Chromosomal and genetic causes of primary ovarian insufficiency in the Utah population

Abstract

Background

Primary ovarian insufficiency (POI) is a heterogeneous disorder, with the relative frequency of known etiologies incompletely defined at a population level.

Objectives

We used chart review and whole-genome sequencing (WGS) to determine known diagnoses, pathogenic variants, and common polygenic risk.

Methods

Women with POI were identified using International Classification of Diseases codes. Established clinical etiologies were documented. A subset of 68 women without an identified cause underwent WGS. Rare pathogenic variants were identified using an electronic clinical decision support framework, and common genetic risk was quantified using a polygenic risk score (PGS) for age at natural menopause.

Results

Of 899 women with POI, 289 carried a diagnosis of Turner syndrome (32.1%). Of the other 610 evaluable women, established clinical etiologies were identified in 8.1%, including FMR1 premutations (2.6%), chromosomal abnormalities (1.5%), and adrenal autoimmunity (1.8%). Among the 68 women with POI who underwent WGS, a likely causal genetic diagnosis was identified in 17.6% (12/68) of cases. Autosomal recessive variants were enriched among women with primary amenorrhea, whereas heterozygous or multiple rare variants in DNA damage repair pathways predominated in secondary amenorrhea. While overall mean PGS did not differ from controls, low PGS (<−0.5 standard deviation) compounded POI risk in a subset of women carrying rare variants.

Conclusion

The population-level data supports rates of established POI etiologies and a gradation of genetic burden in which chromosomal abnormalities and recessive inheritance contribute to POI manifesting as primary amenorrhea, whereas heterozygous rare variants and common polygenic risk can predispose to secondary amenorrhea.

Primary ovarian insufficiency (POI) is defined by the loss of normal ovarian function, resulting from depletion of oocytes before the age of 40 years (1). Clinical features include alterations in menstrual function, infertility, and symptoms of estrogen deficiency. The diagnosis of POI is confirmed by persistently elevated serum follicle stimulating hormone (FSH) levels in a woman younger than 40 years with amenorrhea.

Excluding iatrogenic causes (ie, surgery and chemotherapy), established etiologies include disorders that lead to accelerated follicular depletion such as X chromosome deletions and translocations, FMR1 premutations, and autoimmune diseases (1). Next-generation sequencing has identified an additional 18% to 30% of the etiologies (2-7). However, the relative frequency of these etiologies has been assessed mainly in patients who present to a single provider or specialized center for evaluation. None have considered both rare and common variation.

The primary objective of this study was to determine known causes of POI from a large cohort of women in Utah, using electronic medical records (EMRs) from 2 healthcare systems (8). These systems serve 85% of the population of Utah and therefore better encompass all women seeking healthcare for POI. After identifying the cause of POI through available records, we also performed whole-genome sequencing (WGS) on a subset of women without a known etiology. The goal was to identify deleterious variants in known genes and to calculate a polygenic risk score for age at menopause, allowing us to determine genetic risk from rare and common variation (9).

We hypothesized that patients with POI presenting with primary amenorrhea would have recessive inheritance of deleterious gene variants. We also hypothesized that single or more complex genetic variants, including common and heterozygous rare variants, would be found in patients with secondary amenorrhea. Together, these EMR and genetic data provide an analysis of known risk and etiology for POI at a population level.

Materials and Methods

Subjects

We identified women aged ≤40 years with POI using electronic medical records from 1995 to 2022 at the University of Utah Health and Intermountain Healthcare, which together serve 85% of all residents in Utah (8). Cases of POI were ascertained through International Classification of Diseases (ICD) codes: ICD-9 (256.3, 256.31, 256.39) and ICD-10 (E28.3, E28.31, E28.39, E28.310, E28.319), EMR notes indicating POI diagnoses, and/or laboratory values (elevated FSH >20 IU/L or anti-Mullerian hormone <0.08 ng/mL) in a woman younger than age 40 years with at least 4 months of amenorrhea in the absence of hormonal contraception or intrauterine device at the time of the laboratory draw. Subjects were excluded for hysterectomy, oophorectomy, and endometriosis with pelvic surgery, pelvic radiation, or chemotherapy before the diagnosis of POI. We identified cases of Turner syndrome through ICD codes: ICD-9 (758.6) and ICD-10 codes (Q96).

