Phenotypic and functional characterization of tumor-reactive T cells in malignant pleural effusions

Abstract

Adoptive cell therapy using tumor-infiltrating lymphocytes (TIL) is effective for treating advanced melanoma but requires surgical tumor resection. Malignant pleural effusions (MPE) may provide a more accessible source of tumor-reactive T cells. Synchronously collected MPE, lung metastasis, and blood from a patient with metastatic melanoma were analyzed using high-dimensional flow cytometry, and single-cell RNA/T cell receptor (TCR) sequencing. TCR reactivity to autologous tumor was tested in vitro. The proliferative and cytotoxic capacity of ex vivo expanded T cells was assessed in vitro. MPE contained a higher fraction of CD3⁺ T cells compared with tumor and was enriched for effector CD8⁺ T cells and effector memory CD4⁺ T cells. Compared with TIL, MPE T cells exhibited lower features of T cell exhaustion and higher cytotoxicity signatures. The clonal repertoire of MPE and tumor highly overlapped, including 62.2% of predicted neoantigen-specific (NeoTCR) clonotypes. MHC class I-restricted reactivity was functionally confirmed in two of four selected NeoTCR clonotypes. MPE T cells demonstrated higher proliferative capacity under high-dose IL-2 expansion relative to TIL and achieved comparable MHC class I-dependent tumor killing. Overall, MPE contains polyclonal, tumor-reactive T cells with favorable functional features, supporting MPE as an accessible source for TIL therapy.

Introduction

Tumor-infiltrating lymphocyte (TIL) therapy leverages the polyclonal, antitumor activity of autologous T cells and is now an FDA-approved therapy for treating patients with advanced metastatic melanoma (1–5). However, broad adoption of TIL therapy is limited by the need for surgical tumor harvests, which carries procedural risks and delays treatment initiation. Furthermore, TIL products cannot be successfully generated ex vivo in up to 20% of cases (6). A more accessible biological source with potentially greater proliferative potential could therefore decrease the time to treatment and improve the accessibility of TIL therapies.

Malignant pleural effusions (MPE) are a complication of advanced cancers and can be collected through thoracentesis, a common in-office procedure. Prior studies examining the cellular composition of MPE suggest that activated T cells may be present, including a subset of tumor antigen-specific T cells (7,8). However, the frequency and proliferative potential of tumor antigen-specific T cells compared to those in the tumor is unknown.

Herein, we systematically characterize the phenotype and function of MPE T cells relative to those in the tumor and blood from a patient with metastatic cutaneous melanoma.

Materials and Methods

Human specimen collection

All samples were collected under the Yale University Institutional Review Board protocol HIC# 0608001773. The patient provided written informed consent prior to sample collection. All studies were conducted in accordance with the Declaration of Helsinki.

Tissue sample processing

Tumor, MPE and blood at the time of surgical resection and freshly processed per standard protocols. In brief, whole blood was centrifuged to separate plasma, after which the remaining fraction was diluted 1:1 with PBS (Gibco, #14190250) and carefully layered over Ficoll (Cytvia, #17144003). Samples were centrifuged at 400 g for 30 min at room temperature. The PBMC layer was collected, washed twice with PBS, and counted. Pleural fluid samples were centrifuged at 400 g for 10 min, with the supernatant preserved and cell pellets resuspended in PBS supplemented with 2% FBS (Gemini Bio, #S11150). Fresh tumor specimens were minced into small fragments and enzymatically digested with collagenase (Roche, #10103578001) and DNase I (Roche, #4716728001). The digested material was filtered through a 70-μm cell strainer, centrifuged at 400 g for 5 minutes, and washed with PBS containing 2% FBS. PBMC, pleural cells, and tumor digest were counted then cryopreserved in freezing medium (90% FBS, 10% DMSO (Sigma Aldrich, #D8418)) using a controlled-rate freezing protocol and stored at −80°C. Intact tumor chunks were also cryopreserved in freezing medium (used for IL-2 expansion experiments).

