In Vitro Inhibition of STAT5 Signaling Limits CD8⁺ T-cell Exhaustion and Improves Outcomes in Adoptive T-cell Therapy

The authors show that attenuation of STAT5 signaling enriches progenitor-like T-cell states while preserving functional competence, suggesting that modulating cytokine/JAK–STAT signaling during ex vivo expansion can improve T cell–based immunotherapies.

Abstract

T-cell exhaustion remains a barrier to adoptive cell therapies. Although γc cytokines are used to expand T cells, their impact on T-cell states remains incompletely defined. Using an in vitro model of repetitive T-cell receptor stimulation, we found that high-dose interleukin 2 (IL2) or IL15 promoted exhausted CD8⁺ T-cell (T_EX) differentiation, with increased inhibitory receptor expression and loss of stem-like features. During chronic lymphocytic choriomeningitis virus clone 13 infection, high IL2Rβ expression marked virus-specific CD8⁺ T cells with features of terminally exhausted T cells, whereas IL2Rβ deficiency favored progenitor-like populations. In T_EX, IL2 and IL15 preferentially induced STAT5 phosphorylation over other signaling pathways. Genetic disruption of STAT5 similarly supported progenitor subsets, implicating STAT5 in the progression toward more differentiated T_EX states under chronic stimulation. We therefore asked whether transient STAT5 attenuation during in vitro T-cell expansion could bias differentiation toward progenitor-like populations without compromising functional competence. Transient JAK3 or STAT5 inhibition enriched TCF1⁺Ly108⁺ progenitor–like cells while preserving cytokine production and degranulation capacity. Following transfer into tumor-bearing mice, cells expanded under STAT5-inhibited conditions mediated superior tumor control and prolonged survival. Likewise, transient STAT5 inhibition preserved memory–progenitor phenotypes in human CD22 chimeric antigen receptor T cells without impairing cytotoxicity. RNA sequencing of STAT5 inhibitor–expanded CD8⁺ T cells confirmed reinforced stem-like transcriptional programs and reduced enrichment of effector and exhausted signatures. Together, these findings identify the IL2Rβ–STAT5 axis as a regulator of CD8⁺ T-cell differentiation and support transient STAT5 modulation during ex vivo expansion to improve therapeutic T-cell products.

Introduction

Adoptive cell therapy (ACT) has demonstrated the potential to induce durable therapeutic responses, including remissions, in patients with advanced malignancies (1, 2). However, long-term clinical benefit remains limited for many recipients, largely due to poor persistence and exhaustion of transferred T cells (2, 3). Although this dysfunctional state may mitigate the risk of immunopathology, it impairs effector functions and memory development, compromising durable antitumor responses and predisposing patients to cancer relapse (3–6). Recent evidence has demonstrated that the presence of stem-like CD8⁺ T cells—characterized by heightened self-renewal potential, enhanced proliferation capacity, and memory-associated features—is critical for sustained antitumor activity, response to immune checkpoint therapies, and long-term therapeutic benefit following ACT (7–9). However, these stem-like cells represent only a minor fraction of expanded T cells in ACT products (10, 11). Hence, there is a growing interest in refining ACT approaches to preferentially expand and maintain stem-like T cells while limiting T-cell exhaustion for optimal therapeutic efficacy (5, 12).

Interleukin 2 (IL2) and IL15 are key cytokines that regulate CD8⁺ T-cell fate through shared receptor subunits (IL2Rβ and γc) and overlapping signaling pathways (13–16). Although both cytokines support CD8⁺ T-cell proliferation and survival, their distinct α-chain usage, receptor affinities, and context-specific STAT signaling confer shared yet distinct functions during infection and cancer (17, 18). IL2 robustly drives T-cell expansion and effector differentiation and is therefore widely used in ACT protocols (19). However, IL2 also accelerates the terminal differentiation of CD8⁺ T cells, limiting their persistence and reducing ACT efficacy (20–23). In contrast, IL15 supports memory CD8⁺ T-cell homeostasis and improves ACT durability and efficacy (24) although, under certain conditions, it may favor effector differentiation over long-term memory (25, 26). Despite advances in understanding cytokine signaling, the mechanisms by which the IL2Rβ axis constrains T-cell stemness and promotes exhaustion remain poorly defined. Moreover, few studies have explored how targeting the IL2Rβ axis could be leveraged to improve ACT outcomes. Here, we investigate whether selective modulation of IL2Rβ signaling can preserve stem-like properties in CD8⁺ T cells and enhance their antitumor activity. Using models of chronic antigenic stimulation, we demonstrate that blockade of JAK3 or STAT5 activity—shared downstream pathways of IL2 and IL15—attenuates exhaustion and enhances T-cell stemness. These results support the use of cytokine signaling modulation strategies in ACT and uncover fundamental principles of T-cell fate regulation in chronic antigenic environments.

Materials and Methods

Ethics approval and consent to participate

Ethical protocols for using human peripheral blood mononuclear cells (PBMC) were approved by the Research Ethics Board (REB) at the Sainte-Justine University Hospital (CHUSJ) and Research Center under study MP-21-2011-312. The study was conducted in accordance with the Declaration of Helsinki. Written informed consent was obtained from all healthy donors.

Mice, viral infection, and adoptive cotransfer

CD45.2⁺ P14 T-cell receptor (TCR) transgenic (Tg) mice, provided by Antonio Freitas (Institut Pasteur), which express a TCR specific for the lymphocytic choriomeningitis virus (LCMV) H-2D^b gp33-41 peptide, were bred in-house and backcrossed onto a recombination-activating gene (RAG)-deficient C57BL/6 (B6; RRID: IMSR_JAX:000664) background, as previously detailed (16). OT-I TCR-Tg mice specific for the H-2K^b–restricted ovalbumin (OVA_257–264; SIINFEKL) epitope (RRID: IMSR_JAX:003831) were purchased from The Jackson Laboratory and maintained in-house on a RAG-deficient B6 background. For the adoptive transfers using the LCMV model, CD45.1⁺ B6 mice were purchased from The Jackson Laboratory (RRID: IMSR_JAX:002014) and bred with P14 TCR-Tg mice to obtain CD45.1⁺ CD45.2⁺ P14⁺IL2Rβ^+/+ mice. RAG2^−/− IL2Rβ-deficient P14 TCR-Tg mice (IL2Rβ^−/−, CD45.2⁺) were obtained by breeding the above P14 mice with RAG2^−/− mice and IL2Rβ^−/− mice from The Jackson Laboratory, as previously described (RRID: IMSR_JAX:002816; refs. 16, 27). For experiments assessing the role of STAT5 signaling, the homozygous CD45.2⁺STAT5^flox/flox mice (Stat5^fl/fl; RRID: MMRRC_JAX:032053) were crossed in-house with P14 TCR-Tg mice and subsequently with R26R-EYFP reporter mice (RRID: IMSR_JAX:006148). All mice were housed in specific pathogen-free facilities at the CHUSJ and Research Center in Montreal, Canada. Experiments were conducted under Canadian Council on Animal Care guidelines and were approved by the Institutional Committee for Animal Care at the CHUSJ and Research Center (#2023-4532, # 2025-7562). Female mice (6–12 weeks old) were used for all experiments.

LCMV Cl.13 was produced in BHK-21 cells (RRID: CVCL_1914) and titrated through plaque assays on Vero cells (RRID: CVCL_0059) as previously described (16, 27). To establish a chronic LCMV infection model, 2 × 10⁶ plaque-forming units of LCMV clone 13 (Cl.13) were injected intravenously into recipient mice. For adoptive cotransfer studies, 5,000 P14 IL2Rβ^+/+ (CD45.1⁺ CD45.2⁺) and 5,000 P14 IL2Rβ^−/− (CD45.2⁺) cells were adoptively cotransferred into sex- and age-matched B6 CD45.1⁺ recipient mice 24 hours prior to Cl.13 infection. For experiments assessing the role of STAT5 signaling, P14 CD8⁺ T cells were isolated from CD45.2⁺Stat5^fl/fl mice or Stat5^wt/wt control mice, and cells were treated ex vivo with TAT-Cre recombinase (EMD Millipore; cat. #SCR508, 1.08 μg/mL) for 45 minutes at 37°C to induce deletion of the floxed Stat5 alleles, followed by washing in complete RPMI (cRPMI; Wisent Bioproducts, cat. #350-400 CL). A total of 1 × 10⁴ treated P14 CD8⁺ T cells were then adoptively transferred intravenously into CD45.1⁺ B6 recipient mice 1 day prior to infection. Spleens were harvested, and transferred P14⁺ cells were analyzed by flow cytometry.

