of the respective neoepitope.

MC57 is a methylcholanthrene-induced C57BL/6-derived fibrosarcoma [provided by Pamela Ohashi (University of Toronto, Toronto, Ontario, Canada), with permission of Hans Hengartner (University Hospital Zurich, Zurich, Switzerland)]. Its transfectants MC57-SIY and MC57-mp68 were generated in our laboratory and have been described (22, 23). 8101 originated in a UV-treated C57BL/6 mouse and was generated in our laboratory (24). Fragments of the original 8101 tumor (1–2 mm in size) were frozen. Bulk is a primary tumor cell culture derived from approximately 20 fragments of the original 8101 tumor. The cell line Bulk-mp68 was generated using MFG-mp68-EGFP as described previously (23). All tumor cell lines, RMA-S cells (see ref. 24) and Plat-E packaging cells (see ref. 25) were maintained in DMEM supplemented with 10% FCS (Gemini Bio-Products). Unless otherwise indicated, cell culture reagents were purchased from Invitrogen (Life Technologies). Before use, tumor cell lines were authenticated by sequencing and/or co-culture with antigen-specific T cells. RMA-S and Plat-E cells were shortly passaged after thawing of the initial frozen stock to generate master cell banks. Working batches were passaged no longer than 3 months and authenticated by cellular morphology.

T-cell clones specific for mp68 (H-2K^b:SNFVFAGI) were generated by immunizing C57BL/6 mice with Bulk tumor cells using procedures previously described (26) except that IL2 (6 IU/mL) was used instead of T-cell growth factor for expansion and cloning of the T cells. Specificity of established T-cell cultures was analyzed by specific lysis of RMA-S cells loaded with mp68 (SNFVFAGI) or p68 wild-type peptide (SNFVSAGI, both peptides: 7.8 × 10⁻⁹ mol/L, provided by Steven Meredith, University of Chicago, Chicago, IL) and Bulk tumor cells expressing the antigen at natural levels as described (24). 1D9 TCR sequences were determined by 5′-RACE-PCR following manufacturer's instructions (Life Technologies): TRAV1-CAVRSDTNAYKVIF-TRAJ30, TRBV19-CASSKRLSSYEQYF-TRBJ2-7. The 1D9 TCR was modified to form a second disulfide bond (27) and codon-optimized genes (GeneArt, Life Technologies) were integrated into MP71-PRE as described (MP71-1D9; ref. 22).

C57BL/6 and Rag^−/− (B6.129S7-Rag1^tm1Mom) mice were purchased from the Jackson Laboratory. To generate a mouse that produces 1D9-transgenic T cells (1D9tg), the retrovirus vector plasmid encoding MP71-1D9 was injected into the nuclei of C57BL/6 zygotes as described previously (28). Transgenic mice backcrossed into the Rag^−/− background solely expressed CD8⁺ 1D9tg T cells (1D9xRag^−/−, B6.129S7-Rag1^tm1Mom Tg(MP71-1D9)1Kshs). YFPx1D9xRag^−/− double-transgenic mice were generated as a pure source of fluorescently labeled 1D9tg T cells. YFP-transgenic Rag^−/− mice have been described (29). OT-IxRag^−/− mice (B6.129S7-Rag1^tm1Mom Tg(TcraTcrb)1100Mjb) were obtained by breeding Rag^−/− and TCR OT-I-transgenic mice (provided by Matthew Mescher, University of Minnesota, Minneapolis, MN). Monospecific T cells of OT-IxRag^−/− mice were used for the generation of 1D9-transduced T cells (1D9td). H-2K^bD^b−/−xRag^−/− mice have been described (22). All animals were maintained at the University of Chicago facilities. The Institutional Animal Care and Use Committee at the University of Chicago approved all animal experiments, and all experiments were performed to conform to the relevant regulatory standards.

