Interleukin-2 transcriptomic expression correlates with prolonged survival and with GNAS alterations in patients with advanced cancers Authors: J. Ahmed, D. Nishizaki, Y. Fujiwara, P. DePietro, T.J. Jensen, T. McFall, S. Kato, R. Kurzrock Published: 2026 ABSTRACT BackgroundInterleukin-2 (IL-2) is a critical immunoregulatory molecule. IL-2 RNA expression and correlation with immune markers, gene alterations, and outcomes were analyzed in advanced/metastatic cancers.Patients and methodsWe investigated IL-2 transcript data through the OmniSeq clinical-grade laboratory [https://www.omniseq.com/; University of California San Diego (UCSD) cohort] and The Cancer Genome Atlas (TCGA, https://www.cancer.gov/tcga) databases. UCSD cohort transcriptomic patterns were relativized against a reference population (n = 735) and expressed as percentiles. IL-2 levels were measured by enzyme-linked immunosorbent assay in CRISPR-Cas9–engineered GNAS-mutant Caco-2 cell lines.ResultsAmong 514 patients, 14.6% had high IL-2 RNA expression (≥75th percentile) and 35% had less than first percentile IL-2 expression. IL-2 expression (greater than or equal to first percentile RNA rank) significantly/independently correlated with high expression of immune checkpoints (TIM-3, BTLA), IL-2Rγ subunits, and ERBB2, GNAS, and NF1 aberrations. Increasing IL-2 RNA expression (multivariable linear regression) correlated with increasing likelihood of GNAS alterations (P = 0.02). GNASR201C-mutated (CRISPR) cell lines showed significantly higher levels of IL-2 than controls. IL-2 expression was an independent positive prognostic factor for overall survival (OS) [n = 489 UCSD patients with clinical data, hazard ratio (HR) 0.74, 95% confidence interval (CI) 0.57-0.96, P = 0.02, multivariable] and significantly correlated with longer OS in TCGA cohort (n = 9869, HR 0.84, 95% CI 0.79-0.91, P < 0.001).ConclusionsHigh IL-2 tissue transcript levels correlated with better outcome in advanced cancer patients and those with GNAS alterations; the latter finding was confirmed by increased IL-2 production by CRISPR-edited GNASR201C cells, suggesting a functional interaction between this mutation and IL-2. Introduction Interleukin 2 (IL-2) is produced by activated T cells, particularly T helper and natural killer cells, after antigen recognition. Secreted IL-2 binds to its cognate heterotrimeric receptor (IL-2R), which has three distinct subunits, namely the alpha chain (IL-2Rα or CD25), the beta chain (IL-2Rβ or CD122), and the gamma chain (IL-2Rγ or CD132), and exerts specific effects on several T cell lineages.1,2 Mature IL-2 mRNA expression is subject to controlled regulation at the posttranscriptional level.3 A comprehensive evaluation of IL-2 tumor RNA expression, its clinical correlates, and its prognostic and outcome associations in solid tumors can be informative. The use of anti-programmed cell death protein 1/anti-programmed death-ligand 1 [anti-PD-1/anti-PD-(L)1] immune checkpoint inhibitors (ICIs) has demonstrated significant efficacy in treatment of various cancers.4 Increased immune checkpoint protein expression by tumor cells may lead to decreased T cell activity and subsequently decreased IL-2 secretion. The suppression of IL-2 expression may produce CD8+ T cell anergy.5 On the contrary, ICIs may increase peripheral IL-2 cytokine release.6 However, due to the induction of regulatory T cells (Tregs) by IL-2 binding to high-affinity IL-2R, clinical responses to anti-PD-(L)1 may be stunted. Tregs also express cytotoxic T-lymphocyte antigen 4 (CTLA-4), which suppresses anticancer immunity. Although a synergistic effect of anti-CTLA4 and IL-2 targeted therapy was not seen in melanoma,7,8 more recent clinical trials using novel IL-2 therapeutics combined with ICI-based treatments have shown variable results across tumor types.9, 10, 11, 12 In this study, we investigated IL-2 RNA expression across malignancies and its correlation with tumor genomic alterations and analyzed the association between IL-2 transcript expression and outcomes in patients with advanced solid tumors. Patients and methods Patient data Tissue samples were collected from 514 patients (including 489 patients with clinically annotated data) with advanced or metastatic cancer who were treated at the University of California San Diego (UCSD) Moores Cancer Center for Personalized Therapy.13, 14, 15, 16 The tissue analysis was conducted at OmniSeq, which is a Clinical Laboratory Improvement Amendments-licensed and College of American Pathologists-accredited laboratory (https://www.omniseq.com/). Clinical information included patient’s age, sex, tumor type, PD-L1 immunohistochemistry (IHC), tumor mutational burden (TMB), and microsatellite instability (MSI) status. UCSD_PREDICT (Personalized Cancer Therapy to Determine Response and Toxicity; NCT02478931) is an ongoing institutional review board (IRB)-approved study, under which the current analysis was conducted. Patients provided informed consent for any investigational intervention, and the study adhered to the guidelines and regulations established by the UCSD IRB. RNA extraction and expression analysis RNA transcriptome analysis was carried out on formalin-fixed, paraffin-embedded (FFPE) tissue samples collected from advanced cancer patients and analyzed using the clinically validated gene expression panel at OmniSeq/Labcorp (https://oncology.labcorp.com/os-welcome). Analytical validation of this panel has been described in a prior study.17 Transcriptomic patterns were relativized against a reference population (n = 735) and expressed as percentiles. First, DNA and RNA co-extraction was carried out using truXTRAC FFPE extraction (Covaris, Inc., Woburn, MA). Next, to determine RNA yield, the Quant-iT RNA HS (Thermo Fisher Scientific, Waltham, MA) assay was used on extracted and purified RNA that had been dissolved in 50 μL of water. Using the Oncomine Immune Research Response Assay (Thermo Fisher Scientific), an RNA library was prepared from a 10-ng RNA sample, and RNA sequencing (Ion Torrent S5XL system, Thermo Fisher Scientific) was carried out. For measurement of RNA expression, the quantification of absolute RNA in the sample was carried out with Torrent Suite’s plugin immuneResponseRNA (v5.2.0.0). Finally, for an accurate transcript quantification of differentially expressed genes, an internal dataset of housekeeping gene profile was used to normalize transcript abundance and rank the IL-2 RNA expression into expressed (RNA expression ≥1st percentile) and not expressed (RNA expression less than first percentile). RNA expression levels for IL-2 and checkpoint proteins were also reported as percentiles based on their rank values, with ‘high’ RNA expression defined as 75th-100th percentile RNA rank. To characterize an IL-2R expression signature, tumors were classified as having high-affinity IL-2R expression if IL-2Rα, IL-2Rβ, and IL-2Rγ RNA levels were all in the 75th percentile or higher (Supplementary Table S1, available at https://doi.org/10.1016/j.esmoop.2026.107768). Intermediate affinity was assigned when IL-2Rβ and IL-2Rγ were ≥75th percentile, but IL-2Rα was <75th percentile RNA expression. Low affinity was defined by IL-2Rα being ≥75th percentile while both IL-2Rβ and IL-2Rγ were <75th percentile in RNA expression. Any other expression pattern was categorized as ‘all else’ (Supplementary Table S1, available at https://doi.org/10.1016/j.esmoop.2026.107768). In the linear regression model, RNA expression for IL-2, IL-2R subunits, and STAT5A were also reported as percentiles based on their RNA rank values [‘high’ (75th-100th percentile), ‘moderate’ (25th-74th percentile), and ‘low’ (0-24th percentile) expression] and then assigned integers of 2, 1, or 0, respectively. The Cancer Genome Atlas data analysis The Cancer Genome Atlas (TCGA) pan-cancer mRNA expression data were downloaded from Genomic Data Common Data Portal (https://portal.gdc.cancer.gov/; number of samples = 11 274; Supplementary Figures S1 and S2, available at https://doi.org/10.1016/j.esmoop.2026.107768). IL-2 RNA expression was quantified as transcripts per million and subsequently classified into ‘expressors versus non-expressors’, ‘high (≥50th percentile) versus low (<50th percentile)’, and ‘high (≥75th percentile) versus low (<75th percentile)’. PD-L1, TMB, and MSI status The Dako PD-L1 IHC 22C3 pharmDx assay, PD-L1 28-8 pharmDx assay (Dako North America, Inc., Carpinteria, CA), or VENTANA PD-L1 SP142 assay (Ventana Medical Systems, Inc., Tucson, AZ) was used for PD-L1 IHC analysis. PD-L1 expression was expressed as positive (≥1%) versus not positive (<1%) on IHC. A positive result was defined as a tumor proportion score or combined positive score of ≥1% using the Dako 22C3 or 28-8 assays or an immune cell score of ≥1% using the SP142 assay. Using the RNA input library described above and selecting a minimum tumor