The mitochondrial DNA copy number and ovary-related reproductive disorders: A bidirectional two-sample Mendelian randomization study

Ke Peng: Designed and performed the study. Ke Peng and Xiao Han: Wrote the first draft of the manuscript. Congquan Wu and Qiaowen Lu: Provided critical input and reading. Shujun Gao: Supervised and edited the paper.

This work has no funding.

Three sets of independent variables were examined to explore the potential links between mtDNA‐CN levels and the risk of ORRDs (Figure  2 ). After quality control, the numbers of SNPs in “IVs‐1,” “IVs‐2” and “IVs‐3” were 55, 47, and 78, respectively (Table  S2 ). F ‐statistics analysis revealed that all variances exceeded 10, suggesting the absence of potential weak instrument bias (Table  S2 ). Table  S1 presents the results of the five MR methods, horizontal pleiotropy, and heterogeneity analyses. The IVW method was utilized as the primary outcome, and the results of the IVW method are presented in Table  2 . Figure  S1 presents the results of the leave‐one‐out sensitivity and single SNP analyses and three MR methods with “IVs‐1,” “IVs‐2” and “IVs‐3.” Primary analysis of the association of mtDNA‐CN with risk of ovarian diseases. * P  < 0.05, ** P  < 0.01, *** P  < 0.001. IV, instrumental variables; PCOS, polycystic ovary syndrome; POF, premature ovarian failure. Mendelian randomization results for causality for mtDNA‐CN causing ORRDs. Abbreviations: CI, confidence interval; DSL, DerSimonian‐Laird; IVs, instrumental variables; IVW, inverse variance weighted; mtDNA‐CN, mitochondrial DNA copy number; nSNP, number of single nucleotide polymorphisms; OR, odds ratio; ORRDs, ovary‐related reproductive disorders; PCOS, polycystic ovary syndrome; POF, premature ovarian failure. For “IVs‐1,” the MR analysis showed no causal association between mtDNA‐CN and the risk of ovarian dysfunction, cyst, and POF. Contrarily, mtDNA‐CN was associated with PCOS (OR = 1.16; 95% CI: 1.01–1.34; P  = 0.035) and ovarian endometriosis (OR = 1.41; 95% CI: 1.07–1.85; P  = 0.013) (Figure  3 ). Regarding “IVs‐2,” genetically predicted mtDNA‐CN was not associated with the risk of ORRDs (Figure  3 ). Also, for “IVs‐3,” the absence of association between mtDNA‐CN and ovarian dysfunction, ovarian cyst, POF, and ovarian endometriosis was confirmed. In contrast, we found that mtDNA‐CN was associated with the risk of PCOS (OR = 1.22; 95% CI: 1.10–1.37; P  < 0.001) (Figure  3 ). Additionally, the weighted median (WM) method had a consistent result (OR = 1.24; 95% CI: 1.06–1.37; P  = 0.006) for PCOS (Table  S1 ). The leave‐one‐out sensitivity test indicated robust results, and the MR Egger analysis suggested no pleiotropy (Table  S1 ). Odds ratios for associations (Assoc.) between genetically predicted mtDNA‐CN and ORRDs. DSL, DerSimonian 6 Laird; IVW, Inverse variance weighted; PCOS, polycystic ovary syndrome; POF, premature ovarian failure; P val, P value. A meta‐analysis was performed due to inconsistent results from the MR analyses conducted across the three different IV sets. The results showed no causal association between mtDNA‐CN and most ovary‐associated diseases, including ovarian dysfunction, ovarian cyst, and POF (Figure  3 ). Surprisingly, we found that genetically predicted mtDNA‐CN were linked to an increased risk of PCOS (OR = 1.16; 95% CI: 1.08–1.25; P  < 0.001) and ovarian endometriosis (OR = 1.25; 95% CI: 1.06–1.47; P  = 0.007) (Figure  3 ). Our study included several genetic variants as instrumental variables in the reverse MR analyses of the association between ORRD and mtDNA‐CN. Specifically, we included four SNPs for ovarian dysfunction, six SNPs for ovarian cyst, four SNPs for PCOS, six SNPs for POF, and 24 SNPs for ovarian endometriosis. The genetic variants used as instrumental variables for the ORRD in the reverse MR analyses are presented in Table  S2 . And F ‐statistics analysis revealed that all variances exceeded 10. The IVW method was utilized as the primary outcome, and the results of the IVW method are presented in Table  3 . Unfortunately, there was no strong evidence for associations of ORRDs with mtDNA‐CN (Figure  4 ). No horizontal pleiotropy was observed for all outcomes (Table  S1 ). The replication analyses by using the study by Hagg et al. (“IVs‐2”) and Longchamps's GWAS (“IVs‐3”) for the reverse association were not performed, due to not enough harmonized data. The results of the leave‐one‐out sensitivity and single SNP analyses, as well as three MR methods for the reverse analyses, were presented in Figure  S2 . Table  3 presents the results of the horizontal pleiotropy and heterogeneity analyses. Mendelian randomization results for causality for ORRDs influencing mtDNA‐CN. Abbreviations: CI, confidence interval; IVs, instrumental variables; IVW, inverse variance weighted; mtDNA‐CN, mitochondrial DNA copy number; nSNP, number of single nucleotide polymorphisms; OR, odds ratio; ORRDs, ovary‐related reproductive disorders; PCOS, polycystic ovary syndrome; POF, premature ovarian failure. Odds ratios for associations (Assoc.) between genetically predicted ORRD and mtDNA‐CN. IVW, Inverse variance weighted; PCOS, polycystic ovary syndrome; POF, premature ovarian failure; P val, P value.

