Functionally heterogeneous intratumoral CD4 + CD8 + double-positive T cells can give rise to single-positive T cells
Functionally heterogeneous intratumoral CD4 + CD8 + double-positive T cells can give rise to single-positive T cells
Results Spatial Survey and Transcriptional Atlas Reveals Heterogeneity of DP T Cells in Kidney Cancer Patients. In addition to conventional SP CD4 + or CD8 + T cells, we found that kidney cancer and normal adjacent kidney tissues possess DPs through immunohistochemistry staining of CD4 + and CD8 + in RCC patients (Fig. 1 A and SI Appendix, Fig. S1 A and Table S1), and compared these findings with thymic tissues from healthy individuals. To further characterize the T cells in RCC patients, we performed flow cytometry and single cell sequencing of sorted CD4 + CD8 + DP, CD4 + SP, and CD8 + SP T cells across different tissue compartments (tumor-derived, normal-derived, and peripheral blood) in parallel with peripheral blood mononuclear cells (PBMCs) from healthy donors (Fig. 1 B and SI Appendix, Fig. S1 B). As described previously (25), we confirmed that RCC patients possessed CD4 + CD8 + DP T cells across all matched PBMCs, tumor and adjacent nonmalignant cells (Fig. 1 and SI Appendix, Table S2). While DPs were present at similar frequencies across tissue compartments, these cells were significantly increased in tumor tissues compared to healthy PBMC controls (Fig. 1 C). Tumor samples were also largely composed of conventional CD8 + SP T cells and were relatively depleted of effector FOXP3 - CD4 + SP T cells compared to normal kidney, cancer PBMC, and healthy PBMC controls (Fig. 1 C). Fig. 1. To evaluate T cell differentiation states, we used flow cytometry to assess CCR7 and CD45RA expression, enabling the classification of naïve, central memory, effector memory, and CD45RA-expressing effector cell populations. In tumor and normal tissue, DP T cells were largely skewed toward an effector and effector memory phenotype compared to a higher frequency of the naïve phenotype found in PBMCs, similar to that of single-positive CD4 + Conv and CD8 + T cells (SI Appendix, Fig. S1 C and FACS Panel B in SI Appendix, Table S3). Given the high frequency of DP T cells in RCC patients, we leveraged scRNA-seq to explore the heterogeneity of sorted CD4 + SP, CD8 + SP, and DP T cells. We sequenced and analyzed 105,553 cells containing 39,935 CD4 +, 47,734 CD8 +, and 17,884 DP T cells (26, 27) (Dataset S1) to generate a DP/SP T cell atlas across tissue compartments from RCC patients and healthy donor PBMCs. The CD4 +, CD8 +, and DP T cells sorted from tumor, nonmalignant kidney tissue, and PBMCs formed seven distinct states each with adequate representation from all individual samples (Fig. 1 D and E and SI Appendix, Fig. S2 A). Notably, DP T cells were well represented in all of these distinct states (Fig. 1 E and SI Appendix, Fig. S2 A) revealing phenotypic heterogeneity within this unique T cell subpopulation. Transcriptional profiles of individual T cells allowed the grouping of several canonical T cell states into clusters (Fig. 1 D), which we characterized by cross-labeling these cells against reference gene signatures from published datasets (5, 28, 29) (SI Appendix, Fig. S2 B) and by profiling genes we found to be cluster-specific (SI Appendix, Fig. S2 C and Dataset S5). Clusters identified in the analysis include known T cell states: cytotoxic cells (GZMB, GZMK) expressing PRF1 and granzyme genes; central memory (CM) and naive (NAIVE) cells, which both expressed CCR7 and differed by high versus low expression of ANXA1, respectively; regulatory (IL2RA) cells expressing IL2RA (CD25) and FOXP3; and mucosal-associated invariant T (MAIT) cells expressing SLC4A10 and preferentially displaying the known semi-invariant TCR chain of TRAV1-2 and the combinations of TRAJ-12/20/33 and TRBV-6/20 (SI Appendix, Fig. S3 A). Subclusters of cytotoxic populations, defined as GZMB and GZMK, differed by low vs. high expression of PD-1 (PDCD1), respectively (SI Appendix, Fig. S3 B). Within multiple sorted T cell populations, we identified proliferating (PROLIF) T cells expressing MKI67, which also