The initial list of identified POI and Turner cases was verified through individual chart review by a medical or reproductive endocrinologist (C.K.W. or L.E.V.) for inclusion. Validation factors included FSH and anti-Mullerian hormone levels, type of physician making the diagnosis, and POI symptoms and signs, as outlined previously (8). During chart review, known causes of POI already determined were documented, including Turner syndrome, FMR1 premutations, karyotype abnormalities, and primary adrenal insufficiency or 21 hydroxylase antibody positivity.

Subjects identified by ICD code were recruited for history and DNA sampling of saliva or blood for WGS from 2015 to 2026. In addition to DNA from subjects, DNA from the mother and father of 5 probands and mother of 1 proband was also available. The study was approved by the University of Utah Institutional Review Board (IRB). All subjects provided written, informed consent.

DNA extraction and sequencing

DNA samples from subjects recruited were extracted using the QIAamp DNA prep kits (QIAGEN). Libraries were prepared using the NEBNext Ultra II DNA PCR-free Library Prep Kit (New England Biolabs, Ipswich, MA) and subjected to WGS using the NovaSeq X sequencing platform (Illumina). DNA from all recruited subjects, except those with already completed data or previously collected samples, underwent FMR1 premutation testing through a clinical laboratory if not previously performed (ARUP Laboratories, Salt Lake City, UT).

Alignment and variant calling

Alignment and variant calling were performed by the Utah Center for Genetic Discovery (UCGD) core services. Variants were called through the UCGD pipeline using the Sentieon software package (https://www.sentieon.com) (10). Reads were aligned to the human reference build GRCh38/hg38 using BWA-MEM (Burrows–Wheeler Aligner). SAMBLASTER was used to mark duplicate reads and de-duplicate aligned BAM files. Aligned BAM files underwent insertion or deletion (INDEL) realignment and base recalibration using Realigner and QualCal algorithms from the Sentieon software package3 to produce polished BAM files. Each polished BAM file was processed using the Sentieon Haplotyper algorithm to produce gVCF files (11). Sample gVCF files were combined and jointly genotyped with 728 samples comprised of the 1000 Genomes Project Centre d′E‘tude du Polyporphisme Humain European from Utah (CEU) samples and samples unrelated to reproduction or cancer phenotypes to produce a multisample VCF file. To produce the final VCF variant quality scores, VCF files were recalibrated using the Sentieon VarCal algorithm to estimate the accuracy of variant calls and reduce potential false positive calls.

Quality control

Quality control algorithms were applied to sequence reads (Fastq files), aligned reads (BAM files), and variants (VCF files) (12). Fastp was used to evaluate read quality, read duplication rate, presence of adapter, and overrepresented sequences in Fastq files (13). Indexcov was used to estimate depth and coverage of aligned sequence data using BAM indexes. Further alignment quality metrics were calculated on BAM files with SAMtools stats.

Variant quality metrics were calculated by running BCFtools stats (14-16). The overall quality of VCF call sets was evaluated using Peddy to confirm sex, relatedness, heterozygosity, and ancestry of each individual and identify potential sample-level data quality issues (17).

Identification of damaging gene variants

We used GEM to identify gene variants in each subject that were most likely to be pathogenic (18). GEM is an Electronic Clinical Decision Support System framework that aggregates and adjudicates data from multiple algorithms and clinical datasets to provide rapid and accurate diagnosis of individual genomes (18). GEM generates a Bayes factor–based score that calculates the degree of support for and against a given model (a gene allele is pathogenic vs benign) considering multiple lines of evidence from the following variant analysis tools and data sources: Variant Annotation, Analysis, and Search Tool, Variant Annotation, Analysis, and Search Tool Variant Prioritizer, Phenotype Driven Variant Ontological Re-ranking (PHEVOR) tool, mode of inheritance for disease genes from Online Mendelian Inheritance in Man, pathogenicity of variants (ClinVar), population specific allele frequencies (gnomAD), quality of variants and the overall genome (data), and quality of the genomic location (gnomAD) (19-23). Using this data, GEM identifies potentially pathogenic genotypes and evaluates support for their association with disease. Gene variants for each subject were considered candidates if they had a GEM score ≥1 (strong support for the model of pathogenicity), together with genes having a GEM score ≥0.69 (substantial support for the model of pathogenicity) (21). We also considered genes with a high PHEVOR Tool Bayes factor ≥0.9, which include genes with strong association with POI in conjunction with additional phenotypes identified by associated with Human Phenotype Ontology terms (21). However, we did not require a high PHEVOR score to include recently discovered POI genes. One to 18 gene variants were identified for each subject that were deleterious and in genes related to the subject's phenotype.

