Deidentified clinical samples were provided from the Biospecimen Repository Core Facility (BRCF) at the University of Kansas Medical Center (KUMC) along with relevant clinical information. Tissue specimens were obtained from patients enrolled under the repository’s institutional review board approved protocol (HSC no: 5929) and following the US Common Rule. Tissue from 74 patients who underwent radical or partial nephrectomy within the University of Kansas Health System for suspected RCC was obtained between August 2017 and January 2020. Online supplemental table 1 provides clinical information and basic research notes on how each tumor tissue sample was used. Tissues were delivered from the operating room to surgical pathology where the BRCF staff obtained a portion of the discarded tissues, deidentified the samples, and immediately transported the specimens at room temperature in RPMI-1640 to the basic research laboratory for TIL and tumor cell line (TCL) generation.

Using aseptic techniques in a laminar flow hood, fibrotic, necrotic, and normal adjacent tissues were grossly dissected from viable tumor. Viable tumor was mechanically and enzymatically digested in a cocktail containing RPMI-1640 (without serum), 1 mg/mL type IV collagenase (Sigma), 100 units/mL deoxyribonuclease I (MP Bio), 1× penicillin, streptomycin, and amphotericin B (Antibiotic-Antimycotic; Gibco). Viable tumor and digestion cocktail were added to a Miltenyi C Tube and disrupted using a gentleMACS Dissociator followed by a 30 min incubation at 37°C with gentle rocking; mechanical disruption and 30 min incubations were continued until a slurry was obtained (approximately 1–1.5 hours). The slurry was passed through a 100 µm cell strainer, pelleted, subjected to ammonium chloride–potassium (ACK)-mediated RBC lysis (BioLegend) and washed in phosphate-buffered saline (PBS). This FTD was either used immediately for culture or cryopreserved for future use. Digests of normal adjacent kidney tissue were performed on selected samples using the same protocol.

Prior to digestion, a portion of the RCC sample and normal adjacent tissue were fixed in 10% neutral buffered formalin at room temperature for 48 hours. Tissues were then dehydrated in 70% ethanol for 24 hours. The samples were then embedded in paraffin and processed by the KUMC BRCF staff using a Tissue-Tek Processor (Sakura). Microtomy was performed on a Leica RM2255 Microtome (Leica Biosystems). Automated staining for pan-cytokeratin and the paired box 8 protein (PAX-8) was performed using an intelliPATH instrument (Biocare). Immunocytochemistry was performed using the same protocol with cell pellets of the primary TCLs established from the tumor sample. Images were acquired using MetaMorph Microscopy Image Analysis Software (Molecular Devices) on a Nikon Eclipse 80i Microscope.

Flow cytometry was performed using a BD LSR II in the KUMC Flow Cytometry Core Laboratory. Cell surface staining was performed in 2% FBS in PBS, and BD Horizon Brilliant Stain Buffer (BD Biosciences) was added when multiple BV dyes were used in the same panel. Cells were subsequently washed in protein-free PBS and stained for viability using Ghost Dye 780 (Tonbo). Cells were resuspended in 2% FBS in PBS for analysis. Transcription factor staining was performed using the True Nuclear Transcription Factor Buffer Set (BioLegend), and cytokine staining was performed using the eBioscience IC Fixation Buffer and Permeabilization Buffer (Invitrogen). The flow cytometry panels used in this study can be found in the online supplemental methods. Analyses were performed using FlowJo (V.10.6.1 and V.10.6.2; FlowJo). Analyses included debris exclusion (Forward Scatter-Area vs Side Scatter-Area (FSC-A vs SSC-A)), single events (Forward Scatter-Area vs Forward Scatter-Height (FSC-A vs FSC-H)), and live cell gates before examining other parameters (online supplemental figure 1).

Expanded TILs rested overnight in the absence of CD3/CD28 stimulation and/or cytokines were either left unstimulated, stimulated with 10 µg/mL of OKT3 or added to autologous tumor cells in T-cell media at an effector:target ratio of 5:1 (5×10⁵ TIL:1×10⁵ tumor cells). The cells were cultured in the presence of 0.7 µL/mL monensin (BD GolgiStop) and anti-CD107a-BV421 in a final volume of 100 µL/condition at 37°C and 5% CO₂ for 4 hours after being briefly centrifuged to promote immediate effector/target contact. Cells were then stained with the remainder of the panel using the indicated antibodies, viability dye, and fix/perm buffer set.