Plat-E packaging cells were transiently transfected with MP71-1D9 or MP71-2C (22) by calcium phosphate precipitation. Virus supernatant was harvested and used for transduction of T cells isolated from OT-I-Rag^−/− mice as detailed in Supplementary Methods.

Tumor cells were injected s.c. onto the shaved back of Rag^−/− mice (1 × 10⁷ Bulk, 2 × 10⁶ MC57-mp68 or MC57-SIY). Tumor volumes were measured along 3 orthogonal axes every 3 to 4 days and tumor volume was calculated as abc/2. 1D9 T cells were administered i.v. on day 2 after transduction (transduced splenocytes of one OT-IxRag^−/− mouse per recipient, around 5 × 10⁶ cells). Naive 1D9tg T cells were prepared from spleens of 1D9xRag^−/− mice and administered i.v. without further culturing (splenocytes of one 1D9xRag^−/− mouse per recipient, around 5 × 10⁶ cells). Local radiation of tumors was done as described (30). Briefly, mice were locally irradiated using an X-ray generator (PCM 1000; Pantak) at a dose of 10 Gy. Mice were shielded with a lead cover leaving the subcutaneous tumors exposed through an opening on the side. Mice were irradiated when tumors reached the size of approximately 300 mm³ and T cells were transferred one day later.

Tumors were isolated, single-cell suspensions were generated by enzymatic digestion and CD11b⁺ cells were enriched using magnetic cell sorting as described (31). Sorted cells were used for co-cultures with 1D9td or 2Ctd T cells.

To analyze antigen presentation by indicated tumor and stroma cells, co-cultures were performed with 1D9td and 2Ctd T cells as described (31). T-cell activation was assessed by measuring IFNγ content of 24 hours co-culture supernatants by ELISA (Femto HS High Sensitivity, eBioscience), following the manufacturer's protocol. Cell-mediated lysis of indicated target cells was determined by standard 4.5 hours chromium release assay as described (31).

RNA was isolated from indicated tumors and PCR analysis to detect expression of the wild-type or mutant p68 allele was done as described (23).

Erythrocytes in blood samples were lysed by ammonium chloride treatment. One μg Fc block (anti-mouse CD16/CD32, 2.4G2) was added to samples and cells were incubated with 50 μL PBS containing 1 μg of indicated anti-mouse antibodies: CD4 [GK1.5, allophycocyanin (APC), BioLegend], CD8 (53–6.7, PerCP; BioLegend), TCRvβ6 [RR4-7, fluorescein isothiocyanate (FITC) or APC, BD Biosciences] or H-2K^b:mp68 multimers (PE, NIH tetramer core facility, Bethesda, MD). Samples were washed in PBS and acquired using Calibur or LSR II flow cytometers (BD Biosciences). Data analysis was performed using FlowJo software (TreeStar).

Implantation of windows, cancer cell injection, confocal microscopy, and data analysis were done as described (29, 32). Detailed information are provided in Supplementary Methods.

The 8101 tumor was induced by repeated exposure of a normal C57BL/6 mouse to UV light (24). The tumor on the back of this mouse was excised along with heart and lung tissue to generate a strain of normal fibroblasts (HLF; ref. 24). Approximately 20 fragments of the autochthonous tumor 8101 were minced to generate an uncloned culture of 8101 cancer cells (23), herein after called “Bulk” (Fig. 1A). This culture was only minimally expanded in vitro to provide cancer cells for sequencing and tumor induction in vivo. Additional 20 fragments were kept for individual analysis to capture the spatial heterogeneity of mutations in the autochthonous tumor and to properly select a mutation as target for adoptive T-cell therapy (ATT) that will likely be present in all parts of the tumor (Fig. 1A).