Flow cytometry

Cell suspensions from spleens or in vitro cultures were stained with LIVE/DEAD Aqua (Thermo Fisher Scientific; cat. #L34957) and antibodies against cell surface antigens (Ag) for 30 minutes at 4°C in 1× phosphate-buffered saline (PBS). Subsequently, cells were fixed and permeabilized at 4°C using Cytofix/Cytoperm Fixation/Permeabilization Solution (BD Biosciences; cat. #554722) for 30 minutes, followed by staining with intracellular antibodies in Perm/Wash buffer (BD Biosciences; cat. #554723) for 1 hour at 4°C. H-2Db/gp33-41 biotinylated monomers were purchased from the NIH Tetramer Core Facility and coupled in-house to streptavidin R-phycoerythrin (eBioscience/Thermo Fisher Scientific; cat. #12-4317-87). Anti-mouse antibodies CD45.1 (A20; cat. #553775, RRID: AB_395043), CD8 (53-6.3; cat. # 563795, RRID: AB_2722501), PD-1 (RMP1-30; cat. #748264, RRID: AB_2872692), CD160 (CNX46-3; cat. #562512, RRID: AB_2737632), CD3 (H57-597; cat. #560705, RRID: AB_1727573; cat. #612821, RRID: AB_2870145), LAG-3 (C9B7W; cat. #740959, RRID: AB_2740584), CD44 (IM7; cat. #563058, RRID: AB_2737979), TNFα (MP6-XT22; cat. #554418, RRID: AB_395379), IFNγ (XMG1.2; cat. #557649, RRID: AB_396766), p-STAT1 (pY701; cat. #612597, RRID: AB_399880), p-STAT3 (py705; cat. #557815, RRID: AB_647144), and p-STAT5 (py694s; cat. #612599, RRID: AB_399882) were obtained from BD Biosciences. TIM-3 (RMT3-23; cat. #119721, RRID: AB_2616907), CX3CR1 (SA011F11; cat. #149031, RRID: AB_2565939), Ly108 (330-AJ; cat. #134608, RRID: AB_2188093), 2B4 (M2B4; cat. #133514, RRID: AB_2564341), and CD107 (1D4B; cat. #121631, RRID: AB_2783065) were sourced from BioLegend. IL2 (4A9; cat. #17-7021-82, RRID: AB_469490), CD69 (H1.2F3; cat. #12-0691-81, RRID: AB_465731), TOX (TXRX10; cat. #12-6502-82, RRID: AB_10855034), CD101 (Moushi10; cat. #56-1011-82, RRID: AB_2815246), 2B4 (eBio244F4; cat. #11-2441-85, RRID: AB_657877), CD45.2 (104; cat. #11-0454-85, RRID: AB_465062), granzyme B (NGZB; cat. #46-8898-82, RRID: AB_11219476), and CD16/CD32 (2.4G2; cat. #16-0161-86, RRID: AB_468900) were provided by Invitrogen, and TCF1 (C63D9; cat. #2203S), p-AKT308 (D25E6; cat. #13038S), p-AKT473 (D9E; cat. #4060S), p-ERK (T202; cat. #4370S), p-S6 (D57.2.2E; cat. #4858S), and p-FOXO3 (S253; cat. #9466S) were supplied by Cell Signaling Technology. Data were acquired with an LSRFortessa II with High Throughput Sampler 192 from BD Biosciences. FlowJo software 10.7.1 (RRID: SCR_008520) was used to perform all data analysis. Fluorescence minus one and unstimulated conditions were used to set the gates.

In vitro CD8⁺ T-cell expansion, cytokine stimulation, and small molecule inhibitors

Mouse P14 CD8⁺ T cells were expanded in vitro, as described previously (27). Briefly, 2 × 10⁵ P14⁺ CD8⁺ T cells were cocultured with dendritic cells (1:1) isolated from B6 mice using the EasySep Mouse CD11c Positive Selection Kit II (STEMCELL Technologies; cat. #18780) with a high dose of gp33 peptide (100 ng/mL; GenScript, cat. #RP20257) in cRPMI. On days 2, 4, and 6, P14 CD8⁺ T cells were counted, normalized, and recultured in cRPMI with a lower dose of gp33 peptide (1 ng/mL) either alone or in combination with IL2 (100 ng/mL; PeproTech, cat. #212-12) or IL15 (200 ng/mL; PeproTech, cat. #210-15). To investigate the impact of JAK/STAT inhibitors, tofacitinib (Tofa; 1 μmol/L; Selleck Chemicals, cat. #S5001), JAK3 inhibitor (JAK3I; 3 μmol/L; MilliporeSigma, cat. #420126), or STAT5 inhibitor (STAT5i; 100 μmol/L; MilliporeSigma, cat. #573108) were added on days 4 and 6. On day 7, cells were collected and counted, and their phenotypes were assessed by flow cytometry. The STAT5i used in this study is a chromone-based nicotinoyl hydrazone small molecule inhibitor originally described by Muller and colleagues (28). It inhibits STAT5b SH2-domain binding to phosphotyrosine-containing peptides and was reported to reduce STAT5 tyrosine phosphorylation and downstream transcriptional activity. STAT5i was used at the concentrations indicated in the figure legends.

Phospho-flow analysis

For ex vivo phospho-flow analysis, single-cell suspensions from the spleens of LCMV-infected mice were rested in cRPMI for 4 hours prior to stimulation. Cells were then stimulated with recombinant murine IL2 (100 U/mL) or IL15 (100 U/mL) for 45 minutes at 37°C. For in vitro phospho-flow analysis, cells were harvested on day 6 of culture, rested in cRPMI for 2 hours, and treated with the indicated inhibitors for 2 hours prior to stimulation with IL2 (1 ng/mL), IL15 (20 ng/mL), IFNγ (100 ng/mL; PeproTech, cat. #315-05), or IL21 (100 ng/mL; PeproTech, cat. #210-21) for 15 minutes at 37°C. Following stimulation, cells were fixed with 2% paraformaldehyde in 1× PBS for 30 minutes at room temperature, permeabilized with ice-cold methanol for 30 minutes at 4°C, and stained with phospho-specific antibodies for 1 hour in PBS containing 2% fetal bovine serum (FBS) prior to flow cytometry acquisition.

In vitro killing, cytokine production, and proliferation studies

The M369 plasmid encoding LCMV glycoprotein cDNA, kindly provided by Philippe Bousso, Institut Pasteur, was digested with EcoRI and KpnI. A PCR product encoding the GP33–41 epitope was generated and subcloned into the digested vector (forward: GCCAGAATTCCCACCATGAAGGCTGTCTACAATTTTGCCACCTGTTAAGCGGCCGCTACGTAAATTCCG; reverse: GTGGCAAGCTTATCATCGTG). EL4 cells (provided by Nathalie Labrecque, Institut de recherches cliniques de Montréal, 2018) were nucleofected using the Amaxa 4D-Nucleofector system (Lonza Bioscience) with program VACA-1005. Briefly, 2 × 10⁶ cells were resuspended in a nucleofection solution, mixed with 2 μg of the GP33-expressing plasmid, and nucleofected according to the manufacturer’s instructions. Immediately after nucleofection, cells were transferred into cRPMI medium and incubated at 37°C with 5% CO₂. Puromycin selection (2.5 μg/mL) was initiated 24 hours after nucleofection. EL4 cells were confirmed to be negative for Mycoplasma contamination at the time of cryopreservation using the Mycoplasma PCR Detection Kit (Applied Biological Materials, cat. #G238). Cells were maintained at low passage and used within 3 to 5 passages and 1 week after thawing from frozen stocks. No additional reauthentication was performed because cells were used exclusively from early-passage frozen stocks. For the in vitro killing assay, P14 cells were harvested on day 7 of the culture, washed, and then cocultured with EL4-GP33 target cells [stained with 10 μmol/L carboxyfluorescein succinimidyl ester (CFSE)] and EL4 nontarget cells (stained with 1 μmol/L CFSE; RRID: CVCL_0255) at a 1:1 effector-to-target/nontarget ratio for 3, 6, and 18 hours in cRPMI medium. Specific killing was quantified using LIVE/DEAD Aqua staining and calculated with the following formula: % killing = 100 – (percentage of live target cells/percentage of live nontarget cells) × 100. For intracellular cytokine staining, the in vitro-expanded P14 CD8⁺ T cells were collected on day 7 and restimulated with EL4-gp33 for 6 hours in the presence of BD GolgiPlug Protein Transport Inhibitor (containing brefeldin A; BD Biosciences, cat. #555029). Cells were washed and stained as previously described for flow cytometry. In the proliferation assay, CD8⁺ T cells in the presence or absence of STAT5i were stained with CFSE (10 μmol/L; Thermo Fisher Scientific, cat. #C34554) on day 4 of expansion. Their proliferation was monitored over the following 24, 48, and 72 hours by measuring CFSE dilution via flow cytometry.

Adoptive T-cell transfer in melanoma mouse models

B16.F10 melanoma cells (BCRJ, cat. #0046, RRID: CVCL_0159) expressing the GP33 peptide [B16-GP33; provided by Alain Lamarre, Institut National de la Recherche Scientifique (INRS)–Institut Armand-Frappier, 2019] or OVA peptide (B16-OVA; provided by Nathalie Labrecque, Institut de recherches cliniques de Montréal, 2019) were cultured in Dulbecco’s Modified Eagle Medium (Wisent Bioproducts, cat. #319-005 CL) supplemented with 10% FBS (cat. #16000044), 2 mmol/L L-glutamine (cat. #25030081), and 100 U/mL penicillin–streptomycin (cat. #15140122) and maintained under G418 selection at 0.6 mg/mL (cat. #10131035). All reagents were obtained from Life Technologies/Thermo Fisher Scientific. B16 cell lines were confirmed to be negative for Mycoplasma contamination at the time of cryopreservation using the Mycoplasma PCR Detection Kit (Applied Biological Materials, cat. #G238). Cells were maintained at low passage and used within 3 to 5 passages and 1 week after thawing from frozen stocks. No additional reauthentication was performed because cells were used exclusively from early-passage frozen stocks. GP33 and OVA expression was confirmed by P14- and OT-I–mediated recognition and cytotoxicity, respectively. A total of 0.5 and 1 × 10⁶ P14⁺ CD8⁺ T cells, stimulated with IL2 or IL15 and expanded with or without STAT5i, were harvested on day 7 of culture and subsequently intravenously injected into B6 mice. Tumor cells (2 × 10⁵ cells in 200 μL 1× PBS) were subcutaneously injected 1 day after (B16-OVA model) or 10 days before (B16-gp33 model) T-cell transfer. The survival was monitored daily, and the size of visible tumors was measured every 3 days using calipers to determine the longest and shortest diameters. Tumor volume was calculated using the following formula: (length × width²)/2. An attrition rate of ∼10% to 15% was accounted for to reflect variability in tumor engraftment, and mice were randomly assigned to experimental groups once tumors reached the predefined size. Tumor measurements were performed in a blinded manner to reduce observer bias. Based on pilot longitudinal tumor growth data (α = 0.05, 80% power), we used 9 to 14 mice per group to detect differences in tumor growth kinetics over time.

