Glucagon-like peptide-1 receptor agonism improves lung cancer outcomes and tumor growth control

We investigated the in vitro effect of liraglutide on;-NINJA (KPN1.1) and Lewis lung carcinoma (LLC) cell proliferation. Based on reported steady-state plasma concentrations of liraglutide ranging from 20–40 nM for typical clinical doses (0.6–1.8 mg), we examined a range of concentrations: 0, 10, 30, 90, 270, and 810 nM. Cell proliferation was assessed by measuring the fold change in absorption values after 48 hours of treatment. There was no significant difference in cell proliferation for KPN1.1 cells across liraglutide concentrations, while LLC cells exhibited increased proliferation at higher drug concentrations (, A and B; supplemental material available online with this article;). Kras lox-stop-lox(lsl)-G12D/+ Tp53 fl/fl Supplemental Figure 1 https://doi.org/10.1172/jci.insight.195484DS1↗

We next investigated the effect of liraglutide on tumor growth in vivo. Here, normal-weight (lean) C57BL/6J mice and those with diet-induced obesity (DIO) were challenged with subcutaneous (s.c.) injection of KPN1.1 lung cancer cells. Upon detection of palpable s.c. tumors, mice received daily liraglutide (0.2 μg/g body weight) or a PBS vehicle intraperitoneally. While liraglutide-treated mice exhibited a slight decrease in body weight during the experiment, and control mice showed an increase (weight change: –0.78 g vs. +0.98 g, respectively;, A and B), no statistically significant difference was observed in mean tumor volumes or final tumor weight (0.98 g vs. 1.04 g;= 0.67) in normal-weight/lean mice (). In contrast, liraglutide administration significantly inhibited the progression of tumor growth, as evidenced by reductions in tumor volumes and weight (1.07 g vs. 1.86 g;= 0.03) relative to the vehicle group (). Of note, liraglutide-treated DIO mice demonstrated greater weight reduction compared with their respective controls (13.46 g vs. 4.06 g, respectively;, A and B). Supplemental Figure 2 Figure 1, A and B Figure 1, C and D Supplemental Figure 2 P P

To gain insight into the mechanism underlying the benefits of GLP-1RAs suggested by clinical and preclinical studies, tumor gene expression was compared in liraglutide- and vehicle-treated s.c. tumors harvested in DIO mouse model studies using RNA sequencing (RNA-Seq). Of the 35,141 transcripts assessed, 27 were differentially expressed significantly (adjusted< 0.05 and |logFC| > 0.58,). The transcripts upregulated with liraglutide treatment (, green) included, which induces Gphase cell cycle arrest by upregulating cyclin-dependent kinase inhibitors. Additionally, treatment was associated with increased, a regulator of angiogenesis (), and, which encodes a factor with pro- and antitumor roles that are also involved with immune cell development and differentiation (). Conversely, liraglutide exposure was also associated with decreased, a well-known marker of poor prognosis with a role in various cell signaling pathways (). Furthermore, the drug was linked to reduced expression of the protumor genesand, associated with hypoxia-induced cellular stress and metabolism, and cell proliferation and migration, respectively (,). P Igfbp7 Creb3l1 Nfix Anxa10 Higd1a Rpl39 Figure 2A Figure 2A ^{11 12 13 14 15} 1

Gene set enrichment analysis (GSEA) revealed significant treatment effects on several pathways relevant to cancer outcomes (normalized enrichment score > 2,< 0.05) spanning cell proliferation, cellular respiration, immune signaling, and cell signaling (). Notably, genes linked to signaling triggered by IFN-γ (an antitumor immunity–promoting cytokine) and growth factor (insulin-derived growth factor) signaling were significantly upregulated with liraglutide treatment. Several metabolic pathways important for tumor and leukocyte biology, including fatty acid (sphingolipids) and amino acid metabolism, were enhanced and disrupted, respectively, and genes important for translation-associated processes were also apparently suppressed with treatment, as were those associated with poor lung cancer survival. Finally, multiple cell signaling pathways well appreciated for driving tumor development and progression were also suppressed in GLP-1RA–treated tumors, including those involving,,, andtranscription factors (). Overall, these findings link liraglutide with gene expression changes in the tumor relevant to immunity, metabolism, and tumor progression that are likely to reflect or explain the effects of GLP-1RAs seen in mice and patients. q KRAS EGFR MYC E2F Figure 2B Figure 2B

