Improving diagnostic precision in primary ovarian insufficiency using comprehensive genetic and autoantibody testing

STUDY QUESTION

Is it possible to find the cause of primary ovarian insufficiency (POI) in more women by extensive screening?

SUMMARY ANSWER

Adding next generation sequencing techniques including a POI-associated gene panel, extended whole exome sequencing data, as well as specific autoantibody assays to the recommended diagnostic investigations increased the determination of a potential etiological diagnosis of POI from 11% to 41%.

WHAT IS KNOWN ALREADY

POI affects ∼1% of women. Clinical presentations and pathogenic mechanisms are heterogeneous and include genetic, autoimmune, and environmental factors, but the underlying etiology remains unknown in the majority of cases.

STUDY DESIGN, SIZE, DURATION

Prospective cross-sectional study of 100 women with newly diagnosed POI of unknown cause consecutively referred to Haukeland University Hospital, Bergen, Norway, January 2019 to December 2021.

PARTICIPANTS/MATERIALS, SETTING, METHODS

In addition to standard recommended diagnostic investigations including screening for chromosomal anomalies and premutations in the fragile X mental retardation 1 gene (FMR1) we used whole exome sequencing, including targeted analysis of 103 ovarian-related genes, and assays of autoantibodies against steroid cell antigens.

MAIN RESULTS AND THE ROLE OF CHANCE

We identified chromosomal aberrations in 8%, FMR1 premutations in 3%, genetic variants related to POI in 16%, and autoimmune POI in 3%. Furthermore in 11% we identified POI associated genetic Variants of unknown signifcance (VUS). A homozygous pathogenic variant in the ZSWIM7 gene (NM_001042697.2) was found in two women, corroborating this as a novel cause of monogenic POI. No associations between phenotypes and genotypes were found.

LIMITATIONS, REASONS FOR CAUTION

Use of candidate genetic and autoimmune markers limit the possibility to discover new markers. To further investigate the genetic variants, family studies would have been useful. We found a relatively high proportion of genetic variants in women from Africa and lack of genetic diversity in the genomic databases can impact diagnostic accuracy.

WIDER IMPLICATIONS OF THE FINDINGS

Since no specific clinical or biochemical markers predicted the underlying cause of POI discussion of which tests should be part of diagnostic screening in clinical practice remains open. New technology has altered the availability and effectiveness of genetic testing, and cost-effectiveness analyses are required to aid sustainable diagnostics.

STUDY FUNDING/COMPETING INTEREST(S)

The study was supported by grants and fellowships from Stiftelsen Kristian Gerhard Jebsen, the Novonordisk Foundation, the Norwegian Research Council, University of Bergen, and the Regional Health Authorities of Western Norway. The authors declare no conflict of interest.

TRIAL REGISTRATION NUMBER

NCT04082169

primary ovarian insufficiency (POI), premature ovarian insufficiency, next generation sequencing, genetic, autoimmune

Introduction

Primary ovarian insufficiency (POI) affects ∼1% of women and is defined as premature loss of ovarian function due to follicle depletion or dysfunction (Golezar et al., 2019; Panay et al., 2020). The diagnosis is made in women under the age of 40 years with amenorrhea >4 months in association with serum FSH levels in the postmenopausal reference range, measured twice at least 4 weeks apart (Webber et al., 2016; National Institute for Health and Care Excellence: Clinical Guidelines, 2019).

The diagnosis of POI has life-changing consequences, primarily due to reduced fertility and increased risk of complications related to premature estrogen deficiency. Clinical presentations are heterogeneous and involve multiple etiological factors, including genetic, autoimmune, infectious, radiation, and toxin-related. Most striking is the large proportion of idiopathic cases; in 70–90% of women with POI, the underlying cause remains unknown (Webber et al., 2016; Panay et al., 2020).

Ovarian development and function are complex, implicating several genetic pathways. Numerical and structural anomalies on the X chromosome are well-established causes of POI, explaining 10–15% of cases (Jiao et al., 2017; Panay et al., 2020). Premutations in the fragile X mental retardation 1 gene (FMR1) are found in 2–4% of POI women (Murray et al., 2014). Recently, numerous additional genetic variants associated with POI have been identified (Tucker et al., 2016; França and Mendonca, 2020; Yang et al., 2021). A few studies have performed next generation sequencing (NGS) in women with sporadic POI attributing genetic causes to 10–75% of cases depending on the size of the applied gene panel and criteria for causality (Bouilly et al., 2016; Patiño et al., 2017; Jolly et al., 2019; Yang et al., 2019; França et al., 2020; Bestetti et al., 2021; Eskenazi et al., 2021; Rossetti et al., 2021).