Polygenic risk score calculation

Genetic predisposition for the age at natural menopause was quantified using a PGS that included 290 genome-wide significant variants from the largest study of age at natural menopause (9). Raw reads from WGS were aligned to the human genome (GRCh38) using bwa-mem2 (v2.2.1). DeepVariant (v1.6.0) was then used for variant calling on aligned reads. Generated gVCF files were finally merged and joint-genotyped using GLnexus (v1.4.1). Variants were used to calculate PGS using the previously described scoring matrix (9). For the PGS calculation, sites missing from the VCF were considered invariant.

Results

The study cohort comprised of 899 women; 610 women with POI (67.9%) and 289 women (32.1%) with Turner syndrome, as reported previously (24, 25) (Fig. 1). The women with Turner syndrome were 9.1 ± 8.2 years (median age 12 years, range: birth-67 years) at diagnosis (Fig. 2). For the women diagnosed with POI and excluding Turner syndrome, the age at diagnosis of POI was a mean of 32 ± 8 years, with a range of 12 to 40 years (Fig. 2). The majority of these women had secondary amenorrhea (93.3%). Individuals with primary amenorrhea (6.7%) were diagnosed with POI at a younger age, ranging from 12 to 21 years (16 ± 2 years) (Fig. 2).

Figure 1.

Flow chart indicating subjects identified with primary ovarian insufficiency (POI) and their subsequent breakdown into identifiable categories. VUS, variant of unknown significance; WGS, whole-genome sequencing.

Figure 2.

Age distribution at diagnosis of primary ovarian insufficiency (POI). The number of subjects (y axis) with Turner syndrome (black bars) and other causes of POI presenting as primary amenorrhea (gray bars) and secondary amenorrhea (white bars) are indicated.

A total of 744 women (82.7%) had sufficient information to understand the etiology of POI when including subjects with Turner syndrome. Of the women without Turner syndrome, 455 women had sufficient information to determine whether there was an already identified etiology through chart review. These 455 women were then categorized based on the presence of a known associated condition or identifiable etiology. Known etiologies were identified in 37 (8.1%) of the 455 women with sufficient information on chart review, while no etiology was identified for the other 418. Etiologies included FMR1 premutations, chromosome abnormalities that were not classified as Turner syndrome, adrenal autoimmunity, or previously diagnosed syndromic causes of POI (Table 1). The chromosomal anomalies observed encompassed a range of karyotypic findings, such as pericentric inversion of chromosome 9 (Inv 9 [p12q13]), X chromosome translocations ([46, X,t(X;13)(q13;q32)], [46,X,t (X;15)(q22;q11.2)], [46,X,t (X;1)(q21.1;p36.1)]), and a chromosome 21 deletion mosaicism that were presumed to be causative. There was 1 mother/daughter pair with chromosome abnormalities and 1 sister pair with autoimmune ovarian insufficiency. No familial cases were identified in the other categories including the FMR1 premutations.

Table 1.

Frequency of known diagnoses in a population sample of women with primary ovarian insufficiency from Utah (N = 455)

Known disease	Frequency (n)	Percent of POI patients
FMR1 premutation	12	2.6
Adrenal autoimmunity	8	1.8
Chromosomal abnormality	7	1.5
Galactosemia	5	1.1
Blepharophimosis	2	0.4
Other neurologic disease	2	0.4
Congenital disorder of glycosylation type 1A (PMM2)	1	0.2
Total	37	8.1

Genetic analysis using WGS was performed in a subset of the remaining POI patients who were recruited, for a total of 68 individuals (11.1% of the total cohort). Detailed information about their clinical presentation and characteristics is summarized in Table S1 (26). A total of 12 women had a likely genetic diagnosis based on the WGS (17.6%) (Tables 2 and 3), 2 with primary amenorrhea and 10 with secondary amenorrhea, while 26 women had no known variants to explain their POI. The remaining subjects had a variant(s) that was not inherited in the expected manner, the variant was considered benign or of uncertain significance by American College of Medical Genetics (ACMG) criteria (27), or the phenotype did not match (Table S2) (26).

