Reprogramming of Tumor-reactive Tumor-infiltrating Lymphocytes to Human-induced Pluripotent Stem Cells

Tumor-infiltrating lymphocytes (TIL) that can recognize and kill tumor cells have curative potential in subsets of patients treated with adoptive cell transfer (ACT). However, lack of TIL therapeutic efficacy in many patients may be due in large part to a paucity of tumor-reactive T cells in TIL and the exhausted and terminally differentiated status of those tumor-reactive T cells. We sought to reprogram exhausted TIL that possess T-cell receptors (TCR) specific for tumor antigens into induced pluripotent stem cells (iPSC) to rejuvenate them for more potent ACT. We first attempted to reprogram tumor neoantigen-specific TIL by αCD3 Ab prestimulation which resulted in failure of establishing tumor-reactive TIL-iPSCs, instead, T cell–derived iPSCs from bystander T cells were established. To selectively activate and enrich tumor-reactive T cells from the heterogenous TIL population, CD8+ PD-1+ 4-1BB+ TIL population were isolated after coculture with autologous tumor cells, followed by direct reprogramming into iPSCs. TCR sequencing analysis of the resulting iPSC clones revealed that reprogrammed TIL-iPSCs encoded TCRs that were identical to the pre-identified tumor-reactive TCRs found in minimally cultured TIL. Moreover, reprogrammed TIL-iPSCs contained rare tumor antigen-specific TCRs, which were not detectable by TCR sequencing of the starting cell population. Thus, reprogramming of PD-1+ 4-1BB+ TIL after coculture with autologous tumor cells selectively generates tumor antigen-specific TIL-iPSCs, and is a distinctive method to enrich and identify tumor antigen-specific TCRs of low frequency from TIL. Significance: Reprogramming of TIL into iPSC holds great promise for the future treatment of cancer due to their rejuvenated nature and the retention of tumor-specific TCRs. One limitation is the lack of selective and efficient methods for reprogramming tumor-specific T cells from polyclonal TIL. Here we addressed this limitation and present a method to efficiently reprogram TIL into iPSC colonies carrying diverse tumor antigen reactive TCR recombination.


Introduction
Tumor-infiltrating lymphocytes (TIL) harvested from patient tumors can be expanded and infused into patients, resulting in the recognition and elimination tumor-reactive T cells (18). Unfortunately, TIL in most human cancers is in an exhausted state and has limited potential to expand (12).
Given that the process of reprogramming exhausted T cells to induced pluripotent stem cell (iPSC) and redifferentiating them to T cells holds the promise of reversing T-cell exhaustion and differentiation, the utilization of this technology has been proposed as an alternative to overcome current ACT limitations. The fact that a single T cell forms a single clone of T cell-derived iPSC (T-iPSC) without altering the recombined structure of TCR genes, while gaining unlimited proliferation capability, makes this method conceptually ideal for cloning TCRs. However, it has been challenging to establish T-iPSCs with multiple tumor neoantigen-reactive TCRs directly from polyclonal TIL containing neoantigen-specific T cells along with numerous irrelevant bystander T cells (19,20). Given that the generation of tumor antigen-specific iPSC colonies is technically difficult, most tumor-specific T-iPSCs have been established from highly clonal T-cell populations that were reactive to common tumor antigens (21,22). Therefore, an additional step to selectively reprogram T-iPSC lines with desired TCRs from heterogeneous TIL is needed.
Here we present an optimal method to selectively reprogram tumor antigenspecific T cells from heterogeneous TIL populations by coculturing with autologous tumor cells and sorting the PD1 + 4-1BB + CD8 + T-cell population before reprogramming. This distinctive method successfully identified tumorreactive TCRs including low-frequency ones that were not detected with other conventional methods.

