Causal relationships between leukocyte telomere length and female reproductive system diseases: a bidirectional Mendelian randomization study

Telomeres, located at the ends of linear eukaryotic chromosomes, contain a highly conserved repeat sequence of 1–5 kb in length (TTAGGG) [ 1 ] . These sequences are pivotal in ensuring chromosome stability and maintaining cellular activity [ 1 ] . In most somatic cells, telomeres gradually shorten with an increasing number of cell cycles. To compensate for cellular aging caused by telomere shortening, telomerase can add repeat sequences to the ends of chromosomes [ 2 ] , although it may not fully compensate for the extent of telomere loss. When telomeres shorten to a certain threshold length, they activate a persistent DNA damage response, inducing cell aging and apoptosis [ 3 ] . Moreover, oxidative stress and chronic low-grade inflammation may accelerate the shortening of telomeres [ 4 ] , and the accumulation of senescent cells is believed to play a vital role in aging and age-related diseases [ 5 ] . Whereas telomere shortening is a characteristic feature of normal aging, it tends to be more pronounced in diseases characterized by accelerated aging, such as primary ovarian insufficiency (POI) [ 6 ] , infertility [ 7 ] , diabetes [ 8 ] , cardiovascular diseases [ 9 – 11 ] , and neurodegenerative diseases [ 12 ] . Although constituting significant societal health and financial burdens, the etiologies of different female reproductive system diseases (RSDs), such as polycystic ovarian syndrome (PCOS) [ 13 ] , endometriosis (EMS) [ 14 ] , and leiomyoma of the uterus (LU) [ 15 ] , have yet to be sufficiently clarified. Therefore, identifying biomarkers associated with female RSDs would be of considerable value with respect to diagnosis and treatment. Several observational studies have previously investigated the role of telomere length in the development of female RSDs, among which, it has been demonstrated that in EMS patients, leukocyte telomere length (LTL) is longer compared with that in the general population [ 16 ] . In contrast, the findings of a further population-based study have provided evidence to indicate that shorter LTLs are associated with a greater likelihood of a history of EMS [ 17 ] . Moreover, in other studies, the LTLs in PCOS patients were found to be shorter [ 18 – 20 ] , longer [ 21 , 22 ] , or not significantly different [ 23 – 25 ] compared with those in control group patients, thereby highlighting the inconsistencies in the findings of these studies. Such inconsistency can presumably be ascribed to the fact that the results obtained in conventional observational studies are inevitably affected by reverse causation or unmeasured confounding effects [ 26 ] . Mendelian randomization (MR) is a widely used analytical method that can be used to elucidate the causal effects of different genetic exposures that can be predicted for complex diseases [ 27 ] . This method has also found application in the realm of obstetrics and gynecology [ 28 – 30 ] . Compared with traditional observational studies, MR has an inherently lower risk of the confounder effect, as genetic variants are randomly distributed at conception and thus are not associated with environmental factors. Moreover, this approach can aid in reducing reverse causation, given that disease status cannot alter germline phenotypes. However, despite these advantages, the two MR studies that have examined the correlation between telomere length and endometrial and ovarian cancer (OCa) have yielded inconsistent results [ 31 , 32 ] . To date, there have been no studies that have investigated the effects of telomere length on a broad range of female RSDs. In this study, we adopted an MR approach to examine the associations between genetic liability and telomere length with respect to nine types of female RSD, namely, cervical cancer (CCa), endometrial cancer (ECa), OCa, EMS, LU, uterine prolapse (UP), PCOS, POI, and ovarian cysts (OC).

All authors were involved in the design of this study. J.S.S., R.L., and B.K.Z. obtained and analyzed the data. J.S.S., R.L., and X.J.W. drafted the manuscript. M.Z.D., L.X.W., Y.N.C., H.Y.W., Y.H.L., and F.G. critically revised the manuscript. L.G. and H.F.H. contributed to the discussion and edited the manuscript. H.F.H. is the guarantor of this work and, as such, has full access to all data associated with the study and takes responsibility for the integrity and accuracy of data analysis. All authors have reviewed the manuscript in its final form and approved the submitted version.

