CXCL9 drives lung adenocarcinoma progression and immunotherapy response via activation of the CXCR3/JAK1/STAT3 signaling pathway

Highlights

• •CXCL9 is upregulated in LUAD responders to PD-L1 immunotherapy.
• •High CXCL9 expression predicts poor prognosis in LUAD.
• •CXCL9 promotes LUAD cell proliferation, migration, and invasion.
• •CXCL9 activates CXCR3/JAK1/STAT3 signaling to drive malignancy.
• •JAK1/2 inhibitor Ruxolitinib reverses CXCL9-driven malignant phenotypes.

Abstract

Background

C-X-C Motif Chemokine Ligand 9 (CXCL9) is an immune-related chemokine implicated in tumor progression and immunotherapy response. However, its biological role and regulatory mechanisms in lung adenocarcinoma (LUAD) remain unclear.

Methods

Public immunotherapy cohorts, The Cancer Genome Atlas (TCGA) pan-cancer datasets, and single-cell RNA sequencing data were analyzed to evaluate CXCL9 expression, prognostic value, and immune associations. Functional assays, including wound healing, Transwell invasion, apoptosis analysis, qRT-PCR, and Western blotting, were performed in LUAD cell lines with CXCL9 overexpression or knockdown. Xenograft mouse models were used to assess tumor growth in vivo. Pharmacological inhibition and receptor-blocking experiments were conducted to explore the involvement of the CXCR3/janus kinase 1 (JAK1)/signal transducer and activator of transcription 3 (STAT3) signaling pathway.

Results

CXCL9 was significantly upregulated in programmed cell death ligand 1 (PD-L1) immunotherapy responders and associated with immune-related signaling pathways. High CXCL9 expression correlated with poor prognosis in LUAD. Functional experiments demonstrated that CXCL9 promoted LUAD cell proliferation, migration, invasion, and resistance to apoptosis both in vitro and in vivo. Mechanistically, CXCL9 activated the CXCR3/JAK1/STAT3 signaling pathway, while treatment with the JAK1/2 inhibitor Ruxolitinib effectively reversed these malignant phenotypes.

Conclusion

Our findings identify the CXCL9–CXCR3/JAK1/STAT3 axis as a critical regulator of LUAD progression and immunotherapy response, highlighting CXCL9 as a potential prognostic biomarker and therapeutic target in LUAD.

Graphical abstract

Introduction

Lung adenocarcinoma (LUAD), the most prevalent form of lung cancer globally, has emerged as a growing public health concern. In 2020, there were approximately 2.21 million new cases of lung cancer reported worldwide, with adenocarcinoma accounting for 57% of cases in women and 39% in men. China had the highest number of lung cancer cases and deaths, representing 37.0% and 39.8% of the global burden, respectively [1]. The increasing incidence of LUAD is likely influenced by various factors, including changes in smoking rates, exposure to environmental carcinogens [2], and an aging population. The introduction of immunotherapy, specifically PD-1/PD-L1 inhibitors, has significantly transformed the treatment landscape for advanced LUAD. Multiple studies have demonstrated the effectiveness of anti-PD-1/PD-L1 monotherapy in improving overall survival compared to chemotherapy, particularly in tumors with high PD-L1 expression. In 2019, the FDA approved pembrolizumab as a first-line treatment for stage III or IV NSCLC patients with tumor PD-L1 expression ≥1% [3].

CXCL9, an interferon-gamma-inducible chemokine, plays a crucial role in shaping the tumor immune microenvironment by attracting various immune cells, particularly natural killer (NK) cells and T lymphocytes. Its presence within tumors has been linked to increased infiltration of these immune cells, leading to improved prognosis and therapeutic outcomes in various cancers, including lung cancer [[4], [5], [6]]. Chemokine receptor axes, including CXCR4 and CXCR3, have been increasingly implicated in lung cancer progression, immune cell trafficking, and tumor microenvironment remodeling [7], providing a rationale for investigating CXCL9-associated immune signaling in LUAD. In lung cancer immunotherapy, CXCL9 may enhance the efficacy of immune checkpoint inhibitors. Studies have demonstrated that higher levels of CXCL9 are associated with increased immune cell infiltration and activation within the tumor microenvironment, which is essential for the success of immunotherapeutic strategies [8,9]. Notably, research has shown that patients with LUAD exhibiting elevated CXCL9 expression tend to have better responses to immunotherapy and longer overall survival. This highlights the role of CXCL9 as both a biomarker for immune activity and a potential therapeutic target to improve the effectiveness of existing lung cancer treatments [10,11]. Recent study have similarly employed molecular signatures to predict immune microenvironment remodeling and immunotherapy response across cancers [12,13], supporting the utility of immune-related chemokines such as CXCL9 as candidate biomarkers in LUAD. Emerging evidence, however, indicates that the biological outcome of CXCL9 is highly context-dependent and varies with its cellular source, receptor distribution, and tumor microenvironment [14]. This functional dichotomy is exemplified by the observation that high CXCL9 expression can coincide with both enhanced immunotherapy response and poor prognosis within the same tumor type, including lung adenocarcinoma [15]. Such seemingly contradictory associations suggest that CXCL9 may exert distinct effects through immune-mediated and tumor-intrinsic mechanisms, respectively. Beyond conventional tumor-centric endpoints, patient-reported outcomes (PROs) and health-related quality of life (HRQoL) and psychosocial intervention have become essential metrics for evaluating the overall value of cancer immunotherapy [16,17]. Integrating biomarker-guided patient stratification with supportive care interventions may therefore be critical for optimizing both survival and patient-centered outcomes in LUAD management.

In this study, we aimed to elucidate the clinical significance and biological function of CXCL9 in lung adenocarcinoma and to reconcile its immune-activating and tumor-promoting roles. We performed pan-cancer bioinformatics analyses using TCGA data, immunotherapy cohorts, and single-cell RNA sequencing to assess CXCL9 expression, prognosis, and tumor microenvironment associations. Subsequently, gain- and loss-of-function experiments in HCC827 and HCC44 cells were conducted to determine whether CXCL9 directly promotes malignant phenotypes in LUAD cells. Finally, we investigated the underlying molecular mechanism focusing on the CXCR3/JAK1/STAT3 signaling pathway, its pharmacological inhibition, and the therapeutic potential of targeting the CXCL9/CXCR3/JAK1/STAT3 axis in xenograft models (Fig. 1).

Fig. 1.

Study design and experimental workflow.

Methods

Data and clinical sample usage

The GSE126044 data set consists of 5 PD-L1 treatment responders and 11 non-responders for lung cancer immunotherapy. GSE135222 includes 27 lung cancer patients treated with PD-L1, with 8 responders and 19 non-responders. EMTAB6149 is a single-cell sequencing data set for 5 lung cancer patients, which was utilized to examine the expression patterns of CXCL9 at the single cell level and its potential impact on tumor cells. The TCGA data set comprises 33 solid tumors and was used to analyze the differential expression of the CXCL9 gene and its prognostic significance in various tumors through RNA-seq (Table 1). Utilizing UCSC XENA, we obtained TCGA pan-cancer RNA-seq, clinical phenotype, and prognostic data. Additionally, serum and lavage fluid samples were collected from 28 patients who responded and did not respond to immunotherapy, and their levels of CXCL9 gene expression were measured (ethics committee approval number: KY20260746). The clinical details of these 28 patients are available in Table 1.

Table 1.

TCGA data set comprises 33 solid tumors.

