This research was approved by the Institutional Review Board at our center, and all patients provided written informed consent for donation of biological samples for research use. We collected 45 biospecimens of primary tumor, NAT, and peripheral blood mononuclear cells (PBMC) from 14 treatment naïve patients with renal tumors who underwent partial or full nephrectomy surgery. Routine clinical pathology assessment of the primary tumors revealed 10 ccRCC, one chromophobe RCC, one type I papillary RCC, one oncocytoma, and one sarcoma ( Table 1 ).

Single-cell suspensions were prepared from primary tissues using the gentleMACS™ tissue dissociator and tumor dissociation kit (Miltenyi Biotec, Bergisch Gladbach, Germany) on the “soft tissue” setting. PBMC were processed using Lymphocyte Separation Medium (Corning, Corning, NY) by density gradient centrifugation or SepMate Isolation (StemCell Technologies, Vancouver, Canada). Sections of each tumor specimen were formalin-fixed and paraffin-embedded (FFPE) for subsequent analysis by immunohistochemistry (IHC).

The following antibodies were used to perform mIHC on FFPE samples: anti-human CD4, (clone SP35, Cell Marque, Rocklin, CA, dilution 1:100), anti-human CD3e (clone F7.2.38, Abcam, Eugene, OR, dilution 1:100), anti-human CD8 (clone EP334, Bio SB, Santa Barbara, CA, dilution 1:100), anti-human Carbonic anhydrase IX (CAIX) (clone TH22, Leica Biosystems, Buffalo Grove, IL, dilution 1:25), anti-human Ki67 (clone MIB1, Agilent Technologies, Santa Clara, CA, dilution 1:50). The experimental conditions for CAIX IHC were described previously (14). The staining was performed using the automated OPAL (PerkinElmer, Waltham, MA) workflow on the Leica Bond Rx staining platform (Leica Biosystems, Buffalo Grove, IL) by the Experimental Histopathology Core at our center. Briefly, formalin-fixed paraffin-embedded tissues were sectioned at 4 microns onto positively-charged slides and baked for 1 hour at 60°C. The slides were then dewaxed and stained on a Leica BOND Rx stainer (Leica, Buffalo Grove, IL) using Leica Bond reagents for dewaxing, antigen retrieval and antibody stripping. A high stringency wash was performed after the secondary and tertiary applications using high-salt TBST solution (0.05M Tris, 0.3M NaCl, and 0.1% Tween-20, pH 7.2-7.6). Leica’s PowerVision Poly-HRP anti-Rabbit Detection or Leica’s PowerVision Poly-HRP anti-Mouse Detection species specific polymer was used for all secondary applications.

Antigen retrieval and antibody stripping steps were performed at 100°C with all other steps at ambient temperature. Endogenous peroxidase was blocked with 3% H₂O₂ for 8 minutes followed by protein blocking with TCT buffer (0.05M Tris, 0.15M NaCl, 0.25% Casein, 0.1% Tween 20, pH 7.6 +/- 0.1) for 30 minutes. The first primary antibody was applied for 60 minutes followed by the secondary antibody application for 20 minutes and the application of the tertiary TSA-amplification reagent (OPAL fluor, PerkinElmer) for 20 minutes. The primary and secondary antibodies were stripped with retrieval solution for 20 minutes before repeating the process with the second primary antibody starting with a new application of 3% H₂O₂. The process was repeated until all 5 antibodies were applied. There was no stripping step after the 5^th antibody. Slides were removed from the stainer and stained with Spectral DAPI (PerkinElmer, Waltham, MA) for 5 minutes, rinsed for 5 minutes, and covered with Prolong Gold Antifade reagent (Invitrogen/Life Technologies, Grand Island, NY).

Slides were cured for 24 hours at room temperature, then whole slide images from each slide were acquired on a PerkinElmer Vectra 3.0 Automated Imaging System. Images were spectrally unmixed using PerkinElmer inForm software. Images were then analyzed with HALO image analysis software (Indica Labs, Albuquerque, NM). Individual cells were identified based on nuclear recognition (DAPI stain) and the imaging software then measured fluorescence intensity of the estimated cytoplasmic areas for each cell. A mean intensity threshold above background was used to determine positivity for each fluorochrome within the cytoplasm, thereby defining cells as either positive or negative for each marker. The positive cell data for the CAIX marker was then used to define tumor borders in ccRCC to facilitate nearest neighbor spatial analyses. For each sample, five to ten representative areas with continuous CAIX⁺ or CAIX^- cells from each sample were selected to train a classifier in HALO. The classifier was then applied to samples to define viable tumor areas via a whole slide scan. CAIX expression was visually confirmed within the tumor border after classification.

