Induction of Neoantigen-Specific Cytotoxic T Cells and Construction of T-cell Receptor–Engineered T Cells for Ovarian Cancer

Ovarian cancer is the fifth leading cause of cancer-related death among women, and patients are often diagnosed in an advanced stage (1). The majority of ovarian cancer patients experience relapse/recurrence after primary therapy and develop resistance to chemotherapy. Although the overall survival of ovarian cancer has improved due to advances in chemotherapy and surgery, a cure for metastatic ovarian cancer is still elusive (2–5).

Immune checkpoint inhibitors [anti–programmed cell death 1 (PD-1) antibodies and anti–programmed cell death ligand 1 (PD-L1) inhibitors] have achieved great success for several cancer types (6–9). Accumulating evidences from patients who responded well to immune checkpoint inhibitors imply that tumor regression is achieved by activation of cytotoxic T cells targeting neoantigens, which are generated mostly by nonsynonymous mutations in cancer cells (10–12). In addition, cytotoxic T cells targeting neoantigens were found to be enriched in tumor-infiltrating lymphocytes (TIL) in patients whose tumors responded well to adoptive TIL infusion therapy (13). However, the majority of cancer patients have had no clinical benefit from either immune checkpoint inhibitors or adoptive TIL infusion therapy. Therefore, it is critically essential to develop new strategies to further enhance host immune response. Adoptive T-cell therapy with T-cell receptor (TCR)–engineered T cells has received attention as a promising new strategy (12, 14). We reported TCR-engineered T cells targeting neoantigens could eradicate even a very large solid tumor in a mouse syngeneic tumor model (15).

To generate neoantigen-specific TCR-engineered T cells, it is imperative to establish an efficient pipeline to isolate neoantigen-specific T cells and to obtain their TCR sequence information. We established an in-house pipeline to identify neoantigen-specific TCRs using HLA-matched healthy donor–derived peripheral blood mononuclear cells (PBMC), particularly focusing on the efficiency and timeline for development (16). Our approach takes 2 weeks to induce the neoantigen-specific T cells and obtain TCR information from neoantigen-specific T cells sorted by peptide-loaded HLA-dextramers.

Although ovarian cancer has not been regarded as an immunogenic cancer, the presence of CD8⁺ T cells in ovarian cancers correlates with better prognosis (17, 18). Hamanishi and colleagues reported the clinical efficacy of anti–PD-1 Ab (Nivolumab) in patients with platinum-resistant ovarian cancer (19). Wick and colleagues reported the presence of a neoantigen-reactive T-cell subset in TILs from ovarian cancer patients (20). These reports suggest that cancer neoantigens can be a promising therapeutic target for ovarian cancer (21). Therefore, we have chosen ovarian cancer for this pilot study to validate our in-house pipeline for induction of neoantigen-specific T cells and identification of neoantigen-specific TCRs. We successfully induced neoantigen-specific T cells against a mutation found in ovarian tumor samples and established corresponding TCR-engineered T cells. We propose this method as a promising new therapy for ovarian cancer patients.

Seven patients with ovarian cancer who received surgery and are still being followed at the University of Chicago Medical Center were enrolled after obtaining written-informed consent. Tumor blocks and blood samples were obtained from each patient. The study protocol was approved by the Institutional Review Board of the University of Chicago (approval number 15-1738 and 16-0402). This study was conducted in accordance with the Declaration of Helsinki.

Tumor genomic DNA was extracted from paraffin-embedded tumor tissues using QIAamp DNA FFPE Tissue Kit (Qiagen). Genomic DNA from the patients’ blood was extracted using the AllPrep DNA/RNA mini Kit (Qiagen) as germline control DNA. Whole-exome libraries were prepared from genomic DNAs using a SureSelectXT Human All Exon V5 kit (Agilent Technologies) and sequenced by 100-bp paired-end reads on a HiSeq2500 Sequencer (Illumina).