Table 2.

Known gene variants in women with POI and primary amenorrhea

POI ID	POI diagnosis age (years)	PGS	PGS SD from control mean	Gene	Chromosome (chr): location:reference base:variant base	Protein change	Consequence	gnomAD allele frequency (female/male)	Conservationa	Interpretation ACMG criteriab	Expected inheritance
210	14	−2.41	1.57	MCM8	chr20:5958727:G:T	c.789+1G>T	Splice donor loss	2.48 × 10−6	Not applicable	Pathogenic	Autosomal recessive/autosomal dominant
					chr20:5967868:TA:T	p.Tyr356LeufsTer27	Frameshift	3.45 × 10−5	Not applicable	Pathogenic	Autosomal recessive/autosomal dominant
116	15	−3.14	0.99	MCM8	chr20:5952486_C:T	p.Arg71Ter	Stop	3.1 × 10−6	Not applicable	Pathogenic	Autosomal recessive/autosomal dominant
					chr20:5987286:G:A	p.Arg723Gln	Missense	2.05 × 10−5	Yes	VUS	Autosomal recessive/autosomal dominant
269	15	−4.63	−0.22	EIF2B2	14:75005867:G:T	p.Gly200Val	Missense	4.88 × 10−4	Yes	Pathogenic	Autosomal recessive
				EIF4ENIF1	22:31463734:G:A	p.Arg178Trp	Missense	1.37 × 10−5	Yes	VUS	Autosomal dominant
90	16	−2.83	1.23	USP36	chr17:78798922:A:G	p.Phe1076Leu	Missense	1.91 × 10−4	Yes	VUS	Autosomal recessive
172	16	−3.20	0.94	ATG7	chr3:11308987:T:C	p.Trp113Arg	Missense	4.34 × 10−6	Yes	VUS	Autosomal recessive
153	17	−4.77	−0.33	POR	chr7:75984775:AGAG:A	p.Glu357del	Splice acceptor	9.55 × 10−5	Not applicable	Pathogenic	Autosomal recessive

The columns indicate age at POI, PGS, PGS SD from the control mean PGS, gene, variant location, protein change, effect on the protein, allele frequency in gnomAD (23), conservation across species, interpretation of ACMG criteria, and expected mode of inheritance. Note that 2 additional women with POI had no known gene variants identified that were pathogenic and likely to explain their POI.

Abbreviation: ACMG, American College of Medical Genetics.

^a Conservation—species including rhesus, mouse, dog, elephant, chicken, Xenopus tropicalis, zebrafish, and lamprey.

^b Interpretation using ACMG criteria including pathogenic, likely pathogenic, and VUS (27).

Table 3.

