Tumor sample from a 63-year-old woman with metastasis ESCC from Peking University Cancer Hospital was obtained in our preclinical research with informed consent. The tumor sample was collected in tubes containing sterile saline solution and resected into several fragments for a) TIL culture; b) generation of patient-derived xenograft (PDX) model. This study was approved by the Institutional Review Board of the Peking University School of Oncology, China.

HEK 293-FT (Life Technologies), a packaging cell line used to produce high-titer lentivirus supernatants, were cultured in complete Dulbecco’s Modified Eagle Medium (DMEM, Gibco, USA) supplemented with 10% fetal bovine serum(FBS, Gibco, USA) containing 0.1 mM MEM Non-Essential Amino Acids, 1 mM sodium pyruvate and 2 mM Glutamax (Life Technologies, USA) at 37 °C with 5% CO2.

Flow cytometry staining antibodies were shown as below: CD3-BUV395(Clone: UCHT1), CD4-PE-CF594(Clone: RPA-T4), CD8-PE-Cy7(Clone: RPA-T8), PD-1-BUV737(Clone: EH12.1), CCR7-PE(Clone: 150503), CD45RA-APC(Clone: HI100), CD137-APC(Clone: 4B4–1), CD107A-AF647(Clone: H4A3), CD107b-AF647(Clone: H4B3), Fixable Viability Stain 780 (FVS780). All antibodies were from BD Biosciences, except for anti-mouse TCR-β constant region (clone H5–597, eBioscience, USA). The blocking antibody against HLA class I were from eBioscience (clone: W6/32).

DNA of autologous tumor cells or the patient or donors’ peripheral blood were exstracted with DNeasy Blood & Tissue Kit (QIAGEN, Germany) according to the manufacturer’s protocol. Genotyping of HLA alleles was performed using high-resolution, high-throughput HLA genotyping with deep sequencing (BGI Diagnosis, Shenzhen, China). The HLA types of the patient and donors were outlined in Additional file 2: Table S1.

TILs were generated as previous described [14, 15] with slightly modification. Briefly, tumor sample was minced into approximately 1–2 mm fragments and each fragment was placed into a well of a 24-well plate comprised of T cell media and 50 ng/mL OKT3 antibody (ACRO, USA). T cell media consisted of X-VIVO 15 (Lonza, USA); Glutamax (Life Technologies, USA); IL-2 (50 U/mL, Perprotech, USA), IL-7 (10 ng/mL, Perprotech, USA); IL-15(10 ng/mL, Perprotech, USA). T cells were incubated at 37 °C with 5% CO_2, and passaged to maintain a density of 1 × 10⁶ cells/mL until there were enough TILs used for screening of tumor-specific T cells.

PDX model was established by implanting tumor fragments mixed with matrigel subcutaneously into immunodeficient NOD-SCID mice to generate PDX model. ESCC PDX models could be promising for individual therapy, since we have verified that the molecular characteristics of ESCC PDXs were in accordance with primary patient tumors in our previous study [16].

Autologous tumor cells were generated from tumor specimens based on successfully established PDX model [17]. Dissected tumor samples were dissociated into single-cell suspension, using the human Tumor Dissociation Kit (Miltenyi Biotech, Germany) with gentleMACS Octo Dissociator with Heaters (Miltenyi Biotech, Germany) according to the manufacturer’s instructions. Single-cell suspensions were harvested, washed with 1 × phosphate-buffered saline (PBS), and then resuspended in 6-well plate with complete media (DMEM, 20% FBS, 2 mM Glutamax) at 37 °C with 5% CO₂.

Single-cell pellets of TILs were stained with CD3-BUV395, CD4-PE-CF594, CD8-PE-Cy7, CCR7-PE, CD45RA-APC, PD-1-BUV737 antibody cocktails. Cells were washed with PBS prior to acquisition on BD FACS Aria III flow cytometer. Data files were analyzed using FlowJo vX software (FlowJo, Tree Star), gated on single cells. Dead cells and debris were excluded by staining with Fixable Viability Stain 780 (FVS780). Phenotypes of T cells were gated on total CD3⁺ T cells.

