Self-Maintaining CD103⁺ Cancer-Specific T Cells Are Highly Energetic with Rapid Cytotoxic and Effector Responses

CD103 (integrin α_E) is primarily expressed on tissue-resident memory T cells (Trm) and is widely recognized as a key player in viral tissue-localized immunity, including in the clearing of murine herpes simplex virus skin infection and murine lung influenza infection (1, 2). Increasing interest is seen for CD103′s potential in human cancer control and eradication. It has been shown that enrichment of CD103⁺ tumor-infiltrating T lymphocytes (TIL) is associated with improved outcomes in cancer patients with either melanoma, colorectal, urothelial, or early-stage lung cancers (3–7).

The induction and development of CD103 expression on developing T cells require initial antigenic contact and the presence of microenvironmental TGFβ1 cues, presumably produced by murine Batf3-dependent dendritic cells (DC), human CD1c⁺ conventional DCs, and regulatory T cells (Treg; refs. 8–10). Impaired CD103⁺ Trm populations in tissues occurs in the absence of TGFβ1, as well as the inactivation of its receptor on T cells (11). However, because DCs in the tumor microenvironment are known to be dysfunctional (12–14), it remains unclear how continuous CD103 expression on differentiated cancer-specific CTLs is maintained, especially during prolonged cancer exposure.

Human cancer testis antigens (CTA), such as synovial sarcoma X-2 (SSX-2) and New York-esophageal squamous cell carcinoma-1 (NY-ESO-1) proteins, are a class of antigens that are not expressed in healthy tissue but are upregulated in cancer cells. Although CTA-specific CTLs are promising targets for immunotherapy, they have low-affinity TCRs and sensitivity to tumor antigen/MHC complexes (15, 16). This can result in an overall impaired efficacy of antitumor T-cell responses and thus poor tumor control. Interestingly, CD103 molecules accumulate in the immunologic synapse following T-cell interaction with autologous cancer cells (17, 18). It is, therefore, possible that CD103 could play a role in TCR-mediated antitumor CTL immune responses.

Tumor-reactive TILs with CD103 and CD39 coexpression are shown to have increased inhibitory receptor expression (19). However, because enrichment of CD103⁺ TILs is known to improve clinical outcomes in patients, its possible dysfunctional immunophenotype raises uncertainty about its relevance as a primary target for immunotherapy and its significance to overall antitumor immune activities. Therefore, a comprehensive understanding of the properties of human CD103⁺ CTLs and the potential contribution of CD103 toward antitumor T-cell effector and cytotoxic responses is urgently required.

In this study, we sought to investigate CD103-mediated characteristics and immune mechanisms of human cancer-specific CTLs. We found that differentiated cancer-specific CTLs could self-produce an activated form of TGFβ1 to maintain continual CD103 expression, without reliance on external TGFβ1 cues from other cells. These CD103⁺ cancer-specific CTLs had increased tumor antigen sensitivity and elevated metabolic potential. Using live-cell imaging, we clearly demonstrated that CD103⁺ CTLs had faster cancer recognition, which led to rapid and cumulative E-Cadherin⁺ cancer cell killing following prolonged cancer exposure. We also demonstrated that CD103^highCD39⁺ TILs, but not CD103^lowCD39⁺ TILs, had cumulative inhibitory receptor coexpression ex vivo and elevated caspase-3–mediated apoptosis after prolonged exposure to cancer. Taken together, our study highlights key antitumor features of CD103⁺ cancer-specific CTLs and provides information that will help in the design of better and more effective tissue-localized T cell–based cancer immunotherapies.

This study was approved by the Oxford Radcliffe Biobank (ORB) research ethics committee (reference number: 09/H0606/5+5), based on the guidelines of the Declaration of Helsinki. Tumor and paratumor tissues were collected from 5 lung cancer patients for ex vivo analysis. Criteria chosen of these patients include nonmetastatic, early TNM stage, carcinoma-confirmed cancer patients. Sizes of tissue samples collected are kept consistent between each patient (resection volume of no less than 0.5 × 0.5 × 0.5 cm), and the patients did not have metastatic disease. Samples were collected during patients' surgery, stored in RPMI-1640 on ice, and immediately processed. The tumor samples were confirmed as malignant using immunohistopathology. Written informed consent was obtained from all subjects prior to the inclusion in the study. Paired tumor and paratumor tissue samples from 5 lung cancer patients were used for ex vivo data analysis, as described below. Peripheral blood was obtained from patients with confirmed presence of cancer-specific T cells for the in vitro functional assays. The patients' clinical details are described as in Supplementary Table S1.

All in vitro functional experiments were conducted using clonally expanded cancer-specific CTLs that were generated from cancer patients, using a well-established human antigen-specific T-cell generation technique as previously described (20–22). A detailed flow chart of the generation of human cancer-specific CTLs is shown in Supplementary Fig. S1. Briefly, mononuclear cells (consisting of a mixture of T cells, B cells, and antigen-presenting cells) from individual patients' blood were isolated using a Ficoll-Hypaque density gradient centrifugation. Around 3 million cells were then stimulated with either cancer SSX2_41-49-specific KV9 peptide (KASEKIFYV) or NY-ESO-1_157-165–specific SC9 peptide (SLLMWITQC; 10 μg/mL; PeproTech) in RPMI-1640 supplemented with 10% v/v heat-inactivated human AB serum (National Blood Service), 2 mmol/L L-glutamine, 1% v/v (500 U/mL) penicillin/streptomycin (Sigma-Aldrich), and recombinant human IL2 (200 U/mL; PeproTech; H10) at 37°C for 14 days. B cells and antigen-presenting cells within the isolated mononuclear cell mixture present the peptides to the T cells for activation and proliferation.