tissue content of 30% malignant nuclei, library preparation was carried out using the Comprehensive Cancer Panel (Thermo Fisher Scientific), which has been validated previously.17 Template amplification was carried out using the Ion Chef System (Thermo Fisher Scientific). For accurate nonsynonymous somatic alteration frequency calculation, variant calling was carried out using bioinformatics algorithms to eliminate synonymous alterations, germline variants, and insertion/deletions. TMB was calculated as nonsynonymous mutations per megabase of DNA, excluding single nucleotide variants with allele frequency of <5%. For MSI assessment, minimum tumor tissue content of 20% malignant nuclei was selected. Using the MiSeq sequencer from Illumina (San Diego, CA), a 20-ng tumor DNA sample was used for next-generation sequencing (NGS). The output from this assay reports as ‘unstable’, ‘stable’, or ‘inconclusive’ in silico comparison to normal allele distribution at each gene locus. Genomic profiling Regarding gene aberration data, clinical-grade NGS (N = 324 genes) was carried out at Foundation Medicine (https://www.foundationmedicine.com/) on tumor tissues or blood-derived circulating tumor DNA. Variants of uncertain significance were not included. Among enrolled patients, those who received NGS testing with immunomic analysis from July 2017 to November 2020 were included in this analysis. Isogenic Caco-2 cells were cultured in Eagle’s Minimum Essential Medium (EMEM) under standard conditions (37°C, 5% CO2). Cell culture supernatants were collected and analyzed for IL-2 levels using enzyme-linked immunosorbent assay (ELISA) according to the manufacturer’s protocol (Thermo Fisher Scientific, Cat# BMS221-2). CRISPR-Cas9 generation of GNAS-mutant Caco-2 cell lines Several single guide RNAs (sgRNAs) targeting the GNAS locus were designed by EditCo (Redwood City, CA) with stringent off-target requirements, ensuring a minimum of two mismatches to potential off-target sites. Donor templates were designed with homology arms of sufficient length to maximize homology-directed repair efficiency. Silent mutations were incorporated into donor sequences to disrupt the protospacer adjacent motif site for each guide, preventing post-editing cleavage and enhancing knock-in efficiency. Guide RNAs were synthesized by Synthego (Redwood City, CA) as chemically modified synthetic sgRNAs, and donor templates were provided as single-stranded oligodeoxynucleotides or plasmids. Ribonucleoprotein complexes were formed by complexing sgRNAs with SpCas9 protein and delivered to Caco-2 cells along with donor templates via electroporation. A positive control sgRNA targeting RELA was included in all transfections. After electroporation, cells were allowed to recover for 48 hours before edit assessment. Editing efficiency was evaluated by PCR amplification of the targeted GNAS region followed by Sanger sequencing. Sequence data were analyzed using Synthego’s Inference of CRISPR Edits software to determine insertion/deletion frequencies and knock-in efficiency. Single cells from the edited population were isolated by limiting dilution and seeded for clonal expansion. Wells were monitored every 2-3 days with imaging to confirm clonal origin. No selection agents were employed during clone isolation. Expanded clones were validated by Sanger sequencing to confirm the presence of the desired GNAS mutations. After initial genotyping, clones were further screened to verify edit integrity. Two clones harboring the target GNAS mutations were selected for expansion and subjected to final quality control, including genotype confirmation by Sanger sequencing and mycoplasma screening using the MycoAlert Mycoplasma Detection Kit (Lonza, Basel, Switzerland). All Caco-2 cell lines were maintained in EMEM supplemented with 10% fetal bovine serum, 100 U/ml penicillin, 100 μg/ml streptomycin, and 2 mM l-glutamine at 37°C in 5% CO2. Cell lines were confirmed mycoplasma-free before experimental use. Statistical analysis Patient characteristics, frequency of IL-2 RNA percentile rank score, and proportion of each variable were summarized using descriptive statistics. Univariable logistic regression analysis was carried out to estimate the association between expressors (greater than or equal to first percentile RNA expression) versus non-expressors (less than first percentile RNA expression) of IL-2. To examine the association between any IL-2 expression and clinical and immunomic factors, multivariable logistic regression was carried out introducing the variables with univariable P < 0.2. Kaplan–Meier method was used for survival analysis. Univariable and multivariable Cox regression were carried out to assess variables associated with immunotherapy treatment. IL-2 expression was dichotomized to expressors (greater than or equal to first percentile RNA expression) and non-expressors (less than first percentile RNA expression) of IL-2. Variables with univariable P < 0.1 were analyzed in the multivariable logistic regression analysis. In both logistic regression and Cox regression analyses, genetic alterations present in at least 20 patients were analyzed. To explore the association between the categorized IL-2 RNA expression levels [low (<25th percentile), moderate (25th-74th percentile), and high (≥75th percentile)] and other variables, a linear regression model was used, with the three expression levels assigned integers from 0 to 2. Variables with univariable P < 0.2 were introduced into the subsequent multivariable linear regression model. ELISA data are reported as mean values ± standard deviation and were analyzed using Prism 8 software (GraphPad Software, Inc., La Jolla, CA). Statistical significance for three different clones was calculated via one-way analysis of variance followed by post hoc Tukey’s test for multiple comparisons. Statistical analyses were conducted using R 4.4.1 (R Foundation for Statistical Computing, Vienna, Austria), and a two-sided P < 0.05 was considered significant. Results Distribution of IL-2 RNA expression across cancer types and patient characteristics Of 514 patients with advanced or metastatic disease, 310 were women (60%), the median age was 61 years, and 489 patients had annotated data on survival outcome. Among them, 272 patients did not receive ICIs, whereas 217 patients were treated with an ICI-based regimen. ICIs were administered at a median of second line of therapy (range, first to eleventh line). The majority of patients received ICIs in combination with non-immunotherapy agents (n = 129), while ICI–ICI combinations were administered to 16 patients and single-agent ICI to 201 patients. Only 2 of 217 patients received anti-CTLA-4 as a single therapy. Of the 489 patients analyzed in the survival analysis, 437 patients had an available tissue collection date. The median time from metastatic diagnosis date to tissue collection date was 1 day (interquartile range, −4 to 205). Distribution of IL-2 RNA percentile rank is presented in Figure 1A. Of the 514 patients, 75 patients (14.6%, 75 of 514) showed high expression (≥75th percentile rank score) of IL-2, and 181 patients (35%, 181 of 514) had zero expression. Figure 1B shows the percentage of IL-2 RNA non-expressors (zero expression) in different cancer types. Among them, patients with neuroendocrine (60%), hepatobiliary (47.4%), esophageal (47.1%), carcinoma of unknown primary site (46.2%), sarcoma (45.8%), uterine (45.8%), and ovarian (41.9%) cancers had the highest proportion of IL-2 RNA non-expression.Figure 1IL-2 expression patterns. (A) Histogram of IL-2 RNA percentile rank score in tumor tissue among 514 patients. High expression (≥75th percentile rank score) shown in dark blue bars was seen in 75 patients (14.6%) whereas 181 (35%) samples had no IL-2 RNA expression shown in orange bar. (B) Percentage of IL-2 non-expressors in various cancer types among 514 patients. IL-2 expression patterns. (A) Histogram of IL-2 RNA percentile rank score in tumor tissue among 514 patients. High expression (≥75th percentile rank score) shown in dark blue bars was seen in 75 patients (14.6%) whereas 181 (35%) samples had no IL-2 RNA expression shown in orange bar. (B) Percentage of IL-2 non-expressors in various cancer types among 514 patients. Supplementary Figure S3 (available at https://doi.org/10.1016/j.esmoop.2026.107768) shows the distribution of RNA expression of IL-15, IL-2Rα, IL-2Rβ, and IL-2Rγ. We included IL-15 RNA expression in our analysis, given its shared functional roles with IL-2.18 The RNA expression of IL-15 and IL-2R subunits was relatively evenly distributed across various rank scores. Association between IL-2 expression and clinical factors, immune markers, and gene aberrations IL-2 greater than or equal to first percentile RNA expression was significantly associated with high RNA expression (≥75th percentile RNA rank) of