Tumor cell line

A patient-derived melanoma cell line was established from surgical tumor tissue in 2025 by culturing in Opti-MEM (Gibco, #31985070) supplemented with 5% FBS and 1% Antibiotic-Antimycotic (Gibco, #15240062) in T75 culture flasks and maintained at 37°C in a humidified 5% CO₂ incubator. The culture medium was replaced on day 2 and subsequently twice per week. To reduce fibroblast overgrowth, first-passage cultures were treated with G418 (Gibco, #10131035) at a final concentration of 50 μg/mL for 3–5 days.

Flow cytometry and cell sorting

Cryopreserved samples were thawed in pre-warmed RPMI with 10% FBS, per 10X Genomics protocols CG000233 and CG000447. Cells were stained for flow cytometry by sequentially staining with fixable live/dead dye, a master mix of fluorophore-conjugated surface antibodies (Table S1) and washed with PBS with 2% FBS. Stained cells were analyzed on a Cytek Aurora using the SpectroFlo software (v3.3.0) or sorted for live, CD45⁺CD3⁺ cells on a Bigfoot Spectral Cell Sorterin the Yale Flow Cytometry Core.

Immunofluorescence microscopy

The wild-type B2M gene was cloned into pLX301 (Addgene, #25895). Self-inactivating minimal HIV-1 virus was produced in 293T cells using the packaging construct psPAX2 (Addgene, #12260) and the envelope plasmid pCMV-VSVG (Addgene, #8454) then transduced into lung-metastatic tumor cells. Tumor HLA class I expression was assessed 36-48 hours post-transduction by staining with PE-conjugated anti-HLA-A/B/C (Table S1) and imaged on a ZOE Fluorescent Cell Imaging System (Bio-Rad).

Jurkat T cell activation assay

TCRα and TCRβ genes were cloned into pLX301. Lentivirus was produced and TCRα/TCRβ genes were co-transduced into TCR knockout Jurkat T cells (9). The wild-type B2M gene was transduced into the tumor cells. Following transduction, tumor cells were treated with 12.5 ng/mL IFNy (R&D Systems, #285-IF-100/CF) for 24 hours. TCR-transduced Jurkat cells were then co-cultured with the tumor cells at a ratio of approximately 4:1 (T:APC). After incubation at 37°C for 24 hours, cells were stained with APC anti-CD69 (Table S1), washed three times with phosphate-buffered saline, and analyzed on a CytoFLEX analyzer.

Ex vivo T cell expansion

Previously cryopreserved MPE and tumor chunks were thawed and resuspended in complete RPMI (Gibco, #11875093) supplemented with 5% human serum (GeminiBio, #100-110-100), 1% penicillin/streptomycin, 1% L-glutamine (Gibco, #25030081), 1% non-essential amino acids (Gbico, #11140050), 1% sodium pyruvate (Gibco, #11360070), and 1% HEPES (Gibco, # 15630080) and recombinant (IL-2 rhIL-2; Peprotech, 20002100UG) 6,000 U/mL, as previously described (10). MPE and tumor cells were plated 10⁶ cells/well or one chunk per well in a 6-well plate and incubated at 37°C for five days. Fresh rhIL-2 (6,000 U/mL) was subsequently replaced every 48-72 hours starting on day five. Cell numbers were quantified through manual cell counts with a hemocytometer, and the CD3⁺ and CD8⁺ fraction of live cells was quantified by flow cytometry.

Cytotoxicity assay

Autologous cancer cells were transduced for GFP or B2M, as above. Cells expanded in HD IL-2 for 21 days were sorted for live, CD45⁺CD3⁺CD8⁺ T cells and added to culture with tumor cells at an Effector:Target 1:1 ratio. Following 20 hours of culture in T cell media, cells were stained for viability and annexin V and analyzed by flow cytometry.