GLP-1R is expressed in several leukocyte populations (), but the effects of GLP-1RAs on these cells are apparently complex, and their likely impact on the antitumor immune response is unclear. While many have found the effects of GPL-1R signaling and agonism to be immunosuppressive (), some studies suggested enhanced antitumor immunity (,). To explore whether liraglutide's antitumor effects in our mouse model studies stem from favorable modulation of antitumor immunity, we used flow cytometry to characterize the phenotypic and functional changes among the tumor-infiltrating leukocytes (TILs) associated with treatment in harvested s.c. tumors. ^{16 17 18 19}

Increased CD4T cell and natural killer (NK) cell frequencies in the tumors of GLP-1RA–treated obese mice were seen (). Proportions of conventional (i.e., non-regulatory) CD4T cells (Tcon), marked by the activation marker CD69, were also elevated in liraglutide-treated mice (), potentially indicating an enhanced antitumor immune response. Interestingly, the fraction of intratumor CD8and Tcon compartments with an effector surface marker profile (CD44CD62L) was apparently reduced in favor of CD44CD62LT cells (), suggesting a potential shift toward a memory-like phenotype in these cells. + + + hi lo + + Figure 3, A and B Figure 3C Figure 3, D and E

Since Foxp3regulatory T cells (Tregs) express relatively high levels of GLP-1R (), and a role for its downstream signaling has been reported to maintain these suppressor cells in the periphery (), we assessed liraglutide-triggered changes among tumor Tregs. Surprisingly, proportions of Foxp3CD25cells within the TIL CD4pool were significantly reduced with treatment (). Interestingly, the functionally potent, activated (CD44CD62L) or "eTreg" subpopulation made up a smaller portion of the TIL-Treg pool from liraglutide-treated mice compared with vehicle-treated control mice, in favor of a population with memory-like potential (i.e., CD44CD62Lsurface staining) (). In agreement with a prior study (), GLP-1RA treatment upregulated PD-1 and PD-L1 on tumor Tregs (, A and B); however, increased levels of mTOR activity (indicated by phosphorylated mTOR[p-mTOR]Foxp3T cells) and Tbet expression were found among GLP-1RA–treated Tregs as well (, C and D). This combination indicates an activated but unstable Treg phenotype (–), less capable of restricting robust antitumor immune response suppression. Collectively, these results suggest the ability of GLP-1RAs to undermine immunosuppression in obese mice. + + + + + – + + + + + ^{20 21} Figure 3F Figure 3G ²⁰ Supplemental Figure 3 Supplemental Figure 3 ^{22 25}

Additional effects of GLP-1RA treatment were evident in antigen-presenting cell (APC) populations in the tumor niche. Dendritic cells (DCs) and macrophages from treated mice displayed upregulated MHC II levels, an effect that approached and achieved statistical significance in these populations, respectively (). Conversely, TILs of the GLP-1RA–treated mice harbored markedly fewer APCs producing indoleamine 2,3-dioxygenase (IDO) (), a mediator of tumor-associated immunosuppression and a mechanism of Treg expansion and differentiation (,). The mice treated with liraglutide also displayed a relative scarcity of arginase-expressing myeloid-derived suppressor cells (MDSCs) (). Thus, it is possible that the GLP-1RA triggers APC phenotypes poised to effectively stimulate T cell responses and cell-mediated antitumor immunity. Further supporting this notion, the pool of TIL-isolated CD4T cells recovered from liraglutide-treated mice capable of producing proinflammatory, tumoricidal cytokines ex vivo was expanded (). In all, these findings support a potent immunostimulatory effect of the GLP-1RA liraglutide impacting several leukocyte populations that may be responsible for the improved tumor growth control and clinical outcomes linked to the drug. Figure 3H Figure 3I ^{26 27} Figure 3J Figure 3K +