PD-1, PD-L2, CTLA-4, BTLA, VISTA, TIM-3, IL-2Rα, IL-2Rβ, IL-2Rγ, and STAT5A as well as ERBB2 and GNAS molecular alterations in the univariable logistic regression (all P < 0.05; Table 1). In this logistic regression model (N = 514), 39 patients had ERBB2 aberrations, of which 24 had amplification only; two had amplification and mutations (one had p.R678Q; one had p.S310F and p.S335C), one had copy number gain, one had insertion (S779_P780insGS), and the other 11 had mutations only (three had p.R678Q, three had p.S310F, two had p.V842I, one had p.D769H, one had p.L313V, and one had p.V777L). A total of 27 patients had GNAS alterations including 20 with hotspot GNAS p.R201H/C missense mutations and seven with GNAS amplifications. When we introduced variables with univariable P < 0.2 into subsequent multivariable logistic regression analysis, high expressions (≥75th percentile RNA) of TIM-3 (P = 0.044), BTLA (P = 0.006), and IL-2Rγ (P = 0.045) and molecular alterations in ERBB2 (P = 0.013), GNAS (P = 0.019), and NF1 (P = 0.045) significantly correlated with IL-2 expression (Table 1).Table 1Univariable and multivariable logistic regression for the association of any IL-2 expression (greater than or equal to first percentile RNA rank score) and clinical factors, immune markers, and gene alterationsCharacteristicConditionaUnivariableMultivariablebOR95% CIPOR95% CIPAge (years)≥61 versus <611.280.89-1.840.1871.280.89-1.980.166SexMale versus female1.230.85-1.790.271PD-1 RNA≥75th versus <75th2.441.42-4.180.0012.440.61-2.590.539PD-L1 RNA≥75th versus <75th1.440.82-2.530.209PD-L2 RNA≥75th versus <75th1.811.11-2.970.0181.810.34-1.280.219CTLA-4 RNA≥75th versus <75th3.351.83-6.13<0.0013.350.58-2.690.570LAG-3 RNA≥75th versus <75th1.410.90-2.210.1311.410.45-1.380.409TIM-3 RNA≥75th versus <75th3.221.79-5.79<0.0013.221.02-4.230.044High TIM-3 expression correlates with any IL-2 expressionVISTA RNA≥75th versus <75th1.721.15-2.580.0081.720.54-1.480.664BTLA RNA≥75th versus <75th4.752.52-8.97<0.0014.751.36-6.360.006High BTLA expression correlates with any IL-2 expressionTNFRSF14 (HVEM) RNA≥75th versus <75th1.410.88-2.240.1501.410.53-1.590.749IL-2Rα RNA≥75th versus <75th2.641.59-4.38<0.0012.640.87-3.140.128IL-2Rβ RNA≥75th versus <75th2.491.54-4.01<0.0012.490.68-2.480.428IL-2Rγ RNA≥75th versus <75th2.631.71-4.06<0.0012.631.01-2.870.045High IL-2Rγ expression correlates with any IL-2 expressionIL-15 RNA≥75th versus <75th1.390.92-2.090.1131.390.65-1.750.803STAT5A RNA≥75th versus <75th1.791.14-2.830.0121.790.54-1.680.871Small intestine cancer (n = 12)Yes versus No6.150.79-48.00.0836.150.63-45.90.125Breast cancer (n = 49)Yes versus No1.030.55-1.900.936Lung cancer (n = 20)Yes versus No1.280.48-3.390.619Pancreatic cancer (n = 55)Yes versus No1.030.57-1.860.913Ovarian cancer (n = 43)Yes versus No0.740.39-1.390.342Colorectal cancer (n = 140)Yes versus No1.510.99-2.310.0551.510.86-2.360.167Uterine cancer (n = 24)Yes versus No0.630.28-1.430.268Stomach cancer (n = 25)Yes versus No1.760.69-4.500.234MAPK pathway alterationscYes versus No1.350.90-2.010.1451.350.63-1.580.984TP53 alterationsYes versus No0.980.68-1.410.918JAK/STAT pathway alterationsYes versus No0.810.13-4.910.822PI3K pathway alterationscYes versus No1.070.67-1.710.793Cyclin pathway alterationscYes versus No0.920.58-1.460.713APC alterationsYes versus No1.180.72-1.940.520ARID1A alterationsYes versus No1.030.57-1.860.913SMAD4 alterationsYes versus No0.700.38-1.270.241ERBB2 alterationsYes versus No0.490.25-0.940.0320.490.18-0.820.013ERBB2 alterations negatively correlate with any IL-2 expression (VUS excluded)MYC alterationsYes versus No0.760.35-1.620.475ATM alterationsYes versus No0.780.35-1.720.538MLL2 alterationsYes versus No1.250.51-3.110.624GNAS alterationsYes versus No10.91.45-82.00.02010.91.52-99.60.019GNAS alterations correlate with any IL-2 expression (VUS excluded)NF1 alterationsYes versus No2.230.74-6.780.1562.231.03-10.20.045NF1 alterations correlate with any IL-2 expression (VUS excluded)TMB (mut/Mb)≥10 versus <100.880.43-1.820.736TMB (mut/Mb)≥20 versus <200.470.14-1.580.223PD-L1 IHCdPositive versus negative0.890.60-1.310.552MSIHigh versus not high0.790.28-2.250.655Gene alterations in 20 or more patients were included in this analysis. Gene alterations include mutations, amplifications, deletions, rearrangements, and any other abnormalities that were not a VUS.BTLA, B and T lymphocyte attenuator; CI, confidence interval; CTLA-4, cytotoxic T-lymphocyte antigen 4; HVEM, herpesvirus entry mediator; IHC, immunohistochemistry; IL-2, interleukin-2; IL-2Rα, IL-2β, IL-2γ, interleukin-2 receptor subunits alpha, beta, and gamma; LAG-3, lymphocyte activation gene 3; MSI, microsatellite instability; mut/Mb, mutations per megabase of DNA; OR, odds ratio; PD-1, programmed cell death protein 1; PD-L1, programmed death-ligand 1; PD-L2, programmed death-ligand 2; TC, tumor cell; TIGIT, T-cell immunoglobulin and ITIM domain; TIM-3, T-cell immunoglobulin and mucin domain-containing protein 3; TMB, tumor mutational burden; TNFRSF14, tumor necrosis factor receptor superfamily member 14; VISTA, V-domain Ig suppressor of T cell activation; VUS, variant of uncertain significance.a75th refers to RNA expression rank.bVariables with univariable P < 0.2 were introduced into the subsequent multivariable model.cCyclin pathway alterations include CCND1-3, CDK4, CDK6, CDKN1B, CDKN2A, CDKN2B, CCNE1, and RB1. In MAPK pathway, MEK1 (MAP2K1), RAF (ARAF/BRAF/CRAF), and RAS (KRAS/HRAS/NRAS) molecular alterations were included. No ERK aberrations were seen. Over 85% of aberrations in the MAPK pathway are attributed to RAS mutations (134 patients). PI3K pathway alterations included PIK3CA, PTEN, and AKT alterations.dPD-L1 IHC was deemed positive if the combined positive score was ≥1% using 22C3, the score of tumor-infiltrating immune cells was ≥1% with SP142, or the score of TCs was ≥1% using 28-8. Table 1Univariable and multivariable logistic regression for the association of any IL-2 expression (greater than or equal to first percentile RNA rank score) and clinical factors, immune markers, and gene alterationsCharacteristicConditionaUnivariableMultivariablebOR95% CIPOR95% CIPAge (years)≥61 versus <611.280.89-1.840.1871.280.89-1.980.166SexMale versus female1.230.85-1.790.271PD-1 RNA≥75th versus <75th2.441.42-4.180.0012.440.61-2.590.539PD-L1 RNA≥75th versus <75th1.440.82-2.530.209PD-L2 RNA≥75th versus <75th1.811.11-2.970.0181.810.34-1.280.219CTLA-4 RNA≥75th versus <75th3.351.83-6.13<0.0013.350.58-2.690.570LAG-3 RNA≥75th versus <75th1.410.90-2.210.1311.410.45-1.380.409TIM-3 RNA≥75th versus <75th3.221.79-5.79<0.0013.221.02-4.230.044High TIM-3 expression correlates with any IL-2 expressionVISTA RNA≥75th versus <75th1.721.15-2.580.0081.720.54-1.480.664BTLA RNA≥75th versus <75th4.752.52-8.97<0.0014.751.36-6.360.006High BTLA expression correlates with any IL-2 expressionTNFRSF14 (HVEM) RNA≥75th versus <75th1.410.88-2.240.1501.410.53-1.590.749IL-2Rα RNA≥75th versus <75th2.641.59-4.38<0.0012.640.87-3.140.128IL-2Rβ RNA≥75th versus <75th2.491.54-4.01<0.0012.490.68-2.480.428IL-2Rγ RNA≥75th versus <75th2.631.71-4.06<0.0012.631.01-2.870.045High IL-2Rγ expression correlates with any IL-2 expressionIL-15 RNA≥75th versus <75th1.390.92-2.090.1131.390.65-1.750.803STAT5A RNA≥75th versus <75th1.791.14-2.830.0121.790.54-1.680.871Small intestine cancer (n = 12)Yes versus No6.150.79-48.00.0836.150.63-45.90.125Breast cancer (n = 49)Yes versus No1.030.55-1.900.936Lung cancer (n = 20)Yes versus No1.280.48-3.390.619Pancreatic cancer (n = 55)Yes versus No1.030.57-1.860.913Ovarian cancer (n = 43)Yes versus No0.740.39-1.390.342Colorectal cancer (n = 140)Yes versus No1.510.99-2.310.0551.510.86-2.360.167Uterine cancer (n = 24)Yes versus No0.630.28-1.430.268Stomach cancer (n = 25)Yes versus No1.760.69-4.500.234MAPK pathway alterationscYes versus No1.350.90-2.010.1451.350.63-1.580.984TP53 alterationsYes versus No0.980.68-1.410.918JAK/STAT pathway alterationsYes versus No0.810.13-4.910.822PI3K pathway alterationscYes versus No1.070.67-1.710.793Cyclin pathway alterationscYes versus No0.920.58-1.460.713APC alterationsYes versus No1.180.72-1.940.520ARID1A alterationsYes versus No1.030.57-1.860.913SMAD4 alterationsYes versus No0.700.38-1.270.241ERBB2 alterationsYes versus No0.490.25-0.940.0320.490.18-0.820.013ERBB2 alterations negatively correlate with any IL-2 expression (VUS excluded)MYC alterationsYes versus No0.760.35-1.620.475ATM alterationsYes versus No0.780.35-1.720.538MLL2 alterationsYes versus No1.250.51-3.110.624GNAS alterationsYes versus No10.91.45-82.00.02010.91.52-99.60.019GNAS alterations correlate with any IL-2 expression (VUS excluded)NF1 alterationsYes versus No2.230.74-6.780.1562.231.03-10.20.045NF1 alterations correlate with any IL-2 expression (VUS excluded)TMB (mut/Mb)≥10 versus <100.880.43-1.820.736TMB (mut/Mb)≥20 versus <200.470.14-1.580.223PD-L1 IHCdPositive versus negative0.890.60-1.310.552MSIHigh