Genomic DNA was isolated from the tumor, NAT and PBMC single cell suspensions using QIAamp blood mini kits (Qiagen, Hilden, Germany). Targeted TRB-complementarity determining region 3 (TRB-CDR3) libraries were constructed using the ImmunoSEQ hsTCRB v3.0 kit (Adaptive Biotechnologies, Seattle, WA) on 1µg of genomic DNA isolated from the biospecimens according to the manufacturer’s protocol. Briefly, bias-controlled multiplex PCR was applied to amplify all possible rearranged genomic TCRβ sequences using an equimolar pool of the 45 TCR Vβ forward primers, and an equimolar pool of the TCR Jβ reverse primers. The following thermal cycling conditions were used for amplification: 1 cycle at 95°C for 15 minutes, 25 to 40 cycles at 94°C for 30 seconds, 59°C for 30 seconds, and 72°C for 1 minute, followed by 1 cycle at 72°C for 10 minutes (15). The final PCR products at 200bp length were pooled and sequenced at survey level resolution on the Illumina MiSeq platform (v3 150 cycle) in the Genomics Core Facility at the Fred Hutchinson Cancer Research Center as previously described (16, 17).

To prepare samples for scRNAseq, single-cell suspensions were thawed from cryogenic storage and stained with DAPI, 1:20 dilution of APC-Cy7-labeled anti-CD3 mAb (clone SK7; BD Biosciences, San Jose, CA), 1:20 dilution of FITC-labeled anti-CD8 mAb (clone RPA-T8; BD Biosciences), and 1:50 PE-labeled anti-CD45 mAb (clone HI30; BD Biosciences), and sorted by flow cytometry (BD FACSymphony™, BD Biosciences). For targeted scRNAseq, single CD3⁺CD45⁺DAPI^- cells were sorted into individual wells of a 96-well plate, snap frozen, and then thawed to lyse the cells. Cell lysates were reverse transcribed and amplified incorporating primers targeting immune phenotype genes and TRAV and TRBV gene segments ( Supplementary Table 1 ). PCR amplification and barcoding were performed as described (18). Samples were pooled and sequenced using the Illumina MiSeq reagent kit v2, 500-cycles (Illumina, San Diego, CA) in the Genomics Core Facility at our center. For whole transcriptome scRNAseq, sorted CD3⁺CD45⁺DAPI^- cells from single-cell suspension samples derived from three patients were loaded at 17,000 cells per lane onto a 10X Genomics Controller (10X Genomics, Pleasanton, CA). scRNAseq libraries were constructed according to the manufacturer’s protocol. Pooled 3’ V(D)J and 5’ GEX libraries were sequenced on a NovaSEQ SP100 flowcell (Illumina) to obtain 5,000 reads/cell and 20,000 reads/cell depth, respectively.

TRB-CDR3 repertoire analyses were conducted using the LymphoSeq R package (http://bioconductor.org/packages/LymphoSeq). Files containing unique TRB variable region nucleotide sequences with associated CDR3 amino acid sequence, read count, frequency, and VDJ gene segment names were exported from the Adaptive Biotechnologies ImmunoSEQ analyzer v2. Non-productive sequences were removed. Normalized Shannon entropy (the frequencies of all productive sequences divided by the logarithm of the total number of unique productive sequences) value was inverted (1 - normalized entropy) to produce clonality. The similarity score was defined as the number of shared unique sequences between two samples as a fraction of the total number of unique sequences in the two samples. The TRB CDR3 amino acid sequences in each repertoire were compared with the VDJdb database (19) of annotated TRB sequences to identify sequences that have previously been associated with a T-cell response to a specific microbial or tissue antigen or a specific pathological condition. To track T cell clone abundance across different tissues and time, alluvial plots featuring the 120 most frequent TRB-CDR3 sequences in the NAT, tumor, and PBMC repertoires of a patient were generated. The GLIPH2 R package was used to identify “specificity groups” comprising closely related sequences with predicted similar or identical antigenic specificity within the TRB-CDR3 sequence repertoires (20). T cell packing plot were created using igraph package. Outer circles represent GLIPH2 specificity groups comprising TRB-CDR3 amino acid sequences with predicted similar or identical antigenic specificity; inner circles represent unique TRB-CDR3 amino acid sequences. For targeted scRNAseq libraries, outputs were de-multiplexed using customized R scripts and the CDR3 regions with associated V(D)J region information were extracted with MiXCR alignment (21). Targeted phenotype analysis was performed with the Seurat package (22) using the FindAllMarkers function to find differentially expressed genes from each identity cluster against the remaining cells, with min.pct = 0.25, log fc.threshold = 0.25. Pathway analysis was conducted using the clusterProfiler (23) and ReactomePA R packages (24). The 10X Chromium scRNAseq outputs were de-multiplexed and mapped to the human reference genome (hg19, GRCh38) through the Cell Ranger V4.0 bioinformatics pipeline (10X Genomics). Correction for batch effects was performed as described (25). Dimension reduction and clustering of the transcriptome-wide scRNAseq data were conducted using the PAGA algorithm incorporated in the Monocle3 package (26–28). To extract marker genes for each cell module, a “top markers” command was executed on a single-cell object. The cell classification was performed with the Garnett R package (29) based on marker genes derived from bulk RNA-seq datasets (30) that have a false discovery rate < 0.001, as well as gene markers compiled from multiple RNA sequencing references ( Supplementary Table 2 ).