Sequence alignment and mutation calling were performed as described previously (22). Briefly, sequence reads were mapped to the human reference genome GRCh37/hg19 using Burrows–Wheeler Aligner (v0.7.10; ref. 23). Possible PCR duplicated reads were excluded using Picard v1.91 (http://broadinstitute.github.io/picard/). Read pairs with mismatches more than 5% of read length and with a mapping quality of <30 were also excluded. Finally, single-nucleotide variations were called using the Fisher exact test–based method with the following parameters: (1) base quality of ≥15, (2) sequence depth of ≥10, (3) variant depth of ≥2, (4) variant frequency in tumor of ≥10%, (5) variant frequency in normal of <2%, and (6) Fisher P value of <0.05 (24).

HLA class I genotypes were determined by OptiType algorithm (25) using whole-exome data of normal controls. We then examined the binding affinities of all possible 8- to 11-mer peptides harboring each AA substitution to HLA-A molecule and filtered out with the predicted binding affinity to HLA-A of <500 nmol/L, using NetMHCv3.4 software (22, 26, 27). To exclude the no- or low-expression genes in ovarian cancer, we downloaded RNA sequencing data of The Cancer Genome Atlas (TCGA; ref. 28) ovarian serous cystadenocarcinoma from GDC Data Portal (https://portal.gdc.cancer.gov). The genes with reads per kilobase of exon per million mapped reads (RPKM) less than 100 were excluded from the neoantigen list for testing of ability to induce neoantigen-specific T cells. We could not examine RNASeq data-quantifying transcript expression in the patient's tumor samples, because we could get only paraffin-embedded tumor tissues for DNA extraction, but could not obtain the high-quality tumor RNA for RNASeq analysis.

Induction of neoantigen-specific T cells was performed following the protocol we developed previously (16). To examine the HLA-A genotype of healthy donors, PCR amplicon–based high-resolution HLA-A genotyping on MiSeq (Illumina) was performed in Scisco Genetics, Inc. Briefly, PBMCs from healthy donors were collected using Vacutainer CPT Cell Preparation Tube (BD Biosciences). CD8⁺ T cells were isolated from PBMCs using the Dynabeads CD8 Positive Isolation Kit (Thermo Fisher Scientific). CD8⁻ cells were used to generate monocyte-derived dendritic cells (DC) using plastic adherence methods, and were cultured in CellGro DC (Cellgenix) containing 1% human AB serum (ABS), 500 U/mL IL4 (R&D Systems), and 1,000 U/mL GM-CSF (R&D Systems) for 72 hours in Primaria 6-well plate (Corning, Inc.). Then, 100 U/mL IFNγ (PeproTech) and 10 ng/mL LPS (Sigma-Aldrich) were added in the culture medium to induce the maturation of DCs. DCs were pulsed with 20 μg/mL of the respective neoantigen peptides (Innopep) for 16 hours at 37°C. After the peptide pulse, DCs were treated with 30 μg/mL of mitomycin C (Sigma-Aldrich) at 37°C for 30 minutes and then cocultured with autologous CD8⁺ T cells in CellGro DC/5% ABS with 30 ng/mL IL21 (R&D Systems) on day 1 (each well contained 1.0 × 10⁵ peptide-pulsed DCs, 5 × 10⁵ CD8⁺ T cells). We examined the single to quintuple scales for each neoantigen candidates based on the total number of PBMCs from healthy donor. Three days later (day 4), 5 ng/mL IL7 (R&D Systems) and 5 ng/mL IL15 (Novoprotein) were added in the culture media. On day 6, the cultures were transferred to 12-well plate with CellGro DC/5% ABS with 5 ng/mL IL7 and 5 ng/mL IL15. On day 8, cultures were supplemented with CellGro DC/5% ABS with 10 ng/mL IL7 and 10 ng/mL IL15. On day 11, neoantigen-specific T cells were assessed using peptide-HLA dextramers (Immudex) for each neoantigen peptide by flow cytometry analysis. CD8⁺Dextramer⁺ T cells were sorted out and used for the following TCR sequencing analysis.