Sample and library preparation for RNA sequencing

For the analysis of cells expressing high or low levels of the IL2Rβ chain following infection, total RNA was extracted from CD8⁺ Tet⁺ CD11a⁺ IL2Rβ^high or IL2Rβ^low T cells on day 21 after LCMV Cl.13 infection, using the Weis laboratory RNA extraction protocol (29). Briefly, cells were sorted on a FACSAria flow cytometer cell sorter and directly isolated and lysed into TRIzol for 5 minutes at room temperature. Phase separation was achieved by adding chloroform (1:4), followed by centrifugation at 12,000 × g for 15 minutes at 4°C. The aqueous phase containing RNA was transferred to a new tube, precipitated with isopropanol, and centrifuged at 12,000 × g for 10 minutes at 4°C to pellet the RNA. The RNA pellet was washed with 75% ethanol, centrifuged at 7,500 × g for 5 minutes at 4°C, air-dried, and resuspended in RNase-free water. Library preparation was carried out using the TruSeq Stranded mRNA kit according to the manufacturer’s instructions (Illumina). Poly-A RNA sequencing (RNA-seq) was performed on the HiSeq 2500 High Output V4 system. For the analysis of cells expanded in vitro in the presence of cytokines with or without STAT5i, RNA was isolated from naïve or expanded CD8⁺ T cells using the miRNeasy Mini Kit (QIAGEN) following the manufacturer’s instructions. RNA samples were sent to the Institute for Research in Immunology and Cancer Genomics Platform in Montreal, Canada, for quality assessment and RNA-seq. Poly-A mRNA enrichment was conducted using Dynabeads Oligo(dT) (Thermo Fisher Scientific), and library preparation was carried out with the KAPA RNA HyperPrep Kit (Roche) following the manufacturer’s instructions. RNA-seq was performed using a NextSeq High Output (150 cycles) kit (PE75) on the NextSeq500 (Illumina).

RNA-seq analysis

All RNA-seq FASTQ files were aligned and mapped to the mouse genome (GENCODE M6/Ensembl v81 GRCm38 for the in vivo IL2Rβ^high/low experiment; GENCODE 31/Ensembl v97 GRCm38/mm10 for the in vitro IL2 and IL15, with or without STAT5i, experiments) using the STAR aligner (v2.4.2 for IL2Rβ and v2.7 for IL2 and IL15; RRID: SCR_004463; ref. 30) with two-pass mapping mode and a regenerated genome index. For the IL2Rβ^high/low analysis, reads were trimmed using Cutadapt (RRID: SCR_011841; ref. 31) to remove adapter sequences (GATCGGAAGAGCACACGTCTGAACTCCAGTCAC) and low-quality read ends (quality cutoff: 20) before alignment. The raw gene count matrices were generated using HTSeq (v0.6.1p1 for IL2Rβ^high/low and v0.13.5 for IL2 or IL15 with or without STAT5i; RRID: SCR_005514; ref. 32) in nonstranded mode. Raw counts were filtered to remove low-expressed transcripts, retaining only protein-coding genes. For IL2Rβ^high/low, additional filtering excluded RIKEN genes, prioritizing genes with clear one-to-one orthologous relationships to human genes. Differential expression analysis was conducted using DESeq (v1.26, RRID: SCR_000154; ref. 33) in both cases, with its internal filtering step to detect and exclude outliers. Significance thresholds were defined as log₂ fold change (FC) ≥ 0.5 and false discovery rate (FDR) < 5%, corrected by the Benjamini–Hochberg procedure.

Raw gene count matrices from the public datasets GSE149876 (34, 35), GSE83978 (36, 37), and GSE123235 (38, 39), which profile different subsets of progenitor and terminally exhausted CD8⁺ T cells (T_EX^term) during chronic LCMV Cl.13 infection, were downloaded from the Gene Expression Omnibus (GEO; RRID: SCR_005012) and processed using the same pipeline as described for the IL2 and IL15 with or without STAT5i experiments. For GSE149876 (Beltra and colleagues, day 30 after infection; ref. 34), subsets included progenitor exhausted T cells (T_EX^prog1; Ly108⁺CD69⁺ and T_EX^prog2; Ly108⁺CD69⁻), intermediate exhausted T cells (T_EX^int; Ly108⁻CD69⁻), and terminally exhausted T cells (T_EX^term; Ly108⁻CD69⁺). In GSE83978 (Utzschneider and colleagues, day 28 after infection; ref. 36), subsets were defined as exhausted phenotype (Tcf1⁻) and nonexhausted phenotype (Tcf1⁺). For GSE123235 (Miller and colleagues, day 30 after infection; ref. 38), progenitor exhausted cells were identified as Slamf6⁺Tim-3⁻ and terminally exhausted cells as Slamf6⁻Tim-3⁺. In the gene set enrichment analysis (GSEA), the top 250 upregulated and downregulated genes from the T-cell exhaustion signatures were mapped to the IL2Rβ^high/low and IL2/IL15 transcriptomic profiles. GSEA was additionally performed using the T-cell exhaustion gene signature from Bengsch and colleagues (40) through the integration of transcriptomic data from GSE41867 (41) with Assay for Transposase-Accessible Chromatin using sequencing (ATAC-seq) data from GSE86881 (42) and GSE97646 (43). The upregulated and downregulated genes were obtained from Bengsch and colleagues’ (40) supplementary table (NIHMS969093-supplement-1.xls), representing an exhaustion-specific gene list (murine and human orthologs; ref. 40). These gene lists were refined based on chromatin accessibility patterns identified in the ATAC-seq datasets. GSEA of the differential expression profile for IL2 + STAT5i (IL2S5i) versus IL2 was performed using the significantly upregulated and downregulated gene sets identified in STAT5A_Y694F knock-in (KI) versus wild-type (WT) and STAT5B_Y699F KI versus WT cells after 48 hours of IL2 stimulation, as reported in Supplementary Data 2 of Lin and colleagues (44). All analyses were performed using custom R scripts along with additional packages [ggplot2 (RRID: SCR_014601; ref. 45), pheatmap, ggpubr, and fgsea]. Genes associated with T-cell exhaustion were categorized into pathways based on classifications from the Universal Protein Resource (RRID: SCR_002380; ref. 46)

Human chimeric antigen receptor T-cell culture and cytotoxicity assay

The use of human blood samples was conducted in accordance with ethical standards and approved by the REB of CHUSJ (protocol MP-21-2011-312). PBMCs were isolated from fresh blood from three independent healthy donors on the day of PBMC isolation and processed fresh by Ficoll-Paque density gradient centrifugation. CD3⁺CD8⁺ T cells were purified through negative magnetic selection using a Human CD8⁺ T Cell Enrichment Kit (Miltenyi Biotec, cat. #130-096-495). Purified CD8⁺ T cells were stimulated with anti-Human T-Activator CD3/CD28 Dynabeads at a bead-to-cell ratio of 1:1 (Gibco/Thermo Fisher Scientific, cat. #11161D) in cRPMI. Two days later, the cells were transduced with a second-generation anti-CD22 chimeric antigen receptor (CAR) 28z, which includes the m971 ScFv followed by the CD28 and CD3ζ signaling domains, as previously described (37, 47). CAR expression was evaluated after 72 hours using flow cytometry. Subsequently, T cells were maintained in fresh medium, refreshed every 2 days, and adjusted to a concentration of 1 × 10⁶ cells/mL. New stimulations were applied using new beads at a ratio of 1:1 in combination with recombinant human IL2 (PeproTech, 0.5 μg/mL; cat. #200-02) or IL15 (PeproTech, 0.5 μg/mL; cat. #200-15) for 14 days. Starting from day 7, STAT5i (100 μmol/L) was added to the culture medium every 2 days until day 14. Cells were collected on days 7 and 14 of expansion for flow cytometry staining and analysis as described above. For cytotoxicity assays, expanded CAR T cells were cocultured with the human EGFP-expressing CD22⁺ pre–B-cell acute lymphoblastic leukemia (B-ALL) cell line 697 (RRID: CVCL_0098) at various effector-to-target ratios (4:1, 2:1, 1:1, and 1:2) in RPMI supplemented with 2% FBS, 100 U/mL penicillin–streptomycin, and 20 U/mL IL2 for 24 hours. The 697 cell line was confirmed to be free of Mycoplasma contamination at the time of cryopreservation using the Mycoplasma PCR Detection Kit (Applied Biological Materials, cat. #G238). Target cells were identified by enhanced GFP fluorescence, and CD22 surface expression was confirmed by flow cytometry. The number of surviving target cells was determined by flow cytometry. The percentage of lysis was calculated as follows: % lysis = [1 – (number of live target cells in wells with effector cells/number of live target cells in wells without effector cells)] × 100.