Numbers	Cancer abbreviation	Cancer full name
1	ACC	Adrenocortical carcinoma
2	BLCA	Bladder urothelial carcinoma
3	BRCA	Breast invasive carcinoma
4	CESC	Cervical squamous cell carcinoma and endocervical adenocarcinoma
5	CHOL	Cholangiocarcinoma
6	COAD	Colon adenocarcinoma
7	DLBC	Lymphoid neoplasm diffuse large B-cell lymphoma
8	ESCA	Esophageal carcinoma
9	GBM	Glioblastoma multiforme
10	HNSC	Head and neck squamous cell carcinoma
11	KICH	Kidney chromophobe
12	KIRC	Kidney renal clear cell carcinoma
13	KIRP	Kidney renal papillary cell carcinoma
14	LAML	Acute myeloid leukemia
15	LGG	Brain lower grade glioma
16	LIHC	Liver hepatocellular carcinoma
17	LUAD	Lung adenocarcinoma
18	LUSC	Lung squamous cell carcinoma
19	MESO	Mesothelioma
20	OV	Ovarian serous cystadenocarcinoma
21	PAAD	Pancreatic adenocarcinoma
22	PCPG	Pheochromocytoma and paraganglioma
23	PRAD	Prostate adenocarcinoma
24	READ	Rectum adenocarcinoma
25	SARC	Sarcoma
26	SKCM	Skin cutaneous melanoma
27	STAD	Stomach adenocarcinoma
28	TGCT	Testicular germ cell tumors
29	THCA	Thyroid carcinoma
30	THYM	Thymoma
31	UCEC	Uterine corpus endometrial carcinoma
32	UCS	Uterine carcinosarcoma
33	UVM	Uveal melanoma

Gene expression analysis

Employing the limma package (version 3.56.2), we evaluated the differential expression signatures in 33 types of tumors between cancer and normal tissues. Our screening criteria for differential genes were |LogFC| > 1 and P value < 0.05. In the prognostic analysis, we utilized the survminer package (version 0.5.0) to determine optimal grouping based on gene expression, and subsequently utilized the km algorithm to assess the impact of CXCL9 on prognosis in various tumors. P < 0.05 was used to determine the significance of the prognostic analysis.

CXCL9 function analysis

GO analysis is a comprehensive database utilized for the annotation of genes and gene products, featuring terms categorized into cellular component, molecular function, and biological process. KEGG, on the other hand, is a collection of databases that provides information on genomes, biological pathways, diseases, and drugs. Differentially expressed genes were analyzed with clusterProfiler (version 4.18.4), followed by GO and KEGG pathway enrichment analysis. The Benjamini & Hochberg algorithm was applied to adjust the P value, with a significance threshold of <0.05. Additionally, we utilized the STRING database to analyze the interaction relationships between genes, specifically focusing on those related to immunotherapy. A protein interaction analysis was conducted, using a screening standard of 0.4 and the Degree metric to determine the importance of genes in the interaction network.

To systematically evaluate the relationship between CXCL9 expression and immune-related pathway activity, we performed single-sample gene set enrichment analysis (ssGSEA) using the GSVA R package. Gene sets for chemokine signaling, cytokine–cytokine receptor interaction, JAK-STAT signaling, IL6-JAK-STAT3 signaling, interferon-gamma response, PD-1/PD-L1 checkpoint, and Toll-like receptor signaling were obtained from the MSigDB Hallmark and KEGG collections. Pathway activity scores were calculated for each sample in TCGA pan-cancer, TCGA-LUAD, GSE126044, and GSE135222 datasets. Spearman correlation was then used to assess the association between CXCL9 expression levels and each pathway activity score. Statistical significance was determined using FDR correction, with P < 0.05 considered significant.

Single-cell WGCNA analysis

We utilized weighted gene co-expression network analysis (WGCNA) on scRNA-seq data to detect co-expressed gene modules and their correlations with CXCL9. The SCTransform function in Seurat (version 5.2.1) was employed to normalize the raw count data. Highly variable genes were identified and utilized to construct a signed co-expression network. The soft-thresholding power was determined based on the scale-free topology criterion. The dynamic tree cut algorithm was applied to define modules, with a minimum size of 30 genes. Module eigengenes were computed and used to establish associations between modules and cell types through Pearson correlation. The WGCNA analysis was performed using the WGCNA R package (version 1.73).

Correlation analysis of CXCL9 and TMB

In the TCGA pan-cancer cohort, the correlation between CXCL9 expression and tumor mutational burden (TMB) was evaluated. The CXCL9 expression values were derived from the TCGA pan-cancer RSEM standardized expression matrix of UCSC Xena, and the TMB data used the non-synonymous mutation load (TMB_NONSYNONYMOUS, mutations/Mb) at the sample level. Firstly, the TCGA sample numbering format was unified, and the expression data and TMB annotations were matched preferentially according to the complete sample ID; for samples that could not be directly matched, patient-level matching was further conducted according to the first 12 patient IDs, and the median CXCL9 expression was taken for samples with multiple expression samples from the same patient. To reduce the influence of skewed distribution, the CXCL9 expression values were converted to log2 (CXCL9 + 0.001), and TMB was converted to log10 (TMB_NONSYNONYMOUS + 0.01). Only samples with both valid indicators were retained. Subsequently, a two-sided Spearman rank correlation test was used to evaluate the correlation between CXCL9 expression and TMB in the pan-cancer combined cohort, and stratified analysis was performed according to cancer types; the Spearman correlation coefficient (rho) and P value were calculated for cancer types with a valid sample number of no <10. The P value for the cancer type test was corrected for multiple comparisons using the Benjamini-Hochberg method, with FDR < 0.05 considered statistically significant. The 95% confidence interval of the correlation coefficient was calculated through Fisher z transformation and presented in a forest plot with rho, 95% confidence interval, and effective sample size for each cancer type. The correlation statistics for the pan-cancer scatter plot were based on all valid samples.

Cell culture and administration

Human lung adenocarcinoma cell lines HCC827 (CL-0094, Procell, China), HCC44 (CL-0901, Procell, China) and normal human lung epithelial cells BESA-2B (ZQ0381, Cell Research, China) were cultured in RPMI-1640 medium or DMEM high-sugar culture medium supplemented with 10% fetal bovine serum and 1% penicillin–streptomycin at 37 °C in a humidified incubator with 5% CO₂. HCC827 harbors an activating EGFR exon 19 deletion and displays an epithelial phenotype with sensitivity to EGFR tyrosine kinase inhibitors [18], whereas HCC44 carries a KRAS G12C mutation, and exhibits a mesenchymal phenotype with high EMT/stemness markers and primary resistance to targeted therapies [19]. This combination allowed us to evaluate the functional impact of CXCL9 across genetically and phenotypically distinct LUAD backgrounds

For cytokine stimulation experiments, recombinant human CXCL9 protein (HY-P7253, MCE, USA) was added to HCC44 cells at the 10 ng/mL, and cells were harvested at different time points (0, 15, 30, and 60 min) for downstream analysis of signaling activation. For pharmacological inhibition studies, cells were treated with the JAK1/2 inhibitor Ruxolitinib (HY-50,856, MCE, USA) at a final concentration of 5 μM for the indicated duration prior to analysis. Cells were treated with the CXCR3 antagonist AMG487 (HY-15,319, MCE, USA), a small-molecule inhibitor with reported selectivity for CXCR3 over other chemokine receptors, at 5 μM for 2 h before subsequent assays.

Wild-type HCC44 cells were seeded in 6-well plates and serum-starved for 12 h. Cells were then treated with CM collected from CXCL9-OE or control cells for 30 min. After treatment, cells were harvested and subjected to Western blot analysis for p-STAT3 and β-actin.

CXCL9 knockout and overexpression

We developed a short hairpin RNA (shRNA) targeting the CXCL9 gene and CXCR3 and constructed a plasmid for CXCL9 overexpression (Tsingke, China), allowing for the modulation of CXCL9 expression in cells. The shRNA plasmids were incorporated into lentiviral particles and utilized for cell transduction (Tsingke, China). Stable cell lines with CXCL9 knockdown were selected using puromycin. To achieve CXCL9 overexpression, the complete coding sequence was amplified and inserted into the pcDNA3.1 vector. The CXCL9 overexpression plasmid was then transfected into cells using Lipofectamine 3000. The effectiveness of CXCL9 knockdown and overexpression was confirmed through qRT-PCR and Western blot analysis. These modified cells were subsequently utilized in functional assays to investigate the impact of modulating CXCL9 expression. The shRNA sequence is as follows: shCXCL9 1: CCGGCATCATCTTCCTGGAGCAGTGCTCGAGCACTGCTCCAGGAAGATGATGTTTTTTG; shCXCL9 2: CCGGCTAGATCCGGACTCGGCAAATCTCGAGATTTGCCGAGTCCGGATCTAGTTTTTTG. The siRNA sequence is as follows: siCXCR3–1: 5′- CCUUCCUGCUGGCUUGUAUTT-3′ (sense), 5′-AUACAAGCCAGCAGGAAGGTT-3′ (antisense); siCXCR3–2: 5′- GGUCAUGGCCUACUGCUAUTT −3′ (sense), 5′- AUAGCAGUAGGCCAUGACCTT-3′ (antisense).