Single-cell RNA-sequencing and TCR repertoire analysis

Single-cell RNA-seq and TCR libraries were generated using 10x Genomics GEM-X 5′ v3 and V(D)J on-chip multiplex assay by the Keck Microarray Shared Resource. cDNA libraries were sequenced on an Illumina NovaSeq 6000 instrument in the Yale Center for Genome Analysis (YCGA). scRNAseq libraries were sequenced at a read length of 28 x 10 x 90bp and a depth of 400 million read pairs. scTCRseq libraries were sequenced at a read length of 150x150bp and a depth of 100 million read pairs. FASTQ files were generated and analyzed with CellRanger (version 7.1.0) using the GRCh38 human reference genome for alignment.

Standard quality control, normalization, and batch correction were performed using open-source Scanpy (11) and scVI-tools (12). Principal component analysis and a neighborhood graph were computed on the scVI latent space, and the Leiden algorithm was applied for community detection. Unsupervised transcriptomic clustering was visualized using Uniform Manifold Approximation and Projection (UMAP). Differential gene expression analysis of Leiden clusters was performed with scVI-tools, and distinct clusters were annotated based on established canonical markers and published gene signatures (Table S2–S3).

Trajectory and functional state analysis

Cell fate probabilities and pseudotime trajectories were inferred using Palantir (v1.4.0) (13). The transcriptional states of T cells across samples were characterized using published gene signature scores for stemness, activation, and exhaustion (14). Tumor-reactive clones were identified by applying published neoantigen-reactive T cell signatures (15,16). Pathway enrichment was assessed with GSEApy (v1.1.8) using Hallmark gene sets from the Molecular Signatures Database (MSigDB_Hallmark_2020) (17).

Tumor antigen reactivity prediction

Single-cell gene set enrichment analysis using cell counts using the R-based package AUCell (v1.24.0) using the NeoTCR signatures, as previously described (15). In brief, predicted neoantigen-specific CD8⁺ and CD4⁺ T cells in the tumor were labeled based on enrichment of the NeoTCR8 and NeoTCR4 gene sets, respectively. The top 5% of tumor T cells were considered neoantigen-reactive. In the blood, the top 20 clonotypes from circulating CD8⁺ T cells enriched for the NeoPBL gene set were considered neoantigen-specific.

TCR repertoire analysis

TCR repertoires were analyzed with Scirpy (v0.22.0) (18). Cells with incomplete or ambiguous chain pairing were excluded. Pairwise distances between TCRs were computed from CDR3 amino acid sequences to quantify similarity, and clonotypes were then grouped based on concordance of the primary chain pair (VJ and VDJ sequences). Clonal expansion was defined as the presence of ≥2 cells sharing the same clonotype, and expanded clones were classified as small (2–4 cells), medium (5–20 cells), or large (>20 cells).

Statistical analysis

All statistical analyses were performed using GraphPad Prism (v10.5.0) and Python (v3.10.16). Flow cytometry data were compared across groups using one-way ANOVA with Tukey’s multiple comparisons test. Differences in gene signature scores between compartments were assessed using Wilcoxon signed-rank testing. All p-values are two-sided, with statistical significance defined as p < 0.05.

Results

MPE is enriched with activated, effector T cells which exist in a pro-inflammatory immune microenvironment

Tumor, MPE, and blood were synchronously collected from a 51-year-old, white non-Hispanic, woman with BRAF wild-type, HLA-A*02:01-positive metastatic cutaneous melanoma. The patient had previously received anti-PD-1 therapy (nivolumab) for two months prior to developing immune-related hypopituitarism, hepatitis, and colitis, requiring prolonged courses of immunosuppressives. She subsequently received a bispecific T cell engager targeting gp100 and CD3 (tebentafusp) and achieved disease stability. Following 16 months on tebentafusp, the patient developed disease progression to the lymph nodes, ribs, lungs, pleura, and histologically-confirmed MPE. She subsequently underwent surgical TIL harvest of a lung metastasis in preparation for standard-of-care TIL therapy (lifileucel). She received lymphodepleting chemotherapy followed by lifileucel infusion but clinically declined over subsequent weeks due to disease progression. She ultimately transitioned to hospice and passed away.