Study Approval
All experiments were conducted with the approval of the NIH Clinical Center and NCI Institutional Review Board and performed in accordance with NIH guidelines. All patients whose samples were studied in this article were enrolled into the Surgery Branch selected TIL protocol (03-C-0277) and if living, signed an informed consent form and received a patient information form before participation. All studies were conducted in accordance with The Declaration of Helsinki, The Belmont Report, and the U.S. Common Rule. All mouse experiments were approved under institutional animal study protocol by the Animal Care and Use Committee of the NCI.

Subjects, TIL, and Autologous Tumor Lines
Frozen TIL samples were obtained from the Surgery Branch TIL lab repository. Patient information including age and sex is shown in Supplementary Fig. S1A. TILs were generated as described previously (7,12,13). Briefly, surgically resected tumors were cut into approximately 1-2 mm fragments and cultured in complete media containing high-dose IL2 (6,000 IU/mL). TIL fragment cultures from patients 1913 and 3784 were frozen after a short culture (days 13-16).
TIL fragments from patient 4069 were further screened for neoantigen reactivity by TMG screening (7) and reactive TILs were expanded in the presence of irradiated feeder cells, 50 ng OKT-3 and 3,000 IU IL2 in 50-50 media (RPMI-AIM-V with 5% human AB serum with pen strep and l-glutamine) to reach approximately 100-150 billion cells for infusion and leftover cells were frozen.
All patients had undergone prior therapies including surgery, chemotherapy, and immunotherapy.
Matched patients' melanoma cell lines were established from enzymatically digested tumor specimens and kept in culture in RPMI1640 medium supplemented with 10% FBS (Gibco) at 37°C and in 5% CO 2 . Melanoma tumor cell lines were authenticated based on the identification of patient-specific somatic mutations and HLA molecules (12,13). Cell lines confirmed negative for Mycoplasma tested frequently by using Cambrex MycoAlert Mycoplasma detection assay (Promega) according to manufacturer's instruction.

In Vitro Activation of T Cells
Healthy donor PBMC or TILs were stimulated in vitro with plate-bound αCD3 Ab (100 ng/mL) or αCD3/28 dynabeads (Gibco) in a ratio of cells to beads 1:1 (according to manufacturer's instructions) in complete media (RPMI-1640+10% human serum) in the presence of 300 IU recombinant human IL2 for 4 days.

TIL and Tumor Cell Coculture
TIL with minimal in vitro culture (13-16 days) were thawed in complete media in the absence of cytokines and rested overnight at 37°C and in 5% CO 2 .

Flow Cytometry and Cell Sorting
Fluorescently labeled antibodies used are shown below in Table 1. Cells were stained with propium iodide (PI) or fixable live/dead stain (Invitrogen) to and washed twice before sorting acquisition. The samples were analyzed by BD Fortessa and sorted by BD Aria II. Flow cytometry data were analyzed using FlowJo (RRID: SCR_008520). Data were gated on live cells (PI negative) and single cells. Gates were set based on fluorescence minus one control. The purity of sorted cells was usually more than 99%. The list of antibodies used is shown in Table 1.

Reprogramming of Peripheral Blood T Cells to T-iPSCs
Human whole T cells from PBMC or sorted populations were subjected for reprogramming to T-iPSCs. T cells were stimulated as described above by αCD3 Ab or αCD3/28 dynabeads. Cells were transduced with CytoTune-iPS 2.0 Sendai Reprogramming Kits (Thermo fisher scientific) carrying Yamanaka factors (Oct4, Sox2, Klf4, and c-Myc) and SV40 (large T antigen) in a 24-well plate for 24 hours in complete media without cytokines. The next day, cells were washed with fresh media, spun down, replaced with T-cell media with IL2, and transferred onto laminin coated (iMatrix 511, reprocell catalog no. NP892-012) dishes for attachment. On the following day, media was changed to human embryonic stem cell media (StemFit) containing 50 ng/mL of bFGF (R&D) and continue to culture until embryonic stem (ES) cell-like colonies started to appear. Fresh media was added everyday with bFGF, and iPSC colonies were stained with alkaline phosphatase (AP) and counted typically at days 20-25.
When ES cell-like colonies appeared around days 20-25, undifferentiated iPSC colonies were chosen manually using the microscope based on their morphology and individual iPSC colonies were transferred onto matrigel (BD) coated 6-well plates and allowed to expand in StemFit media with bFGF (50 ng/mL). Rock inhibitor (

