Associations of genetically proxied inhibition of HMGCR and PCSK9 with gastrointestinal cancer: a drug target Mendelian randomization study

Gastrointestinal cancer is a significant contributor to cancer-related mortality globally, with colorectal cancer (CRC) being the most prevalent and lethal type. Pancreatic cancer (PC) ranks second in mortality, with a 5-year relative survival rate of only 12% in the United States, followed by hepatocellular carcinoma (HCC), gastric cancer (GC), and esophageal cancer (EC)[1]. Existing studies indicate that lipid and cholesterol metabolism play a critical role in oncogenic signaling pathways and tumor cell development, particularly in gastrointestinal cancer[2]. Currently, the effects of lipid-lowering drugs on gastrointestinal cancers remain inconsistent and controversial, potentially influenced by the specific drug, dosage, and treatment duration[3]. Numerous observational studies have explored the association between lipid-lowering drugs and gastrointestinal cancer outcomes, but the results have been mixed. For instance, a retrospective study from Korea indicated that the combination of lipid-lowering drugs reduced the overall risk of gastrointestinal cancer, particularly among long-term users, suggesting a potential cancer preventive effect of 3-hydroxy-3-methylglutaryl-coenzyme A reductase (HMGCR) and proprotein convertase subtilisin/kexin type 9 (PCSK9) inhibitors[4]. However, no significant effect of statins was observed in CRC, and there was even a trend toward increased incidence of PC[5,6]. Mendelian randomization (MR) analysis of drug targets is a widely accepted method for inferring the effects of drugs on diseases[7]. This study attempts to employ MR analysis to elucidate the causal relationship between lipid-lowering drugs and gastrointestinal cancer. In addition, our study did not use artificial intelligence (AI), complying with the TITAN guidelines[8]. We confirm that no generative AI tools were used in manuscript preparation, data analysis, or figure generation. Methods To model the effects of lipid-lowering drugs, we employed two genetic tools: expression quantitative trait loci (eQTL) of drug target genes and genetic variants associated with low-density lipoprotein cholesterol (LDL-C) located within or near these genes. The study included two lipid-lowering drug targets (HMGCR and PCSK9) and five gastrointestinal cancers: EC, GC, HCC, PC, and CRC. Coronary heart disease (CHD) risk served as a positive control (Fig. 1A). The data related to LDL-C, CHD, EC, GC, HCC, PC, and CRC were sourced from the MRC-IEU genome-wide association studies (GWAS) database (https://gwas.mrcieu.ac.uk/). The eQTL data for HMGCR were obtained from the eQTLGen Consortium (https://www.eqtlgen.org/), while the PCSK9 data were sourced from the GTEx V8 Consortium (https://gtexportal.org/) (Supplemental Digital Content Table S1, available at: https://links.lww.com/JS9/F383). In the primary MR analysis, instrumental variables (IVs) were selected based on single nucleotide polymorphisms (SNPs) located within a ± 100 kb window of the HMGCR or PCSK9 regions, demonstrating significant genome-wide associations with LDL-C (p < 5.0 × 10 − 8). All IVs and exposures showed a strong correlation (F > 10) and were confirmed to be independent (r2 < 0.3) through linkage disequilibrium (LD) analysis[9,10]. Ultimately, 19 significant SNPs for HMGCR and 33 significant SNPs for PCSK9 were retained (Supplemental Digital Content Table S2, available at: https://links.lww.com/JS9/F383). Additionally, we performed two-sample MR using cis-eQTLs of HMGCR and PCSK9 as part of our secondary analysis. The number of common SNPs associated with the expression levels of these genes in blood was 25 for HMGCR and 16 for PCSK9, respectively (Supplemental Digital Content Table S3, available at: https://links.lww.com/JS9/F383). Given that only 16 cis-eQTLs exist for the common PCSK9 gene in blood, we did not remove LD. Figure 1.: (A) Study overview and design of drug target MR analysis; (B) IVW-MR association between expression of gene HMGCR or PCSK9 and CHD, GC, or HCC outcomes; (C) IVW-MR association between LDL-C mediated by gene HMGCR or PCSK9 and gastrointestinal cancers outcomes. GC, gastric cancer; HCC, hepatocellular carcinoma; CHD, coronary heart disease; HMGCR, 3-hydroxy-3-methylglutaryl-coenzyme A reductase; IVW, inverse variance weighted; LDL-C, low-density lipoprotein cholesterol; MR, Mendelian randomization; PCSK9, inhibitors and proprotein convertase subtilisin/kexin type 9. We employed inverse variance weighted (IVW), weighted median, and MR-Egger for analysis, with the IVW method being the most commonly used due to its ability to provide reliable causal estimates even in the presence of heterogeneity[11]. Heterogeneity