To characterize the cellular composition of immune cells across tissues, spectral flow cytometry was used to assess the expression of 25 markers in MPE, lung metastasis (“Tumor”), and blood (Figure 1A; Figure S1–S4). Overall, MPE had the highest percentage of CD45⁺ immune cells (94.8% in MPE vs. 89.8% in blood vs. 14.7% in tumor; Figure S5) and CD3⁺ T cells (74.9% in MPE vs. 58.6% in tumor vs. 48.7% in blood; Figure 1B). Of CD3⁺ T cells, the tumor had the highest proportion of CD8⁺ T cells (55.8% in tumor vs. 29.9% in MPE vs. 31.6% in blood), which were predominantly composed of CD45RA⁻CCR7⁻ effector memory (Tem) and CD45RA⁺CCR7⁻ terminally differentiated (Temra) subsets (Figure 1C). Tumor CD8⁺ T cells were also most enriched with CD103⁺CD69⁺ resident memory T cells (Trm; 24.6% in tumor vs. 8.4% in MPE vs. 0.3% in blood; Figure 1D) whereas MPE CD8⁺ T cells had the highest proportion of CX3CR1⁺KLRG1⁺ effector T cells (5.6% in MPE vs. 2.9% in tumor vs. 2.8% in blood; Figure 1E). Co-inhibitory receptors PD-1, LAG-3, TIGIT, and TIM-3 was highest in tumor CD8⁺ T cells, consistent with a terminally exhausted state (Figure 1F; Figure S6A).

FIGURE 1. Immunophenotyping of T cells across tissue compartments.

(A) Design and analytical pipeline of multiparametric flow cytometer, 5’ droplet-based scRNA-seq with paired TCR sequencing, and functional co-culture assay to characterize T cells in blood, malignant pleural effusion (MPE), and a lung metastasis (tumor). Created in BioRender. Lu, B. (2026) https://BioRender.com/5yuf9ru. (B) Proportion of immune cell subtypes across tissue compartments as a percentage of CD45⁺ cells. (C) Proportion of CD8⁺ or CD4⁺ CD45RA⁺CCR7⁺ naïve (Tn), CD45RA⁻CCR7⁺ central memory (Tcm), CD45RA⁻CCR7⁻ effector memory (Tem), CD45RA⁺CCR7⁻ terminally differentiated (Temra), and CD4⁺CD25^highCD127^low regulatory (Treg) T cells across tissue compartments as a percentage of CD3⁺ cells. (D) Frequency of CD8⁺CD69⁺CD103⁺ T cells across tissue compartments. (E) Frequency of CD8⁺CX3CR1⁺KLRG1⁺ T cells across tissue compartments. (F) Histograms and mean fluorescence intensity of PD-1, LAG-3, TIGIT, and TIM-3 on CD8⁺ T cells. (G) Frequency of CD25⁺ and (H) CD127⁺ on conventional (Tconv) T cells. (I) Histograms and mean fluorescence intensity of CD4⁺ Tconv across tissue compartments.

Despite a similar proportion of CD4⁺ T cells between MPE and blood, MPE CD4⁺ conventional T cells (Tconv) displayed a more activated and effector phenotype. CD4⁺ Tconv cells in MPE contained the highest proportion of Tem cells (72% in MPE vs. 63% in tumor vs. 20.1% in blood), the highest expression of the activation marker CD25 (14.3% in MPE vs. 9.3% in blood vs. 8.8% in tumor), intermediate expression of the activation marker CD69 (5.61% in MPE vs. 0.5% in blood vs. 57.8% in tumor), and intermediate expression of PD-1, LAG-3, and TIM-3, which is overall consistent with an activated, effector phenotype (Figure 1C, 1G, 1H, 1I; Figure S6B).

While the proportion of CD4⁺CD25^highCD127^low regulatory T cells (Treg) was numerically highest in MPE (6.9% in MPE vs. 3.8% in tumor vs. 4.2% in blood), MPE Tregs possessed lower levels of activation markers (CD69, HLA-DR) and co-inhibitory receptor (PD-1, LAG-3, TIM-3) expression (Figure 1C, Figure S5J–M, Figure S7). Taken together, these findings suggest that Tregs in MPE have a less activated and less suppressive phenotype than tumor Tregs.