Cell migration, invasion, and apoptosis

In the wound healing assay, tumor cells were cultured until 90% confluence in six-well plates. A linear scratch was created using a sterile pipette tip, followed by washing with PBS to remove detached cells. The cells were then incubated in serum-free medium and imaged at 0 and 48 h to evaluate migration into the scratch area. For the Transwell invasion assay, tumor cells were suspended in serum-free medium and added to the upper chamber of a Matrigel-coated Transwell insert. The lower chamber contained complete medium as a chemoattractant. After 24 h of incubation, non-invading cells were removed from the upper side of the insert and invading cells on the lower side were fixed, stained, and counted under a microscope. Cell apoptosis was evaluated using an Annexin V-FITC/propidium iodide (PI) Apoptosis Detection Kit (40302ES60, Yeasen, China) according to the manufacturer’s instructions.

Xenograft tumor model

All animal experiments were approved by Medical Ethics Committee of the First Affiliated Hospital of the Air Force Medical University (approval number: 261,126) and conducted in accordance with institutional guidelines for animal welfare. Six-week-old male BALB/c nude mice were purchased from an accredited animal facility and maintained under SPF conditions with controlled temperature (22–25 °C), humidity (50–60%), and a 12 h light/dark cycle. Mice had free access to food and water.For tumor xenograft establishment, HCC44 cells stably overexpressing CXCL9 (CXCL9 OE) or corresponding vector control (Control) cells were harvested during the logarithmic growth phase, washed with phosphate-buffered saline (PBS), and resuspended in serum-free RPMI-1640 medium. A total of 5 × 10⁶ cells in 100 μL suspension were subcutaneously injected into the right flank of each mouse. Tumor growth was monitored every 2–3 days using digital calipers, and tumor volume was calculated using the formula: Volume = (length × width²) / 2. For pharmacological intervention, mice bearing established tumors were administered the JAK1/2 inhibitor Ruxolitinib (180 mg/kg) (CXCL9 OE+ Ruxolitinib). Tumor size and body weight were recorded throughout the experiment. At the experimental endpoint, mice were euthanized, and tumors were excised, photographed, and weighed. Tumor tissues were either snap-frozen for protein analysis or fixed in 4% paraformaldehyde for histological and immunohistochemical evaluation.

Western blot

Total protein was extracted from cultured cells or tumor tissues using RIPA lysis buffer (R0010, Solarbio, China) supplemented with protease and phosphatase inhibitor. Protein concentration was determined using a BCA Protein Assay Kit (PC0020, Solarbio, China) according to the manufacturer’s instructions. Equal amounts of protein (20 μg per sample) were separated by 10% SDS–PAGE and subsequently transferred onto PVDF membranes (ISEQ00010, Merck Millipore, Germany). Membranes were blocked with 5% non-fat milk or 5% BSA in TBST for 1 h at room temperature and then incubated overnight at 4 °C with primary antibodies against CXCL9 (1:1000, 22,355-1-AP, Proteintech, China), CXCR3 (1:500, 26,756-1-AP, Proteintech, China), JAK1 (1:1000, 66,466-1-Ig, Proteintech, China), phospho-JAK1 (1:2000, PJAK1-BIOTIN, Thermo Fisher, USA), STAT3 (1:2000, 10,253-2-AP, Proteintech, China), phospho-STAT3 (1:5000, 60,479-1-Ig, Proteintech, China), BIRC5 (1:500, 10,508-1-AP, Proteintech, China), and β-actin (1:10,000, EM21002, HUABIO, China). After washing with TBST, membranes were incubated with HRP-conjugated secondary antibodies for 1 h at room temperature. Protein bands were visualized using an enhanced chemiluminescence (ECL) detection system (61,006–9Q, Clinx, China) and captured using a chemiluminescence imaging system.

Quantitative real-time PCR (qRT-PCR)

Total RNA was extracted from cells using TRIzol reagent (15596026CN, Invitrogen, USA) following the manufacturer’s protocol. RNA concentration and purity were assessed using a NanoDrop spectrophotometer. Complementary DNA (cDNA) was synthesized from 1 μg of total RNA using a reverse transcription kit (RT-01,022, FOREGENE, China). Quantitative real-time PCR was performed using SYBR qPCR SuperMix Plus (M00041, LAMROLBIO, China) on a real-time PCR system. Relative gene expression levels were calculated using the 2^−ΔΔCt method. The CXCL9 primer sequences are as follows: CXCL9: F-5′-CCAGTAGTGAGAAAGGGTCGC-3′, R-5′-AGGGCTTGGGGCAAATTGTT-3′; CXCR3: F-5′-GTGGACATCCTCATGGACCT-3′, R-5′-CGGAACTTGACCCCTACAAA-3′.

Immunohistochemistry (IHC)

Tumor tissues harvested from xenograft models were fixed in 4% paraformaldehyde (E672002–0500, sangon, China) for 24 h, embedded in paraffin, and sectioned into 4 μm. Tissue sections were deparaffinized in xylene (10,023,418, Sinopharm Chemical Reagent Co., Ltd, China) and rehydrated through a graded series of ethanol solutions. For antigen retrieval, sections were heated in citrate buffer (pH 6.0) (C1031, Solarbio, China) using a microwave or pressure cooker for 10–15 min, followed by cooling to room temperature. Endogenous peroxidase activity was blocked by incubation with 3% hydrogen peroxide for 10 min. After washing with phosphate-buffered saline (PBS), sections were blocked with 5% bovine serum albumin (BSA) for 30 min at room temperature to reduce nonspecific binding. Sections were then incubated overnight at 4 °C with primary antibodies against Ki67 (1:500, 84,432-1-RR, Proteintech, China), CXCL9, and phospho-STAT3 (p-STAT3). After washing with PBS, sections were incubated with HRP-conjugated secondary antibodies for 60 min at room temperature. Immunoreactivity was visualized using a diaminobenzidine (DAB) substrate kit (G1212, Servicebio, China), followed by counterstaining with hematoxylin. Finally, sections were dehydrated, cleared, and mounted for microscopic examination. Images were captured using a light microscope (Nikon SI, Nikon, Japan). The staining intensity and percentage of positive cells were evaluated semi-quantitatively using ImageJ software.

Enzyme-linked immunosorbent assay (ELISA)

HCC44 CXCL9 overexpressing (OE) and vector-control cells were seeded in complete RPMI-1640 medium and cultured for 24 h. The medium was then replaced with serum-free RPMI-1640, and cells were incubated for an additional 24 h. Conditioned medium (CM) was collected, centrifuged at 3000 × g for 10 min to remove cellular debris, and stored at −80 °C. CXCL9 protein concentration was measured using a human CXCL9 ELISA kit (ab219047, Abcam, USA) according to the manufacturer’s instructions.

Statistical analysis

For comparisons of normally distributed data, t-tests were used. Conversely, employ non-parametric tests for data that deviates from a normal distribution, to compare differences. For continuous variables, initially partition the data into high and low expression groups based on the optimal cutoff value, subsequently employing the log-Rank test to assess prognostic outcomes. All analyses were deemed statistically significant at a significance level of P < 0.05.

