Infertility is defined as an inability to achieve pregnancy after 12 months of unprotected intercourse (1). With the growing global trend for aging population and low fertility rates, infertility has garnered extensive attention in recent years. According to statistics, infertility is estimated to affect as many as 186 million people worldwide (2), accounting for ~15% of couples of childbearing age (3, 4). The incidence and prevalence of infertility vary from country to country, but the overall prevalence has been continuously rising in all countries. The reported rates of infertility are 10-15% in America (5, 6), 15% in Denmark (7), 11.9-17.6% in China (8, 9), 13-20% in Iran (10) and 13.5% in Korea (11). It is estimated that approximately half of infertility cases are attributed to genetic factors, although the underlying causes remain currently unidentified and poorly characterized for most patients (12). Several studies have reported the association between the development of female infertility and genetic polymorphisms (13–15), which indicates that genetic factors may have a substantial role in the development of infertility. As to idiopathic infertility, although its etiology remains elusive, a growing body of evidence indicates a key role of the genetic component involved in the development and progression of the disease. A recent article in the New England Journal of Medicine reported that approximately 17% of the women with idiopathic infertility has pathogenic or likely pathogenic variants for genetic diseases (16).

The fragile X mental retardation gene (FMR1) is localized on chromosome X (Xq27.3), encoding for the FMR1 protein (FMRP), and contains a trinucleotide (CGG) repeat element in the 5'untranslated region (17, 18). A premutation (55–200 CGG trinucleotide repeats in 5' UTR of FMR1 gene) leads to over-representation of FMR1 transcription, aberrant translation products and decline in ovarian function. Premutation (PM) carriers with the above symptoms can be diagnosed as Fragile X primary ovarian insufficiency (FXPOI) (19, 20).

FXPOI is the most commonly known genetic cause for 46, XX POI (21), which is closely related to impaired fecundity. As a consequence, research on the FMR1 gene has extended to other endpoints of reproductive health, including infertility and ovarian reserve (22–27). One study from Latvia reported that the FMR1 gene high-normal alleles are associated with secondary infertility (27), while another study from Switzerland failed to find the association between FMR1 CGG repeat length expansions and infertility (28). Gleicher et al. noted that there is a mild shift toward higher CGG counts at both FMR1 alleles in infertile women in the American population (29). In addition to repeat numbers, the stability of the CGG repeat sequence is also affected by AGG interruptions within the repeat region (30). Since prior studies have shown that AGG interruptions alter the stability of non-canonical structures formed by pathogenic CGG repeats (31, 32), it is possible that AGG interruptions may also play protective roles in fragile X-associated disorders. Recently, Quilichini J et al. have suggested that more than 2 AGG interruptions may be associated with the development of POI, which further confirms the importance of AGG interruptions as a genotypic characteristic for female fecundity (22). The association between FMR1 allele and infertility has been preliminarily explored in the European and American population, however, the relationship between the pattern of AGG interruptions and infertility, especially in Asian population, has not yet been elucidated. The aims of the study, therefore, were to investigate: 1) the relationship of CGG repeat length and AGG interruption patterns at both FMR1 alleles with primary/secondary infertility in the Asian population and 2) whether FMR1 alleles can affect hormone levels in patients with primary/secondary infertility.

A total of 266 patients and 276 controls undergoing FMR1 testing participated in our study. The basic characteristics and FMR1 allelic data of patients in the PI, SI and control groups were presented in Table 1 and Supplementary Figure 1 . The most prevalent CGG repeat length at allele 1 was 29 CGG in all groups (48.6% in the PI group, 46.8% in the SI group and 47.1% in the control group). The mean number of CGG repeat length at allele 1 was 28.7 ± 2.3 in the PI group, 29.1 ± 3.1 in the SI group and 28.6 ± 2.5 in control patients, the values were not significantly different among the three groups (p=0.478). As for allele 2, the most prevalent CGG repeat length was 29 CGG (30.6%) in the SI group and 30 CGG in the other two groups (33.1% for PI group and 34.1% for control group). The mean number of CGG repeat length at allele 2 in the PI (31.8 ± 3.8) and SI (33.1 ± 6.7) group was higher than that in the control group (30.9 ± 3.3), however, only the difference between SI and control group was statistically significant (Bonferroni corrected p=0.003). Only one PM allele (87 CGG repeat length) was found in the SI group, while it was not observed among the control patients. Age, education level and smoking status showed no significant differences among the three groups (p=0.076, p=0.699, p=0.546, respectively). In addition, there was no significant difference in the etiology distribution between the PI and SI groups (p=0.950).