We performed bidirectional MR analysis to assess the causal relationships between LTL and RSDs using publicly available data, the methodology of which is illustrated in Fig. 1 . Using this approach, the genetic variants selected to estimate the causal effect must satisfy the following three key assumptions [ 27 , 33 ] : (I) the instrumental variables (IVs) are associated with the exposure; (II) the IVs are independent of any known or unknown confounders that mediate the exposure to the outcome; and (III) the outcome is associated with the genetic instrument only through the effect of exposure. A simplified representation of the study basis. I, II, and III denote assumptions I, II, and III, respectively. Data regarding the variables for genetic variants associated with LTL were obtained from a genome-wide association study (GWAS) meta-analysis ( n  = 472,174) [ 34 ] . The data were obtained from the UK Biobank, and telomere length measurements were obtained using an established quantitative polymerase chain reaction (PCR) assay. Data from GWASs of CCa ( n  = 239,158), ECa ( n  = 240,027), OCa ( n  = 246,520), UP ( n  = 234,822), and OC ( n  = 218,469) were obtained from a large genome-wide meta-analysis combining summary data from the UK Biobank and the FinnGen consortium [ 35 ] . The summarized statistics for the PCOS data were collected from a genome-wide meta-analysis ( n  = 141,355), obtained from the FinnGen consortium and the Estonian Biobank [ 36 ] . In addition, we obtained aggregate data regarding EMS ( n  = 77,257), LU ( n  = 123,579), and POI ( n  = 118,482) from the FinnGen consortium. All GWAS data samples were derived from European populations. Supplementary Table 1, http://links.lww.com/RDM/A75 , provides a summary of all datasets included in this investigation. To select single-nucleotide polymorphisms (SNPs) that fit the aforementioned three assumptions, we used TwoSampleMR (version 0.5.6) in the R (version 4.3.1). Initially, we selected the IVs that are associated with exposure. We screened SNPs associated with LTL (or female RSDs in the reverse analysis) at a genome-wide significance level ( P <5 × 10 −8 ). However, when SNPs that strongly predicted several female RSDs were extracted at the P <5 × 10 −8 level, the number of available SNPs was low or even zero. Consequently, we adjusted the threshold of significant SNPs to P <5 × 10 −6 to obtain SNPs that predict female RSDs, including CCa, ECa, OCa, and OC in the reverse MR analysis. Secondly, we ensured that each SNP is independent. We assessed the related linkage disequilibrium to ascertain the presence of SNPs in a state of linkage disequilibrium, and ensured the independence of the SNPs by excluding those within a 1000-kb window with an r 2   0.8) as a substitute. To avoid a weak tool bias, variance ( R 2 ) and F statistics were employed to assess the robustness of SNPs. An F statistic value larger than 10 was deemed sufficiently significant to ensure that the association between SNPs and exposure assessed in the MR analyses was not influenced by a weak tool bias [ 27 ] . Thirdly, we extracted the exposure SNPs from the outcome GWAS and excluded those associated with the outcome ( P <5 × 10 −8 ). Harmonization was conducted to align the alleles of exposure and outcome SNPs, and we discarded palindromic SNPs or SNPs with incompatible alleles. Finally, putative pleiotropic effects were eliminated by retrieving the secondary phenotype of each SNP from PhenoScan V2 [ 37 ] . We used the Phenoscan V2 online tool to input the SNPs that were strongly associated with exposure into the Phenoscanner website in the form of rsID, and the website automatically generated the traits associated with the input SNPs. If the trait is directly related to the outcome, the SNP corresponding to this trait is regarded as a confounding SNP. After removing the confounding SNPs, the MR analysis was repeated. The remaining SNPs were subsequently aggregated into the outcomes GWAS database, and palindromic SNPs with intermediate allele frequencies were removed. To analyze the causal association between LTL and female RSDs, we utilized several MR methods in this study to ensure the stability and reliability of our results. The inverse variance weighted (IVW) method was used for the primary analysis to assess causal relationships. As supplementary analyses, we used the weighted median (WM), MR-Egger (MRE) regression, simple mode (Sm), and weighted mode (Wm) methods [ 38 , 39 ] . The IVW method is an extension of the Wald ratio estimator based on meta-analytic principles that can provide an accurate estimate in an ideal state in which all included SNPs are presumed to be valid IVs without pleiotropy [ 40 ] . The IVW assigns effect estimation weights to each IV based on its effect estimation and variance, such that IVs with smaller variances have greater weights [ 41 , 42 ] . The WM method is the median of the distribution function obtained by ranking the SNP effect values of all individuals according to their weights. When at least 50% of the information is derived from effective IVs, WM can obtain robust estimates [ 43 ] . The MRE method considers the possible heterogeneity of IVs and provides a corrected estimate of causal effects [ 44 ] , whereas the SM method selects a strongly correlated SNP as the IV and uses the effect estimation of this SNP to estimate the causal effect of exposure on the results, and the Wm method for causal effect estimation uses the summary data of multiple genetic instruments [ 45 ] . We proceeded with the application of MR-PRESSO to identify any outlier SNPs that might be present [ 38 ] . The MR-Egger intercept was also utilized to investigate the possibility of directional pleiotropy [ 39 ] . To examine the heterogeneity of IVs in the GWAS dataset of outcomes, we used MR Egger and IVW to derive Cochran’s Q statistics [ 46 , 47 ] . In addition, we generated scatter plots and funnel plots to assess the causal effects of individual hypotheses and to verify the consistency of the results. Moreover, we carried out “leave-one-out” analyses, wherein each instrument SNP was excluded one by one. All statistical analyses were performed using R version 4.3.1. For the MR analyses, we used TwoSampleMR (version 0.5.6) [ 48 ] and MRPRESSO (version 1.0) [ 38 ] .