Autoimmune disease often coexists with POI, especially hypothyroidism and primary adrenal insufficiency (Addison’s disease) (Silva et al., 2014; Kirshenbaum and Orvieto, 2019). The exact molecular mechanism behind this association is not fully known, but autoimmune oophoritis with an immune infiltrate selectively affecting the theca cell layer of the follicle has been demonstrated (Hoek et al., 1997). A diagnostic ovarian biopsy is not recommended, instead serum autoantibodies against steroidogenic cell enzymes are used as surrogate markers, including autoantibodies against 21-hydroxylase (21OH), side chain cleavage (SCC) enzyme, 17alpha-hydroxylase (17OH), as well as NACHT leucine-rich-repeat protein 5 (NALP5), all of which are highly expressed in the ovaries (Winqvist et al., 1993; Reato et al., 2011; Brozzetti et al., 2015). The reported prevalence of autoimmune POI varies considerably from 0% to 30%, depending on patient selection and assay methods (Bakalov et al., 2005; Silva et al., 2014; Jiao et al., 2017; Kirshenbaum and Orvieto, 2019).

Recommended diagnostic evaluations to assess the cause of spontaneous POI include chromosome analysis and testing for FMR1 premutation, while screening for autoantibodies against thyroid peroxidase (TPO), and 21OH or adrenocortical antibodies can be considered if an immune disorder is suspected (Yeganeh et al., 2019). Currently, extended investigations with chromosomal microarray (CMA) and panel-based NGS, as well as other autoantibody assays, are not included in the routine assessment of POI (Webber et al., 2016; Yeganeh et al., 2019).

The discrepancy in the frequency of etiological diagnosis, and the large proportion of idiopathic POI cases, may reflect variation in study populations or insufficient investigation tools. Therefore, it is essential to improve diagnostic precision to predict outcomes and tailor future treatments for women with POI.

Here, we aimed to identify the etiology of as many POI women as possible by careful clinical phenotyping and use of the latest diagnostic tools including an NGS gene panel of 103 POI-related genes and specific autoantibody assays, with the ultimate aim to provide personalized diagnosis and follow-up.

Materials and methods

Patients

Women with newly diagnosed POI of unknown cause referred for evaluation to the Endocrinology or Gynecology outpatient clinics at Haukeland University Hospital, Bergen, Norway, between January 2019 and December 2021, were recruited and systemically evaluated in a prospective cross-sectional trial. Women <40 years were included if they had primary amenorrhea or secondary amenorrhea for at least 4 months, and serum FSH in the menopausal range (>21 IU/l), on at least two occasions more than one month apart. Patients who were <16 years or who had previously undergone gonadotoxic treatment (chemotherapy or radiation) or ovarian surgery were excluded from participation. Of the 105 consecutive patients identified with spontaneous POI, five patients did not meet the study eligibility criteria. Written informed consent was obtained from all participants. The study was approved by the Ethics Committee 2018/1206/REK Midt.

Clinical assessment

Clinical, reproductive, and familial data as well as anthropometrics (weight in kilograms (kg), height in centimeters (cm), blood pressure (mmHg)) were recorded by the same examiner by standard recommended methods (Webber et al., 2016; Panay et al., 2020).

Bone mineral density assessment

Bone mineral density (BMD) was assessed using a Dual-Energy X-ray Absorptiometry (DXA) (General Electric Lunar iDXA). BMD results from the femoral neck and lumbar spine (L₂–L₄) were expressed as Z and T scores. Low bone mass was defined as a Z-score < −2 SD below the age-adjusted mean and osteoporosis was defined as a T-score < −2.5 SD (Kiriakova et al., 2019; Panay et al., 2020).

Laboratory assessment

FSH and LH, prolactin, adrenocorticotropin (ACTH), thyroid-stimulating hormone, free thyroxine (fT4), vitamin D, sex hormone-binding globulin, and anti-Müllerian hormone (AMH) were analyzed using electro-/chemiluminescent immune assays and the steroid hormones (progesterone, testosterone, androstenedione, dehydroepiandrosterone sulfate (DHEA-S), and cortisol 17β-estradiol) were measured using liquid chromatography-tandem mass spectrometry (LC-MS/MS (Hormone laboratory, Haukeland University Hospital, Bergen, Norway)).

Genetic investigations

The genetic evaluation included chromosomal karyotyping and CMA, FMR1 analysis, and an NGS approach based on clinical whole exome sequencing (WES) (Fig. 1).

Figure 1.

Flow chart of diagnostic workup in primary ovarian insufficiency. POI, primary ovarian insufficiency; NGS, next generation sequencing; WES, whole exome sequencing; TPO, thyroxide peroxidase; 21OH, 21-hydroxylase; 17OH, 17-hydroxylase; SCC, side-chain-cleavage enzyme; NALP5, NACHT leucine-rich-repeat protein 5.

Karyotyping

G-banded chromosome analysis was performed according to standard protocols. At least 10 metaphases were screened for structural and numerical anomalies. In individuals with suspected mosaicism for structural anomalies on the X chromosome (monosomy X in 1/30 single metaphases), additional karyotyping was performed using cultured skin fibroblasts (Barch et al., 1997).

Chromosomal microarray

Submicroscopic genomic copy number variations (CNVs), such as deletions and duplications, and long continuous stretches of homozygosity (LCSH) were investigated by the CytoscanHD array (Thermo Fisher Scientific) following the supplier’s protocol and analysis software.

FMR1 premutation testing

CGG repeat numbers in the 5′ untranslated region of the FMR1 gene were determined by the AmplideX^®FMR1 PCR kit from Asuragen, as recommended by the manufacturer. Alleles containing between 55 and 200 CGG repeats were defined as premutations.