was examined and assessed by MR-Egger and IVW. Cochran’s Q-value was utilized to evaluate heterogeneity among genetic instruments, with P > 0.05 indicating no evidence of heterogeneity[12]. In cases of heterogeneity, the IVW random effects model was employed to evaluate the effect size of the MR. We performed the MR-Egger regression equation and MR-PRESSO analysis to assess the potential horizontal pleiotropy of the SNPs used as IVs to ensure the robustness of the findings, with P > 0.05 indicating no horizontal pleiotropy[13]. If the MR-PRESSO analysis identified outliers, the MR analysis was repeated after their removal. We also performed a co-localization analysis to identify evidence of shared causal variants between exposure and outcome, where PPH4/(PPH3 + PPH4) > 0.8 was considered significant[14]. In the summary-data-based MR (SMR) analysis, we employed available eQTLs for the drug target genes as proxies for exposure to each lipid-lowering drug. Results As expected, both the HMGCR drug target [OR (95%CI) = 0.618 (0.519, 0.737), P = 8.29E-08] and the PCSK9 drug target [OR (95%CI) = 0.435 (0.370, 0.512), P = 7.20E-24] significantly reduced the risk of CHD in the IVW-MR analysis (Supplemental Digital Content Table S4, available at: https://links.lww.com/JS9/F383). Based on the IVW analysis results, gene proxy inhibition of HMGCR was strongly associated with a higher risk of GC [OR (95%CI) = 2.373 (1.583, 3.560), P = 2.89E-05], validated by co-localization [PPH4/(PPH3 + PPH4) = 0.857]. In addition, genetically predicted PCSK9 inhibition had a positive effect on HCC risk [OR (95%CI) = 2.147 (1.048, 4.399), P = 0.037] with co-localization support [PPH4/(PPH3 + PPH4) = 0.952] (Supplemental Digital Content Table S5, available at: https://links.lww.com/JS9/F383). Secondary analysis also indicated that higher levels of circulating HMGCR may have the potential to increase the risk of GC [OR (95%CI) = 1.222 (1.141, 1.309), P = 1.01E-08], which is consistent with the results of our preliminary analysis. In addition, we observed suggestive evidence regarding an inverse association between PCSK9 expression and the risk of susceptibility to HCC [OR (95%CI) = 0.699 (0.620, 0.789), P = 6.47E-09]. The conclusion drawn here is contrary to the results of our preliminary analysis, which may be due to the presence of LD. Therefore, we again performed MR analysis between the positive control CHD and the cis-eQTLs of HMGCR [OR (95%CI) = 0.928 (0.898, 0.960), P = 1.68E-05] and PCSK9 [OR (95%CI) = 1.182 (1.145, 1.219), P = 6.61E-26], and finally found the same phenomenon (Fig. 1B). In our subsequent SMR analysis, no significant correlation was found between HMGCR and PCSK9 expression and disease outcomes (Supplemental Digital Content Table S6, available at: https://links.lww.com/JS9/F383). There is no evidence of any noteworthy association between HMGCR inhibitors or PCSK9 inhibitors and EC, PC, and CRC (Fig. 1C). Homogeneity was supported with Cochran’s Q-value (p > 0.05), negating heterogeneity concerns. MR-Egger regression equation further denoted an absence of horizontal pleiotropy (P > 0.05). MR-PRESSO analysis detected no influential outliers amid IVs. Leave-one-out sensitivity analyses showed that the association between HMGCR inhibition or PCSK9 inhibition and CHD and gastrointestinal cancer was not substantially driven by any individual SNP (Fig. 2). Figure 2.: Sensitivity analysis of HMGCR or PCSK9 on CHD and gastrointestinal cancers. Leave-one-out analysis of HMGCR on (A) CHD, (B) CRC, (C) EC, (D) GC, (E) HCC, (F) PC. Leave-one-out analysis of PCSK9 on (G) CHD, (H) CRC, (I) EC, (J) GC, (K) HCC, (L) PC. CHD, coronary heart disease; CRC, colorectal cancer; EC, esophageal cancer; GC, gastric cancer; HCC, hepatocellular carcinoma; HMGCR, 3-hydroxy-3-methylglutaryl-coenzyme A reductase; PC, pancreatic cancer. PCSK9, inhibitors and proprotein convertase subtilisin/kexin type 9. Discussion Gastrointestinal cancer is characterized by a prolonged onset, an insidious early stage, and complex pathogenesis, presenting significant challenges for developing effective therapeutic drugs. While cancer treatment options are continually improving, developing new drugs to enhance prognosis remains challenging, making breakthrough progress increasingly difficult. Repurposing existing drugs is more economical and time-efficient, particularly since some lipid-lowering drugs have been associated with cancer risk[15]. As one of the most commonly prescribed drug classes, HMGCR inhibitors have garnered significant attention due to their pleiotropic effects, primarily regulating tumor cell development, survival, migration, invasion, metastasis, and apoptosis through downstream signaling pathways mediated by the prenylation of the reticular