Causal Relationships between Homocysteine and the Polycystic Ovary Syndrome: A Mendelian Randomization Analysis

Background Polycystic ovary syndrome (PCOS) is an endocrine disease attributed to multiple genetic variants and environmental factors. We aimed to find the causal association of homocysteine (Hcy) with PCOS. Methods A two-sample Mendelian randomization (MR) analysis was performed. We selected 14 single-nucleotide polymorphisms (SNPs) as instrumental variables to predict the risk of PCOS from genome-wide association studies (GWAS). The summary statistics of PCOS were obtained from 3 large genome-wide association studies in the European population, involving 4,138 cases and 20,129 controls, 3,609 cases and 229,788 controls, 994 cases and 165,817 controls, separately. Results The IVM analyses revealed that plasma Hcy levels were not causally associated with the risk of PCOS in the meta-analysis (combined effect = 1.032, 95% confidence interval (CI): 0.885–1.203, p=0.688). Conclusions There was no sufficient evidence to support the causal association of the Hcy with the risk of PCOS.


Introduction
Polycystic ovary syndrome (PCOS) is a commonly encountered endocrinopathy in women of reproductive age, arising from the interaction between multiple genetic variants and environmental factors [1].At present, there are three distinct diagnostic criteria for PCOS, namely, the National Institutes of Health (NIH) criteria, the Rotterdam criteria, and the Androgen Excess and PCOS (AE-PCOS) Society criteria, each with a prevalence of 6%, 10%, and 10% for PCOS, respectively [2][3][4].Te Rotterdam diagnostic criteria, predominantly adopted by experts worldwide, requires the presence of two out of three traits: [1] oligo-and/ or anovulation, [2] hyperandrogenism, and [3] polycystic ovaries [3,5].PCOS is associated with an elevated risk of developing obesity, metabolic syndrome, type 2 diabetes, and cardiovascular disease [6].
To gain deeper insights into the nature of the observed associations, Mendelian randomization (MR) is recommended for advanced study, utilizing genetic variations as instrumental variables (IVs) for exposures.MR is an evolving methodological approach that employs IVs, including single-nucleotide polymorphisms (SNPs) identifed through genome-wide association studies (GWAS), to delineate causal relationships between exposures and disease outcomes.It is less susceptible to the aforementioned shortcomings, given that genetic variations are randomly allocated and remain steady throughout life, owing to Mendel's frst and second laws of heritability [28,29].However, MR studies rely on the following assumptions: [1] selected IVs must be reliably associated with the exposure factor; [2] selected IVs can only afect the outcome through the exposure; and [3] selected IVs cannot be associated with confounding factors.In recent years, MR methods have been widely applied to elucidate the etiology and consequences of PCOS using GWAS data [30], but no relevant studies for Hcy and PCOS have been undertaken so far.In the current study, a two-sample MR was performed to evaluate the causal role of hyperhomocysteinemia on PCOS using SNPs as IVs, Hcy as the exposure, and PCOS as the outcome in European populations.A brief overview of the MR study design is illustrated in Figure 1.

IVs Selection.
We obtained instrumental variables (IVs) from a genome-wide association study (GWAS) comprising 10 European population cohorts with a total of 44,147 white individuals [31] and 18 SNPs from 13 loci were found to be associated with plasma Hcy levels at a genome-wide signifcance (p < 5 × 10 −8 ).Using the European population of 1000 Genomes as a reference panel, linkage disequilibrium was estimated across SNPs (r 2 < 0.05), and three SNPs (rs12134663, rs957140, and rs12921383) were removed from our study to make the remaining SNPs independent.To ensure that the specifed SNPs only infuenced the outcome through the exposure, we used the PhenoScanner V2 website (https://www.phenoscanner.medschl.cam.ac.uk) to kick out the SNP associated with any potential confounders of PCOS and rs548987 was removed because of pleiotropic efect on BMI.Tree [32] SNPs (rs7422339, rs234709, and rs2851391) were not available in the FinnGen datasets and did not get a replacement with proxy SNPs (r 2 > 0.8).Lastly, a total of 14 (11 in FinnGen) SNPs were selected (Table 1), explaining 5.9% of the variation in the Hcy plasma levels [31], and these SNPs were selected by the previous studies to predict the serum level of Hcy [32][33][34].In advance, the F statistic was calculated to detect the strength of the IVs, and F < 10 meant a weak IV scenario [35].