Among non-T cell populations, MPE was enriched for CD14⁺CD16⁺ monocytes (59.3% in MPE vs. 27.4% in blood vs. 1.9% in tumor; Figure 1B, Figure S5B). In addition, NK cells within MPE were enriched for CD56^highCD16⁻ NK cells (20.2% in MPE vs. 2.0% in blood vs. 0.5% in tumor; Figure S5C) which are associated with effector cytokine production (19). These results suggest that MPE possesses a unique immunological milieu.

MPE T cells exist in an intermediate transcriptional state between blood and tumor T cells

To define transcriptional features of T cells between tissue compartments, we performed scRNAseq paired with scTCRseq on sorted CD45⁺CD3⁺ T cells. In total, we analyzed 9,819 high-quality T cells, including 3,156 from MPE, 3,553 from tumor, and 3,110 from blood. Overall, thirteen distinct transcriptomic clusters were identified, including seven CD8⁺ T cell clusters, five CD4⁺ T cell clusters, and one proliferating cluster (Figure 2A–B, Figure S8, Table S2–3).

FIGURE 2. Transcriptional profiles and dynamics of T cells across tissues.

(A) UMAP visualization of single-cell transcriptomic data from 9,819 T cells isolated from tumor, malignant pleural effusion and peripheral blood. (B) Heatmap of canonical lineage genes across cell clusters. (C) Stacked bar plot of cluster proportions across tissues. (D) UMAP visualization of T cell clusters by tissue type. (E) Differential gene expression along pseudotime trajectory. (F) Empirical cumulative distribution function (ECDF) of compartment-specific CD8⁺ T cells along the pseudotime trajectory. (G) Violin plot of CD8⁺ T cells per-sample scattered along pseudotime. Between tissue differences in gene signature scores were determined using pairwise Mann-Whitney U tests (p < 0.05). (H) Gene set enrichment analysis (MSigDB Hallmark 2020) of differentially expressed genes between tumor and MPE samples (FDR < 0.05).

Compositional analysis of transcriptional cluster distributions between tissues were overall consistent with the flow cytometry findings: naïve and early activated CD8⁺ and CD4⁺ T cell populations (C1, C2, C9, C10) were most enriched in the blood (Figure 2C–D, Table S3); CD8⁺ NK-like (C5) T cells and CD4⁺ Tem (C12) cells were enriched in MPE (C5: 20.3% of MPE T cells, C12: 33.6% of MPE); and CD8⁺ effector (C3), CD8⁺ exhausted (C4) T cells, and CD4⁺ effector T cells (C11) were enriched in tumor (C3: 19.6% of tumor T cells, C4: 7.3% of tumor T cells, C11: 26.7% of tumor T cells). Proliferating T cells (C8) were rare in blood (0.2%) and most abundant in tumor (3.4%) (Figure 2C–D, Table S3).

To understand the relative transcriptional states along a continuous differentiation trajectory, we next performed pseudotime trajectory analyses of CD8⁺ and CD4⁺ T cells separately (Figure 2E, Figure S9). Dynamic gene expression analysis of CD8⁺ T cells demonstrated a continuum between genes associated with naïve (e.g. CCR7, SELL, TCF7) and activation (GZMK, CCL5, HLA-DRB1) states (Figure 2E). Notably, unsupervised cell ordering showed that T cells segregated by tissue across pseudotime. Blood T cells predominantly occupying earlier pseudotime values and tumor T cells distributed at the terminal end of the dynamic trajectory (Figure 2F–G, Figure S9). Gene set enrichment analysis revealed that both MPE and tumor were enriched for inflammatory pathways indicative of T cell activation (Figure 2H). However, tumor T cells were also enriched for chronic stimulation and exhaustion pathways, including apoptosis and hypoxia (Figure 2H). Overall, these results are consistent with our immunophenotyping analyses suggesting MPE are enriched for activated, effector T cells with lower features of exhaustion relative to tumor.