In this study, we eventually obtained 136, 121, 123, 124, 121, 126, 109, 131, and 133 SNPs as the IVs for LTL to assess the associations between LTL and CCa, ECa, OCa, EMS, LU, UP, PCOS, POI, and OC, respectively (Supplementary Table 2, http://links.lww.com/RDM/A76 ). For the reverse MR analysis of the effects of RSDs on LTL, we selected 15, 14, 19, 11, 20, 8, 4, 11, and 24 SNPs associated with CCa, ECa, OCa, EMS, LU, UP, PCOS, POI, and OC, respectively (Supplementary Table 2, http://links.lww.com/RDM/A76 ). In each case, an absence of bias from weak IVs was indicated by an F statistic value greater than 10. As depicted in Fig. 2 , when LTL was considered as the exposure, the IVW analysis revealed a significant positive causal relationship between the genetically predicted LTL and LU (odds ratio [ OR ] = 1.73, 95% confidence interval [ CI ]: 1.52–1.98, P  = 4.9E-16). The IVW results were also supported by those obtained using WM ( OR  = 1.60, 95% CI : 1.32–1.92, P  = 9.6E-07), MRE ( OR  = 1.83, 95% CI : 1.42–2.38, P  = 1.0E-05), Sm ( OR  = 1.56, 95% CI : 1.03–2.37, P  = 0.039), and Wm ( OR  = 1.64, 95% CI : 1.18–2.28, P  = 0.004). Genetic liability to LTL was also found to be positively associated with OC, and the correlation was supported by IVW ( OR  = 1.31, 95% CI : 1.19–1.45, P  = 1.5E-07), WM ( OR  = 1.29, 95% CI : 1.07–1.55, P  = 0.007), MRE ( OR  = 1.42, 95% CI : 1.17–1.72, P  = 4.3E-04), and Wm ( OR  = 1.32, 95% CI : 1.08–1.63, P  = 0.028) analyses. Additionally, using the IVW method, we obtained evidence to indicate that a longer LTL is associated with a heightened risk of EMS ( OR  = 1.25, 95% CI : 1.06–1.46, P  = 0.008). Mendelian randomization analysis of the associations between leukocyte telomere length and nine female reproductive system diseases. Numbers in bold red font indicate that the relationship has nominal statistical significance ( P values <0.05). CI : confidence interval; IVW: inverse variance weighted; MRE: Mendelian randomization Egger; nSNPs: number of single-nucleotide polymorphisms; OR : odds ratio; SE : standard error. Although we also observed that a longer LTL is associated with an increased risk of ECa ( OR  = 1.59, 95% CI : 1.01–2.48, P  = 0.046) using the MRE method, we failed to detect any causal relationships using the IVW method ( P  = 0.059), the main outcome in MR analysis. Moreover, we identified no causal relationships between LTL and the other five assessed diseases, namely, CCa (IVW: P  = 0.650), OCa (IVW: P  = 0.487), UP (IVW: P  = 0.482), PCOS (IVW: P  = 0.415), and POI (IVW: P  = 0.541). The effects of each SNP on LTL compared with its effect on each of nine assessed female RSDs is illustrated in scatter plots (Supplementary Fig. 1, http://links.lww.com/RDM/A67 ), and a forest plot was generated to display the causal effects of each LTL-associated SNP on the nine female RSDs (Supplementary Fig. 2, http://links.lww.com/RDM/A68 ). In our reverse MR analysis, we detected no causal effects of the nine female RSDs on LTL, with values of 1 being obtained for OR s and all P values being statistically non-significant when conducting analysis using the IVW method, with consistent results being obtained using the other MR methods (Fig. 3 ). The effects of SNPs on RSDs compared with those on LTL are presented in scatter plots shown in Supplementary Fig. 3, http://links.lww.com/RDM/A69 , and the causal effects of each RSDs-associated SNP on LTL are depicted in forest plots shown in Supplementary Fig. 4, http://links.lww.com/RDM/A70 . Mendelian randomization analysis of nine female reproductive system diseases on leukocyte telomere length. CI : confidence interval; IVW: inverse variance weighted; MRE: Mendelian randomization Egger; nSNPs: number of single-nucleotide polymorphisms; OR : odds ratio; SE : standard error. On the basis of MR-Egger intercept analysis (Supplementary Table 3, http://links.lww.com/RDM/A77 ), we obtained no evidence of directional pleiotropy, thereby indicating that it is unlikely that the relationship between exposure and outcome would be influenced by potential confounding factors via different pathways. Moreover, sensitivity analyses provided no indication of heterogeneity, as evidenced by the results obtained using Cochran’s IVW and MR Egger Q tests (Supplementary Table 3, http://links.lww.com/RDM/A77 ). Funnel plots revealed the impact distribution of single SNPs (Supplementary Fig. 5, http://links.lww.com/RDM/A71 , Supplementary Fig. 6, http://links.lww.com/RDM/A72 ), whereas the results obtained using leave-one-out sensitivity analysis tended to indicate that causality is unlikely to be influenced by any individual SNP (Supplementary Fig. 7, http://links.lww.com/RDM/A73 , Supplementary Fig. 8, http://links.lww.com/RDM/A74 ). Accordingly, these findings highlight the reliability of our conclusions.