Autologous tumor cells were pretreated with complete media containing DAC cocktails (DAC (10 μM, Sigma Aldrich, USA), IFNγ (100 U/mL, ACRO, USA) and TNF-α(10 ng/mL, ACRO,USA) for 48 h, which restored cell surface expression of HLA molecules through enhancing the mRNA expression of HLA-related molecular including TAP or LMP genes or inhibiting DNA methylation [18–20]. Both IFN-γ enzyme-linked immunospot (ELISPOT) and enzyme-linked immunosorbent assays (ELISA) were used to screen tumor-reactive TILs cocultured with pretreated ATCs.

Human IFN-γ ELISPOT Kit (with precoated plates, Abcam, USA) was performed as the protocol’s procedure. Briefly, 2 × 10⁴ T cells, rested in cytokines-free media overnight, and 1 × 10⁴ PBS-washed autologous tumor cells were incubated together for approximately 20 h in the absence of exogenous cytokines at 37 °C with 5% CO₂. The number of colored spots was determined by ImmunoSpot plate reader and associated software (Celluar Technologies, USA).

1 × 10⁶ responder cells (T cells) and 1 × 10⁵ target cells (ATCs) were incubated together in a 96-well plate in the absence of exogenous cytokines for 18-24 h. Coculture supernatant was transferred into a new 96-well plate, and IFN-γ concentration was measured by using commercially available human IFN-γ ELISA kit (ExCell Bio, China) as manufacturer’s protocols. In HLA blocking experiment, ATCs were pretreated with HLA-blocking antibody (clone W6/32, 50 μg/mL). Following a 3.5-h incubation, 5 × 10⁴ target cells (ATCs) and 1 × 10⁵ responder cells (T cells) were cocultured overnight for assessment of IFN-γ releasing level with the standard IFN-γ ELISA procedure [21].

CFSE-based cytotoxity assay was performed as previously described [22, 23] with slightly alteration. Target cells were labeled with 5 μM CFSE (BD Biosciences) for 10 min and then cocultured with TCR-Ts at 37 °C for 4 h, at E: T ratio of 1:1 and 4:1. After the coculture, 1 μg/mL propidium iodide (PI, BD Biosciences) was added for assigning the ratio of cell death, and the samples were analyzed by flow cytometry.

For enrichment of tumor-reactive TILs, 2 × 10⁶ TIL-F1 and TIL-F4 were stimulated with 2 × 10⁵ autologous tumor cells pretreated with DAC cocktails in T-cell media for 1 week, respectively, after which they were stimulated with autologous tumor cells one more time. Moreover, both pre- and post-stimulated TILs were cocultured with 1 × 10⁵ autologous tumor cells (E: T = 5:1) to evaluate their ability of specifically identifying and killing autologous tumor cells using flow cytometry. Data were analyzed using FlowJo vX software (Treestar Inc) after gating on live cells (FVS780 negative). Meanwhile, CD8⁺CD137⁺ T cells of pre- and post-stimulated TIL-F1 were sorted into 96-well PCR plates by single-cell sorting (Additional file 1: Figure S3). And then 96-well PCR plates were immediately put into liquid nitrogen and conserved in minus 80°C prior to running single-cell RT-PCR.

For single-cell PCR, all the primer sequences were listed in Additional file 2: Table S2 as previously described [24] except that the primers of TRBC were optimized, the final concentration of each Vα and Vβ region primers was 5 μM, and final concentration of TRAC and TRBC primers was 20 μM. The single-cell RT-PCR reaction condition for first-step RT-PCR was as follows: 30 min at 50 °C for RT reaction; 95°Cfor 15 min and 30 cycles of 94 °C for 30s, 52 °C for 30s, 72°C for 1 min; 72°C for 10 min. For the second cycle, 2 μL of cDNA product was used as template for TCRα/β separately in total 20 μL 2nd-PCR TRA/TRB mix (Additional file 2: Table S3), containing multiple internal primer sequences (INT) of Vα/Vβ and one primer for Cα/Cβ, with PrimeSTAR® HS DNA Polymerase (Takara Bio, Japan). The cycling program was 98 °C for 1 min; 98 °C for 10s,52°C for 10s,72°C for 45 s × 43;72 °C for 10 min.The second PCR products were analyzed with TRA sequencing primer for TCRα or TRB sequencing primer for TCRβ (Additional file 2: Table S3). PCR products were purified and sequenced by Sanger sequencing method. The TCR sequences were analyzed using IMGT/V-Quest tool (http://www.imgt.org/).

