The flow chart of patient recruitment was shown in Figure 1 . Between January 2019 and November 2024, 306 patients who received anti-therapies at Department of Oncology, the Affiliated Hospital of Yangzhou University (n=206), Nothern Jiangsu People’s Hospital Affiliated Yangzhou University (n=58) and Baoying Traditional Chinese Medicine Hospital (n=42) were initially included. The inclusion criteria were as follows: 1) patients aged over 18; 2) pathologically diagnosed as lung cancer; 3) patients receiving ICIs or combined with other anti-cancer therapies including chemotherapy, radiotherapy, and targeted therapy. The exclusion criteria were as follows: 1) multiple existing primary tumors; 2) incomplete medical records; 3) data missing in the follow-up; 4) insufficient ICI therapy (less than two cycles); 5) hyperprogression. Hyperprogression was defined as Evaluation Criteria in Solid Tumors (RECIST) version 1.1 progressive disease at the first computed tomography (CT) scan during ICI therapy and an absolute increase in the tumor growth rate exceeding 50% per month (20). As a result, a total of 235 patients were finally included. This study was approved by the local ethics committee (No. 2022-YKL11-class 05). The informed consents were acquired from patients for using their tissue samples and medical information in scientific researches.

For ICI therapy, the following drugs were used: sintilimab (n=84), tirelizumab (n=60), camrelizumab (n=29), pembrolizumab (n=13), durvalumab (n=12), serplulimab (n=11), toripalimab (n=11), atezolizumab (n=8), nivolumab (n=4), adebrelimab (n=2) and envafolimab (n=1). The management of drug toxicities was carried out according to National Comprehensive Cancer Network (NCCN) guidelines (21). Corticosteroids could be used for treating most high-grade toxicities and permanent discontinuation of ICI drugs was suggested in case of severe toxicities. For monitoring drug toxicities, patients received physical examination and laboratory detection every two or three weeks. In case of pulmonary or cardiovascular toxicity, CT or ultrasound examination are recommended. For chemotherapy, the following drugs were used: platinum (n=175), etoposide (n=55), paclitaxel (n=48), pemetrexed (n=46), nab-paclitaxel (n=37), docetaxel (n=7) and gemcitabine (n=5). In case that severe toxicities were observed, the drug dose were reduced by 25-50% or another regimen was recommended. For targeted therapy, the following drugs were used: bevacizumab (n=14), anlotinib (n=11) and apatinib (n=2). A total of 22 patients received radiotherapy. Chemotherapy combined with immunotherapy was the most commonly used, the representative chemotherapeutic agents were pemetrexed (500 mg/m²) plus cisplatin (75 mg/m²) or carboplatin (area under the curve=5 mg/mL/min) once every 3 weeks.

The therapy response was evaluated every two or three cycles, based on serum tumor biomarkers and radiological examination. The RECIST version 1.1 was used to determine Complete Response (CR), Partial Response (PR), Progressive Disease (PD) and Stable Disease (SD). The clinical outcome was determined using OS and PFS. The OS was defined as the period from ICI initiation to death from any cause, while PFS was defined as the period from ICI initiation to PD.

The STGs were obtained from the DrugBank (https://go.drugbank.com) (22), Comparative Toxicogenomics Database (https://ctdbase.org/) (23), Swiss Target Prediction (http://www.swisstargetprediction.ch/) (24) and TargetNet (http://targetnet.scbdd.com/calcnet/index/) (25) databases.

The transcriptome data of Lung Adenocarcinoma (LUAD) and Lung Squamous Cell Carcinoma (LUSC) from The Cancer Genome Atlas (TCGA, https://portal.gdc.cancer.gov/v1/) database were downloaded. The favorable prognostic genes were identified using the univariate cox regression method with the following inclusion criteria: HR<1 and p<0.05. Then, the shared genes between the favorable prognostic genes and STGs were identified and their prognostic significance was validated using the Kaplan-Meier survival model (p<0.01). The associations between the identified genes and clinical features were determined using UALCAN database (https://ualcan.path.uab.edu/index.html) (26).