Known gene variants in women with POI and secondary amenorrhea

POI ID	POI Age (years)	PGS	PGS SD from control mean	Gene	Chromosome (chr): location:reference base:variant base	Protein change	Consequence	gnomAD allele frequency	Conservationa	Interpretation ACMG criteriab	Expected inheritance
151	23	−3.81	0.44	ATM	chr11:108365423:G:A	p.Gly3029Asp	Missense	3.9 × 10−4	Yes	VUS	Autosomal recessive/autosomal dominant
				MSH5	chr6:31761553:A:G	Thr707Ala	Missense	Novel	Yes	LP	Autosomal recessive
231	31	−4.54	−0.15	FANCM	chr14:45164413:G:A	p.Gly546Ser	Missense	1.18 × 10−4	Yes	VUS	Autosomal recessive
				FANCM	chr14:45189123:C:T	p.Gln1701Ter	Stop	9.55 × 10−4	Not applicable	P	Autosomal recessive
134	31	−3.89	0.38	MCM8	chr20:5978005:C:T	p.Leu509Phe	Missense	3.72 × 10−6	Yes	VUS	Autosomal recessive/autosomal dominant
				USP36	chr17:78813799:C:A	p.Gln413His	Missense	Novel	Yes	VUS	Autosomal recessive
166	33	−5.16	−0.65	BRCA1	chr17:43092196:TCTTG:T	p.Gln1111AsnfsTer5	Frameshift	1.12 × 10−5	Not applicable	P	Autosomal recessive/autosomal dominant
169	35	−4.43	−0.06	MCM8	chr20:5994359:G:A	p.Gly831Ser	Missense	1.37 × 10−4	Yes	VUS	Autosomal recessive/autosomal dominant
				MARF1	chr16:15615842:G:A	p.Pro1081Ser	Missense	Novel	Yes	VUS	Autosomal recessive
				CLPP	chr19:6366325:T:C	p.Ile208Thr	Missense	Novel	Not chicken	VUS	Autosomal recessive
131	38	−4.52	−0.12	FANCA	chr16:89792033:GCCAA:G	p.Val372AlafsTer42	Frameshift	1.16 × 10−4	Not applicable	P	Autosomal recessive/autosomal dominant
				RECQL4	chr8:144513409:G:A	p.Arg758Ter	Stop	2.3 × 10−5	Not applicable	P	Autosomal recessive/autosomal dominant
232	38	−4.68	−0.26	RAD51D	chr17:35101355:TG:T	p.His250ThrfsTer2	Frameshift	2.29 × 10−5	Not applicable	P/LP	Autosomal dominant
				SLX4	chr16:3594575:C:T	p.Asp680Asn	Missense	1.24 × 10−5	Yes	VUS	Autosomal recessive/autosomal dominant
129	40	−4.44	−0.07	BRIP1	chr17:61780325:G:A	p.Ser624Leu	Missense	2.42 × 10−5	Yes	VUS	Autosomal dominant
				MCM8	chr20:5985988:C:T	p.Arg674Trp	Missense	2.11 × 10−5	Yes	VUS	Autosomal recessive/autosomal dominant
170	40	−6.95	−2.09	BLM	chr15:90763016:C:T	p.Gln645Ter	Stop	9.86 × 10−5	Not applicable	P	Autosomal recessive
229	40	−3.24	0.90	EIF4ENIF1	chr22:31463112:G:A	p.Arg203Cys	Missense	3.04 × 10−5	Yes	VUS	Autosomal dominant
				FANCD2	chr3:10094295:C:T	p.Arg1299Cys	Missense	3.35 × 10−5	Yes	VUS	Autosomal recessive

Gene variants shown include those that are likely to be causal along with additional variants in the same subject. The columns indicate age at POI, PGS, PGS SD from the control mean PGS, gene, variant location, protein change, effect on the protein, allele frequency in gnomAD (23), conservation across species, interpretation of ACMG criteria, and expected mode of inheritance.

Abbreviations: ACMG, American College of Medical Genetics; LP, likely pathogenic; P, pathogenic.

^a Conservation—species including rhesus, mouse, dog, elephant, chicken, Xenopus tropicalis, zebrafish, and lamprey.

^b Interpretation using ACMG criteria including P, LP, and VUS (27).

In the women with primary amenorrhea and genetic analysis (n = 8), causal, recessive MCM8 gene variants were identified in the 2 women, with parental DNA confirming biallelic inheritance in one of these women (Table 2). One subject had a 46,XX chrom9 inversion (p12q13), which did not appear to be causal. The only other variant identified was a missense variant in USP36, in which a previous candidate variant caused POI at a later age (Table 2) (2). The other women with primary amenorrhea had only a single-gene variant that would not be expected to be causal (Table 2) or no gene variant of interest (n = 2). Therefore, a total of 27.6% of the total number of women with primary amenorrhea (n = 29) had a known diagnosis (n = 1 galactosemia, n = 1 PMM2, n = 2 MCM8, and n = 4 chromosomal abnormalities).

In women with secondary amenorrhea, several pathogenic variants in genes previously implicated in POI were identified (Table 3). One subject had 2 variants in FANCM, published previously and confirmed biallelic by parental DNA (28). Three subjects with POI at ages 31, 35, and 40 years had a deleterious missense MCM8 variant, along with variants in other candidate genes. Three subjects had single deleterious variants in DNA damage repair genes including frameshift variants in RAD51D (28) and BRCA1, and a stop-gain variant in BLM. The latter 2 women also had polygenic risk scores that would predispose to early age at menopause. One woman had a stop-gain variant in RECQL4 and a FANCA frameshift (28). A deleterious single missense variant was also identified in FANCD2 in 1 woman. One woman had variants in ATM and MSH5, genes also involved in DNA damage and repair pathways.