TCRα/β chains were synthesized (GenScript) and cloned into the lentivirus vector. TCR was constructed in a β/α chain order and its constant regions were replaced by mouse counterparts modified with interchain disulfide bond and hydrophobic substitution as previously described, which not only was convenient for detection of TCR-T, but also improved TCR pairing and TCR/CD3 stability [17, 25]. However, since murine constant region of TCR-T could potentially be immunogenic in clinical application, human constant region could be essential for construction of TCR in order to improve the longevity of TCR-T persistence and enhance their therapeutic efficacy in patients.

Transduction of PBLs was conducted as previously delineated [17, 21, 26] with slightly modification. In brief, human peripheral blood mononuclear cells (PBMCs) were separated by centrifugation on a Ficoll-Paque Plus (GE Healthcare, USA) and then stimulated in T-cell media with 50 ng/ml OKT3 and 1 μg/ml anti-CD28 for 2 days before transduction. TCR lentivirus were generated by cotransfection of 293-FT cells with lentivector and packaging plasmids (ratio of pLP1: pLP2:pVSV-G is 2:2:1) using PEI MAX 40000 (Polysciences Inc. USA) [27]. The lentiviral supernatants were harvested at 48 and 72 h after transfection and concentrated using optimized ultracentrifugation approaches with 20,000 g, 90 min at 4 °C [28]. Activated T cells were transduced by concentrated lentivirus at the presence of 8 μg/mL polybrene (Sigma-Aldrich, USA). The transduction efficiency was assessed by flow cytometry using mouse TCR-β chain constant region staining.

NOD/SCID mice were used to establish patient-derived xenograft approved by the Institutional Review Board of the Peking University School of Oncology, China. Tumor treatment was on days 5 and 12 following tumor inoculation and consisted of two intravenous injections of 4 × 10⁶ T cells as well as a single intraperitoneal injection of 5 mg/kg DAC on day 5. Tumor size was determined by caliper measurement of perpendicular diameters of each tumor and was calculated using the following formula: tumor volume (mm³) = [(length) × (width) × (width)]/2.

Statistical analysis was performed using GraphPad Prism 7.0 (GraphPad Software, CA) and SPSS software (version 24; IBM SPSS, Armonk, NY, USA). Statistical comparison was conducted with Student’s t test, and two-way repeated measures ANOVA. All tests were two-sided and p value < 0.05 was considered statistically significant. All in vitro experiments were performed more than three independent experiments.

Tumor specimen was obtained from a 63-year-old woman with ESCC. Clinical characteristics and HLA types of the patient are outlined in Additional file 2: Table S1. In order to screen tumor-reactive TILs, we obtained four TIL fragments (TIL-F1 to TIL-F4) from different areas in one resected lesion.

To evaluate spatial heterogeneity of TILs, we measured the phenotypic characteristics of four TIL fragments derived from different anatomical sites of tumor sample by flow cytometry. The percentages of CD3⁺ T cells in all four TIL fragments were similar and approximately 99% (Additional file 1: Figure S1). However, percentages of CD4⁺ TILs hugely varied from 30.6 to 87.7%, and percentages of CD8⁺ TILs from 9.67 to 63.6%, suggesting significant difference in distribution of CD4⁺ and CD8⁺ TILs among different anatomical sites (Fig. 1a and b). The level of PD-1 expression hugely varied in four TIL fragments, with higher proportions in TIL-F1 and TIL-F2 (35.8 and 30.7%, respectively; Fig. 1c and d). The percent of effector-memory T cells (CCR7⁻CD45RA⁻) was highest in all four TIL fragments, followed by effector T cells (CCR7⁻CD45RA⁺), as showed in Fig. 1e and Additional file 1: Figure S2.Fig1Fig. 1

Phenotype and functional screening of different tumor infiltrating lymphocytes (TILs) fragments. a Flow cytometry analysis revealed percentages of CD4⁺ and CD8⁺ T cells from TIL-F1 to TIL-F4. b CD4/CD8 ratio. c The percentages of PD-1⁺T cells in four TIL fragments. d Comparision of PD-1 expression. e Comparison of memory-phenotype T cells. f IFN-γ ELISPOT analysis of all four TIL fragments cocultured with autologous tumor cells (ATCs). TILs with no targets are negative controls. Medium well is the blank negative control and OKT-3 well is the positive control. Column histogram summarized the number of positive spots. g IFN-γ ELISA measurement of all four TIL fragments cocultured with ATCs. T cells with no targets are negative controls. The results shown are representative of independent experiment done with repeated 3 times