iPSC Colony Count and Measurement
Bright field images of colonies grown in 6-well plates were collected using a Zeiss AxioObserver Z1 microscope equipped with a 10x plan-apochromat (numerical aperture 0.45) objective lens, a condenser lens with 0.55 N.A., motorized stage and Hamamatsu ORCA Flash4 v2 sCMOS camera. Tile imaging mode in the Zen software was used to collect multiple images covering 80% of the good area, for each well in the 6-well plate. The tile images were stitched using the Zen software and the resultant image was analyzed using the image analysis module of the Zen software. The images were background subtracted, contrast inverted, and an intensity threshold applied to segment the cell colonies from the background. The area of the individual colonies was measured.

TCRα and TCRβ Analysis
For bulk TIL, sorted T cells and iPSC (about 1E+5 cells per group) were spun down, washed with PBS once and snap frozen in 50-100 μL of 1 mol/L HEPES buffer (GIBCO#15630-080).
For TIL-iPSCs, cells were trypsinized, dissociated into single-cell suspension and counted in an automated cell counter machine (Countess). Approximately 1E+5 cells from 15-20 individual iPSC lines were mixed and pooled together into a 15 mL tube (master tube), spun down and washed once with PBS and snap frozen in 100 μL of 1 mol/L HEPES buffer. All the remaining colonies were trypsinized, collected, and frozen down (mother dish). Genomic DNA extraction and Immunoseq TCRβ survey sequencing were performed by Adaptive Biotechnologies. The results were analyzed using IMMUNOSEQ ANALYZER. For pooled samples the TCRs that are more than 0.5% of productive TCRs were considered to be valid. To identify the TCRα-TCRβ paired sequence, genomic DNA was extracted from individual TIL-iPSC clones and analyzed by Adaptive Biotechnologies.

Construction of Candidate TCRs
To test the reactivity of candidate TCRs, retroviral vector constructs were synthesized and transduced into healthy donor T cells using the method described previously (7,23). Medium was replaced 8 hours after transfection and cells were incubated for further 48 hours in complete media. To capture the viral particles, retroviral supernatants were spun at 2,000 × g for 2 hours at 32°C in 6-well non-tissue culture-treated plates coated with Retronectin (Takara Bio). Healthy donor peripheral blood lymphocytes were used as donor T cells for transduction. T cells were activated using complete media + 50 ng/mL OKT3 (Miltenyi Biotec) of media for 48 hours and 2 million cells were added per well of virus-coated 6well plate, spun for 10 minutes at 300 × g at 32°C, then incubated overnight at 37°C. Surface murine TCRβ constant region + cells were transduced with the cloned TCR pair successfully.

Generation of Immature DCs as Antigen-presenting Cells
Monocyte-derived, immature DCs were generated on tissue culture tissue culture flasks. Briefly, apheresis samples were thawed, washed, counted approximately 5-10E+6 cells/mL with AIM-V media (Life Technologies) and then incubated at approximately 1E+6 cells/cm 2 in an appropriately sized tissue culture flask and incubated at 37°C, 5% CO 2 . After 90 minutes, nonadherent cells were vigorously washed with AIM-V media and collected, and then incubated with AIM-V media for another 60 minutes. The flasks were then vigorously washed again with AIM-V media and then the adherent cells were incubated with DC media. DC media comprised of RPMI-1640 containing 5% human serum, 100 U/mL penicillin, and 100 μg/mL streptomycin, 2 mmol/L l-glutamine, 800 IU/mL GMCSF (Leukine), and 200 U/mL IL4 (Peprotech). On days 2-3, fresh DC media was added to the cultures with 200 U/mL of GMCSF and 200 U/mL IL4. DCs were used on days 5-6 after starting the culture.