After 14 days, CD103⁺ and CD103⁻ cancer-specific CTLs were sorted using a PE-conjugated HLA class I tumor peptide tetramer [HLA-A2 KV9 or HLA-A2 SC9 tetramers; in-house preparation as in refs. 22, 23), FITC-CD8 (clone: SK1; BD Biosciences; RRID:AB_2739852), and BV711-CD103 (Ber-ACT8, BioLegend; RRID: AB_2629650) using a BD Aria IIlu (BD Biosciences). The purity of the sorted population was confirmed at 95% purity. The isolated cells were then clonally expanded and cultured in vitro. The sorted cells were cultured with 2 million irradiated cone blood cells (Oxford NHS Blood and Transplant Service) for 14 days in H10 medium (as described above), supplemented with recombinant human IL2 (200 U/mL; PeproTech). The T cells were confirmed for antigen specificity using HLA-A2 KV9 and SC9 tetramers staining, ran on Attune Nxt flow cytometry (Thermo Fisher), analyzed using FlowJo v.10 (TreeStar Inc.). The validated T cells were then stored in several batches for future assays and used for all in vitro functional assays. T-cell cultures were tested for Mycoplasma monthly, maintained for 2 months for every assay, passaged once after every thaw, and reauthenticated again using the tetramers method after every thawing and passage.

All the following in vitro functional assays were performed by culturing CTLs with target cells [HCT116 (CCL-247; obtained from ATCC, in 2017) or THP-1 (TIB-202; from ATCC, in 2017)]. Each commercial cell line was authenticated every 6 months, with specific surface markers known to be expressed by each cell line, ran on Attune Nxt flow cytometer (Thermo Fisher), and analyzed on FlowJo v.10 (TreeStar Inc.). Antibodies used include APC–E-Cadherin (67A4, BioLegend; RRID: AB_756069) for HCT116 or APC-CD33 (WM53; BioLegend; RRID: AB_314352) for THP-1. Each cell line was cultured in vitro as per the manufacturer's guidelines. Briefly, HCT116 cells were cultured at 2 million cells in McCoy 5′A medium (Thermo Fisher), whereas THP-1 was cultured in RPMI-1640 supplemented with 10% v/v fetal calf serum (Thermo Fisher) for 4 days prior to any T-cell functional assays. Each cell line was tested for Mycoplasma upon receipt from the vendor and every month.

Prior to culturing, target cells were externally loaded with corresponding peptides (SSX-2 or NY-ESO-1, 10 μg/mL) at 37°C for 1 hour to enable the loading and presentation of the peptides on the MHC complex on the target cell surface. HCT116 is a commercially available cancer cell line that is known to be HLA-A2⁺, E-Cadherin⁺ (the ligand for CD103) and does not express SSX-2 and NY-ESO-1 proteins (24–27). Another condition was also included by using THP-1 as target cells, which is E-Cadherin⁻ HLA-A2⁺ SSX-2⁻ NY-ESO-1⁻.

To assess cytokine production, 0.2 million CTLs were either treated with blocking anti-CD103 (15 μg/mL; Ber-ACT-8, BioLegend) or isotype control (mouse IgG1κ; 15 μg/mL; BioLegend) for 20 minutes. Fifty microliters of supernatants were collected and diluted at ratio 1:5 in RPMI, after 48 hours of coculturing T cells with peptide-loaded HCT116 target cells at an effector-to-target (E:T) ratio of 1:10 at 37°C. IFNγ was quantified using the Bio-Plex Pro Human Cytokine Assay (Bio-Rad) and ran on Bio-Plex 200 and analyzed using Bio-Plex Manager (Bio-Rad). Concentration was calculated automatically by the Bio-Plex Manager. Briefly, fluorescence intensity was measured by the machine and concentration was quantified based on standard (diluted by 1:100) for five times, with known original concentration. Standard was provided by the vendor (Bio-Rad) and purchased before every Bio-Plex assay. Cytokine concentration was then corrected using control (supernatant without T cells present). TGFβ1 presence in the supernatant was evaluated using the Human TGF-beta-1 Quantikine ELISA kit (R&D Systems), according to the manufacturer's instructions. Briefly, 50 μL of supernatant was used every time, diluted 1:20 in RPMI-1640, ran on MultiSkan FC Microplate Photometer (Thermo Fisher). The same control and concentration correction method was used for the ELISA as per used for the Bio-Plex assay.

CTLs (0.2 million) were treated with monensin (10 μg/mL) and brefeldin A (19 μg/mL; provided by BD Biosciences) for 15 minutes prior to coculturing with peptide-loaded HCT116 target cells at an E:T ratio of 1:10 at 37°C for 5 hours. Cells were then fixed with Cytofix/Cytoperm (BD Biosciences) and stained with conjugated antibodies including Alexa Fluor488-IFNγ (B27; BD Biosciences; RRID: AB_396827), PE-LAP (TW4-6H10; BioLegend; RRID: AB_10639862), and APC-CD107a (H4A3; BD Biosciences; RRID: AB_1727417). Samples were ran on Attune Nxt flow cytometer (Thermo Fisher), and analyzed on FlowJo V.10 (TreeStar Inc.). Evaluation of responses following anti-CD103 treatment was carried out using the antibody blocking treatment mentioned above.

