Drug repositioning and ovarian cancer, a study based on Mendelian randomisation analysis

Background The role of drug repositioning in the treatment of ovarian cancer has received increasing attention. Although promising results have been achieved, there are also major controversies. Methods In this study, we conducted a drug-target Mendelian randomisation (MR) analysis to systematically investigate the reported effects and relevance of traditional drugs in the treatment of ovarian cancer. The inverse-variance weighted (IVW) method was used in the main analysis to estimate the causal effect. Several MR methods were used simultaneously to test the robustness of the results. Results By screening 31 drugs with 110 targets, FNTA, HSPA5, NEU1, CCND1, CASP1, CASP3 were negatively correlated with ovarian cancer, and HMGCR, PLA2G4A, ITGAL, PTGS1, FNTB were positively correlated with ovarian cancer. Conclusion Statins (HMGCR blockers), lonafarnib (farnesyltransferase inhibitors), the anti-inflammatory drug aspirin, and the anti-malarial drug adiponectin all have potential therapeutic roles in ovarian cancer treatment.


Introduction
Ovarian cancer is the second most common gynaecological cancer and the gynaecological malignancy with the highest mortality rate worldwide (1).Due to the concealed location of the disease, most cases are found at an advanced stage.Although the effect of surgery or chemotherapy is better in early stage patients, they are prone to relapse and drug resistance.Therefore, there is a need to develop new therapeutic agents (2).But drug discovery and development is an expensive, time-consuming, and risky business.In this context, drug repositioning strategies are becoming increasingly attractive.Drug repositioning aims to identify new indications for old drugs, with the advantage that data on pharmacokinetic properties and toxicity are already available.Thus, in principle, it is possible to reduce research costs and accelerate drug use/distribution (3).
Mendelian randomization studies are an etiological research method based on Mendel's law of independent assignment, using genetic factors as instrumental variables for risk factors to make causal inferences about disease or health-related outcomes.Its advantage is that it can effectively avoid the influence of various confounding factors (20).In this study, we used the Mendelian randomization study method to analyse the previously reported therapeutic effects of conventional drugs on ovarian cancer.

Materials and methods
All analyses used publicly available database resources and therefore did not require ethical approval from an institutional review board.The entire MR analysis process illustrated in Figure 1.

GWAS data source of ovarian cancer
Genetic association data for ovarian cancer (ieu-a-1120) were obtained from the Ovarian Cancer Association Consortium (OCAC), which performed a GWAS analysis of 22,406 cases of invasive epithelial ovarian cancer pathology and 40,941 healthy population controls, and through genetic analysis identified 17273705 SNP loci (21) (Table 1).
We additionally selected a dataset on ovarian cancer (ieu-b-4963) from the UKB database, which contains 1,218 ovarian cancer cases and 198,523 controls from healthy populations, and identified 9,822,229 SNP loci by genetic analysis.Both data sets were from European populations.(Table 1) (https://gwas.mrcieu.ac.uk).

Selection of drugs and targets
All publicly available literatures (published in English) were retrieved from China National Knowledge Infrastructure (CNKI) and Pubmed database from inception to December 2023 with the following keywords as search terms: "Drug repositioning" and "ovarian cancer", and drugbank database was used to search the targets of related drugs.Drugs with the same mechanism of action and target were grouped into one group, and drugs with unclear targets were excluded.

Selection of drug targeting sites and exposure factors
We used relevant drug targets as proxies for each drug exposure at 100 kb loci downstream of each of the drug target gene regions above and below the drug target gene region.p<5×10 -8 , r 2 <0.3 were obtained to screen for SNP loci closely related to the drug target as exposure factors, and targets for which no SNP loci were retrieved were excluded.To reduce the weak instrumental variable bias, the F-statistic was calculated separately for each SNP, and then weak instruments with an F-statistic < 10 were filtered.

MR analysis
In univariate MR analysis, IVW is often used as the primary statistical method for MR analysis-.When horizontal pleiotropy is not present, inverse variance weighted (IVW) can provide relatively stable and accurate causality by using meta-analytical methods in combination with estimates of the causal effects of each SNP.The results of IVW were supplemented using the MR Egger, (weighted median, WME) method.Heterogeneity was estimated using Cochran's Q statistic and p < 0.05 was used as the threshold parameter for identifying heterogeneity.Directional pleiotropy was detected using and MR-Egger test and outliers, and p < 0.05 was used as the threshold parameter for identifying horizontal pleiotropy.Leave-one-out sensitivity analyses were used to evaluate the magnitude of the effect of SNPs on the estimation results of causal associations.

Result
After removing the drugs with completely consistent targets and those without clear targets, 31 drugs and 110 targets were obtained.
(see Supplementary Table 1) The causal associations between different kinds of drug targets and ovarian cancer were analysed by MR, and there were 77 instrumental variable SNPs with Fstatistics >10, suggesting that there was a low possibility of weak instrumental variable bias.The information of the 77 SNPs is shown in Table 2.

Results of sensitivity analyses
The results of MR-PRESSO analysis showed that there were no outliers for all SNPs included in this study.Heterogeneity test was conducted using the IVW method and MR-Egger regression, when the test P < 0.05, heterogeneity exists among SNPs; when the P value > 0.05, heterogeneity does not exist among SNPs (Supplementary Figure 1).MR-Egger regression was used to test for polytropy, when the test P < 0.05, it means that there is polytropy; when the P value > 0.05, it means that there is no polytropy.(Supplementary Figure 2) Leave-one-out analysis was used to remove one SNP at a time, and the remaining SNPs were subjected to IVW analysis.(Supplementary Figure 3, Supplementary Table 3).