Statistical analysis

Statistical analyses were conducted with GraphPad Prism (RRID: SCR_002798). Statistical analysis of data was performed using Mann–Whitney and unpaired Student t tests. Simultaneous comparisons of more than two groups were performed using one-way analysis of variance (ANOVA) or two-way ANOVA, as indicated in the figure legends. Statistical significance of survival curves was assessed using the log-rank (Mantel–Cox) test. Significance was set as follows: *, P < 0.05; **, P < 0.01; ***, P < 0.001; and ****, P < 0.0001. In all cases, P < 0.05 was considered statistically significant. Data in bar graphs and scatter plots are displayed as mean ± SEM.

Results

Sustained IL2 and IL15 stimulation promotes exhausted T cell–like differentiation in vitro

IL2 and IL15 cytokines have shared yet distinct functions in CD8⁺ T-cell immunity. Both cytokines share common receptor chains (IL2Rβ and IL2Rγ) and activate similar signaling molecules (e.g., JAK3 and STAT5) in T cells (18, 48). To compare the impact of IL2 and IL15 on CD8⁺ T-cell differentiation, we developed an in vitro system of continuous IL2 and IL15 stimulation in the context of transient or continuous TCR stimulation modeling acute and chronic infection, respectively. Naïve P14 CD8⁺ T cells were stimulated with gp33 peptide–loaded dendritic cells (100 ng/mL). On days 2, 4, and 6, CD8⁺ T cells were normalized in number and expanded in the presence or absence of gp33 (1 ng/mL) and increasing concentrations of IL2 or IL15 (Fig. 1A). Consistent with prior observations (27), both IL2 and IL15 promoted dose-dependent induction of 2B4, CD160, and, to a lesser extent, LAG-3 and TIM-3 (Fig. 1B). Although continuous TCR stimulation promoted higher expression of these inhibitory receptors (IR), 2B4 and TIM-3 were predominantly regulated by cytokine signals. In contrast, we observed a gradual loss of stem-like features (e.g., CD62L, Ly108, and TCF-1) with heightened levels of both IL2 and IL15, particularly in the context of transient peptide priming (Fig. 1C and D). This suggested that constant IL2 and IL15 stimulation, when provided at high levels, drives cells toward more differentiated T-cell fates.

Figure 1.

IL2 and IL15 promote exhaustion-like differentiation in vitro. A, Schematic of the experimental design for in vitro experiments. P14 CD8⁺ T cells were cocultured with gp33-pulsed dendritic cells (100 ng/mL). On days 2, 4, and 6, P14 cells were normalized and cultured without (priming) or with (continuous) gp33 peptide alone (1 ng/mL) or in combination with varying concentrations of IL2 (blue) or IL15 (green; 5, 10, 100, and 200 ng/mL). Cells were analyzed on day 7. B, Histograms of indicated IRs. C, Contour plots for CD44 and CD62L expression. D, Histograms of the mean fluorescence intensity of Ly108 and TCF1. E, Volcano plots of the DE genes in continuously stimulated CD8⁺ T cells and IL2 (left, 100 ng/mL) or IL15 (right, 200 ng/mL) compared with naïve CD8⁺ T cells (n = 3 replicates per group). Upregulated and downregulated genes associated with T-cell exhaustion are labeled and categorized into pathways (log₂ FC ≥ 0.5 and FDR < 5%). F, Preranked GSEA of DE gene profiles comparing IL2- or IL15-stimulated CD8⁺ T cells to naïve CD8⁺ T cells against a T-cell exhaustion gene signature derived from Bengsch and colleagues (upregulated genes, red line; downregulated genes, blue line; ref. 40). Data are representative of three independent experiments (B–D). ES, enrichment score.

To more precisely evaluate if our in vitro approach recapitulates features of T-cell exhaustion, CD8⁺ T cells stimulated with high levels of cytokines and continuous TCR stimulation were collected at the end of the culture for bulk RNA-seq and compared with naïve CD8⁺ T cells (n = 3 independent cultures per group). At an FDR of 5% and using a log₂ FC ≥ 0.5, we found that 3,072 genes were upregulated and 2,869 genes were downregulated in IL2 stimulated cells, whereas 3,094 genes were upregulated and 2,941 genes were downregulated in IL15-stimulated cells, compared with naïve cells (Fig. 1E; Supplementary Tables S1 and S2). Continuous cytokine and TCR stimulation led to elevated expression of transcripts associated with T_EX^term, including high levels of IR-encoding genes such as Lag3, Havcr2 (TIM-3), Ctla4, Entpd1 (CD39), CD38, Tigit, CD244a (2B4), and CD160. Additionally, these cells showed increased expression of transcription factors (TF) including Prdm1 (BLIMP-1), Eomes, Hif1a, Nfkb1, and Id2; cytolytic molecules including Gzma and Gzmb; and cell surface receptors including Tnfrsf9 (CD137), Tnfrsf4 (CD134), and IL2Rβ (IL2Rβ/CD122). In contrast, the expression of TFs associated with T-cell stemness, including Satb1, Bach2, Tcf7 (TCF-1), Lef, and Klf2, as well as chemokines or cell surface proteins related to stemness, including IL7r (IL7Ra/CD127), Ccr7, S1pr1, Itgae (CD103), Sell (CD62L), and Slamf6 (Ly108), was diminished in IL2- or IL15-stimulated CD8⁺ T cells. Notably, gene expression profiles of CD8⁺ T cells cultured with IL2 or IL15 under continuous TCR stimulation closely resembled the gene signatures of exhausted T (T_EX) cells obtained from chronically infected mice (Fig. 1F; ref. 40). Collectively, these results indicate that sustained TCR and IL2 or IL15 stimulation promote the differentiation of T_EX-like cells in vitro, supporting the use of this in vitro model to investigate the cellular and molecular mechanisms of exhaustion.

Loss of IL2Rβ-dependent signals favors progenitor-like T_EX states during chronic LCMV infection

We then took advantage of the murine LCMV Cl.13 model to better understand the impact of IL2 and IL15 signals on CD8⁺ T-cell fates. This model causes a chronic infection in mice and exhaustion of virus-specific CD8⁺ T cells by approximately 1 month of infection (27). Using this model, we previously found that the expression of IL2Rβ (common to IL2 and IL15) on Ag-specific CD8⁺ T cells positively correlated with features of T-cell exhaustion (e.g., IR expression) and extended these findings to virus-specific CD8⁺ T cells in patients with chronic hepatitis C (27). To interrogate how IL2Rβ expression influences T-cell commitments toward exhaustion, we used the same model and sorted CD8⁺ T cells specific for the gp33-41 epitope of the LCMV based on their level of IL2Rβ expression on day 21 post-infection (p.i.) with LCMV Cl.13. Bulk RNA-seq was then performed to compare gene expression profiles between cells with low (IL2Rβ^low) or high (IL2Rβ^high) IL2Rβ expression. At a FDR of 5% and using a log₂ FC ≥ 0.5, 1,293 differentially expressed (DE) genes were upregulated and 1,408 DE genes were downregulated in IL2Rβ^high compared with IL2Rβ^low cells (Fig. 2A; Supplementary Table S3). Comparative GSEA further showed an enrichment for terminally exhausted signature genes (when compared with Slamf6⁻ Tim3⁺ CD8⁺ T cells from day 28 p.i.; ref. 38) among DE genes in IL2Rβ^high cells, whereas progenitor exhausted signature genes (from Slamf6⁺ Tim3⁻ CD8⁺ T cells from day 28 p.i.; ref. 38) were downregulated (Supplementary Fig. S1A and S1B). The upregulated genes in IL2Rβ^high cells include those encoding IRs such as Pdcd1 (PD1), CD244a (2B4), CD160, Lag-3, Tigit, Havcr2 (TIM-3), CD101, CD38, Entpd1 (CD39), CD200R2, and CD200R4; costimulatory molecules like Tnfrsf9 (CD137/4-1BB) and Tnfrsf7 (CD27); cytotoxic proteases such as Gzma and Gzmb; and TFs including Tox, Prdm1 (Blimp), Eomes, Ikzf2, and Batf (Fig. 2A; Supplementary Table S3). Conversely, genes associated with T-cell stemness were downregulated in IL2Rβ^high gp33-41–specific cells. These include surface receptors such as Il7r (IL7Ra/CD127), Ccr7, Slamf6 (Ly108), Sell (CD62L), and S1pr1; the effector molecule TNF; and TFs such as Id3, Tcf7 (TCF-1), Lef1, Satb1, and Klf2 (Fig. 2A; Supplementary Table S3). These observations suggest a role for sustained IL2Rβ-dependent signals in fostering more differentiated exhausted states at the expense of T-cell stemness in the context of persistent antigenic stimulation.

Figure 2.