Predicted tumor-reactive T cells are shared between tumor and MPE

To understand the clonotypic relationship across compartments, we performed scTCRseq and identified 5,717 unique TCR clonotypes. T cells in the tumor exhibited the highest clonality (Figure 3A–C, Figure S10), with expanded clonotypes (n ≥ 2) comprising 36.8% of tumor, 22.2% of MPE, and 5.3% of blood T cells (Figure 3C). Among 196 expanded clonotypes identified in the tumor, 90 clonotypes (45.9%) were also detected in MPE, whereas only 28 clonotypes (14.3%) were detected in blood (Figure 3D–F). Furthermore, clone sizes in MPE were strongly correlated with those in tumor (Figure 3F). Taken together, these data indicate that the MPE T cell repertoire largely reflects the clonal repertoire in tumor.

FIGURE 3. Expanded and tumor-reactive clones are shared across tumor and MPE.

(A) Projection of expanded clonotypes onto the T cell transcriptomic UMAP, categorized by clone size. (B) Alpha diversity, representing the minimum number of unique clonotypes that account for 50% of all cells. (C) Clonal expansion across blood, MPE, and tumor (D) Distribution of the top 25 tumor-enriched clonotypes across compartments. (E) Heatmap of clonotype overlap between blood, MPE, and tumor. (F) Scatterplot of clone sizes of clonotypes found in tumor and MPE. Each point represents a unique clonotype. Dot sizes correspond with the overall number of clones between tissues. (H) Projection of neoantigen-reactive TCR clonotypes onto the UMAP. NeoTCRs were identified using published gene signatures (neoTCR4, neoTCR8, and neoTCRpbl). (I) Cluster composition of NeoTCR8 clonotype shared between tumor and MPE. (J–L) Tissue-specific gene signature scores for exhaustion (J), cytotoxicity (K), and stemness (L) for shared NeoTCR8 clonotypes in tumor and MPE. Between tissue differences were determined by Wilcoxon signed rank testing (p < 0.05).

To identify predicted neoantigen-specific T cells, we applied previously described, experimentally-validated gene signatures to the tumor (NeoTCR8 for CD8⁺ and NeoTCR4 for CD4⁺) (18) and blood (NeoPBL) (16). This approach identified 118 clonotypes across the three neoantigen gene signatures (collectively referred to as NeoTCR) (Figure 3G), including 37 (31.4%) expanded clonotypes (n ≥ 2). Tracing TCRs across tissue compartments, we identified clonally-related NeoTCR T cells in the MPE, including 23/37 expanded NeoTCR clonotypes in tumor (62.2%) (Figure 3H). Notably, NeoTCR4 clonotypes were overall lowly expanded in the tumor, and only eight NeoTCR4 sister clones could be detected in MPE, all of which were singletons (Figure 3H). Overall, 14.9% and 6.0% of tumor and MPE CD8⁺ T cells, respectively, were predicted to be tumor-reactive, compared to only 0.6% of CD8⁺ T cells in the blood.

Given the robust clonal overlap of NeoTCR CD8⁺ T cells in tumor and MPE, we next compared the cluster distribution and transcriptional states of these cells between tissues. In tumor, NeoTCR8 cells were primarily within the Tex and Teff clusters, whereas sister clones in MPE were found within the NK-like cluster (Figure 3I). We also compared the expression of exhaustion, cytotoxicity, and stemness signatures (Figure 3J–3L) (14). NeoTCR8 clonotypes in MPE had higher stemness (p<0.0001) and cytotoxicity scores (p=0.02), while NeoTCR8 in the tumor had significantly higher exhaustion scores (p<0.0001). Notably, NeoTCR8 clonotypes in MPE also had a higher expression of CD39⁻CD69⁻ stem-like signature (p < 0.05; Figure S11A), which has been associated with durable responses with TIL therapy. In contrast, NeoTCR clonotypes in tumor more highly expressed a CD39⁺CD69⁺ terminal differentiation signature associated with poor responses (p<0.001; Figure S11B). These findings suggest that predicted tumor-specific T cells in MPE are in a less terminally exhausted differentiation state than tumor.