NGS and variant interpretation

We used a two-tiered NGS-based strategy to identify genetic variants causative of POI. First, libraries for WES were prepared from genomic DNA and gene panel-based filtration of variants identified by WES were performed using the expert-curated POI (Version 1.60) panel from Genomics England PanelApp, and an in-house designed panel for general hypogonadism. In total, 103 genes were included based on ovarian biological studies and were categorized ontologically as follows: ovarian development, meiosis and DNA repair, follicle function, metabolism and immune regulation, hypothalamic–pituitary–ovarian axis (Martin et al., 2019; Rossetti et al., 2021; Ruth et al., 2021).

Secondly, since LCSH regions are associated with increased risk of recessive disorders, in women with considerable genomic regions of LCSH (defined as at least two LCSH regions ≥5Mb on two or more chromosomes), we searched for potentially causal variants in POI candidate genes among the entire protein-encoding exome (Fan et al., 2021).

Libraries for WES were prepared using SEQCap EZ HyperCap (v1.2) or KAPA HyperCap (v3.0) library preparation kits, with SeqCap EZ MedExome- or KAPA HyperExome target enrichment kits, respectively (all kits and reagents from Roche). Libraries were sequenced on Illumina NextSeq500 or Illumina NovaSeq 6000 instruments using paired-end 150 bp or 100 bp reads, respectively. Data processing, alignment to GRCh37 and variant calling were performed essentially as described (Bredrup et al., 2021), except that GATK v.3.8.1 was used according to GATK’s Best Practices guidelines (McKenna et al., 2010; Van der Auwera et al., 2013). Variant annotation and interpretation were performed using Alissa Interpret (Agilent Technologies) and Alamut Visual 2.15 (Sophia Genetics). As this study was performed in a clinical setting, only rare variants were assessed; defined as variants with allele frequencies ≤0.1% (autosomal dominant model) or ≤2.0% (autosomal recessive model) in the public database gnomAD (versions v2.1.1 and v3.1.2).

Formal clinical classification of the identified variants was based on the American College of Medical Genetics and Genomics (ACMG) guidelines (Richards et al., 2015). Based on 28 criteria, these guidelines classify variants according to a 5-tier system: Class 5 (Pathogenic), Class 4 (Likely Pathogenic), Class 3 (Variants of Unknown Significance (VUS)), Class 2 (Likely Benign), Class 1 (Benign). However, not all 28 classification criteria were available in the current study (e.g. family data to assess co-segregation of variants and disease). Furthermore, the ACMG guidelines are primarily intended for classifying highly penetrant variants in patients with rare, monogenic diseases. Modified disease-specific guidelines have been developed, but not yet for POI (Loong et al., 2022). Hence, most of the variants identified in the current study were placed in the VUS category. Based on the strength of available evidence, and according to guidelines published from the Association of Clinical Genomic Science in the UK (https://www.acgs.uk.com/media/11631/uk-practice-guidelines-for-variant-classification-v4-01-2020.pdf), we differentiate between hot/warm VUS’es and cool/cold/ice-cold VUS’es. Hot/warm VUS’es are variants with a high level of supporting evidence of causality and where additional evidence might be obtained to allow re-classification as likely pathogenic. In the following a hot/warm VUS will be referred to as possibly pathogenic or VUS+, to discriminate from cold VUS’es less likely to be pathogenic. Summaries of all variant interpretations are also included in Supplementary Data File S1.

Autoimmune antibody assays

Assays for autoantibodies against 21OH, SCC, 17OH, and NALP-5 were performed by radio-binding ligand assays as described previously (Oftedal et al., 2008). Positive cut-offs were calculated using positive and negative controls with index thresholds of 2 SD above the mean of healthy subjects (>57, >200, >102, and >65 for 21OH, SCC, 17OH, and NALP-5, respectively) (Falorni et al., 2015). Autoantibodies against TPO were analyzed by electrochemiluminescence immunoassay (Roche Cobas). Individuals with positive 21OH and SCC autoantibodies were classified as autoimmune POI.

Statistical analysis

Continuous data in normality distribution were expressed as mean and SD; otherwise, data were presented as median within the quartile range [IQR]. Categorical data was expressed in percentage and compared by Chi-square test with Yates Continuity Correction.

Correlations between continuous variables such as age at menarche, timing of POI, years since amenorrhea, hormone levels and BMD, were investigated using linear regression analysis and Spearman rank correlation.

Differences between groups related to timing POI, family history of POI, fertility, symptoms of estrogen deficiency and POI etiology, were investigated using Student’s t-test, one-way analysis of variance or Mann–Whitney U test.

Results

Population characteristics

One hundred women with POI of unknown cause were included in the study. Demographic, clinical, and reproductive characteristics are presented in Table 1. Fifteen women (15%) had primary amenorrhea, and 85% had secondary amenorrhea with a median age of 33 [29–37] years (Fig. 2). No significant association was found between age at menarche and timing of POI diagnosis.

Figure 2.

Age at primary ovarian insufficiency. Primary amenorrhea: 15%. Secondary amenorrhea: 85%.

Table 1.

Clinical characteristics of women with premature ovarian insufficiency (POI).