Peptide Pulsing
On day 5, DCs were harvested and then resuspended at 5E+5 cells/mL 200 U/mL of GMCSF and 200 U/mL IL4. Peptides were dissolved in DMSO and pulsed onto the antigen-presenting cells (APC) at 10 μg/mL and incubated for 4 hours at 37°C with 5% CO 2 . After 4 hours of peptide pulsing, APCs were washed twice with PBS prior to setup the coculture with T cells.

Patient-derived Xenograft
Fresh tumor specimens from patients were chopped into small fragments of 2 mm in dimension. One fragment was implanted subcutaneously at the flank of an NSG mouse using a 20-gauge needle. Tumor growth was monitored weekly.

Coculture of PDX-derived Tumor Cells with Neoantigen-specific TCR
One day prior to coculture, tumor cells were seeded into a 96-well plate in concentration of 1E+5 cells in 100 μL of culture media per well. Next day, 1E+5 T cells in 100 μL of RPMI-1640 +10% human AB serum per well were added to the tumor cells. The plate was incubated at 37°C, in 5% CO 2 for 16 hours.
Supernatants from the coculture wells were collected for IFNγ ELISA assay and nonadherent cells were harvested for surface 4-1BB expression by FACS.

Determination of Proliferation and Apoptosis by Live Cell Imaging
Incucyte, live cell analysis system (Sartorius, Essen Bioscience), placed inside a conventional cell culture incubator at 37°C in 5% CO 2 , was used for real-time imaging of RFP-expressing tumor cells from each patient's tumors (TC-4069, TC-1913, and TC-3784). A total of 5,000 tumor cells were seeded in 100 μL of complete culture medium in the 96-well plate and rested overnight in a CO 2 incubator to settle down. Next day, mTCRβ + sorted TCR transduced (healthy donor PBL) T cells were counted, washed once with PBS and resuspended in complete medium and added at 100 μL per well on top of tumor cells at the effector-to-target ratio of 2:1 ratio. The plates were then placed into the Incucyte real-time cell imaging device and the red cell count per image was measured over time. Throughout the assay, both phase and fluorescent images were collected using phase contrast and red fluorescence channels with a 10 × object.
Images were taken every 3 hours for 48-72 hours, and each condition was run in quadruplicate. Images were analyzed using IncuCyte 2021B software and data were generated using the GraphPad Prism statistical software (GraphPad Software). The data were reported as mean ± SD and each experiment was performed at least twice from two different healthy donor PBL transduced with appropriate TCRs.

Statistical Analysis
All data were analyzed and presented as mean ± SD. Statistical analysis was performed using GraphPad Prism program 8.1.1 (RRID:SCR_002798 GraphPad Software Inc.). The replication information is described in each figure.

Data Availability
Genomic and TCR sequencing were performed by Adaptive Biotechnologies.
Derived data from these analyses are available from the corresponding author upon reasonable request. Raw data from the other assays presented in this study may also be requested from the corresponding author.

TCR Stimulation is Necessary for Reprogramming of T Cells to Generate T-iPSCs
To identify the optimal method of selective reprogramming of T cells with desired neoantigen-specific TCRs, we first optimized methods for reprogramming PB T cells. Given that cell-cycle progression is necessary for somatic cells to be reprogrammed to iPSCs (25,26) and TCR engagement induces T-cell proliferation, we examined the necessity of TCR stimulation for efficient reprogramming of T cells. PB T cells were subjected to reprogramming with or without TCR stimulation by anti CD3/28 Ab. The efficiency of reprogramming toward an iPSC state was determined by the number of colonies successfully stained by the pluripotent stem cell marker, AP ( Fig. 1A; ref. 27). Interestingly, T-iPSCs were established only when the T cells were stimulated (Fig. 1B). Moreover, although it is possible to generate a very limited number of colonies using only the four Yamanaka factors (c-myc, KLF4, Sox2, and OCT3/4), the addition of SV40 large T antigen greatly enhanced the efficiency of reprogramming (Fig. 1B). Therefore, TCR stimulation is essential for T-cell reprogramming.
Considering that the majority of TIL are differentiated cells and CD8 + T cells also contain other T-cell subsets (e.g., Naïve, CM, EM, and EMRA), we sought to elucidate whether TCR stimulation is necessary for the reprogramming of any T-cell subset (Fig. 1C). While there was a tendency that less-differentiated T cells have higher probability of reprogramming, all the subsets were able to be reprogrammed to T-iPSCs by prestimulation by αCD3/28 beads while only EMRA was reprogrammed by plate-bound OKT-3 (αCD3 Ab) stimulation ( Fig. 1D-F) suggesting that bulk TIL containing different subsets can be reprogrammed to TIL-iPSCs by optimal TCR stimulation by αCD3/28 beads.