versus not high0.790.28-2.250.655Gene alterations in 20 or more patients were included in this analysis. Gene alterations include mutations, amplifications, deletions, rearrangements, and any other abnormalities that were not a VUS.BTLA, B and T lymphocyte attenuator; CI, confidence interval; CTLA-4, cytotoxic T-lymphocyte antigen 4; HVEM, herpesvirus entry mediator; IHC, immunohistochemistry; IL-2, interleukin-2; IL-2Rα, IL-2β, IL-2γ, interleukin-2 receptor subunits alpha, beta, and gamma; LAG-3, lymphocyte activation gene 3; MSI, microsatellite instability; mut/Mb, mutations per megabase of DNA; OR, odds ratio; PD-1, programmed cell death protein 1; PD-L1, programmed death-ligand 1; PD-L2, programmed death-ligand 2; TC, tumor cell; TIGIT, T-cell immunoglobulin and ITIM domain; TIM-3, T-cell immunoglobulin and mucin domain-containing protein 3; TMB, tumor mutational burden; TNFRSF14, tumor necrosis factor receptor superfamily member 14; VISTA, V-domain Ig suppressor of T cell activation; VUS, variant of uncertain significance.a75th refers to RNA expression rank.bVariables with univariable P < 0.2 were introduced into the subsequent multivariable model.cCyclin pathway alterations include CCND1-3, CDK4, CDK6, CDKN1B, CDKN2A, CDKN2B, CCNE1, and RB1. In MAPK pathway, MEK1 (MAP2K1), RAF (ARAF/BRAF/CRAF), and RAS (KRAS/HRAS/NRAS) molecular alterations were included. No ERK aberrations were seen. Over 85% of aberrations in the MAPK pathway are attributed to RAS mutations (134 patients). PI3K pathway alterations included PIK3CA, PTEN, and AKT alterations.dPD-L1 IHC was deemed positive if the combined positive score was ≥1% using 22C3, the score of tumor-infiltrating immune cells was ≥1% with SP142, or the score of TCs was ≥1% using 28-8. Univariable and multivariable logistic regression for the association of any IL-2 expression (greater than or equal to first percentile RNA rank score) and clinical factors, immune markers, and gene alterations Gene alterations in 20 or more patients were included in this analysis. Gene alterations include mutations, amplifications, deletions, rearrangements, and any other abnormalities that were not a VUS. BTLA, B and T lymphocyte attenuator; CI, confidence interval; CTLA-4, cytotoxic T-lymphocyte antigen 4; HVEM, herpesvirus entry mediator; IHC, immunohistochemistry; IL-2, interleukin-2; IL-2Rα, IL-2β, IL-2γ, interleukin-2 receptor subunits alpha, beta, and gamma; LAG-3, lymphocyte activation gene 3; MSI, microsatellite instability; mut/Mb, mutations per megabase of DNA; OR, odds ratio; PD-1, programmed cell death protein 1; PD-L1, programmed death-ligand 1; PD-L2, programmed death-ligand 2; TC, tumor cell; TIGIT, T-cell immunoglobulin and ITIM domain; TIM-3, T-cell immunoglobulin and mucin domain-containing protein 3; TMB, tumor mutational burden; TNFRSF14, tumor necrosis factor receptor superfamily member 14; VISTA, V-domain Ig suppressor of T cell activation; VUS, variant of uncertain significance. 75th refers to RNA expression rank. Variables with univariable P < 0.2 were introduced into the subsequent multivariable model. Cyclin pathway alterations include CCND1-3, CDK4, CDK6, CDKN1B, CDKN2A, CDKN2B, CCNE1, and RB1. In MAPK pathway, MEK1 (MAP2K1), RAF (ARAF/BRAF/CRAF), and RAS (KRAS/HRAS/NRAS) molecular alterations were included. No ERK aberrations were seen. Over 85% of aberrations in the MAPK pathway are attributed to RAS mutations (134 patients). PI3K pathway alterations included PIK3CA, PTEN, and AKT alterations. PD-L1 IHC was deemed positive if the combined positive score was ≥1% using 22C3, the score of tumor-infiltrating immune cells was ≥1% with SP142, or the score of TCs was ≥1% using 28-8. Overall survival correlation with IL-2 expression, clinical factors, immune markers, and gene alterations On univariable Cox regression analysis, IL-2 greater than or equal to first percentile RNA rank; CTLA-4, LAG-3, PD-1, and IL-15 ≥75th percentile RNA rank; and diagnoses of sarcoma and ovarian cancer were associated with longer overall survival (OS; P < 0.05). Diagnosis of pancreatic and stomach cancers and MAPK pathway and TP53 alterations were associated with shorter OS (P < 0.05; Table 2 and Supplementary Figure S4, available at https://doi.org/10.1016/j.esmoop.2026.107768). However, when we introduced variables with univariable P < 0.1 into subsequent multivariable Cox regression analysis, age ≥61 years [hazard ratio (HR) 1.31, 95% confidence interval (CI) 1.02-1.67, P = 0.03] and pancreatic (HR 1.69, 95% CI 1.12-2.54, P = 0.01) and stomach cancers (HR 2.29, 95% CI 1.29-4.04, P = 0.004) were significantly associated with shorter OS among all patients. In addition, IL-2 ≥1st percentile RNA rank (HR 0.74, 95% CI 0.57-0.96, P = 0.02), sarcomas (HR 0.50, 95% CI 0.26-0.97, P = 0.04), and ovarian cancer (HR 0.58, 95% CI 0.37-0.92, P = 0.02) were significantly associated with longer OS among all patients.Table 2Univariable and multivariable Cox regression analysis (N = 489) for correlations between various factors and OSVariableConditionNUnivariableMultivariableaHRCIPHRCIPAge (years)<61248≥612411.250.98-1.590.071.311.02-1.670.03Older age correlates with shorter survivalSexFemale293Male1961.070.84-1.370.56IL-2b<1st175≥1st3140.760.59-0.970.030.740.57-0.960.02Any IL-2 RNA expression correlates with longer survivalCTLA-4b<75thc409≥75th800.690.48-0.990.0450.800.50-1.280.36LAG-3b<75th381≥75th1080.720.53-0.990.041.000.68-1.46>0.99PD-L1b<75th427≥75th621.060.71-1.570.78PD-L2b<75th394≥75th950.800.57-1.120.19PD-1b<75th403≥75th860.650.46-0.930.020.830.53-1.290.40CD4b<75th387≥75th1021.000.73-1.360.97VISTAb<75th330≥75th1591.110.85-1.440.44TIM-3b<75th402≥75th870.760.54-1.060.11BTLAb<75th402≥75th870.730.52-1.030.070.880.57-1.350.55TNFRSF14 (HVEM)b<75th390≥75th990.850.62-1.170.32IL-15b<75th345≥75th1440.720.54-0.950.020.890.66-1.200.45IL-2Rαb<75th387≥75th1020.990.72-1.350.94IL-2Rβb<75th375≥75th1140.920.68-1.250.60IL-2Rγb<75th336≥75th1531.030.80-1.330.82STAT6b<75th313≥75th1760.990.77-1.280.96SarcomaNo465Yes240.490.26-0.930.030.500.26-0.970.04Sarcomas correlate with longer survivalNeuroendocrine tumorNo474Yes150.450.19-1.100.080.520.21-1.320.17Small intestine cancerNo478Yes111.520.62-3.680.36Breast cancerNo446Yes430.690.43-1.110.13Lung cancerNo469Yes201.190.68-2.090.54Pancreatic cancerNo437Yes522.271.61-3.21<0.0011.691.12-2.540.01Pancreatic cancer correlates with shorter survivalOvarian cancerNo447Yes420.610.40-0.940.030.580.37-0.920.02Ovarian cancer correlates with longer survivalColorectal cancerNo356Yes1331.030.79-1.340.85Uterine cancerNo467Yes221.090.60-2.000.77Stomach cancerNo466Yes232.071.21-3.560.012.291.29-4.040.004Stomach cancer correlates with shorter survivalTMB (mut/Mb)<10397≥10310.800.48-1.340.40PD-L1 IHCNegative344Positive1441.240.95-1.630.12MSINot high443High130.450.17-1.210.12MAPK pathway alterationsd,eNo339Yes1501.321.03-1.710.031.170.88-1.550.29TP53 alterationsdNo249Yes2401.321.04-1.680.031.170.90-1.510.25PI3K pathway alterationsd,eNo401Yes880.990.72-1.350.94JAK pathway alterationsdNo484Yes51.260.47-3.380.65Cyclin pathway alterationd,eNo396Yes931.591.18-2.150.0021.310.94-1.830.11APC alterationsdNo410Yes790.970.70-1.340.834ARID1A alterationsdNo438Yes510.950.64-1.410.797SMAD4 alterationsdNo442Yes471.641.13-2.390.011.270.86-1.870.23ERBB2 alterationsdNo452Yes370.940.59-1.480.777MYC alterationsdNo460Yes291.771.11-2.830.0171.520.94-2.460.09ATM alterationsdNo463Yes261.170.71-1.940.542MLL2 alterationsdNo469Yes200.730.38-1.430.361GNAS alterationsdNo469Yes200.820.46-1.460.497NF1 alterationsdNo472Yes171.000.53-1.870.988ImmunotherapyNo272Yes2171.100.87-1.410.42Gene alterations in 20 or more patients were included in this analysis. Gene alterations include mutations, amplifications, deletions, rearrangements and any other abnormalities that were not a VUS. OS is defined as the date of advanced/metastatic disease to the date of the last follow-up (see also Supplementary Figure S4, available at https://doi.org/10.1016/j.esmoop.2026.107768).BTLA, B and T lymphocyte attenuator; CI, confidence interval; CTLA-4, cytotoxic T-lymphocyte antigen 4; HR, hazard ratio; HVEM, herpesvirus entry mediator; IHC, immunohistochemistry; IL-2, interleukin-2; IL-2Rα, IL-2β, IL-2γ, interleukin-2 receptor subunits alpha, beta, and gamma; LAG-3, lymphocyte activation gene 3; MSI, microsatellite instability; mut/Mb, mutations per megabase of DNA; OR, odds ratio; OS, overall survival; PD-1, programmed cell death protein 1; PD-L1, programmed death-ligand 1; PD-L2, programmed death-ligand 2; TIGIT, T-cell immunoglobulin and ITIM domain; TIM-3, T-cell immunoglobulin and mucin domain-containing protein 3; TMB, tumor mutational burden; TNFRSF14, tumor necrosis factor receptor superfamily member 14; VISTA, V-domain Ig suppressor of T cell activation; VUS, variant of uncertain significance.aVariables with univariable P < 0.1 were introduced into multivariable analysis.bRNA