Association between melatonin receptor gene polymorphisms and polycystic ovarian syndrome: a systematic review and meta-analysis

Abstract Background: Polycystic ovarian syndrome (PCOS) is a kind of common gynecological endocrine disorder. And the mutations of melatonin receptor (MTNR) genes are related to the occurrence of PCOS. But previous researches have shown opposite results. So, the object of our systematic review and meta-analysis is to investigate the relationship between MTNR 1A/B polymorphisms and PCOS. Methods: PubMed, Embase, Ovid, the Cochrane Library, Web of Science and three Chinese databases (VIP, CNKI and Wanfang) were used to retrieve eligible articles published between January 1980 and February 2020. And we used the odds ratio (OR) and its 95% confidence interval (CI) to investigate the strength of the association by six genetic models, allelic, codominant (homozygous and heterozygous), dominant, recessive and superdominant models. Review Manager 5.3, IBM SPSS statistics 25 and Stata MP 16.0 software were used to do this meta-analysis. Results: Our meta-analysis involved 2553 PCOS patients and 3152 controls, for two single nucleotide polymorphisms (rs10830963 C> G in MTNR1B and rs2119882 T> C in MTNR1A) and significant associations were found in some genetic models of these single nucleotide polymorphisms (SNPs). For rs10830963, strongly significant was found in the heterozygote model (GC vs. CC, P=0.02). Additionally, a slight trend was detected in the allelic (G vs. C), homozygote (GG vs. CC) and dominant (GG+GC vs. CC) model of rs10830963 (P=0.05). And after further sensitivity analysis, a study with high heterogeneity was removed. In the allelic (P=0.000), homozygote (P=0.001), dominant (P=0.000) and recessive (GG vs. GC+CC, P=0.001) model, strong associations between rs10830963 and PCOS were found. Moreover, for rs2119882, five genetic models, allelic (C vs. T, P=0.000), codominant (the homozygote (CC vs. TT, P=0.000) and heterozygote model (CT vs. TT, P=0.02), dominant (CC + CT vs. TT, P=0.03) and recessive model (CC vs. CT + TT, P=0.000) showed significant statistical associations with PCOS. Conclusion: MTNR1B rs10830963 and MTNR1B rs2119882 polymorphisms are associated with PCOS risk. However, the above conclusions still require being confirmed by much larger multi-ethnic studies.


Background
Melatonin secreted by the pineal gland during the dark phase of the sleep/wake cycle, is a kind of hormone which periodically regulates several physiological functions [1], including glucose homeostasis [2] and insulin secretion [3]. When the melatonin secretion or the corresponding Melatonin Receptor (MTNR)

Study protocol
Five English databases (PubMed, Embase, Ovid, the Cochrane Library and Web of Science) and three Chinese databases (VIP, CNKI and Wanfang) were searched in our study. And we used three subject heading terms, 'Polycystic ovarian syndrome' , 'melatonin' and 'melatonin receptor' , as well as the synonyms of these terms confirmed by Medical Subject Headings (MeSH) referred to in the Supplementary Material S1 (Search strategies) [41], where other free words were registered. The latest search results were updated on 14 February 2020. And there were no language restrictions in these search strategies. Finally, all references were exported to EndNote X9 software for further research.

Definitions and results
As the definition of PCOS used in the present study, all the eligible studies adopted the 2003 Rotterdam diagnostic inclusion criteria of PCOS [42]: (1) thin ovulation or anovulation. (2) Clinical manifestations of hyperandrogen or hyperandrogenemia. (3) Polycystic ovarian changes, that is, ≥12 follicles with a diameter of 2-9 millimeters on ovary, or ovarian volume > 10 ml. As long as two of the above criteria are met and other diseases are excluded, PCOS can be diagnosed. The results were the genotype distribution in MTNR gene and the odds ratio (OR) of PCOS.