DNA was extracted from primary TCLs which were established from three consecutive RCC samples using the QIAmp DNA Mini Kit. DNA was quantified using a Tecan NanoQuant Plate on a Tecan Infinite M200 Pro and then underwent a multiplex PCR reaction targeting exons of 275 solid tumor-related genes using a commercially available kit (Qiagen). The prepared libraries were then subjected to NGS on a NextSeq 500 instrument (Illumina) to generate fastQ files (text-based format for storing nucleotide sequences). The reads were mapped to the Homo_sapiens_sequence hg19 reference using QIAGEN Biomedical Genomics Workbench pipeline. Metrics for reported variants such as depth of coverage, variant allele frequency, and average quality cut-offs were evaluated individually for pathogenic and likely pathogenic hits compared with uncertain significance variants. The Variant Call Format files were generated and loaded into QIAGEN Clinical Insight to assess individual calls for pathogenicity and quality control parameters.

DNA was extracted from TILs expanded in each condition from three specimens using the QIAmp DNA Mini Kit. Samples were provided to Adaptive Biotechnologies for TCRB CDR3 sequencing. Adaptive Biotechnologies performed quality checks and data normalization using their established pipeline and uploaded the data for analysis using the ImmunoSEQ Analyzer 3.0 platform.

Statistical significance was considered a p-value less than 0.05 determined by using a ratio paired t-test on Graph Pad Prism 8.2.1.

Using our RCC tumor digestion protocol involving gross tumor dissection, mechanical and enzymatic digestion, and ACK RBC lysis, we observe both viable presumptive lymphocytes (green arrows) and viable presumptive tumor cells (red arrows) in the preparation using trypan blue exclusion (figure 1A). We confirmed that there was both a viable CD4+ and CD8+ T-cell infiltrate liberated within tumor digests from the 10 ccRCC samples which pre-expansion phenotypes were analyzed and found the average CD4/CD8 ratio to be 1.24 (figure 1B,C). We also found that 50% of the samples had greater than 10% of double-positive T cells (CD3+/CD4+/CD8+) prior to expansion (figure 1B,C). Double-positive T cells within the tumor microenvironment have been proposed to be a clonal/oligoclonal population of dysfunctional, tumor-specific T cells capable of being reactivated by checkpoint inhibitors; however, the significance of these double-positive T cells outside of the thymus is largely unknown.17 18 We next characterized the differentiation state of pre-expansion TILs and showed that an average of 62% of ccRCC-infiltrating CD8+ T cells have an effector memory (T_em=CD45RA-/CD62L-) phenotype prior to expansion, a subset which is known to have proliferative potential, in contrast to the average 29% with an effector phenotype (T_eff=CD45RA+/CD62L-, AKA T_emra) which have a minimal ability to divide and are prone to apoptosis (figure 1D,E). In summary, we consistently liberated viable CD8+ T_em cells from RCC tumor specimens which should have the capacity for in vitro expansion. Therefore, in subsequent experiments we directly compared PreREP to other methods of TIL expansion to determine if the reported high failure rates for expanding TILs from whole RCC FTDs with IL-2 alone could be overcome with alternative protocols. Figure 1F depicts the schema for the three TIL expansion protocols which were explored.