To assess the neoantigen-specific T cells, the cells were incubated with each neoantigen-specific dextramer for 10 minutes at room temperature and then incubated with anti-human CD8 Ab (clone HIT8a; BD Biosciences) at 4°C for 20 minutes. Negative control dextramer (WI3233, Immudex) was used to examine general background or unspecific staining on the donor analyzed. Anti-human CD137 Ab (clone 4B4-1, Miltenyi Biotec) was used to examine the cell surface expression of CD137. Anti-mouse TCRβ monoclonal Ab (H57-597; eBioscience) was used to assess the cell surface expression of engineered TCRs. CD8⁺Dextramer⁺ T cells were analyzed and sorted by flow cytometry (FACS LSRII, Aria Fusion; Becton Dickinson). Data analysis was performed using Flow Jo software (Treestar).

TCR sequencing was performed using the methods described previously (26, 29, 30). In brief, we extracted total RNAs from sorted CD8⁺Dextramer⁺ T cells by flow cytometry. The cDNAs with common 5′-RACE adapter were synthesized from total RNA using a SMART library construction kit (Clontech). The TCRA and TCRB cDNAs were amplified by PCR using a forward primer for the SMART adapter and reverse primers corresponding to the constant region of each of TCRA and TCRB. After adding the Illumina index sequences with barcode using the Nextera XT Index Kit (Illumina), the prepared libraries were sequenced by 300-bp paired-end reads on Illumina MiSeq platform, using a MiSeq Reagent v3 600-cycels kit (Illumina). Obtained sequence reads were analyzed using Tcrip software (30).

Both TCRA and TCRB sequences were codon-optimized, synthesized by GeneArt (Life Technologies), and cloned into pMP71-PRE as described previously (31). To increase TCR surface expression, we used TCRs with mouse-constant regions (32). Transient retroviral supernatants were generated, and PBMCs from healthy donors were transduced as described previously (33).

C1R (lacking HLA-A and HLA-B, B lymphoblast) was purchased from the American Type Culture Collection. C1R cells stably expressing HLA-A2 (HLA*02:01) (C1R-A02) or HLA-A24 (HLA-A24:02) (C1R-A24) were prepared by the transfection of the vectors encoding HLA-A*02:01 or HLA-A*24:02 gene. C1R cells were cultured under the recommendation of the depositor.

The plasmid DNAs designed to express a part of RFC5- or BRAP-mutated proteins (50 AA length and the mutation was placed in the center) and GFP were codon-optimized and synthesized by GeneArt. Linearized plasmid DNA was used as in vitro transcription template to produce mRNA using the mMESSAGE mMACHINE T7 Kit and Poly(A) Tailing Kit (Ambion). Electroporation was done with Gene Pulser Xcell (Bio-Rad) at 300 V and 250 μF. Immediately after electroporation, cells were returned to culture medium and incubated for 12 to 16 hours at 37°C and 5% CO₂.

Enzyme-Linked ImmunoSpot (ELISPOT) assay to detect IFNγ-secreting T cells was performed using a Human IFN-γ ELISpotPRO kit (MABTECH) according to the manufacturer's instruction. Briefly, antigen-presenting cells (APC) were pulsed with respective peptides at 37°C for 20 hours and 5% CO₂. T cells were pretreated with IL2 (35 U/mL) for 16 hours and then cocultured with the peptide-pulsed APCs (2 × 10⁴ cells/well) at 37°C for 20 hours in 96-well plate. Spots were captured and analyzed by an automated ELISPOT reader, ImmunoSPOT S4 (Cellular Technology Ltd.), and the ImmunoSpot Professional Software package, Version 5.1 (Cellular Technology Ltd.).

To measure the secreted cytokine levels in the supernatant, we used OptEIA Human IFN-γ ELISA set (BD Biosciences), OptEIA Human IL2 ELISA set (BD Biosciences), OptEIA Human TNF ELISA set (BD Biosciences), and Human IL-4 ELISA development kit (MABTECH). Briefly, APCs were pulsed with respective peptides at 37°C for 20 hours and 5% CO₂. T cells were pretreated with IL2 (35 U/mL) for 16 hours and then cocultured with the peptide-pulsed APCs (2 × 10⁴ cells/well) at 37°C for 20 hours in 96-well plate. The supernatant was transferred into a new 96-well plate, and each protein concentration was measured according to the manufacturer's instruction.