For quantifying immune cell infiltration, the following algorithms were used: EPIC, MCP-COUNTER, TIMER, XCELL, QUANTISEQ, CIBERSORT and CIBERSORT-ABS. The immune subtypes were classified as follows: C1 (Wound healing), C2 (IFN-gamma dominant), C3 (Inflammation), C4 (lymphocyte depletion), C5 (Immunologically Quiet), and C6 (TGF-beta dominant). The tumor immune micro-environment cell composition database (TIMEDB) was used to investigate the correlation of STGs with immune cells (https://timedb.deepomics.org/) and the details were provided in Supplementary Table S1 (27). For further confirming the correlation, the cellular distribution of STGs were analyzed using single-cell sequencing data from TISCH database (http://tisch.comp-genomics.org/home/).

The tumor and adjacent normal tissues were collected from 42 NSCLC patients who received ICI-based therapy at Department of Oncology, the Affiliated Hospital of Yangzhou University. The paraffin embedded tissue samples were cut into 4μm-thick sections, then dewaxed using xylene and rehydrated using gradient alcohol. After antigen retrieval and blocking endogenous peroxidase activity, the sections were incubated with the primary antibody against RAR-related orphan receptor alpha (RORA) (Proteintech, USA, 1:200) or CD8 (Cell Signaling Technology, USA, 1:1000) overnight. The sections were then incubated with the secondary antibody (Cell Signaling Technology, 1:1000) for 30 min and staining was visualized using Diaminobenzidine Kit (Thermo Fisher Scientific, USA). For staining assessment for RORA expression, the scoring system was calculated based on staining Intensity (SI) and Percentage of Positive stained cells (PP). SI was classified as follows: negative (score 0), weak (score 1), moderate (score 2) and strong (score 3). PP was classified as follows: ≤5% (score 0), 6-25% (score 1), 26-50% (score 2), 51-75% (score 3) and ≥76% (score 4). A final score was calculated by multiplying the scores of SI and PP. High RORA expression was defined as a final score more than 6. For staining assessment for CD8 expression, more than 30 positively stained cells per field (×400) in the tumor stroma was defined as CD8⁺ T cell rich tumor tissues, while the opposite case was defined as CD8⁺ T cell deficient tumor tissues.

A549 cells were obtained from the Type Culture Collection of the Chinese Academy of Sciences (Shanghai, China) and were cultured in complete F-12K medium (Procell, USA) supplemented with 10% fetal bovine serum (FBS) (Thermo Fisher Scientific), 1% L-alanyl-L-glutamine and 1% penicillin/streptomycin. The cells were maintained at 37°C in a humidified incubator with 5% CO₂.

The sequences for RORA knockdown (KD) and negative control (NC) were as follows: KD: (Forward: 5’-GCUUCUACCUGGACAUACA-3’, Reverse: 5’-UGUAUGUCCAGGUAGAAGC-3’); NC: (Forward: 5’-UUCUCCGAACGU GUCACGU-3’, Reverse: 5’-ACGUGACACGUUCGGAGAA-3’). The plasmid construction, lentiviral packaging and transfection was performed according to our previous study (28).

Total protein from A549 cells was extracted using cell lysis buffer (Servicebio, China). Equal amounts of protein were separated by 10% SDS–polyacrylamide gel electrophoresis and transferred onto a nitrocellulose membrane. The membrane was blocked at room temperature for 30 min, followed by incubation with a primary antibody against RORA (Proteintech, 1:1000) at 4°C overnight. Then, the membrane was incubated with horseradish peroxidase-conjugated goat anti-rabbit IgG secondary antibody (Beyotime, China, 1:1000) at room temperature for 1h. Finally, the membrane was visualized using an enhanced chemiluminescence detection kit (Beyotime).

A549 cells were seeded in a 96-well plate and incubated for 24 hours. The cells were then treated with different concentrations of Atorvastatin (MedChemExpress, USA) and incubated for 24, 48, 72, or 96 hours. After washing with PBS, 10 µL of CCK-8 solution (Beyotime) was added. After 1h incubation, absorbance at 450 nm was measured using a microplate reader.

A549 cells were plated into six-well culture dishes and maintained under humidified conditions at 37°C with 5% CO₂ for two weeks. After incubation, the colonies were fixed in 4% paraformaldehyde for 20 minutes, followed by staining with crystal violet solution (Beyotime) for 20 minutes and counted.