IL2Rβ deficiency signaling preserves progenitor exhausted states during chronic LCMV infection. A, Volcano plot showing DE genes between IL2Rβ^high and IL2Rβ^low CD8⁺ Tet⁺ CD11a⁺ T cells on day 21 p.i. Genes previously reported by Miller and colleagues (GSE123235; ref. 38) as significantly upregulated in T_EX^prog (Slamf6⁺Tim-3⁻, green) and in T_EX^term (Slamf6⁺Tim-3⁺, orange) are highlighted; all other genes are shown in gray (log₂ FC ≥ 0.5 and FDR < 5%; n = 3 biological replicates per group). B–G, Flow cytometric analysis of splenic P14⁺ IL2Rβ^+/+ cells (P14 WT, black line or circles) and P14⁺ IL2Rβ^−/− cells (red line or circles) on day 28 p.i., following adoptive cotransfer in B6 mice and LCMV Cl.13 infection. B, Representative histogram. C, Quantification of MFI for the indicated surface markers. D, Frequency of PD1^high and TCF1⁺ cells. E, Representative contour plot of Ly108 and CD69 expression. F and G, Frequency (F) and absolute number (G) of T_EX^Prog1 (CD69⁺Ly108⁺), T_EX^Prog2 (CD69⁻Ly108⁺), T_EX^int (CD69⁻Ly108⁻), and T_EX^term (CD69⁺Ly108⁻) populations. Data are representative of (C, F, and G) or pooled from (D) two independent experiments (total of 24 mice). Error bars indicate means ± SEM. *, P < 0.05; **, P < 0.001; ****, P < 0.0001, by Mann–Whitney (C and D) or one-way ANOVA (F and G). MFI, mean fluorescence intensity.

To specifically address whether IL2Rβ expression on CD8⁺ T cells supports the transition from progenitor to more differentiated exhausted states in vivo, we cotransferred 5 × 10³ congenically distinct populations of WT (CD45.1⁺/CD45.2⁺) and IL2Rβ-deficient (IL2Rβ^−/−, CD45.2⁺) CD8⁺ T cells Tg for the gp33-41 epitope of LCMV (P14 mice) into B6 (CD45.1⁺) recipient mice 1 day prior to infection with LCMV Cl.13 (Supplementary Fig. S1C). On day 28 p.i., when exhaustion was fully established (35, 49, 50), donor cells isolated from the spleen were analyzed by flow cytometry. We observed a greater accumulation of WT over IL2Rβ^−/− P14 cells (∼37-fold; Supplementary Fig. S1D). Phenotypically, P14 IL2Rβ^−/− cells displayed significantly decreased expression of all IRs tested—that is, 2B4, TIM-3, Lag3, and CD101—compared with P14 WT, and a reduced frequency of PD1^high cells (Fig. 2B–D). At this time point during a chronic LCMV infection, distinct but developmentally related subsets of T_EX cells have been identified. These include TCF1-expressing progenitors (Ly108⁺ CD69⁺/T_EX^prog1 and Ly108⁺ CD69⁻/T_EX^prog2) that continuously repopulate more differentiated populations, including an intermediate or effector-like subset (Ly108⁻ CD69⁻ or Ly108⁻ CX3CR1⁺/T_EX^int) and a T_EX^term population (Ly108⁻ CD69⁺ or Ly108⁻CX3CR1⁻) that both lose TCF1 expression (34, 50). In the absence of IL2Rβ, increased expression of TCF1 suggested a developmental bias toward progenitor/stem-like exhausted cells (Fig. 2B and D). Consistently, the proportion of distinct T_EX subsets was redistributed toward a higher frequency of Ly108-expressing T_EX progenitors in IL2Rβ-deficient cells compared with P14 WT (both Ly108⁺ CD69⁻ T_EX^prog2 and Ly108⁺ CX3CR1⁻ stem-like cells; Fig. 2E and F; Supplementary Fig. S1E and S1F). In contrast, the frequency of T_EX^int (defined either as Ly108⁻CD69⁻ or CX3CR1⁺) and T_EX^term (Ly108⁻CD69⁺ or Ly108⁻CX3CR1⁻) cells was reduced in P14 IL2Rβ^−/− compared with their WT counterpart. Absolute numbers reflected this developmental bias, with dramatic reductions of T_EX^int and T_EX^term in the absence of IL2 and IL15 signals, whereas the number of T_EX progenitor cells was comparatively less affected (Fig. 2G; Supplementary Fig. S1G). These findings point to IL2Rβ-dependent signals as a key axis promoting more differentiated T_EX cell states at the expense of a progenitor or stem-like program during chronic antigenic stimulation.

Transient STAT5 inhibition downstream of IL2 and IL15 enriches progenitor-like T-cell states

We next interrogated the signals downstream of IL2 and IL15 that dictate CD8⁺ T_EX-cell differentiation. IL2 and IL15 engage multiple signaling pathways in CD8⁺ T cells, including AKT/mTOR, ERK, and JAK/STAT pathways, with preferential activation of STAT5 over STAT1 and STAT3 (48, 51). However, activation of these pathways may be altered in T_EX cells compared with conventional T_EFF or T_MEM cells, with evidence that AKT/mTOR and ERK signaling are attenuated in the former population (52). Leveraging the LCMV Cl.13 model, we sought to define which downstream pathways promote T_EX exit from a stem-like progenitor state toward terminal exhaustion. To this end, we isolated spleens from mice infected with LCMV Cl.13 on days 8 and 28 p.i. and assessed the activation of key signaling molecules in P14 CD8⁺ T cells following ex vivo restimulation with IL2 or IL15. Naïve CD8⁺ T cells were also included as controls. Consistent with prior work (22, 23), neither IL2 nor IL15 markedly enhanced signaling through the AKT/mTOR or ERK pathways in T_EX cells, as assessed by phosphorylation of AKT, ERK, Foxo3a, and S6 (Fig. 3A; Supplementary Fig. S2A). Induction of STAT1 and STAT3 phosphorylation was likewise limited. In contrast, we observed robust induction of phospho-STAT5 during both the early (day 8 p.i.) and late (day 28 p.i.) phases of LCMV Cl.13 infection, predominantly in response to IL15 and, to a lesser extent, IL2 (Fig. 3A). These observations point to STAT5 as the preferential downstream effector of IL2- and IL15-dependent signaling in T_EX cells in vivo.

Figure 3.

STAT5 inhibition favors the differentiation of T_EX^prog cells. A, MFI of phosphorylated signaling proteins in CD8⁺ T cells isolated from LCMV Cl.13–infected mice on days 8 and 28 p.i. following ex vivo stimulation with IL2 (left, blue) or IL15 (right, green; n = 2 independent experiments; total of 24 mice). B, Representative flow cytometry plot showing CX3CR1 and Ly108 expression in adoptively transferred Tat-Cre–treated P14 STAT5^+/+ and P14 STAT5^fl/fl CD8⁺ T cells analyzed on day 28 after LCMV Cl.13 infection. C–I, P14 cells were cocultured with gp-33–pulsed dendritic cells in the absence of cytokine or in the presence of IL2 (100 ng/mL; blue) or IL15 (200 ng/mL; green), with or without STAT5i (red; 100 μmol/L unless otherwise specified) from day 4 to day 7. C, MFI of indicated IRs on days 4, 6, and 7. D, Representative histograms of Ly108, CD62L, and TCF1 expression on day 7. E and F, Representative contour plots of CD44 and CD62L and CD69 and Ly108 expression on day 7. G, CFSE dilutions of cells labeled on day 4 and analyzed after 24, 48, and 72 hours. H, Absolute numbers of T_EX^prog1 (CD69⁺Ly108⁺), T_EX^prog2 (CD69⁻Ly108⁺), T_EX^int (CD69⁻Ly108⁻), and T_EX^term (CD69⁺Ly108⁻) populations on day 7. I, Representative histogram of BCL2 expression on day 7. Data are representative of two to five independent experiments (A, B, D–G, and I) or pooled from three independent experiments (C and H). Error bars indicate means ± SEM. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001, by Mann–Whitney test (A), two-way ANOVA (C), or one-way ANOVA (H). MFI, mean fluorescence intensity.

To interrogate the potential role of STAT5 in T_EX differentiation, we took advantage of a conditional STAT5 knockout model to abrogate STAT5 signaling throughout T-cell differentiation in vivo. To do so, naïve WT P14 Rosa26-EYFP STAT5^+/+ and STAT5-floxed P14 Rosa26-EYFP STAT5^fl/fl CD45.2⁺ cells were subjected to transient Cre recombination prior to adoptive transfer into congenically distinct B6 CD45.1⁺ recipient mice. On day 28 p.i. with LCMV Cl.13, we observed high expression of Ly108 and CD62L in the absence of STAT5, whereas TIM-3 expression was significantly reduced (Supplementary Fig. S2B). Importantly, P14 STAT5^fl/fl cells were predominantly composed of T_EX progenitors (both Ly108⁺ CD69⁺ T_EX^prog1 and Ly108⁺ CD69⁻ T_EX^prog2) and failed to express the effector-associated marker CX3CR1, in contrast to their WT counterparts (Fig. 3B; Supplementary Fig. S2C). This supports a role for STAT5 signaling in promoting progression toward more differentiated T_EX cell states in vivo.