The present study represents a groundbreaking effort in employing bidirectional two‐sample MR to investigate the association between mtDNA‐CN and ORRDs comprehensively. The findings suggest that genetically predicted mtDNA‐CN is a significant indicator that is strongly correlated with PCOS and ovarian endometriosis, while no potential associations were found with other ORRDs. The study did not reveal any causal relationship between genetically predicted ORRDs and mtDNA‐CN. These insights enhance our understanding of mitochondrial function in patients with ORRDs and contribute to a complete understanding of the condition. Previous observational studies have examined mtDNA‐CN levels in PCOS, ovarian endometriosis, and POF patients. While some findings related to PCOS and ovarian endometriosis remain controversial. Additionally, no studies have investigated the relationship between mtDNA‐CN and ovarian dysfunction or ovarian cysts. Our research fills these gaps and is the first to explore the reverse relationship between mtDNA‐CN and ovarian‐related reproductive disorders (ORRDs). ORRDs may be associated with oxidative stress and inflammation, which could influence mtDNA‐CN levels. However, we did not find evidence suggesting that genetic predisposition to ORRDs affects mtDNA‐CN levels. PCOS is a common endocrine disorder affecting women of reproductive age. It is characterized by chronic anovulation, infertility, hyperandrogenism, insulin resistance, and metabolic syndrome. In addition to metabolic imbalances, there is increased production of reactive oxygen species (ROS). 24 Recent research has suggested that mtDNA‐CN and mtDNA mutations were found to be part of an important regulatory mechanism in the etiology of PCOS. 22 , 40 Lee et al. were the first to report a reduction in mtDNA‐CN in the peripheral blood of women with PCOS. 22 Subsequently, Ding et al. identified mutations in mitochondrial tRNA genes associated with PCOS through sequencing. They conducted a series of experiments to support their findings, demonstrating that RNA Leu A3302G, MT‐tRNALeu C3275T, MT‐tRNAGln T4363C, among others, are linked to both PCOS and reduced mtDNA‐CN. 41 , 42 Recently, Tharayil et al. also reported a decrease in relative mtDNA‐CN, negatively correlated with the waist‐to‐hip ratio (WHR) in women with PCOS. 43 These studies have consistently demonstrated a correlation between reduced mtDNA‐CN and PCOS. The underlying mechanism may be that mitophagy eliminates dysfunctional mitochondria, reducing mtDNA‐CN. 18 However, A study conducted by Chen et al. revealed that mtDNA‐CN was significantly higher in Chinese patients with PCOS. 44 They thought such an increase could reflect an adaptive response to mitochondrial dysfunction or a compensatory mechanism to meet the heightened energy demands associated with the condition. 44 This is consistent with our findings. Additionally, another study found no correlation between skeletal muscle mtDNA‐CN and the risk of developing PCOS. 45 The variation in these findings may stem from differences in the sources of mtDNA‐CN data and the methodologies employed for measurement. As a result, the relationship between mtDNA‐CN and PCOS remains ambiguous; it is not clear whether a causal link exists or whether PCOS affects mtDNA‐CN levels. Nevertheless, identifying elevated mtDNA‐CN levels in the peripheral blood of women of reproductive age suggests further evaluation of ovarian function for early screening and prevention. Given these studies' relatively small sample sizes, future large‐scale, multicenter research involving women with PCOS is crucial to validate these findings. Endometriosis is a chronic gynecologic condition similar to malignancy and often affects the pelvic cavity, particularly the ovary. 46 , 47 Pathologic symptoms include ectopic endometrium growth, periodic bleeding, and the formation of ovarian endometriosis cysts (OECs). 