	(n 100)
Age at menarche (years)	13 [12–14]
Primary amenorrhea (%)	15
Age secondary amenorrhea (years)	32 [23–37]
<20 (%)	4
20–29 (%)	20
30–39 (%)	61
Family history of POI1 (%)	20
Ever pregnant2 (%)	51
Age first pregnancy (years) (n = 51)	26 [24–30]
Age last pregnancy (years) (n = 32)	32 [30–34]
In a relationship/partner (%)	64
BMI (Kg/m2)	24.2 [21.7–27.1]
Height (cm)	167 (162–170)
Smoke (%)	7
Ethnicity
European (%)	75
African (%)	17
Asian (%)	8
Other disease
Osteoporosis (n = 76) (%)	8
Osteopenia (n = 76) (%)	10
Hypothyroidism (%)	17
Type 1 diabetes (%)	4
Multiple sclerosis (%)	5
Laboratory3
FSH IU/l (<214)	70.4 [50.1–104.0]
17β-estradiol pmol/l (74–19404)	43.0 [13.5–85.0]

Median [Quartile 1–Quartile 3] or frequency/percent (%).

1

Mother and/or sister with POI.

2

Egg donation in three women.

3

Before starting hormone replacement therapy (HRT).

4

Normal range.

A family history of POI among first-degree relatives was reported in 20% (mothers 13%, sisters 10%, mother and sister 3%), including two sisters in the study. In addition, 17% reported second-degree relatives with POI or early menopause (grandmothers, aunts, or cousins). Women with a family history of POI developed POI 3 years later than those without (P = 0.019).

Half of the women (51%) had been pregnant, and 33% were multiparous. Three women had given birth after egg donations. Previous pregnancy was associated with POI diagnosis 5 years later compared to those who had never been pregnant (P < 0.03).

Most women reported symptoms of estrogen deficiency (81%) with a great degree of overlap between symptoms (Fig. 3). Vasomotor symptoms were more common in women with secondary amenorrhea than primary amenorrhea (71% versus 40%). Symptoms were not related to time since amenorrhea, use of hormone replacement therapy (HRT) or levels of 17β-estradiol, However, there was a border-line association between higher FSH levels and vasomotor symptoms (P = 0.057).

Figure 3.

Reported symptoms in women with primary ovarian insufficiency. Venn diagram showing overlap of symptoms.

DXA scans were available from 76 women. Nearly one in four women (23%) had reduced bone mass. Low bone mass (osteopenia) was found in 12% and osteoporosis in 11%. Women with primary amenorrhea (n 12) had lower BMD compared with those with secondary amenorrhea (n 64) (lumbar spine Z score −1.50 (−3.90 to −0.20) versus −0.50 (−4.5 to 3.20), P = 0.023, and T score −3.0 (−4.7 to 3.2) versus −1.35 (−3.90 to 0.0), P = 0.014, respectively). Younger age at POI diagnosis was associated with lower BMD (Spearman Rank Order r = −0.300, P = 0.012), also after adjusting for duration of POI, BMI, and HRT use (P < 0.01).

Women who had not yet started HRT (45%) exhibited typical postmenopausal endocrine profiles with FSH levels 70.4 [50.1–104.0] IU/l (premenopausal reference <21 IU/l) and 17β-estradiol 43.0 [13.5–85] pmol/l (premenopausal reference 74–1940 pmol/l). S-AMH was not detectable or below the age-specific range (<1.0 pmol/l) in 93% of women, and in the lower range (median 2.3, minimum 1.1 to maximum 10.3 pmol/l) in 7%. S-AMH values did not correlate with age, previous pregnancy status, family history of POI, or other hormone values.

Etiology

Chromosome analysis

Chromosomal anomalies were found in eight women (8%) (Table 2 and Fig. 4). Standard chromosome karyotyping analysis revealed X chromosome anomalies in five women (5%). Of these, three involved deletions and/or duplications, while two women had numerical anomalies (one had low-grade mosaicism for monosomy X, and one had trisomy X (47, XXX)).

Figure 4.

Primary ovarian insufficiency associated gene panel and ovarian function. Genetic variants found and classified: Red: pathogenic, likely or possibly pathogenic variant. Blue: variants of unknown significance (VUS). Black: no variants found in these genes. The figure was generated using Servier Medical Art.

Table 2.

Chromosomal anomalies detected by karyotyping and chromosomal microarray.

ID	Karyotype (ISCN)	Chromosomal microarray	Phenotype
			POI age (years)	Menarche age (years)	Fertility	Ethnicity	Height (cm)	Familial POI	Additional phenotype
P25	46, X, del(X)(q21.2)	arr[GRCh37] Xq21.2q28(85917017_154542526)x1	25	13	G1/P1#	European	166	–	Hypothyroidism Vitiligo
P34	46, X, der(X)(pter-q26.3::p11.3-pter)	arr[GRCh37] Xp22.33p11.3(287730_44199601)x3, arr[GRCh37] Xq26.3q28(136980232_155233731)x1	33	12	G0	African	165	–	Dysmorphic (asymmetric) facial features
P39	46, X, del(X)(q21)	Not performed	30	12	G0	European	150	–	Vitiligo
P41	46, XX [26], 47, XXX [2], 45, X[2]	Not performed	31	11	G0	Asian	158	–	Hypothyroidism
P69	46, XX	arr[GRCh37] 22q13.2(42154189_42578609)x3	37	12	G2/P2	European	165	Sister	–
P75	46, XX	arr[GRCh37] 8p23.3p23.2(2179649_3142724)x3	30	14	G0	African	162	–	Osteoporosis
P93	46, XX	arr[GRCh37] 8p12(28933652_30847526)x1	37	16	G4/P4	European	160	–	–
P96	47, XXX	arr[GRCh37] Xp22.33q28(168547_155233731)x3	29	12	G0	European	164	–	Hypogammaglobulinemia. Alopecia. Nail dystrophy. WPW syndrome. ADHD.

