Lipid-lowering drugs affect lung cancer risk via sphingolipid metabolism: a drug-target Mendelian randomization study

Background: The causal relationship between lipid-lowering drug (LLD) use and lung cancer risk is controversial, and the role of sphingolipid metabolism in this effect remains unclear. Methods: Genome-wide association study data on low-density lipoprotein (LDL), apolipoprotein B (ApoB), and triglycerides (TG) were used to develop genetic instrumental variables (IVs) for LLDs. Two-step Mendelian randomization analyses were performed to examine the causal relationship between LLDs and lung cancer risk. The effects of ceramide, sphingosine-1-phosphate (S1P), and ceramidases on lung cancer risk were explored, and the proportions of the effects of LLDs on lung cancer risk mediated by sphingolipid metabolism were calculated. Results: APOB inhibition decreased the lung cancer risk in ever-smokers via ApoB (odds ratio [OR] 0.81, 95% confidence interval [CI] 0.70–0.92, p = 0.010), LDL (OR 0.82, 95% CI 0.71–0.96, p = 0.040), and TG (OR 0.63, 95% CI 0.46–0.83, p = 0.015) reduction by 1 standard deviation (SD), decreased small-cell lung cancer (SCLC) risk via LDL reduction by 1 SD (OR 0.71, 95% CI 0.56–0.90, p = 0.016), and decreased the plasma ceramide level and increased the neutral ceramidase level. APOC3 inhibition decreased the lung adenocarcinoma (LUAD) risk (OR 0.60, 95% CI 0.43–0.84, p = 0.039) but increased SCLC risk (OR 2.18, 95% CI 1.17–4.09, p = 0.029) via ApoB reduction by 1 SD. HMGCR inhibition increased SCLC risk via ApoB reduction by 1 SD (OR 3.04, 95% CI 1.38–6.70, p = 0.014). The LPL agonist decreased SCLC risk via ApoB (OR 0.20, 95% CI 0.07–0.58, p = 0.012) and TG reduction (OR 0.58, 95% CI 0.43–0.77, p = 0.003) while increased the plasma S1P level. PCSK9 inhibition decreased the ceramide level. Neutral ceramidase mediated 8.1% and 9.5% of the reduced lung cancer risk in ever-smokers via ApoB and TG reduction by APOB inhibition, respectively, and mediated 8.7% of the reduced LUAD risk via ApoB reduction by APOC3 inhibition. Conclusion: We elucidated the intricate interplay between LLDs, sphingolipid metabolites, and lung cancer risk. Associations of APOB, APOC3, and HMGCR inhibition and LPL agonist with distinct lung cancer risks underscore the multifaceted nature of these relationships. The observed mediation effects highlight the considerable influence of neutral ceramidase on the lung cancer risk reduction achieved by APOB and APOC3 inhibition.


Introduction
The incidence and mortality rates of lung cancer have consistently increased in accordance with global patterns of malignant tumor (Sharma, 2022).Despite continuous advancements and refinements in lung cancer treatment, the pursuit of effective prevention methods to mitigate the risk of developing this disease remains paramount (Kordiak et al., 2022).Observational studies have suggested a potential association between statin use and a reduced risk of lung cancer (Khurana et al., 2007;Yang et al., 2015;Kwon et al., 2019).However, a case-control study conducted within the UK population reported an increased risk of lung cancer associated with long-term statin exposure (Vinogradova et al., 2011).Furthermore, recent Mendelian randomization (MR) studies have challenged the notion of a causal relationship (Carter et al., 2020;Min et al., 2023).Investigations of lipidomic profiles revealed that the use of statins, PCSK9 inhibitors, and fenofibrate, instead of NPC1L1 inhibitors, results in a reduction in plasma ceramide and sphingomyelin levels (Ng et al., 2014;Tarasov et al., 2014;Ng et al., 2015;Croyal et al., 2018).Nonetheless, findings regarding the impact of statins on plasma sphingosine-1-phosphate (S1P) levels have exhibited inconsistency (Egom et al., 2013;Therond and Chapman, 2022).Preclinical investigations have elucidated the significant role of sphingolipid metabolism in the pathogenesis of lung cancer (Ogretmen, 2018;Meng et al., 2021;Tang et al., 2023), although the precise relationship remains unclear (Lin et al., 2022).Among the sphingolipids, ceramide and S1P have garnered particular attention within lung cancer studies (Goldkorn et al., 2013).Notably, ceramidases play a pivotal role in the conversion of ceramide to S1P (Pyne and Pyne, 2010).While a case-control study has uncovered associations between higher concentrations of plasma sphingosine-1-phosphate and ceramides and the increasing risk of lung cancer, these findings contradict the preclinical results that suggest that ceramides promote apoptosis in lung cancer cells (Ogretmen and Hannun, 2004;Carpinteiro et al., 2008;Alberg et al., 2013).Evidence on how lipid-lowering drugs (LLDs) may influence the risk of lung cancer by modulating sphingolipid metabolites remains elusive, along with the potential underlying mechanisms.
To explore the intricate relationships among LLDs, sphingolipid metabolites (including plasma ceramide, ceramidase, and S1P), and the risk of lung cancer, we employed a drug-target MR analysis (Holmes et al., 2021) to address a critical knowledge gap and provide valuable insights into potential chemoprevention in lung cancer.