To study the intensity and spatial heterogeneity of T-cell infiltration of renal tumors, we collected renal tumor, NAT, and PBMC from 14 consecutive nephrectomy cases. Pathologic analysis demonstrated 10 ccRCC tumors, and four non-clear cell renal tumors (chromophobe RCC, papillary RCC, oncocytoma, and sarcoma) ( Table 1 ). mIHC with CD3, CD4 and CD8 T-cell markers revealed distinct patterns of T-cell infiltration in the different tumor subtypes ( Figures 1A, B ). We observed significantly higher CD3⁺CD4⁺ (> 8.6-fold) and CD3⁺CD8⁺ (> 22.2-fold) T-cell infiltration into ccRCC compared to the non-ccRCC tumors ( Figure 1C ), consistent with previous reports (31, 32). We used the ccRCC marker CAIX to distinguish regions of viable tumor from necrotic tumor and stroma to characterize T-cell distribution with respect to viable tumor domains. Eight of 10 ccRCC cases were stained strongly positive for CAIX. Five of eight CAIX⁺ tumors had increased CD4⁺ and CD8⁺ T-cell infiltration into the CAIX⁺ areas ( Figures 1D, E top panels), whereas in three tumors, the opposite pattern was observed, with CD4⁺ and CD8⁺ T cells being mostly excluded from the CAIX⁺ tumor regions ( Figures 1D, E bottom panels). Two of the five T-cell infiltrated CAIX⁺ ccRCC tumors showed significant heterogeneity of intra-tumoral T-cell infiltration, with frequent CD8⁺ and CD4⁺ cells observed in some tumor regions with discrete CAIX expression but far fewer such cells in other contiguous CAIX-expressing areas ( Figures 1F–I ). These data are consistent with a recent related study (9), and indicate that immune infiltration in ccRCC at the time of diagnosis is characterized by marked spatial heterogeneity in T cell distribution and density.

To evaluate the TCR diversity and clonal composition of T cells infiltrating into renal tumors, we performed deep sequencing of the hypervariable CDR3 domain of the TRB gene using genomic DNA extracted from T cells purified from renal tumor, NAT, and PBMC from 14 subjects. In total, we identified 389,813 unique TRB-CDR3 sequences, and an average of 2,549 unique sequences (range 25-6,693) per tumor ( Figure 2A ). Analysis of the most frequently observed intra-tumoral TRB-CDR3 sequences, each serving as a marker for an expanded T-cell clone (a clonotype), demonstrated that nine out of 14 patients harbored a single T-cell clone dominating the TIL repertoire and comprising 4.6% to 24% of total TRB-CDR3 sequences ( Supplementary Figure 1A ). The TIL TCR repertoire in an additional patient (patient 1) showed a pattern characterized by several dominant clonotypes expanded at similar frequencies. The TIL TCR repertoire from the cohort of ten ccRCC tumors demonstrated significantly higher clonality than was observed in the four non-ccRCC tumors ( Figure 2B ). The T-cell clonality in ccRCC tumors was also higher than in NAT ( Figure 2C ), identifying clonal expansion as a phenotype of the tumor-associated T-cell repertoire in ccRCC. Moreover, CD8⁺ but not CD4⁺ T-cell tumor infiltration as revealed by IHC ( Figures 2D, E ) correlated significantly with clonality of the TIL repertoire, highlighting the dominance of CD8⁺ T-cell clonal expansion in the ccRCC TME.