Antigen Nonspecific TCR Stimulation of TIL Results in Reprogramming of Non-tumor Antigen-specific T Cells
Given that T-iPSC can be generated from any T-cell subgroup we next tested whether this method can be applied to reprogramming tumor neoantigenspecific T cells identified in a patient with pancreatic cancer (Pt. 4069). To increase the probability of generating tumor neoantigen-specific TIL-iPSC clones, an almost monoclonal T-cell population (frequency 92.7%) expressing the TCR (TCRBV11-02*02) specific for the patient's neoantigen ZFYVE27 R6H ( Fig. 2A) were activated by αCD3 Ab and reprogrammed as reported previously (21  T-cell clone (TCRBV04-02*01) which was a very minor population and not detected by TCR deep sequencing in the original TIL products (Fig. 2B).
To establish neoantigen-specific TIL-iPSCs another TCR stimulation method, αCD3/28 beads stimulation was used before reprogramming. To acquire as many TCRs as possible from TIL-iPSCs, a total of 96 clones were isolated (Supplementary Fig. S1A). About 10-15 iPSC clones were pooled and mixed into one master tube (seven in total), total genomic DNA was extracted from each master tube, and TCRs were identified by TCRβ survey sequencing ( Supplementary  Fig. S1B). This modification of the stimulation method allowed TIL-iPSCs with preidentified neoantigen-specific TCR (TCRBV11-02*02) to be established along with 26 other TCRs ( Supplementary Fig. S2A). We investigated whether the newly found TCRs (NF TCR) are specific to the patient's tumor or neoantigens. For that purpose, we chose one of the master tubes (master tube 6) having 14 different iPSC clones and their TCR α and β sequence was examined individually. We found five new TCR pairs along with the preidentified reactive one (PIR-TCR) in this tube ( Fig. 2C; Supplementary Fig. S2A). We subsequently cloned these TCRs (Fig. 2C) Supplementary  Fig. S2D). Furthermore, we have also investigated whether this PIR-TCR can inhibit PDX-derived tumor cells proliferation by Incucyte live-cell analysis system (Sartorius, Essen Bioscience). Consistent with previous findings, PIR-TCR can suppress tumor growth (Fig. 2H).
Taken together, these data demonstrated that nonspecific stimulation by CD3 cross-linking is not always successful in specifically generating tumor-reactive TIL-iPSCs especially for those clones with low frequency, suggesting that a new method to selectively reprogram tumor-reactive T cells is needed.