HCT116 target cells were initially stained with 0.5 μmol/L carboxyfluorescein diacetate succinimidyl ester (CFSE; Thermo Fisher) before peptide loading at four different concentrations for 1 hour (2 μg/mL, 1 μg/mL, 0.4 μg/mL, and 0.08 μg/mL), and cocultured with CTLs at E:T ratio of 1:10 at 37°C for 6 hours. For the prolonged cancer exposure assay, target cells were cocultured with CTLs hourly for up to 12 hours. Cells were then stained with 7-AAD (BD Biosciences) and either APC-E-Cadherin (67A4, BioLegend; RRID: AB_756069) for HCT116 or APC-CD33 (WM53; BioLegend; RRID: AB_314352) for THP-1. Assessment of cell death was based on the CFSE⁺7-AAD⁻ population present. Apoptosis-induced death was measured hourly by the presence of activated caspases in CTLs following 12-hour exposure to target cells, using the FITC-active caspase-3 apoptosis kit (BD Biosciences), ran on Attne Nxt flow cytometer (Thermo Fisher) and analyzed on FlowJo v.10 (TreeStar Inc.).

Briefly, the T-cell receptor was sequenced by isolating mRNA from CTLs using the RNAeasy Mini Kit (Qiagen) and cDNA was synthesized from 500-ng mRNA using the SMARTer RACE cDNA Amplification kit (Takara), as per manufacturer's instruction. Briefly, cDNA was PCR-amplified for the CDR3 region of both alpha and beta TCR chains using the PCR Advantage Kit (Takara) using the following primers for TRAC: 5′- GGAACTTTCTGGGCTGGGGAAGAAGGTGTCTTCTGG-3′ and for TRBC: 5′- TGCTTCTGATGGCTCAAACACAGCGACCT-3′ and run on a 1% agarose gel for PCR band confirmation (TRAV band at 700 bp and TRBV band at 500 bp). PCR product was then purified using NucleoSpin Gel and PCR clean-up kit (Macherey-Nagel) and transformed into TOP10 competent cells (Thermo Fisher) before being plated on LB agar media at 50 μL per plate, as per manufacturer's instructions. Colony PCR was performed using 200 ng of initial PCR product to amplify the product before isolating the plasmid DNA using the Spin Miniprep kit (Qiagen). The purified plasmid DNA was then sent for sequencing at a 100 nmol/L concentration, performed by Timothy Rostron and John Frankland of the Sequencing Facility, Weatherall Institute of Molecular Medicine, University of Oxford. Sequencing data of T-cell clones are described in the Results section and Supplementary Table S2.

Three days before coculture of cancer-specific CTLs with 2 μg/mL peptide-loaded HCT116 cells, or THP-1 cells were seeded in a 24-well plate at 0.25 million cells/mL in DMEM, supplemented with 10% v/v fetal calf serum (Sigma-Aldrich), 2 mmol/L L-glutamine, and 1% v/v (500 U/mL) penicillin/streptomycin (Sigma-Aldrich). Target cells were incubated at 37°C for 2 hours, to allow cancer cell lines to form a monolayer and produce cytokines and chemokines. One million cells/mL CTLs were seeded on a 5-μm pore size transwell insert (Corning) and rested in 37°C incubator to let the cells to settle to the bottom of the insert. After 1 hour, inserts were transferred to the 24-well plates containing the target cells and further incubated at 37°C for 2 hours. Following this, cells in each well were harvested, stained with LIVE/DEAD Fixable Aqua Dead Cell Stain Kit (Thermo Fisher) at 4°C for 20 minutes, with subsequent PerCP/Cy5.5-CD8 (SK1; BD Biosciences; RRID:AB_2687497) staining at the same temperature and duration before quantified using Attune Nxt flow cytometer (Thermo Fisher), and analyzed using FlowJo v.10 (TreeStar Inc.). Migration was calculated as the total number of live CTLs present in the 24-well plate. Evaluation of migration following anti-CD103 treatment was performed by initially treating CTLs with 15 μg/mL anti-CD103 or isotype control for 20 minutes before coculture. Imaging was performed using IncuCyte (Essence Bioscience) and quantified using ImageJ (NIH).

CTLs were starved of IL2 3 days prior to the experiment. After which, 7.5 million cells/mL CTLs were seeded on a Seahorse 96-well microplate (Seahorse Biosciences) containing XF complete media (for mitochondrial stress: DMEM plus 2 mmol/L L-glutamine, 10 mmol/L glucose, and 1 mmol/L sodium pyruvate; for glycolytic stress: DMEM plus 2 mmol/L L-glutamine). For the mitochondrial stress assay, assay was performed using Seahorse XF cell mito stress test kit, per the manufacturer's instruction. Briefly, cells were injected initially with 1 mmol/L oligomycin (inhibits ATP synthase complex V), then 1.5 mmol/L carbonyl cyanide-4 (trifouromethoxy) phenylhydrazone (FCCP; disrupts mitochondrial membrane potential), and finally with 100 μmol/L rotenone (complex I inhibitor)/1 mmol/L antimycin A (complex III inhibitor) at 20-minute intervals. For the glycolysis stress assay, assay was performed using Seahorse XF glycolysis stress test kit, per manufacturer's instruction. Briefly, cells were injected initially with 10 mmol/L glucose, then 1 μmol/L oligomycin, and finally with 50 mmol/L 2-deoxy-glucose (2-DG; inhibitor of glycolysis) at 20 minutes intervals. Drug injection and analysis were performed by the Seahorse XF96 extracellular Flux analyzer (Seahorse Biosciences), with Seahorse Wave software (Agilent). The oxygen consumption rate (OCR) was recorded as pmol/min, whereas the extracellular acidification rate (ECAR) was recorded as mpH/min.