In conclusion, our MR analyses in this study provide evidence to indicate that a longer LTL is associated with a heightened risk of developing LU, OC, and EMS. However, RSDs have no causal effects on LTL. Consequently, LTL could serve as a potential biomarker for these three diseases. We accordingly believe that these results warrant further validation by other researchers. Moreover, we propose that further studies should undertake more in-depth analyses of the precise roles and underlying mechanisms of LTL in the development of leiomyoma of the uterus, OC, and EMS.

In recent years, the association between telomere length and the risk of female RSDs has received considerable attention. To the best of our knowledge, this study is the first large-scale, two-sample, bidirectional MR study to examine the potential causal association between telomere length and nine female RSDs. Our findings in this study indicate that genetically predicted telomere length is positively correlated with the risk of uterine fibroids, EMS, and OC, whereas conversely, the results of reverse MR analysis revealed that none of the nine female RSDs have a causal influence on telomere length. The findings of previous studies have provided evidence to indicate that longer telomere lengths are associated with enhanced cell proliferation and repair capabilities, and may also contribute to suppressing apoptosis [ 49 ] . Telomere length has also been established to be closely associated with the maintenance of stem cell division and growth [ 50 ] . In addition, studies have indicated that a longer telomere length is positively correlated with estrogen levels [ 51 ] . Leiomyomas of the uterus are the most common benign pelvic tumors in women, with an incidence rate of up to 70% [ 52 ] . Although three previous observational studies have provided evidence of shorter telomere lengths in patients with LU [ 53 – 55 ] , our findings in the present study indicate that a longer telomere length is associated with an increased risk of developing this disease, which is a novel finding in this field of research. Moreover, no previous studies have used MR analysis to assess the causal relationship between telomere length and the risk of LU. We speculate that longer telomere lengths may promote the growth of LU by maintaining the division and growth of uterine smooth muscle stem cells, thereby promoting the proliferative capacity of these cells. Moreover, as previously mentioned, estrogen levels have been shown to be positively correlated with telomere length [ 51 ] , and thus excessive estrogen stimulation may promote an increase in telomere length, thereby contributing to the growth and proliferation of uterine smooth muscle cells and thus enhancing the risk of LU. However, other mechanisms may be implicated in this association, and hence, further research is needed to verify this hypothesis. OC are common gynecological abnormalities that can occur at any age, although they are most prevalent during the reproductive years [ 56 ] . Following ovulation, one or both ovaries may develop a fluid-filled sac referred to as an OC, which can be detected through ultrasound and is mostly stable [ 56 ] . However, as the condition progresses, various complications such as bleeding, cyst rupture, and pelvic pain may occur [ 57 ] . The etiology of OC remains unclear, and to date, no studies have been published regarding the association (if any) between telomere length and OC. We are thus probably the first group to assess the causal relationship between LTL and OC and have accordingly found that a longer LTL is a risk factor for the development of OC. It can thus be speculated that an increase in LTL may be one of the mechanisms underlying the development of OC. In this regard, longer telomeres may promote the active proliferation of granulosa and theca cells in the follicles and inhibit the apoptosis of follicular cells, leading to abnormal hormone secretion and ovulatory disorders, which in turn result in the formation of OC. Further research is, however, needed to confirm the role of a longer LTL in OC development. With respect to EMS, previous observational studies have reported inconsistent results regarding telomere lengths in patients with EMS compared with control group patients, variously indicating shorter [ 18 – 20 ] , longer [ 21 , 22 ] , or no significant differences [ 23 – 25 ] in telomere length. Our findings in the present study provide evidence in support of the contention that longer telomeres