Results

CXCL9 is upregulated across multiple cancer types and predicts poor prognosis in LUAD

Analysis of CXCL9 expression across 33 cancer types from The Cancer Genome Atlas (TCGA) revealed markedly elevated CXCL9 transcript levels in tumor tissues compared with matched normal tissues in multiple malignancies, including bladder urothelial carcinoma (BLCA), breast invasive carcinoma (BRCA), colon adenocarcinoma (COAD), lung adenocarcinoma (LUAD), and stomach adenocarcinoma (STAD), esophageal cancer (ESCA), glioblastoma (GBM), head and neck cancer (HNSC), kidney clear cell carcinoma (KIRC), liver cancer (LIHC), pheochromocytoma and paraganglioma (PCPG), prostate cancer (PRAD), rectal cancer (READ), endometrioid cancer (UCEC), as indicated by the boxplot distribution (Fig. 2A). In representative cancer types, CXCL9 expression was significantly upregulated in tumor tissues versus normal controls in LUAD, BRCA, COAD, and KIRC (Fig. 2B). Univariate Cox proportional hazards regression analysis was performed to evaluate the prognostic value of CXCL9 across TCGA cancer types. Elevated CXCL9 expression was significantly associated with overall survival in several cancers, including LUAD, underscoring its potential as a prognostic indicator (Fig. 2C). Kaplan–Meier survival analysis stratified by CXCL9 expression levels further confirmed that high CXCL9 expression was associated with significantly poorer overall survival in LUAD, whereas no significant prognostic association was observed in BRCA, COAD, or KIRC. These findings specifically highlight the clinical relevance of CXCL9 in lung adenocarcinoma progression and patient outcome (Fig. 2D).

Fig. 2.

Pan-cancer analysis of CXCL9 expression and its prognostic significance. (A) CXCL9 expression levels across 33 cancer types from TCGA database. Red boxes indicate tumor tissues and cyan boxes indicate normal tissues. (B) Representative boxplots showing CXCL9 expression in tumor versus normal tissues in LUAD, BRCA, COAD, and KIRC. (C) Forest plot summary of univariate Cox regression analysis for CXCL9 across various cancers. (D) Kaplan–Meier overall survival curves stratified by CXCL9 expression in LUAD, BRCA, COAD, and KIRC. HR, hazard ratio; OS, overall survival.

CXCL9 is upregulated in immunotherapy responders and enriched in immune activation pathways

Two PD-L1 data sets (GSE126044, GSE135222) were collected for analysis of potential genes affected by immunotherapy. GSE126044 showed 1119 genes with increased expression and 666 genes with decreased expression, while GSE135222 had 185 genes with high expression and 169 genes with low expression in the immune response group (Fig. 3A). Cross analysis of these data sets revealed 85 genes that were differentially expressed in both (Fig. 3B). Functional enrichment analysis identified Toll-like receptor (TLR) signaling pathway, natural killer cell proliferation, and Chemokine signaling pathway as important in PD-L1 immunotherapy (Fig. 3C).Protein interaction analysis of the 85 genes showed an interaction network among 67 genes, with CXCL9 having the highest degree of interaction (Fig. 3D). CXCL9 was highly expressed in the response groups of both data sets (Fig. 3E). To further investigate the relationship between CXCL9 and immunotherapy, serum and alveolar lavage fluid from 9 responders and 9 non-responders were collected and analyzed for CXCL9 gene expression. Results showed high expression of CXCL9 in the immune response group (Fig. 3F). Serum and BALF CXCL9 levels were strongly positively correlated (Fig. 3F). To further explore the mechanistic link between CXCL9 and the immune signaling pathways enriched in immunotherapy responders, we performed correlation analysis between CXCL9 expression and the activity of seven immune-related pathways. As shown in S1A, CXCL9 expression was significantly and positively correlated with chemokine signaling, cytokine–cytokine receptor interaction, interferon-gamma response, PD-1/PD-L1 checkpoint, Toll-like receptor signaling, JAK-STAT signaling, and IL6-JAK-STAT3 signaling across TCGA pan-cancer, TCGA-LUAD, and the two anti-PD-L1 immunotherapy cohorts.

Fig. 3.

CXCL9 is upregulated in immunotherapy responders and enriched in immune activation pathways. (A) Volcano plots of differentially expressed genes in GSE126044 and GSE135222. CXCL9 is highlighted in red. (B) Venn diagram showing the overlap of differentially expressed genes between the two cohorts. (C) GO and KEGG enrichment analyses of the 85 common differentially expressed genes. (D) PPI network of the shared genes; CXCL9 is marked with a red rectangle. (E) CXCL9 expression in NR versus R in GSE126044 and GSE135222. (F) CXCL9 protein levels in serum and BALF from NR and R groups, and the correlation between serum and BALF CXCL9 levels. *P < 0.05, **P < 0.01.

CXCL9 is predominantly expressed by monocytes/macrophages and associated with immune microenvironment remodeling in NSCLC

The analysis revealed the significant role of CXCL9 in LUAD, prompting us to conduct differential expression analysis on TCGA-LUAD samples with high and low CXCL9 expression levels. This analysis identified 1891 differentially expressed genes, including 413 highly expressed genes (Fig. 4A Left). Functional enrichment analysis of these genes showed a strong association with immune-related signaling pathways, such as cytokine receptor interaction, Th17 cell differentiation, and Chemokine signaling pathway (Fig. 4A Right). Further examination of CXCL9 expression in single-cell sequencing data revealed its predominant expression in monocytes and macrophages, with some expression in malignant cells (Fig. 4B). Although CXCL9 expression was most prominent in monocytes and macrophages, given that tumor-intrinsic CXCL9 can act in an autocrine/paracrine manner through CXCR3 to promote malignant phenotypes [20], we specifically focused on elucidating the functional role of CXCL9 in LUAD cells. To better understand the role of CXCL9 in tumor cells, we performed differential expression analysis on cells expressing CXCL9 and those not expressing it, identifying 875 differentially expressed genes (Fig. 4C Left). Differential expression analysis between CXCL9-positive and CXCL9-negative malignant cells further supported an association with immune-related programs, raising the possibility that CXCL9-producing tumor cells may simultaneously participate in and respond to the inflammatory microenvironment. Functional enrichment analysis of these genes showed a strong association with immune response (Fig. 4C Right). To identify key genes related to CXCL9, we performed WGCNA analysis on malignant cells expressing CXCL9, resulting in 14 distinct modules (Fig. 4D). Analysis of genes in each module revealed that CXCL9 and the brown module had the highest number of genes. Further investigation of the brown module identified key core genes, primarily belonging to the IFIT and CXCL families (Fig. 4E). Cell–cell communication inference analysis exhibited prominent interaction edges connecting fibroblasts, endothelial cells, macrophages, and malignant cells. Quantitative heatmaps of interaction number and interaction strength further delineated the dominant signaling axes, with fibroblasts and monocyte-derived cells serving as major senders and receivers within the CXCL9-associated communication landscape (Fig. 4F). Given the immunotherapy focus of this study, we next investigated the relationship between CXCL9 expression and tumor mutational burden (TMB). Across the entire TCGA pan-cancer dataset, CXCL9 expression showed a significant positive correlation with TMB (S1B). When analyzed by individual cancer type, CXCL9 expression was positively correlated with TMB in multiple malignancies, including LUAD (S1C)

Fig. 4.

Single-cell landscape of CXCL9 expression and its immune microenvironment associations in NSCLC. (A) Volcano plot and circos plot of CXCL9-related genes and their enriched KEGG pathways. (B) UMAP projection of cell types in the NSCLC_EMTAB6149 dataset (left) and CXCL9 expression feature plot (right). (C) Differential gene expression (left) and GO enrichment analysis (right) comparing malignant cells with high versus low CXCL9 expression. (D) Soft-threshold power selection and hdWGCNA dendrogram for fibroblast modules. (E) Circos plot of CXCL9 co-expression network (left) and brown module hub gene network (right). (F) Cell–cell communication circle plot (top), SPP1 signaling pathway network (middle), and interaction heatmaps showing the number and strength of interactions among cell types (bottom).