We hypothesized that inhibitory signals from tumor and/or stromal cell populations within the FTD could be suppressing mitogenic stimulation from tumor antigens and exogenous IL-2 and that this could be responsible for the reported failures for generating TIL cultures from whole RCC FTDs with IL-2 supplementation alone. First, we observed that the adherent cultured cells from the ccRCC samples have a striking morphology which differs from adjacent normal renal cortical tissue (figure 2A). There are prominent vacuoles within the cytoplasm of the adherent cells, likely filled with lipids and glycogen, which give the parent tissue its characteristic “clear cell” morphology. This was our first clue that tumor cell growth dominated the adherent population as opposed to other stromal populations. Also, primary adherent cell cultures from ccRCC uniformly express cytokeratin reflecting the epithelial origin of these tumors as well as the transcription factor PAX-8, which is expressed in the proximal tubules of the kidney that is thought to be the cell of origin of ccRCC (figure 2B). Next, to further characterize the adherent cells, we evaluated gene mutations using a NGS gene panel (275 genes) on DNA isolated from three consecutive adherent ccRCC cultures. We detected pathogenic VHL mutations, the most commonly mutated gene in ccRCC,19 in all three of the adherent primary cultures. Mutant PBRM1 (2 of 3), SETD2 (1 of 3), and TP53 (1 of 3) were also detected in the adherent cell cultures, all with greater than 98% mutant allele frequency, confirming a near pure population of clonal tumor cells grows out after five or more passages (online supplemental table 2). Next, we asked if these adherent tumor cells expressed immunosuppressive ligands in vitro, which could be responsible for failed TIL generation from whole FTDs. We found that adherent cells in culture are able to express programmed cell death ligand 1 (PD-L1), a well-characterized inhibitory immune checkpoint ligand. Two of three samples displayed at least a 10-fold increase in mean fluorescence intensity (MFI) relative to the isotype when cultured in FM; however, in response to autologous stimulated TIL-conditioned media (TIL CM) for 24 hours all three samples massively upregulated PD-L1 with an average 80-fold increase in MFI relative to the isotype control (figure 2C). This is likely due to IFN-ɣ present in the TIL CM (online supplemental figure 2) because PD-L1 has been shown to be an IFN-inducible ligand in many cell types.20 21 Galectin-9 (Gal-9) is a ligand for the T-cell immunoglobulin and mucin-domain containing-3 (TIM-3) receptor and it has been shown to be a negative prognostic factor in ccRCC likely due to the suppression of type 1 T-cell responses by linking the inhibitory TIM-3 receptor to the supramolecular activation cluster on the surface of T cells.22–24 We found that adherent RCC cells during in vitro culture express Gal-9, as shown by the average fivefold increase (n=3) in MFI relative to the isotype control (figure 2D). IFN-ɣ is known to upregulate Gal-9 in certain cell types25; however, we observed a slight decrease in MFI when RCC cells were cultured in TIL CM. This could be due to complex interplay of Gal-9 expression and secretion because Gal-9 is known to be secreted by non-classical pathways involving multivesicular bodies which mediate exosome biogenesis; however, little is known about the regulation of this secretion.26 IFN-ɣ is also known to increase expression of the antigen-presenting major histocompatibility complexes (MHCs/human leukocyte antigen (HLA)). We show that RCC cells uniformly express conventional class I HLA molecules; however, interestingly only two of three RCC cultures upregulated HLA-A/B/C in response to autologous TIL CM and the other showed a slight decrease in expression (online supplemental figure 3). Structurally related to the conventional class I MHC molecules is the non-classical immunosuppressive HLA-G, which under normal physiological conditions is known to be expressed in the placenta and promote maternal immune tolerance to paternal antigens expressed by the developing fetus.27 However, RCC is known to ectopically express this molecule to evade the immune system.28 Although expression levels were considerably lower than conventional class I HLA molecules and the other immunosuppressive ligands evaluated, HLA-G was detectable on the surface of the adherent RCC cells and slightly upregulated in response to TIL CM with an average 1.7-fold increase in MFI relative to the isotype (figure 2E).

In addition to the immunosuppressive properties of the adherent RCC cells themselves, myeloid derived suppressor cells (MDSCs: CD33+/HLA-DR-) are known to impact T-cell expansion and function. Therefore, we examined if monocytic MDSCs (mMDSCs: CD33+/HLA-DR-/CD14+), granulocytic MDSCs (pmnMDSCs: CD33+/HLA-DR-/CD15+), and/or early MDSCs (CD33+/HLA-DR-/CD14-/CD15-) were depleted by removing adherent cells prior to expansion. We found that MDSCs were a rare subset within three RCC FTDs and that CD33+ myeloid cells in general were depleted during the ACD step involving overnight culture in T-cell media (ie, rare in both the adherent and non-adherent population relative to day 0); however, some of the persisting MDSCs were detected in the adherent population indicating that panning does reduce T-cell exposure to this non-desirable cell type during TIL expansion (online supplemental figures 4 and 5 ).