Study design
The results obtained from MR analysis, using data from genome-wide association studies (GWAS) to analyze the effects of exposure factors, closely resemble the findings of randomized controlled trials.This resemblance can be attributed to the random allocation of genetic variants during meiosis (Yarmolinsky et al., 2022).The expression and functionality of drug targets can be profoundly affected by genetic variations.Moreover, the impact of drugs can be anticipated through the genetic variability present in the genes encoding their protein targets (Chauquet et al., 2021).Similar to the previous study, instrumental variables (IVs) for LLDs were extracted from GWAS summary statistics related to low-density lipoprotein (LDL), apolipoprotein B (ApoB), and triglycerides (TG), enabling an analysis of the causal relationship between LLD exposure and the risk of developing lung cancer (Williams et al., 2020;Xiao et al., 2023).To clarify the mediating role of sphingolipid metabolites, we employed a two-step MR analysis.This involves utilizing the outcome variable from the initial MR analysis as the exposure variable for the subsequent MR analysis (Adams and Boutwell, 2021;Xu et al., 2022).Our analysis investigates the impact of LLDs on plasma sphingolipid metabolite (ceramide, S1P, and ceramidase) levels, alongside establishing a causal link between these sphingolipid metabolites and lung cancer risk.Additionally, we quantify the mediating influence of sphingolipid metabolites on the association between LLDs and lung cancer risk.
To generate IVs for LLDs, we used the LLD target genes, involving four targets of drugs for decreasing LDL (APOB, HMGCR, NPC1L1, and PCSK9), six targets of drugs for decreasing ApoB (ANGPTL3, APOB, APOC3, LPL, PPARA, and PCSK9), and five targets of drugs for decreasing TG (ANGPTL3, APOB, APOC3, PPARA, and LPL) (Xiao et al., 2023).We confined the single-nucleotide polymorphisms (SNPs) to those located within a 100-kb range surrounding the LLD target genes that exhibited genome-wide significance in their association with the lipid traits (p < 5.0 × 10 −8 ).To maximize the IV strength for each LLD target, SNPs were allowed to be in weak linkage disequilibrium (r 2 < 0.30, window size = 10,000 kb) with each other (Xie et al., 2023).Given the absence of any genetic variation in PPARA during the selection process, it has been excluded from subsequent evaluations (Li et al., 2023).
IV strength was assessed using the F-statistic, with values > 10 indicating non-weak IVs (Duan et al., 2021).Additionally, for a positive control analysis, GWAS summary statistics for coronary heart disease (CHD) obtained from CARDIoGRAM (ieu-a-8, n = 86,995) were utilized as the outcome measure (Huang et al., 2021).

Statistical analysis
MR analysis for the effects of LLDs on the risk of lung cancer and sphingolipid metabolite levels was calculated.The clumping procedure (r 2 < 0.001, window size = 10,000 kb) was performed to remove linkage disequilibrium, and SNPs that exhibited genomewide significance (p < 5.0 × 10 −8 ) were selected as IVs for ceramide, S1P, and acid/neutral ceramidases (Fang et al., 2023) (Supplementary Table S3).
When only one SNP was available as the IV, we utilized the Wald ratio method.Otherwise, we employed five methods: inverse variance-weighted (IVW), MR-Egger, simple mode, weighted median, and weighted mode.A reliable causal relationship was defined as follows: estimates from the IVW method or the Wald ratio method were statistically significant (p < 0.05) and were consistent with the direction of causal estimates computed by the MR-Egger method (Wang et al.).
To explore the mediating effect of sphingolipid metabolites regarding the effect of LLDs on lung cancer risk, a two-step MR analysis was performed.The proportion mediated by sphingolipid metabolites was calculated using the following formula (Chen et al., 2022;Zhao et al., 2022): , where β1 represents the estimated effect of LLDs on sphingolipid metabolites, β2 represents the estimated effect of sphingolipid metabolites on lung cancer, and β3 represents the estimated direct effect of LLDs on lung cancer.The product of β1 and β2 is considered the indirect effect, while (β3 + β1 × β2) represents the total effect.The proportion is calculated when β1, β2, and β3 exhibit statistical significance, as this quantifies the proportion of mediation by sphingolipid metabolites within the total effect (Chen et al., 2022).To account for multiple statistical tests and reduce false-positive results, we controlled the false discovery rate (FDR) using the Benjamini-Hochberg method.Adjusted p < 0.05 was considered statistically significant (Liu et al., 2023).
To assess the sensitivity of the MR results, we conducted MR-Egger intercept tests and calculated Cochran's Q statistic to evaluate pleiotropy and heterogeneity.A significance level of p < 0.05 was applied to determine statistical significance.All MR analyses were performed using the "TwoSampleMR" package (version 0.5.6) in R software (version 4.1.1),and forest plots were generated using the "forestploter" package.