In order to investigate the association of the repeats on FMR1 gene and infertility, linear regression analysis with GLM model for the CGG repeats on both alleles were performed ( Table 2 ). The associations of the CGG repeat length at allele 1 with PI or SI remained nonsignificant after adjusting for potential confounding factors including age, education, smoking status, cohort and the CGG repeat length at FMR1 allele 2 (aOR 1.204, 95% CI 0.716- 2.027, p=0.484 for PI; aOR 1.216, 95% CI 0.706- 2.097, p=0.481 for SI). In addition, the number of CGG repeats at allele 2 were still associated with SI when adjusting for age, education, smoking status, cohort and the CGG repeat length at FMR1 allele 1 (aOR 7.366, 95% CI 2.983-18.188, p<0.001).

To further clarify the specific range of repeat numbers which is associated with infertility, participants were divided into four consecutive FMR1 groups according to the repeat lengths at both alleles. The results were shown in Table 3 and Figure 1 . A different distribution pattern was observed among PI, SI and control groups in both alleles (p=0.005, p<0.001, respectively). For allele 1, when compared to controls, the proportion of patients with 35–44 CGG repeat at FMR1 allele 1 was higher in the PI group and SI group (2.8% vs 0.4%, 4.8% vs 0.4%, respectively; Figure 1A ). For allele 2, there was also a higher proportion of patients with 35–44 repeats in the PI group and SI group than in the control group (23.9% vs 15.2%, 37.9% vs 15.2%, respectively; Figure 1B ). The results of multivariate logistic regression were shown in Table 4 . After adjusting for age, education, smoking status, cohort and the CGG repeat length at FMR1 allele 2, the CGG repeat length at FMR1 allele 1 was not associated with PI (p=0.108), while the association between the CGG repeat length at FMR1 allele 1 and SI showed a trend toward significance (p=0.064). For allele 2, the differences in the proportion of 35–44 repeats between groups remained significant when adjusting for age, education, smoking status, cohort and the CGG repeats at FMR1 allele 1 (aOR=1.751, 95% CI 1.028- 2.985; p=0.039 for PI; aOR=3.657, 95% CI 2.193- 6.098; p<0.001 for SI, respectively). Furthermore, the proportion of the other sized CGG repeats did not differ between groups after adjusting for potential confounding factors (p=ns for all).

As we found the association between 35–44 CGG repeat length at both FMR1 alleles and SI in the above analysis, we further investigated the serum FSH and AMH levels between <35 and 35–44 CGG repeats subgroups ( Figure 2 ). The levels of serum FSH did not differ significantly between groups (p=0.80 for allele 1, p=0.51 for allele 2, respectively; Figures 2A, C ). The mean value of AMH levels in 35–44 CGG repeats subgroup at allele 1 was lower than that in <35 CGG repeats subgroup, although the difference was not statistically significant (p=0.47, Figure 2B ). A higher AMH levels was observed in 35–44 CGG repeats subgroup at allele 2 as compared with <35 CGG repeats subgroup, similarly, there was no statistically significant difference (p=0.05, Figure 2D ). Subsequently, we investigated the relationship between FMR1 alleles and reproductive hormone levels in patients with infertility ( Figure 3 ). After adjusting for confounding factors including age, education, smoking status, infertility type and the CGG repeats of the other FMR1 allele, there was no significant difference in FSH levels with the change of CGG repeat length at FMR1 allele 1 or 2 (p=0.110, p=0.429, respectively; Figures 3A, C ). Decreased AMH levels were associated with an increase in CGG repeat at allele 1 after adjusting for confounding factors including age, education, smoking status, infertility type and the CGG repeats of FMR1 allele 2 (Adjusted R² = 0.178, p=0.003; Figure 3B ). However, no significant correlation was found between the number of CGG repeats at allele 2 and AMH levels (Adjusted R² = 0.150, p=0.086; Figure 3D ).

