Causal association of circulating cytokines with the risk of lung cancer: a Mendelian randomization study

Background Lung cancer is the deadliest and most prevalent malignancy worldwide. While smoking is an established cause, evidence to identify other causal factors remains lacking. Current research indicates chronic inflammation is involved in tumorigenesis and cancer development, though the specific mechanisms underlying the role of inflammatory cytokines in lung cancer pathogenesis remain unclear. This study implemented Mendelian randomization (MR) analysis to investigate the causal effects of circulating cytokines on lung cancer development. Methods We performed a two-sample MR analysis in Europeans utilizing publicly available genome-wide association study summary statistics. Single nucleotide polymorphisms significantly associated with cytokine were selected as genetic instrumental variables. Results Genetically predicted levels of the chemokine interleukin-18 (IL-18) (OR = 0.942, 95% CI: 0.897–0.990, P = 0.018) exerted significant negative causal effects on overall lung cancer risk in this analysis. Examining specific histologic subtypes revealed further evidence of genetic associations. Stem cell factor (SCF) (OR = 1.150, 95% CI: 1.021–1.296, P = 0.021) and interleukin-1beta (IL-1β) (OR = 1.152, 95% CI: 1.003–1.325, P = 0.046) were positively associated with lung adenocarcinoma risk, though no inflammatory factors showed causal links to squamous cell lung cancer risk. Stratified by smoking status, interferon gamma-induced protein 10 (IP-10) (OR = 0.861, 95% CI: 0.781–0.950, P = 0.003) was inversely associated while IL-1β (OR = 1.190, 95% CI: 1.023–1.384, P = 0.024) was positively associated with lung cancer risk in ever smokers. Among never smokers, a positive association was observed between lung cancer risk and SCF (OR = 1.474, 95% CI: 1.105–1.964, P = 0.008). Importantly, these causal inferences remained robust across multiple complementary MR approaches, including MR-Egger, weighted median, weighted mode and simple mode regressions. Sensitivity analyses also excluded potential bias stemming from pleiotropy. Conclusion This MR study found preliminary evidence that genetically predicted levels of four inflammatory cytokines—SCF, IL-1β, IL-18, and IP-10—may causally influence lung cancer risk in an overall and subtype-specific manner, as well as stratified by smoking status. Identifying these cytokine pathways that may promote lung carcinogenesis represents potential new targets for the prevention, early detection, and treatment of this deadly malignancy.


Introduction
As the deadliest cancer worldwide, lung cancer was responsible for 1,817,000 deaths in 2022, accounting for 18.7% of all cancerrelated fatalities (1).Due to typically late-stage diagnosis once symptoms arise, the 5-year survival rate for lung cancer patients is dismally low at around 20% (2).Smoking is a well-known predominant risk factor, yet evidence to elucidate other potential causes remains scarce (3,4).Identifying additional contributory risk factors to promote early detection and treatment thus represents a critical need in combating this devastating disease.
Over the past two decades, a sizable body of evidence has firmly established chronic infection and inflammation as key promoters of carcinogenesis (5)(6)(7)(8).The inflammatory tumor microenvironment, comprised of leukocytes releasing cytokines, chemokines, reactive oxygen species and other cytotoxic mediators, can drive tumor progression through processes such as genotoxicity, aberrant tissue repair, heightened cellular proliferation, invasion and metastasis (5,9).Critical transcription factors including STAT3 and NF-kB have been implicated in inflammation-fueled carcinogenesis (10,11).Tumors can further manipulate the inflammatory milieu to suppress anti-tumor T cell responses (12).However, despite substantial links between cytokines and cancer, the precise underlying mechanisms, particularly in lung cancer, remain to be fully deciphered.
Mendelian randomization (MR) analysis utilizes genetic variation as an instrumental variable to infer the potential causal relationship between modifiable exposures and disease outcomes (13).As genetic variation is randomly assigned and not susceptible to reverse causation, this ingenious approach circumvents confounding and reverse causation bias that plague traditional observational studies (14).In this study, we implemented MR to investigate the causal role of circulating cytokines in lung cancer pathogenesis.Using genetic instruments as surrogates for cytokine levels, we evaluated whether inflammation actively drives lung cancer development or simply represents an epiphenomenon.
Elucidating these causal pathways will provide novel mechanistic insights and could reveal previously unrecognized therapeutic targets for this pernicious lung disease.