46 , 47 Mitochondria possess a crucial role in the cellular physiology of endometriosis. 48 Mitochondria function abnormality was associated with endometriosis. Notably, our research reveals that a genetic predisposition to mtDNA‐CN increases the risk of ovarian endometriosis, but ovarian endometriosis has no association with mtDNA‐CN levels. Huo et al. found that cell free‐mtDNA levels are higher in the follicular fluid (FF) of patients with OECs and are inversely associated with mtDNA expression in mural granulosa and cumulus granulosa cells. 20 However, Derbaly et al. found that the mtDNA‐CN was significantly lower in women with endometriosis compared to healthy controls, 49 which aligns with the findings of Beeram et al. 50 The mtDNA‐CN demonstrated high sensitivity and specificity in differentiating endometriosis from healthy individuals, with a sensitivity of 95% and a specificity of 93.75%. 49 Furthermore, mutations in mtDNA are more prevalent in individuals with endometriosis, contributing to a higher risk of the condition. 51 The differential level of mtDNA‐CN in different cells or tissues in endometriosis and its potential mechanisms affecting the development of endometriosis need to be further investigated. POF, or premature ovarian insufficiency (POI), is a complex condition characterized by various symptoms, including amenorrhea (the absence of menstruation), reduced estrogen levels, and elevated female gonadotropin levels. 52 POI is believed to develop gradually and can be categorized into different stages: decreased fertility (occult POI), elevated follicle‐stimulating hormone (FSH) levels (biochemical POI or bPOI), and eventually amenorrhea (overt POI). 53 Increasing evidence suggests that mitochondrial dysfunction significantly contributes to ovarian aging. 54 However, the precise mitochondrial mechanisms underlying ovarian aging and infertility are not yet fully understood. Research indicates that women with POI exhibit the lowest levels of mtDNA‐CN, whereas those maintaining normal ovarian function show the highest levels. 55 Although the connection between mtDNA‐CN and ovarian function remains incompletely understood, earlier studies have suggested that cell‐free cf‐mtDNA in FF may serve as a valuable biomarker for predicting in vitro fertilization (IVF) outcomes in women without reproductive disorders. 21 , 23 While our study did not establish a causal link between mtDNA‐CN and POF, it is possible that the limited size of our database sample restricts the applicability of our findings. Further research with larger sample sizes is essential to investigate the potential relevance of peripheral blood mtDNA‐CN or FF mtDNA‐CN in patients with POF or POI. The present study offers several benefits. First, the MR design effectively reduces residual confounding and reverse causality, enhancing causal inference accuracy in relationships. Second, we conducted a thorough analysis of causal associations between mtDNA‐CN and ORRD risk using three groups of independent IVs. We also conducted a meta‐analysis to reconcile inconsistent results obtained from these IV sets. Lastly, the causal associations uncovered in our MR analysis offer potential candidates with mtDNA‐CN for subsequent functional studies. This may aid in developing innovative approaches for preventing and treating ORRDs. Nonetheless, our study was subject to certain limitations that warrant consideration. First, the study primarily relied on European subjects, thereby limiting the applicability of our findings to non‐European populations. To validate our conclusions, additional research involving diverse ethnic groups is necessary. Additionally, the genetic tools employed in the study only accounted for a small percentage of the mtDNA‐CN phenotypic diversity, which may have hindered our ability to estimate the association accurately. Finally, while both genders were represented in the mtDNA‐CN data, the ORRDs were limited to females, potentially introducing bias into our outcomes. Our findings indicate that mtDNA‐CN is causally associated with an increased risk of PCOS and ovarian endometriosis at the genetic level may play a causal role. Consequently, mtDNA‐CN has the potential to serve as a non‐invasive biomarker for the early detection of both conditions, particularly when used in conjunction with other biomarkers and imaging techniques. However, the causal relationship between ORRD and mtDNA‐CN remains uncertain. Thus, further research is essential to deepen our understanding of the link between mtDNA‐CN and ORRD.