To reduce the impact of model-specific or drug-specific factors on tumor dynamics, we repeated our experiment examining the effects of GLP-1RA administration on murine tumor growth, this time in an independent autochthonous mouse model with conditionalactivation anddeletion (KP) using the GLP-1RA semaglutide. To induce obesity, tumor-free KP mice were prefed a high-fat Western diet (60% kcal from fat) supplemented with sugar water (23.1 g/L D-fructose and 18.9 g/L D-glucose) for 13 weeks prior to tumor initiation (). Lung adenocarcinoma was initiated via intratracheal delivery of Cre recombinase; this precipitates spontaneous tumor growth, which more closely mimics the geographic pattern of human lung cancer metastases compared with s.c. implantation. The KP mice with DIO and induced lung adenocarcinoma were treated with semaglutide (= 7) or vehicle control (= 7). KrasG12D Trp53 n n Figure 4A

Semaglutide treatment significantly extended overall survival (HR: 0.21, 95% CI: 0.057–0.79,= 0.012) compared with vehicle-treated control (). Longitudinal microcomputed tomography (μCT) revealed significantly reduced tumor burden in the semaglutide groups (), mirroring the results of our heterotopic models, where GLP-1RA reduced tumor growth and progression in obesity-associated lung cancer. P Figure 4B Figure 4C

We then performed a retrospective examination of clinical outcomes in a database of patients undergoing surgery for NSCLC at our center between 2015 and 2024. A total of 1,177 patients met criteria for inclusion in the postsurgical cohort (a flowchart of the cohort is represented in). Within this cohort, 71 patients were GLP-1RA users. Demographic characteristics, including age, clinical stage, race, and smoking history were generally well balanced between the GLP-1RA users and non-users; however, GLP-1RA users had significantly higher mean body mass index (BMI) at the time of resection than the non-users (35.07 vs. 30.6;= 0.001). Demographic data are reported in. Most patients across both groups had clinical stage I disease 970 (82.4%), while 193 (16.4%) had stage II, and 14 (1.2%) had stage III lung cancers. Supplemental Figure 4A Supplemental Table 1 P

In an unmatched, univariate Cox regression model including the entire cohort, GLP-1RA use was associated with improved recurrence-free survival (RFS) (HR: 0.41; 95% CI: 0.17–1.02;= 0.027). No such association was noted for overall survival (OS) (HR: 0.70; 95% CI: 0.43–1.16;= 0.09). P P

In a propensity-matched analysis, where propensity scores were calculated using age, sex, race, BMI, cancer stage, smoking status, and histology as covariates and GLP-1RA use as the dependent variable, a matched cohort of 629 was identified that included 560 GLP-1RA non-users and 69 GLP-1RA users. This cohort did not include any patients with stage III disease (see). Univariate Cox regression of the matched cohort showed that GLP-1RA use was associated with improved RFS (HR: 0.41, 95% CI: 0.16–1.04,= 0.026;). GLP-1RA use was not significantly associated with OS (HR: 0.78, 95% CI: 0.44–1.38,= 0.20;). Results of univariate analyses for all the covariates in both unmatched and matched cohorts are presented in. Supplemental Table 1 Figure 5A Figure 5B Table 1 P P

In light of the findings from the preclinical model and expanding use of immunotherapy in the treatment of lung cancer, we set out to test whether the apparent benefits of GLP-1RA use extend to patients on ICI therapy. A cohort of 300 patients with advanced lung cancer treated with ICI at our center between 2015 and 2023 was identified, of which 10 patients were prescribed concurrent GLP-1RA. A flowchart of the cohort is represented in. All GLP-1RA users in the cohort had diabetes (reflecting FDA-approved indications during our study period), whereas 25% of non-users were diabetic. Most patients (78.7%) had stage IV disease; the remainder of the cohort was comprised of stage III NSCLC not amenable to local treatments. Treatment included immunotherapy alone for 158 patients (52.7%) and a combination of immuno- and chemotherapy for 142 (47.3%). Patients receiving GLP-1RA were significantly younger; other characteristics such as sex, BMI, race, smoking status, tumor histology, and treatment modality were similar between groups (). Supplemental Figure 4B Supplemental Table 2