Positive control analysis
We investigated the causal associations between the IVs for inhibiting the LLD targets and CHD.Notably, the effect of ANGPTL3 inhibition, whether via ApoB or TG reduction, did not show statistically significant results in the positive control analysis.Consequently, ANGPTL3 inhibition was excluded from the subsequent MR analysis.The inhibition of the remaining six LLD targets exhibited a significant causal association, with a reduced risk of CHD. Figure 1 and Supplementary Table S4 show detailed results, illustrating the magnitude and significance of these associations.S5-S10 show the results of our MR analysis, estimating the effects of LLDs on the risk of lung cancer (overall) and in five subgroups (lung cancer in ever-smokers, lung cancer in never-smokers, LUSC, LUAD, and SCLC).

The effects of sphingolipid metabolites on lung cancer risk
Our MR analysis revealed that an increase in the plasma neutral ceramidase level by 1 SD decreased the risk of overall lung cancer (OR 0.95, 95% CI 0.91-0.99,p = 0.032), lung cancer in ever-smokers (OR 0.92, 95% CI 0.88-0.98,p = 0.024), LUAD (OR 0.93, 95% CI 0.88-0.99,p = 0.043), and LUSC (OR 0.93, 95% CI 0.86-0.99,p = 0.041) (Figure 5A; Supplementary Table S15).However, there was no evidence of a causal relationship between the plasma acid ceramidase level and risk of lung cancer (Figure 5B; Supplementary Table S16).The plasma ceramide levels exhibited a positive correlation with heightened lung cancer risk in  never-smokers; however, this relationship did not achieve statistical significance following correction (OR 1.49, 95% CI 1.10-2.01,p = 0.054) (Figure 5C; Supplementary Table S17).As the IVs of S1P did not intersect with the SNPs present in the GWAS data for lung cancer, the execution of this MR analysis was rendered unfeasible.The MR analysis exhibited neither pleiotropy nor heterogeneity.

Analysis of the mediating effects of sphingolipid metabolites
The aforementioned results suggested that neutral ceramidase plays a mediating role in the effect of LLDs on lung cancer risk.Table 3 shows the proportions mediated by neutral ceramidase regarding the negative effect of APOB inhibition on lung cancer risk in smokers via ApoB and TG reduction (8.1% and 9.5%, respectively) and the negative effect of APOC3 inhibition on LUAD risk via ApoB reduction (8.7%).Figure 6 illustrates schematic diagrams of the direct and indirect effects of LLDs on lung cancer and the mediation by neutral ceramidase.