The role of mitochondria in cellular energy metabolism, differentiation, proliferation, aging, and death has been established. 1 , 2 Mitochondria possess a circular, double‐stranded, haploid, and intron‐free genome, known as mitochondrial DNA (mtDNA), which comprises 37 genes. 3 In a single cell, there are thousands of copies of mtDNA, and the quantity of mtDNA transcripts is primarily determined by the mtDNA copy number (mtDNA‐CN). 4 In normal physiological conditions, mtDNA content within cells remains stable, providing the necessary energy for cell survival. 5 However, damage to mtDNA can serve as a biomarker for mitochondrial dysfunction, resulting in reduced cell metabolic activity. 6 Oxidative stress levels likely influence changes in mtDNA‐CN due to environmental oxidants and gene–environment interactions, both of which are recognized as risk factors for various diseases. 7 Research has shown that mtDNA‐CN levels are associated with the development of multiple illnesses. 8 , 9 , 10 Consequently, mtDNA‐CN could be an accessible biomarker reflecting mitochondrial function and overall health. The ovary, a female gonad, ovulates and secretes sex hormones‐its reproductive and endocrine functions, and the follicle is its basic functional unit. 11 Recent advancements in endocrinology and diagnostic techniques have significantly improved the detection rates of infertility attributed to ovarian factors. Common conditions include polycystic ovary syndrome (PCOS), ovarian tumors, premature ovarian failure (POF), and ovarian endometriosis. These disorders can disrupt hormone secretion and impede ovulation, potentially leading to infertility that may pose serious health risks. 12 , 13 These diseases still lack clinically validated biomarkers or a comprehensive set of biomarkers that develop non‐invasive screening and diagnostic methods. Several studies have shown that oxidative stress and mitochondrial dysfunction are the primary contributors to many ORRDs. 14 , 15 , 16 , 17 , 18 Currently, there are only a limited number of epidemiologic studies exploring the connection between mtDNA‐CN and ORRD. 7 , 10 , 19 , 20 , 21 , 22 , 23 Among these, PCOS has been the most extensively examined, yet the findings have been inconsistent. Some studies report a negative association between mtDNA‐CN and PCOS, while others reveal a positive association. 22 , 24 Most of these investigations have involved small populations with few cases, underscoring the necessity for replication in larger cohorts. Furthermore, none of these studies have established whether the relationship between mtDNA‐CN and ORRDs is causal. A more comprehensive understanding of the role of mtDNA‐CN may provide valuable opportunities for the early prevention and treatment of ORRDs. Mendelian randomization (MR) is a sophisticated technique that leverages data from genome‐wide association studies (GWAS) to mimic the conditions of a randomized controlled trial. 25 , 26 , 27 This mimicry is achieved through allelic randomization during meiosis, establishing a fixed genotype at conception. 28 By employing genetic variants as instrumental variables (IVs), MR is generally more adept at minimizing residual confounding and reverse causation than traditional observational methods. 28 , 29 As a result, the MR approach enhances the causal connections between exposures and outcomes. 25 , 26 Previous research has investigated the causal relationship between mtDNA‐CN and various health risks, such as Crohn's disease, 30 cardiometabolic disease, 31 and pan‐cancer. 32 While some of these studies utilized MR designs, one study revealed no causal association between mtDNA‐CN and ovarian cancer. 32 Nonetheless, MR design has yet to be applied to examine the relationship between mtDNA‐CN and other ORRDs, including ovarian dysfunction, ovarian cysts, PCOS, POF, and ovarian endometriosis. In this study, we performed a bidirectional two‐sample MR analysis to evaluate the causal relationship between mtDNA‐CN and the risk of ORRDs, including ovarian dysfunction, cysts, PCOS, POF, and endometriosis.

The authors declare no conflict of interest.