#

Pregnant after egg donation.

ADHD, attention deficit hyperactive disorder; G, Gravida; ISCN, International System for Human Cytogenetic Nomenclature; P, Para; WPW syndrome, Wolf–Parkinson–White syndrome.

See Supplementary Data File S1 for interpretation of findings.

CMA mapped the breakpoints of the larger structural X chromosome rearrangements (Table 2). Furthermore, CMA detected smaller copy number variants in three women with normal karyotypes: a 1.9 Mb deletion on Chromosome 8 involving several POI candidate genes (KIF13B, LEPROTL1, RBPMS, and TEX15), an intragenic duplication on 8p23.3p23.2 involving Exons 27–70 of the gene CSMD1 (involved in ovarian development), and an intergenic duplication on 22q13.2 involving MEI1 (involved in meiosis) and 13 other protein-coding genes (Supplementary Data File S1).

In 14 women, CMA revealed at least two regions of homozygosity on different chromosomes (Supplementary Data File S1). As LCSH regions are associated with increased risk of recessive disorders, we searched for rare homozygous variants in novel POI candidate genes across the entire protein-coding exome in these women (see below).

Skin biopsy and chromosome karyotyping in cultured fibroblasts was done on suspicion of mosaicism (45, X) from karyotyping of blood in four women, but were all normal.

FMR1 premutation analysis

FMR1 premutations were detected in three women (3%) (Table 3). Two had a family history of POI on the paternal side of the family, indicating paternal inheritance of the FMR1 premutation. There were no family members with known fragile X-syndrome in any of the three families.

Table 3.

FMR1 premutations.

ID		Phenotype
	CGG repeat number	POI age (years)	Menarche age (years)	Fertility	Ethnicity	Height (cm)	Familial POI	Additional phenotype
P1	30/85	29	11	G0	European	170	Pat.gr.mother	–
P63	26/78	32	12	G1/P1	European	163	–	–
P74	23/78	28	12	G0	European	167	Pat.gr.mother	–

POI, premature ovarian insufficiency; FMR1, Fragile X mental retardation 1 gene; G, Gravida; P, para; Pat. gr.mother, paternal grandmother.

Gene panel-based NGS and extended WES analysis

Altogether, panel-based NGS and WES revealed variants in POI associated genes in 38% of the women (Fig. 5).

Figure 5.

Etiology in women with primary ovarian insufficiency.

In 16 women (16%), we identified genetic variants classified as pathogenic and likely or possibly pathogenic (the latter also referred to as warm/hot VUS’es). These included variants in genes associated with ovarian development (EIF4ENIF1, NANOS3, SOHLH2, SOX8), or meiosis and DNA repair (BUB1B, MCM8, MCM9) (Table 4). Furthermore, we identified two patients with ultra-rare homozygous variants in MTOR and SMC3, two genes not previously associated with POI, but with highly important roles in ovarian development and meiosis and DNA repair, respectively. In one patient, we detected a rare missense variant in the TP63 gene, encoding the transcriptional regulator p63 that is critical for maintenance of the oocyte genome integrity. In two female relatives with primary amenorrhea, we found a two base pair deletion in the ZSWIM7 gene (NM_001042697.2: c.231_232del, p.(Cys78Phefs*21)), introducing a frameshift and premature stop codon expected to lead to complete loss-of-function.

Table 4.

Primary ovarian insufficiency (POI) associated genetic variants detected by panel-based next generation sequencing (NGS) and extended whole exome sequencing (WES) analysis.