possessed high levels of PD-1 (SI Appendix, Fig. S3 B). Both cytotoxic and PROLIF states showed high expression of a gene signature associated with terminally exhausted CD8 + T cells (SI Appendix, Fig. S3 C) (30, 31). In addition, GZMK and PROLIF cells showed increasing intratumoral prevalence (SI Appendix, Fig. S3 D), supporting a relationship between PD-1 expression, terminal exhaustion, and increased exposure to cognate antigens within the TME (32). Furthermore, principal component analysis (PCA) of cell compositional profiles from individual samples revealed that the major source of phenotypic variance are mainly associated with tissue location and second, by sorted T cell subpopulations (SI Appendix, Fig. S3 E). Overall, this dataset reveals the phenotypic heterogeneity within the DP T cell population, and some similarities to CD4 + and CD8 + SP T cells. Tumor-Infiltrating DP T Cells Possess Lytic Capacity Against Autologous Tumor Cells. Using our scRNAseq atlas, we compared state-specific enrichment for each sorted T cell population through cell frequencies in each tissue compartment in RCC patients. We observed a significant increase in CD4 + T cells in IL2RA and GZMK cell states and CD8 + T cells in GZMK and PROLIF cell states within the TME (SI Appendix, Fig. S3 F), consistent with previous findings observed in other cancer types (6, 33). In addition, we also observed decreased frequencies of GZMB in CD8 + T cells and NAIVE in CD4 + T cells when comparing tumor to PBMC (SI Appendix, Fig. S3 F). Focusing on the DP T cell population, we observed noticeable differences in their distribution across the multiple tissue compartments collected from RCC patients (Fig. 2 A). Comparable to CD4 + and CD8 + SP T cells, we also noted a significantly higher frequency of the GZMK state for DP T cells localized in the tumor (Fig. 2 A), supporting previous observations of cytotoxic DP T cells (22). Notably, GZMK intratumoral DP T states also expressed high levels of PD-1 (SI Appendix, Fig. S3 B), which could sensitize these cells to PD1-blockade (10). To investigate protein-level expression, we performed flow cytometry on whole digested tumor, normal kidney, and PBMCs from patients with renal cell cancer. Intratumoral DP T cells expressed high levels of cytolytic effector molecules, with granzyme K notably enriched, consistent with our scRNA-seq data (Fig. 2 B). These intratumoral DP T cells spanned a broad continuum of CD4 and CD8 expression, including CD8hiCD4lo and CD4hiCD8lo subsets (SI Appendix, Fig. S1 B). Further stratification of these subsets revealed a fourfold enrichment of granzyme K expression in CD8hiCD4lo cells compared to a more modest bias for granzyme B expression in CD4hiCD8lo cells (SI Appendix, Fig. S4 A), which parallels the expression patterns observed in CD8 + and CD4 + single-positive (SP) T cells in our previous studies (6). Importantly, intratumoral DP T cells had a 20-fold increase in expression of 4-1BB and CD39 than their counterparts in normal kidney or PBMCs (Fig. 2 C), suggesting that these TILs are highly antigen-experienced (34, 35). Nonetheless, surface marker expression varied considerably across patients, highlighting substantial interindividual heterogeneity within the DP compartment. The differential expression of granzyme K (GZMK) and granzyme B (GZMB) in these cytotoxic T cells further suggests that these molecules may contribute to distinct cytokine response profiles, offering insight into the functional diversity of intratumoral T cells (36). To substantiate the functional impact of DP CD4 + CD8 + in RCC tumors, we isolated intratumoral DP TILs and cocultured them with autologous tumor cells using a time-lapse imaging assay to track tumor cell death in real time. We found that DP TILs were cytotoxic and induced a fourfold increase in autologous tumor cell death in tumor cocultures compared with tumors alone (Fig. 2 D). This level of cytotoxicity was on par with CD8 TILs incubated with the same autologous tumor (Fig. 2 D), exhibiting comparable rates of tumor death. In addition, cocultures were preincubated with anti-MHC class I and II antibodies to further