Tumor-reactive T Cells were Reprogrammed to iPSCs by Stimulating TIL with Autologous Tumor Cells and Enriching Reactive Populations with PD-1 and 4-1BB Expression
Considering that T cells require TCR stimulation for reprogramming and tumor antigen-specific TILs are activated in the presence of autologous tumor cells, we sought to explore whether a coculture system could mediate selective reprogramming of tumor-reactive cells. To further enhance the probability of specific reprogramming of tumor-reactive TILs, we sorted on the basis of the activation markers PD-1 and 4-1BB following coculture ( Fig. 3A; refs. 12,23,28,29). For this study, we utilized melanoma TIL that had undergone minimal expansion in vitro, where autologous tumor cell lines were available and tumor-reactive TCRs were preidentified (13). TILs from patient 1913 were cocultured with their autologous tumor cell line for 16 hours and PD-1 + 4-1BB + CD8 + T cells were sorted and infected with Sendai virus containing four Yamanaka factors and SV40 large T antigen to facilitate the reprogramming ( Fig. 3A; Supplementary Fig. S3A). As a control, TILs stimulated with αCD3/28 beads were reprogrammed as well. After 3 weeks, typical ES cell-like colonies appeared and 221 colonies were manually collected from the dish prestimulated with autologous tumor cells for iPSC line establishment ( Supplementary   Fig. S1A). About 20 iPSC clones were pooled and mixed per master tube (13 in total), total genomic DNA was extracted from each master tube, and TCRs were identified by TCR β sequencing (Supplementary Fig. S1B). For control TIL-iPSCs generated by αCD3/28 beads stimulation, all the colonies were collected without cloning each individual colony and TCRs were identified by TCR sequencing. A total of nine different TCRs were detected from these 13 master tubes ( Fig. 3B; Supplementary Fig. S3B). Four of them were present in all samples and the remaining five TCRs were detected only in one of the master tubes ( Fig. 3B). Six tumor-specific TCRs had previously been identified from this patient and all six were detectable in the starting TIL and sorted PD-1 + 4-1BB + population in various frequencies (Fig. 3C), Therefore, tumor antigen specific T cells can be successfully reprogrammed by autologous tumor cell coculture. The frequency of the six preidentified (PIR) TCRs were enriched in the PD-1 + 4-1BB + population compared with the starting bulk population, showing that the enrichment of PD-1 + 4-1BB + populations before reprogramming is a feasible strategy to pre-enrich antigen-specific clones ( Fig. 3D and E). Three out of six PIR-TCR were present in relatively high frequency (>2%) in the PD-1 + 4-1BB + population and these TCRs were detected in all master tubes suggesting a high number (at least 13) of iPSC clones carrying each PIR-TCRs (Fig. 3B).
Another three PIR-TCRs were detected in PD-1 + 4-1BB + population in relatively low frequency (<2%) and were not detected in any of the master tubes. One of the TCR which was not present in established TIL-iPSCs (TCR-V) was detected in the remaining dish after picking up iPSC colonies, suggesting the selection of more colonies may have resulted in detection.
TIL-iPSCs established by αCD3/28 beads stimulation were almost clonal (Supplementary Fig. S3C) but this TCR did not appear in the list of tumor-reactive clones by previous studies (Fig. 3C; (Fig. 2C). T cells were cocultured with autologous DC pulsed with wild-type (WT) peptide (16 mer or 9 mer) or mutant peptide (9 mer). D shows the representative FACS plots and E shows the percentage of 4-1BB + cells in each condition. Representative data of three independent experiments from 3 different healthy donor T cells. Mock indicates empty vector transduced T cells used as negative control. F and G, CD137 (4-1BB) upregulation assay of T cells transduced with candidate TCRα and TCRβ pairs identified from TIL-iPSCs (Fig. 2C). T cells were cocultured with autologous, or HLA-matched allogeneic tumor cells derived from the PDXs model. TCR-4 of Fig. 3B, which was detected in all 13 master tubes, is reactive or not.
For that purpose, TCRα and TCRβ sequences of those T-iPSC clones were identified (Fig. 4A) by immunosequencing. TCRα and TCRβ pairs were cloned into a gamma retrovirus vector (pMSGV1) and transduced into healthy donor PB T cells and tested for specific recognition of autologous tumor cells. While T cells transduced with the TCR from beads stimulated T-iPSC did not express 4-1BB nor produced IFNγ, those with the TCR from tumor cell-stimulated T-iPSC expressed 4-1BB and produced IFNγ as PIR-TCR ( Fig. 4B-D). Moreover, the T cells transduced with the TCR from TC-stimulated T-iPSC (Fig. 3B, TCR-4) showed cytotoxicity against autologous tumor cells similarly to PIR TCR-III, but those with the TCR from beads stimulated T-iPSC did not (Fig. 4E). These results demonstrated that TIL-tumor cell coculture before reprogramming is a better method to reprogram tumor antigen-specific TIL than nonspecific T-cell stimulation by αCD3/28 beads.