Discussion
Currently, the role of drug repositioning in the treatment of ovarian cancer is gradually attracting attention, but there is a lack of large-scale, multicentre, prospective studies, and the retrospective studies are affected by a variety of confounding factors, leading to huge discrepancies in the results of current studies.Drug-targeted Mendelian randomisation studies are based on the downstream products of target proteins (biomarker), SNPs (pQTL or eQTL) with significant effects on biomarker near the gene encoding the target protein are used as instrumental variables, and Mendelian randomisation analyses are performed with disease as the endpoint to validate the effects of the target proteins on the disease under study.
The main target of action of statins is HMGCR, but they can also act on other targets, such as HDAC2, DPP4, ITGAL and so on.Among these, ITGAL encodes integrin a, which is the target of fatsoluble statins.Previously, it was thought that only fat-soluble statins have an anti-ovarian cancer effect and can prolong the generation time of patients (23).It has been suggested that this may be due to the fact that lipid-soluble drugs can cross the cell membrane more easily and thus exert anti-tumour effects (24).The results of the present study suggest that lipid-soluble statins may exert anti-tumour effects not only by interfering with HMGCR, but also, possibly, by inhibiting ITGAL.
Summary of the univariable MR analysis results for the association between drug-related targets and the risk of Ovarian cancer (ieu-a-1120) using the IVW, MR Egger, and Weighted median methods.
As statins are key enzymes in the MVA pathway, it has long been questioned whether similar anti-tumour effects could be achieved by inhibiting other enzymes in the pathway (25), or whether better therapeutic effects could be achieved by combining drugs (26).It has also been suggested that the anti-tumour effects of statins may not be related to lipid lowering but to their intermediate metabolites such as FPP and GGPP, which provide isoprenyl groups to isoprenylate a variety of small GTPase-binding proteins such as Ras, MEK, PI3K, and anchor them to the cell membrane.This is how it works, and some studies suggest that zoledronic acid and lonafarnib may reduce the production of FPP and GGPP, thereby exerting an anti-ovarian cancer effect.However, it has also been suggested that the anti-tumour effect of statins can only be reversed by the addition of GGPP (24).In this study, we also found that the main targets of zoledronic acid, GGPS1 and FDPS, were not associated with the development of ovarian cancer, but the targets of action of lonafarnib, the two subunits of farnesyltransferase, FNTA and FNTB, were both associated with the development of ovarian cancer, which seems to support this view.However, it is difficult to explain why FNTA is negatively associated with ovarian cancer prognosis, while FNTB is positively associated with ovarian carcinogenesis.
The mechanism by which aspirin improves the prognosis of ovarian cancer is usually thought to be related to COX inhibition, with low-dose aspirin primarily inhibiting COX1 and higher doses inhibiting COX2 activity.In the past, the literature has suggested that high doses of aspirin reduced mortality in patients of ovarian cancer (19).More recent literature suggests that the source of the statistics may amplify this dose-dependent difference (27).Our study shows that PTGS1 (COX1) is significantly positively correlated with ovarian cancer prognosis, while PTGS2 is not significantly correlated with ovarian cancer prognosis.In addition aspirin may inhibit several targets such as HSPA5, NEU1, CCND1, CASP1, CASP3, which are negatively correlated with ovarian cancer prognosis.Therefore, whether simply acting on PTGS1 targets can obtain better anti-ovarian cancer effects will be a question we need to verify in the future.
Originally used in antimalarial therapy, quinacrine has now been shown to have anti-tumour effects (28,29), particularly in inducing apoptosis in P53-deficient or mutated tumour cells (30).PLA2G4A encodes phospholipase A2, and quinacrine may inhibit ovarian cancer by inhibiting PLA2G4A which in turn reduces arachidonic acid metabolites and thus inhibits ovarian cancer development.

Limitation of the study
Currently, drug target Mendelian randomisation analysis usually selects single-target drugs, such as PCSK9 inhibitors, while statins are considered to be single-target drugs for HMGCR (20), but statins are not actually single-target drugs, and their targets of action are not limited to HMGCR.there are also drugs for which it is difficult to specify which target is the main target of their action.Therefore, in this study, all known targets were screened for target selection.However, whether this affects the results is also a problem that needs to be addressed in the current Mendelian randomization analysis of drug targeting.
In addition, when selecting genetic association data for ovarian cancer, we selected two datasets, but found that the conclusions drawn from the two databases were not consistent.The number of ovarian cancer patients and SNP loci found in ieu-a-1120 data far exceeds that in ieu-b-4963.It suggests that the results of drug-targeted Mendelian randomisation analysis are affected by multiple factors (sequencing technology, sequencing depth and number of cases).
Third, the genetic information related to ovarian cancer is mainly from European populations, but the results of drug repositioning may be affected by regional and racial differences as reported in the relevant literature.Further studies are needed to clarify the differences between different races.
Therefore, the results of this study are only for reference, and we hope to provide ideas for the selection of therapeutic targets for ovarian cancer in the future.

Conclusion
This study provides MR evidence for the potential value of conventional drugs in preventing ovarian cancer.Genetic links between drug targets and ovarian cancer may accelerate drug development prioritisation and reduce the failure rate of clinical trials and provide new ideas for future drug development.

FIGURE 3
FIGURE 3Summary of the univariable MR analysis results for the association between drug-related targets and the risk of Ovarian cancer (ieu-b-4963) using the IVW, MR Egger, and Weighted median methods.

TABLE 1 Ovarian
Cancer GWAS Data Sources and Descriptions from Open GWAS.

TABLE 2
Basic information on 77 relevant Instrumental variables.