ID	Gene (REFSEQ)	cDNA	Protein	Zygosity	ACMG class	Phenotype
						POI age (years)	Menarche age (years)	Fertility	Ethnicity	Height (cm)	Familial POI	Additional phenotype
Pathogenic, likely or possibly pathogenic genetic variants
P5	SOX8(NM_014587.4)	c.1163C>T	p.(Ala388Val)	Het	VUS+	PA	–	G0	European	170	–	No pubarche Osteoporosis
P12	SOHLH2(NM_017826.3)	c.298_301del	p.(Val100LeufsTer2)	Het	VUS+	31	15	G1/P1	African	173	–	–
P33	MCM9(NM_017696.2)	c.3031C>A	p.(Pro1011Thr)	Het	VUS+	32	13	G3/P1	African	172	Sister	Sickle cell anemia
P35	NANOS3(NM_001098622.2)	c.292G>A	p.(Glu98Lys)	Het	VUS+	38	14	G1/SpA1	European	167	Aunt	–
P37	SOX8(NM_014587.4)	c.968C>T	p.(Ser323Leu)	Het	VUS+	33	15	G1/SpA1	African	155	–	Osteopenia
	SOX8(NM_014587.4)	c.1064C>T	p.(Ser355Leu)	Het	VUS+
P50	MTOR(NM_004958.4)	c.259A>G	p.(Ile87Val)	Hom	UC	PA	–	G0	Asian	176	–	Migraine
P51	MCM8(NM_032485.5)	c.832C>T	p.(Arg278Cys)	Hom	VUS+	36	15	G1/SpA1	Asian	163	–	–
P62	BUB1B(NM_001211.5)	c.2995C>T	p.(Arg999Trp)	Het	VUS+	35	11	G0	European	162	Mother	–
P71*	ZSWIM7(NM_001042697.2)	c.231_232del	p.(Cys78PhefsTer28)	Hom	P	PA	–					–
P72*	ZSWIM7(NM_001042697.2)	c.231_232del	p.(Cys78PhefsTer28)	Hom	P	PA	–
P78	SMC3(NM_005445.4)	c.969 + 4A>C	p.?	Hom	UC	35	13	G0	Asian	154	–	–
P85	EIF4ENIF1(NM_019843.4)	c.5A>G	p.(Asp2Gly)	Het	VUS+	33	11	G5/P3/ProvAb2	African	165	–	–
P86	BUB1B(NM_001211.5)	c.3029C>T	p.(Ser1010Phe)	Het	VUS+	34	15	G4/P3/SpA1	African	166	–	Epilepsy
P90	EIF4ENIF1(NM_019843.4)	c.490C>T	p.(Arg164Trp)	Het	VUS+	23	11	G1/P1	European	166	–	–
P97	TP63(NM_003722.5)	c.290G>T	p.(Arg97Leu)	Het	LP	37	14	G1/P1	European	184		–
P102	SOX8(NM_014587.4)	c.443A>G	p.(Lys148Arg)	Het	VUS+	35	14	G4/P2/SpA2	European	160	Cousin	Osteopenia
Monoallelic variants in genes associated with autosomal recessive inheritance
P9	SOHLH1(NM_001012415.2)	c.743C>T	p.(Ser248Leu)	Het	VUS	34	13	G3/P2/SpA1	European	160	Sister	–
P21	AARS2(NM_020745.3)	c.1534G>C	p.(Asp512His)	Het	VUS	39	11	G3/P3	European	167	–	–
P22	STAG3(NM_001282716.1)	c.2351G>C	p.(Cys784Ser)	Het	VUS+	35	13	G1/P1	European	172	Pat.gr. mother	Pollen allergy
P24	FSHR(NM_000145.3)	c.1330G>A	p.(Ala444Thr)	Het	VUS	21	12	G0	European	162	Pat.gr. mother	Osteoporosis Vitiligo
P26	STAG3(NM_001282716.1)	c.3305_3306insGCC	p.Ile1102delinsMetPro	Het	VUS	38	12	G0	European	169	–	Hypothyroidism
P27	FSHR(NM_000145.3)	c.910A>G	p.(Met304Val)	Het	VUS	23	12	G0	African	167	–	–
	PSMC3IP(NM_013290.6)	c.13C>G	p.(Arg5Gly)	Het	VUS
P30	C14ORF39(NM_174978.3)	c.748A>G	p.(Arg250Gly)	Het	VUS	39	12	G3/P3	European	172	–	Type 1 diabetes
P45	STAG3(NM_001282716.1)	c.3515 + 5G>A	p.?	Het	P	30	14	G3/P1/SpA2	European	168	–	Psoriasis
P47	C14ORF39(NM_174978.3)	c.748A>G	p.(Arg250Gly)	Het	VUS	23	12	G2/P1/ProvA1#	European	180	–	–
P62	SGO2(NM_001160033.1)	c.217C>T	p.(Arg73Ter)	Het	P	35	11	G0	European	162	Mother	–
P106	HFM1(NM_001017975.4)	c.2449A>G	p.(Lys817Glu)	Het	VUS	15	12	G0	European	180	–	–
Incidental carrier findings
P4	LMNA(NM_170707.2)	c.1487C>T	p.(Thr496Met)	Het	VUS	32	9	G1/P1	African	164	–	–
P14	EIF2B5(NM_003907.3)	c.190C>G	p.(Pro64Ala)	Het	VUS+	33	13	G1/ProvA1	African	161	–	–
P36	EIF2B4(NM_001318966.1)	c.29G>T	p.(Gly10Val)	Het	VUS	18	16	G0	African	166	–	Osteopenia
P46	POLG(NM_002693.3)	c.1639G>T	p.(Ala547Ser)	Het	VUS	38	17	G2/P2	European	172	–	Epilepsy
P62	GALT(NM_000155.4)	c.563A>G	p.(Gln188Arg)	Het	P	35	11	G0	European	162	Mother	–
P64	HARS2(NM_012208.4)	c.591T>G	p.(Cys197Trp)	Het	VUS	37	12	G0	European	153	–	MS. Osteopenia. Hypothyroidism
P66	PMM2(NM_000303.3)	c.422G>A	p.(Arg141His)	Het	P	PA	–	G0	European	177	–	–
P75	AIRE(NM_000383.2)	c.1367G>A	p.(Arg456His)	Het	VUS	30	14	G0/P0	African	162	–	Osteoporosis
P94	AIRE(NM_000383.2)	c.769C>T	p.(Arg257Ter)	Het	P	37	12	G2/P2	European	168	–	Ulcerative colitis
P99	GALT(NM_000155.4)	c.197C>T	p.(Pro66Leu)	Het	VUS	31	13	G1/P1	European	159	–	Hypothyroidism
P104	POLG(NM_002693.3)	c.2243G>C	p.(Trp748Ser)	Het	P	38	12	G5/P2/SpA2/ProvA1	European	160	Sister	Hypothyroidism