HCT116 or THP-1 cells were seeded at 0.05 million cells/mL per well in μ-slide 8-well glass bottom (Ibidi; cat number: 80841) in DMEM supplemented with 10% v/v fetal calf serum, 2 mmol/L L-glutamine, and 1% v/v (500 U/mL) penicillin/streptomycin and incubated at 37°C for 2 days to allow cells to attach to the glass, grow, and produce cytokines and chemokines. Peptide (1 μg/mL) was then added 1 hour before imaging, with the seeded target cell culture supplemented with 2 mmol/L calcium chloride and Alexa-Fluor 647-Annexin V (5 μL/well; cat. number: A23204; Thermo Fisher). The imaging chamber was then loaded onto a spinning disk confocal microscope (Zeiss Cell Observer Spinning Disc Confocal; Zeiss), equipped with a 10× air objective, enclosed in an incubator chamber at 37°C and 5% CO₂. CTLs were then added at 2 million cells/mL and imaging was carried out on each individual well in parallel, using brightfield and 632-nm excitation at 20% of maximum. Imaging was carried out for 12 hours at 1 frame per 4 minutes. Annexin V–positive staining is illustrated in the figures as blue.

For the killing duration assay, target cells and CTLs were prepared as described above. CTLs were then labeled with calcium-sensitive probe Fluo4-AM (10 μg/mL; F14201, Sigma-Aldrich) and 2.5 μmol/L probenecid (Sigma-Aldrich) and incubated at 37°C for 5 minutes. CTLs were then washed twice in fresh medium containing probenecid. The imaging chamber was set up as described above. Fluo4-AM–labeled CTLs were added to each well of the imaging slide before imaging was carried out using brightfield and 488-nm excitation at 20% of maximum. Imaging was carried out for 2 hours at 4 frames per minute. Fluo4-AM–positive staining is illustrated in the figures as red. All images were analyzed using ImageJ (NIH).

Freshly collected tissues were cut into 1-mm³ pieces and digested with the human tumor dissociation kit in C-tubes (Miltenyi Biotec) following the protocol provided by the manufacturer. The C-tube was fixed on a rotator (Miltenyi Biotec) and rotated at low speed whereas the tumor tissue was enzymatically digested at 37°C. GentleMACS Octodissociator (Miltenyi Biotec) was used to physically disrupt the tissue between incubation with enzymes. At the end of dissociation, the cells were filtered through 100-μm strainer to obtain the single-cell suspension. The single cells were pelleted and washed once with 50 mL RPMI (Thermo Scientific Heraeus, 40R) at 300 × g for 10 minutes at room temperature. The cells were resuspended in R10. Then the mononuclear cells were isolated with LymphoPrep (STEMCELL Technologies Inc.) by following the protocol provided by the manufacturer. Mononuclear cells isolated from paired tumor and paratumor tissue were first stained with the LIVE/DEAD Fixable Aqua Dead Cell Stain Kit (Thermo Fisher) before being labeled with conjugated antibodies, with each step of incubation at 4°C for 20 minutes. All samples were acquired on BD LSR Symphony (BD Biosciences) and analyzed using FlowJo v.10 software (TreeStar Inc.).

Antibodies used for the surface staining include BV510-CD16 (3G8; BioLegend; RRID:AB_2562085), BV510-CD56 (HCD56; BioLegend; RRID:AB_2561944), BV510-CD19 (HIB19; BioLegend; RRID:AB_2561668), BV510-CD14 (M5E2; BioLegend; RRID:AB_2561946), BUV395-CD3 (SK7; BD Biosciences; RRID:AB_2744382), PerCP/Cy5.5-CD8 (SK1; BD Biosciences; RRID:AB_2687497), BUV737-CD4 (SK3; BD Biosciences; RRID:AB_2713927), PE-CD56 (MEM-188; BioLegend; RRID:AB_314447), APC-NKG2a (131411; R&D Systems; RRID:AB_356987), PE-Texas Red-CD39 (A1, BioLegend; RRID:AB_2564318), BV785-CD103 (Ber-ACT8; BioLegend; RRID:AB_ 2734364), BB515-Tim3 (7D3, BD Biosciences; RRID:AB_ 2744368), BV650-PD-1 (EH12.2H7; BioLegend; RRID:AB_2566362), APC-R700-TIGIT (1G9, BD Biosciences; RRID:AB_2739254), PE/Cy7-CD27 (M-T271; BioLegend; RRID:AB_2562258), BV421-CCR7 (G043H7; BioLegend; RRID:AB_11203894), BV711-CD45RA (HI100; BioLegend; RRID:AB_11218999), and FITC-CD28 (CD28.2; BD Biosciences; RRID:AB_396071). In vitro integrin-β8 staining on CD103⁺ and CD103⁻ CTLs (isolated from tumor tissue of cancer patient) following exposure to target cells at an E:T ratio of 1:10 for 12 hours was performed using APC-integrin beta 8 (416922; R&D Systems; RRID: AB_2044685).

Preparation of IHC slides was as previously described (28, 29). Briefly, adjacent tissue sections of formalin-fixed, paraffin-embedded tumor tissue from 5 lung cancer patients were stained immunohistochemically for E-Cadherin (NCH-38; Dako; RRID:AB_2341210), CD103 (EPR4166(2); AbCAM, UK; RRID:AB_1142856), and CD8 (C8/144B; Agilent Technologies; RRID:AB_2075537), and automated staining was carried out using the Leica BOND-MAX autostainer (Leica Microsystems) following antigen retrieval using Epitope Retrieval Solution 2 (Leica Microsystems). Analysis was conducted using the Visiopharm Integrator System platform made for analyzing digitized serial slides, with the protocol implemented as Analysis Protocol Packages (APP). Several -APPs were designed to quantify slides stained individually with the antibodies following tissue section realignment. Realignment was performed at both large scale and finer detailed level, to get the best match between the 3 adjacent slides. Region of interest (ROI) was conducted, and first auxiliary APP runs were performed using threshold classification that identifies the tissue regions. The second auxiliary APP runs done on the E-Cadherin slides used threshold classification identifying positive and margin regions. ROI was then superimposed on aligned CD8- and CD103-stained slides for subsequent analysis. An HDAB-DAB color deconvolution band was used to detect positively stained cells on CD103 and CD8 slides. To enhance the stained cells, while suppressing the background variation, several preprocessing steps were included. This included the color deconvolution bands were input to a threshold classifier. The thresholding classification method defines a threshold for a given feature and assigns one class to all pixels with a feature value above or equal to that value, and another class for the rest.