In total, 2,759 expanded TCR clonotypes (defined as unique TRB-CDR3 amino acid sequences with a frequency > 1% in TIL repertoires) were discovered. To address the possibility that expanded clones within TIL samples represented “bystander” T cells recognizing non-tumor-associated antigens, we compared the TRB-CDR3 amino acid sequences to the VDJdb database of annotated TRB sequences (19) to identify sequences that have previously been associated with a T-cell response to a specific microbial or tissue antigen or pathological condition. Only 194 unique TRB-CDR3s (0.05%) from the 14 TIL repertoires matched CDR3s from TCRs that have previously been associated with T-cell responses to one of the 13 defined pathogens or public tumor-antigen targets (non-RCC), with individual clonal expansion no greater than 1.3% of the corresponding sample repertoire ( Supplementary Figure 1B ). Clonotypes in TIL repertoires potentially reactive with a specific non-tumor related target account for ≤ 7% of the TIL repertoire in each patient ( Supplementary Figure 1C ). TRB-CDR3s were also screened against a TCR sequence library with greater than 11.7 million unique TRB-CDR3 sequences pooled from PBMC collected from 55 control subjects without cancer as a tool to interrogate shared specificity for non-cancer-associated target antigens (33, 34). 63.9% of TIL-derived, expanded TRB-CDR3s with clonal frequency over 1% were patient-unique and not identifiable in this control sequence library ( Supplementary Figure 1D ). Taken together, these data exclude expanded clones specific for common pathogens as the explanation for the unique clonal architecture of RCC TIL. Instead, expanded T cell clones in ccRCC carry TCRs with primarily private specificities that are not commonly shared with healthy subjects supporting the idea that the clonally expanded T cells may be reacting against RCC-associated targets.

The TCR repertoires of TIL from each tumor had a subset of sequences in common with the repertoires isolated from autologous NAT or PBMC. To quantify the sequence overlap, a similarity score was calculated representing the number of shared unique sequences between two samples normalized to the total number of unique sequences. Most of the subjects with higher sequence similarity between TIL and NAT or PBMC repertoires had “hot” tumors with dense T-cell infiltration in CAIX⁺ tumor regions revealed by IHC, whereas the subjects with low repertoire similarity between the three tissues had “cold” tumors characterized by CAIX⁺ areas with few T cells ( Figure 2F ). The TRB-CDR3 sequences in TIL repertoires were more frequently shared with PBMC than NAT ( Supplementary Figure 2A ). In five ccRCC patients, the TIL repertoire also contained clonotypes expanded with a frequency ≥ 1% that were not identified in either PBMC or NAT. For example, in subject 2, three clonotypes that were detected only in the RCC tumor comprised ~60% of the entire TIL repertoire (blue TRB-CDR3 sequence: CASSLRGTNYGYTF; red TRB-CDR3 sequence: CASGLTGSVEQFF; green TRB-CDR3 sequence: CASSWGAEAFF) ( Figure 2G ). A follow-up PBMC sample was obtained from subject 7 one year after initial nephrectomy. TCR repertoire analysis revealed that multiple expanded TRB-CDR3 clonotypes observed in tumor, NAT, and PBMC at frequencies up to 3% at the time of surgery were maintained in peripheral blood at high frequency (up to 3.8%) one year later ( Supplementary Figure 2B ). Thus, intensely infiltrated tumors exhibited higher TRB-CDR3 amino acid sequence similarity between the tumor, NAT and PBMC. Furthermore, this TIL-derived TCR repertoire of expanded TRB-CDR3 sequences was a variable and possibly stable composite of sequences shared on T cells trafficking in blood or normal tissues and sequences unique to the TME.

We speculated that among expanded TRB-CDR3 sequences found in TIL repertoires, unique TRB-CDR3s may share specificity targeting the same tumor-associated antigens. To explore this possibility, we used the GLIPH2 algorithm (20) to group RCC TIL-derived TRB-CDR3s by their predicted specificity based on CDR3 sequence motifs ( Figure 2H ). When compared with peripheral blood and NAT-derived repertoires, GLIPH2 specificity clustering of TIL repertoires contained significantly fewer unique TRB-CDR3 sequences per specificity group ( Figures 2I, J ), and in general these were comprised of a single or at most 2-3 unique sequences. Thus, GLIPH2 analysis suggested either selectivity or exclusion of T cells entering the TME amongst a larger pool of cells with shared antigen specificity circulating in the peripheral blood and capable of trafficking into NAT.