enhance the risk of developing EMS. Longer telomere lengths may inhibit the apoptosis of ectopic endometrial cells and promote the proliferation and differentiation of progenitor or stem cells, thereby facilitating the infiltration, survival, and growth of ectopic endometrial cells. Moreover, the findings of previous studies have indicated that estrogen plays a significant role in the pathology of EMS [ 58 ] , and consequently, it is plausible that an excessive production of this hormone may also stimulate the growth and proliferation of ectopic endometrial tissue, thereby increasing the risk of EMS. In addition to the aforementioned findings indicating putative associations between LTL and RSDs, we also obtained results indicating no evident associations when performing forward MR analysis. Human papillomavirus infection is the primary cause of CCa [ 59 ] , and consistent with our findings in this study, previously conducted studies have presented evidence to indicate that LTL is not associated with the risk of CCa [ 60 ] . Similarly, in the case of ECa, which may be associated with prolonged estrogen stimulation and genetic mutations, previous studies have indicated no significant association with telomere length [ 61 ] , consistent with our findings in the present study. Conversely, whereas it has previously been suggested that LTL is associated with the prognosis of OCa [ 62 ] , thus indicating that LTL may influence the progression of this cancer, our findings provided no evidence to indicate a causal relationship between LTL and OCa. Accordingly, it is conceivable that whereas LTL may not be a causal factor promoting the development of OCa, it may influence the severity of this disease. UP is primarily a disease characterized by changes in anatomical structure [ 63 ] , and the present study is the first to examine the association between telomere length and UP, with our findings indicating no causal relationship between the two. Moreover, no studies have yet sought to determine whether the length of telomeres has an influence on the anatomical structure of organs. Although the etiology of PCOS has yet to be sufficiently established, our findings tend to indicate that LTL is not associated with PCOS, which is consistent with the result obtained in a previous observational study [ 24 ] . Observational studies conducted to date have, nevertheless, present evidence to indicate an association between a shortened telomere length and POI [ 6 ] , although our findings would tend to indicate that LTL is not causally associated with POI. The aforementioned disparities between the findings of the present and previous studies could be attributable to the fact that, unlike MR-based analyses, observational studies are often affected by confounding factors. Notably, the results of our reverse MR analysis revealed no causal association between the nine female RSDs and LTL, thus indicating that the status of these RSDs has no appreciable effects on LTL. We found that the trends in results obtained in previous observational studies tended to be inconsistent and may also differ from those obtained when performing MR analysis. This can be explained by the fact that the results obtained in traditional observational studies are inevitably influenced by reverse causation, sample size, or unknown confounding factors. Contrastingly, MR analysis has the distinct advantage of being largely unaffected by these factors [ 27 ] . Moreover, in this study, for the purposes of conducting two-sample bidirectional MR analyses, we used GWAS cohort data comprising tens of thousands, hundreds of thousands, and even millions of samples. Such large sample sizes can contribute to improving statistical power, thereby increasing the likelihood that the study results will be more dependable. However, despite our important findings, this study has inevitable limitations. The GWAS data used in this study were obtained only from individuals of European ancestry, and consequently, the results are not representative of other ethnicities or geographic regions. Accordingly, caution needs to be exercised when generalizing our observations to non-European populations. Thus, to gain a more comprehensive understanding of the associations between telomere length and the female reproductive system, future research should focus on expanding the scope of analysis to include a wider range of ethnic groups, which would thereby contribute to broadening the applicability of the study’s conclusions.

