CXCR6 by increasing retention of memory CD8⁺ T cells in the ovarian tumor microenvironment promotes immunosurveillance and control of ovarian cancer

Human studies

Tissue samples were collected from patients undergoing primary tumor debulking surgery for ovarian cancer under a protocol approved by the institutional review board (protocol # I215512). For correlative analysis of immune markers and patient survival, a publicly available ovarian cancer TCGA Firehose legacy 2020 mRNA-seq database (n=307) was used. Out of this, only high-grade (stages IIIA–IV, n=280) patients were used for further analysis (online supplemental figure 1A). Correlation and survival analysis were done in either whole set (n=280) or in a subset of patients stratified for high CD8 levels (top 75 percentile, n=210, was used, while the bottom 25 percentile, n=70, was not included). For survival analysis with CXCR6, a comparison between the top and bottom 25 percentile was done either in whole or patients stratified for high CD8.

Mice studies

Wild type (Wt.) and CXCR6 knockout (KO) C57BL/6 were purchased from The Jackson Laboratory (Bar Harbor, Maine, USA). OT1 PL RAGKO (Mice that expresses TCR for OVA_257-264 peptide and has no endogenous CD4⁺, CD8⁺, T, and B cells due to KO in RAG gene) was a kind gift from Dr Shrikant. OT1 PL RAGKO mice were crossed with CXCR6KO to generate OT1 mice that were deficient for CXCR6. All the aforementioned crossing and expansion were done under breeding protocol 1145M, and experiments involving the mice were performed under protocol 1371M. All animals were maintained under pathogen-free conditions in the Laboratory Animal Shared Resource at Roswell Park Comprehensive Cancer Center (RPCCC). All animal experiments were carried out according to protocol guidelines reviewed and approved by the Institute Animal Care and Use Committee of RPCCC (Buffalo, New York).

Adoptive cell transfer

OT1 CD8⁺ T cells from Wt. or the CXCR6KO mice were purified from spleens using EasySep Mouse CD8⁺ T Cell Isolation Kit (#19853; STEMCELL Technologies, Vancouver, Canada). Isolated CD90.1⁺ OT1 CD8⁺ T cells were injected into CD90.2⁺ C57BL/6 recipient mice at a concentration of 10⁶ cells/100 μL PBS through the retro-orbital route.

Vaccination

Attenuated strain MG1 of Maraba virus27 28 expressing full-length ovalbumin (OVA) was manufactured at McMaster University, their titer determined, shipped on dry ice to RPCCC, and stored at −80°C before use. They were injected at 10⁷ Plaque Forming Units (PFU)() intraperitoneally for use as vaccine/adjuvant to activate OT1 T cells that were adoptively transferred 1 day before to C57BL/six mice.

Tumor model

IE9-mp1, a derivative of widely used ID8 mouse ovarian surface epithelial cell line29 obtained by one passage in mice and genetically engineered to express OVA,30 was the tumor cell line used for in vivo mice study. Cell line was cultured in RPMI-1640 (10–040 CM)+10% fetal bovine serum (35–010-CV) in the presence of 1× penicillin/streptomycin (100 IU/100 μg/mL, 30–002 CI) at 37°C, 5% CO₂ conditions. All reagents used for cell line culture were from Corning, Corning, New York. Mice were injected I.P. with 1×10⁷ cells of IE9-mp1. All the control and treated mice were euthanized when their abdominal circumference is ≈10 cm, which was considered as the experimental endpoint.

Real-time qPCR analysis

RNA from tumor and cells were isolated by RNeasy kit (#74106, Qiagen,). cDNA conversion of isolated RNA was done by using iScript cDNA synthesis kit (#170889, BIO-RAD). cDNA was quantified by using predesigned KiCqStart SYBR Green primers (#KSPQ12012, Sigma Aldrich; sequences of the primers provided in online supplemental table 1) specific for the various human chemokine receptor genes and IQ SYBR green reagent (#170882, BIO-RAD) on CFX96 real-time PCR detection system (BIO-RAD). qPCR data were analyzed on CFX manager V.3.1 (BIO-RAD).