interrogate MHC-restricted antigen recognition. The cytotoxic activity of DP TILs was partially dependent on recognition of MHC class I, with moderate attenuation in tumor killing after preincubation with a blocking antibody to pan-MHC class I (Fig. 2 D). A more pronounced inhibition of tumor killing was seen after preincubation with a pan-MHC class II antibody (Fig. 2 D), suggesting higher dependence on MHC II-mediated cytotoxicity. Interestingly, there was no additive effect of pan-MHC class I and II antibodies when used in combination (Fig. 2 D). Given the spectrum of CD4 and CD8 coreceptor expression observed in DP T cells (SI Appendix, Fig. S1 B), it is plausible that this population comprises multiple functionally distinct subpopulations, with subsets preferentially dependent on either MHC class I or II. The absence of an additive effect when both class I and II antibodies may suggest a degree of overlapping or converging mechanisms of recognition. Altogether, these findings confirm that patient-derived DP T cells not only express cytolytic proteins but can functionally mediate autologous tumor cell killing in an MHC class I- and II-dependent manner. DP T Cells Are Clonally Expanded. Given that TCR rearrangement is completed at the DP stage of thymocyte maturation, prior to differentiation to the CD4 + SP or CD8 + SP stage (18, 19, 37), we investigated whether DP TCRs were also expressed in single-positive CD4 + or CD8 + T cells within patients. In parallel with scRNA-seq, we performed scTCR-seq to understand the clonal relatedness between these three distinct T cell populations. To assess clonal relatedness between DP and SP T cells within RCC patients, we restricted the analysis to cell barcodes derived from individual patients (23). Using this approach, we inferred one or more complementarity-determining region 3 (CDR3) regions of alpha- or beta-chains in 40,730 CD4 +, 48,882 CD8 +, and 18,198 DP T cells. In total, we obtained 38,684 T cells (Dataset S3) with complementary matched transcriptome profiles and paired TRA and TRB CDR3 nucleotide sequences. To investigate potential lineage relationships between DP and SP T cells, we classified TCR clonotypes by donor based on their expression within the DP T cell subpopulation (Fig. 3 A). Across individuals, CD4 + and CD8 + SP T cells rarely shared TCR clonotypes with one another, suggesting largely distinct clonal origins for these two populations (SI Appendix, Fig. S5 A and Dataset S4). In contrast, DP T cells frequently harbored TCR clonotypes that overlapped with both CD4 + and CD8 + SP T cells from the same individual. Indeed, 48.7% of T cells with paired scRNA-seq and scTCR-seq data (18,844 out of 38,684 cells; mean 41.2 ± 16.7% across individuals; SI Appendix, Fig. S5 A and Dataset S4) expressed a TCR also found in a DP T cell, consistent with prior observations (22). This trend held across tissue compartments in RCC patients: T cells bearing TCRs linked to DP T cells were particularly enriched in tumor and PBMC samples, comprising 46.4% of T cells in these compartments (12,963 out of 27,921; SI Appendix, Fig. S5 A). Based on these findings, we stratified the T cells into three clonotype-defined groups: DP TCRs (T cells bearing DP TCRs), CD4 + Only TCRs, and CD8 + Only TCRs (Fig. 3 B). Fig. 3. With our schema, we can leverage these distinct clonotypic groups and track specific T cell lineage expression patterns and identify functional capabilities unique to cells expressing specifically DP TCRs (Fig. 3 B). Among T cells bearing DP TCRs in RCC patients, we observed strong enrichment for cytotoxic states, with 42.2 ± 8.1% expressed in GZMB and 41.1 ± 7.0% expressed in GZMK T cells, collectively accounting for >80% of the clonotypic group (SI Appendix, Fig. S5 B). We observe this similar trend in CD8 + Only TCRs, which harbored mainly in GZMB (22.3 ± 2.2%) and GZMK (34.4 ± 8.3%) T cells. In contrast, CD4 + Only TCRs were predominantly expressed in naïve (38.1 ± 13.4%) and central memory (24.3 ± 4.5%) T cells, suggesting divergent functional biases across clonotype-defined groups. Assessment of clonal architecture across tissue compartments