Whole-exome next-generation sequencing of Bulk identified 3,710 mutations (average coverage depths: 146×) using HLF DNA as baseline representing normal tissue from the host (Fig. 1B, analysis pipeline see: Supplementary Fig. S1 and detailed data in Supplementary Table S1). RNAseq showed that 33% of these mutations (total number: 1,207) were expressed in Bulk (Fig. 1B). Because mutant peptide epitopes with the highest affinity to MHC-I (IC₅₀: ≤4 nmol/L) appear to be required for eradication of large established cancer by adoptively transferred T cells (31), we first wanted to determine which of the 1,207 mutations would be predicted as preferred epitope. We made use of an open access computer algorithm (NetMHC 3.4; ref. 21) to predict mutant peptides of Bulk that bind to MHC-I. A total of 1,106 potential neoepitopes could be determined when screening the expressed mutated genes for peptides (octa-, nona-, or decamers) that harbor a mutation and bind to H-2K^b or H-2D^b (Fig. 1B, IC₅₀: ≤500 nmol/L). Only 194 of these mutant peptides were found to bind either one of the MHC molecules with higher affinity (Fig. 1B, IC₅₀: ≤50 nmol/L). Only 15 mutations (0.4% of all mutations detected in Bulk) resulted in neoepitopes that had an IC₅₀ of ≤4 nmol/L. Among those, highest RNA expression was found for the mutant gene of the DEAD box helicase p68 encoding a serine to phenylalanine exchange at position 551 (Fig. 1C, in red, mp68 in the following).

The above results suggested that mp68 should be an excellent target for adoptively transferred T cells to eradicate established 8101 tumors. To prevent treatment failure, it is essential that the vast majority of cancer cells express the targeted antigen. To this end, we sequenced the DNA of 20 tumor fragments that were randomly selected from the original tumor. Each fragment contained 3,900 ± 561 mutations (mean ± SD, average coverage depths: 81× (±10), Supplementary Table S2). Mutations of the fragments clustered as illustrated in Fig. 1D, demonstrating differences in clonotypic composition (analysis pipeline see: Supplementary Fig. S2). This spatial diversification is confirmed by a large number of mutations uniquely present in individual fragments (numbers given at the top of each branching). Twenty five percent of all mutations (total number: 1,792) were shared by all fragments and represent the “mutational trunk.” One of these mutations was p68^S551F (mp68) described above. The average read frequency of the mutant allele (30 ± 8%; mean ± SD) suggested that the mp68 gene was represented appropriately for expression. Another mutation, p53^S238A, was previously used to mark the common ancestry of the 8101 tumor (33) and was detected in all but 2 of the tumor fragments (Fig. 1D).

As reported above, algorithm analysis of exome and RNAseq data selected mp68 as highest ranking neoepitope. This is precisely the same neoepitope, we had previously found to be the immunodominant rejection antigen in 8101 by direct immunology (24). To obtain T cells recognizing this tumor-specific antigen, we generated a number of mp68-specific T-cell clones from T cells of tumor-free mice that had been immunized with the uncloned 8101 cancer cells. Tumor-reactive T cells were expanded by repeated stimulation in vitro with irradiated 8101 tumor cells and subsequently cloned by limiting dilution. Figure 1E shows that several clones achieved half maximal lysis of RMA-S cells at low peptide concentration (7.8 pmol/L SNFVFAGI), whereas these T cells were not reactive against the corresponding p68 wild-type peptide (SNFVSAGI, specific lysis <10%, data not shown). These high avidities are in agreement with the previously determined EC₅₀ of another mp68-specific T-cell clone (24). Decisive for selecting 1D9 T cells as source of the TCR was the high reactivity against Bulk tumor cells expressing mp68 at natural levels (Fig. 1F). Thus, we determined the 1D9 TCRα- and β-chain sequences and cloned them into a retrovirus vector (Supplementary Fig. S3A) for transduction of CD8⁺ T cells (1D9td). The vector was also used to generate a 1D9 TCR-transgenic mouse (1D9tg, 1D9xRag^−/− mice, Supplementary Fig. S3B) for comparison because many previous studies used TCR-transgenic T cells for adoptive therapy to target surrogate or mutated antigens on tumors (34). However, only TCR-transduced T cells are realistic for patient therapy. As recipient T cells for transduction, we used CD8⁺ T cells from OT-IxRag^−/− mice, which only express the OT-I-TCR specific for the ovalbumin peptide SIINFEKL, an irrelevant target for our model. Transduction efficacy was measured using TCRvβ6-specific antibodies (representative staining is shown in Supplementary Fig. S3C). The specific lysis of Bulk tumor cells by 1D9td T cells (Fig. 2A) was comparable with the lysis determined for the original 1D9 T-cell clone (Fig. 1F).