To further explore shared T-cell specificity for common antigen targets between RCC tumors, GLIPH2 analysis was applied to pooled TIL repertoires selected from subjects with an HLA allele in common. GLIPH2 clustering of pooled TRB-CDR3 sequences from the most frequent HLA-A (HLA-A*02:01, 6 tumors) or -B alleles (HLA-B*35:01, three tumors) among the 10 ccRCC tumors revealed only rare TRB specificity grouping with sequences from more than one patient ( Supplementary Figures 2C, D ) and no clusters with sequences from more than two patients. This outcome is consistent with RCC TIL repertoires being dominated by private antigen specificities.

To investigate the phenotypes and functional capacity of clonally expanded T cells within the TIL repertoires from renal tumors, we implemented a targeted scRNAseq platform with primer pairs designed to amplify the CDR3 regions of the TRA and TRB genes as well as 23 genes defining T-cell phenotypes ( Supplementary Table 1 ). A total of 1,471 single CD3 ⁺ cells were flow-sorted from tumor, NAT, or PBMC samples collected from five patients with ccRCC and one patient with an oncocytoma ( Figure 3A ). A range of 55-151 single CD3⁺ cells were characterized from each of the six tumor samples. Compared with NAT and PBMC, CD3⁺ T cells from TIL demonstrated robust expression of cytokine genes IL2, IL12A, IL10, and IL13; cytotoxicity genes IFNG, GZMB, PRF1, TNFA, and TGFB; the proliferation marker MKI67, the transcription factors GATA3 and TBET; and the activation-induced/exhaustion genes CTLA4 and EOMES, but low-level expression of the transcription factors RORC and RUNX1 ( Supplementary Figure 3A ), suggesting that the frequency of Th17 and CD4 T cells is lower in RCC tumor than in the two the normal tissues.

Despite the limited sampling depth of this targeted scRNAseq platform when compared with our high-throughput, DNA-based TRB repertoire analyses of the same samples, an average of 45.5% (range, 27.4 – 65.3% across six tumors) of the TRA/TRB-CDR3 sequence pairs identified in tumor-infiltrating T cells were observed in two or more cells, representing “expanded” clones ( Figure 3B ). Moreover, expression of functional markers (IFNG, GZMB and PRF1), effector transcription factors (TBET and RUNX1) and activation/exhaustion markers (CTLA4 and EOMES) were upregulated in expanded T-cell clonotypes (defined by TRA/TRB-CDR3 sequence pairs) that were detected multiple times in tumors while T cells that were observed only once were characterized by higher levels of IL2 and RORC expression ( Supplementary Figure 3B ). These observations confirm and extend recent single-cell analyses of RCC-derived TIL that indicate that clonally expanded T cells are enriched for effector T cells with an exhausted phenotype (12).

The most frequently observed expanded TRA/TRB-CDR3 sequence pairs (clonotypes) were detected in cells from two or more tissues ( Figure 3C ). However, many expanded TRA/TRB-CDR3 sequence pairs were only observed in TIL (at up to 14 instances per tumor sample). Compared with expanded T-cell clonotypes that were found in two or more tissues, expanded T-cell clones that were exclusively found in tumor demonstrated elevated expression of the proliferation marker MKI67 ( Figure 3D ) and the antigen-induced marker CTLA4 ( Figure 3E ). In all five ccRCC tumors analyzed by targeted scRNAseq, expression of MKI67 was primarily observed in T-cell clones found exclusively in tumor for every case ( Supplementary Figure 3C ). Thus, despite high expression of CTLA4, indicating late stage of antigen-experienced differentiation, a subset of tumor-associated expanded T cell clones remained capable of continued proliferation and expansion within the TME.