In an unmatched analysis, univariate Cox regression modeling demonstrated that GLP-1RA use was associated with improved progression-free survival (PFS) (HR: 0.39; 95% CI: 0.13–1.14;= 0.044) and OS (HR: 0.40; 95% CI: 0.15–1.04;= 0.031). P P

In a propensity-matched analysis, propensity scores were calculated using age, sex, race, smoking status, BMI, tumor histology, first-line therapy, and stage at the start of ICI therapy as covariates, with GLP-1RA use as the dependent variable. The matched cohort (= 89) had 79 non-users and 10 GLP-1RA users (see). Univariate Cox regression of the matched cohort showed that GLP-1RA use was associated with improved PFS (HR: 0.31, 95% CI: 0.1–0.94,= 0.019;) and OS (HR: 0.41, 95% CI: 0.16–1.01,= 0.027;). Results of univariate analyses for all covariates in both unmatched and matched cohorts are presented in. n P P Supplemental Table 2 Figure 6A Figure 6B Table 2

Male mice were used in these lung cancer preclinical models to facilitate creation of DIO models, as male C57BL/6J mice more uniformly display significant weight gain on high-fat diets (relative to normal-diet controls) than females (–). This approach is consistent with established protocols in the field and facilitates direct comparison with previous studies. ^{45 47}

The murine lung cancer cell lines KPN1.1 () (provided by N. Joshi, Yale School of Medicine, New Haven, Connecticut, USA) and LLC (ATCC, CRL-1642-LUC2) were cultured in vitro and harvested in the logarithmic phase by trypsinization, followed by centrifugation. Cell concentration was adjusted and added to strip-well plates at 3000 cells per well, which were then incubated at 37°C. After 24 hours, the cells were treated with liraglutide at 10, 30, 90, 270, and 810 nM concentrations. To measure the cell proliferation rate, the wells were washed, and 95 μL of cell culture medium and 5 μL of CCK-8 reagent (Dojindo Laboratories) were added to each well. The strips were allowed to incubate at 37°C for 2 hours, and the absorbance value (OD value) was measured and recorded using a microplate reader at a wavelength of 450 nm at 0, 24, 48, 72, and 96 hours. The cell proliferation rate was calculated as follows: cell proliferation rate = (OD value at a given concentration at 48-hour time point – blank)/(OD value on day 0 – blank). ⁴⁸

Age-matched obese and non-obese male C57BL/6J mice were purchased (The Jackson Laboratory, strain numbers 380050 and 380056) or generated in-house by feeding normal-weight mice with high-fat (60% calories from fat, 5.4 kCal/g) or control chow (~7.2% calories from fat, 3.9 kcal/g) (Bio-Serv, products S3282 and F4031, respectively). For s.c. tumor challenge experiments, KPN1.1 cells or LLC cells were injected into the flanks of 7- to 17-week-old obese and non-obese mice (75,000 cells per mouse). Mice with palpable tumors (typically detected 6–7 days after implantation) received intraperitoneal injections of liraglutide (0.2 μg per gram of mouse body weight) or phosphate-buffered saline (PBS) vehicle daily. The mice were euthanized after 19–22 days of cell inoculation, and tissues were harvested for further analysis.