We performed bidirectional two‐sample MR analyses based on the latest summary statistics of genome‐wide association studies (GWASs) to investigate the associations between mtDNA‐CN and ORRD as well as to test whether ORRD leads to change in the level of mtDNA‐CN. 33 , 34 To test the causal effect of risk factors on an outcome, MR utilizes single nucleotide polymorphisms (SNPs) as instrumental variables. This method is favored as it eliminates any reverse causation bias and confounding factors, as SNPs are randomly allocated during meiosis according to Mendel's laws. However, for MR analysis to be successful, three key assumptions must be met: (1) the SNPs are linked to the exposure, (2) confounders do not influence the SNPs in the exposure‐outcome relationship (known as the independence assumption) and (3) the SNPs only affect the outcome through the exposure. Genome‐wide association studies (GWAS) summary data were collected by sourcing IVs from three distinct studies: (1) IVs‐1 was extracted from the study by Chong et al., 35 using automatic mitochondrial copy (AutoMitoC) to detect mtDNA levels in blood samples. The work involved 395 781 UK Biobank (UKB) participants 35 ; (2) IVs‐2 was extracted from the study by Hägg et al. 36 The work, which assessed mtDNA‐CN using the weighted intensities of probes targeting the mitochondrial genome, included 295 150 UKB participants; and (3) IVs‐3 was extracted from the study by Longchamps et al., 2 which included 465 809 white individuals from the Cohorts for Heart and Aging Research in Genomic Epidemiology (CHARGE) consortium and the UK Biobank (UKB). GWAS data on ORRDs was from FinnGen ( https://www.finngen.fi/en ), and all subjects were of European descent. All the summary data used in the study are publicly available, and the detailed information is shown in Table  1 , and links are provided in Table  S3 . A brief description of ORRDs' GWAS summary statistics. Abbreviations: GWAS, genome‐wide association study; ORRDs, ovary‐related reproductive disorders; PCOS, polycystic ovary syndrome; POF, premature ovarian failure. The data were published and ethical approval was granted so no further ethics committee approval or informed consent was required. Strict selection criteria were employed to choose qualified instrumental variables. We included SNPs at the genome‐wide significance level ( P  < 5e −8 ) for most ORRDs and all mtDNA‐CN. However, to identify enough SNPs (number ≥3) in common between exposure and outcome, we also selected SNPs with genome‐wide suggestive significance P values ( P  < 5e −6 ), including POF. None of the instrumental SNPs were in linkage disequilibrium (LD, r 2  < 0.001 within a clumping window of 10 000 kb), as this situation may cause a misleading outcome. A clumping process was employed, wherein SNPs were clumped based on LD in the given genome region. Independent SNPs were identified through clumping, using a threshold of r 2  < 0.001 and a window size of 10 000 kb. We calculated the proportions of phenotypic variation interpreted by IV and assessed the intensity of the selected SNPs with the F statistic ( F  = beta 2 /se 2 ) to present the strength of the instruments. 37 SNPs with strong instrumentation were identified as having an F ‐statistic >10. The step‐by‐step approach devised for conducting MR is presented as a flowchart in Figure  1 . The two‐sample IVW method was used as the main MR analysis to estimate their causal relationship. We performed a sensitivity analysis to assess the significance of our results. We used the MR‐Egger regression to confirm the presence of horizontal pleiotropy. If horizontal pleiotropy was detected among the selected SNPs, the analyses were repeated after removing these pleiotropic SNPs. 38 We performed Cochran's Q test to evaluate the heterogeneity among SNPs associated with each microbial taxon. The leave‐one‐out sensitivity analysis was used to assess the effects of a single SNP on the overall estimates. To facilitate the integration of the IVs from the three sets, a meta‐analysis using the DerSimonian and Laird methods was conducted to consolidate the findings. 39 A two‐sided P value less than 0.05 was set as statistically significant. All the analyses were performed using TwoSampleMR package (version 0.5.10) and MR‐PRESSO package (version 1.0) of the R program (version 4.3.2). Furthermore, a meta‐analysis was performed using STATA (version 17.0). Basic assumption of Mendelian randomization. IV, instrumental variables; PCOS, polycystic ovary syndrome; POF, premature ovarian failure.

Figure S1. Figure S2. Table S1. Table S2. Table S3. Data S1.