CXCL9 promotes proliferation, migration, and invasion of lung adenocarcinoma cells in vitro and in vivo

To functionally characterize CXCL9 in lung adenocarcinoma, stable CXCL9-knockdown HCC827 cells (shCXCL9) and CXCL9-overexpressing HCC44 cells (CXCL9 OE) were generated. Quantitative RT-PCR confirmed a reduction in CXCL9 mRNA in HCC827-shCXCL9 cells and a increase in HCC44-CXCL9-OE cells. Western blot analysis corroborated these findings at the protein level (Fig. 5A). Wound-healing scratch assays demonstrated that CXCL9 knockdown in HCC827 cells significantly attenuated cell migration at 48 h compared with control cells. Conversely, CXCL9 overexpression in HCC44 cells markedly enhanced migratory capacity (Fig. 5B). Transwell invasion assays revealed that CXCL9 knockdown significantly decreased the number of invasive HCC827 cells, whereas CXCL9 overexpression substantially increased invasiveness in HCC44 cells (Fig. 5C). Annexin V/PI flow cytometric analysis showed that CXCL9 overexpression in HCC44 cells significantly reduced the apoptotic rate compared with control cells. Reciprocally, CXCL9 knockdown in HCC827 cells significantly increased apoptosis (Fig. 5D). In vivo tumorigenicity was assessed using a nude mouse xenograft model. Ectopic overexpression of CXCL9 in HCC44 cells resulted in significantly larger tumor volumes compared with vector-control cells over a 15-day observation period (Fig. 5E). Western blot analysis of xenograft tumor lysates revealed that CXCL9 overexpression upregulated the expression of BIRC5 and p-STAT3, while total STAT3 levels remained unchanged. These molecular changes are consistent with the enhanced proliferation and survival observed in CXCL9-overexpressing tumors (Fig. 5F). To further investigate the molecular alterations associated with CXCL9-driven tumor growth in vivo, we examined the expression of the downstream transcriptional targets of STAT3, including CCL2, SNAI1, and SOX2 by Western blot. Consistent with the pro-tumorigenic phenotype, CXCL9-overexpressing tumors displayed significantly elevated protein levels of CCL2, SNAI1, and SOX2 compared with control tumors (S2C). Immunohistochemical staining of tumor sections confirmed that CXCL9-overexpressing tumors exhibited markedly increased positivity for Ki67 (proliferation index), p-STAT3, and CXCL9 itself compared with control tumors (Fig. 5G).

Fig. 5.

CXCL9 promotes proliferation, migration, and invasion of lung adenocarcinoma cells in vitro and in vivo. (A) Validation of CXCL9 knockdown in HCC827 cells and CXCL9 overexpression in HCC44 cells by qRT-PCR and Western blot. (B) Wound-healing assay showing the effect of CXCL9 knockdown or overexpression on cell migration at 0 h and 48 h. Scale bar, 200 μm. (C) Transwell invasion assay in HCC827 and HCC44 cells. Scale bar, 100 μm. (D) Annexin V/PI staining and quantification of apoptosis rates. (E) Representative images of xenograft tumors and growth curves of tumor volume. (F) Western blot analysis of BIRC5 and p-STAT3 in xenograft tumors. (G) Immunohistochemical staining of Ki67, p-STAT3, and CXCL9 in tumor sections and corresponding quantitative analysis. *P < 0.05,**P < 0.01, ***P < 0.001, ****P < 0.0001.

CXCL9 activates the JAK1/STAT3 signaling pathway in lung adenocarcinoma cells

GSEA of RNA sequencing data from HCC44-CXCL9-overexpressing cells revealed significant enrichment of the "IL6-JAK-STAT3 signaling" hallmark gene set (Fig. 6A). Western blot analysis validated the bioinformatic prediction at the protein level. In HCC827 cells, CXCL9 knockdown significantly reduced the phosphorylation levels of JAK1 and STAT3 without altering total JAK1 or STAT3 protein abundance. Conversely, in HCC44 cells, CXCL9 overexpression significantly increased p-JAK1 and p-STAT3 levels (Fig. 6B). To determine whether CXCL9-mediated JAK1/STAT3 activation regulates downstream effectors implicated in tumor aggressiveness, we evaluated the protein expression of CCL2, SNAI1, and SOX2 in HCC827-shCXCL9 and HCC44-CXCL9-OE cells. Western blot analysis revealed that CXCL9 knockdown significantly reduced the protein levels of CCL2, SNAI1, and SOX2 in HCC827 cells, whereas CXCL9 overexpression markedly increased their expression in HCC44 cells (S2 A-B). These findings suggest that CXCL9 modulates EMT- and stemness-associated markers, supporting its role in promoting migration, invasion, and tumor progression. To determine the kinetics of CXCL9-induced STAT3 activation, HCC44 cells were treated with recombinant CXCL9 and harvested at serial time points (0, 15, 30, and 60 min). Western blot analysis demonstrated a time-dependent increase in p-STAT3 levels following CXCL9 stimulation, with peak activation observed at 30–60 min, establishing a rapid and sustained signaling response (Fig. 6C). Pharmacological inhibition experiments were performed using Ruxolitinib, a potent JAK1/2 inhibitor. In HCC44-CXCL9-overexpressing cells, co-treatment with Ruxolitinib effectively abolished the elevated p-STAT3 levels induced by CXCL9 overexpression, reducing phosphorylation to near-baseline levels (Fig. 6D). Functional rescue experiments using Transwell assays showed that Ruxolitinib treatment reversed the enhanced migratory phenotype conferred by CXCL9 overexpression. While CXCL9 OE cells exhibited significantly increased invasion, the addition of Ruxolitinib reduced migratory cell numbers to levels comparable to or below the control group, confirming that the pro-migratory effect of CXCL9 is mediated through JAK1/STAT3 signaling (Fig. 6E). Annexin V/PI apoptosis analysis further demonstrated functional rescue by Ruxolitinib. CXCL9 overexpression significantly decreased the apoptotic rate, whereas co-treatment with Ruxolitinib restored apoptosis to control levels. Collectively, these findings establish that CXCL9 exerts its pro-migratory and anti-apoptotic functions, at least in part, through activation of the JAK1/STAT3 signaling axis (Fig. 6F).

Fig. 6.

CXCL9 activates the JAK1/STAT3 signaling pathway in lung adenocarcinoma cells. (A) GSEA plot (left) and enriched hallmark pathways (right) in HCC44-CXCL9-overexpressing cells. (B) Western blots of p-JAK1, JAK1, p-STAT3, and STAT3 in HCC827-shCXCL9 and HCC44-CXCL9-OE cells. (C) Time-course analysis of p-STAT3 activation following CXCL9 treatment in HCC44 cells. (D) Western blot showing that Ruxolitinib blocks CXCL9-induced p-STAT3 upregulation. (E) Transwell migration assay and (F) Annexin V/PI apoptosis analysis in HCC44 cells treated with CXCL9 overexpression alone or in combination with Ruxolitinib. *P < 0.05, ** P < 0.01, ***P < 0.001, **** P < 0.0001.

Inhibition of JAK1/STAT3 reverses CXCL9-promoted tumor growth in vivo

CXCL9 overexpression resulted in substantially larger tumor masses compared with the control group, whereas concurrent Ruxolitinib administration markedly reduced tumor size (Fig. 7A-B). Growth curve analysis demonstrated that CXCL9-overexpressing tumors grew significantly faster than control tumors from Day 10 onward. Importantly, Ruxolitinib treatment significantly suppressed the accelerated growth kinetics of CXCL9-overexpressing tumors, resulting in volumes comparable to or smaller than the control group at endpoint (Fig. 7C). Endpoint tumor volume measurements quantified at Day 16 confirmed that CXCL9 OE tumors were significantly larger than control tumors, while CXCL9 OE + Ruxolitinib tumors were significantly smaller than CXCL9 OE tumors (Fig. 7D). Consistent with volume data, endpoint tumor weight measurements revealed that CXCL9 overexpression significantly increased tumor weight, and Ruxolitinib co-treatment completely abrogated this effect (Fig. 7E). Immunohistochemical analysis of tumor tissues demonstrated that CXCL9 overexpression significantly increased the expression of Ki67 and p-STAT3 compared with control tumors. Notably, Ruxolitinib treatment markedly attenuated both Ki67 and p-STAT3 staining intensity in CXCL9-overexpressing tumors (Fig. 7F).

Fig. 7.

Inhibition of JAK1/STAT3 reverses CXCL9-promoted tumor growth in vivo. (A) Photographs of nude mice from each group. (B) Tumor images from Control, CXCL9 OE, and CXCL9 OE + Ruxolitinib groups. (C) Tumor volume growth curves over 16 days. (D and E) Endpoint tumor volume and tumor weight. (F) Representative IHC images of Ki67 and p-STAT3 expression in tumor tissues. *P < 0.05, **P < 0.01, ***P < 0.001, **** P < 0.0001.