revealed greater clonal restriction among DP TCRs relative to CD4+ Only TCRs (SI Appendix, Fig. S5 C). We also observed the same clonal restrictions in healthy individuals’ TCR repertoire from their PBMC, albeit to a lesser degree (SI Appendix, Fig. S5 C). Next, we determined whether clonal expansion of T cells expressing DP TCRs was phenotypically constrained to specific cell states in each tissue compartment from RCC patients. DP TCRs in peripheral PROLIF T cells were highly restricted compared to the other clonotype subgroups (Fig. 3 C). In nonmalignant tissues, we observed a more modest clonal restriction pattern in DP TCRs compared to other clonotype subgroups from GZMB and GZMK T cells, but these differences were not statistically significant (Fig. 3 C). Interestingly, we also observed DP TCRs in intratumoral NAIVE T cells showing a >fivefold clonal restriction compared to CD8 + and CD4 + Only TCR subgroups (Fig. 3 C). Most importantly, DP TCRs in intratumoral cytotoxic GZMK T cells had a >threefold clonal restriction compared to other clonotype subgroups (Fig. 3 C). Intratumoral DP TCR Gene Signature Predicts CPI Response. To assess the relevance of intratumoral DP clonotypes on patient outcomes, we applied differential gene expression analysis on our scRNA-seq data and generated transcriptional programs from intratumoral T cells bearing each clonotype subgroup (DP, CD8 + Only, and CD4 + Only TCRs) (SI Appendix, Fig. S6 A and Table S4). We evaluated these gene signatures to predict clinical response from CPI using bulk RNA-seq data from pretreated locally advanced or metastatic renal cell carcinoma tumors from a phase 2 trial with the following treatment arms: atezolizumab (anti-PD-L1) monotherapy, combination atezolizumab with bevacizumab and sunitinib (4). In advance-staged RCC patients (n = 263) with pretreatment RNA-seq data, our DP TCR gene signature showed greater progression-free survival (PFS) in the anti-PD-L1 treatment arms by hazard ratios of individual genes (Fig. 3 D), compared to CD8 + and CD4 + only TCR gene signatures. The DP TCR gene signature showed significantly improved survival in the combination treatment arm compared to CD4 + and CD8 + only TCR gene signatures (SI Appendix, Fig. S6 B; P < 0.02). Focusing on the phenotypic profile of DP, CD8, and CD4 T cells bearing DP TCRs within the tumor, we found that the majority exhibited a GZMK cytotoxic phenotype (mean = 63.4 ± 20.9%; Fig. 3 E). By flow cytometry, intratumoral DP T cells possess elevated levels of inhibitory receptors PD-1 (PDCD1) and TIM3 comparable to CD8 + T cells (Fig. 3 F), supporting the importance of this underappreciated T cell population in CPI response (10). This elevated expression of inhibitory receptors was particularly enriched within the GZMK phenotype (SI Appendix, Fig. S3 B), reinforcing their potential relevance in checkpoint blockade response. These results highlight the relevance of intratumoral DP clonotypes in association with clinical response to anti-PD-L1 treatment. DP T Cells Undergo Extrathymic Maturation in the TME. To examine whether tumors could represent an extrathymic site of T cell development in vivo, we explored the syngeneic MC38 colorectal tumor mouse model. (SI Appendix, Fig. S7 A). We validated that this model indeed induced DP T cells within the tumor (SI Appendix, Fig. S7 B), with 1.6% of these T cells showing reactivity against the MC38 tumor MHC-I class antigen Adpgk by tetramer staining (SI Appendix, Fig. S7 C and D). To examine the origin of intratumoral DP T cells, we implanted MC38 tumors into syngeneic Rag2-GFP mice and profiled T cells on day 14 (Fig. 4 A and SI Appendix, Fig. S7 E). Rag2 expression is required for TCR recombination, marking the early stages of T cell development and recent thymic emigrants (38, 39). As expected, of the DP T cells found within the thymus, the majority were GFP + (Fig. 4 B and SI Appendix, Fig. S7 F). Surprisingly, we also observed GFP + thymic emigrant DP T cells in tumors and spleens at day 14 (Fig. 4 B and SI Appendix, Fig. S7 F). GFP + DP T cells were highly proliferative compared to GFP - DP T cells across tumor, thymus, and