The study was supported by the National Key Research and Development Program of China (2021YFC2700701, 2022YFC2703803, 2022YFC2703001), the National Natural Science Foundation of China (82088102, 82071731, 82171613, 8227034, 81601238), Chinese Academy of Medical Sciences Innovation Fund for Medical Sciences (2019-I2M-5-064), the Science and Technology Commission of Shanghai Municipality (21Y11907600), Shanghai Municipal Commission of Health and Family Planning (20215Y0216), Collaborative Innovation Program of Shanghai Municipal Health Commission (2020CXJQ01), Clinical Research Plan of Shanghai Hospital Development Center (SHDC2020CR1008A), Shanghai Clinical Research Center for Gynecological Diseases (22MC1940200), Shanghai Urogenital System Diseases Research Center (2022ZZ01012), Shanghai Frontiers Science Research Base of Reproduction and Development, The Science and Technology Commission of Quzhou Municipality (2022K54), Open Fund Project of Key Laboratory of Reproductive Genetics, Ministry of Education, Zhejiang University (KY2022035), and Open Fund Project of Guangdong Academy of Medical Sciences (YKY-KF202202).

Supplementary information is linked to the online version of the paper on the Reproductive and Developmental Medicine website.

All authors declare no conflicts of interest. He-Feng Huang is Editorial Board member of Reproductive and Developmental Medicine. The article was subject to the journal’s standard procedures, with peer review handled independently of these Editorial Board member and their research groups.

The authors acknowledged the United Kingdom Biobank and the FinnGen consortium for contributing the data used in this work. We thank all the genetics consortiums for making the GWAS summary data publicly available.

The original contributions presented in the study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding author.