#

Pregnant after egg donation.

*

P71 and P72 did not consent to publishing ethnic, anthropometric, family or clinical information.

ACMG, American College of Medical Genetics and Genomics; G, Gravida; Het, heterozygous; Hom, homozygous; P, pathogenic; MS, multiple sclerosis; P, para; Pat. gr.mother, paternal grandmother; PA, primary amenorrhea; ProvA, provoked abortion; VUS+, possibly pathogenic; REFSEQ, reference sequence gene annotation; SpA, spontaneous miscarriage; UC, unclassified; VUS, variants of unknown significance.

See Supplementary Data File S1 for interpretation of findings.

Furthermore, in 11 (11%) other women we identified monoallelic variants in genes primarily associated with autosomal recessive inheritance (Table 4). These genes were also involved in ovarian development (SOHLH1), or meiosis and DNA repair (C14ORF39, HFM1, PSMC3IP, SGO2, STAG3), but variants in genes involved in follicle function (FSHR), and metabolism and immune regulation (AARS2) were identified as well. Among the detected variants were both hot and cold VUS’es and likely pathogenic variants. Since the presence of POI in these women may be caused by these variants along with an unidentified gene variant on the other allele, we searched extensively for potential second variants in these genes in the NGS and CMA raw data, without finding any.

Finally, in 11 (11%) additional women we also detected monoallelic variants in genes associated with autosomal recessive disease not consistent with the clinical phenotype (Table 4). We therefore consider these variants incidental potential carrier findings.

Autoantibodies

A total of 23% of women presented with known autoimmune disorders, including hypothyroidism (17%), Type 1 diabetes (T1D, 4%), and multiple sclerosis (MS, 5%). Two women had MS and hypothyroidism, and one woman had MS and T1D.

In total, 3% of POI individuals were considered to have a likely autoimmune etiology based on the presence of autoantibodies against 21OH and SCC (Table 5). All had AMH levels above the detection threshold (AMH, 1.0, 0.2, and 10.3 pmol/l, respectively). None had clinical characteristics or hormone levels consistent with Addison’s disease, as all had normal ACTH and cortisol levels.

Table 5.

Autoimmune premature ovarian insufficiency (POI).

Patient ID	Autoantibodies (index titer)				Phenotype
	21OH (Pos > 57)	SCC (Pos > 200)	17OH (Pos > 102)	TPO (Pos > 33)	POI age (years)	Menarche age (years)	Fertility	Ethnicity	Height (cm)	Familial POI	Additional phenotype
P17	211	433	111	111	39	13	G1/P1	European	162	–	Multiple Sclerosis. Hypothyroidism.
P80	135	458	167	29	14	12	G0	European	154	–	Osteoporosis
P99	809	765	88	37	31	13	G1/P1	European	159	–	Hypothyroidism. Cystic ovaries.

21OH, 21-hydroxylase; 17OH, 17-hydroxylase; SCC, side-chain-cleavage enzyme; TPO, thyroxide peroxidase.

Apart from the patients with autoantibodies described above, isolated autoantibodies against 17OH were found in three and NALP5 were found in four additional women without other biochemical or clinical markers of autoimmunity. Positive TPO autoantibodies were detected in 21% of all the women.

Phenotype–genotype correlations

No significant between-group differences related to the cause of POI were found regarding the timing of menarche or POI debut, frequency of complications and symptoms of estrogen deprivation, family history, or hormone levels. Women with X chromosomal anomalies were slightly shorter than idiopathic POI (height: 161 cm versus 168 cm, P = 0.042). More POI women with sub-Saharan African ethnicity had a POI associated genetic variant compared to POI women with European ethnicity (59% versus 33%, P = 0.043). Women with autoimmune POI experienced a trend toward later amenorrhea compared to non-autoimmune POI (median 34 versus 32 years). AMH levels were also higher in these women (2.6 versus 0.1 pmol/l).

Discussion

This extensive clinical study identified a likely underlying cause of POI in approximately one-third of the women. Adding a POI-related NGS gene panel and exome sequencing to the recommended diagnostic investigations increased the frequency of an etiological diagnosis. No specific clinical or biochemical markers predicted the underlying cause of POI, actualizing the discussion of which tests should be included as diagnostic screening in clinical practice.