spleen (Fig. 4 C and SI Appendix, Fig. S7 G) and expressed significantly higher levels of CCR7 and CD62L compared to GFP - DPs (Fig. 4 C and SI Appendix, Fig. S7 G). These differences were completely absent between GFP + and GFP - SP TILs (Fig. 4 C) and suggest that tumor infiltrating thymic emigrant DPs are immature relative to tumor-infiltrating SPs. We also observed lower expression of CD69 in GFP - DP TILs compared to GFP + DP TILs (Fig. 4 C). Validating this result in RCC patients, we also observed rare RAG2 expression in DP TILs in RCC patient tissues (Fig. 4 D). Immunostaining of RCC patient compartments also confirmed expression of S1PR1 in RAG2 -positive DP TILs (Fig. 4 D) which is an essential gene mediating clonal eviction of DP T cells (40). Combined with the observation of higher CD69 expression in GFP + DP TILS from our Rag2-GFP mouse model, these results indicate that a subpopulation of DP T cells infiltrating tumors may be clonally evicted immature T cells that may undergo further differentiation. Fig. 4. In Vivo DP-to-SP T Cell Transitioning in the TME. Although prior studies have described the reexpression of CD4 or CD8 coreceptors in SP T cells within tumor-bearing mice (22), we instead focused on whether DP T cells could transition to SP T cells in vivo using adoptive cell transfer (ACT) experiments. Sorted SP CD4 +, CD8 +, or DP T cells from donor CD45.1 + C57BL/6J mice were injected intravenously into CD45.2 + Rag1 KO C57BL/6J mice, both of which harbored MC38 tumors (Fig. 5 A). Four days after transfer, we observed that transferred CD4 + SP T cells mainly remained as CD4 + SP in tumors, with minimal T cells transitioning to DP (Fig. 5 B and SI Appendix, Fig. S8 A – C). We observed that >30% of adoptively transferred CD8 + TILs transitioned into DP T cells. Remarkably, we also found that >45% of adoptively transferred DP T cells transitioned to CD8 + SP T cells within the tumor (Fig. 5 B and SI Appendix, Fig. S8 A – C). Given our previous results suggesting clonally evicted thymic T cells in the TME, these DP T cells may have undergone differentiation through TCR stimulation into self-tolerant mature CD8 + T cells, as supported by prior studies (40). Fig. 5. To determine whether DP-to-SP transitions occur in RCC patients, we leveraged TCR sequences to trace clonal lineage across tissues and T cell subpopulations. We isolated expanded intratumoral TCRs (expressed in >1 cell; excluding TCRs expressed in MAIT T cells) exclusively expressed in either DP or SP T cells and tracked their lineage and phenotype trajectories across tissue compartments in RCC patients. As expected from prior studies showing SP-to-DP transitions within tumors (22), we detected intratumoral SP TCRs in a small subset of mature DP T cells localized in adjacent nonmalignant tissues and peripheral blood (Fig. 5 C). More compellingly, we also found intratumoral DP TCRs in mature SP T cells outside the tumor, with a notable bias toward CD8 + SP identity—suggesting that intratumoral DP T cells can differentiate into SP T cells. To better understand the nature of CD8-to-DP transition, we analyzed previously published single-cell data from CD8 + Pmel + T cells transitioning to CD4 + CD8 + DP cells in B16 tumor-bearing mice (22). DP T cells clustered into phenotypic states shared with their CD8 + precursors which includes naïve/memory-like (Ccr7 and Sell), cytotoxic (Gzmk and Nkg7), and proliferative (Mki67) populations (SI Appendix, Fig. S9 A and B and Dataset S5). Pseudotime trajectory analysis using Monocle (41) revealed distinct branch points aligned with these functional populations (Fig. 5 D and SI Appendix, Fig. S9 C and D). Differential expression analysis across the pseudotime continuum identified upregulation of transcription factors Bcl11a, Spi1 (PU.1), and Mef2c (Fig. 5 E and Dataset S6)—all highly conserved core regulatory genes in T cell development and fate specification (42). Taken together, our findings demonstrate that intratumoral DP T cells retain the capacity to transition extrathymically into SP T cells, underscoring unexpected plasticity in T cell fate within the tumor microenvironment.