Flow cytometry of human and mice samples

Human tumors were minced into small pieces, suspended in RPMI media, and homogenized into single-cell suspension using Miltenyi gentleMACS dissociater. The single-cell suspension was passed through a 100 µm cell strainer, washed, and spun on Ficoll to remove debris and dead cells. TILs processed previously were further processed for downstream analysis. For flow-based sorting of human Trm and non-Trm cells, processed TILs were stained for live/dead dye (Zombie Aqua fixable viability kit, #423 102 BioLegend) in PBS, followed by staining with BioLegend antibodies specific for CD3 (Pe-Cy7, HIT3a, #300316), CD8 (BV421, RPA-T8, #300106), CD103 (Alexa Flour 488, Ber-ACT8, #350208) and sorted on BD FACSAria II sorter into CD8⁺CD103⁻ (non-Trm) and CD8⁺CD103⁺ (Trm) cells and further processed for RNA analysis. Mice tumors were similarly processed, except for the Ficoll step, and were straightaway stained with flow antibodies. Human tumor samples were stained with live/dead dye in PBS, followed by staining with BioLegend antibodies specific for CD3, CD8, CD103, CCR5 (APC-cy7, J418F1, #359110), and with BD Biosciences antibodies specific for CXCR3 (PE, IC6, #560928), CXCR4 (APC,12G5, #560936), and CXCR6 (BV786, 13B1E5, #743602) in FACS buffer. Mice tissue samples were stained with live/dead dye, followed by antibodies specific for CD8 (FITC, 53–6.7, #553031), CD90.1 (PE, OX7, #554898), CD103 (A700, M290, #565529) from BD biosciences, CXCR6 (BV421, SAO51D1, #151109), CD44 (APC-Cy7, IM7, #103208), CD62L (BV786, MEL-14, #104440), CD69 (PE-Cy7, H1.2F3, #104512), and CXCR3 (BV650, CXCR3-173, #126531) from BioLegend. Stained samples were acquired on the BD LSRII flow cytometry system using BD FACSDiva software. Postrun analyses were done using the FCS Express V.7 Research Edition software (De Novo Software, Pasadena, California, USA).

Confocal microscopy

Human or mouse tissue sections were marked on the edge with ImmEdge pen (H-4000, Vector Labs) and fixed in 4% Paraformaldehyde (PFA) for 15 min. Human and mouse tissues were blocked with 5% human or mouse serum, respectively, for 1 hour at room temperature (RT). This was followed by incubation with fluorophore-conjugated (1:100 dilution) or non-conjugated primary antibodies (1/500 dilution) for 3 hours at RT. Slides were washed four times in 1× PBS. To visualize antigens targeted with non-conjugated primary antibodies, fluorophore-conjugated secondary antibodies (1:1000 dilution, Cell Signaling Technology) were added along with nuclear dye (Hoechst 33422, #H1399; Thermo Fisher Scientific) for 30 min at RT, washed 4 four times in 1× PBS, and sections were mounted with coverslips in prolonged gold mounting medium. Tissues were analyzed in Leica SP8 DMRE spectral confocal microscope, and Image J software was used for postimage analysis. The full details of antibodies and secondary antibodies used for confocal staining of human and mouse tissues are given in online supplemental table 2).

Cytotoxic T lymphocyte induced activated caspase3/7 assay

Wt. OT1 (5×10⁵) and CXCR6KO OT1 were incubated by itself or with 5×10⁴ of OVA expressing IE9-mp1 target cell line or OVA non-expressing ID8 cell in the presence of 4 µM CellEvent Caspase-3/7 Green Detection Reagent (#C10423, Thermo Fisher Scientific) in triplicate for 24 hours. Caspase-3/7 fluorescence is measured on Biotek Synergy HT microplate reader at emission 528 nM. Images of the wells with each condition were taken on ZOE fluorescent cell imager (BIO-RAD).

Statistical analysis

All statistical analyses involved in the study were performed using GraphPad Prism 8 software. The type of statistic used is described in the figure legends, respectively. For all analyses, p<0.05 was considered statistically significant; *, **, ***, and **** indicate p<0.05, p<0.01, p<0.001 and p<0.0001, respectively. All data were expressed as mean±SEM except in some cases, where box and whisker plots were used. In the box and whisker plots, the horizontal line represents the median and the whiskers connect the highest and lowest observations. Log-rank (Mantel-Cox) test was used to calculate the mice survival.

Chemokine receptor CXCR6 is highly expressed on human tumor-resident memory CD8⁺ T cells

To characterize the potential chemokine/chemokine receptor axes involved in the localization of Trm CD8⁺ T cells in the ovarian cancer tumor microenvironment (TME), we profiled the expression of the known chemokine receptors in Trm (CD3⁺CD8⁺CD103⁺) versus non-Trm (CD3⁺CD8⁺CD103⁻) cells obtained from patients with ovarian cancer tumors using Realtime- quantitative polymerase chain reaction (RT-QPCR) and flow cytometry. We used CD103 for identifying Trm cells, as it is considered a canonical marker for Trm cells and predicts favorable outcomes in patients with ovarian cancer.31 32 While mRNA expression of most chemokine receptors demonstrated no clear differences between Trm and non-Trm, Trm cells consistently showed significant upregulation of CXCR6 mRNA compared with non-Trm cells (figure 1A). CCR7 mRNA showed a reverse trend with more expression in non-Trm cells (figure 1A). Flow cytometry analysis further confirmed the high expression of CXCR6 protein in CD8⁺CD103⁺ Trm cells (figure 1B), in agreement with previous reports.33 Although CXCR3 was high in CD103⁺ Trm cells, no significant difference was noted in its expression when compared with non-Trm CD8⁺ T cells. Analysis with confocal microscopy further substantiated the coexpression of CXCR6 and CD103 on CD8⁺ TILs (figure 1C).