The pattern of AGG interruption on the FMR1 alleles was also investigated in the study. The top five highest-frequency AGG interruption patterns in the PI, SI and control group were presented in Figure 4 . For allele 1, the top three ranked patterns were 9-A-9-A-9, 9-A-9-A-8 and 9-A-9-A-10 in the PI and SI group, 9-A-9-A-9, 9-A-9-A-10 and 9-A-19 in the control group ( Figure 4A ). Patients in the PI and SI groups showed a 3 AGG pattern (9-A-6-A-9-A-9, 9-A-6-A-9-A-8) for the fourth or fifth ranked AGG interruption pattern while controls carried a continuous stretch of 23 without any AGG interruptions and 19–9 CGG repeats for the same ranks. For allele 2, the top 3 AGG interruption patterns were 9-A-9-A-10, 9-A-9-A-9, 9-A-6-A-9-A-9 in the PI group, 9-A-9-A-9, 9-A-6-A-9-A-9 and 9-A-9-A-10 in the SI group and 9-A-9-A-10, 9-A-9-A-8, 9-A-9-A-9 in the control group ( Figure 4B ). To summarize, the pattern of 3 AGG interruptions (9-A-6-A-9-A-8, 9-A-6-A-9-A-10, 9-A-6-A-9-A-9) appeared earlier in the PI and SI group than in the control group, which indicated that patients with 3 AGG pattern accounted for a higher proportion in the PI and SI group rather than in the control group. The results were corroborated by further analysis of the number of AGG interruptions among different groups ( Supplementary Table 1 ). Although the difference was not statistically significant, there was a higher proportion of 3 AGG interruptions at allele 2 in the SI group as compared to the control group (30.6% vs 11.3%, p=0.099). The proportion of 3 AGG interruptions at allele 1 in the SI group was also higher than that in control group (6.1% vs 0%, p=0.146). We then compared the number of AGG interruptions at allele 1 and allele 2 between patients with <35 and 35–44 CGG repeat length ( Figure 5 ). Patients with 35–44 CGG repeat length showed a higher carrier rate of 3 AGG interruptions at both allele 1 and allele 2 (p<0.001 for both).

Primary ovarian insufficiency (POI), or primary ovarian failure (POF), is now more recognized as a pathological state of ovarian hypofunction, menstrual irregularity, infrequent or absent ovulation. Female fecundity can be impaired by any condition that affects ovulation rate, ultimately leading to infertility (37). FMR1 gene is the most prominent candidate gene of POI, as a consequence, it has been postulated that a relationship exists between FMR1 gene and female infertility (38). However, there is relatively little published studies on the effects of FMR1 CGG repeat size on female infertility, especially in the Asian population. In the study, we firstly analyzed the relationship of FMR1 CGG repeats with PI or SI, we found a higher number of CGG repeats at allele 2 were associated with SI. Patients in the SI group seem to have a higher repeat size (35–44 CGG) at both alleles. Our results first confirmed the relation between higher FMR1 repeat number and SI in the Asian population. Previously, only one publication based on the European population reported the relationship of high normal (35–54 CGG) sized FMR1 repeat length with SI (27). Another study conducted by De Geyter et al. reported that the distribution of FMR1 CGG repeat length did not differed between infertile women and control women (28). Differences in study inclusion criteria may explain the inconsistent results between two European population-based studies. In the former study, the inclusion criteria of patients were SI while whole infertile patients, including patients with PI or SI, were involved in the study by De Geyter et al. (28). Though collectively termed as infertility, these two diseases have vastly different characteristics and molecular pathogenesis. Therefore, it was reasonable for the inconsistent findings in the above two studies. In fact, inconsistent findings of the PI and SI group were also observed in our study, which suggest that FMR1 gene may be more closely associated with SI rather than PI. The underlying reason remains unclear, but it could be attributed to the impact of the FMR1 gene on ovarian function occurring during adulthood rather than affecting the follicle pool (39). The mutation of the FMR1 gene may drive premature depletion of follicles, thereby shortening the reproductive lifespan, ultimately leading to secondary infertility. Since it does not affect the primordial follicle pool, it is unlikely to be a direct contributor to primary infertility.