The classification rule is defined as:

where T is the user-selected threshold (cutoff value), A and B are the labels/classes to which the pixel is assigned. For postprocessing steps, a method for cell separation that is based on shape and size was used (to exclude the effect of slide preparation). Cell areas that were too small (if a cell was found to be dysregular or deemed as having cellular membrane damaged) were removed, and we applied an unbiased counting of frames to avoid the cells intersecting with neighboring tile boundaries, which were counted twice (or more). Cells present at “within” were defined as cells located within the E-Cadherin⁺ regions, cells present at “clustering” were defined as cells located within 1.5 cm from the E-Cadherin⁺ regions, and cells present at “distal” were defined as cells located over 1.5 cm from the E-Cadherin⁺ regions.

All graph generation and statistical analyses were conducted using GraphPad Prism v.7 software. Unless stated otherwise, data are summarized as median ± SEM. The number of patients used for each ex vivo analysis is shown in the figure legends. Each in vitro CTL functional assay was replicated 3 times. Statistically significant differences between 2 groups were assessed using a two-tailed paired t test, with Wilcoxon adjustments for nonparametrically distributed variables. For comparisons between more than 2 paired groups of tissues or treatments, one-way ANOVA with Tukey multiple comparison test was performed. Statistical analysis of the metabolic assays, as well as comparisons between different blocking treatments, was carried out using two-way ANOVA with the Tukey multiple comparison test. All tests were two-sided, and differences were considered as statistically significant as P < 0.05. Data inquisition is upon request.

Following initial antigen contact, naïve T cells develop and acquire CD103 expression in the presence of TGFβ1, known to be produced by DCs and Tregs (8–10). However, tumor-infiltrating DCs are dysfunctional in cancer patients (12–14). Thus, it remains unclear how CD103 expression is maintained on matured, differentiated cancer-specific CTLs. Paired CD103⁺ and CD103⁻ cancer-specific CTLs were sorted and clonally expanded (SSX-2–specific CTLs; NY-ESO-1–specific CTLs) as described in our previous studies (refs. 20, 21, 23; Supplementary Fig. S1). Both paired CD103⁺ and CD103⁻ cancer-specific CTLs had a predominantly late-stage effector memory phenotype (CD28⁻CD27⁻CD45RO⁺CD45RA⁻CCR7⁻) and expressed the standard Trm marker CD69 (Supplementary Fig. S2A and S2B). Each pair of CTLs had the same TCR usage and TCR surface density (Supplementary Fig. S2C and S2D). Specifically, paired CD103⁺ and CD103⁻ SSX-2–specific CTLs used TRAV 8-6 TRAJ 30 and TRBV 6-1 TRBJ 2-7, whereas paired CD103⁺ and CD103⁻ NY-ESO-1–specific CTLs used TRAV 12-2 TRAJ 31 and TRBV 12-4 TRBJ 1-2. The standardization of TCR and phenotype between paired CTLs helps (i) to eliminate other confounding factors that could affect CTLs functions, such as different TCR affinity and recognition strength, and (ii) to ensure any observational changes to the CTL function is due to the action of CD103 on T cells.

As TGFβ1 is known to induce CD103 expression on CTLs (8–10), we therefore assessed the TGFβ1 on CD103⁺ cancer-specific CTLs. Significantly high TGFβ1 was produced by CD103⁺ cancer-specific CTLs but not by the CD103⁻ subset following coculture with peptide-loaded HCT116 cells (Fig. 1A). A significant production of TGFβ1 by CD103⁺ CTLs was observed, even after taking into consideration the cytokine production by HCT116 cells (Fig. 1A).

Previous studies show that HCT116 do not express NY-ESO-1 and SSX-2 proteins (24–27). In our current study, we confirmed that cancer-specific CTLs did induce IFNγ and TNFα in cocultures with HCT116 cells in the absence of external peptide loading (Supplementary Fig. S3). Therefore, HCT116 cells itself did not stimulate SSX-2–specific and NY-ESO-1–specific T-cell responses due to the lack of SSX-2 and NY-ESO-1 expression.

Because the presence of CD103 on cancer-specific CTLs may contribute to self-TGFβ1 production, we next evaluated whether interfering with the interaction of CD103 and its ligand E-Cadherin on HCT116 cells could affect TGFβ1 production. Following anti-CD103 treatment, we found that the TGFβ1 production still remained high (Fig. 1B). The TGFβ1 self-producing capacity by CD103⁺ CTLs was validated using intracellular cytokine staining analysis. The majority of CD103⁺ CTLs expressed intracellular TGFβ1, which was not seen with CD103⁻ CTLs (Fig. 1C and D). TGFβ1 was also expressed by CD103⁺ CTLs after treatment with anti-CD3/CD28 in the absence of target cells (Fig. 1D). These observations, therefore, suggest that CD103⁺ CTLs independently expressed and produced their own TGFβ1, which is likely responsible for sustaining its CD103 expression.