The invasive and migratory capabilities of A549 cells were evaluated using a transwell chamber assay. Briefly, A549 cells were seeded into the upper chamber with serum-free medium, with the lower chamber filled with complete medium. After 24h incubation, the cells were fixed with paraformaldehyde (Beyotime), stained with crystal violet and counted under a microscope. For the invasion assay, a matrix gel was added to the upper chamber before the experiment.

The statistical analysis was performed using SPSS software (version 25.0), GraphPad prism (version 8.0) and R project (version 4.3.0). The clinical correlation analysis was performed using the chi-square test. The survival curves were plotted using the Kaplan-Meier model and the survival difference was compared using the log-rank test. The factors affecting OS or PFS were identified using the univariate and multivariate analysis based on cox proportional hazards regression model. Comparisons between two groups were evaluated using t-test, while differences among more than two groups were assessed by one-way analysis of variance (ANOVA). Statistical significance was defined as a p-value < 0.05.

The baseline clinical characteristics of included patients were shown in Table 1 . 202 patients (86.0%) were males. The median age at initial diagnose was 69 years, ranging from 48 to 88 years. 199 patients (84.7%) were pathologically diagnosed as NSCLC, while the rest (15.3%) were diagnosed as small cell lung cancer (SCLC). 36 patients (15.3%) have previously received surgical treatment. 123 patients (52.3%) had a history of smoking. 216 patients (91.9%) received ICI drugs combined other anti-therapies, while the rest (8.1%) received ICI monotherapy. There were 13 patients (5.5%) with EGFR mutation site, 40 patients (17.0%) with other mutation sites, and 182 patients (77.4%) without sequencing information. There were 64 patients (27.2%) with negative PD-L1 expression, 94 patients (40.0%) with positive PD-L1 expression, and 77 cases (32.8%) without information about PD-L1 expression. 80 patients (34.0%) received statins, and the most commonly used drug type was Rosuvastatin (n=58), followed by Atorvastatin (n=12) and Simvastatin (n=10). The correlation analysis demonstrated that statin use was significantly correlated with coronary artery disease(p<0.001), while no significant difference was observed in the result clinical features between the statin users and non-statin users (all p>0.05).

In the entire cohort, the statin users had a significantly better OS (p=0.001, Figure 2A ) and PFS (p<0.001, Figure 2B ) than the non-statin users. The univariate analysis revealed smoking history, ECOG, tumor stage and statin use were significantly correlated with both OS and PFS (all p<0.05, Figures 2C, D ). The multivariate analysis confirmed statin use together with smoking history, ECOG and tumor stage were independent factors affecting OS and PFS (all p<0.05, Figures 2E, F ). In terms of treatment response ( Supplementary Figure S1 ). After the first evaluation, the statin users had significantly higher objective response rate (ORR) and disease control rate (DCR) than the non-statin users.

In the subgroup analysis, the statin use was associated with improved OS and PFS in patients aged below 70 years ( Supplementary Figures S2A, B ). In contrast, among patients aged 70 years or older, the statin use was associated with improved PFS, whereas no statistically significant difference was observed in OS ( Supplementary Figures S2C, D ). Regarding smoking status, statin use was linked to improved OS and PFS in smokers ( Supplementary Figures S3A, B ). In the non-smokers, statin use was significantly associated with improved PFS instead of OS ( Supplementary Figures S3C, D ). In patients with NSCLC, the statin use was associated with improved OS and PFS ( Supplementary Figures S4A, B ), and similar results were observed in patients with SCLC ( Supplementary Figures S4C, D ). In patients with stage III, no significant differences in outcome were observed between the statin and non-statin group ( Supplementary Figures S5A, B ). In patients with IV stage, the statin group had significantly better prognosis than the non-statin group ( Supplementary Figures S5C, D ). Finally, the use of rosuvastatin was found to associate with improved OS and PFS ( Supplementary Figures S6A, B ).