activating system[16]. Several traditional observational studies over the past few decades have produced inconsistent conclusions regarding the effect of HMGCR-targeting drugs on GC risk. A population-based case–control study found that the use of any statin was significantly associated with a reduced risk of GC compared to healthy controls[17]. Conversely, an analysis of a multicenter cohort study involving 8,798 Korean patients with newly diagnosed GC showed no survival benefit associated with long-term statin use (>545 days) compared to prior short-term use[18]. Some studies ever suggest that low LDL-C may either increase the risk of GC or show no association[19]. This finding aligns with our study, which indicates that genetically proxied HMGCR inhibition increases the risk of GC. Additionally, co-localization analysis demonstrated that HMGCR and GC share the same genetic region. A secondary analysis of cis-eQTLs for HMGCR also revealed significantly improved prognosis in GC patients with elevated HMGCR expression. These results suggest that HMGCR inhibitors should be avoided in the clinical treatment of cardiovascular disease patients with GC due to their potential role in promoting GC. Although this appears to contradict conventional research conclusions, statin use does not equate to HMGCR inhibition, necessitating further investigations to elucidate the precise mechanism by which HMGCR regulates GC. Furthermore, our drug target MR analysis found that genetically proxied PCSK9 inhibition significantly increased the risk of HCC. As another lipid-lowering drug target, PCSK9 not only increases plasma LDL-C concentration by enhancing lysosomal degradation of LDL receptors in hepatocytes, but also participates in regulating cell apoptosis and inflammatory responses, and even affects the antitumor efficacy of anti-PD-1/PD-L1 immunotherapy to some extent[20,21]. Although the protective effect of PCSK9 inhibitors in atherosclerosis is well established, their effect on HCC remains uncertain. Immunohistochemical analysis revealed that PCSK9 expression levels in the serum of HCC patients were lower than those in adjacent liver cirrhosis tissues[22]. Interestingly, PCSK9 knockout mice exhibited increased susceptibility to HCC, with evidence suggesting that PCSK9 inhibits HCC growth by interacting with GSTP1 and suppressing the JNK signaling pathway[23,24]. The aforementioned conclusions, together with our co-localization analysis results, further validate the association between PCSK9 and an increased risk of HCC. This indicates that caution should be exercised when using PCSK9 inhibitors in treating cardiovascular disease patients, as they may induce the development of HCC. However, a retrospective study involving 105 patients who underwent radical resection for HCC found that those with high PCSK9 expression had poorer postoperative overall survival and disease-free survival[25]. The reason for the inconsistent results of the above studies can be attributed to the fact that the incidence of HCC is the result of multi-gene interaction, and observational studies do not allow causal inference and cannot overcome confounding factors. As a genetic epidemiological method, MR studies offer distinct advantages, including the ability to avoid reverse causality and eliminate confounding bias, thereby supplementing the conclusions of observational studies and addressing the limitations of short follow-up periods. However, it is important to acknowledge several limitations of this study. First, the number of effective eQTLs for PCSK9 in blood tissues was insufficient to allow for LD removal during statistical analysis, potentially affecting the statistical power of the results regarding PCSK9 inhibition. Second, the inclusion of participants of both European and East Asian ancestry may introduce potential bias and spurious associations due to population stratification. Third, the limited power of the GWAS data used for GC and HCC weakens the evidence for shared causal variants identified in this study’s co-localization analysis, suggesting that future discussions should be revisited once more robust GWAS results become available. Finally, drug target MR studies can only reflect long-term exposure, and the causal relationship between short-term drug use and disease prognosis remains unclear. Conclusions To sum up, we provide genetic evidence that HMGCR inhibition increases GC risk and PCSK9 inhibition elevates HCC risk. Crucially, our findings caution against uncritical prescription of these lipid-lowering agents in high-risk populations, particularly patients with pre-existing gastric or hepatic pathologies. Further randomized controlled trials are essential to validate these findings and to explore the underlying mechanisms.