Method Study design
In this two-sample MR analysis, we utilized single nucleotide polymorphisms (SNPs) as instrumental variables (IVs).To ensure validity, SNPs were chosen based on three critical assumptions: (1) significant associations with exposure factors, satisfying relevance; (2) effects on outcomes solely via exposures, not through alternative pathways, fulfilling exclusivity; and (3) independence from confounding factors (15).

Data resource
SNPs associated with circulating cytokines were obtained from the latest genome-wide association studies (GWAS), as listed in Supplementary Table 1.Summary data from a large-scale cytokine GWAS meta-analysis were used to generate genetic instruments for cytokines (16).This meta-analysis measured cytokine levels in plasma and blood samples from 8,293 Finnish individuals across three population-based cohorts.
To investigate causal effects of circulating cytokines on lung cancer risk, we obtained lung cancer GWAS summary statistics from the IEU open database.Lung cancer data were obtained from the International Lung Cancer Consortium (ILCCO) (https:// ilcco.iarc.fr/),including 29,836 cases and 55,586 controls (17).The study also provided associations between instrumental SNPs and different histologic subtypes of lung cancer, including lung adenocarcinoma (case: 11,245, control: 54,619), squamous cell carcinoma (case: 7,704, control: 54,763), and small cell lung cancer (case: 2,791, control: 20,580).Subgroup analyses were performed according to smoking status, including smokers (23,223 cases and 16,964 controls) and never-smokers (2,355 cases and 7,504 controls), and all populations were restricted to European ethnic groups.

SNPs selection
To identify valid SNPs for MR, we implemented several filtering steps: First, we selected independent SNPs strongly associated with different cytokine levels (P <5×10 −6 ) to maximize instrument availability (18,19).Second, we used clumping to prune correlated SNPs in linkage disequilibrium (r 2 <0.001, 10,000 kb) and avoid biased results.Third, we excluded pleiotropic SNPs associated with potential confounders including smoking, diabetes and anxiety using PhenoScanner (20).Fourth, we retained only concordant SNPs between exposure and outcome datasets as valid instruments.Finally, we excluded weak instruments with Fstatistics <10 (calculated as F = R 2 ×(N-2)/1-R 2 ) to minimize bias (21).

Statistical analysis
After SNP filtering, our primary MR analysis utilized inversevariance weighted (IVW) estimation to evaluate the overall causal effects, given its accuracy when all instruments are valid (22).Complementary approaches including weighted median, MR Egger, weighted mode and simple mode were also implemented (23,24).To probe potential horizontal pleiotropy, we performed MR Egger regression and Mendelian randomization pleiotropy residual sum and outlier (MR-PRESSO) testing (25,26).Heterogeneity was assessed via Cochran's Q and MR Egger regression, and result robustness was verified through leave-oneout analysis.Furthermore, we conducted Steiger testing to evaluate possible reverse causation (27).
To correct for multiple hypothesis testing, we applied Bonferroni correction and set statistical significance at P < 0.0012 based on the number of cytokines analyzed.P-values ranging from 0.0012 to 0.05 were considered suggestive evidence for potential causality (28,29).All Mendelian randomization analyses were implemented in R utilizing the TwoSampleMR and MR-PRESSO packages.A P-value below 0.05 was deemed statistically significant.