percentile rank scores: ≥1st for IL-2 implies ≥1st percentile RNA expression rank, and ≥75th for variables refers to ≥75th percentile RNA expression rank.c75th refers to RNA expression rank.dGene alterations include mutations, amplifications, deletions, rearrangements, and any other abnormalities that were not a VUS. Gene alterations in 20 or more patients were included in this analysis.eCyclin pathway alterations include CCND1-3, CDK4, CDK6, CDKN1B, CDKN2A, CDKN2B, CCNE1, and RB1. In MAPK pathway, MEK1 (MAP2K1), RAF (ARAF/BRAF/CRAF), and RAS (KRAS/HRAS/NRAS) molecular alterations were included. No ERK aberrations were seen. Over 85% of aberrations in the MAPK pathway are attributed to RAS mutations (134 patients). PI3K pathway alterations included PIK3CA, PTEN, and AKT alterations. Table 2Univariable and multivariable Cox regression analysis (N = 489) for correlations between various factors and OSVariableConditionNUnivariableMultivariableaHRCIPHRCIPAge (years)<61248≥612411.250.98-1.590.071.311.02-1.670.03Older age correlates with shorter survivalSexFemale293Male1961.070.84-1.370.56IL-2b<1st175≥1st3140.760.59-0.970.030.740.57-0.960.02Any IL-2 RNA expression correlates with longer survivalCTLA-4b<75thc409≥75th800.690.48-0.990.0450.800.50-1.280.36LAG-3b<75th381≥75th1080.720.53-0.990.041.000.68-1.46>0.99PD-L1b<75th427≥75th621.060.71-1.570.78PD-L2b<75th394≥75th950.800.57-1.120.19PD-1b<75th403≥75th860.650.46-0.930.020.830.53-1.290.40CD4b<75th387≥75th1021.000.73-1.360.97VISTAb<75th330≥75th1591.110.85-1.440.44TIM-3b<75th402≥75th870.760.54-1.060.11BTLAb<75th402≥75th870.730.52-1.030.070.880.57-1.350.55TNFRSF14 (HVEM)b<75th390≥75th990.850.62-1.170.32IL-15b<75th345≥75th1440.720.54-0.950.020.890.66-1.200.45IL-2Rαb<75th387≥75th1020.990.72-1.350.94IL-2Rβb<75th375≥75th1140.920.68-1.250.60IL-2Rγb<75th336≥75th1531.030.80-1.330.82STAT6b<75th313≥75th1760.990.77-1.280.96SarcomaNo465Yes240.490.26-0.930.030.500.26-0.970.04Sarcomas correlate with longer survivalNeuroendocrine tumorNo474Yes150.450.19-1.100.080.520.21-1.320.17Small intestine cancerNo478Yes111.520.62-3.680.36Breast cancerNo446Yes430.690.43-1.110.13Lung cancerNo469Yes201.190.68-2.090.54Pancreatic cancerNo437Yes522.271.61-3.21<0.0011.691.12-2.540.01Pancreatic cancer correlates with shorter survivalOvarian cancerNo447Yes420.610.40-0.940.030.580.37-0.920.02Ovarian cancer correlates with longer survivalColorectal cancerNo356Yes1331.030.79-1.340.85Uterine cancerNo467Yes221.090.60-2.000.77Stomach cancerNo466Yes232.071.21-3.560.012.291.29-4.040.004Stomach cancer correlates with shorter survivalTMB (mut/Mb)<10397≥10310.800.48-1.340.40PD-L1 IHCNegative344Positive1441.240.95-1.630.12MSINot high443High130.450.17-1.210.12MAPK pathway alterationsd,eNo339Yes1501.321.03-1.710.031.170.88-1.550.29TP53 alterationsdNo249Yes2401.321.04-1.680.031.170.90-1.510.25PI3K pathway alterationsd,eNo401Yes880.990.72-1.350.94JAK pathway alterationsdNo484Yes51.260.47-3.380.65Cyclin pathway alterationd,eNo396Yes931.591.18-2.150.0021.310.94-1.830.11APC alterationsdNo410Yes790.970.70-1.340.834ARID1A alterationsdNo438Yes510.950.64-1.410.797SMAD4 alterationsdNo442Yes471.641.13-2.390.011.270.86-1.870.23ERBB2 alterationsdNo452Yes370.940.59-1.480.777MYC alterationsdNo460Yes291.771.11-2.830.0171.520.94-2.460.09ATM alterationsdNo463Yes261.170.71-1.940.542MLL2 alterationsdNo469Yes200.730.38-1.430.361GNAS alterationsdNo469Yes200.820.46-1.460.497NF1 alterationsdNo472Yes171.000.53-1.870.988ImmunotherapyNo272Yes2171.100.87-1.410.42Gene alterations in 20 or more patients were included in this analysis. Gene alterations include mutations, amplifications, deletions, rearrangements and any other abnormalities that were not a VUS. OS is defined as the date of advanced/metastatic disease to the date of the last follow-up (see also Supplementary Figure S4, available at https://doi.org/10.1016/j.esmoop.2026.107768).BTLA, B and T lymphocyte attenuator; CI, confidence interval; CTLA-4, cytotoxic T-lymphocyte antigen 4; HR, hazard ratio; HVEM, herpesvirus entry mediator; IHC, immunohistochemistry; IL-2, interleukin-2; IL-2Rα, IL-2β, IL-2γ, interleukin-2 receptor subunits alpha, beta, and gamma; LAG-3, lymphocyte activation gene 3; MSI, microsatellite instability; mut/Mb, mutations per megabase of DNA; OR, odds ratio; OS, overall survival; PD-1, programmed cell death protein 1; PD-L1, programmed death-ligand 1; PD-L2, programmed death-ligand 2; TIGIT, T-cell immunoglobulin and ITIM domain; TIM-3, T-cell immunoglobulin and mucin domain-containing protein 3; TMB, tumor mutational burden; TNFRSF14, tumor necrosis factor receptor superfamily member 14; VISTA, V-domain Ig suppressor of T cell activation; VUS, variant of uncertain significance.aVariables with univariable P < 0.1 were introduced into multivariable analysis.bRNA percentile rank scores: ≥1st for IL-2 implies ≥1st percentile RNA expression rank, and ≥75th for variables refers to ≥75th percentile RNA expression rank.c75th refers to RNA expression rank.dGene alterations include mutations, amplifications, deletions, rearrangements, and any other abnormalities that were not a VUS. Gene alterations in 20 or more patients were included in this analysis.eCyclin pathway alterations include CCND1-3, CDK4, CDK6, CDKN1B, CDKN2A, CDKN2B, CCNE1, and RB1. In MAPK pathway, MEK1 (MAP2K1), RAF (ARAF/BRAF/CRAF), and RAS (KRAS/HRAS/NRAS) molecular alterations were included. No ERK aberrations were seen. Over 85% of aberrations in the MAPK pathway are attributed to RAS mutations (134 patients). PI3K pathway alterations included PIK3CA, PTEN, and AKT alterations. Univariable and multivariable Cox regression analysis (N = 489) for correlations between various factors and OS Gene alterations in 20 or more patients were included in this analysis. Gene alterations include mutations, amplifications, deletions, rearrangements and any other abnormalities that were not a VUS. OS is defined as the date of advanced/metastatic disease to the date of the last follow-up (see also Supplementary Figure S4, available at https://doi.org/10.1016/j.esmoop.2026.107768). BTLA, B and T lymphocyte attenuator; CI, confidence interval; CTLA-4, cytotoxic T-lymphocyte antigen 4; HR, hazard ratio; HVEM, herpesvirus entry mediator; IHC, immunohistochemistry; IL-2, interleukin-2; IL-2Rα, IL-2β, IL-2γ, interleukin-2 receptor subunits alpha, beta, and gamma; LAG-3, lymphocyte activation gene 3; MSI, microsatellite instability; mut/Mb, mutations per megabase of DNA; OR, odds ratio; OS, overall survival; PD-1, programmed cell death protein 1; PD-L1, programmed death-ligand 1; PD-L2, programmed death-ligand 2; TIGIT, T-cell immunoglobulin and ITIM domain; TIM-3, T-cell immunoglobulin and mucin domain-containing protein 3; TMB, tumor mutational burden; TNFRSF14, tumor necrosis factor receptor superfamily member 14; VISTA, V-domain Ig suppressor of T cell activation; VUS, variant of uncertain significance. Variables with univariable P < 0.1 were introduced into multivariable analysis. RNA percentile rank scores: ≥1st for IL-2 implies ≥1st percentile RNA expression rank, and ≥75th for variables refers to ≥75th percentile RNA expression rank. 75th refers to RNA expression rank. Gene alterations include mutations, amplifications, deletions, rearrangements, and any other abnormalities that were not a VUS. Gene alterations in 20 or more patients were included in this analysis. Cyclin pathway alterations include CCND1-3, CDK4, CDK6, CDKN1B, CDKN2A, CDKN2B, CCNE1, and RB1. In MAPK pathway, MEK1 (MAP2K1), RAF (ARAF/BRAF/CRAF), and RAS (KRAS/HRAS/NRAS) molecular alterations were included. No ERK aberrations were seen. Over 85% of aberrations in the MAPK pathway are attributed to RAS mutations (134 patients). PI3K pathway alterations included PIK3CA, PTEN, and AKT alterations. IL-2 RNA expression is an independent prognostic factor for longer OS across cancers Kaplan–Meier survival analysis demonstrated that, among the 489 patients, those whose tumors expressed IL-2 transcripts showed significantly longer OS from the date of diagnosis in all patients (N = 489; log-rank P = 0.028; Figure 2A). In the subgroup of patients who never received immunotherapy, IL-2 RNA expression showed numerically longer median OS compared with those with any IL-2 RNA expression (3.63 versus 2.63 years), but the difference was not statistically significant (N = 272; log-rank P = 0.24; Figure 2B). As mentioned above, IL-2 expression greater than or equal to first percentile RNA rank was significantly and independently associated with longer OS among all patients in the UCSD cohort (HR 0.74, 95% CI 0.57-0.96, P = 0.02; Table 2 and Supplementary Figure S4, available at https://doi.org/10.1016/j.esmoop.2026.107768). To account for immortal bias, we carried out a left-truncated OS analysis (Supplementary Figure S5, available at https://doi.org/10.1016/j.esmoop.2026.107768) among 489 patients. This analysis showed