Cytotoxic assay was performed using a CytoTox 96 Non-Radioactive Cytotoxicity Assay kit (Promega) according to the manufacturer's instruction. Briefly, C1R-A02 cells were pulsed with respective peptides (1 μmol/L) at 37°C for 20 hours and used as target cells. Effector cells and target cells were incubated in 96-well plate at 5:1, 10:1, 20:1, and 50:1 ratios for 4 hours at 37°C and 5% CO₂. Experiments were conducted in triplicate. Maximum lactate dehydrogenase (LDH) release from target cells was measured by the addition of lysis solution. The spontaneous LDH release of effector and target cells was measured by separate incubation of the respective population. After 4-hour incubation, plate was centrifuged at 250 × g for 4 minutes. Supernatant was transferred to a new 96-well plate. The substrate (CytoTox 96 nonradioactive cytotoxicity assay kit; Promega) was added to each well, and the plate was incubated for 30 minutes in the dark at room temperature. Stop solution was added to terminate the reaction, and absorbance at 490 nm was recorded. The percentage of cytotoxic activity was calculated according to the following formula: % Cytotoxicity = [(Experimental – Effector Spontaneous – Target Spontaneous)/(Target Maximum – Target Spontaneous)] × 100.

The Student t test was performed for comparison of the percentage of cytotoxic activity between C1R-A02 pulsed with mutant and corresponding wild-type peptides. Statistical analyses were done using GraphPad Prism version 6.0 (GraphPad software). P value of <0.05 was considered to be statistically significant.

Through whole-exome sequencing of genomic DNAs from paraffin-embedded ovarian cancer tissues and corresponding normal cells of 7 ovarian cancer patients, we identified a total of 463 nonsynonymous mutations (35–97 nonsynonymous mutations in individual patients, Table 1, Supplementary Table S1). We then predicted the binding affinity of peptides including AA substitutions to individual HLA-A molecules that were estimated from the whole-exome sequence data of normal DNAs and obtained neoantigen candidate epitope sequences which showed IC₅₀ of 500 nmol/L or lower (Table 1; Supplementary Table S2). From these candidate peptides, we focused on neoantigen candidates which showed IC₅₀ of <50 nmol/L and further filtered potential peptides by high expression levels in the TCGA transcriptome data (median RPKM >100; ref. 28). We selected 14 neoantigen candidates (1–3 for each tumor sample, Table 2) to examine their ability to induce neoantigen-specific T cells using PBMCs isolated from healthy donors.

Among the 14 peptides we synthesized, 11 neoantigens were expected to be recognized by HLA-A2–restricted cytotoxic T lymphocyte (CTL), and the remaining three were by HLA-A24–restricted CTL. We obtained PBMCs from two healthy donors, one having HLA-A*02:01 and the other having HLA-A*24:02. After 11 days of coculturing CD8⁺ T cells with autologous DCs with/without each neoantigen peptide, we searched for CD8⁺ T cell's binding to each peptide-HLA-dextramer complex by flow cytometry. We detected CD8⁺HLA-dextramer⁺ T cells for three neoantigens were recognized by HLA-A2–restricted CTL (Fig. 1A). The proportions of CD8⁺HLA-dextramer⁺ T cells were 0.088% (1,348 cells) for RFC5_K160N, 0.0097% (248 cells) for BRAP_R543C, and 0.20% (300 cells) for GINS1_I87V.

We extracted total RNAs from sorted CD8⁺HLA-dextramer⁺ T cells for three peptides and performed TCRα and β sequencing using our previously published method (ref. 30; Fig. 1B). We observed dominant TCRA and TCRB sequences that accounted for >50 % in all three cases (Supplementary Table S3). Although the total cell numbers were very small, the dominant TCR frequencies of 50% or higher may imply that these T cells were likely to be expanded by neoantigen stimulation. We speculated that T cells with these dominant TCRA and TCRB pairs were likely to be neoantigen-specific and generated TCR-engineered T cells with these TCRA- and TCRB-paired sequences for further functional analysis. Low-frequency TCR sequences might derive from impurities arising from sorting a low-frequency population. It is also possible that low-frequency TCR sequences may also be neoantigen-specific. However, in this study, we focused on the validation of the most dominant TCR clones from a very small number of sorted T cells.