After characterization of mp68 as an ideal target for ATT (high binding affinity to MHC-I, expressed in all regions of the primary tumor, and recognized by a TCR derived from a high-avidity T cell clone), we next assessed whether Bulk tumors could be rejected by mp68-specific TCR gene therapy. Thus, Rag^−/− mice bearing Bulk tumors were treated by adoptive transfer of approximately 5 × 10⁶ 1D9 T cells. ATT led to initial regression of Bulk tumors, but all tumors eventually relapsed (Fig. 2B). The efficacy of 1D9tg and 1D9td T cells was similar. Whole-exome sequencing of tumor reisolates (Reis#1 and Reis#2) revealed a mutational pattern, which was almost identical to the original Bulk tumor (Fig. 2C). However, the variant allelic frequency (VAF) of mp68 seemed to be reduced (Fig. 2D, Supplementary Table S1). This reduction was apparently related to the escape from mp68-directed T-cell therapy as the proportion of the p53^S238A mutation remained rather unchanged. Furthermore, Reis#1 and Reis#2 failed to express the mp68 RNA (Fig. 2E). RNA derived from in vitro-cultured Bulk tumor cells, an mp68-negative 8101 variant (PRO1A; ref. 33) and a Bulk tumor reisolate from a mouse not treated with 1D9 T cells were used as controls. These results demonstrated that primary tumors escape mp68-specific ATT by loss of mp68 expression, which is consistent with previous studies (23).

The failure of eradicating 8101 tumors by 1D9 T-cell therapy raised the question whether 8101 was a cancer exceptionally difficult to treat or whether the destructive force of just CD8⁺ T cells grafted with a single type of TCR targeting a single tumor-specific mutant peptide was insufficient to achieve eradication of any large solid tumor. We therefore made use of the MC57 tumor model in which we had previously eradicated large solid tumors by adoptive transfer of CD8⁺ TCR-transgenic T cells targeting surrogate antigens (31, 35). MC57 cancer cells were transduced with a retroviral vector to express the mp68 peptide coupled to GFP. This allowed us to express high levels of the targeted antigen in >99.9% of MC57-mp68 cancer cells (Supplementary Fig. S4). 1D9td T cells recognized MC57-mp68 but not mock-transfected MC57 cells in standard cytotoxicity assays (Fig. 3A). Remarkably, a single adoptive transfer of approximately 5 × 10⁶ 1D9 CD8⁺ T cells eradicated an established solid MC57-mp68 tumor in almost all of the Rag^−/− mice (8/9 rejected, Fig. 3B, left). 1D9tg T cells were similarly effective as 1D9td T cells. The growth of MC57-mp68 was not affected by transfer of T cells transduced with a TCR of unrelated specificity (2C TCR, Mock, Fig. 3B, right). These experiments proved that T cells modified to express TCRs with a single specificity for a single mutant antigen can destroy very large and long-established solid cancers.

Given the importance of stromal destruction for preventing relapse from T-cell therapy in the MC57 and other experimental models (36), we asked whether recognition of mp68 on stromal cells by 1D9 T cells was required for eradication of MC57-mp68. Surprisingly, large MC57-mp68 tumors growing in Rag^−/− mice devoid of MHC-I expression (H-2K^bD^b−/−) were rejected by 1D9 T cells (4/4, Fig. 3C), indicating that 1D9 T cells received sufficient stimulation directly from MC57-mp68 tumor cells in absence of stromal antigen cross-presentation. However, large amounts of IFNγ were released when stromal cells (mostly macrophages) from MC57-mp68 tumors were used as stimulators (Fig. 3D). The levels of IFNγ were comparable with those stimulated by stromal cells from tumors expressing the SIY surrogate tumor antigen, which is consistent with our previous studies (31).