Shared NeoTCR8 clonotypes show functional reactivity to autologous cancer cells

We next sought to functionally confirm the tumor reactivity of NeoTCR8 clonotypes. We first established an autologous cancer cell line from the lung metastasis (Figure 4A). We noted a paucity of HLA class I expression on the cancer cell line (Figure 4B) and identified a single-nucleotide variant resulting in a missense mutation in the B2M gene (c.92C>G) (20) on whole-exome sequencing analysis. HLA class I expression was restored by transducing the B2M gene but not HLA-A2 (Figure 4B). We next selected four TCR’s that were predicted to be tumor-reactive based on NeoTCR8 expression, had the highest degree of clonal expansion, and were shared between tumor and MPE (Figure 4C–D). Paired TCRα and TCRβ sequences were cloned into CD8⁺ TCR knockout Jurkat T cells. Two of the four selected TCRs (TCR1, TCR4) resulted in significant increases in CD69 when specifically co-cultured with B2M-transduced, HLA class I-expressing cells but not control or HLA-A2 transduced cancer cells (Figure 4E–F, Figure S12).

FIGURE 4. Functional specificity and reactivity of MPE T cells.

(A) Diagram of ex vivo tumor and TIL expansion and co-culture assay. Created in BioRender. Lu, B. (2026) https://BioRender.com/7j96zw5 (B) Brightfield and immunofluorescence images of cancer cells with or without HLA-A2 or B2M transduction. (C) Dimensionality reduction UMAP plot highlighting CD8⁺ T cells selected for TCR reactivity experiments. (D) Stacked bar plot of TCR abundance and tissue distribution. Representative (E) flow cytometry plots and (F) bar graphs showing CD69 expression. (G) Growth dynamics of MPE and tumor CD3⁺ T cells during ex vivo expansion normalized to baseline (day 0) proportions. (H) Summary bar plot of cancer cell viability following 20-hours of in vitro culture with or without tumor or MPE CD8⁺ T cells. Between-group differences were measured by ordinary one-way ANOVA with Tukey’s multiple comparisons test in (F) and (H) (**p < 0.01, ****p < 0.0001). Error bars represent standard error to the mean for groups where technical replicates were available in (F), (G), and (H).

Ex vivo expanded MPE T cells exhibit cytotoxic capacity against autologous cancer cells

To assess the feasibility of using MPE as a biological source for adoptive cell therapy, we next cultured unselected cells from MPE and TIL in high-dose IL-2 (HD IL-2), similar to conventional TIL expansion methods (Figure 4A) (10) and quantified the growth dynamics of CD3⁺ T cells. Given the loss of B2M in the native tumor, we also tested the effects of adding B2M-transduced autologous cancer cells on day 5 of HD IL-2 expansion to further enrich for tumor-reactive T cells. When cultured with HD IL-2, MPE T cells expanded more readily than tumor T cells (Figure 4G), with further increases in CD8⁺ T cells among groups pulsed with B2M-transduced cancer cells (Figure S13). Notably, tumor T cells only expanded when pulsed with B2M-expressing cancer cells.

To examine how the cytotoxic capacity of expanded MPE T cells compared with tumor T cells, we next co-cultured sorted CD8⁺ T cells with CFSE-labeled, autologous cancer cells transduced with GFP or B2M. Because tumor T cells failed to expand in the native tumor, we examined MPE and tumor CD8⁺ T cells pulsed with B2M-transduced cancer cells during ex vivo. Following 20 hours in co-culture, we observed significantly higher cancer cell killing when B2M-transduced tumor cells were co-cultured with either tumor or MPE CD8⁺ T cells (Figure 4H). Importantly, no differences in cytotoxic capacity were observed between tumor and MPE CD8⁺ T cells. Taken together, these data demonstrate that MPE T cells possess higher proliferative capacity with similar cytotoxicity relative to tumor T cells.