a statistically significant but shorter median survival time (HR 0.78, 95% CI 0.61-0.997, P = 0.047) compared with the unadjusted analysis (Figure 2A).Figure 2Survival analysis (UCSD and TCGA datasets) and linear regression analysis between higher IL-2 RNA expression and pathogenic GNAS alterations. (A) Median OS (calculated from the date of advanced/metastatic disease to the last follow-up date or death) among all patients (N = 489) in the UCSD dataset was 3.63 years (95% CI 3.41-4.37) for IL-2 expressors and 2.59 years (95% CI 2.17-3.68) for IL-2 non-expressors (log-rank P = 0.028; see also Supplementary Figure S5, available at https://doi.org/10.1016/j.esmoop.2026.107768). (B) Median OS for patients who did not receive immunotherapy (n = 272) was 3.63 years (95% CI 3.41-5.28) for IL-2 expressors and 2.63 years (95% CI 2.08-4.07) for IL-2 non-expressors (log-rank P = 0.24). Among patients who received immunotherapy (N = 217), (C) median PFS (defined as the date of cycle 1 day 1 of immunotherapy to the date of earliest occurrence of radiological or clinical disease progression or death) was 0.39 years (95% CI 0.31-0.56) for IL-2 expressors and 0.38 years (95% CI 0.31-0.48) for IL-2 non-expressors (log-rank P = 0.25), and (D) median OS was 1.38 years (95% CI 1.21-2.23) for IL-2 expressors and 1.36 years (95% CI 0.95-1.81) for IL-2 non-expressors (log-rank P = 0.16). Among the TCGA pan-cancer cohort, (E) IL-2 RNA expressors (≥1% RNA percentile rank) had longer OS compared with non-expressors (HR 0.84, 95% CI 0.79-0.91, P < 0.001). (F) OS was longer for high (≥50th percentile RNA rank) versus low (<50th percentile RNA rank) RNA expressors (HR 0.85, 95% CI 0.79-0.91, P < 0.001). (G) Longer OS for high (≥75th percentile RNA rank) RNA expressors versus low (<75th percentile RNA rank) expressors (HR 0.79, 95% CI 0.72-0.86, P < 0.001). (H) Linear regression analysis bubble plot showing correlations between higher IL-2 RNA expression and GNAS pathogenic alterations. The bold blue line represents a univariable regression line for the association between IL-2 RNA expression levels and GNAS alterations (y = 0.48x + 0.57). The blue symbol X indicates the mean value of IL-2 expression levels for each GNAS aberration status. β refers to the slope of the regression line, representing the difference in mean values of IL-2 expression level between GNAS alteration statuses. The regression slope was 0.48 (95% CI 0.15-0.80, P = 0.004) in the univariable analysis and 0.38 (95% CI 0.07-0.68, P = 0.02) in the multivariable analysis (see also Table 3). (I) IL-2 secretion in GNAS-wt and GNASR201C isogenic Caco-2 cell lines (n = 8 samples per clone). Raw optical density measurements are shown for each sample. Statistical analysis was carried out using one-way analysis of variance with Tukey’s post hoc test for multiple comparisons. ∗P < 0.05. CI, confidence interval; HR, hazard ratio; OS, overall survival; PFS, progression-free survival; TCGA, The Cancer Genome Atlas; UCSD, University of California San Diego. Survival analysis (UCSD and TCGA datasets) and linear regression analysis between higher IL-2 RNA expression and pathogenic GNAS alterations. (A) Median OS (calculated from the date of advanced/metastatic disease to the last follow-up date or death) among all patients (N = 489) in the UCSD dataset was 3.63 years (95% CI 3.41-4.37) for IL-2 expressors and 2.59 years (95% CI 2.17-3.68) for IL-2 non-expressors (log-rank P = 0.028; see also Supplementary Figure S5, available at https://doi.org/10.1016/j.esmoop.2026.107768). (B) Median OS for patients who did not receive immunotherapy (n = 272) was 3.63 years (95% CI 3.41-5.28) for IL-2 expressors and 2.63 years (95% CI 2.08-4.07) for IL-2 non-expressors (log-rank P = 0.24). Among patients who received immunotherapy (N = 217), (C) median PFS (defined as the date of cycle 1 day 1 of immunotherapy to the date of earliest occurrence of radiological or clinical disease progression or death) was 0.39 years (95% CI 0.31-0.56) for IL-2 expressors and 0.38 years (95% CI 0.31-0.48) for IL-2 non-expressors (log-rank P = 0.25), and (D) median OS was 1.38 years (95% CI 1.21-2.23) for IL-2 expressors and 1.36 years (95% CI 0.95-1.81) for IL-2 non-expressors (log-rank P = 0.16). Among the TCGA pan-cancer cohort, (E) IL-2 RNA expressors (≥1% RNA percentile rank) had longer OS compared with non-expressors (HR 0.84, 95% CI 0.79-0.91, P < 0.001). (F) OS was longer for high (≥50th percentile RNA rank) versus low (<50th percentile RNA rank) RNA expressors (HR 0.85, 95% CI 0.79-0.91, P < 0.001). (G) Longer OS for high (≥75th percentile RNA rank) RNA expressors versus low (<75th percentile RNA rank) expressors (HR 0.79, 95% CI 0.72-0.86, P < 0.001). (H) Linear regression analysis bubble plot showing correlations between higher IL-2 RNA expression and GNAS pathogenic alterations. The bold blue line represents a univariable regression line for the association between IL-2 RNA expression levels and GNAS alterations (y = 0.48x + 0.57). The blue symbol X indicates the mean value of IL-2 expression levels for each GNAS aberration status. β refers to the slope of the regression line, representing the difference in mean values of IL-2 expression level between GNAS alteration statuses. The regression slope was 0.48 (95% CI 0.15-0.80, P = 0.004) in the univariable analysis and 0.38 (95% CI 0.07-0.68, P = 0.02) in the multivariable analysis (see also Table 3). (I) IL-2 secretion in GNAS-wt and GNASR201C isogenic Caco-2 cell lines (n = 8 samples per clone). Raw optical density measurements are shown for each sample. Statistical analysis was carried out using one-way analysis of variance with Tukey’s post hoc test for multiple comparisons. ∗P < 0.05. CI, confidence interval; HR, hazard ratio; OS, overall survival; PFS, progression-free survival; TCGA, The Cancer Genome Atlas; UCSD, University of California San Diego. From the large TCGA dataset analysis (Supplementary Figures S1 and S2, available at https://doi.org/10.1016/j.esmoop.2026.107768), 47% of patients (4613 of 9869) had zero values for IL-2 RNA expression. Survival analysis demonstrated longer OS among IL-2 expressors (any IL-2 RNA expression greater than zero) compared with non-expressors (HR 0.84, 95% CI 0.79-0.91, P < 0.0001), with median OS being 92.9 months (95% CI 86.6-103.2) versus 81.1 months (95% CI 72.0-88.2; Figure 2E); OS was also longer for high (≥50th percentile) compared with low (<50th percentile) RNA expressors [HR 0.85, 95% CI 0.79-0.91, P < 0.0001, median OS 96.2 months (95% CI 86.6–104.6) versus 81.1 months (95% CI 72.0-88.1); Figure 2F], and also longer for high (≥75th percentile) compared with low (<75th percentile) RNA expressors [HR 0.79, 95% CI 0.72-0.86, P < 0.0001, median OS 104.6 months (95% CI 95.6-148.1) versus 81.1 months (95% CI 76.6-88.1); Figure 2G]. A Cox regression model with a random intercept of cancer type showed that, in the pan-cancer setting (TCGA cohort), any IL-2 expressor had a longer survival (HR 0.89, 95% CI 0.82-0.96, P = 0.004) compared with zero IL-2 expressors. The pooled HR for all cancers was 0.91 (95% CI 0.84-0.99, P < 0.05; Supplementary Figure S6, available at https://doi.org/10.1016/j.esmoop.2026.107768). IL-2 RNA expression was not predictive of survival outcome among patients who received ICIs IL-2 expression (≥1% versus <1% RNA percentile rank) did not predict outcome [either progression-free survival (PFS) or OS] among patients who received immunotherapy (n = 217, from the start of immunotherapy, P = 0.25 and 0.16, respectively; Figure 2C and D). Given that the immunotherapy-treated subset included only 217 patients, an interaction analysis (using Cox model with delayed entry and time-dependent immunotherapy exposure) may have had limited statistical power to detect a modest predictive effect of IL-2 status. Affinity based on IL-2R RNA expression and its association with Treg markers Patients with a signature of a high-affinity IL-2R (IL-2Rα, IL-2Rβ, and IL-2Rγ all being ≥75th percentile RNA rank; Supplementary Table S1, available at https://doi.org/10.1016/j.esmoop.2026.107768) tended to show higher levels of Treg marker expression (CD3, CD4, IL-2Rα, IL-7R, FOXP3, and PTPRC), but not MKI67, a marker of cellular proliferation (Supplementary Figure S7, available at https://doi.org/10.1016/j.esmoop.2026.107768).19,20 Further, it was noted that the expression of Treg markers (excluding MK167) correlated positively with IL-2R affinity (data not shown). However, the signature of high-affinity IL-2R (IL-2Rα, IL-2Rβ, and IL-2Rγ all being ≥75th percentile RNA rank) was not a prognostic factor across cancers among all patients (data not shown). GNAS molecular alterations were positively associated with IL-2 expression levels (IL-2 expression low 0-24th, moderate 25th-74th, high ≥75th percentile) We categorized IL-2, IL-2 receptor subunits, and STAT5A expression