To further investigate whether stromal cross-presentation of mp68 contributed to stromal destruction, we made use of longitudinal confocal microscopy (32) to follow the course of tumor destruction. MHC-I-positive and -negative Rag^−/− mice with approximately 2-week-old MC57-mp68 tumors were treated with 1D9 T cells (Fig. 4A). In both situations, T cells rapidly vanished from the circulation after i.v. injection, but appeared in the tumor usually 2 to 3 days after transfer (designated as day 0 in Fig. 4A). T cells appeared to be arrested in their movement in close vicinity of tumor vessels. Within the following 24 hours, the blood flow completely ceased with or without cross-presentation (evidenced by immobile DiD-labeled erythrocytes (round) as compared with erythrocytes dashing through the tumor vessels (lines) on day 0. The extravasation of DiD-labeled erythrocytes indicated that the tumor-endothelial barrier was destroyed. This evidence of vascular leakage was the microscopic correlate to the dark-red discoloration of the tumor observed after successful ATT. Nevertheless, destruction of cancer cells (indicated by loss of green fluorescence observed in viable cancer cells) was clearly accelerated when 1D9 T cells could recognize cross-presented mp68 on CD11b⁺ tumor stromal cells (day 1 in both Fig. 4A and Fig. 4B and C). This finding correlated with observations from in vitro experiments showing that high levels of IFNγ were only released by 1D9 T cells after stimulation with mp68 cross-presenting stromal cells (Fig. 3D). Stimulation by cancer cells alone was less efficient and led to 10× lower IFNγ release. As control for the experiment, 1D9 T cells were transferred into a mouse bearing a MC57 tumor expressing an irrelevant antigen (SIYRYYGL coupled to GFP). Interestingly, 1D9 T cells did neither show homeostatic expansion (Supplementary Fig. S5) nor infiltration in the tumor (Supplementary Fig. S6). In contrast, 1D9 T cells showed strong expansion in mice bearing MC57-mp68 tumors, irrespective of H-2K^b expression by the host (Supplementary Fig. S5). Together, these data indicate that direct recognition of highly and uniformly expressed mp68 sufficed for the infused 1D9 T cells to localize to the solid tumor, find the cancer cells as stimulators to proliferate and to release sufficient amounts of cytokines to destroy the tumor vasculature and eradicate the cancer. Nevertheless, when the neoantigen was not cross-presented, cancer cell destruction was significantly delayed. Thus, stromal cross-presentation could become a determining factor for eradication of antigen-loss variants when the target antigen is not uniformly expressed by all cancer cells. This is suggested by previous experiments of adoptive therapy of tumors expressing surrogate antigens after transfection rather than transduction and containing significant numbers of antigen-negative cancer cells (see ref. 36).

Because antigen cross-presentation accelerated tumor destruction and has facilitated tumor rejection in other models (35), we assessed whether the escape of 8101 tumors by outgrowth of antigen-loss variants correlated with lack of stromal cross-presentation of mp68. Indeed, CD11b⁺ stromal cells isolated from Bulk tumors stimulated far less IFNγ release by 1D9 T cells than CD11b⁺ stromal cells isolated from MC57-mp68 [Fig. 5A (left) compared to Fig. 3D]. Thus, the level of mp68 in Bulk tumor cells may not have provided sufficient antigen for direct or cross-presentation to stimulate sufficient cytokine release to cause stromal destruction and antigen-negative variants could escape (Fig. 2B). We then generated Bulk tumor cells that were transduced to express higher levels of mp68 antigen (Bulk-mp68, Supplementary Fig. S4B). Bulk-mp68 tumor cells were killed effectively in vitro (Fig. 5B) and stimulated approximately 20x higher IFNγ release by 1D9 T cells (Fig. 5A, right) by direct antigen presentation of the Bulk-mp68 cancer cells or stroma cells derived from Bulk-mp68 tumors. Most importantly, 1D9 TCR gene therapy eradicated 4 of 5 Bulk-mp68 tumors established for over a month and at least 1 cm in diameter (Fig. 5C). Thus, resistance of primary tumors to 1D9 treatment was overcome by increasing the level of mp68 antigen that is naturally expressed by Bulk tumor cells. However, because the antigen was uniformly expressed on the transduced Bulk tumor cells, there were probably no variants that could require stromal presentation to be eradicated. In any case, proper expression of the mp68 antigen allowed eradication of large Bulk tumors by adoptive transfer of peripheral T cells engineered to express a single type of TCR and specific for a single AAS.