We constructed the retroviral vectors encoding for the RFC5_K160N-specific TCR, BRAP_R543C-specific TCR, and GINS1_I87V-specific TCR cDNAs, and transduced each of them into PBMCs from a healthy donor. Subsequently, we examined whether these TCR-engineered T cells bound to the HLA-dextramer loaded with the mutant or wild-type peptide. As shown in Fig. 2, RFC5_K160N TCR-engineered T cells, BRAP_R543C TCR-engineered T cells, and GINS1_I87V TCR-engineered T cells bound to the HLA-dextramer with the mutant peptide, but not to that with the wild-type peptide. On the other hand, GINS1_I87V TCR-engineered T cells bound to both mutant and wild-type peptide-HLA dextramers (Fig. 2).

As predicted from results of the HLA-dextramer–binding assay, ELISPOT assay showed IFNγ secretion in RFC5_K160N and BRAP_R543C TCR-engineered T cells in a mutant-specific and peptide dose–dependent manner using C1R-A02 cells with HLA-A2 (HLA-A*02:01) as APCs (Fig. 3A), but IFNγ secretion was observed in both mutant-peptide– and wild-type-peptide–pulsed C1R-A02 cells when we used GINS1_I87V TCR-engineered T cells. Furthermore, CD137 upregulation (Fig. 3B) was observed in RFC5_K160N and BRAP_R543C TCR-engineered T cells upon coculture with C1R-A02 cells loaded with graded amounts of the mutant peptide. On the other hand, CD137 upregulation was not observed when the engineered T cells were cocultured with C1R-A02 cells loaded with physiologically relevant amounts of the corresponding wild-type peptide. However, BRAP_R543C TCR-engineered T cells responded to 10⁻⁵ mol/L of the wild-type peptide. As expected, CD137 upregulation in GINS1_I87V TCR-engineered T cells was observed when C1R-A02 cells were pulsed with both wild-type and mutant peptides. We also quantified the levels of several cytokines (IFNγ, TNFα, IL2, and IL4) with an ELISA assay, and all results are comparable with the IFNγ ELISPOT and CD137 assays (Fig. 3C; Supplementary Fig. S1).

To further validate these TCR-engineered T cells, we explored peptide-dependent cytotoxic activity of T cells against C1R-A02 cells loaded with either a mutant or a corresponding wild-type peptide. RFC5_K160N TCR-engineered T cells revealed cytotoxic activity exclusively against mutant peptide–pulsed C1R-A02 cells (Fig. 4A). However, BRAP_R543C TCR-engineered T cells showed some levels of cytotoxicity against wild-type peptide–loaded C1R2-A02 cells when the ratios of engineered T cells/C1R-A02 cells were very high. As we expected from other functional analysis, GINS1_I87V TCR-engineered T cells also showed cytotoxicity to C1R-A02 cells pulsed with the mutant peptide as well as the wild-type peptide.

To verify that the mutant epitopes are endogenously processed and presented on the cell with MHC molecules, we transfected each of the plasmid constructs designed to express a part of RFC5 and BRAP proteins including the AA substitution into C1R-A02 cells. When RFC5_K160N and BRAP_R543C TCR-engineered T cells were mixed with these transfected C1R-A02 cells, we observed high levels of IFNγ secretion, indicating that the mutant epitope was expectedly processed and presented on the cell surface (Fig. 4B).

Finally, we examined HLA-restricted activity of three TCR-engineered T cells. We used C1R-A24 cells that express HLA-A24 (HLA-A*24:02) as APCs and evaluated IFNγ secretion levels. As shown in Supplementary Fig. S2, IFNγ secretion was almost exclusively observed when these three TCR-engineered T cells were cocultured with C1R-A02 cells pulsed with mutant peptides, and not when these three TCR-engineered T cells were cocultured with C1R-A24 cells pulsed with mutant peptides or with both cell lines without peptides, clearly indicating that the mutant peptides were recognized by TCRs through the presentation on HLA-A02 molecules.

In this study, we validated our previously developed time-efficient protocol for identification of specific TCRs against neoantigens which were identified from whole-exome sequence results of 7 clinical ovarian cancers. We stimulated CD8⁺ cells derived from HLA-matched healthy donors with 14 candidate neoantigen peptides and obtained T cells reacting with the peptide–HLA complex for three neoantigen peptides. We sequenced TCRs of these sorted cells and identified possible TCRα and β pairs. We generated three neoantigen-specific TCR-engineered T cells and confirmed their reactivity to the respective neoantigens, although one of them showed cross-reactivity to the wild-type peptide.