CXCL9 activates STAT3 signaling through its receptor CXCR3

CXCR3 is a receptor for CXCL9, to elucidate the receptor mechanism underlying CXCL9-induced STAT3 activation, HCC44-CXCL9-overexpressing cells were treated with AMG487, a selective CXCR3 antagonist. Western blot analysis showed that AMG487 treatment significantly suppressed the elevated p-STAT3 levels in CXCL9-overexpressing cells, reducing phosphorylation to levels comparable to the control group (Fig. 8A). qRT-PCR and Western blot analyses showed that CXCR3 was expressed at both the mRNA and protein levels in these LUAD cell lines, confirming their capacity to respond to CXCL9 stimulation (Fig. 8B). Genetic knockdown of CXCR3 was performed in HCC44 cells using specific siRNA. Western blot confirmed that CXCR3 protein expression was efficiently reduced (Fig. 8C). In CXCL9-overexpressing HCC44 cells, simultaneous knockdown of CXCR3 by siRNA significantly abrogated CXCL9-induced p-STAT3 upregulation. While CXCL9 OE + scramble siRNA cells maintained high p-STAT3 levels, CXCL9 OE + CXCR3-siRNA cells exhibited markedly reduced p-STAT3 (Fig. 8D). To validate the proposed autocrine CXCL9–CXCR3 signaling loop, we first measured CXCL9 protein secretion in the conditioned medium (CM) of HCC44-CXCL9-OE cells. ELISA analysis revealed that HCC44-CXCL9-OE cells secreted significantly higher levels of CXCL9 protein compared with vector-control cells (S3A). To further determine whether secreted CXCL9 is biologically active, we treated wild-type HCC827 and HCC44 cells with CM derived from HCC44-CXCL9-OE or control cells. Western blot analysis showed that CM from CXCL9-overexpressing cells markedly increased STAT3 phosphorylation in recipient cells, whereas CM from control cells had minimal effect (S3B). Collectively, these findings demonstrate that LUAD cells can actively secrete CXCL9 protein and activate the JAK1/STAT3 pathway through an autocrine and/or paracrine mechanism. Taken together, the convergent findings from pharmacological blockade with AMG487 and genetic silencing of CXCR3 strongly support the conclusion that CXCL9-induced STAT3 activation is mediated specifically through CXCR3, rather than through alternative receptors or non-specific effects.

Fig. 8.

CXCL9 activates STAT3 signaling through its receptor CXCR3. (A) Western blot and quantification of p-STAT3 in HCC44 cells overexpressing CXCL9 and treated with DMSO or AMG487. (B) Western blot and qRT-PCR analysis of CXCR3 mRNA expression in BEAS-2B, HCC827 and HCC44 cells. (C) Efficiency of CXCR3 knockdown by siRNA in HCC44 cells confirmed by Western blot. (D) Western blot and quantification showing that CXCR3 silencing abrogates CXCL9-induced p-STAT3 upregulation. *P < 0.05, ** P < 0.01, ***P < 0.001, **** P < 0.0001.

Discussion

In this study, we systematically investigated the clinical significance, biological function, and underlying molecular mechanisms of CXCL9 in lung adenocarcinoma (LUAD). Through integrated bioinformatics analyses of pan-cancer databases, anti-PD-1 immunotherapy cohorts, and single-cell RNA sequencing data, we identified CXCL9 as a prognostic biomarker associated with poor overall survival in LUAD. Functionally, gain and loss of function experiments demonstrated that CXCL9 promotes malignant phenotypes, including enhanced proliferation, migration, and invasion, as well as suppressed apoptosis, in LUAD cells both in vitro and in vivo. Mechanistically, we elucidated that CXCL9 activates the JAK1/STAT3 signaling cascade via its cognate receptor CXCR3, thereby driving tumor progression. Importantly, pharmacological inhibition of JAK1/2 by Ruxolitinib effectively reversed CXCL9-mediated oncogenic effects, highlighting the therapeutic potential of targeting the CXCL9/CXCR3/JAK1/STAT3 axis in LUAD.

During our pan-cancer analysis of CXCL9, we found that CXCL9 has inconsistent effects on tumors in different tumor types. Numerous studies have highlighted the tumor-suppressive effects of CXCL9. For instance, research indicates that CXCL9 can enhance the infiltration of cytotoxic T cells into tumors, thereby promoting anti-tumor immunity. One study demonstrated that increased expression of CXCL9 in NSCLC models led to improved responses to immune checkpoint inhibitors by facilitating T cell activation and recruitment to the tumor microenvironment [21]. Similarly, engineered chemokine-based approaches have been shown to inhibit colorectal cancer growth by remodeling the tumor microenvironment and enhancing immune cell recruitment, underscoring the broad potential of chemokine-mediated microenvironment modulation in cancer immunotherapy [22]. Another investigation found that CXCL9 expression was associated with favorable clinical outcomes in patients with glioblastoma, suggesting its role in inhibiting tumor growth through immune-mediated mechanisms [20]. Additionally, a study reported that CXCL9 can induce apoptosis in cancer cells via the activation of specific signaling pathways, further supporting its potential as a therapeutic target in cancer treatment [23]. Conversely, CXCL9 has also been implicated in promoting tumor progression under certain conditions. Research has shown that while CXCL9 can attract immune cells to the tumor site, it may also facilitate a pro-tumorigenic environment by promoting chronic inflammation. For example, a study on pancreatic cancer revealed that specific mutations could enhance the expression of CXCL9, leading to an immunosuppressive microenvironment that supports tumor growth and metastasis. Furthermore, another investigation found that high levels of CXCL9 were correlated with advanced disease stages in various cancers [[24], [25], [26], [27], [28], [29]], suggesting its involvement in fostering an environment conducive to tumor progression rather than suppression. The context-dependent function of CXCL9 may be further shaped by broader molecular and stress-related cues within the tumor microenvironment. For example, Ning et al. reported that chronic hepatitis B patients with rtA181T-mutated HBV infection exhibit increased mutation rates of tumour suppressor genes and a higher risk of hepatocellular carcinoma, illustrating how viral molecular alterations can reshape tumour-associated inflammatory responses and drive cancer progression [30]. Similarly, Ma et al. demonstrated that piR-16,404 drives ferroptotic liver injury through dysregulation of the CASTOR1/mTORC1/GPX4 axis, highlighting how stress-associated signaling pathways can modulate tissue injury and local inflammation [31]. lncRNA-mediated upregulation of ENTPD1 in osimertinib-resistant NSCLC promotes immunosuppressive microenvironment remodeling and resistance, suggesting that CXCL9-related inflammatory circuits may also be co-opted during treatment adaptation [32]. Collectively, these studies support the notion that the biological outcome of CXCL9 signaling is not intrinsic but is heavily influenced by the surrounding inflammatory, mutational, and metabolic context, which may account for its divergent roles across different malignancies

Our single-cell analysis revealed that monocytes and macrophages are the predominant CXCL9-expressing populations in NSCLC, with lower but detectable expression in malignant cells. The in vitro and in vivo experiments in the present study were therefore designed to specifically test the tumor-intrinsic consequences of CXCL9 signaling, namely that LUAD cells expressing CXCL9 can activate CXCR3/JAK1/STAT3 in an autocrine or paracrine manner to promote proliferation, migration, invasion, and survival. This does not exclude a major contribution of immune cell-derived CXCL9, which may recruit T cells and NK cells to shape the anti-tumor immune response, as supported by our immunotherapy cohort analyses. In fact, CXCL9 derived from macrophages or other stromal cells could also signal through CXCR3 expressed on tumor cells, creating a paracrine axis that complements tumor-intrinsic CXCL9 [33]. Future studies using cell-type-specific genetic manipulation, tumor–immune co-culture systems, or immunocompetent models will be essential to dissect the relative contributions of tumor-derived versus immune-derived CXCL9 to LUAD progression and immunotherapy response.