Together, these data suggest that adherent cells left in culture during TIL expansion will express multiple inhibitory ligands that have the potential to counteract the stimulatory signals from exogenous IL-2. This can be visualized by comparing cultures of autologous TILs expanded by panning, PreREP, and FTD+ beads over the first 6 days of culture, in which TILs expanded after ACD rapidly form proliferating clusters of T cells requiring splitting at day 4 and again at day 6, relative to both other protocols in which adherent cells are left in culture and splitting was not indicated in the first 6 days (figure 2F).

Within the validation cohort we observed a 94% success rate (15/16 consecutive samples) for establishing TIL cultures from histologically confirmed ccRCC using our panning protocol and a 0% success rate for establishing TIL cultures from chromophobe (0/1) and papillary (0/2) RCC. Digests of six of the ccRCC samples in the validation cohort contained at least 1.2×10⁷ total viable cells (samples designated by “*” in online supplemental table 1), which was the minimum number of total cells within the FTD needed to compare autologous TIL expansion using all three protocols (4×10⁶ cells/protocol). No more than 6×10⁶ total viable cells were used to initiate TIL cultures using each protocol to improve comparability between samples. Importantly, five of the six samples used to compare protocols were from late-stage (TNM (tumor, node, metastases) stage 3) ccRCC making this cohort clinically relevant because early-stage RCC is often cured by nephrectomy,29 whereas late-stage disease has a poor prognosis and is the main focus of efforts to improve outcomes using immunotherapies.

Successful TIL generation was considered to be a cell culture which required splitting at least four times (1 well of a 24-well plate to an upright 175 cm² flask) using the panning protocol or three times (2 wells of a 24-well plate to an upright 175 cm²) using the PreREP or FTD+ beads protocols in the 14-day time frame. Average TIL yield was 23.6×10⁷ (range=5.5×10⁷–4.8×10⁸) using the panning protocol, 6.6×10⁷ (range=4.6×10⁶–2.5×10⁸) using the PreREP protocol, and 18.1×10⁷ (range=4.5×10⁷–3.7×10⁸) using the FTD+ beads protocol (figure 3A). Using the described criteria for successful TIL generation, of the six samples used for comparing the protocols, 6/6 panning, 3/6 PreREP, and 6/6 FTD+ beads were successful; however, there were enough viable cells in the three failed PreREP cultures to perform immunophenotyping and functional assays—which was the rationale for using a minimum of 4×10⁶ total cells to initiate each culture. An average of 93.2%, 82.6%, and 93.0% of cells were CD3+ after expansion using the panning, PreREP, and FTD+ beads protocols, respectively (figure 3B), and there were no statistically significant differences in the percentage of CD4+ or CD8+ T cells generated by each protocol (figure 3C). However, TILs expanded from two donors via PreREP had greater than 20% double-negative (DN) T cells which parallels Markel et al’s report using the PreREP protocol for generating TILs from RCC,30 but the biological significance of these postexpansion DN TILs is largely unknown.