Polygenic risk scores (PGS) for age at natural menopause were similar in our POI cohort compared with the 1000 Genomes controls (−4.20 ± 1.16 vs −4.36 ± 1.24; P = .23). Some of the subjects had a PGS score 0.5 or more standard deviations below the control mean, −4.98 or lower, as previously found in women in the UK Biobank with menopause between ages 34 and 40 years (Table 3) (9). There was no correlation between the PGS score and age of POI diagnosis (r = −0.083; P = .41). The women with a BRCA1 frameshift and a stop-gain variant in BLM had deleterious PGS <−0.5 standard deviation (SD) below the mean, which may have contributed to the age at POI diagnosis. Samples available from the mother and sisters of the BLM proband indicated a PGS score that could explain some of the risk for POI in the proband and early menopause in the mother (Fig. 3). One additional subject with age at menopause of 35 years had a PGS score of −0.97 SD below the mean.

Figure 3.

Family pedigree showing age at menopause, polygenic risk score, and rare genetic variants. Women (circles) and men (squares) are indicated. Age at menopause, mean (standard deviation), and rare gene variants are indicated below. The proband is indicated in black.

Discussion

In this large, population-based cohort of women with POI from Utah, 32% of cases were categorized as Turner syndrome, with defining X chromosome abnormalities. In the remaining cases of POI, we found that established etiologies accounted for a minority of cases, with known diagnoses identified in 8.1% of women after exclusion of Turner syndrome. Among women without an established etiology who underwent WGS, likely pathogenic variants were identified in 12/68 (17.6%), most frequently in genes involved in DNA repair and meiotic recombination. Autosomal recessive variants were enriched among women with primary amenorrhea, whereas women with secondary amenorrhea more often harbored heterozygous or multiple rare gene variants. In addition, polygenic risk scores for age at natural menopause appeared to contribute to POI in some women.

The largest group of women with POI were those with Turner syndrome (n = 289, 32.1%). The diagnosis of Turner syndrome carries a separate ICD code and can present with phenotypic characteristics and comorbid complications that make the diagnosis more evident (29). Despite the difference in ascertainment, the proportion of women with Turner syndrome we identified is similar to that in previous studies, which ranged from 13% to 30% (30, 31).

The most common diagnoses when Turner syndrome was removed were FMR1 premutations, chromosomal abnormalities and galactosemia. FMR1 premutations cause POI through a gain-of-function mechanism related to increased CGG repeats in the 5′ untranslated region that may sequester proteins important for mRNA processing (32). In our cohort, 2.9% of patients carried a fragile X premutation, which aligns with the 1.8% to 7% prevalence reported in sporadic POI (32, 33). We found lower rates than would be expected for familial cases, with rates of 10% to 15% observed among women with a family history of FMR1 ovarian insufficiency; but only 6.3% of POI cases are familial in our data (8, 32). The lower rates of the FMR1 premutation in our data compared to familial POI and the absence of familial FMR1 premutations may be explained by ascertainment bias in studies of FMR1 families, incomplete or missing documentation and diagnoses, or the use of unrelated coding in the EMR. Additionally, patients without a known family history of POI might not undergo confirmatory testing and referral patterns favoring reproductive endocrinology or endocrinology clinics rather than genetic counseling could result in underrepresentation of FMR1-related cases. Of note, we did not find any missed FMR1 premutations in our subjects undergoing WGS.

Chromosomal abnormalities were identified in 1.8% of our POI cohort, a rate lower than up to 32% reported in the literature (34), likely reflecting our exclusion of patients with Turner syndrome. We found a similar rate of structural alterations in the Xq arm compared to previous studies (0.9 vs 0.4-1.7%) (35-38). Deletions, duplications, inversions, isochromosomes and translocations are known to impair ovarian function, especially when involving critical regions Xq13-21 and Xq23-27 (39). Notably, 2 of our 3 identified translocations involved these critical regions. We had a small number of autosomal abnormalities, including the chromosome 9 pericentric inversion (Inv 9 [p12q13]). Chromosome inversions are seen at a higher rate in women with infertility than in the population, but are not always, associated with infertility (40, 41). We did perform WGS on one of the women with a chromosome 9 pericentric inversion, with no candidate gene variants or copy number variants (CNVs) identified on chromosome 9, although we could have missed a noncoding region at the breakpoint. The only other POI gene variant found in the woman was in USP36, which is known to cause POI only in a recessive form (2). Finally, a chromosome 21 mosaic has been described in a female infertility population, suggesting that it may be the causal anomaly in our subject (42).