Study
Outcomes.Genetic associations with PCOS were gathered from 3 large genome-wide association studies (GWAS) in the European population.Te study conducted by Day et al. included 4,138 cases of PCOS based on NIH or Rotterdam criteria and 20,129 controls collected from six cohorts (self-report-based data from 23andMe was excluded) [36].Using the ICD code of PCOS (ICD-10 code E28.2, ICD-9 code 256.4,or ICD-8 code 256.90), the study performed by Tyrmi et al. comprised 3,609 cases and 229,788 controls from Finland and Estonia [37], whilst the FinnGen study release 7 (R7) consisted of 994 cases and 165,817 controls (https://fnngen.gitbook.io/documentation/v/r7/)[38].A description of all data sources is summarized in Table 1.Written consent was provided by participants, and the relevant ethical review boards approved all the studies involved in the MR analysis.Considering that the data used herein were previously published, ethics committee approval was waived.

Statistical Analyses.
Te inverse-variance-weighted (IVW) method was applied to evaluate the association between plasma Hcy levels and the risk of PCOS.A p value of <0.05 was considered a statistically signifcant diference [39].For sensitive analysis, the weighted median estimator, which can yield consistent results even when half of the instrumental variables are invalid, was used [40].Te MR-Egger regression method and the MR-PRESSO method were also used to evaluate the directional pleiotropy by calculating the intercept of the association between Hcy and PCOS [41].Te MR-Steiger fltering was used to estimate potential reverse causal impact.
Leave-one-out sensitivity analysis was performed for the heterogeneity test to explore the infuence of specifc SNPs on the association.All statistical analyses were conducted using the R Studio (version 4.0.2) and the package "MendelianRandomization."

Causal Associations with PCOS.
Te F statistics of all the selected SNPs exceeded 10 (Table 1), indicating they were not weak instruments.
As depicted in Figure 2, the IVM analyses revealed that plasma Hcy level was not causally associated with the risk of PCOS in the meta-analysis (combined efect � 1.032, 95% confdence interval (CI): 0.885-1.203,p � 0.688).As anticipated, consistent results were observed in the FinnGen