Because the active form of TGFβ1 was an essential component for CD103 induction, we then assessed whether integrin β8 (known to cleave the inactive form of TGFβ1) was upregulated on TGFβ1⁺CD103⁺ CTLs. TGFβ1⁺CD103⁺ CTLs had significant upregulation of membranous β8 compared with the CD103⁻ CTLs following peptide-loaded target cell exposure for 12 hours (Fig. 1E and F). The upregulation of integrin β8 on CD103⁺ CTLs' cell surface likely suggests that this marker may contribute to CD103 expression maintenance on matured T cells by facilitating production of the TGFβ1 active form.

It is well established that human CTA-specific T cells have low-affinity TCRs (15, 16). Thus, we next assessed whether CD103 presence on T cells could assist in improving tumor antigen sensitivity and strengthening TCR-mediated effector responses. Following coculture of CTLs with HCT116 cells preincubated with various doses of peptides, the CD103⁺ cancer-specific CTLs demonstrated significantly elevated IFNγ responses compared with the CD103⁻ CTLs and was sensitive up to 0.4 μg/mL antigen stimulation (Fig. 2A). The reduced IFNγ production by CD103⁺ CTLs following treatment with anti-CD103 (Fig. 2A) clearly indicated that CD103 helped to improve tumor antigen sensitivity of cancer-specific CTLs.

CD107a marker is an indicator of degranulation and potential granule-mediated cytotoxic responses of T cells. Therefore, we evaluated whether the enhanced TCR tumor antigen sensitivity affected degranulation by CD103⁺ CTLs. Consistent with the IFNγ response, we observed that CD107a expression was also significantly elevated on CD103⁺ CTLs and was sensitive up to 0.4 μg/mL antigen stimulation (Fig. 2B), and degranulation was impaired following anti-CD103 treatment of CD103⁺ CTLs (Fig. 2B). We then assessed whether the enhanced degranulation capacity could improve CD103⁺ CTL antitumor cytotoxicity. As expected, significantly higher numbers of HCT116 target cells were killed by CD103⁺ cancer-specific CTLs, even with 0.4 μg/mL antigen stimulation (Fig. 2C).

Because E-Cadherin is a known ligand of CD103, we next asked whether their interaction was necessary to improve tumor antigen sensitivity. To do this, we cocultured CD103⁺ CTLs with the E-Cadherin–negative THP-1 cancer cell line. As shown in Fig. 2D, CD103⁺ CTLs had significantly reduced degranulation in cocultures with THP-1 cells compared with the E-Cadherin-positive HCT116 cells. We further found that in the presence of THP-1 cells, CD103⁺ CTLs showed significant reduction of target cell killing (Fig. 2E). This further confirmed that CD103 and its interaction with E-Cadherin on target cells contributed to enhanced TCR recognition and tumor antigen sensitivity, thus resulting in enhanced granule-mediated cytotoxicity.

We next assessed whether tumor antigen–sensitive CD103⁺ cancer-specific CTLs could perform more efficient single-cell cancer killing. Using live-cell imaging, we measured the timing of individual cytotoxic events following CTL exposure to peptide-loaded HCT116 cells. CD103⁺ CTLs initiated significantly faster target cell killing compared with the CD103⁻ CTLs, which required a longer period of time to initiate killing (Fig. 3A and B; Supplementary Movies S1–S4), and higher numbers of CD103⁻ CTLs were required to initiate the killing of individual target cells (Fig. 3A; Supplementary Movies S1–S4). These observations suggest that the presence of CD103 helps mediate a quicker cytotoxic T-cell response. Indeed, the efficient individual target cell killing by CD103⁺ CTLs became more apparent as early as 4 hours after target cell exposure (Fig. 3C and D). The cytotoxic activities of CD103⁺ CTLs resulted in significantly higher HCT116 cell death over a 12-hour time period compared with CD103⁻ CTLs (Fig. 3E; Supplementary Fig. S4).

To confirm that the CD103 and E-Cadherin interaction was contributing to the enhanced CTL cytotoxicity, we also cocultured CD103⁺ CTLs with peptide-loaded, E-Cadherin–negative THP-1 cells. No difference in THP-1 cell death following coculture with either CD103⁺ CTLs or CD103⁻ CTLs was seen across a 12-hour time period (Supplementary Fig. S5). Taken together, these observations clearly highlight the importance of CD103 and its interaction with E-Cadherin in elevating CTL antitumor responses.

Previous studies have highlighted the importance of aerobic glycolysis (glucose oxidation) not only as the major energy source for activated T-cell function, but also as a regulator of cytokine production (30, 31). Thus, we next investigated the metabolic capacity of CD103⁺ cancer-specific CTLs by evaluating their glycolytic activity. CD103⁺ CTLs had significantly higher basal glycolytic rates compared with the CD103⁻ CTLs (Fig. 4A and B). CD103⁺ CTLs also showed significantly elevated maximal glycolytic capacity (Fig. 4A and C). This suggested that CD103⁺ CTLs could utilize and convert glucose at a faster rate to produce energy, which could potentially be used to induce more efficient and rapid immune responses.

However, in the tumor microenvironment, enhanced glucose consumption by cancer cells is known to deprive TILs of glucose for their cellular functions (32). It is, therefore, vital for CD103⁺ cancer-specific CTLs to be able to efficiently catalyze multiple metabolites, especially during conditions of stress. Therefore, we evaluated the total metabolite consumption capacity of activated CD103⁺ CTLs by assessing the total OCR as a measure of the overall T-cell mitochondrial activity. We found that the CD103⁺ CTLs had a significantly higher basal metabolic activity compared with the CD103⁻ CTLs (Fig. 4D and E). CD103⁺ CTLs also had a larger spare mitochondrial capacity (Fig. 4D and F), suggesting their potential in utilizing various metabolites at higher levels compared with the CD103⁻ CTLs in order to meet energy requirements.