After removing women with a known diagnosis, 418 women remained and of those, 68 subjects underwent WGS. Of those, a known genetic variant was identified in two-thirds of cases (n = 42; 61.8%). However, only 17.6% were thought to be causal. Our data are consistent with previous studies suggesting that 12% to 30% of subjects will have a known gene using next-generation sequencing (2, 4, 43, 44). Thus, an etiologic diagnosis could be found in a significant number of women with POI using WGS.

We found autosomal recessive variants associated with POI in women with primary amenorrhea, as we previously demonstrated (2). Two of 8 subjects genotyped each had 2 MCM8 gene variants, which equates to 25% of the women with primary amenorrhea in the subset who underwent genotyping. The pathogenic variants in MCM8 disrupt double-strand DNA repair and meiotic recombination, leading to early depletion of ovarian follicles and primary gonadal failure (45). Previous case reports of MCM8-related POI in consanguineous or familial backgrounds also presented with primary amenorrhea (46-50). The identified MCM8 variants included 3 loss-of-function variants—2 in one woman and 1 in the other—and a rare missense variant located in a highly conserved amino acid in the ATPases Associated with diverse cellular Activities (AAA) core domain (51). While some prior cases have described associated features such as small for gestational age, small stature or benign papular, bluish skin lesions (pilomatricomas) with homozygous stop mutations, these findings were absent in our patients (49).

In addition to the recessive MCM8 variants in women with primary amenorrhea, we identified single pathogenic gene variants in the MCM8 gene. Single copies of protein-truncating MCM8 variants have a clear effect on menopause age with one mother developing POI at age 29 years owing to a single loss-of-function variant, whereas the daughter had primary amenorrhea and POI with 2 copies of the same variant (49). Single missense variants in the AAA+ core domain have been identified as causal for a woman with POI at age 21 years, in the same location in which 2 of our subjects have MCM8 variants (52). Functional studies will be needed to ultimately determine pathogenicity.

In the remaining subjects with primary amenorrhea, only one gene variant was identified, which is unlikely to explain the early menopause. One woman had a missense variant in ATG7, a gene previously identified as an autosomal dominant cause of POI (53). However, both women experienced POI in their early 30s and had missense variants in the ubiquitin E1-like activating enzyme domain (3, 53), while the current variant is not in a critical region. The other women had a single variant in genes causing POI in a recessive manner (54, 55). The final subject with a pericentric inversion on chromosome 9, which would not be expected to cause gonadal dysgenesis, also carried a USP36 missense variant, previously demonstrated to cause POI in a woman aged 39 years, but not primary amenorrhea (2). Further in-depth molecular characterization is underway to identify the underlying genetic cause in this cohort.

Our data did include women with cancer and POI reported previously, validating the potential pathogenicity of a subset of the DNA damage variants identified (28). One previously reported subject carried 2 deleterious variants in the FANCM gene, 1 stop and 1 missense mutation, each inherited from a separate parent (28). A systematic review found that the phenotype of biallelic FANCM variants that included one p.Gln1701Ter and another protein-terminating variant developed POI at age 20-30 and breast cancer at age 43-62 years, similar to our subject, and demonstrated chromosome instability (56). The proposed mechanism involves defective DNA damage repair and impaired meiotic recombination, leading to accelerated follicular atresia (57). The relationship between FANCM and POI underscores the critical role of DNA repair pathways in ovarian reserve maintenance.

Several subjects carried variants in 2 different DNA damage repair genes that do not have direct contact but act in the same pathways. One subject, previously published, carried both a FANCA frameshift mutation and a RECQL4 stop-gain mutation. Similar to FANCM, FANCA encodes a DNA repair protein critical for the resolution of interstrand cross-links (58). The RECQL4 gene encodes a RecQ helicase, an enzyme essential for maintaining genomic stability through its roles in DNA repair, recombination, and replication (59). Both genes have been implicated in syndromic forms of POI and are integral to genome stability during meiosis and DNA repair (60), although no direct interaction between the 2 in POI pathogenesis has been described. Similarly, other genes with variants, including ATM and MSH5; BRIP1 and MCM8; and RAD51D and SLX4, all repair DNA through homologous recombination but do not interact with each other (45, 58, 61-63). Current evidence supports the concept that defects in multiple DNA repair genes can independently contribute to POI (60), highlighting the polygenic nature of ovarian insufficiency. Further research is needed to determine whether these effects are additive or synergistic.