T Cells with Tumor Antigen-specific TCRs of Extremely Low Frequency were Reprogrammed to TIL-iPSCs
To confirm the TIL-tumor cell coculture can selectively reprogram tumorreactive TIL, another melanoma TIL from patient 3784 was similarly cocultured with an autologous tumor cell line for 16 hours. PD-1 + 4-1BB + CD8 + T cells were then sorted and reprogrammed into iPSCs (Fig. 5A). A total of 178 iPSC lines were established ( Supplementary Fig. S1A) from this patient and TCRβ sequences were identified as described previously (Supplementary Fig. S1B). Of them, we found the preidentified tumor antigen-specific iPSC clone (clone 159) among the 178 colonies that were picked up and established. Furthermore, to demonstrate a successful reprogramming, we have examined clone 159 pluripotency by immunofluorescent staining of key genes associated with pluripotency ( Supplementary Fig. S4A), chromosomal karyotyping for genomic stability (Supplementary Fig. S4B) and their ability to differentiate into three embryonic germ layers ( Supplementary Fig. S4C). Together, these data confirm that antigen specific TIL iPSC attain true pluripotency.
This patient is a complete responder for TIL therapy and eight tumor-reactive TCRs were preidentified in this patient's sample using various methods (12,13). Although most of the PIR-TCR were found at relatively low frequency in the starting bulk TIL and PD-1 + 4-1BB + populations (Fig. 5B), one major TCR, which was enriched from 7.18% to 10.3% (1.43 times) in the PD-1 + 4-1BB + compared with bulk population, was detected in two iPSC master tubes out of nine tubes (Fig. 5B). Even though only one PIR-TCR was reprogrammed (Fig. 5B, PIR-TCR IV), we identified a total of 25 different TCRβ chains from nine master tubes (Supplementary Table S1). Interestingly, most of the TCRs identified in TIL-iPSC clones were undetectable or at very low frequency in the starting TIL or PD-1 + 4-1BB + population. Some were highly enriched in the PD-1 + 4-1BB + population compared with the bulk population (Fig. 5C).
Because our previous experiments demonstrate the ability to generate several different tumor antigen-specific TCRs from a bulk TIL population by a coculture system before reprogramming, we sought to determine whether the unknown minor T-cell clones were preferentially reprogrammed to TIL-iPSCs because they were tumor antigen specific. To test this hypothesis, we sequenced TCRα and TCRβ chains from several TIL-iPSC lines and identified the TCRα and TCRβ chains (Fig. 6A). Three TCR pairs from T-iPSC clones whose TCRβ were not detected in the PD-1 + 4-1BB + population, one TCR pair from a clone whose TCRβ was very low frequency (0.16%; Supplementary Table S1), and the PIR-TCR IV (Fig. 5B) as positive control were tested for specificity against autologous tumor cells. TCRα and TCRβ pairs were cloned into a gamma retrovirus vector (pMSGV1) and transduced into healthy donor PB T cells and tested for specific recognition of autologous tumor cells. Three out of four newly tested TCR pairs (NF TCR-1, 2, and 4) showed reactivity against autologous tumor cells as PIR-TCR IV (our positive control) by 4-1BB expression, cytokine production and target cell killing (Fig. 6B-E). These results demonstrate that TIL-tumor cell coculture before reprogramming can selectively enrich antigenspecific T cells and subsequently reprogram tumor-reactive T cells even from very minor clones.