We transitioned our single cell analyses to the 10X Genomics microfluidic partitioning platform to gain higher throughput and deeper sampling of the T cell transcriptomes in tumor, NAT, and PBMC samples than achieved by our targeted scRNAseq platform. Using this platform, we profiled a total of 22,823 CD3 ⁺ T cells (average: 7,607; range: 5,366-8,923) isolated from tissues from three additional subjects ( Figure 4A ) representing a 28-fold increase in throughput per patient, on average, compared with our targeted scRNAseq analyses. We identified 12,450 cells associated with TRA/TRB-CDR3 pairs (average: 4,150; range: 2,757-5,191), an 18-fold per patient increase over our targeted scRNAseq analyses ( Supplementary Figure 4A ). The linear correlation of TRB clonal frequency by whole transcriptome scRNAseq compared with bulk DNA template-based TCR repertoire profiling suggested that the T-cell composition could be accurately sampled from different tissues by whole transcriptome scRNAseq of paired TRA/TRB recombination regions ( Supplementary Figures 4B, C ). It was also possible to accurately associate the expression phenotype and tissue distribution for a given TRB-CDR3 amino acid sequence detected in bulk DNA template-based TCR repertoire sequencing by matching the TRB-CDR3 sequences. Moreover, we compared the number of highly abundant clones that were identified by bulk DNA template-based TCR repertoire profiling with the two scRNAseq platforms. The whole transcriptome scRNAseq platform consistently detected more highly abundant clones than targeted scRNAseq, and the latter platform demonstrated more interpatient variability ( Suppementary Figure 4D ).

Dimensionality reduction and t-distributed stochastic neighbor embedding (tSNE) clustering analysis of the CD3⁺ TRA/TRB- expressing T cells from tumor, NAT, and PBMC revealed distinct clusters containing cells from tumor compared with cells from PBMC, with cells from NAT located primarily within or along the margins of the tumor cluster ( Figure 4B ). Consistent with our observations from targeted scRNAseq, the whole transcriptome scRNAseq data showed that the tumor-infiltrating T cells are phenotypically distinct in comparison to T cells isolated from NAT or peripheral blood. TILs were enriched for CD8⁺ cells expressing high levels of effector/cytokine (GZMB, IFNG, IL2 and IL12A) and repetitive antigen encounter/exhaustion (PDCD1, CTLA4) genes ( Supplementary Figure 5A ).

From the whole transcriptome scRNAseq data analyzed for TRA/TRB-CDR3 pairs, we detected 53.5% CD3⁺ cells with expanded TCR clonotypes (i.e., TRA/TRB-CDR3 pair detected more than one time). We defined large, medium, and small clonotypes (large clones, >20 cells, 50% of total expanded cells; medium clones, 5-20 cells, 25% of total expanded cells; small clones, 2-4 cells with identical TRA/TRB CDR3 pairs, 25% of total expanded cells). Singletons refer to the T-cell clonotypes with a TRA/TRB-CDR3 pair observed only one time ( Figures 4C, D ). The tumor-infiltrating T cells with a larger clonal size were enriched for CD8 expression, whereas CD4 expression was associated with singletons and small clonal size. Regarding the top differentially expressed marker genes, the naïve markers IL-7R and LTB were primarily expressed on singletons whereas the chemokine CCL5, the cytotoxicity regulator CST7, and the cytotoxicity genes NKG7, GZMB and GZMH were more highly expressed on expanded T cells ( Supplementary Figure 5B ).

This observation led us to examine pathways that were differentially expressed in these cells. Indeed, gene expression profiling for markers of T-cell activation, leukocyte transendothelial migration, and T-cell proliferation revealed that the size-based segregation of expanded TIL clonotypes was associated with unique expression profiles ( Figures 4E–G ). Large clones had upregulated canonical T-cell activation pathway genes including MAPK1, ZAP70 and LCK and had decreased expression of the JUN/FOS axis, PIK3R1, ITK, FYN, and the costimulatory molecule CD28 ( Figure 4E ) consistent with antigen driven expansion. The large T-cell clones had upregulated expression of the small GTPases RHOA, RAP1B, RAC1 and RAC2 in the T-cell transendothelial migration pathway, consistent with a “go” signal for enhanced migration (35) ( Figure 4F ). In contrast, the singletons and small clones had increased RHOH expression which has been shown to inhibit chemokine-induced T-cell migration (36). The large clonotypes expressed proliferation genes including FKBP1B, CCND3, P2RX7, while the singletons expressed proliferation cytokines IL23A, TNFRSF4 and TNFRSF14 ( Figure 4G ). Overall, this result suggested that the magnitude of clonal expansion associates with T-cell subtypes with different phenotypes and functional capacity. The T-cell clonotypes with a larger clonal size exhibit a CD8 cytotoxic phenotype, while the clonotypes with smaller clonal size remain naïve or of CD4 origin. Distinct T-cell activation and migration pathways are associated with cytotoxic or naïve T cells with different clone size.