We next asked whether STAT5 activity, or its upstream kinase JAK3, could be transiently manipulated to alter the developmental trajectory of CD8⁺ T cells in response to IL2 or IL15, without fully abrogating cytokine signaling. This question is particularly relevant because STAT5 activity has been implicated in differentiation toward effector-like T_EX states and in sustaining their functional capacity in vivo (53). We therefore leveraged our in vitro exhaustion model to test this hypothesis. Using concentrations of IL2 (100 ng/mL) and IL15 (200 ng/mL) that generated comparable T_EX phenotypes at the protein and transcriptional levels (Fig. 1), small-molecule STAT5i or JAK3I (Tofa) were added on days 4 and 6 of culture. These inhibitors reduced STAT5 phosphorylation to varying degrees, whereas pSTAT1 and pSTAT3 signals were low overall and were further diminished in the presence of Tofa (Supplementary Fig. S3A and S3B). STAT5i also transiently increased pAKT levels without affecting pS6 and pERK (Supplementary Fig. S2D). Notably, the transcriptional changes induced by STAT5i closely aligned with those observed in Stat5a- and Stat5b-tyrosine phosphorylation KI CD8⁺ T cells after IL2 stimulation, as shown by significant enrichment of KI-upregulated genes and reciprocal depletion of KI-downregulated genes in the IL2S5i versus IL2 comparison (Supplementary Fig. S2E). Together, these findings support that STAT5i primarily attenuates signaling and transcriptional programs downstream of STAT5.

Having established this, we next assessed how transient STAT5 or JAK3 inhibition affected the acquisition of exhaustion-associated versus stem-like features. Blocking STAT5 or JAK3 activity with small molecule inhibitors over the final 72 hours of culture limited the induction of cytokine-responsive IRs—that is, 2B4, TIM-3, and CD160 (Fig. 3C; Supplementary Fig. S3B). The effect of STAT5 inhibition on IR expression was dose-dependent (Supplementary Fig. S2F). STAT5 inhibition also reinforced stem-like features in P14 CD8⁺ T cells, including the expression of Ly108, CD62L, and the canonical TF Tcf1 (Fig. 3D; Supplementary Fig. S2G and S2H). Indeed, the addition of STAT5i on day 4 promoted the emergence of cells with a progenitor-like phenotype (Ly108⁺CD69⁺) that also coexpressed central memory–associated markers (CD62L⁺CD44⁺) in a dose-dependent manner (Fig. 3E and F; Supplementary Fig. S2G and S2H). Conversely, STAT5i markedly limited the accumulation of T_EX^term (Ly108⁻CD69⁺) as well as effector/effector-like populations (CD62L⁻CD44⁺ T_EFF and Ly108⁻CD69⁻ T_EX^int) in contrast to IL2- and IL15-expanded conditions without STAT5i (Fig. 3E and F; Supplementary Fig. S2G and S2H). JAK3 inhibition using either JAK3I or Tofa produced comparable effects on CD8⁺ T-cell differentiation, with Tofa exerting a greater impact on the preservation of progenitor-like cells (Supplementary Fig. S3).

Consistent with the results obtained in vivo in the absence of IL2Rβ expression on CD8⁺ T cells, blockade of JAK3–STAT5 signaling in vitro modestly limited cell expansion by day 7 (1.6- to 1.7-fold; Supplementary Fig. S2I). This was associated with reduced proliferation of CD8⁺ T cells in the presence of STAT5i (Fig. 3G). However, despite this reduction in overall expansion, the absolute number of stem-like Ly108⁺ P14 cells recovered at the end of culture was increased in the presence of STAT5i, with marked increases in T_EX^prog1 and T_EX^prog2 populations (Fig. 3H). In addition, STAT5 inhibition was associated with increased expression of the antiapoptotic molecule Bcl2 (Fig. 3I). Collectively, these findings indicate that IL2 and IL15 promote the acquisition of exhaustion-associated features at the expense of stem-like biology under conditions of repetitive TCR stimulation. In contrast, transient pharmacologic inhibition of STAT5 restrains these cytokine-driven differentiation effects and enriches progenitor-like T-cell states, supporting its potential utility during the ex vivo generation of ACT products.

Transient STAT5 inhibition reinforces progenitor-like transcriptional programs

To understand the underlying molecular impact of STAT5 inhibition on in vitro expanded CD8⁺ T cells, we performed RNA-seq and compared P14 CD8⁺ T cells expanded in vitro with IL2 or IL15, with or without STAT5 inhibition, and harvested on day 7 of culture. Naïve CD8⁺ T cells were also analyzed as a reference. A total of 211 and 351 genes were upregulated, and 71 and 246 genes were downregulated when comparing IL2S5i to IL2 and IL15 + STAT5i (IL15S5i) to IL15, respectively (log₂ FC ≥ 0.5 and FDR < 0.05; Fig. 4A and B; Supplementary Tables S4 and S5). Notably, we observed concordance in the gene programs altered by STAT5 inhibition for both the IL2 and IL15 conditions (Fig. 4C). This suggested that at least in these in vitro settings, STAT5 induces a comparable differentiation program downstream of both IL2 and IL15. Differential gene expression analysis revealed that genes associated with T-cell memory and/or T-cell stemness, including Sell (CD62L), Slamf6 (Ly108), Ccr7, Tcf7 (TCF-1), Id3, and Bach2, were upregulated in the presence of STAT5i compared with IL2 or IL15 conditions alone (Fig. 4D; ref. 54). Genes encoding critical proteins regulating egress, migration, and/or homing of memory CD8⁺ T cells and resident memory T cells were also upregulated, such as S1pr1, Klf2, and Itgae (CD103). Genes associated with cytotoxic effector molecules such as Gzma, Gzmb, perforin 1 (Prf1), and Tnfsf10 (TRAIL) were downregulated in IL2S5i- and/or IL15S5i-expanded CD8⁺ T cells. Additionally, genes related to terminal exhaustion, including Havcr2 (TIM-3) and Entpd1 (CD39), were also downregulated in these cells (55).

Figure 4.

Stem-like transcriptional programs emerge following STAT5 inhibition. A and B, Volcano plots showing DE genes in (A) IL2S5i versus IL2 or (B) IL15S5i versus IL15 CD8⁺ T cells on day 7 of culture. Downregulated genes are shown in blue for IL2S5i versus IL2 and green for IL15S5i versus IL15; upregulated genes are shown in red (log₂ FC ≥ 0.5 and FDR < 5%, n = 3 biological replicates per condition). Labels indicate DE genes associated with T-cell exhaustion or stemness. C, Scatter plot comparing DE genes in IL2S5i versus IL2 (x-axis) and IL15S5i versus IL15 (y-axis). DE genes associated with T-cell exhaustion or stemness are labeled. D, Heatmap showing hierarchical clustering of experimental conditions (n = 3 per condition, x-axis) and top DE genes from A–C associated with T-cell exhaustion or stemness (right, y-axis). Gene expression was log-normalized and scaled across all samples, with genes grouped by pathway (left, y-axis). E, Heatmap comparing gene expression profiles of naïve and cytokine-stimulated CD8⁺ cells, with or without STAT5i, to T_EX^prog1, T_EX^prog2, T_EX^int, and T_EX^term signatures from Beltra and colleagues (GSE149876; ref. 34). F, GSEA comparing DE gene profiles of IL2S5i versus IL2 and IL15S5i versus IL15 CD8⁺ T cells against published progenitor and T_EX gene signatures (34, 36, 38). P values are indicated as P < 0.05, P < 0.01, and P < 0.005.

The RNA-seq analysis also revealed that CD8⁺ T cells expanded with IL2 and IL15 were transcriptionally closer to the more differentiated exhausted subsets, specifically T_EX^int and T_EX^term populations, observed in chronically infected mice [Fig. 4E (right) side; ref. 34]. In contrast, CD8⁺ T cells expanded with STAT5i displayed transcriptional profiles closer to T_EX^prog1 and T_EX^prog2 progenitors [Fig. 4E (left) side]. GSEA further showed positive enrichment of gene sets (red) associated with Slamf6⁺Tim-3⁻ T_EX^prog (38), TCF1⁺ memory-like T cells (36), and TCF1⁺ progenitor T_EX cells (34) from chronic LCMV infection, in IL2 or IL15S5i-expanded cells (Fig. 4F). Conversely, genes associated with terminally exhausted Slamf6^–Tim-3⁺ (38), TCF1⁻ T cells (36), and TCF1⁻ T_EX^term (34) were negatively enriched (blue) in STAT5i-expanded CD8⁺ T cells. Together, these observations confirm that transient reduction in STAT5 signaling downstream of IL2 and IL15 supports stem-like programs in in vitro expanded CD8⁺ T cells.

Attenuated STAT5 signaling enhances cytokine production while preserving cytotoxic potential

Repetitive Ag signals during in vitro expansion protocols not only drive programs that resemble those of in vivo exhausted CD8⁺ T cells, but these cells also undergo substantial functional impairment (56, 57). The ability of STAT5 inhibition to reduce some features of exhaustion in vitro, as well as the demonstrated impact it had on the transcription of effector molecules such as granzymes, perforin and TRAIL, prompted us to evaluate the functionality of CD8⁺ T cells expanded under conditions of attenuated STAT5 signaling. To do so, P14 CD8⁺ T cells expanded for a week under constant TCR signaling and IL2 or IL15, in the presence or absence of STAT5i, were restimulated for 6 hours at the end of the culture (day 7) with a murine EL4 cell line expressing the cognate LCMV D^bgp33-41 peptide (EL4-gp33). We found that cells expanded with STAT5i (under either IL2 or IL15 conditions) produced significantly more IFNγ than their nontreated counterparts (IL2 or IL15 alone) and expressed higher levels of the degranulation marker CD107, whereas the overall production of TNFα and IL2 was comparable across conditions (Fig. 5A and B). More detailed analysis revealed that STAT5 inhibition affected the quality of effector cells on a per-cell basis, with an increased frequency of cells expressing high levels of TNFα, IFNγ, and CD107a and an increased frequency of polyfunctional cells secreting all three effector proteins simultaneously (Fig. 5C). In contrast and as expected from the transcriptomic data, STAT5i-expanded cells, particularly in the context of IL15 stimulation, had reduced but not absent expression of granzyme B, in terms of both overall levels and frequency of granzyme B^high cells (Fig. 5A–C), suggesting potential impairment in immediate cytolytic potential.