In the analysis of sequencing data sets involving PD-L1 responders and non-responders, it was observed that the mutated genes primarily function in the Toll-like receptor signaling pathway. The correlation between the Toll-like receptor (TLR) signaling pathway and PD-L1 immunotherapy is significant in cancer treatment, particularly in enhancing immune responses. Studies have shown that TLRs, which play a crucial role in initiating immune responses, can upregulate PD-L1 expression on tumor cells, thereby influencing immune evasion mechanisms [34]. For example, a study revealed that TLR2 activation resulted in increased PD-L1 expression in head and neck squamous cell carcinoma [35], suggesting that microbial stimuli can modulate immune suppression through this pathway. Furthermore, TLR agonists have been found to enhance the effectiveness of PD-1/PD-L1 blockade therapies by promoting antitumor immunity, as they can increase the infiltration of CD8+ T cells within tumors. Additionally, the combination of TLR-targeted therapies and PD-L1 inhibitors has shown promise in overcoming resistance to immunotherapy, indicating a synergistic effect that could potentially improve patient outcomes. These findings underscore the significance of comprehending TLR signaling in optimizing PD-L1-based immunotherapies.

Our research on CXCL9 overexpression and knockdown has revealed its potential to promote the proliferation and invasion of lung cancer tumor cells. Studies have also shown that CXCL9 may contribute to tumor aggressiveness by promoting the formation of metastatic niches. For example, the activation of the CXCR3 receptor, which is expressed on certain tumor-initiating cells, by CXCL9 and CXCL10 has been linked to increased cancer cell proliferation and lung metastasis [36]. Additionally, high levels of CXCL9 have been associated with improved postoperative survival and reduced tumor aggressiveness in colorectal carcinoma [37], suggesting a complex role where it may have both inhibitory and promoting effects on tumor dynamics depending on the context. Furthermore, CXCL9′s involvement in recruiting immune cells to the tumor microenvironment may facilitate tumor growth and metastasis, making it a potential biomarker for prognosis while also acting as a mediator of metastatic processes. These findings highlight the need for further investigation into the multifaceted roles of CXCL9 in the progression of lung cancer.

The JAK/STAT pathway is a well-established regulator of tumor progression, mediating cell proliferation, survival, immune evasion, and metastasis [38]. Previous studies have shown that activation of JAK1 leads to phosphorylation of STAT3, which subsequently translocates to the nucleus and induces transcription of genes involved in cell survival [39]. Consistent with these reports, our data revealed that CXCL9 overexpression significantly increased phosphorylation levels of JAK1 and STAT3 without affecting total protein expression, suggesting that CXCL9 primarily regulates pathway activation rather than protein abundance.

Importantly, our time-course experiments further demonstrated that recombinant CXCL9 induces rapid and sustained STAT3 phosphorylation, peaking at 30–60 min, indicating a direct signaling effect. This is consistent with previous studies showing that chemokine-mediated signaling can rapidly activate the JAK/STAT pathway through receptor-dependent mechanisms, particularly via chemokine receptor engagement that triggers downstream phosphorylation and transcriptional responses [40]. Moreover, we identified CXCR3 as the functional receptor mediating this process, as both pharmacological inhibition with AMG487 and genetic knockdown of CXCR3 abrogated CXCL9-induced STAT3 activation. These results support the concept of a CXCL9–CXCR3–JAK1/STAT3 signaling axis, which has been increasingly recognized as a key regulatory pathway in tumor biology. The progression of lung cancer is driven by a complex interplay of genetic, epigenetic, and environmental factors. In this broader context, recent work by Xu et al. demonstrated that the piRNA PMLCPIR promotes PM2.5-induced lung cancer, highlighting how non-coding RNAs and environmental insults can cooperate to dysregulate signaling pathways and accelerate lung malignancy [41]. Our findings extend this mechanistic landscape by identifying the CXCL9–CXCR3–JAK1/STAT3 axis as a chemokine-mediated signaling cascade that similarly promotes LUAD cell proliferation, migration, invasion, and survival.

One of the important mechanistic findings from our in vivo experiments is that the expression levels of BIRC5, CCL2, SNAI1, and SOX2 are significantly increased in xenograft tumors with excessive CXCL9 expression. BIRC5 is a typical downstream target of the STAT3 signaling pathway and is a potent inhibitor of apoptosis, which can promote cell proliferation and survival [42], CCL2 [43], SNAI1 [44], and SOX2 [45] are downstream regulatory molecules of the JAK1/STAT3 axis. The upregulation of the expression levels of BIRC5, CCL2, SNAI1, and SOX2 in tumors with excessive CXCL9 expression provides crucial evidence for the proposed CXCL9/CXCR3/JAK1/STAT3 axis, which links the STAT3 activation driven by CXCL9 with a functionally related downstream effector molecule.

Ruxolitinib, a selective JAK1/2 inhibitor, has been widely studied in both inflammatory diseases and cancer [46]. Mechanistically, Ruxolitinib suppresses phosphorylation of downstream STAT proteins, thereby inhibiting transcription of pro-survival and pro-inflammatory genes [47]. In agreement with these studies, our results showed that Ruxolitinib effectively abolished CXCL9-induced STAT3 phosphorylation and reversed the enhanced migratory and anti-apoptotic phenotypes in CXCL9-overexpressing cells. Furthermore, in vivo experiments demonstrated that Ruxolitinib significantly reduced tumor growth and decreased expression of Ki67 and p-STAT3 in xenograft tumors, confirming its therapeutic efficacy in this model.

In addition, previous studies have demonstrated that aberrant activation of the JAK/STAT pathway can regulate immune cell infiltration, cytokine secretion, and immune escape within tumors [48]. Emerging therapeutic strategies aim to exploit regulated cell death pathways and advanced drug delivery systems to improve cancer treatment. For example, Wang et al. demonstrated that arsenic trioxide-based nanoparticles enhance chemotherapy by activating pyroptosis, highlighting the potential of nanomedicine-based delivery to potentiate conventional cytotoxic agents [49]. In addition, Yang et al. developed a novel GSH depletor that simultaneously induces ferroptosis and cuproptosis in hepatocellular carcinoma, underscoring the therapeutic value of targeting cellular redox metabolism [50]. Furthermore, studies on CXCL9 and other CXCR3 ligands have shown that these chemokines play important roles in recruiting immune cells, particularly T lymphocytes and NK cells, and may thereby influence the efficacy of immunotherapy [51]. Interestingly, research on Ruxolitinib has revealed that this JAK1/2 inhibitor can modulate CXCL9/CXCR3 axis-related signaling and alter cytokine production, suggesting that it may exert dual effects on both tumor cell signaling and immune regulation [47]. Therefore, targeting the CXCL9–CXCR3–JAK/STAT axis may have broader implications beyond tumor-intrinsic signaling, potentially affecting tumor–immune interactions.

Our functional data establish that CXCL9 activates STAT3 through CXCR3 and JAK1, however, the upstream events that drive CXCL9 induction in LUAD remain to be fully characterized. CXCL9 is a canonical IFN-γ inducible gene, and its transcription is tightly regulated by JAK1/STAT1 signaling and interferon regulatory factor 1 (IRF1) [52]. In addition, TLR signaling and downstream NF-κB activation can cooperate with IFN-γ to amplify CXCL9 expression in both tumor and immune cells [53]. The enrichment of TLR signaling, chemokine signaling, cytokine–cytokine receptor interaction, and Th17 cell differentiation in our bioinformatics analyses suggests that LUAD cells with high CXCL9 are situated within a highly inflammatory network. In this context, CXCL9 may function not only as a downstream effector of IFN-γ/STAT1 and TLR/NF-κB inputs but also as an upstream activator of STAT3 through CXCR3, thereby creating a feed-forward loop that reinforces tumor cell survival and immune remodeling. The WGCNA-identified IFIT and CXCL family co-expression modules further support the integration of CXCL9 within an interferon-responsive gene program. Future studies using upstream regulator prediction tools and pathway interaction networks will be needed to formally dissect the hierarchical relationships among these immune signaling pathways and the CXCL9–CXCR3–JAK1/STAT3 axis.

Our findings add CXCL9 to the growing list of prognostic biomarkers in lung malignancies. For example, Xiao et al. recently identified membrane progesterone receptor alpha (mPRα/PAQR7) as a prognostic biomarker across human cancers and validated its predictive value in lung squamous cell carcinoma [54]. While mPRα/PAQR7 and CXCL9 both associate with patient outcome, CXCL9 offers the additional advantage of being linked to immunotherapy response and a druggable signaling axis (CXCR3/JAK1/STAT3). In parallel, the integration of artificial intelligence and multi-omics approaches is reshaping biomarker-driven cancer stratification, as highlighted by Hussain et al. in the context of triple-negative breast cancer [55]. These advances suggest that future NSCLC prognostic models may benefit from combining CXCL9 with other molecular features through machine-learning frameworks, ultimately improving patient selection for immunotherapy and targeted JAK/STAT inhibition.