It is generally believed that the pathogenic mechanism of CGG repeat expansions may be associated with two pathways: either the detrimental effects arising from the overexpression of RNA containing long CGG repeat sequences, or the toxicity caused by products generated through repeat-associated non-AUG translation. The pathogenic products derived from the FMR1 gene impair female fertility by affecting follicular development, inducing follicular depletion and compromising ovarian reserve. One study has demonstrated that the transcriptional activity of CGG repeats ranging from 41–60 is higher than that of repeats with fewer than 40 CGG repeats, which indicates that repeat sequences which have not yet reached the premutation threshold or are in the vicinity of this threshold may still exert potential toxic effects. Although no direct evidence has confirmed a definitive causal relationship between high normal sized CGG repeat sequence and RNA toxicity or FMRpolyG-associated pathology, several studies have provided substantial evidence supporting the potential reproductive toxicity of the alleles (27, 40, 41). In the study, we speculate that lower reproductive function of patients with 35–44 CGG repeat length may be related to reduced ovarian reserve, however, we failed to find decreased AMH levels or increased FSH levels in infertile patients with 35–44 CGG repeat length at allele 2. Patients with 35–44 CGG repeats at allele 1 seem to have a lower AMH levels as compared to patients with <35 CGG repeats, the difference was not statistically significant possibly because of the small sample size. Our results also suggest a modest negative correlation between AMH levels and CGG repeat length at allele 1. Previous studies suggest that FMR1 repeat length had limited predictive ability for ovarian reserve and most of these findings have been reported in American and European populations (28, 42, 43). Considering the unique role of FMR1 allele 1 in the Asian population (44, 45), we supposed that the CGG repeat length on allele 1 appears to have a stronger correlation with AMH levels than allele 2 in Asian women.

Our study represents the first investigation of the association between the pattern of AGG interruption at the FMR1 alleles and female infertility. The AGG interruption at the FMR1 alleles generally includes the following four patterns: 0, 1, 2, or 3 in the expansion of CGG repeat tracts, which can affect allele instability during transmission. Several studies have reported that there was a trend towards the absence of AGG interruption patterns in patients with POF or RSA as compared to that in control women (44, 46). This led us to ask whether a higher proportion of pure CGG repeat tract will be found in infertile patients. Unexpectedly, we did not observe a dramatic increase of 0-AGG interruption patterns in PI or SI group compared to controls. The SI group showed a higher proportion of 3 AGG interruptions at both alleles than in controls. Patients with 35–44 CGG repeat length also showed a higher carrier rate of 3 AGG interruptions at both alleles. It has been shown that the most common AGG pattern of 3 interruptions is 9-A-6-A-9-A-9 (36 CGG repeats) (47, 48), thus, the findings further confirm the unique associations between 35–44 CGG repeat length and SI.

At present, carriers with high normal sized (35–44 CGG) repeat length receive little attention during genetic counseling practice. Female carriers with high normal sized (35–44 CGG) repeat length should also be informed on the potential risk of subfertility, and family members should be offered fragile-X screening and genetic counselling. Further, with the growing demand for extending the effective reproductive age of females, it is necessary for clinicians to increase the attention towards the fertility distress of females with high normal sized (35–44 CGG) repeat length, so as to timely provide fertility guidance for these carriers.

The major limitations of the current study are as follows. Firstly, this study was conducted as a retrospective case-control study with single-ethnicity participants, which may affect the extrapolation of results. Secondly, the sample size of patients who underwent reproductive hormones testing was small, limited sample size affected the accuracy of the results.

Overall, our study suggests that in the Chinese population, the proportions of patients with 35–44 CGG repeat at both FMR1 alleles were higher in patients with SI when compared with controls. The finding reveals a positive association between high normal sized CGG repeat pattern and secondary infertility, the underlying mechanism merits further investigation. In addition, we also find that the CGG repeat length on FMR1 allele 1 are mildly correlated with serum AMH levels in infertile patients.