To further validate anti-tumor activity of TIL-F1 and isolate tumor-reactive TCRs, TIL-F1 was stimulated with ATCs for twice, and single CD8⁺ CD137⁺ TIL-F1 before and after stimulation (namely PRE and POST) were both sorted into 96-well plates and were amplified using single-cell PCR to obtain their TCRs, presented in Fig. 2. Therefore, the most dominant TCR of post-stimulated CD8⁺ CD137⁺ TIL-F1 could be most likely tumor-reactive TCR, which was transduced into donor PBLs to generate tumor-reactive TCR-Ts. To exclude possibility of non-specific amplification of TIL-F1 stimulated with ATCs, TIL-F4 as a negative control was also stimulated with ATCs based on the same stimulation procedure as TIL-F1. The flow cytometry-based assays indicated that the percentage of tumor-reactive T cells in post-stimulated TIL-F1 was significantly higher than that in pre-stimulated TIL-F1 and yet percentages of tumor-reactive T cells in pre- and post-stimulated TIL-F4 were low and similar (Fig. 3).Fig2Fig. 2

Flowchart for isolation of TCRs from CD137 positive TIL-F1 and function verification of corresponding TCR-Ts. TIL-F1 was stimulated with ATCs for twice, and single CD8⁺ CD137⁺ TIL-F1 before and after stimulation (namely PRE and POST) were both sorted into 96-well plates and were amplified using single-cell RT-PCR to obtain their TCRs. Subsequently, the most dominant TCR in CD8⁺ CD137⁺ TIL-F1 before and after stimulation was cloned into lentiviral vector and introduced into donor peripheral blood lymphocytes (PBLs) to generate tumor-reactive T cell receptors-engineered T cells (TCR-Ts). Finally, we evaluated whether this TCR-Ts could specifically identify and recognize ATCs in vivo and in vitro

We isolated approximately 100 CD8⁺ CD137⁺ T cells from pre- and post-stimulated TIL-F1 co-cultured with ATCs. 43 and 42 TCR α/β chain pairs were identified by single-cell TCR sequencing from pre- and post-stimulated CD8⁺ CD137⁺ T cells, respectively (Table 1 and Table 2). The percentages of the first, second and third ranked TCRs in pre-stimulated CD8⁺ CD137⁺ TIL-F1 were 58.54, 14.63, 9.76%, respectively (Table 1, Fig. 4a). After stimulation, the percentage of only the first ranked TCR (namely TCR1) in pre-stimulated TIL-F1 significantly increased (Table 2, Fig. 4b), which suggested that the first ranked TCR could be most likely tumor-reactive TCR.Tab1

TCR genes of CD8⁺CD137⁺T cells in pre-stimulated TIL-F1

To optimize expression of introduced TCR in T cells, TCR was constructed in a β/α chain order and constant regions was replaced by mouse counterparts modified with interchain disulfide bond and hydrophobic substitution as previously described (Fig. 4c) [17, 25]. The first ranked TCR was lentivirally transduced into PBLs to generate TCR-Ts with transduction efficiency of more than 50% (Fig. 4d).

The ability of these TCR-Ts to specifically identify and mediate effector functions in response to autologous tumor cells in vitro was evaluated with cytokine production and T cell cytotoxicity assays. The TCR-Ts showed high levels of IFN-γ secretion and specific cytotoxicity to ATCs (Fig. 4e, f and Additional file 1: Figure S5). Besides, we found that TCR-Ts displayed significantly diminished levels of IFN-γ with HLA class I blocking antibodies, which indicated the recognition effect of TCR-Ts was mainly restricted by HLA class I presentation (Additional file 1: Figure S4). To further explore the potential cytotoxicity of TCR-Ts to mediate regression of tumor in vivo, a single intravenous injection of TCR-Ts with intraperitoneal DAC was administered into patient-derived xenograft (PDX) models [18, 19]. Of note, administration of TCR-Ts induced a statistically significant regression of tumors than untreated group, untransduced T cells treated group or DAC treated group (Fig. 4g, p < 0.001). These findings revealed that tumor-reactive TCR-Ts could exert an antitumor activity against ESCC in vivo and in vitro.