We categorized all unique clonotypes (unique TRB/TRA-CDR3 pair sequences) based on their tissue association into seven possible tissue distribution patterns (tissue specific: tumor, NAT, PBMC; shared by two tissues: tumor-NAT (TN), tumor-PBMC (TP), PBMC-NAT (PN); or shared by all tissues: ALL). We then similarly segregated the expanded clonotypes ( Figures 5A–C ). Among tumor-associated clonotypes (tumor, TN, TP, ALL), 70% of the 565 tumor-associated, expanded T-cell clonotypes were shared with one or more other tissues. Clonotypes shared by all tissues (226/565, 40%) comprised the most frequently observed tissue distribution pattern. The clonotype size differed among tumor-associated TCRs with different patterns of tissue distribution. The TCRs shared by all tissues had the highest average clonal size ( Figure 5D ) reflecting the phenotype of large clonotypes described above. Effector genes such as GZMB, GZMH and KLRG1 were the marker genes in clonotypes shared by all tissues, while the clonotypes present only in tumor showed upregulated T-cell exhaustion genes (HAVCR2, LAG3, CTLA4, TIGIT, PDCD1, and ICOS) and activation genes (SELL and TNFRSF9) ( Figure 5E ).

We then correlated the tissue distribution pattern of expanded T-cell clonotypes with functional T-cell subtype classification ( Figures 5F, G ). Tumor-restricted clonotypes were enriched for a T effector memory (T_em) population (yellow) compared with expanded clonotypes present only in NAT or PBMC. In contrast, T cells with clonotypes shared by all tissues were highly enriched for a T effector (T_eff) population (red). We observed significantly higher levels of exhaustion (ICOS, TIM3, CTLA4, PD1, LAG3, TIGIT) and activation (TNFSR9, CD69, CD25) markers for clonotypes restricted to the tumor ( Figure 5H ). In addition, the tumor-restricted clonotypes expressed lower effector markers (GZMB, PRF1, IFNG) compared with clonotypes shared with other tissues, indicating possible antigen-experience and exhaustion of TCR clonotypes in the TME. Most of the singleton cells were central memory T cell (T_cm), stem-cell like memory T cell (T_scm), and CD4⁺ T cells comprised of Th1, Th2, Th17 and regulatory T cells (T_reg). In general, CD4⁺ T cells tended to have lower clonal size or cell counts, regardless of their phenotype.

Collectively, these data point to the presence of two distinct populations of expanded T cells in tumor: 1), a T_eff population expressing a “go” signal and interstitial migration genes with large clonal size refreshed from T cells in the circulation, and 2), a T_em population with the TCRs only present in tumor, expressing markers of antigen experienced markers and exhaustion genes, and high expression of RHOH as an inhibitory signal for chemokine mediated migration.

We speculated that chemokine signaling plays an important role in controlling T-cell subtype distribution in the TME. T-cell clonotypes restricted to tumor expressed a characteristic set of chemokines and chemokine receptors including CCR1, CCR5 and CXCL5. In contrast, the T-cell clonotypes shared by all tissues expressed a unique group of chemokine ligands and receptors including the pro-inflammatory chemokine receptor CX3CR1 ( Figure 5I ). This result suggests the presence of tumor-specific homing cues that preferentially recruit T cells with a pro-inflammatory phenotype into the TME. When analyzed by T-cell subtypes, chemokine ligand and expression profiles varied widely. For example, the expression of CCR4 was upregulated in Th2 cells, CCR8 was upregulated in Treg cells, and CXCL16 was upregulated in naïve CD4⁺ T cells, suggesting the presence of cell type-specific homing cues ( Supplementary Figure 5C ).