levels into ‘high’ (≥75th percentile RNA rank), ‘moderate’ (25th-74th percentile RNA rank), and ‘low’ (0-24th percentile RNA rank), assuming these categories as integers (2, 1, and 0, respectively) to model expression linearly (see also ‘Patients and Methods’ section). In the univariable linear regression analysis, the presence of gastrointestinal cancer, IL-2R subunits, and STAT5A expression ≥75th percentile RNA rank and presence of MAPK pathway and GNAS alterations were associated with higher expression of IL-2 (P < 0.05; Table 3). For multivariable linear regression analysis, we included molecular alterations present in 20 or more patients in our dataset. In the multivariable linear model, the expression levels of IL-2Rα, IL-2Rβ, and IL-2Rγ, and STAT5A demonstrated a significant association with the expression levels of IL-2, similarly categorized. Additionally, GNAS aberrations showed a significant association with the expression levels of IL-2 (β 0.38, 95% CI 0.07-0.68, P = 0.02; Table 3 and Figure 2H).Table 3Univariable and multivariable linear regression analysis comparing variables shown with IL-2 RNA expression levelsVariableNUnivariableMultivariableaβ95% CIPβ95% CIPAge (1: ≥61 years versus 0: <61 years)256 versus 2580.06−0.07, 0.180.39Sex (1: Male versus 0: Female)204 versus 3100.02−0.11, 0.150.81GI cancer (1: Yes versus 0: No)272 versus 2420.130.00, 0.260.045−0.08−0.21, 0.060.27IL-2Rα (2: high versus 1: moderate versus 0: low)111 versus 270 versus 1330.320.23, 0.41<0.0010.110.00, 0.210.04 (When IL-2 increases by one unit, IL-2Rα increases significantly by 0.11 units on average)IL-2Rβ (2: high versus 1: moderate versus 0: low)124 versus 272 versus 1180.350.27, 0.44<0.0010.170.06, 0.270.002 (When IL-2Rβ levels increase by one unit, IL-2 expression levels significantly increase by 0.17 units on average)IL-2Rγ (2: high versus 1: moderate versus 0: low)160 versus 233 versus 1210.310.22, 0.39<0.0010.170.06, 0.270.001 (When IL-2Rγ increases by one unit, IL-2 expression levels increase significantly by 0.17 units on average)STAT5A (2: high versus 1: moderate versus 0: low)b121 versus 274 versus 1190.340.25, 0.42<0.0010.160.06, 0.250.002 (When STAT5A increases by one unit, IL-2 expression levels increase significantly by 0.16 units on average)MAPK pathway aberrations (1: Yes versus 0: No)b,c127 versus 3570.190.05, 0.330.0060.10−0.03, 0.240.13TP53 aberration (1: Yes versus 0: No)b254 versus 260−0.11−0.24, 0.010.08−0.08−0.20, 0.040.18PI3K pathway aberration (1: Yes versus 0: No)b,c94 versus 4200.07−0.10, 0.230.42JAK pathway aberration (1: Yes versus 0: No)b5 versus 509−0.20−0.84, 0.450.55Cyclin-related alteration (1: Yes versus 0: No)b,c120 versus 394−0.03−0.18, 0.110.65APC aberration (1: Yes versus 0: No)b84 versus 430−0.08−0.25, 0.090.34ARID1A aberration (1: Yes versus 0: No)b55 versus 4590.01−0.20, 0.210.94SMAD4 aberration (1: Yes versus 0: No)b49 versus 465−0.14−0.35, 0.080.21ERBB2 aberration (1: Yes versus 0: No)b39 versus 475−0.17−0.41, 0.070.16−0.14−0.36, 0.080.22MYC aberration (1: Yes versus 0: No)b29 versus 485−0.15−0.43, 0.120.27ATM aberration (1: Yes versus 0: No)b27 versus 487−0.08−0.36, 0.200.58MLL2 aberration (1: Yes versus 0: No)b23 versus 4910.15−0.15, 0.460.33GNAS aberration (1: Yes versus 0: No)b20 versus 4940.480.15, 0.800.0040.380.07, 0.680.02 [GNAS pathogenic alterations correlate with higher levels of IL-2 expression (by 0.38 units on average)]NF1 aberration (1: Yes versus 0: No)b20 versus 494−0.10−0.42, 0.230.56Gene alterations in 20 or more patients were included in this analysis. Gene alterations include mutations, amplifications, deletions, rearrangements and any other abnormalities that were not a VUS. Expression levels categorized as 0: low (<25th), 1: moderate (25th-74th percentile), or 2: high (≥75th) as a linear variable (see also Figure 2H).β refers to the slope of the regression line, representing the amount of change in IL-2 levels when each variable increases by one unit. Positive β value indicates positive correlation.CI, confidence interval; IL-2, interleukin-2; IL-2Rα, IL-2β, IL-2γ, interleukin-2 receptor subunits alpha, beta, and gamma; GI, gastrointestinal; HR, hazard ratio.aVariables with univariable P < 0.2 were introduced into the subsequent multivariable analysis.bGene alterations in 20 or more patients were included in this analysis.cCyclin pathway alterations include CCND1-3, CDK4, CDK6, CDKN1B, CDKN2A, CDKN2B, CCNE1, and RB1. In MAPK pathway, MEK1 (MAP2K1), RAF (ARAF/BRAF/CRAF), and RAS (KRAS/HRAS/NRAS) molecular alterations were included. No ERK aberrations were seen. Over 85% of aberrations in the MAPK pathway are attributed to RAS mutations (134 patients). PI3K pathway alterations included PIK3CA, PTEN, and AKT alterations. Table 3Univariable and multivariable linear regression analysis comparing variables shown with IL-2 RNA expression levelsVariableNUnivariableMultivariableaβ95% CIPβ95% CIPAge (1: ≥61 years versus 0: <61 years)256 versus 2580.06−0.07, 0.180.39Sex (1: Male versus 0: Female)204 versus 3100.02−0.11, 0.150.81GI cancer (1: Yes versus 0: No)272 versus 2420.130.00, 0.260.045−0.08−0.21, 0.060.27IL-2Rα (2: high versus 1: moderate versus 0: low)111 versus 270 versus 1330.320.23, 0.41<0.0010.110.00, 0.210.04 (When IL-2 increases by one unit, IL-2Rα increases significantly by 0.11 units on average)IL-2Rβ (2: high versus 1: moderate versus 0: low)124 versus 272 versus 1180.350.27, 0.44<0.0010.170.06, 0.270.002 (When IL-2Rβ levels increase by one unit, IL-2 expression levels significantly increase by 0.17 units on average)IL-2Rγ (2: high versus 1: moderate versus 0: low)160 versus 233 versus 1210.310.22, 0.39<0.0010.170.06, 0.270.001 (When IL-2Rγ increases by one unit, IL-2 expression levels increase significantly by 0.17 units on average)STAT5A (2: high versus 1: moderate versus 0: low)b121 versus 274 versus 1190.340.25, 0.42<0.0010.160.06, 0.250.002 (When STAT5A increases by one unit, IL-2 expression levels increase significantly by 0.16 units on average)MAPK pathway aberrations (1: Yes versus 0: No)b,c127 versus 3570.190.05, 0.330.0060.10−0.03, 0.240.13TP53 aberration (1: Yes versus 0: No)b254 versus 260−0.11−0.24, 0.010.08−0.08−0.20, 0.040.18PI3K pathway aberration (1: Yes versus 0: No)b,c94 versus 4200.07−0.10, 0.230.42JAK pathway aberration (1: Yes versus 0: No)b5 versus 509−0.20−0.84, 0.450.55Cyclin-related alteration (1: Yes versus 0: No)b,c120 versus 394−0.03−0.18, 0.110.65APC aberration (1: Yes versus 0: No)b84 versus 430−0.08−0.25, 0.090.34ARID1A aberration (1: Yes versus 0: No)b55 versus 4590.01−0.20, 0.210.94SMAD4 aberration (1: Yes versus 0: No)b49 versus 465−0.14−0.35, 0.080.21ERBB2 aberration (1: Yes versus 0: No)b39 versus 475−0.17−0.41, 0.070.16−0.14−0.36, 0.080.22MYC aberration (1: Yes versus 0: No)b29 versus 485−0.15−0.43, 0.120.27ATM aberration (1: Yes versus 0: No)b27 versus 487−0.08−0.36, 0.200.58MLL2 aberration (1: Yes versus 0: No)b23 versus 4910.15−0.15, 0.460.33GNAS aberration (1: Yes versus 0: No)b20 versus 4940.480.15, 0.800.0040.380.07, 0.680.02 [GNAS pathogenic alterations correlate with higher levels of IL-2 expression (by 0.38 units on average)]NF1 aberration (1: Yes versus 0: No)b20 versus 494−0.10−0.42, 0.230.56Gene alterations in 20 or more patients were included in this analysis. Gene alterations include mutations, amplifications, deletions, rearrangements and any other abnormalities that were not a VUS. Expression levels categorized as 0: low (<25th), 1: moderate (25th-74th percentile), or 2: high (≥75th) as a linear variable (see also Figure 2H).β refers to the slope of the regression line, representing the amount of change in IL-2 levels when each variable increases by one unit. Positive β value indicates positive correlation.CI, confidence interval; IL-2, interleukin-2; IL-2Rα, IL-2β, IL-2γ, interleukin-2 receptor subunits alpha, beta, and gamma; GI, gastrointestinal; HR, hazard ratio.aVariables with univariable P < 0.2 were introduced into the subsequent multivariable analysis.bGene alterations in 20 or more patients were included in this analysis.cCyclin pathway alterations include CCND1-3, CDK4, CDK6, CDKN1B, CDKN2A, CDKN2B, CCNE1, and RB1. In MAPK pathway, MEK1 (MAP2K1), RAF (ARAF/BRAF/CRAF), and RAS (KRAS/HRAS/NRAS) molecular alterations were included. No ERK aberrations were seen. Over 85% of aberrations in the MAPK pathway are attributed to RAS mutations (134 patients). PI3K pathway alterations included PIK3CA, PTEN, and AKT alterations. Univariable and multivariable linear regression analysis comparing variables shown with IL-2 RNA expression levels Gene alterations in 20 or more patients were included in this analysis. Gene alterations include mutations, amplifications, deletions, rearrangements and any other abnormalities that were not a VUS. Expression levels categorized as 0: low (<25th), 1: moderate (25th-74th percentile), or 2: high (≥75th) as a linear