TCGA analysis reveals a positive correlation of CXCR6 with CD103 and survival of patients with ovarian cancer

Analysis of the whole database of patients with high-grade serous ovarian cancer (n=280) or patients stratified for high CD8 (n=210) showed a significant correlation between CD103 and CXCR6 mRNA markers (figure 2A). Details of TCGA database stratification are provided in online supplemental figure 1A-C). Furthermore, survival analysis using TCGA data revealed that the CXCR6 marker to be positively associated with increased survival in CD8 high but not in the whole dataset (figure 2B).

Mice treated with adoptive transfer of OT1 and vaccination with OVA-expressing Maraba virus (Mrb-OVA)-treated mice display higher survival

To directly test whether CXCR6 plays a functional role in the localization and efficacy of tumor-resident memory responses, we developed a prophylactic vaccination model using a mouse intraperitoneal ovarian cancer model as previously described.34 35 We used the prophylactic model instead of the therapeutic setting for the following reasons: (1) the continued presence of tumors may promote more of the effector rather than memory responses; (2) response to the tumor may not necessarily be derived from memory CD8⁺ T cells; and (3) the OVA-expressing IE9-mp1 tumor model is highly aggressive, and the cure rate is low,35 which makes priming and challenge with the tumor difficult. This model (figure 3A) uses ACT of naïve OT1 CD8⁺ T cells into C57BL/6 recipient, followed by priming with Mrb-OVA vaccination to activate OT1 cells. We chose OT1 T cells for adoptive transfer to allow for careful monitoring and tracking of T-cell responses as shown in previous studies.13 36 Mrb-OVA was used as the vaccine as it was shown to induce robust antitumor CD8⁺ T-cell responses in our previous study.35 Tumor challenge was done 31–30 days after adoptive transfer of OT1 and vaccination with Mrb-OVA. This is to give enough time for the vaccine-induced effector response to subside and allow memory generation. This increases the probability that antitumor immunity following tumor challenge would come from vaccine-induced memory cells. In this model, the OT1+Mrb OVA combination provides better tumor control than individual treatments with either OT1 or Mrb-OVA (online supplemental figure 2A-C). As shown in figure 3B, the combination of ACT and vaccination with Mrb-OVA improved tumor protection compared with no treatment as indicated by a median survival of 70 days vs 33 days.

CXCR6 is highly expressed in antigen-specific CD8⁺ T cells that reside in the tumor, but not those in circulation

Analysis of endpoint tumors from mice that were treated with OT1+Mrb OVA showed that CXCR6 expression was seen predominantly in tumor-specific CD90.1⁺ OT1 (figure 3C), but less frequently in CD90.1⁻ CD8⁺ T cells. Similarly, analysis of endpoint tumors in Mrb-OVA alone treated mice (online supplemental figure 2) revealed that the endogenous tumor-specific CD8⁺ T cells (marked by OVA tetramer staining) showed high CXCR6 expression compared with non-specific (OVA tetramer negative) T cells (online supplemental figure 3A). When CXCR6 expression was compared between OT1 CD8⁺ T cells in tumor, spleen, and lymph node tissues of OT1+Mrb OVA-treated mice that had reached endpoint, only tumor-infiltrating OT1 T cells showed high CXCR6 expression (figure 3D and F). Analysis of CD103 in CD90.1⁺ TILs revealed high expression by CXCR6⁺ cells, rather than in the CXCR6⁻ subset (figure 3E and F).

CD44 and CD62L marker-based stratification showed that CD90.1⁺ TILs were predominantly of CD44⁺CD62L⁻ effector memory phenotype, whereas those in lymph nodes and spleen were more of central memory type (figure 3F). The observation that Trm cells resembled more of the effector/effector memory than the central memory phenotype is in agreement with previous observations.17 37 38 Comparison of OT1 CD8⁺ T cells prior to transfer (top), 7 days post-transfer in recipient blood (middle) and in endpoint tumors (lower) revealed again that CXCR6 is acquired by T cells only in tumors and marks CD44⁺CD62L⁻CD103⁺ tumor-resident memory cells, but not circulatory cells (online supplemental figure 3B).