Discussion
In this drug-target MR analysis, we assessed the effects of LLDs on lung cancer risk via six LLD targets (APOB, APOC3, HMGCR, LPL, NPC1L1, and PCSK9), comprising 12 lipid-lowering pathways.Our findings revealed that APOB inhibition decreased the lung cancer risk in ever-smokers and SCLC.APOC3 inhibition decreased the LUAD risk but increased the SCLC risk.Additionally, HMGCR inhibition increased the risk of SCLC, whereas the LPL agonist decreased the risk of SCLC.
We delved deeper into the mediating mechanisms involving sphingolipid metabolites within these processes.Our investigation revealed that inhibiting APOB, APOC3, and HMGCR increased plasma neutral ceramidase levels.Notably, we observed that neutral ceramidase played a protective role in various forms of lung cancer, including overall lung cancer, lung cancer in eversmokers, LUAD, and LUSC.These findings highlight the potential therapeutic significance of neutral ceramidase as a mitigating factor in the development and progression of various types of lung cancer.In addition, we found that APOB and PCSK9 inhibition led to a reduction in plasma ceramide levels, while the LPL agonist increased plasma ceramide and S1P levels.Nonetheless, no substantiated evidence has emerged to establish a causal link between ceramides and the risk of lung cancer.
Several MR studies have explored the potential preventive effects of LLDs on malignancies (Carter et al., 2020;Liu et al., 2021a;Min et al., 2023); however, these investigations have primarily focused on statins and have not included stratified analysis of different subtypes of lung cancer.In contrast to previous MR studies, our study expands the range of LLD targets under consideration.In our positive control analysis, we observed that ANGPTL3 inhibition did not achieve statistical significance.This finding aligns with the results reported by Wang et al. (2021).We made the decision not to conduct further analysis for ANGPTL3 inhibition in subsequent MR analysis.
In contrast to previous MR analysis of the effects of statins on lung cancer risk, our study reveals novel findings that HMGCR inhibition increases the risk of SCLC, as does APOC3 inhibition.This aligns with the findings reported by Vinogradova et al., where the link between statins and heightened lung cancer risk persisted even after adjusting for cardiovascular factors.However, their investigation lacked a subgroup analysis focusing on lung cancer types (Vinogradova et al., 2011).Moreover, evidence hints at the potential for long-term statin use to heighten cancer risk in women (Friedman et al., 2008).
The mechanistic underpinnings behind the potential of statins in increasing SCLC risk remain elusive.We posited that this effect might stem from off-target consequences associated with prolonged statin use (Liu et al., 2021b;Jiang et al., 2023).
Immunohistochemical assessments of lung cancer tissue samples have unveiled lower protein expression of APOC3 in SCLC compared to normal lung tissue.This discrepancy might propose a pathway through which APOC3 inhibitors heighten the vulnerability to SCLC (Shi et al., 2016).Furthermore, an investigation under the UK's Early Access to Medicines Scheme reported instances of lung cancer metastasis after the prolonged utilization of APOC3 inhibitors (volanesorsen) (Jones et al., 2023).Our findings strongly advocate for heightened vigilance concerning the potential for APOC3 inhibitors to increase lung cancer risk.
While alipogene tiparvovec acts as an LPL agonist, its limited market presence and high cost (Senior, 2017) have left the landscape of observational inquiries into its long-term cancer risk ambiguous.Nonetheless, our study propounds a shielding effect of LPL agonists against SCLC susceptibility.Moreover, Cerne et al. (2007) reported diminished LPL gene expression within lung cancer tissues as opposed to normal tissues, potentially reinforcing the notion of LPL agonists mitigating lung cancer risk.
The observed causal relationships between APOB inhibition and reduced lung cancer risk in ever-smokers and LUAD risk are particularly noteworthy, as neutral ceramidase plays a significant mediating role in these relationships.In recent years, there has been increasing research into the role and mechanisms of neutral ceramidase in malignancies (Coant and Hannun, 2019).Preclinical studies have demonstrated that inhibiting neutral ceramidase can prevent the occurrence and progression of colon cancer (Garcia-Barros et al., 2016), whereas inhibiting acid ceramidase can inhibit the proliferation of non-small-cell lung cancer and enhance its sensitivity to cisplatin (Yildiz-Ozer et al., 2018;White-Gilbertson et al., 2019).However, observational studies have presented contrasting findings, with high levels of acid ceramidase being associated with improved prognosis in breast and ovarian cancers (Ruckhaberle et al., 2009;Hanker et al., 2013;Sanger et al., 2015).The conflicting results from clinical observational studies and preclinical studies suggest that ceramidases may have pleiotropic effects on the occurrence, treatment, and prognosis of cancers.These findings highlight the complex and multifaceted nature of ceramidases in cancer-related processes.In our MR analysis, we identified a causal relationship between increased neutral ceramidase and a reduced risk of lung cancer in ever-smokers.This suggests that neutral ceramidase may play a critical role in the development of smoking-induced lung cancer, with APOB inhibition potentially acting as a protective factor.Further research is warranted to elucidate the precise