Causal effects of circulating cytokines on lung cancer and different subtypes
After implementing the described filtering procedures, 4-16 SNPs were retained as instruments for the circulating cytokines (Supplementary Table 1).High F-statistics confirmed all selected SNPs were robust instruments (all F values >10).We then leveraged these SNPs to infer causality between cytokines and lung cancer.MR estimates from the various analytical methods are visualized in Figure 1 (Supplementary Table S2).Specifically, IVW analysis revealed interleukin-18 (IL-18) exerted significant negative causal effects on lung cancer risk (OR = 0.942, 95% CI: 0.897-0.990,P = 0.018), while eotaxin had significant positive causal effects (OR = 1.061, 95% CI: 1.002-1.123,P = 0.043) (Figures 1A, B).Eotaxin, however, has a statistical efficacy of only 67%, thus it remains to be verified.Consistent estimates were yielded by MR Egger and weighted median methods.Scatter plots visualized the effects for each method across datasets (Supplementary Figure 1).
MR Egger regression and MR-PRESSO global testing showed no evidence of horizontal pleiotropy (Table 1 and Supplementary Tables 6-9).Importantly, Cochran's Q statistics and MR Egger regression revealed no significant heterogeneity among the individual SNP instruments (P > 0.05).MR-PRESSO global test p-values exceeding 0.05 for all cytokinelung cancer associations, including different pathological types, further ruled out pleiotropy.Leave-one-out analysis confirmed the robustness of the causal estimates (Supplementary Figures 4-6).
Among the other 35 examined cytokines, none demonstrated significant correlation with lung cancer risk or pathological types in IVW or secondary MR analyses (Tables 1-4).Heterogeneity testing showed no significant heterogeneity for any of the cytokines (Tables 5-8).Across all cytokines, MR Egger regression consistently revealed no evidence of pleiotropy (Tables 5-8).MR-PRESSO outlier testing validated the significant MR findings, except for monocyte chemoattractant protein-3 (MCP-3) in lung cancer (including different pathologic types) (Supplementary Tables 6-9), and macrophage migration inhibitory factor (MIF) and IL-5 in smallcell lung carcinoma, where limited SNPs were available (Supplementary Table 9).Furthermore, Steiger p-values <0.05 verified the detected causal direction was correct for all cytokines (Table 9-12).

Subgroup analyses
Additional subgroup analyses stratified by smoking status were conducted to investigate whether the causal effect of cytokines on lung cancer risk was modified by smoking.In ever smokers, we found that interferon gamma-induced protein 10 (IP-10) (OR = 0.861, 95% CI: 0.781-0.950,P = 0.003) was inversely associated with lung cancer risk, while IL-1b (OR = 1.190, 95% CI: 1.023-1.384,P = 0.024) was positively associated with lung cancer risk (Supplementary Tables 10, 11).Scatter plots visualized the effects for each method across datasets (Supplementary Figures 7, 8).Heterogeneity tests showed no significant heterogeneity across cytokines, except for tumor necrosis factor (TNF)-a (Supplementary Table 12).MR-Egger regression revealed no evide nc e of horizont al pl eiotropy for an y cyt okin e (Supplementary Table 12).The MR-PRESSO outlier test

FIGURE 2
Odds ratio cyclic heatmap of the MR analysis results for circulating cytokines and risk of squamous cell lung cancer.

A B
(A) Odds ratio cyclic heatmap of the MR analysis results for circulating cytokines and risk of lung cancer.(B) Forest plot of results with statistical significance.Frontiers in Oncology frontiersin.orgconfirmed significant MR findings, except for TNF-b, TNF-a, and MCP-3, due to limited SNP availability (Supplementary Table 13).

Discussion
In this MR study, we investigated potential causal relationships between 41 circulating cytokines and lung cancer risk.Our analysis provides preliminary evidence that genetically predicted levels of certain cytokines may influence cancer susceptibility.We identified four inflammatory mediators -SCF, IL-1b, IL-18, and IP-10involved in genetic susceptibility to lung cancer overall and in

A B
(A) Odds ratio value cyclic heatmap of the MR analysis results for circulating cytokines and risk of small cell lung carcinoma.(B) Forest plot of results with statistical significance.

A B
(A) Odds ratio value cyclic heatmap of the MR analysis results for circulating cytokines and risk of lung adenocarcinoma.(B) Forest plot of results with statistical significance.