Discussion

The polyclonal nature and enrichment for tumor antigen-specific T cells within the tumor microenvironment is thought to underly the success of TIL therapy (21). However, chronically stimulated T cells undergo progressive dysfunction – termed T cell exhaustion – and lose effector and proliferative capacity (14,22,23). An optimal biological source for TIL therapy therefore balances both tumor antigen-specificity and proliferative capacity.

In this study, MPE was highly enriched for effector T cells with lower exhaustion features compared with tumor T cells and shared highly overlapping clonal repertoire with tumor T cells, including predicted tumor-reactive clonotypes. Furthermore, predicted tumor-reactive clonotypes in MPE were enriched for a stem-like phenotype associated with durable responses to TIL therapy, whereas predicted tumor-reactive clonotypes in tumor were dominated by a terminally differentiated phenotype associated with poorer outcomes (23). Importantly, MPE T cells demonstrated greater proliferative potential, a less exhausted transcriptional state, and similar cytotoxicity following ex vivo expansion compared to TILs.

It is unclear whether the observed differences in MPE and tumor are driven by tissue-specific effects of the local microenvironment or related to the recent thoracenteses the patient underwent in the weeks prior to surgical sampling, which may have promoted migration of a new pool of T cells which have yet to undergo prolonged antigen exposure. Whether these cells would eventually undergo the same degree of terminal exhaustion if left undisturbed is uncertain. However, if the beneficial functional properties of MPE T cells are linked to the timing of their migration, the accessibility of MPE presents an opportunity for repeated sampling for cellular therapy.

Of clinical relevance, the patient did not benefit from TIL therapy. This lack of benefit is likely driven by a loss-of-function B2M mutation in the tumor, which would impair MHC class I surface expression and prevent T cell recognition of tumor cells. Although the precise timing of this alteration cannot be determined, the patient’s prolonged stable disease on tebentafusp — which relies on gp100 presentation in the context of the HLA-A*02:01 MHC class I allele and bispecific T cell engagement — followed by subsequent disease progression raises the possibility that this mutation arose as a mechanism of acquired immune escape prior to TIL harvest.

While our findings are encouraging, validation in larger cohorts and across additional cancer types is essential to confirm their generalizability and functional effectiveness of a TIL product derived from MPE in vivo. Furthermore, we found that antitumor T cells exist in the MPE in an activated transcriptional state, and by tracing clonotypes shared with tumor cells, we infer that they likely recognize the same antigens. However, we cannot exclude the possibility that additional, unrecognized antitumor T cells exist in alternative phenotypic states within the MPE. Additionally, while our functional assays demonstrated tumor-killing capacity, further studies are needed to validate the functional in vivo effectiveness of a TIL product derived from MPE.

In summary, our study demonstrates that MPE harbors tumor-reactive T cells with high cytotoxic tumor-killing capacity, high proliferative potential, and a lower degree of terminal exhaustion relative to tumor. These findings therefore support ongoing efforts to use MPE as a more accessible biological source for cell therapies for patients with pleural metastases (24).

Synopsis.

Here, the authors demonstrate the phenotypic, clonotypic, and functional characteristics of malignant pleural effusion (MPE)-derived T cells compared with synchronously collected tumor and blood-derived T cells, suggesting the use of MPE-derived tumor-reactive T cells for adoptive TIL therapy.

List of abbreviations:

TIL

Tumor-infiltrating lymphocytes

MPE

Malignant pleural effusions

scRNA-seq

Single-cell RNA sequencing

scTCR-seq

T cell receptor sequencing

Data Availability

Raw single-cell RNA, TCR, and whole exome sequencing data files are available through the National Center for Biotechnology Information (NCBI) dbGaP phs004859.v1.p1. Processed data is available through the NCBI Gene Expression Omnibus repository accession GSE342167. All biological materials used in this study are available from the corresponding authors upon reasonable request.