RNA sequencing was performed on excised, subcutaneous KPN1.1 tumors from obese mice treated with GLP-1RA. At the time of mouse tumor harvest, a fragment was flash frozen and was later used for RNA isolation using the miRNeasy mini kit (Qiagen). The sequencing libraries were prepared with the RNA HyperPrep Kit with RiboErase (HMR) kit (Roche Sequencing Solutions), from 500 ng total RNA. Following the manufacturer's instructions, the first step depletes rRNA from total RNA. After ribosomal depletion, the remaining RNA was DNase digested to remove any gDNA contamination. Samples were then purified, fragmented, and primed for cDNA synthesis. Fragmented RNA was then reverse transcribed into first-strand cDNA using random primers. The next step removes the RNA template and synthesizes a replacement strand, incorporating dUTP in place of dTTP to generate double-stranded cDNA. Pure Beads (KAPA BIOSYSTEMS) were used to separate the double-stranded cDNA from the second-strand reaction mix, resulting in blunt-ended cDNA. A single "A" nucleotide was then added to the 3′ ends of the blunt fragments. Multiple indexing adapters, containing a single "T" nucleotide on the 3′ end of the adapter, were ligated to the ends of the double-stranded cDNA, preparing them for hybridization onto a flow cell. Adapter-ligated libraries were amplified by PCR, purified using Pure Beads, and validated for appropriate size on a 4200 TapeStation D1000 Screentape (Agilent Technologies, Inc.). The DNA libraries were quantified using a KAPA Biosystems qPCR kit, and were pooled together in an equimolar fashion, following experimental design criteria. Each pool was denatured and diluted to 350 pM with 1% PhiX control library added. The resulting pool was then loaded into the appropriate NovaSeq Reagent cartridge for 100-bp paired-end sequencing and sequenced on a NovaSeq 6000 following the manufacturer's recommended protocol (Illumina Inc.). Raw reads that passed the Illumina RTA quality filter were demultiplexed and preprocessed using FastQC for sequencing base quality control. Reads were then mapped to the latest version of a mouse reference (GRCm39/mm39) using Bowtie (v1.0.1) () and TopHat (v2.0.13) aligner (). Mapped reads were quantified at the gene level as a raw counts matrix using featureCounts from Subread (). Samples were filtered after quality assessment, and a total of 10 samples were used in subsequent analyses (= 5/group). Raw feature counts were normalized, and differential expression analysis was carried out using DESeq2 (), and visualized using volcano plots. Differential expression rank order was used for subsequent GSEA (), performed using the cluster profile package in R, and visualized via lollipop plots. Gene sets queried included the Hallmark, Canonical pathways, and GO Biological Processes Ontology collections available through the Molecular Signatures Database (MSigDB) (). ^{49 50 51 52 53 54} n

Flow cytometry analysis was performed on the s.c. KPN1.1 tumors excised from obese mice treated with GLP-1RA. Tumors were cleaned of skin and fat before mechanical (GentleMACS dissociation) and enzymatic digestion in a collagenase/hyaluronidase mix (Stem Cell Technologies) following the manufacturer's protocol. The resulting cell suspensions were filtered (100 μm cell strainer), washed, and pelleted before RBC lysis in 1 mL ACK buffer (Thermo Fisher Scientific). Cells were then washed with PBS containing 1% fetal bovine serum (FBS) and 2 mM EDTA before incubation with fluorochrome-conjugated antibodies diluted in like buffer. After surface immunostaining, intracellular markers (i.e., FOXP3) were stained after fixation and permeabilized using eBioscience's FOXP3 staining kit. For intracellular cytokine staining, cell suspensions were restimulated in media containing PMA and ionomycin in the presence of GolgiPlug (BD) for 5 hours at 37°C before surface staining, fixation/permeabilization, and internal staining with anti–IFN-γ and –TNF-α antibodies. Samples were run on an Aurora spectral flow cytometer (Cytek) before data analysis via OMIQ (). https://www.omiq.ai/↗

Effect of semaglutide on survival in an autochthonous KRAS-driven murine lung cancer model.;(KP) mice () (The Jackson Laboratory, 032435) were fed a high-fat Western diet (60% kcal fat; Research Diets, D12492) with sugar water (23.1 g/L D-fructose + 18.9 g/L D-glucose) ad libitum for 13 weeks to induce DIO. Lung tumors were initiated via intratracheal adenoviral Cre recombinase (Ad-Cre; 2.5 × 10PFU; University of Iowa Viral Core) (–).Tumor-bearing DIO-KP mice were randomized to receive intraperitoneal semaglutide (60 nmol/kg/0.25 μg/gm body weight, 3 times per week; S9697, Selleck Chemicals) or saline (pH 7.4; Cytiva) (). Imaging was conducted through the Oncology Precision Therapeutics and Imaging Core (OPTIC) at the Columbia University Herbert Irving Comprehensive Cancer Center. High-resolution μCT was performed to assess lung tumor burden, as previously described (). Kras LSL-G12D Trp53 fl/fl ⁵⁵ 7 ^{56 59 60 61}

Two distinct single-center clinical cohorts were manually curated and analyzed.