We observed that T-cell clonotypes in TIL were often composed of a range of phenotypic subpopulations, suggesting phenotypic differentiation within the TME ( Figure 5J ). A subset of clonotypes were composed of individual cells mapping to both the CD4 and CD8 lineages. To exclude possible doublets, we filtered out cells with more than one pair of TRA/TRB-CDR3s detected. Globally, we detected 266 CD4⁺CD8⁺ double positive T cells out of 12,039 total cells (2.2%) analyzed with whole transcriptome scRNAseq platform, and 221 (83%) of them were from tumor, representing 4.5% of the 4,844 tumor-derived T cells. Interestingly, Interestingly, all TCR clonotypes that included CD4⁺CD8⁺ double positive T cells were also expressed by additional single positive cells ( Supplementary Figure 6A ). Among 56 tumor-infiltrating T cell clonotypes that contained double positive cells, 26 were restricted to tumor, 20 were shared by all tissues and 10 were shared with NAT. Tumor-derived double positive T cells had a higher rate of clonal expansion (89.6%) compared with overall TIL (70.6%). Further analysis identified CD4 and CD8 co-expression by 10 out of 52 top expanded tumor-derived clonotypes with clonal frequency over 1%. Tumor-associated double positive expanded clonotypes frequently displayed a T_em phenotype and expressed elevated exhaustion markers PDCD1, TIM3, LAG3 and CTLA4 ( Figures 5K, L ) similar to tumor-restricted CD8⁺ expanded clonotypes. The top differentially expressed gene pathways in tumor-infiltrating double positive cells versus single co-receptor expressing cells from shared clonotypes were related to MHC antigen presentation, CD28 stimulation and TCR signaling, interferon signaling, and PD-1 signaling ( Supplementary Figures 6B, C ). Among all 56 unique double positive T cell associated clonotypes, only one TCR sequence had identity with a defined antigen specificity (TRB: CASSPGQGGGYTF; TRA: CAVHNTDKLIF; antigen: M. Tuberculosis) (20, 37). Taken together, our data indicate the phenotype of CD4⁺CD8⁺ double positive T cell clonotypes is consistent with target antigen recognition in the TME (38).

Our analyses of both the targeted and high-throughput whole transcriptome scRNAseq data sets demonstrated that the proliferation marker MKI67 was primarily expressed on expanded, tumor-restricted TCR clonotypes ( Figure 6A ). The MKI67 marker was associated with clonotypes with a T_em phenotype ( Figures 5F , 6B ). These data suggest that T cell proliferation was occurring within the TME. To evaluate the in situ localization of T cells positive for Ki67, we performed mIHC on eight CAIX⁺ ccRCC tumors. This analysis revealed that proliferating Ki67⁺CD3⁺CD8⁺ and Ki67⁺CD3⁺CD4⁺ T cells were in close proximity to CAIX-expressing tumor cells ( Figures 6C, D ). Taken together, these data suggest direct interactions of proliferating, expanded T_em cells with tumor cells within the TME.

We have demonstrated that the TCR repertoire isolated from RCC tumors is primarily composed of private sequences unique to the individual tumor. To screen for rare TCR clonotypes that may be shared across multiple tumors, we compiled a TCR repertoire database including the 14 tumors analyzed in this report pooled with publicly available TCR repertoire datasets representing a total of 53 RCC tumors (16, 39, 40). We then compared all expanded TRB-CDR3 sequences identified from our targeted scRNAseq as well as our high-throughput whole transcriptome scRNAseq analyses against this RCC tumor-derived TCR repertoire database. As seen with the repertoire comparison to normal donors, the majority of TCR clonotypes identified were private to individual tumors (65.8% of the TRB-CDR3 sequences). However, we also identified four TRB-CDR3s with unique TRA pairings that were detected in over 10 cells/tumor and shared by 10 to 14 RCC tumors representing the best examples of putative “public” TCR clonotype sequences ( Figure 7A ). Two of the four TCRs (TRB-CDR3 sequences CASSLQGADYGYTF and CASSPGGDEQFF, both from patient 14) were only found in the tumor tissue with clonotype-bearing T cells adopting a uniform T_em phenotype and nucleotype (unique DNA sequences) across all cells ( Figure 7B ). Two additional TCRs (TRB-CDR3 sequences CASSLVSGELFF, patient 13, and CASSLGRGYGYTF, patient 14) were detected in all three tissues. Single-cell phenotypes showed heterogeneity but were dominated by a T_eff subpopulation in all tissues. Interestingly, the CASSLVSGELFF CDR3 protein sequence originated from two distinct nucleotide sequences, each appearing in more than one cell, suggesting the existence two precursor T-cell clones responding to a common selective pressure that drove convergent expansion of T cells with different clonal origin. Based on DNA-template repertoire analysis, the TRB-CDR3 sequences for these TCRs had a high prevalence among normal donors (56-71% of 55 normal donors) but showed preferential clonal expansion in RCC patients. The average clonal frequency was 1000-fold greater in RCC tumor than normal donors ( Figure 7C ). These data reveal the preferential clonal expansion of TCRs frequently present in healthy individuals that may be responding to RCC antigen encounter.