Strønen and colleagues initially reported the successful application of healthy donor–derived T cells for targeting tumor-specific neoantigens (34). They selected 57 neoantigen candidates from 3 melanoma patients and succeeded in inducing neoantigen-specific T cells for 11 of the 57 peptides. They concluded that T cells obtained from healthy donors have a broad T-cell repertoire, which could increase a chance of inducing neoantigen-specific T cells. The use of HLA-matched healthy donor T cells may be effective because advanced cancer patients often suffer from myelosuppression caused by multiple regimens of chemotherapy and have a smaller T-cell diversity than healthy donors. Moreover, we intended to shorten the process of isolating neoantigen-specific T cells; this is critically important because advanced cancer patients are usually unable to wait a long period to have access to new treatments. Our protocol requires only 2 weeks from the priming of T cells with neoantigens to the identification of neoantigen-specific TCRs (16). We previously reported that our next-generation sequencer-based TCR repertoire analysis is able to accurately identify V-(D)-J combinations including CDR3 sequences in various types of tumor samples (26, 29, 30, 35–41). In this study, our TCR repertoire analysis facilitated the rapid identification of neoantigen-specific TCRs and allowed cloning of TCRαβ pairs and subsequent production of TCR-engineered T cells.

Most of the neoantigen-targeting studies have been performed using melanoma samples, because melanoma has the highest somatic mutation burden among human cancers, and hundreds of neoantigen candidates can be identified for each patient (34, 42–46). Only few studies have been performed in cancer types with a relatively lower number of somatic mutations, although a large body of evidence has suggested the importance of neoantigens. We selected ovarian cancer as a pilot study to validate our protocol because ovarian cancer is one of the common female cancers and harbors a number of somatic mutations that is average (1 mutation/Mb compared with 12 mutations/Mb for melanoma) among various cancer types (1, 45). In addition, the presence of neoantigen-reactive T-cell subtypes in TILs from ovarian cancer patients has been reported (20). In this study, we successfully identified 3 neoantigen-specific TCRαβ pairs from 14 predicted neoantigen candidates. Although our sample size was small, the success rate of obtaining neoantigen-specific T cells was similar to that in previous report for melanoma (34).

To exclude the risk of severe autoimmune adverse events, the neoantigen-specific TCR should recognize the mutant peptide (neoantigen peptide) exclusively, and not the corresponding wild-type epitope peptide. Our findings of cross-reactivity in GINS1_I87V TCR-engineered T cells as well as BRAP_R543C TCR-engineered T cells indicate the importance of careful validation for the specificity of TCRs against neoantigens. The neoantigens can be distinguished from corresponding wild-type peptides by TCR either when the mutation changes the motif (structure) recognized by TCR or when the mutation changes an AA at the HLA anchor position and significantly enhances the affinity to HLA molecules (47). The anchor positions for HLA-A2 (HLA-A*02:01) are usually at the position-2 (P2) and P9. Both the wild-type and mutant GINS1 peptides (I87V is located at the P9, Table 2) are predicted to have similar and high affinity to the HLA molecule. Probably, these two peptides bound equally to HLA-A2 and did not influence the recognition by TCR. We have previously observed cross-reactivity to wild and mutant peptides in a case of a different AA substitution in the anchor position by another TCR-engineered T-cell clone (the affinity to HLA-A was also similar between mutant and wild-type peptides; ref. 16). On the other hand, AA substitutions in RFC5 and BRAP proteins occurred in P4 and P3 positions of the epitope peptides, respectively. Because these P3 and P4 positions are critical positions for recognition by TCR, RFC5_K160N- and BRAP_R543C-specific TCR-engineered T cells revealed recognition specifically to the mutated peptides, but not to the wild-type peptides. These results suggest we should consider the position of the amino substitution in neoantigen peptides. When the AA change occurs at the anchor position, and the affinity of the wild-type and mutant peptides to HLA molecules is predicted to be similar, these neoantigen peptides might not be a feasible target. Furthermore, we have to be cautious about TCR-engineered T cells like BRAP_R543C TCR-engineered T cells. Under the very high peptide concentration (probably nonphysiologic condition) of the wild-type peptide, we observed CD137 upregulation in BRAP_R543C TCR-engineered T cells. Because the affinities of mutant- and wild-type peptides to HLA-A2 were predicted to be 5 and 25 nmol/L, respectively, it is very likely that the wild-type peptide still binds very effectively. It is also possible that BRAP_R543C TCR may have a weak affinity for the wild-type peptide sequence. Hence, under the conditions of a higher ratio of TCR-engineered T cells/APCs with high concentration of the wild-type peptide (Fig. 4A), BRAP_R543C TCR-engineered T cells could show the some level of cytotoxicity to C1R-A02 cells with the wild-type peptide. Although the high level of antigen presentation and the high ratio of effector/target cells used in this study are not physiologic, we have to be extremely careful to the safety of the TCR-engineered T cells in a clinical setting.