Tsao et al. revealed an association between low serum eotaxin-1 and
shorter progression-free survival in non-small cell lung cancer patients on vandetanib (34,35).The statistical power of eotaxin was calculated to be only 67% and had a non-significant p-value after Bonferroni correction.In summary, although eotaxin is closely linked to pathogenesis and prognosis in some cancers, further research is warranted to elucidate its precise role in lung cancer.SCF activates the c-Kit signaling pathway to stimulate lung cancer cell proliferation, migration and invasion (36).In vitro experiments show SCF promotes lung cancer cell proliferation, migration and metastasis (37)(38)(39).Moreover, clinical studies find elevated SCF and c-Kit expression correlates with lung cancer progression and metastasis (40,41).IL-1b advances lung cancer progression through numerous mechanisms including inducing angiogenic factors like VEGF, activating oncogenic signaling, and promoting immunosuppression and inflammation (42)(43)(44).Mechanistically, IL-1b spurs tumor growth by activating MAPK/NF-kB pathways, recruiting immunosuppressive cells, promoting inflammation, angiogenesis, invasion and metastasis, downregulating tumor suppressors, upregulating oncogenes, and enabling immune evasion and inhibition of apoptosis (45).Preclinical studies also demonstrate beneficial effects of IL-1b knockdown and inhibition.No inflammation or angiogenesis occurred in IL-1b-deficient mice, and IL-1b-deficient mice showed no local tumor growth or lung metastasis compared to wild-type mice (46).Moreover, IL-1b antibody treatment inhibited tumor progression and boosted antitumor immunity in mice by reducing inflammation and promoting M1 macrophage maturation (47).VEGF is a pivotal driver of angiogenesis in lung cancer, promoting tumor growth and metastasis (48).Furthermore, IL-1b induces VEGF secretion, which also spurs tumor expansion and spread (49).In lung cancer, VEGF overexpression critically supports angiogenesis and correlates with disease progression and prognosis (50).Multiple studies demonstrate VEGF inhibition slows lung cancer growth and improves chemotherapy efficacy (51,52).Anti-VEGF monoclonal antibody therapies including bevacizumab are now widely used in lung cancer treatment (53).Similarly, the calculated statistical power of VEGF is only 21%.In contrast, increased genetic predisposition to higher interleukin-18 (IL-18) and interferon gamma-induced protein 10 (IP-10) levels are negatively associated with lung cancer risk.IL-18, a pro-inflammatory cytokine, plays a significant role in anti-tumor immunity through several mechanisms.Firstly, IL-18 activates natural killer (NK) cells and certain subsets of T cells, enhancing their cytotoxic activity against tumor cells by upregulating cytotoxic molecules such as perforin and granzymes, which are crucial for direct tumor cell killing (54).Secondly, IL-18 stimulates the production of key anti-tumor cytokines, notably interferongamma (IFN-g), which has potent anti-tumor effects, including the inhibition of tumor cell proliferation, induction of tumor cell apoptosis, and enhancement of antigen presentation to help the immune system better recognize and target tumor cells (55).Thirdly, IL-18 facilitates the infiltration of immune cells, p a r t i c u l a r l y T c e l l s a n d N K c e l l s , i n t o t h e t u m o r microenvironment, thereby enhancing the overall immune response against the tumor (56).Additionally, IL-18 promotes the differentiation of T cells towards a Th1 phenotype; Th1 cells produce IFN-g and activate macrophages and other immune cells that contribute to anti-tumor immunity, making this shift towards a Th1-dominated response crucial for effective anti-tumor activity (57,58).Furthermore, IL-18 often works synergistically with other cytokines and immune signals to amplify the immune response, In contrast, increased genetic predisposition to higher levels of interleukin-18 (IL-18) and interferon gamma-induced protein 10 (IP-10) is negatively associated with lung cancer risk.IL-18, a proinflammatory cytokine, plays a significant role in anti-tumor immunity through several mechanisms.Firstly, IL-18 activates natural killer (NK) cells and certain subsets of T cells, enhancing their cytotoxic activity against tumor cells by upregulating cytotoxic molecules such as perforin and granzymes, which are crucial for direct tumor cell killing (54).Secondly, IL-18 stimulates the production of key anti-tumor cytokines, notably interferon-gamma (IFN-g), which has potent anti-tumor effects, including the inhibition of tumor cell proliferation, induction of tumor cell apoptosis, and enhancement of antigen presentation to help the immune system better recognize and target tumor cells (55).Thirdly, IL-18 facilitates the infiltration of immune cells, particularly T cells and NK cells, into the tumor microenvironment, thereby enhancing the overall immune response against the tumor (56).Additionally, IL-18 promotes the differentiation of T cells towards a Th1 phenotype; Th1 cells produce IFN-g and activate macrophages and other immune cells that contribute to anti-tumor immunity, making this shift towards a Th1-dominated response crucial for effective anti-tumor activity (57,58).Finally, IL-18 often works synergistically with other cytokines and immune signals to amplify the immune response, significantly boosting IFN-g production and enhancing the cytotoxic activity of NK cells and T cells when combined with IL-12 (59, 60).IL-18 also has anti-angiogenic and pro-lymphangiogenic properties contributing to its anti-tumor activity (61).However, Jiang et al. found that IL-18 may promote metastasis by inhibiting E-cadherin expression (62) (63,64).IP-10 attracts immune cells to tumors, inhibits angiogenesis, and reduces tumor burden (65).Previous studies confirm IP-10 participates in anti-tumor immunity by promoting immune cell migration to tumors, inducing apoptosis, and suppressing angiogenesis (66)(67)(68)(69).In xenograft models of lymphoma, squamous cell carcinoma, and lung adenocarcinoma, CXCL10 production negatively correlated with tumor growth and significantly reduced tumor-associated angiogenesis (70).In advanced endometrial cancer, CXCL10 was shown to antagonize fibroblast growth factor action, thereby inhibiting angiogenesis.In estrogen receptor-positive breast tumors, CXCL10 inhibits vascular endothelial growth factor levels to reduce tumor burden (71).
We utilized Mendelian randomization to evaluate associations between circulating cytokines and lung cancer risk.Contrary to a prior study by Bouras et al. showing positive CTACK-nonsmoking lung cancer and negative IL-18-lung cancer/adenocarcinoma relationships (72), we found no evidence for CTACKnonsmoking lung cancer or IL-18-adenocarcinoma associations, potentially attributable to divergent instrument variable selection and GWAS data pooling.Notably, our analysis revealed novel causal links between SCF, IL-1b, IL-18, and IP-10 in overall lung cancer as well as specific histological subtypes, highlighting important etiological roles for these cytokines.However, some limitations should be considered.First, the relaxed IV significance threshold of P < 5×10 -6 introduces possible false positives and bias, although the consistent F-statistics >10 suggest weak instrument bias is less likely.Second, the single Finnish ethnicity limits generalizability to other populations.Third, no cytokines were statistically significantly associated with cancer risk or subtypes after Bonferroni correction, including six inflammatory factors with suggestive correlations -Eotaxin, SCF, IL-1b, VEGF, IL-18, and IP-10.However, excluding Eotaxin and VEGF, the other factors (SCF, IL-1b, IL-18, and IP-10) had statistical power over 80% but still require validation of these potential associations in larger cohorts and GWAS.Fourth, while efforts were made to mitigate confounding, pleiotropy cannot be completely ruled out.Finally, while not addressed here, inflammatory factors may influence lung cancer progression and survival rather than development.Therefore, further studies should analyze the role of inflammatory factors in lung cancer aggressiveness.