Activation of CD103 and its cytoplasmic tail domain in T cells is known to enhance T-cell retention in tumor tissue (33–35). We therefore assessed whether CD103 could mediate better recognition, homing, and retention of T cells to cancer cells. To do this, we initially placed either the CD103⁺ or CD103⁻ cancer-specific CTLs across a transwell barrier, away from HCT116 cells, and allowed any movement to occur within 2 hours period. Within the first hour, the CD103⁺ CTLs migrated in significantly larger numbers across the barrier toward HCT116 cells (Fig. 5A). Treatment of the CD103⁺ CTLs with anti-CD103 significantly reduced the number of T cells migrating toward target cells to a level similar to that of the CD103⁻ CTLs (Fig. 5B). To validate the necessity of CD103 in T-cell mobility and homing, we exposed CD103⁺ CTLs to E-Cadherin–negative THP-1 cells. No significant difference in the numbers of CD103⁺ CTLs and CD103⁻ CTLs migrating toward THP-1 was found (Supplementary Fig. S6). These in vitro data indicated the importance of CD103 in the movement and migration of CTLs toward E-Cadherin–positive cancer cells.

In support of our in vitro observations, we carried out ex vivo IHC on cross-sectional tumor tissues from 5 lung cancer patients. We found that CD103⁺CD8⁺ TILs localized at the E-Cadherin–rich tumor islets, whereas CD103⁻CD8⁺ TILs were mainly in the E-Cadherin–absent tumor regions. CD103⁺CD8⁺ TILs significantly clustered around, as well as within, the E-Cadherin–rich tumor islets at a significantly higher frequency and density (Fig. 6A and B). In contrast, the CD103⁻CD8⁺ TILs were much denser, distal from the E-Cadherin⁺ tumor islets (Fig. 6A and B). Altogether, these observations highlighted that CD103⁺ CTLs had better homing and clustering capacity to cancer cells expressing E-Cadherin.

It was previously shown that TILs coexpressing CD103 and CD39 are tumor-reactive TILs, with increased inhibitory receptor expression (19). However, it remains unclear whether CD103⁺ TILs have different subpopulations with distinctive exhaustion immunophenotypes. We first categorized ex vivo TILs derived from paired tumor and paratumor tissues from 5 lung cancer patients according to their CD103 and CD39 expression. Only the CD103⁺ TILs derived from the tumor expressed high CD39 (CD39^high) but not the CD103⁺ T cells derived from the paratumor tissue (Fig. 7A). In contrast, the CD103⁺CD39^lo T-cell subpopulation could be found in both paired tissue samples (Fig. 7A). This indicated the presence of multiple CD103⁺ TIL subpopulations and suggests potential differences in their immunophenotype. Only the CD103⁺CD39^high TIL subpopulation had higher expression of PD-1, Tim3, and TIGIT (Fig. 7B; Supplementary Fig. S7). This subpopulation had a significantly higher frequency of PD-1⁺Tim3⁺TIGIT⁺ cells (Fig. 7C). In contrast, the CD103⁺CD39^low TIL subpopulation lacked Tim3 and NKG2A expression, with only 40% PD-1⁺TIGIT⁺ cells (Fig. 7D). These observations, therefore, suggested that only some CD103⁺ TILs acquired Tim3 expression and had high coexpression of different inhibitory receptors.

Upregulation of inhibitory receptors on TILs is one of the marks of T-cell activation and exhaustion. To further assess whether prolonged antigen exposure of CD103⁺ CTLs in cancer could impair T-cell survival, we cocultured CD103⁺ CTLs with HCT116 cells for a 12-hour period in vitro. We found that prolonged cancer exposure resulted in significantly increased CD103⁺ CTLs expressing activated caspase-3 compared with the CD103⁻ CTLs (Fig. 7E). Because activated caspase-3 presence is known to be an indicator of cellular apoptosis, we evaluated CD103⁺ CTL cell death after prolonged cancer exposure. As expected, a significantly higher number of CD103⁺ T-cell death was observed following prolonged cancer exposure compared with CD103⁻ T-cell death (Fig. 7F). These observations suggest that CD103⁺ CTLs were more prone to apoptosis following prolonged cancer exposure.

The association of CD103⁺ TIL enrichment with improved clinical outcomes in cancer patients (3, 4) highlights the need to have a clear and comprehensive understanding of the CD103-mediated characteristics of cancer-specific CTLs and their implications for future immunotherapies. We demonstrated that mature differentiated CD103⁺ cancer-specific CTLs could self-regulate their CD103 expression by producing activated TGFβ1, without needing to rely on external cues from other TGFβ1-producing cells. We further showed that the presence of CD103 improved TCR antigen sensitivity and enabled faster cancer recognition and more rapid and efficient cancer cell killing. CD103⁺ CTLs were also found to have an elevated energetic potential and faster migration capacity. However, our data also demonstrated that CD103⁺ cancer-specific CTLs are more susceptible to apoptosis following prolonged cancer exposure.

It is well established that DCs, and most likely Tregs, are the main players in producing TGFβ1 for inducing CD103 expression on developing CTLs (8–10). Both TGFβ1 and antigen exposure are required to induce a CD103 tissue residential T-cell signature (36, 37). We found that mature CD103⁺ cancer-specific CTLs, but not CD103⁻ CTLs, could self-produce activated TGFβ1 to continually sustain prolonged CD103 expression, without needing to rely on other external TGFβ1-producing cells. The self-regulation of CD103 maintenance indicates a possibly unique subset of cancer-specific T cells that could be present in the tumor microenvironment with unique and effective antitumor immune properties. However, the frequency of this CD103⁺TGFβ1⁺ self-regulatory, cytotoxic T-cell population and its importance, especially during cancer growth, merits further investigation.