Despite our discoveries, several limitations of the present study should be acknowledged. First, although our integrated analyses combined TCGA datasets, public immunotherapy cohorts, and single-cell sequencing data, the clinical validation cohort was relatively small, comprising only nine responders and nine non-responders to PD-L1 immunotherapy. No formal power calculation was performed a priori, as this analysis was exploratory and hypothesis-generating. Consequently, the observed association between CXCL9 expression and PD-L1 immunotherapy response should be interpreted with caution until validated in larger, prospective, multicenter cohorts. We note, however, that the direction of the association was supported by independent analyses of two public immunotherapy datasets (GSE126044 and GSE135222), which partially mitigates the constraint imposed by the small in-house cohort. Future studies should prioritize prospective enrollment of immunotherapy-treated LUAD patients across multiple centers to perform adequately powered validation of CXCL9 as a predictive biomarker and to evaluate its interaction with established predictors such as PD-L1 expression and tumor mutational burden. Second, a conceptual limitation of our study is the use of nude mouse xenograft models for the in vivo experiments. Because CXCL9 is fundamentally a T-cell-recruiting chemokine and our study emphasizes its role in immune-related signaling and immunotherapy response, the lack of intact T-cell immunity in nude mice represents an important constraint. Nude mice were nevertheless chosen to facilitate engraftment of human HCC44 cells and to specifically investigate tumor-intrinsic CXCL9–CXCR3–JAK1/STAT3 signaling in vivo. An immunocompetent or syngeneic model was not used because human LUAD cell lines are rejected in immunocompetent mice, and a fully syngeneic model would require murine lung adenocarcinoma cells, which may differ from human LUAD in CXCL9 regulation and immune context. Future studies should validate our findings in immunocompetent syngeneic models, genetically engineered mouse models (GEMMs), or humanized immune system mice to comprehensively assess both tumor-intrinsic and immune-mediated effects of CXCL9. Third, although we demonstrated that CXCL9 promotes tumor progression through activation of the CXCR3/JAK1/STAT3 axis, additional downstream effectors and upstream regulatory mechanisms may also contribute to the biological effects of CXCL9 in LUAD and warrant further investigation. Fourth, our single-cell analysis indicated that CXCL9 is predominantly expressed by monocytes and macrophages, whereas our functional experiments focused on tumor-intrinsic CXCL9 signaling in LUAD cell lines. Therefore, the relative contributions of tumor-derived versus immune-derived CXCL9, and the extent to which paracrine CXCL9 from immune cells affects tumor cell behavior in vivo, remain to be formally determined. Finally, Smoking status [56], EGFR mutation background [57] and tumor stage [58] are all known to influence immune cell infiltration, and immunotherapy responsiveness in LUAD. Although the integrated datasets used in this study contained partial clinical annotations, we did not perform stratified analyses across all of these variables due to limited subgroup sizes and incomplete treatment history records. Therefore, our results should be interpreted as exploratory, and future studies in well-annotated, larger cohorts are warranted to clarify whether the prognostic and predictive value of CXCL9 is independent of these confounders.

Conclusion

In conclusion, our study identified CXCL9 as a key regulator associated with PD-L1 immunotherapy response and tumor progression in lung adenocarcinoma. Integrated bioinformatics analyses and experimental validation demonstrated that CXCL9 is highly expressed in immunotherapy responders and promotes LUAD cell proliferation, migration, invasion, and anti-apoptotic ability. Mechanistically, CXCL9 secreted by LUAD cells binds CXCR3 and activates JAK1/STAT3 signaling, while inhibition with Ruxolitinib effectively reversed its oncogenic effects both in vitro and in vivo (Fig. 9). Collectively, these findings provide novel insights into the molecular mechanisms underlying immunotherapy response and highlight the CXCL9/CXCR3/JAK1/STAT3 axis as a potential prognostic biomarker and therapeutic target for LUAD.

Fig. 9.

CXCL9 promotes lung cancer progression through activation of the CXCR3/JAK1/STAT3 signaling axis.

Consent for publication

All authors have collectively provided verbal consent for the publication of this work for academic and research purposes.

Ethical approval statement

All patient samples used in this study were derived from a research project that received ethical approval, with informed consent signed by all participants. The collection and handling of samples were conducted in accordance with the relevant human experimental ethics regulations and approved by the Ethics Committee of The First Affiliated Hospital of Fourth Military Medical University (Ethics number: KY20260746). Throughout the study, ethical standards were strictly followed to ensure patient privacy and data security. This study has been reviewed by the Laboratory Animal Welfare and Ethics Committee of The First Affiliated Hospital of Fourth Military Medical University and meets animal ethics and animal welfare requirements. Ethics number: 261126).

Clinical trial number

Not applicable.

Funding

This work was supported by National Natural Science Foundation of China, General Program (NO. 82473215) , Key Project in Shaanxi Province Natural Science Basic Research Program (NO. 2024JC-ZDXM-45) and Shaanxi Sanqin Special Support Program Innovation and Entrepreneurship Team - Precise Diagnosis, Treatment and Standardized Management of Lung Cancer.

Data availability statement

The datasets generated and analyzed during the present study are available from the corresponding author on reasonable request.

CRediT authorship contribution statement

Qing Ju: Writing – review & editing, Writing – original draft, Validation, Supervision, Software, Resources, Project administration, Formal analysis, Data curation, Conceptualization. Xiangxiang Chen: Writing – review & editing, Writing – original draft, Supervision, Resources, Project administration, Investigation, Funding acquisition, Formal analysis, Data curation. Ning Chang: Visualization, Validation, Supervision, Project administration, Methodology, Investigation, Data curation. Ying Zhou: Visualization, Supervision, Resources, Formal analysis, Data curation. Dan Qiu: Supervision, Resources, Project administration, Investigation. Yong Zhang: Writing – review & editing, Supervision, Software, Methodology, Investigation, Funding acquisition, Data curation, Conceptualization. Jian Zhang: Writing – review & editing, Visualization, Supervision, Software, Project administration, Methodology, Funding acquisition, Data curation.

Declaration of competing interest

All of the authors declare that there is no conflict of interest.

Appendix. Supplementary materials

Supplementary 1: (A) Heatmap showing Spearman correlation coefficients between CXCL9 expression and the activity of seven immune-related pathways across TCGA pan-cancer, TCGA-LUAD, GSE126044, and GSE135222 datasets. Asterisks indicate statistical significance. (B) Pan-cancer scatter plot showing the correlation between CXCL9 expression (log2 RSEM) and tumor mutational burden (TMB, log10 non-synonymous mutations/Mb). Spearman correlation coefficient and P value are indicated. (C) Forest plot showing Spearman correlation coefficients between CXCL9 expression and TMB across individual cancer types. Error bars represent 95% confidence intervals. Asterisks indicate statistical significance*P < 0.05, ** P < 0.01, ***P < 0.001.

Supplementary 2: (A) Western blot analysis and quantitative analysis of CCL2, SNAI1, and SOX2 protein levels in HCC827 cells following CXCL9 knockdown. (B) Western blot analysis and quantitative analysis of CCL2, SNAI1, and SOX2 protein levels in HCC44 cells following CXCL9 overexpression. (C) Western blot analysis and quantitative analysis of CCL2, SNAI1, and SOX2 protein levels in xenograft tumor tissues from the control and CXCL9-overexpression groups. β-Actin was used as a control. *P < 0.05, **P < 0.01, ****P < 0.0001.

Supplementary 3: (A) ELISA quantification of CXCL9 protein concentration in conditioned medium (CM) from HCC44-CXCL9-overexpressing (OE) and vector-control cells. (B) Western blot showing p-STAT3 activation in HCC44 cells treated with CM from CXCL9-OE or control cells. *P < 0.05, ** P < 0.01, ***P < 0.001.