Discussion
Although iPSC technology is a new avenue to generate an unlimited amount of undifferentiated and nonexhausted populations of de novo T cells, the development of a method capable of reprogramming polyclonal tumor antigen-specific TIL is still necessary (21,22,30).
Several reports have already been published demonstrating the importance of enriching tumor and neoantigen-specific T cells for successful cell therapies against cancers (9,31). However, it is always challenging to start with a polyclonal (bulk) population of T cells where mutation or antigen-specific TCR is unknown. Therefore, we started with some extensively studied patients where the mutation and their appropriate TCRs are known. Here we describe the first report of selectively establishing multiple tumor antigen specific TIL-iPSCs from patient TIL with minimal in vitro culture.    Table S1 for the details of the TCR clones (TCR 2, 3, 6, 9, 11, 12, 15, and 17).
TIL-iPSCs even from highly enriched populations of tumor-reactive TILs. Consequently, we employed a tumor mediated stimulation to selectively reprogram tumor-reactive T-cell clones. To further enhance specific reprogramming of tumor-reactive TIL, we enriched PD-1 + 4-1BB + cells after tumor cell coculture. This population was chosen due to previous demonstration that PD-1 + TIL contains neoantigen-specific TCRs (12,13) and that tumor neoantigen-specific TCRs were identified from 4-1BB + cells after TIL-TMG-expressing DC coculture (23). Most of the preidentified tumor antigen-specific T-cell clones were modestly enriched in the PD1 + 4-1BB + population compared with starting TIL (Fig. 3C-E and 5B), and some of them were reprogrammed into TIL-iPSCs. Moreover, many TIL-iPSCs of unknown specificity were established from very minor clones, some of which were not detectable in starting TIL or the sorted PD1 + 4-1BB + population by TCR sequencing. We confirmed the majority of tested TCRs were reactive and have cytotoxicity against autologous tumor cells.
These data show that not only does the frequency of the reactive T-cell clones play a role, but also some other factors of T cells such as differentiation status, Recently another group reported establishment of TIL-iPSCs from TILs sorted by the expression of CD107a or 4-1BB after coculturing with autologous tumor organoids (32). However, their method is to first establish reactive TIL lines by nonspecific stimulation by αCD3 Ab before reprogramming to TIL-iPSCs, which we report here as not optimal (Pt. 4069, Fig. 2). The key points for selective reprogramming of tumor-reactive T cells are transducing reprogramming factors to TILs directly after TIL-tumor cell coculture and avoiding nonspecific TCR stimulation.
The frequency in the starting TIL population has been the most important factor for successful cloning of tumor antigen-specific T cells and their TCR genes. Even using novel technology to identify TCRα and TCRβ chains from a single cell, the detection limit would be one in a thousand cells, due to the throughput capacity (33)(34)(35)(36). Our method is distinctive in that it has the potential to detect extremely minor tumor-reactive T cells from bulk TIL. Those TILs are specifically activated and reprogrammed into iPSC when they come in contact with their physiologic neoantigens. However, nonspecific activation by αCD3 Ab or bystander stimulation by other stimuli such as viral antigens does not selectively generate antigen-specific T cells.
The caveat of this method is the availability of autologous tumor cells. As an alternative, it may be possible to use novel technology like tumor organoids or autologous DC transduced with TMGs which contain the patient's tumor specific mutations as minigenes. Another caveat of not only our strategy but the whole field of differentiating T cells from iPSCs is the lack of a robust method to generate less differentiated/naïve-like T cells from human iPSCs, which is an area of research under focus by a number of groups (22,(37)(38)(39). Given that the potency of the differentiated T cells from human iPSCs still be controversial, in this study we focused on the method to selectively reprogram tumor-reactive T cells in TIL to TIL-iPSCs. To further demonstrate TCR antitumor specificity and function, we use healthy donor's PB T cells to transduce and evaluate the cloned TCR from TIL-iPSCs. If TIL-iPSCs with tumor-reactive TCRs established by this method are induced to generate less-differentiated naïve-like T cells, these rejuvenated T cells will have polyclonal tumor-reactive TCR population, high expansion capacity, and potential to persist longer in vivo, which may revolutionize the current adoptive cell-based therapies against cancers (40)