Our in vitro data showed that CD103⁺ cancer-specific CTLs had higher energy potential compared with CD103⁻ CTLs. However, in the tumor microenvironment, enhanced glucose consumption by cancer cells can deprive TILs of the glucose needed for proper T-cell function (32). This therefore suggests that CD103⁺ CTLs need to be able to utilize different metabolites to meet their high energy demand. Previous work by Kupper and colleagues has highlighted the capacity of virus-specific Trm cells to take up exogenous free fatty acids, and this accounts for the long-term survival of T cells and further helps in mediating protective immunity in virus-infected tissues (38). In light of this, it merits further investigation on whether CD103⁺ cancer-specific Trm cells are able to utilize a variety of metabolites and how this affects their effector function and survival, especially in the glucose-deprived tumor microenvironment.

It is well established that human cancer-specific T cells have low-affinity TCRs and, therefore, have less efficient TCR-mediated activating signaling (15). Because CD103 is known to accumulate in the immunologic synapse (18), it is likely that increasing CD103 and E-Cadherin interactions could strengthen the affinity of TCR binding and, therefore, increase the sensitive engagement with the tumor antigen/MHC complex on cancer cells. We demonstrated that the CD103 and E-Cadherin interaction improved the efficiency of TCR-mediated effector responses by CD103⁺ CTLs. In addition to improve TCR sensitivity, it is also likely that CD103 could assist in improving TCR-mediated signaling. Previous work has highlighted that CD103 could enhance the phosphorylation of 2 key downstream proteins of TCR signaling, ERK1/2 and PLCγ1 (18). Therefore, the specific mechanisms of CD103 in mediating TCR downstream signaling necessitates further investigation.

Coexpression of CD103 and CD39 has been demonstrated to mark tumor-reactive TILs (19). A study also indicates that regulation of TILs in tumor tissue can lead to competition between antitumoral Trm activity and protumoral exhaustion activity (39). Here, we demonstrated varying subpopulations of CD103⁺ TILs that are present in the tumor microenvironment and that only the CD39^highCD103⁺ TIL subpopulation coexpressed multiple inhibitory receptors. Our data also showed that prolonged exposure of CD103⁺ CTLs to cancer could result in more T-cell apoptosis, which goes hand in hand with our previous finding that the overall frequency of CD103⁺ TILs diminishes as tumor size progresses (21). Based on our current study, the CD103⁺CD39^low TIL subpopulation as an alternative tumor-reactive subpopulation could be a critical Trm population that can be improved using novel tissue-localized immunotherapy strategies. In summary, our data shed light on the working and characteristics of CD103⁺ cancer-specific CTLs, which could provide significant insights for future design of novel tissue-localized antitumor T-cell immunotherapies.

No potential conflicts of interest were disclosed.

Conception and design: M. Abd Hamid, M. Fritzsche, Y. Peng, T. Dong

Development of methodology: M. Abd Hamid, H. Colin-York, L. Cerundolo, Y. Peng, T. Dong

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H. Colin-York, M. Browne, J.-L. Chen, X. Yao, C. Waugh, D. Maldonado-Perez, E. Bowes

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Abd Hamid, H. Colin-York, N. Khalid-Alham, S. Rosendo-Machado, C.P. Conlon, Y. Peng, T. Dong

Writing, review, and/or revision of the manuscript: M. Abd Hamid, V. Cerundolo, C.P. Conlon, M. Fritzsche, Y. Peng, T. Dong

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): X. Yao, C. Waugh, M. Fritzsche

Study supervision: M. Fritzsche, Y. Peng, T. Dong

Other (provided tissue samples, chief investigator of Research Ethics Committee under which project completed): C. Verrill

The authors thank the Wolfson Imaging Centre Oxford (WIMM, Oxford) for providing microscope facility support and sequencing of the TCR by John Frankland and Timothy Frankland (Sequencing Facility, WIMM, Oxford). The authors acknowledge the contribution to this study made by the Oxford Centre for Histopathology Research and the Oxford Radcliffe Biobank, which are supported by the National Institute of Health Research (NIHR) Oxford Biomedical Research Centre (BRC), Sally Ann-Clark (WIMM, Oxford) for flow cytometry help, Marahaini Musa [Department of Oncology, UK; Universiti Sains Malaysia (USM), Malaysia] for assistance, Guihai Liu and Bella Bu for mRNA extraction, and Clare Hardman and Janina Nahler for advice on transwell migration assay. The authors acknowledge the editing support of Life Science Editors. The views expressed are those of the author(s) and not necessarily of the National Health Service (NHS), the NIHR, or the Department of Health, UK. This work was supported by the Chinese Academy of Medical Sciences (CAMS) Innovation Fund for Medical Sciences (CIFMS), China (grant number: 2018-I2M-2-002); Medical Research Council (MRC), UK (MR/L018942/1 and MRC Human Immunology Unit Core); and the John Fell Fund (R45603/CN002). M. Abd Hamid is funded by the Malaysia's King Scholarship, M. Fritzsche and H. Colin-York are funded by Wellcome Trust (212343/Z/18/Z) and Engineering and Physical Science Research Council (EPSRC; EP/S004459/1), and C. Verrill's and N. Khalid-Alham's research time is supported as well as D. Maldonado-Perez is funded by the NIHR Oxford BRC (Molecular Diagnostics Theme/Multimodal Pathology Subtheme).

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.