Significantly fewer CD103+/CD8+ T cells were observed when TILs were expanded by panning relative to both other protocols with an average of 10.6%, 31.8%, and 24.6% of CD8+ T cells expressing CD103 in the TIL products expanded by panning, PreREP, and FTD+ beads, respectively (figure 3D,E). CD103 is a marker of tissue-resident memory T cells (T_RM) and binds to E-cadherin, playing a role in retaining these T cells in epithelial tissues.31 32 Although T_RM cells may play an important role in the immune response to epithelia-derived tumors, we believe they are not optimal for adoptive transfer because they are not known to recirculate33 and therefore are unlikely to contribute to systemic responses after adoptive transfer. Another T-cell subset that we believe is not optimal for adoptive transfer is regulatory T cells (T_regs), which are known to suppress antitumor effector functions. We show that panning produces a significantly smaller proportion of T_regs (T_regs=CD4+/CD25+/FOXP3+) with an average of 5.9%, 18.2%, and 13.7% of CD4+ T cells expressing CD25 and FOXP3 after expansion by panning, PreREP, and FTD+ beads, respectively (figure 3F,G). The high levels of T_regs produced by both other protocols may be a result of simultaneous activation and PD-1/PD-L1 signaling that occurs as a result of leaving adherent cells in the culture. This combination of stimulatory and inhibitory signaling has been implicated in T_reg induction.34 There were no statistically significant differences in the expression of CD28, CD27, KLRG1, CD62L, or IL-7Ra as a result of expansion method; however, a slight (average 2.1% vs 3.8%), but statistically significant increase in CCR7 expression was observed when TILs were expanded by FTD+ beads relative to panning (online supplemental figure 6). Although there is minimal significance when evaluating each of these markers individually, when we evaluate expression of all of these markers simultaneously using t-Distributed Stochastic Neighbor Embedding, we see differential clustering (online supplemental figures 7‒12).

There is strong evidence that coexpression of multiple inhibitory immune checkpoints correlates with a state of decreased function and “exhaustion”35 36; therefore, we also characterized the presence and relative abundance of immune checkpoint molecules on TILs following expansion using each of the three protocols. We report that panning, compared with PreREP and FTD+ beads, generates significantly fewer PD-1+ T cells (an average of 7.2%, 31.2%, and 15.4%, respectively) (figure 4A,B). We show that on average over 75% of RCC TILs expanded in all conditions express TIM-3 (figure 4A,C) with no significant differences in the percentage of TIM-3+ T cells between methods. However, the percentage of PD-1+/TIM-3+ double-positive T cells is also significantly lower in the TIL products expanded by panning relative to PreREP and FTD+ beads with an average of 6.0%, 23.0%, and 12.9% of T cells expressing both PD-1 and TIM-3, respectively (figure 4A,D). We did not observe a significant difference in the percentage of T cells expressing the lymphocyte-activation gene-3 (LAG-3) immune checkpoint (figure 4E); however, we also see a significant decrease in T-cell immunoreceptor with IG and ITIM domains (TIGIT+) T cells when panning is used, as compared with PreREP and FTD+ beads, with an average of 50.8%, 67.9%, and 62.1% of T cells expressing TIGIT, respectively (figure 4F,G). Taken together, our results demonstrate a decrease in coexpression of inhibitory immune checkpoints when TILs are expanded by panning, which may indicate a decreased relative state of exhaustion.

TILs are known to respond to a variety of tumor-associated antigens or tumor-specific neoantigens created by somatic mutations within protein coding genes which are transcribed, translated, degraded and ultimately presented as mutated peptide fragments on MHCs. TILs recognize these peptide fragments/MHCs through their unique T-cell receptors (TCRs). The most diverse region of the TCR is the complementarity-determining region 3 (CDR3) on the beta chain and it is statistically improbable that two different T-cell clones share the same TCRB CDR3 sequence; therefore, this region can be used to identify unique T-cell clones. We preformed TCRB CDR3 sequencing on autologous TILs expanded by panning, PreREP, and FTD+ beads from three donors. Donor 68 was selected because it was the outlier in which TIL yields were nearly identical using all three methods (22.4×10⁷, 24.7×10⁷, and 22.8×10⁷, respectively) and donors 57 and 62 were selected because TILs from these samples fit the trends in yield, phenotype and function which we are reporting. We found that TILs expanded using pan-CD3 stimulation via anti-CD3/CD28 beads (panning and FTD+ beads) results in a TIL product with a greater absolute number of T-cell clones (figure 5A). An average of 13,822, 4279, and 11 381 clones were detected in the TIL products expanded by panning, PreREP and FTD+ beads, respectively, based on productive TCRB CDR3 DNA sequences, which result in a unique amino acid sequence (ie, unique but synonymous DNA sequences were considered identical clones). The top 20 most abundant clones detected in the TIL products expanded by the three methods accounted for an average of 16.5%, 54.8%, and 17.2% of the total TCR repertoires, respectively (figure 5B). In agreement with these trends, the productive Simpson clonality was greatest in the TIL products expanded by PreREP (figure 5C). Finally, we show that although the overall clonal diversity within the TIL products differed, the most abundant clones expanded by each method tend to be identical (figure 5D). This is exemplified by TILs expanded from donor 68 in which gross TIL yields were nearly identical and the frequency of the most abundant clones were also highly comparable.