Eligibility criteria
Included references needed to meet the following criteria: 1. Assess the association between SNP in MTNR and PCOS risk, 2. Contain particular data, such as the frequency of MTNR genotypes.
Excluded studies should meet the following criteria: 1. Abstracts and reviews,

Statistical analysis
Review Manager 5.3 software (available from the Cochrane Community, https://community.cochrane.org/help/toolsand-software/revman-5) was performed for data analysis [45]. Each study was evaluated by the Hardy-Weinberg equilibrium (H-W equilibrium, HWE) with a chi-square test (χ 2 test) by IBM SPSS statistics 25 software [46]. Odds ratios (ORs) and their corresponding 95% confidence intervals (CIs) were utilized to test the association between each SNPs, rs2119882 and rs10830963 of MTNR1A/B and PCOS with six genetic models used to test the influence of genotype, that is, for rs10830963 (C> G): allelic (G vs. C), codominant (homozygote (GG vs. CC) and heterozygote (GC vs. CC) models), dominant (GG + GC vs. CC), recessive (GG vs. GC + CC) and superdominant (GG + CC vs. GC) models were calculated; and for rs2119882 (T> C): allelic (C vs. T), codominant (homozygote (CC vs. TT) and heterozygote (CT vs. TT) models), dominant (CC + CT vs. TT), recessive (CC vs. CT + TT) and superdominant (CC + TT vs. CT) models were calculated [47,48]. Our study used Cochrane Q-test for heterogeneity testing and the I 2 , a test statistic, for quantification [49]. In the Q-test, when P≥0.1 or I 2 ≤ 50%, the fixed-effects model could be established to analyze. Otherwise, the random-effects model could be used. Next, a Z-test was performed to evaluate the statistical significance of combined ORs in quantitative synthesis. And all analysis will use 0.05 as significantly statistical level. Finally, by eliminating the heterogeneous merger studies one by one, the effect of a single heterogeneous study on the entire model was further discussed through a sensitivity analysis. And Stata MP 16.0 software was used to analyze the publication bias by Egger's and Begg's tests.

Study selection
In total, 1028 references were identified after searching from databases (refer to Supplementary Material S1, Search strategies) and their recommended links. Then, all searched reference lists were imported into EndNote X9, which helps to remove duplicate studies. The majority of articles were excluded for unrelated titles and abstracts and other eight articles were excluded because of which studied populations were not PCOS patients, genotype data were not valid, or references were unrelated. Additionally,  study [50] was removed because it was a family-based transmission disequilibrium test without control group. Finally, five articles met the quality requirements of meta-analysis and the flow diagram of our study selection is shown in Figure 1.

Study characteristics
The eligible references in the present study involved two genes and six polymorphisms, MTNR1B gene (rs10830963, rs10830962, rs4753426, rs1562444, rs12792653) and MTNR1A (rs2119882). Four references, including Xu et al.   [54], reported rs2119882, including 1680 patients and 1472 controls. Other SNPs had not been included in subsequent analysis due to insufficient data. All included studies with an NOS score more than 7, indicated a good level of quality ( Table 1). The majority populations of the eligible references were Chinese and the methods used to detect gene polymorphisms were polymerase chain reaction (PCR) and sequencing. The studies characteristics are described in Table 2. Allele frequencies and genotypes of each study, and the outcome of HWE was listed in Table 3.

Quantitative synthesis
The results of the association between the MTNR polymorphisms and the PCOS risk are shown in Table 4. The forest plots of each genetic model are described in Figures 2 and 3.

Analysis of rs10830963 polymorphisms in MTNR1B
To study the relationship between rs10830963 (C>G) and PCOS risk, we included four case-control studies (Table  3) and analyzed them with six genetic models [55] (Table 4, Figure 2). Among them, in the heterozygote model, the variation was significantly associated with PCOS occurrence . And in last model, the superdominant model, no statistically significant association were found (OR: 0.99, 95% CI: 0.85-1.14, P=0.87, I 2 = 0%). All studies were evenly distributed on both sides of the center line, and no obvious publication bias was found.

rs2119882 polymorphisms analysis in MTNR1A
To study the association between rs2119882 and PCOS risk, we included two case-control studies (Table 3) and used six genetic models for analysis (Table 4, Figure 3). In five genetic models, allelic, homozygote, heterozygote, dominant, recessive and superdominant, significant statistical associations were found

Sensitivity analysis
For genetic models with high heterogeneity (I 2 and P≤0.01), a sensitivity analysis was conducted by removing all the eligible references one by one in each model to assess the influence of each reference heterogeneity on the whole model analysis [56]. More details in forest plots and funnel plots are shown in Figure 4 and

Publication bias
Due to the limited number of available references [44], publication bias was investigated only for the relationship between rs10830963 and PCOS risk by Stata MP 16.0 software. The result of heterozygote model (GC vs. CC) is  shown in Figure 5. In Egger's test, P was 0.469, while similarly in Begg's test P was 0.497. Because both the P-values were more than 0.1, so no publication bias was found. Relationship between rs10830963 and PCOS risk was relatively stable and convincing.