For clarifying the mechanisms underlying the impact of statins on ICI drugs, we focused on STGs. As shown in Supplementary Figure S7A , target genes of three representative statin drugs (Atorvastatin, n=718; Rosuvastatin, n=550; Simvastatin, n=1058) were initially obtained from four online drug databases. Then, a total of 504 shared target genes were identified ( Supplementary Figure S7B ) and the details were provided in Supplementary Table S2 . Meanwhile, a total of 457 favorable prognostic genes were identified through the univariate cox regression using the TCGA database and the details were provided in Supplementary Table S3 . As shown in Supplementary Figure S7C , a total of 11 overlap genes were finally determined between 504 STGs and 457 favorable prognostic genes. For further confirming the prognostic significance of these 11 genes, the Kaplan-Meier survival model was utilized. As a result, high expression of transcription factor RORA was found to significantly associate with better survival of patients with lung cancer (p=0.003, Figure 3A and Supplementary Figure S8 ). The expression of RORA was also found to be significantly reduced in tumor tissues as compared with normal lung tissues from patients with lung cancer ( Figure 3B ). The receiver operator characteristic (ROC) curve analysis revealed the diagnostic area under the curve (AUC) for distinguish tumors from normal tissues is 0.809 ( Figure 3C ). In addition, the analysis based on the UALCAN database confirmed that RORA expression was significantly reduced in tumor tissues from lung cancer patients within TNM stage I-IV as compared with that in normal lung tissues ( Figure 3D ). The cellular assay demonstrated RORA overexpression ( Supplementary Figure S9A ) significantly inhibited the proliferation ( Supplementary Figure S9B ), colony formation ( Supplementary Figure S9C ), invasion ( Supplementary Figure S9D ), and migration ( Supplementary Figure S9E ) of lung cancer cells.

As shown in Supplementary Figure S10A , the RORA expression was positively correlated with infiltration of numerous immune cells such as B cells, CD4⁺ and CD8⁺ T cells. Using the TCGA cohort including 867 lung cancer patients, high RORA expression was predominantly correlated with immune C1 type (Wound Healing), followed by immune C2 type (IFN-gamma Dominant) and immune C3 type (Inflammatory) ( Supplementary Figure S10B ). For further confirming the correlation between RORA and immune cells, two NSCLC single-cell datasets were utilized. As shown in Figure 4A , RORA expression was positively correlated with NK cells in NSCLC_GSE127465 dataset and CD8⁺ T cells in NSCLC_GSE131907 dataset. This result was also confirmed by gene localization analysis in Figure 4B .

Since the bioinformatics analysis has closely linked RORA expression with CD8⁺ T cells, we next validated this correlation in tissue samples from a cohort including 42 ICI-treated patients with lung cancer and patient information was provided in Supplementary Table S4 . The expression of RORA and CD8 in normal and tumor tissues were detected using IHC and representative staining images were shown Figure 5A . The positive expression of RORA were mainly detected in the cell nucleus while that of CD8 were mainly found in the tumor stroma. According to the evaluation criterion, 16 and 26 cases were defined as high and low RORA expression respectively, while 13 and 29 cases defined as CD8⁺ T cell richness and deficiency. The correlation analysis demonstrated that RORA expression was positively correlated with CD8⁺ T cell abundance (r=0.884, p<0.001, Figure 5B ). In addition, a significant positive correlation was also observed between statin use and CD8⁺T cells richness (p=0.032). Finally, high RORA expression in tumor tissues was found to associate with better OS ( Figure 5C ) and PFS ( Figure 5D ) in lung cancer patients receiving ICI based therapy.

For further confirming the correlation between statins and RORA, we selected Atorvastatin for cellular assays. The CCK-8 assay determined that the optimum inhibitory concentration of Atorvastatin was 40 μM ( Figure 6A ). Using this concentration, Atorvastatin treatment effectively impaired the viability of lung cancer cells within 96h ( Figure 6B ). The western blot demonstrated that Atorvastatin treatment significantly increased the protein expression of RORA ( Figure 6C ). The function assays demonstrated Atorvastatin treatment effectively inhibited the colony formation ( Figure 6D ), invasion ( Figure 6E ), and migration ( Figure 6F ) of lung cancer cells, while RORA knockdown could partly rescue this inhibitory impact.