We have previously shown that chemotherapeutic drugs or local irradiation can synergize with ATT to achieve eradication of cancers that fail to be rejected because they express low levels of the targeted antigen (30). The efficacy of the combined treatment correlated with antigen loading onto the stroma from cancer cells damaged or stressed by the chemo- or radiotherapy. We followed the same experimental design (30) to improve the effectiveness of mutation-specific TCR gene therapy of heterogeneous tumors expressing the original autochthonous mp68 neoepitope. Indeed, the combination with local irradiation of Bulk tumors significantly improved therapeutic outcome after adoptive transfer of approximately 5 × 10⁶ 1D9 T cells (rejection in 4/8 mice, Fig. 5D) whereas treatment of Bulk tumors with either 1D9 T cells or local irradiation alone was ineffective.

Most human cancers have been present for years and the smallest lesions when first detected are usually at least approximately 1 cm in diameter harboring approximately 10⁹ cancer cells. These two facts give guidance for shaping preclinical models. First, the size of the targeted cancer cell population determines the likelihood of escape and tumors in mice must have a size that provides them with a realistic chance of escaping. Escape remains the biggest threat to any cancer therapy, including ATT (37). The time of establishment of experimental tumors is the second critical variable. Tumor transplantation invariably induces acute inflammation that greatly helps the host (or experimentalist) to eradicate transplanted tumors at early stages (i.e., during the first two weeks after transplantation; ref. 38). When a tumor is targeted a month after transplantation, therapeutic procedures often become ineffective even when tumors remain rather small (39). Therefore, it is important that TCR gene therapy described in this study eradicated tumors that were not only large (∼1 cm in diameter) but had also been established for a month or longer.

Our studies of the 8101 tumor demonstrate that “reverse immunology,” the sequencing of a tumor's genetic information and the computational prediction of neoepitopes, predicts the neoantigen mp68 we had previously identified as immunodominant rejection antigen of 8101. Although our study showed consistency between direct and “reverse immunology”, tracing of suitable neoepitopes without guidance by specific T cells can be misleading. For example, predicting the binding of cysteine-containing peptides to MHC is problematic (40) and some neoepitopes predicted to bind may be destroyed by proteasomal cleavage (25). Predicting neoepitopes in cancer by combining sequencing with mass spectometry analysis of eluted peptides (41) or multimer staining of specific T cells (12) could improve reliability.

We suggest that the clinically relevant situation is likely to be mirrored by tumors composed of cancer cells expressing the tumor-specific mutations at natural levels, resembled by Bulk used in this study. Even though we targeted an immunodominant rejection antigen expressed in every analyzed part of the original tumor, previously unnoticed antigen-negative variants or cancer cells that lost antigen expression foiled cancer eradication by T-cell therapy. We targeted a mutation in p68, a coactivator of p53 (1), which probably played a role as driver in the formation of the cancer but sustained expression was not required for tumor maintenance. Nevertheless, mp68 provided a subnanomolar affinity neoepitope expressed in comparison with other neoepitopes of the tumor at comparatively high levels. Also, mp68 was found in all analyzed regions of the cancer, which suggested that this neoepitope was not affected by usual spatial tumor heterogeneity. Better targets would obviously be driver mutations essential for cancer cell function and survival and expressed at higher levels so that antigen escape does not occur (3, 42–44). However, such ideal targets may be rare. Given the remarkable genetic and phenotypic diversity of cancers (45), we therefore believe that escape by antigen-negative variants will likely remain an important problem in targeting tumors, experimentally or clinically.