Discussion
As one among common endocrine disorders in gynecological reproduction, the PCOS pathogenesis is still unclear.
Family studies revealed aggregation in PCOS [57], but no precise genetic mechanism can explain the etiology. Because these patients have hyperinsulinemia and IR, genes like MTNR1A/B related to insulin secretion possibly play a crucial role in PCOS progression. At the same time, recently several researches have been conducted or been registered [58] to determine the relationship between MTNR genes and PCOS susceptibility. But these researches have shown opposite results, so our study aims to further elaborate the relationship between MTNR1A/B gene and the risk of PCOS. Melatonin works by activating two important G protein-coupled receptors, MT1 and MT2 [59]. By querying this National Center for Biotechnology Information (NCBI) database, it is known that the MTNR1A gene that encodes MT1 is located on the chromosome 4q35.2, and the MTNR1B gene of MT2 is located on chromosome 11q14.3. Rs10830963 is located in the intron region between the two exons encoding MTNR1B, while rs2119882 is located in the MTNR1A promoter region. Although these relative positions of SNPs in MTNR are different, the results for all genetic models in this two SNPs are still similar. However, the limitation of our study is that we only discussed from the perspective of MTNR gene while other SNPs are ignored. Because the pathogenesis of PCOS is complex and different genome-wide association studies (GWASs) conclusions are more heterogeneous [60], multiple SNPs mutations may affect metabolic and endocrine profiles in the same PCOS patient, such as insulin receptor gene (INSR) and luteinizing hormone/choriogonadotropin receptor (LHCGR) [61]. Therefore, further research is expected to combining multi-racial GWAS results of arrays by bioinformatics [62,63] and take other gene receptor into consideration to analysis-related molecular pathways [64].
In our meta-analysis, we find that in the heterozygote model, MTNR1B rs10830963 polymorphism is significantly associated with the PCOS occurrence. Through the sensitivity analysis, after excluding Wang et al. study [39], multiple model analyses showed that rs10830963 also has a significant correlation (the allelic, homozygote, dominant and recessive models). This may be due to the large heterogeneity of the present study. First, the study was conducted 10 years ago and its population was not in H-W equilibrium, possibly due to the bias in collecting participants in a limited geographical region. Second, the article does not report the period for recruiting the participants. Therefore, the study cannot rule out the effect of selection bias on the population. Finally, the SNPs are in a strong linkage disequilibrium and there may be a certain association between linkage markers. In addition, due to the large heterogeneity in different ethnicity, our study still needs large sample, multi-center and multi-ethnic studies to confirm.
Moreover, obese women in PCOS patients can be complicated with T2DM [65]. At the same time, their IR level is higher than normal, which can effectively predict impaired glucose tolerance and the occurrence of metabolic syndrome [66]. MTNR is also a candidate gene for IR and gestational diabetes mellitus [38]. So, further studies can take the clinical heterogeneity of PCOS into consideration and divide patients into different groups, such as obese and non-obese groups by body mass index [67], in following subgroup and phenotype analysis of SNPs.
As a result of weak endocrine regulation of melatonin through mutation receptor, to some extent, our meta-analysis supports the feasibility and acceptability of appropriate melatonin supplementation for PCOS [68,69]. At the same time, some studies have suggested that inositol can also be used to an novel and additional treatment of PCOS to improve symptoms [70][71][72], because as an insulin sensitizer it can decrease the levels of androgen, improve glucose metabolism [73], and restore spontaneous ovulation [74,75]. But due to the limited study of melatonin and inositol combination therapy [76,77], the clinical outcomes of IVF are still uncertain [78,79]. Therefore, the follow-up systematic review and meta-analysis is recommended to further confirm the combination therapy effect of melatonin and other drugs for PCOS.

Conclusion
The SNPs rs10830963 and rs2119882 of MTNR1B and MTNR1A, are associated with a much higher PCOS risk in Chinese population. And above conclusions still require confirmation by future study including much larger multi-ethnic studies and other SNPs instead of the above conclusions.

Data Availability
The analytical data are all included in the published articles and the final supplementary materials.

Competing Interests
The authors declare that there are no competing interests associated with the manuscript.