Additional single-gene variants were identified in women with secondary amenorrhea. One subject carried a BRCA1 frameshift mutation. BRCA1 plays a crucial role in homologous recombination and DNA repair within oocytes (64). Loss of its function results in increased double-strand DNA breaks and oocyte apoptosis, potentially leading to premature depletion of the ovarian reserve (64). Some, but not all, studies have reported associations between BRCA1 mutations and diminished ovarian reserve (65-67). Large studies do demonstrate earlier age at menopause and a greater number of women with POI and early menopause in carriers of a BRCA1 variant (68, 69). Important for this subject, her PGS was −0.65 SD below the mean, which has been suggested to contribute to age at menopause before age 40 years (9). Taken together, it may be important to incorporate the PGS into the analysis of POI rare genetic risk to predict age at menopause.

There was a FANCD2 variant found in a woman with POI at age 40 years, with another missense variant in EIF4ENIF1 that could contribute to POI based on its high pathogenicity, but these genes do not operate in the same pathway (58, 70). Further functional analyses will be needed to better characterize the causality of these variants.

Of the POI cohort, 1 woman was of non-European ancestry and was excluded from PGS analyses. This exclusion was performed to minimize population stratification as the PGS model had poor performance in women not of European ancestry. Previous data suggested that the bottom 10th percentile of PGS could explain menopause down to 34 years (9). We found 3 subjects whose PGS might contribute to age at menopause in this age range. In contrast, women with primary amenorrhea in our cohort did not exhibit markedly low PGS values, consistent with the presence of rare, likely high-impact variants causing ovarian insufficiency independent of polygenic risk.

A combination of rare and common variants may explain POI in some cases. One woman with POI at age 35 and a low PGS (−5.56, SD −0.97) had an unaffected mother with a lower PGS (−6.50, SD −1.73). In the familial analysis of the PGS, the proband had menopause at age 40 years, the earliest age at menopause in the family. She had the lowest PGS and an additional stop-gain variant in BLM, which has not been associated with earlier menopause in the heterozygous state and appears to be a founder variant in families in Utah (71). The mother and a sister were found to harbor a variant in NOBOX, which encodes an oocyte-specific transcription factor that is critical for early folliculogenesis and regulation of oocyte-expressed genes (72). Although NOBOX was historically considered to cause autosomal dominant POI, more recent data (73) support autosomal recessive inheritance. The contribution of rare variants and common risk to variability in menopausal timing or POI susceptibility remains uncertain, but common variants in combination with a rare variant may best explain early menopause based on the small number of subjects analyzed herein.

Our population data largely confirm the rates for common causes of POI, which have previously been described in POI cohorts, confirming important information regarding the prevalence of known POI causes for clinical care. In our WGS subset, our findings support a model in which the autosomal recessive variants are associated with primary amenorrhea and POI before puberty, whereas heterozygous or less deleterious variants may predispose to later-onset POI. Furthermore, patients with multiple variants across different DNA repair pathways may have an additive or synergistic risk, suggesting a spectrum of genetic burden. Emerging evidence further supports polygenic contributions to POI. Studies demonstrating possible synergistic effects between variants in 2 or more genes and common variants will be needed to determine whether they explain the wide variance in clinical phenotype for women with POI.

Funding

The work in this publication was supported by the Eunice Kennnedy Shriver National Institute of Child Health and Human Development R01HD099487 (C.K.W.). We also acknowledge partial support for the Utah Population Database through grant P30CA042014 from the National Cancer Institute. The computational resources used were partially funded by the Office of Research Infrastructure Programs, National Institutes of Health shared instrumentation grants S10OD034321 and S10OD021644. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Disclosures

M.Y. is a stock holder or has received stock option awards from Fabric Genomics Inc. B.M. and M.Y. have received consulting fees from Fabric Genomics Inc.

Data availability

The data supporting the current study have not been deposited in a public repository because they require IRB approval and Utah Resource for Genetic and Epidemiology Research approval but are available from the corresponding author upon request.