Discussion
To the best of our knowledge, this is the frst study based on MR methods involving adequate sample sizes to examine the correlation between plasma Hcy levels and PCOS, and our result did not identify plasma Hcy levels as a risk factor for PCOS.Te selection of homocysteine (Hcy) as a biomarker for assessing the risk of polycystic ovary syndrome (PCOS) as an outcome disease warrants careful consideration of potential confounders.While our study aimed to elucidate the relationship between Hcy levels and PCOS risk, it is imperative to acknowledge the multifactorial nature of PCOS etiology.Factors such as age, body mass index (BMI), hormonal imbalances, insulin resistance, and lifestyle variables can signifcantly infuence both Hcy levels and PCOS development [42].For instance, higher BMI has been associated with elevated Hcy levels and increased PCOS risk [43].Furthermore, hormonal imbalances, particularly elevated androgen levels and insulin resistance, are common features of PCOS and may independently impact Hcy metabolism.In addition, dietary factors rich in folate, vitamin B6, and vitamin B12 have been linked to lower Hcy levels and may potentially mitigate the risk of developing PCOS [44].Terefore, future studies investigating the association between Hcy and PCOS should comprehensively account for these confounding variables to accurately assess the true relationship between Hcy levels and PCOS risk.Accumulating evidence uncovered these years suggests that higher Hcy levels may play an instrumental role in the etiology of PCOS.A recent meta-analysis, enrolling 1718 PCOS cases and 1399 controls, demonstrated that PCOS patients had statistically signifcantly higher Hcy levels after individually adjusting for obesity, insulin resistance, and testosterone level [13].However, heterogeneity was significant in this meta-analysis, and subgroup analyses for these confounders were not performed [13].Another metaanalysis performed by Murri et al., including 2090 cases and 1421 women without PCOS, found that Hcy levels were 23% higher in PCOS women than in healthy controls [14].Interestingly, the observed diference between the Hcy levels and PCOS in this study was not signifcant.Tis inconsistency may be ascribed to several factors as follows: (1) both studies exhibited considerable heterogeneity, thereby compromising the reliability of their results; (2) publication bias is also an indispensable factor.Publishers are prone to accepting studies with statistically signifcant results.Negative results regarding the association between Hcy levels and PCOS were identifed in limited articles in PubMed and other databases; (3) a large number of past studies have been restrained by small sample sizes, with the biggest metaanalysis merely involving 2090 cases, which is signifcantly lower than our study population; and (4) Hcy levels are correlated with many confounding variables, encompassing age, BMI, smoking status, concomitant subclinical infammatory diseases, and insulin resistance [11,18,45], which may have infuenced the results of the studies.Obesity, a prevalent feature among women with PCOS, has 4 International Journal of Endocrinology been reported to increase Hcy levels through various pathways, including impaired methylation and increased oxidative stress [46].Moreover, insulin resistance, a hallmark of PCOS, may contribute to elevated Hcy levels by altering the activity of enzymes involved in Hcy metabolism.
Earlier studies have provided evidence that obesity is a risk factor for PCOS [36].However, Hcy levels are typically elevated in obese patients [11], indicating that a high-fat diet might aggravate hepatic enzyme activity involved in homocysteine metabolism [47].While some confounding variables were adjusted in observational studies, they may not fully account for reverse causation.Elevated Hcy levels may stem from various factors associated with obesity, such as dietary patterns and metabolic dysregulation.Indeed, the consumption of high-fat diets among obese individuals exacerbates the activity of hepatic enzymes involved in Hcy metabolism, resulting in elevated Hcy levels [48].However, it is critical to recognize that altered liver enzyme activity is a direct consequence of obesity.Our study explored their potential impact on diseases such as polycystic ovary syndrome (PCOS), thereby elucidating the complex interplay between metabolic disorders and Hcy metabolism.Te robust evidence generated by the MR methodological approach strongly validates this fnding.International Journal of Endocrinology Nevertheless, some limitations in our study should not be overlooked.First, PCOS patients were not stratifed by the abovementioned diagnostic criteria, given that Tyrmi et al. and the FinnGen (R7) study [37,38] exclusively provided information based on ICD codes without detailed diagnostic information.Besides, the study of Tyrmi et al. contained data from FinnGen data freeze release 6 (R6) and Estonian Biobank, leading to a partial overlap of the outcome data with the FinnGen (R7) study.Second, our study only enrolled individuals of European ancestry to mitigate population structure bias; consequently, our conclusions may not be generalizable to a global population.Finally, PhenoScanner was used to identify potential pleiotropy of the IVs, but the possibility of missing potential pleiotropic efects cannot be excluded, considering that various phenotypes associated with genetic variants are still being explored.

Conclusion
Our two-sample MR analysis found that plasma Hcy levels were not causally associated with PCOS, which will ultimately help improve prevention and management strategies for this common gynecological condition.

Figure 1 :
Figure 1: Study design of the Mendelian randomization analysis between homocysteine and the risk of polycystic ovary syndrome.PCOS: polycystic ovary syndrome and Hcy: homocysteine.Te "×" means our study design avoids such a situation.

Figure 2 :
Figure 2: Associations of homocysteine with polycystic ovary syndrome.CI: confdence interval and OR: odds ratio.

Figure 3 :
Figure 3: Forest plots and scatter plots for Hcy and PCOS in FinnGen.

Figure 4 :
Figure 4: Forest plots and scatter plots for Hcy and PCOS in Tyrmi et al.

Figure 5 :
Figure 5: Forest plots and scatter plots for Hcy and PCOS in Day et al.

Table 1 :
Characteristic of the instrumental variables (IVs) associated with plasma homocysteine (Hcy) and the associations with polycystic ovarian syndrome (PCOS).

Table 2 :
Weighted median and MR-Egger analysis for genetic associations between exposures and PCOS risk.