We propose several complementary solutions to solve the problem of escape from TCR gene therapy. First, targeting multiple independent neoepitopes on the same cancer could reduce the chance for therapy-induced selection of escape variants (26, 46). For example, the 8101 primary cancer contained another neoepitope that was retained by mp68-loss variants and was recognized by a different cytolytic T-cell clone (24). Second, tumor-specific CD4⁺ T cells facilitate the accumulation of CD8⁺ T cells in the tumor microenvironment (47) and also cooperate with CD8⁺ T cells during the effector phase in bystander killing of cancer variants thereby preventing tumor escape (48). Thus, a properly chosen combination of two mutation-specific TCRs, one restricted to MHC-I and the other to MHC-II is likely to be a fruitful approach. Certainly, focusing on highly expressed mutant proteins that are cross-presented by the tumor stroma, and thereby sensitize the stroma for destruction will greatly increase the killing of antigen-negative variants as bystanders (30, 35). Finally, TCR gene therapy could clinically be combined with adjuvant therapy, like chemotherapy or irradiation as used in this study. For example, stereotactic irradiation can be given to many sites of tumor growth in patients. Even though such treatments may fail to kill a significant number of tumor cells, irradiation may increase trafficking of T cells to irradiated tumors (49). Furthermore, chemo- or radiotherapy can increase the amount of antigen released from cancer cells (30) as can be shown when a high affinity TCR probe is available for measuring the amount of peptide:MHC on stromal cells. This seems to be of particular importance when targeting neoepitopes whose expression levels do not support sufficient cross-presentation and stimulation of T cells. Synergy has also recently been reported between radiotherapy and checkpoint blockade (50). Together, mutation-specific TCR gene therapy could benefit from various additional clinical treatments.

No potential conflicts of interest were disclosed.

Conception and design: M. Leisegang, B. Engels, K. Schreiber, H. Schreiber

Development of methodology: B. Engels, K. Schreiber, A. Arina, J. Duraiswamy, W. Uckert, H. Schreiber

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): B. Engels, K. Schreiber, P.Y. Yew, C. Idel, A. Arina, Y. Nakamura

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Leisegang, B. Engels, K. Schreiber, K. Kiyotani, C. Idel, J. Duraiswamy, R.R. Weichselbaum, Y. Nakamura, H. Schreiber

Writing, review, and/or revision of the manuscript: M. Leisegang, B. Engels, K. Schreiber, C. Idel, J. Duraiswamy, R.R. Weichselbaum, Y. Nakamura, H. Schreiber

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K. Schreiber, P.Y. Yew, C. Idel, W. Uckert, H. Schreiber

Study supervision: Y. Nakamura, H. Schreiber

The authors thank Kordelia Hummel for excellent technical support and David Binder for help with the irradiation experiments. The authors thank Vytas Bindokas for guidance in the longitudinal imaging studies, Linda Degenstein at the University of Chicago Transgenics/ES Cell Technology Mouse Core Facility as well as the University of Chicago Flow-Cytometry Core Facility. The authors also thank Thomas Blankenstein and Michael Spiotto for providing suggestions regarding the article. This work was supported by the Cancer Center at the University of Chicago.

NIH grants R01-CA22677 (to H. Schreiber, Y. Nakamura, and W. Uckert) and R01-CA37156 (to H. Schreiber), Collaborative Research grant of the Berlin Institute of Health (to H. Schreiber and W. Uckert), Einstein-Stiftung Berlin (to H. Schreiber)

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.