If somatic mutations are observed in a part of tumor tissues, a subset of tumors may not be targeted by TCR-engineered T cells. Hence, it would be ideal to identify and target neoantigens that are commonly presented in all the tumor cells. The future clinical application of this approach might need to collect multiple biopsies from a patient's tumor(s) to identify such trunk mutations. However, we think that it is practically very difficult to obtain the multiple portions of an original tumor and metastatic regions from advanced cancer patients because of the risk of complications during the biopsy processes.

Although we focused MHC Class I–restricted neoantigens in this study, our strategy can be applied to MHC Class II–restricted neoantigens. The binding affinities of each peptide to MHC Class II molecules can be examined (48–50). Recently, the adoptive CD4⁺ T-cell therapy using an MHC Class II–restricted TCR that recognized MAGE-A3 (cancer germline antigen) was investigated in a clinical trial (51). We would also like to identify MHC Class II–restricted neoantigen-specific TCRs and apply them to adoptive TCR-engineered CD4⁺ T-cell therapy.

In conclusion, for the data using 7 ovarian cancer cases, we successfully induced 3 neoantigen-specific T cells and identified their TCR sequences within 2 weeks (from peptide stimulation to TCR identification). This pilot study supports the feasibility of our time-efficient protocol for the use in solid cancers with relatively lower mutational loads such as ovarian cancer. In addition, we suggest some considerations for the selection of neoantigen candidates as well as the specificity of TCR against the mutant peptide versus the wild-type peptide.

J.-H. Park is an employee of and a consultant/advisory board member for OncoTherapy Science, Inc. M. Harada is an employee of OncoTherapy Science, Inc. Y. Nakamura reports receiving commercial research grants from and is a consultant/advisory board member for OncoTherapy Science, Inc. No potential conflicts of interest were disclosed by the other authors.

Conception and design: T. Matsuda, Y. Nakamura

Development of methodology: T. Matsuda, L. Ren, Y. Ikeda, Y. Nakamura

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T. Matsuda, M. Harada, E. Lengyel, G.F. Fleming

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T. Matsuda, M. Leisegang, J.-H. Park, Y. Ikeda, K. Kiyotani

Writing, review, and/or revision of the manuscript: T. Matsuda, M. Leisegang, J.-H. Park, T. Kato, K. Kiyotani, E. Lengyel, G.F. Fleming, Y. Nakamura

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): M. Leisegang, L. Ren, T. Kato, Y. Ikeda, Y. Nakamura

Study supervision: M. Leisegang, Y. Nakamura

The authors appreciate Drs. Rui Yamaguchi, Seiya Imoto, and Satoru Miyano at the University of Tokyo for developing the algorithm of TCR repertoire analysis and helpful support in data management. The supercomputing resource (http://sc.hgc.jp/shirokane.html) was provided by Human Genome Center, the Institute of Medical Science, the University of Tokyo. The authors also thank Kimberley Borutta for excellent technical support.

This work was supported in part by a Team Science Award of UCCCC (The University of Chicago Medicine Comprehensive Cancer Center), Einstein Foundation Berlin, and a research grant from OncoTherapy Science, Inc.

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.