Figure 5.

STAT5 inhibition enhances cytokine production while preserving T-cell cytotoxicity. A–E, P14 CD8⁺ T cells were expanded in vitro with continuous gp33 peptide and IL2 (100 ng/mL; blue) or IL15 (200 ng/mL; green) in the presence or absence of STAT5i (100 μmol/L; red), as previously described in Fig. 3C-I, and harvested on day 7. A, Representative histograms and (B) quantification of MFI for the indicated cytokines. C, Frequency of cells expressing high levels of the indicated cytokines. D and E, In vitro cytotoxicity assay in which day 7 cells were cocultured with CFSE-labeled EL4-gp33 target cells (CFSE^high) and Ag-negative EL4 nontarget cells (CFSE^low) at a 1:1 effector-to-target ratio for 3, 6, or 18 hours. D, Representative contour plot showing the frequency of CFSE^high and CFSE^low cells at indicated time points. E, Percentage of EL4-gp33 target cells lysed by P14 CD8⁺ T cells expanded in vitro for 7 days. Data are representative of three independent experiments. Error bars indicate means ± SEM. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001, by one-way ANOVA (B and C) or two-way ANOVA (E). FSC-A, forward scatter area; MFI, mean fluorescence intensity.

To directly test cytotoxicity, we performed a time course in vitro killing assay using P14 CD8⁺ T cells harvested after 7 days of expansion. An equal number of P14 CD8⁺ T cells expanded with IL2 or IL15, with or without STAT5i, were cocultured with EL4-gp33 target cells and Ag-negative EL4 control cells, allowing Ag-specific killing to be quantified over time. At 3 and 6 hours, STAT5i-expanded CD8⁺ T cells exhibited reduced target cell killing compared with cells expanded with IL2 or IL15 alone, consistent with reduced granzyme B expression in these populations (Fig. 5D and E). However, by 18 hours, EL4-gp33 target cells were nearly completely eliminated across conditions, indicating that all groups ultimately achieved comparable target cell elimination. These results indicate that despite reduced early killing, CD8⁺ T cells expanded under attenuated STAT5 signaling retain cytotoxic potential while exhibiting enhanced degranulation and cytokine production.

STAT5i-expanded CD8⁺ T cells improve tumor control after adoptive transfer

We next evaluated the antitumor potential of CD8⁺ T cells expanded under conditions of limited STAT5 signaling. We first tested these cells in a prophylactic model by transferring OT-I CD8⁺ T cells expanded for 7 days with cognate peptide and IL2, with or without STAT5i during the final 3 days of culture, into B6 mice 1 day before subcutaneous (s.c.) injection of B16-OVA melanoma cells (Fig. 6A). Mice that received IL2 or IL2S5i–expanded OT-I cells exhibited delayed tumor growth compared with mice that did not receive T-cell transfer (P = 0.0068 and P < 0.0001, respectively; Fig. 6B). ACT-treated mice also showed prolonged survival compared with controls, with a median survival of 28 days versus 22 days (Fig. 6C). Notably, ACT with IL2S5i–expanded cells resulted in greater tumor growth reduction than ACT with IL2-expanded cells although this did not reach statistical significance (P = 0.079). IL2S5i–expanded cells also conferred the greatest survival benefit, with increased survival compared with IL2-expanded cells (P = 0.0401) and 22% of mice surviving on day 60 (Fig. 6B and C). All animals in the other groups, including mice receiving IL2-expanded ACT, eventually succumbed to tumor challenge. These findings suggest that STAT5i-expanded CD8⁺ T cells improve tumor control after transfer in vivo, at least in the prophylactic setting.

Figure 6.

STAT5i-expanded CD8⁺ T cells improve tumor control after adoptive transfer. A, Schematic of the prophylactic tumor model. OT1 CD8⁺ T cells stimulated with the OVA_257–264 peptide (black) and IL2 (100 ng/mL; blue), with or without STAT5i (100 μmol/L; red) for 7 days, and were adoptively transferred into B6 mice 1 day before s.c. injection of 2 × 10⁵ B16-OVA tumor cells. Tumor size and animal survival were monitored. B, Individual tumor growth curves and (C) survival of mice following B16-OVA inoculation. D, Schematic of the therapeutic tumor model. B6 mice were inoculated with 2 × 10⁵ B16-GP33 tumor cells. P14 CD8⁺ T cells stimulated in vitro for 7 days with gp33 peptide (black) and IL15 (200 ng/mL; green), with or without STAT5i (100 μmol/L; red), were adoptively transferred into tumor-bearing mice on day 10. Tumor size and animal survival were monitored. E, Individual tumor growth curves and (F) survival of mice following B16-GP33 inoculation. Data in (B) and (C) are pooled from two independent experiments with 9 to 11 mice per group. Data in (E) are pooled from two independent experiments with 11 to 14 mice per group. Data in (F) are representative of one of two experiments with 6 to 7 mice per group. Statistical significance was determined by two-way ANOVA (B and E) and the log-rank (Mantel–Cox) test (C and F). Median survival is indicated on each graph. *, P < 0.05; **, P < 0.01; ****, P < 0.0001.

We next asked whether CD8⁺ T cells expanded in vitro retained the capacity to mediate tumor control in a therapeutic setting in which tumors were already established. P14 CD8⁺ T cells were expanded for 7 days with cognate gp33-41 peptide and IL15, in the presence or absence of STAT5i, and injected into B16-GP33 tumor-bearing mice on day 10 after tumor inoculation (Fig. 6D). Consistent with the prophylactic model, CD8⁺ T cells expanded under attenuated STAT5 conditions had the greatest effect on tumor growth (P < 0.0001 for IL15S5i vs. control; P = 0.0047 for IL15 vs. control; Fig. 6E). Similarly, mice that received IL15S5i-expanded cells exhibited prolonged survival, with a median survival of 32 days compared with 28 days for mice receiving IL15-expanded cells and 26 days for controls (P = 0.0447 for STAT5i-treated vs. controls; Fig. 6F). Notably, only the IL15S5i group included mice surviving beyond 40 days, with 16% survival, whereas all other mice died by day 32. Together, these experiments indicate that transient STAT5 inhibition during in vitro CD8⁺ T-cell expansion enhances the in vivo efficacy and antitumor activity of adoptively transferred cells in this particular tumor context.

Restricted STAT5 signaling favors memory–progenitors in human CAR T cells while preserving cytotoxicity

Finally, we asked whether reduced STAT5 signaling could similarly promote memory–progenitor features in human CAR T cells. To investigate this question, we modulated STAT5 activity during the expansion of human CAR T cells using protocols that favor CAR T-cell exhaustion (58). Human CD8⁺ T cells were activated, transduced with a CD22-targeting CAR, and expanded with anti-CD3/CD28 and IL2 or IL15, in the presence or absence of STAT5i added during the later phase of culture (Fig. 7A; refs. 37, 39, 47). Using CD39 and CD69 coexpression, as well as IR expression, to evaluate the phenotypic characteristics of the cells, we found that human CD22-CAR T cells expanded in the presence of IL2 and IL15 differentiated into a CD39⁺CD69⁺ highly differentiated subset (Fig. 7B–D) and expressed high levels of PD-1, TIM-3, 2B4, and CD39 (Fig. 7E and F). In contrast, CAR T cells expanded under restricted STAT5 signaling displayed reduced expression of CD69, an early activation marker, and CD39, an ecto-ATPase expressed by effector memory and T_EX cells (Fig. 7B). This led to an increased frequency and absolute number of memory–progenitor cells (CD39⁻CD69⁻; Fig. 7C and D), which have been shown to persist and support anticancer responses in patients with melanoma treated with ACT (8). Other memory-associated markers were similarly increased in the presence of STAT5i, including the L-selectin CD62L (Fig. 7E), whereas the expression of IRs was reduced in this setting (Fig. 7E and F). Increased memory–progenitor differentiation and limited IR expression also correlated with reduced CAR T-cell expansion after day 10 of culture (Fig. 7G).

Figure 7.

Human CAR T cells acquire memory–progenitor features under STAT5 limiting conditions. A, Schematic of the experimental design. Human CD8⁺ T cells were activated using human CD3/CD28 Dynabeads for 2 days prior to transduction with an anti-CD22 CAR. CAR T cells were expanded for 2 weeks using CD3/CD28 beads and either human IL2 (0.5 μg/mL; blue) or human IL15 (0.5 μg/mL; green) every 2 days, with or without STAT5i (100 μmol/L; red) from day 7 onward. B, Representative flow cytometry plots showing CD69 and CD39 expression on day 14 of culture. C and D, Frequency (C) and cell number (D) of memory progenitors (CD69⁻ CD39⁻) and terminally differentiated (CD69⁺ CD39⁺) cells. E, Representative histograms of the indicated surface markers. F, Frequency of cells expressing high levels of the indicated IRs. G, Total CAR T-cell number obtained on days 10 and 14 of culture when stimulated with IL2 (blue) or IL15 (green) with or without STAT5i (red). H, Percentage of enhanced GFP⁺ CD22⁺ target cells lysed after 24 hours by CAR T cells expanded with IL2, with or without STAT5i, at the indicated effector-to-target ratios. Data in (B–G) are representative of two independent experiments. Data in (H) are pooled from three independent experiments with nine technical repeats. Error bars indicate means ± SEM. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001 by two-way ANOVA.