Conclusion
This MR study found preliminary evidence that genetically predicted levels of four inflammatory cytokines-SCF, IL-1b, IL-18, and IP-10-may causally influence lung cancer risk overall, in specific histologic subtypes, and stratified by smoking status.The identification of these cytokine pathways, which may promote lung carcinogenesis, represents potential new targets for the prevention, early detection, and treatment of lung cancer.Overall, these findings reveal putative causal effects of circulating cytokines in lung cancer pathogenesis, illuminating cytokine-mediated immunological mechanisms affecting susceptibility across lung cancer subtypes and smoking exposure groups.

Abbreviations
Conversely, experimental studies by Xiong et al. and Chen et al. demonstrated that IL-18 inhibited tumor proliferation and growth, enhanced apoptosis, and normalized the Th1/Th2 imbalance tumor necrosis factor-beta TRAIL TNF-related apoptosis inducing ligand VEGF vascular endothelial growth factor

TABLE 1
Primary results of MR analysis on squamous cell lung cancer.

TABLE 2
Primary results of MR analysis on lung cancer.

TABLE 3
Primary results of MR analysis on lung adenocarcinoma.

TABLE 5
Heterogeneity and pleiotropy analyses for lung cancer.

TABLE 6
Heterogeneity and pleiotropy analyses for squamous cell lung cancer.

TABLE 7
Heterogeneity and pleiotropy analyses for lung adenocarcinoma.

TABLE 8
Heterogeneity and pleiotropy analyses for small cell lung carcinoma.