variable (see also Figure 2H). β refers to the slope of the regression line, representing the amount of change in IL-2 levels when each variable increases by one unit. Positive β value indicates positive correlation. CI, confidence interval; IL-2, interleukin-2; IL-2Rα, IL-2β, IL-2γ, interleukin-2 receptor subunits alpha, beta, and gamma; GI, gastrointestinal; HR, hazard ratio. Variables with univariable P < 0.2 were introduced into the subsequent multivariable analysis. Gene alterations in 20 or more patients were included in this analysis. Cyclin pathway alterations include CCND1-3, CDK4, CDK6, CDKN1B, CDKN2A, CDKN2B, CCNE1, and RB1. In MAPK pathway, MEK1 (MAP2K1), RAF (ARAF/BRAF/CRAF), and RAS (KRAS/HRAS/NRAS) molecular alterations were included. No ERK aberrations were seen. Over 85% of aberrations in the MAPK pathway are attributed to RAS mutations (134 patients). PI3K pathway alterations included PIK3CA, PTEN, and AKT alterations. GNASR201C mutation drives cell-autonomous IL-2 production We generated isogenic Caco-2 cell lines including wild-type GNAS controls and two clones harboring R201C mutations using CRISPR-Cas9 genome editing. All mutations were validated by Sanger sequencing. To assess differences in IL-2 production, we collected conditioned media from the three isogenic lines and measured IL-2 levels by ELISA. We observed a significant increase in IL-2 production in both clones harboring the GNASR201C mutation compared with wild-type controls (Figure 2I). Because these cell lines are genetically identical except for the introduced mutation, these findings suggest that the GNASR201C alteration drives cell-autonomous IL-2 production. Discussion In our study, we noted that any expression of IL-2 RNA, in a diverse patient population, was associated with high RNA expression of several checkpoints that are known to be negative regulators of T-cell activity or produce T-cell anergy, including high expression of TIM-3 and BTLA, in the multivariable analysis.21,22 It is known that exhausted T cells have low sensitivity to cytokines and PD-1 inhibition.23 IL-2 RNA expression was independently prognostic for longer OS from time of diagnosis of advanced/metastatic disease in our patient cohort. We aimed to validate our findings using TCGA RNA sequencing and patient survival data, which also showed that IL-2 expression (compared with no expression), as well as higher IL-2 expression (compared with lower expression), were prognostic for longer survival. Consistent with findings in our patient cohort, no prognostic effect for OS was observed in high-affinity IL-2R expressors (see Supplementary Table S1, available at https://doi.org/10.1016/j.esmoop.2026.107768, for high-affinity IL-2R determination) compared with others in the TCGA dataset (data not shown). In our dataset, IL-2 level did not correlate significantly with outcome after ICI therapy. Specifically, in the subset of patients treated with ICIs in our study, IL-2 RNA expression did not predict PFS or OS from the start of immunotherapy. Our data are consistent with a prior pooled analysis of 24 clinical trials that demonstrated no independent correlation between peripheral plasma IL-2 and survival benefit in patients treated with ICIs.6 IL-2 signal transduction involves activation of several downstream pathways upon binding of IL-2 to its receptor.24 We noted a significant and independent association between GNAS alterations and IL-2 RNA expression (P = 0.02) on multivariable linear regression analysis. This correlation indicates that either an increase in IL-2 RNA expression levels increases the likelihood of GNAS aberrations or perhaps the presence of GNAS aberrations is associated with higher likelihood of increased IL-2 expression. This finding has not been reported in the past. Pathogenic GNAS missense mutations (p.R201C/H) and GNAS amplifications frequently co-occur with MAPK alterations.25 In our dataset, 18 of 27 patients with GNAS alterations had any MAPK pathway aberrations, with 14 of these having KRAS alterations; however, MAPK alterations were not significantly associated with IL-2 RNA expression in our analysis. Other studies have also shown correlations between specific mutations and cytokine levels. For example, TP53 mutations correlate with increased levels of IL-6, and RAS mutations are correlated with high IL-1β levels, perhaps by inducing transcription factor [nuclear factor (NF)-IL6/CREB, NFB1, NF-κB, and NF-IL6P] binding at multiple IL-1β promoter elements, consistent with the activation of an upstream signaling pathway.26,27 Pathogenic GNAS mutations and amplifications occur in <10% of patients, with higher frequencies in mucinous carcinomas than in non-mucinous tumors.28,29 The common hotspot alterations in the TCGA database were similar to those in our patient cohorts, including the GNAS p.R201C/H missense mutations and amplifications. There is limited data on the impact of GNAS alterations on the immune microenvironment in cancer cells. Our findings demonstrate that the introduction of GNASR201C mutation into isogenic Caco-2 cells led to a significant increase in IL-2 secretion compared with GNAS wild-type cells, supporting that this alteration alone can promote cell-autonomous cytokine production. The R201C variant is a well-known activating mutation that persistently engages the Gαs–adenylyl cyclase–cAMP–PKA pathway, thereby driving transcription of downstream targets under the control of CREB and NF-κB.30, 31, 32 However, increased cAMP in cells generally increases Tregs and suppresses IL-2 production, which is different from the findings in our cohort of patients.32 In appendiceal and other cancers, GNAS mutations have been linked to tumorigenesis, and our findings from GNAS altered cell lines build on this work by implicating IL-2 as a novel downstream effector. A study analyzing 1348 patients using a clinical-grade NGS panel identified a significant association (P ≤ 0.0028) between GNA∗ alterations (GNAS, GNAQ, GNA11) and molecular aberrations in AURKA, SRC, CBL, and LYN. Cbl-b, a ubiquitin E3 ligase, plays a crucial role in T cell receptor/CD28 co-stimulation.33 Notably, Cbl-b knockout (Cbl-b−/−) T cells exhibit full activation in vivo, including robust IL-2 production.34,35 Another study in Cbl-bH257L T cell-mutated mice showed increased IL-2 secretion with anti-CD3 antibody stimulation.36,37 Our study has certain limitations. It includes experience from a single institution in a heterogeneous group of patients who received ICIs as different lines of treatment. The UCSD cohort was limited in size. Future studies should examine larger cohorts with different IL-2 cut-offs (see Figure 1). The associations between IL-2 expression and immune markers, genomic alterations, and clinical outcomes are exploratory in nature and intended to be hypothesis-generating. A limitation of the study is that bulk RNA expression cannot differentiate if this expression derives from tumor or immune cell infiltration. Future studies should address this issue using single cells or other analyses. However, the origin of IL-2 RNA detected in tumor specimens is likely attributable to one or more cellular compartments within the tumor, including infiltrating immune cells of the tumor microenvironment (TME) or the malignant cells themselves. The TME is a heterogeneous mixture of resident and tumor cells that maintain close communication through their secretion products, and IL-2 is a key immunoregulatory cytokine within this milieu.38 Critically, our isogenic cell lines harboring the GNASR201C mutation produced significantly higher levels of IL-2 protein compared with their wild-type counterparts, demonstrating that this gain-of-function variant augments IL-2 expression and secretion in an R201C-dependent manner. Although the exact condition of the TME in individual cases was not assessed, it is known that ICIs activate the non-naïve, quiescent T cell population, and IL-2 signaling is crucial for T cells to exit from this quiescent state.39 Further, it remains unclear if the favorable prognosis correlated with IL-2 RNA expression (longer OS among all patients) is applicable to specific cancer types in our cohort, as the number of patients with individual cancer types was relatively small in our cohort. Finally, future studies should examine IL-2 RNA expression at the single cell level. Conclusion IL-2 RNA expression was a robust prognostic marker among patients with solid tumors, predicting longer OS. Interestingly, IL-2 transcript expression was significantly associated with GNAS alterations, and this association was confirmed by high IL-2 production in CRISPR-edited cells bearing GNASR201C mutations compared with wild-type controls. Further studies are needed to understand the immunophenotype associated with GNAS pathogenic alterations and the role of IL-2 in solid tumors bearing these mutations.