Analysis of additional Trm markers, namely, CD69 and CXCR3, revealed that they were found in both circulatory and resident memory CD90.1⁺CD8⁺ T cells (figure 3F and online supplemental figure 3B). This discrepancy between the current study and other studies concerning CD69 as a Trm marker can be attributed to the different tissue microenvironment as suggested by Walsh et al.39 Expression of CXCR3 on both resident memory and circulatory memory cells in murine studies further supported our observations in human Trm. Consistent expression of CXCR3 on both subsets and selective late expression of CXCR6 on resident memory cells suggest that while CXCR3 may direct default peripheral migration of both memory subsets to the tumor, CXCR6 selectively promotes tissue residency of memory cells. The aforementioned observations strongly attest to the validity of CXCR6 as a Trm marker with functional implications in resident memory response to ovarian tumors.

Cells negative for CD45 and EpCAM contribute more to CXCL16 expression than hematopoietic or epithelial cells to help Trm cells localize within the ovarian TME.

As CXCL16 is the main identified ligand for CXCR6,40 we wanted to identify the cells that produce the CXCL16 in the murine ovarian TME. We stained tissues from endpoint tumors from OT1+Mrb-OVA-treated mice with antibodies for mouse CXCL16 and EpCAM (tumor cell marker) and CD45 (immune cells). Confocal microscopy analysis (figure 4A) revealed multiple sources of CXCL16, as revealed by CXCL16 (Green) staining in EpCAM⁺ (gray), CD45⁺ (red), and in the cells that were negative for EpCAM and CD45. However, quantitative analysis revealed that the latter EpCAM⁻CD45⁻ cells were the dominant producers of CXCL16, as they contributed to 72% of CXCL16 expression compared with EpCAM⁺ and CD45⁺ cells that contributed 13% and 14%, respectively (figure 4A). Additional analysis among CD45⁺ cells revealed that F4/80⁺ cells were the major producers of CXCL16, as they contributed about 67% to CXCL16 expression within immune cells (figure 4B). To further substantiate the CXCL16 role in Trm localization, we stained for F4/80, CD103, and CXCL16 markers in endpoint tumor tissues and analyzed them using confocal immunofluorescence microscopy. The analysis revealed that most CD103⁺ T cells were found proximal to CXCL16-positive cells (figure 4C), with most of them within 20 µm distance. The aforementioned observations suggest that multiple sources of cells express CXCL16 in the ovarian TME to support the local accumulation of CXCR6⁺ Trm cells.

KO of CXCR6 in tumor-specific CD8⁺ T cells enhances circulatory but reduces resident memory response in tumors, leading to diminished protection against ovarian cancer

To determine if CXCR6 plays a critical functional role in Trm generation and tumor immunity, CXCR6KO OT1 mice were generated by crossing OT1 PL RAGKO mice with CXCR6KO mice. In vitro testing confirmed no clear phenotypical or functional differences between Wt. and CXCR6KO OT1 T cells, as no differences were noted in marker expression (Online supplemental figure 4A) or their lytic capacity against IE9-mp1 target cells (online supplemental figure 4B). Wt. or CXCR6KO (KO) CD90.1⁺ OT1 cells were adoptively transferred into C57BL/6 recipient mice (CD90.2⁺) on day −31, followed by vaccination with Mrb-OVA at day −30 and tumor implantation on day 0 (figure 5A). Blood analysis on day −23 (7 days after ACT+vaccination) revealed the C57BL/6 recipients that received the adoptive transfer of KO T cells showed significantly higher circulatory T-cell response than those that received Wt. T cells (figure 5B). This was backed by similar findings in the spleen of C57BL/6 recipient mice that received KO T cells, as they again showed a higher percentage of CD90.1⁺ T cells compared with Wt. recipients (figure 5C and online supplemental figure 5). In contrast, quantification of TILs in endpoint tumors revealed that C57BL/6 recipients of CD8⁺ T cells from KO mice had few or no detectable CD90.1⁺ TILs compared with that received from Wt. mice, where CD90.1⁺ TILs were consistently detected (figure 5D). This was also true for CD103⁺ TILs, as they were also at low frequency or undetectable in recipient mice that received KO compared with Wt. OT1 (figure 5E). This difference in transferred T-cell persistence in tumors was reflected by differences in tumor progression (figure 5F) and survival (figure 5G), as C57BL/6 recipients of Wt. CD8⁺ T cells controlled tumors significantly better than KO T-cell recipients. Together, the observations confirm that loss of CXCR6 reduces resident memory responses and the associated protection against ovarian cancer.

Patient consent for publication

Not applicable.