To determine whether transient STAT5 inhibition impaired the antitumor function of expanded CAR T cells, we performed an in vitro cytotoxicity assay. To do so, we cocultured IL2-expanded CD22 CAR T cells generated in the presence or absence of STAT5i with CD22 pre–B-ALL leukemic targets for 24 hours. Using various effector-to-target ratios, we found that CD22 CAR T cells expanded with or without STAT5i efficiently eliminated leukemic targets, particularly at high effector-to-target ratios (Fig. 7H). Together, these experiments indicate that transient STAT5 inhibition during human CAR T-cell expansion favors a memory–progenitor phenotype while preserving in vitro antitumor cytotoxicity. Collectively, these findings indicate that sustained IL2 and IL15 exposure promotes progression toward more differentiated T_EX states in murine and human CD8⁺ T cells, whereas transient STAT5 inhibition during ex vivo expansion enriches progenitor-like states while preserving functional competence and enhancing antitumor activity after transfer.

Discussion

To improve ACT efficacy, it is essential to clarify how cytokine signaling shapes T-cell differentiation under chronic antigenic stimulation. Here, we developed and validated a controlled in vitro model of sustained cytokine exposure during continuous antigenic stimulation, a setting designed to capture key features of T_EX-cell differentiation (57). Under these conditions, high concentrations of IL2 and IL15 promoted progression toward more advanced T_EX differentiation states. In vivo, IL2Rβ^high virus-specific CD8⁺ T cells were enriched for terminal exhaustion–associated signatures, whereas genetic loss of IL2Rβ or STAT5 favored progenitor-like T_EX populations. These results extend our prior work showing that IL2Rβ/CD122 expression marks more differentiated T_EX cells during chronic LCMV infection and that IL2Rβ-dependent signals promote PD-1^high terminal differentiation, IR expression, and loss of memory potential (27). They are also consistent with tumor studies showing that sustained IL2 signaling can promote CD8⁺ T-cell exhaustion through persistent STAT5 activation and downstream tryptophan-associated transcriptional reprogramming (23). Together, these findings indicate that when delivered at high intensity and under chronic TCR stimulation, IL2 and IL15 converge on similar transcriptional programs that accelerate differentiation away from progenitor-like states and toward more advanced T_EX programs.

These findings should be interpreted in the context of the bifunctional role of STAT5 in exhausted and tumor-reactive CD8⁺ T cells. Recent studies have shown that IL2 or enforced STAT5 activity can promote effector-like differentiation, persistence, and antitumor function while limiting exhaustion-associated transcriptional programs (53, 59, 60). IL2 combined with PD-1 blockade can mobilize stem-like exhausted CD8⁺ T cells toward functional effector states, whereas constitutive or engineered STAT5 activation can oppose TOX-driven exhaustion programs and sustain durable effector-like _TEX populations during chronic Ag exposure (53, 59). In tumors, MDM2-dependent stabilization of STAT5 further supports CD8⁺ T-cell survival and antitumor immunity, highlighting a beneficial role for STAT5 in specific cellular contexts (61). Our data are consistent with a role for STAT5 in effector differentiation, as STAT5 inhibition reduced expansion and granzyme B expression and reduced immediate cytotoxicity. Nonetheless, time-restricted STAT5 attenuation during ex vivo expansion preserved cytokine production, degranulation, delayed but effective target-cell killing, and improved tumor control after transfer. Together, these findings support a model in which STAT5 signaling can promote exit from progenitor-like states but may also sustain effector differentiation and antitumor function when engaged in the appropriate temporal and cellular context.

These observations have direct implications for ACT manufacturing, in which cytokine signals must support expansion without prematurely driving T cells into more advanced differentiated states. Thus, our data do not support permanent STAT5 inhibition in therapeutic T cells but rather the temporal tuning of STAT5 activity during a defined manufacturing window. This concept aligns with emerging strategies that modulate the quality, intensity, or timing of cytokine signals rather than eliminating them. For example, the engineered IL2 partial agonist H9T alters STAT5 signaling and expands CD8⁺ T cells while limiting terminal differentiation and improving antitumor activity (22). In parallel, CAR constructs incorporating an IL2Rβ-derived signaling domain deliver Ag-dependent JAK–STAT signals, thereby improving CAR T-cell persistence and antitumor efficacy (62). In our system, STAT5 inhibition reduced T-cell proliferation and overall expansion, highlighting a trade-off between cell yield and functional quality. This is particularly relevant for ACT manufacturing, in which sufficient cell numbers are required for clinical efficacy, but prolonged or cytokine-intensive expansion can compromise product fitness by promoting differentiation and loss of progenitor-like features (63, 64). Similar principles have been observed with chronic type I IFN signaling, in which sustained JAK-dependent inflammatory signaling promotes terminal CD8⁺ T-cell differentiation, whereas delayed inhibition after PD-1 blockade enhances progenitor-like states and T-cell plasticity (65). Together, these studies suggest that optimal therapeutic benefit may require neither maximal nor absent STAT5 activity but appropriately timed and calibrated cytokine signaling.

T-cell stemness is an important determinant of ACT efficacy because progenitor-like CD8⁺ T cells exhibit greater self-renewal, persistence, proliferative capacity, and responsiveness to inflammatory or checkpoint-dependent signals (66, 67). However, progenitor bias alone is not sufficient for tumor control. Effective ACT requires that stem-like T cells retain the capacity to differentiate into effector or effector-like populations after transfer (7, 49, 68). In our study, STAT5i-expanded cells not only acquired TCF1/Ly108 or memory–progenitor features but also retained degranulation, cytokine production, and target cell killing capacity. Importantly, these cells improved tumor control in both prophylactic and established tumor settings, indicating that progenitor-like enrichment did not occur at the expense of antitumor activity. This improved antitumor activity may reflect enhanced responsiveness to tumor-derived cues, more efficient conversion into effector-like progeny, delayed terminal differentiation, enhanced persistence, or a combination of these mechanisms, which remain to be defined. Consistent with this concept, prior studies have shown that modulation of other signaling pathways downstream of the TCR or cytokine receptors, including RAS–MAPK and PI3K–AKT, can generate less differentiated T cells with improved antitumor activity (63–65). Together, these findings support a broader ACT-manufacturing principle in which calibrated modulation of intracellular signaling can preserve stem-like properties while maintaining the effector potential required for tumor control.

In summary, this work establishes a controlled in vitro model to dissect cytokine-driven CD8⁺ T-cell differentiation under chronic stimulation. We show that sustained high-dose IL2 and IL15 promote progression toward more differentiated T_EX states, whereas transient attenuation of STAT5 signaling during ex vivo expansion enriches TCF1⁺ progenitor-like states while preserving functional competence and improving antitumor activity after transfer. These findings identify the intensity, timing, and duration of IL2Rβ–STAT5 signaling as important determinants of CD8⁺ T-cell fate. More broadly, they support a model in which cytokine/JAK–STAT signals can sustain effector function or promote terminal differentiation depending on their temporal and cellular context, providing a framework to modulate cytokine signaling during ex vivo expansion and improve the functional quality of therapeutic T-cell products.

Supplementary Material

Figure S1

IL2Rβ deficiency favors progenitor TEX states during chronic LCMV infection

Figure S2

STAT5 inhibition enhances stem-like features in CD8+ T cells stimulated

Figure S3

JAK inhibition favors stem-like progenitor differentiation in vitro

Table S1

Table S1

Table S2

Table S2

Table S3

Table S3

Table S4

Table S4

Table S5

Table S5

Data Availability

The RNA-seq data generated during this study have been deposited in the National Center for Biotechnology Information GEO under accession numbers GSE335610 (in vivo IL2Rβ^high/low virus-specific CD8⁺ T cells during chronic LCMV infection) and GSE335822 (CD8⁺ T cells cultured in vitro with IL2 or IL15 in the presence or absence of STAT5 inhibition). Corresponding analyses are provided in Supplementary Tables S1–S5. Other data are available from the corresponding author upon reasonable request.

Authors’ Disclosures

É. Haddad reports personal fees from Takeda Pharmaceuticals, Rocket Pharmaceuticals, Pharming, and Octapharma during the conduct of the study. H. Decaluwe reports grants from the Canadian Institutes of Health Research and Fonds de Recherche du Québec - Santé during the conduct of the study. No disclosures were reported by the other authors.

Authors’ Contributions

M. Shourian: Conceptualization, formal analysis, investigation, visualization, methodology, writing–original draft, writing–review and editing. B. Bourdin: Visualization, writing–original draft, project administration, writing–review and editing. N. El-Hachem: Formal analysis. C. Ruisseaux: Investigation. A. Vallée: Investigation. A. Ballestero Vidal: Investigation. O. Gingras: Formal analysis. H. Roméro: Investigation. E.O. Kwarteng: Investigation. É. Haddad: Resources, supervision, writing–review and editing. V.-P. Lavallée: Supervision, writing–review and editing. J.-C. Beltra: Resources, investigation, writing–review and editing. H. Decaluwe: Conceptualization, resources, formal analysis, supervision, funding acquisition, validation, methodology, writing–original draft, project administration, writing–review and editing.