mechanisms underlying their effects and to reconcile the disparities observed among different types of cancer.Ceramidases play essential regulatory roles in the onset and advancement of various cancer forms (Sanger et al., 2015;Garcia-Barros et al., 2016) and are pivotal in sphingolipid metabolism for converting ceramides into sphingosine, S1P, and fatty acids (Parveen et al., 2019).The sphingolipid metabolism pathway proves indispensable in the metastatic process of lung cancer (Pyne et al., 2018;Coant and Hannun, 2019); its potential as a novel therapeutic target holds significant promise for augmenting the efficacy of tumor treatments (Vijayan et al., 2019), and animal experiments have demonstrated that NPC1L1 inhibition can modulate sphingolipid metabolism (Yamanashi et al., 2020).Although observational studies have indicated an association between statin use and reduced plasma ceramide levels, there is limited information on the impact on ceramidase levels.
In alignment with the findings of a previous study, our study identifies increased plasma ceramide levels as a contributing factor to heightened lung cancer risk among non-smokers, albeit with nominal statistical significance (Alberg et al., 2013).Preclinical studies have shown that exposure to cigarette smoke leads to increased ceramide levels in the lung tissue of mice (Filosto et al., 2011;Goldkorn et al., 2014;Lavrynenko et al., 2020); the contribution of ceramides to the heightened risk of smokingrelated lung cancer necessitates prospective investigations for conclusive evidence.Consistent with previous observational studies, we found that PCSK9 inhibition reduced plasma ceramide levels (Tarasov et al., 2014), and APOB inhibition had the same effect.In addition, our results revealed that the inhibition of LPL could increase plasma S1P levels.However, owing to the nonconvergence of the IVs for S1P in the lung cancer GWAS dataset, an analysis of the association between S1P and lung cancer risk was precluded.Notably, Alberg et al. (2013) observed an increased risk of lung cancer with increased S1P levels.
The strength of our study resides in its augmentation of LLD targets and lipid-lowering pathways.It amplifies the scrutiny of distinct lung cancer subgroups.More importantly, we delve into the mediating effect of sphingolipid metabolites, thus elucidating the intricate mechanism underpinning how LLD influences lung cancer risk.However, it is important to acknowledge the limitations of our study.First, MR analyses could not be performed on two of the target genes of LLD: ANGPTL3 and PPARA.This limitation arises from the inability of IVs to establish positive control associations or the unavailability of suitable genetic variants.Furthermore, given the intricate pharmacological mechanisms underlying these drugs, our MR analysis was unable to assess their potential off-target effects.Second, IVs employed as proxies for lifetime exposure to LLD may be limited in their ability to capture the effects of short-term drug exposure.Additionally, our study is constrained to examining causal directions and does not enable a precise estimation of dosage effects or the cumulative impact of multiple medications.Third, although the F-statistic for the genetic variants exceeds 10, indicating a low probability of weak instrument bias, statistical power is constrained due to the limited availability of only two SNPs as IVs for acidic and neutral ceramidases (Larsson et al., 2019).Moreover, the relatively small sample size utilized for the assessment of sphingolipids raises concerns regarding potential selection bias.As a result, the interpretation of our findings should be exercised with caution.Fourth, despite sensitivity analyses not yielding statistically significant outcomes, it remains essential to acknowledge the potential impact of confounding variables and horizontal pleiotropy, given the possibility of SNPs residing in a state of weak linkage disequilibrium (r 2 < 0.30).Fifth, the absence of GWAS summary statistics for sphingolipids and lung cancer across diverse populations precludes validation.MR studies cannot substitute randomized controlled trials, and further clinical and pharmacoepidemiological studies are essential for triangulating.Sixth, the study population predominantly comprised individuals of European ancestry, which may limit the generalizability of our findings to other racial and ethnic groups.

Conclusion
We elucidated the intricate interplay between LLDs, sphingolipid metabolites, and lung cancer risk.Associations of APOB, APOC3, and HMGCR inhibition and the LPL agonist with distinct lung cancer risks underscore the multifaceted nature of these relationships.The observed mediation effects highlight the considerable influence of neutral ceramidase on lung cancer risk reduction achieved by APOB and APOC3 inhibition.

FIGURE 2
FIGURE 2Heatmap of the causal relationship between LLDs, lung cancer, and ceramidases.

FIGURE 6
FIGURE 6Overview of the direct and indirect effects of LLD on lung cancer risk, and the effects of neutral ceramidase.The effects of LLD on lung cancer risk involving mediating effects of neutral ceramidase are highlighted in red.LDL, low-density lipoprotein; ApoB, apolipoprotein B; TG, triglycerides.
Table 2 lists the demographic characteristics of detailed lung cancer GWAS summary statistics.

TABLE 1
Information on lipid-lowering drug targets included in the studies.

TABLE 2
Demographic characteristics of GWAS statistics for lung cancer.

TABLE 3
Proportions mediated by neutral ceramidase regarding the effects of LLD on lung cancer risk.