A potential concern when expanding TILs using immediate pan-CD3 stimulation via anti-CD3/CD28 beads is that the increased clonal diversity of the TILs may be the result of expanding bystander T cells which are not specific for tumor antigens. This could create a dilute TIL product relative to the PreREP protocol. However, on the contrary the increased clonal diversity could be attributed to tumor-reactive T-cell clones which are eliminated as a result of the suppressive signaling that occurs when leaving adherent cells in the TIL culture. Surprisingly, there was not a significant difference in the proportion of TILs which degranulated (CD107a+) after co-culture with autologous tumor cells as a result of the expansion method, although the average percentage was greatest using the PreREP protocol with a mean of 2.2%, 3.8%, and 1.9% of TILs expanded by panning, PreREP, and FTD+ beads degranulating in response to autologous tumor cells, respectively (figure 6A,B). Of note, there were as many as 3.2% of TILs degranulating in the unstimulated condition after PreREP expansion and up to 1.8% using the FTD+ beads protocols. This observation may indicate tumor cells and/or antigen are still in the culture at the 14-day mark using these protocols. Furthermore, over 80% of TILs expanded by all methods expressed granzyme B (presumably within the CD107a+ granules) with no significant differences in GZMB expression (online supplemental figure 13). Both a CD8+ and CD8- response was observed; the CD8- response is a result of CD4+, CD56+, and CD4-/CD56- TIL with no discernable trend in CD8- responses as a result of expansion method (online supplemental figure 14A,B). There was also no significant difference in the percentage of TILs which produced IFN-ɣ in response to autologous tumor cells with an average of 1.5%, 2.0%, and 1.0% of TILs producing this effector molecule after expansion by panning, PreREP, and FTD+ beads, respectively. A similar phenomenon of IFN-γ production in the unstimulated populations, which were not subjected to panning, was noted (figure 6C,D). Interestingly, in response to autologous tumor cells, TILs generated by all protocols were rarely polyfunctional (ie, simultaneously degranulated and produced IFN-ɣ) (figure 6E). As shown, only ~16.5% (0.89/5.39) of the TILs which responded in co-culture with autologous tumor cells produced IFN-γ and degranulated, even though of the cells capable of degranulating and/or producing IFN-γ (OKT3 positive control)~40.8% (17.3/42.4) were polyfunctional. This could be a result of bystander TILs, which are only activated in the positive control, having increased levels of polyfunctionality or it may be that the tumor cells are suppressing a polyfunctional response by an unknown mechanism. These results parallel the report by Markel and colleagues who showed that non-pooled PreREP RCC TIL cultures displayed either lytic activity or IFN-ɣ secretion, but not both.30 The average percentage of tumor-reactive T cells generated—defined as CD107a+ and/or IFN-γ+ when co-cultured with autologous tumor cells—was greatest using the PreREP protocol (figure 6F). An average of 3.4%, 5.3%, and 2.8% of TIL were tumor reactive in the final TIL products expanded by panning, PreREP and FTD+ beads, respectively. However, when we multiplied the percentage of tumor-reactive TIL (figure 6F) by the absolute number of TILs generated in each condition from each sample (figure 3A), we observed an average of 8.0×10⁶, 2.7×10⁶, and 4.6×10⁶ total tumor-reactive TILs generated by panning, PreREP, and FTD+ beads, respectively (figure 6G). These results indicate that although a greater percentage of bystander TIL is potentially being generated when anti-CD3/CD28 beads are added to the initial TIL culture, the combination of ACD and bead-based TIL stimulation generates an increased total number of RCC-reactive TIL in a 14-day time frame.