In a postsurgical cohort, medical charts for patients undergoing surgical resection at the Roswell Park Cancer Center (RPCC) between 2015 and 2024 were manually curated. Patients were included for analysis if they had (a) histologically confirmed NSCLC and (b) a BMI greater than 25. Patients were defined as GLP-1RA users if they had GLP-1RA prescribed for more than 6 months during the period between resection and any event (defined as disease recurrence, death, or loss to follow-up).

In an advanced-disease cohort, medical charts for patients undergoing treatment for advanced or metastatic lung cancer at the RPCC between 2015 and 2023 were manually curated. Patients were included for analysis if they had (a) histologically confirmed advanced or metastatic NSCLC, (b) received at least 3 doses of an anti–PD-(L)1 and/or anti–CTLA-4 ICI (c), received no definitive local therapy (surgery or radiotherapy), and (d) had a BMI greater than 25. Patients were again defined as GLP-1RA users if they had GLP-1RA prescribed for more than 6 months during the period between the start of ICI treatment and any event (defined as disease progression, death, or loss to follow-up).

In both cohorts, patient demographics, including age, sex (male or female), race (White and non-White), smoking status (current, former, never smokers), and BMI were collected.

Visceral fat indexes and abdominal circumference were not routinely available in this retrospective database. For the surgical cohort, RFS was calculated from the date of resection until disease recurrence, death, or censoring at loss to follow-up. For the advanced-disease cohort, PFS was used, calculated from the date of treatment initiation until disease progression, death, or censoring at loss to follow-up. OS was defined from the point of resection in the postsurgical cohort and start of treatment in the advanced-disease cohort until the date of death or censoring at loss to follow-up.

In the retrospective cohort analysis, results are reported as RFS, PFS, and OS as appropriate to the clinical scenario. To reduce confounding from baseline differences in small sample sizes, propensity scores were initially estimated using logistic regression with GLP-1RA use as the dependent variable. Propensity score matching was subsequently performed using greedy nearest-neighbor matching with a fixed control-to-treated ratio of 10:1 and a caliper width of 0.2 on the propensity score. The matching procedure was implemented in SAS version 9.4 (IBM) using PROC PSMATCH. Survival analysis was done using Cox's proportional hazards regression on the matched cohort, accounting for the matched sets through stratification or clustering. Group differences were assessed using independenttests and χtests. All statistical tests were 2-sided with a level of significance of 5%. OS and PFS analyses were conducted using Cox's proportional hazards modeling for univariate models. The statistics for survival analysis are expressed in terms of HR, and 95% CI with a 1-sidedvalue at a 5% level of significance. For animal model studies and in vitro experiments, the group differences were calculated usingtests or 2-way ANOVA tests. Statistical analyses were performed using SPSS version 28.0.1.0 (IBM) or GraphPad Prism version 9. Flow cytometry analysis was performed using OMIQ. t P t 2

The retrospective study protocol was reviewed and approved by the Institutional Review Board of the RPCC (BDR 115119), and individual patient consents were waived. All animal experiments conducted were reviewed and approved by the Institutional Animal Care and Use Committee (IACUC) of RPCC (protocols 1487 and 1349) or by Columbia University IACUC guidelines (protocol AABV8661).

RPCC IRB protocols prohibit the sharing of clinical data to protect patient privacy. The data from mouse experiments are available in thefile. Flow cytometry data will be made available as raw data files upon request. Gene sequencing data have been deposited in the NCBI Gene Expression Omnibus (GEO GSE302656). Supporting Data Values

A part of this study was submitted and presented at the Academic Surgical Congress – Annual Meeting 2025. This manuscript or any part of it is not under consideration for publication elsewhere.