We found X-chromosomal anomalies in the lower range of what has previously been reported (Panay et al., 2020). As chromosomal anomalies may result in more extreme phenotypes, including syndromic features, short stature, and primary amenorrhea, the frequency in our study is likely underestimated as many of these individuals are diagnosed already in childhood (Jiao et al., 2017). Karyotyping is suitable for detecting large chromosomal aberrations, such as deletions, duplications, balanced, and unbalanced rearrangements. Here, we identified structural anomalies that included Xq13–Xq21 and Xq23–27, underlining the importance of this region for the POI phenotype (Rizzolio et al., 2006). Apart from confirming the findings on the X-chromosome and estimating the breakpoints, CMA revealed three potentially clinically relevant CNVs involving genes involved in ovarian biology and meiosis and DNA repair. The 1.9 Mb deletion on Chromosome 8 was intriguing, as it contained at least four genes implicated in different ovarian processes: KIF13B, LEPROTL1, RBPMS, and TEX15. RBPMS encodes a transcriptional regulator demonstrated to be important for oocyte development (Aguero et al., 2016; Kaufman et al., 2018). Based on the finding of this deletion on Chromosome 8, we therefore suggest RBPMS as a novel POI-associated gene. Others have also found CMA useful in identifying candidate POI genes (Tšuiko et al., 2016; Bestetti et al., 2021). In our hands, CMA was valuable in mapping LCSH regions, and thus for selecting patients for whole exome analysis. Still, chromosomal karyotyping during the diagnostic workup of POI was the most useful test for detecting chromosomal anomalies or mosaicism.

The prevalence of FMR1 premutation in our cohort (3%) was in accordance with previous reports (Murray et al., 2014). As none of the three families had family members with known fragile X syndrome, a negative family history of fragile X should not be used to guide who will be offered FMR1 testing or not. POI occurs in 20% of FMR1 premutation allele carriers, and the risk of developing POI increases at CGG repeat numbers above 80. In addition to explaining POI, these females are at high risk of giving birth to children with fragile X syndrome, and prenatal diagnostics and preimplantation genetic testing should be discussed. Screening for FMR1 is therefore considered both useful and important in POI (Hunter et al. 1993–2023; Monaghan et al., 2013).

Using an NGS POI-related panel and additional WES, we found that 27% had at least one variant that may be involved in POI. Hundreds of genes have been suggested to be associated with POI based on their involvement in processes related to different aspects of ovarian function (França and Mendonca, 2022). In our study, it is striking that the vast majority of genetic variants were found in genes involved in ovarian and follicular development, and in meiosis and DNA repair. This is consistent with other recent studies employing NGS-based analyses to identify genetic causes of POI (Bestetti et al., 2021; Huang et al., 2021; Heddar et al., 2022).

We identified variants in EIF4ENIF1, NANOS3, SOHLH2, and SOX8 encoding transcription factors, transcriptional regulators, and RNA-binding proteins playing key roles in ovarian/gonadal development (Panula et al., 2016; Shin et al., 2017; Portnoi et al., 2018; Shang et al., 2022).

Further, we also found variants in BUB1B, MCM8, MCM9, MTOR, SMC3, TP63, and ZSWIM7; all involved in chromosomal stability, homologous recombination during meiosis, and repair of DNA breaks (Park et al., 2013; Touati et al., 2015; Prakash et al., 2021). The TP63 gene encodes a transcriptional regulator p63 that is critical for maintenance of the oocyte genome integrity (Suh et al., 2006). Pathogenic variants in the TP63 gene have lately been established as a cause of isolated POI, through a gain-of-function mechanism (Tucker et al., 2022; Huang et al., 2023). Recently, pathogenic variants in the ZSWIM7 gene have been reported in both males and females with infertility (Hussain et al., 2022; McGlacken-Byrne et al., 2022). ZSWIM7 encodes Zinc finger SWIM domain-containing protein 7 (ZSWIM7), also known as SWIM domain-containing SRS2-interacting protein 1 (SWS1), which is critical for meiotic homologous recombination (Abreu et al., 2018). Both female and male ZSWIM7 knockout mice are infertile due to impaired meiotic DNA recombination (Li et al., 2021). The same deletion that we report in two relatives with POI was initially described in unrelated male patients with azoospermia but has also been seen in a Turkish woman with POI (Alhathal et al., 2020; Li et al., 2021; Yatsenko et al., 2022).

In pace with increasing knowledge of the underlying molecular mechanisms of POI, custom-built gene panels and variant classification must be dynamic. In particular, we suggest that ZSWIM7 should be included into POI-related gene panels based on our findings of a homozygous loss-of-function variant in ZSWIM7 in two relatives with primary amenorrhea.

However, it can be challenging to establish definite causality, as family investigations and functional characterizations are needed to corroborate the pathogenicity of variants. The inheritance pattern can be autosomal dominant or recessive. In 11 women, we detected monoallelic variants (ranging from VUS’s to pathogenic) in genes associated with autosomal recessive POI. No second variant in any of those genes could be detected, but cannot be ruled out, and should be followed up in the future with more specialized methods such as whole genome- and/or long-range sequencing. Recently, there has also been increased focus on oligogenic inheritance with possible synergistic effects of variants in several genes explaining the variance in the observed phenotype (Rossetti et al., 2021). Thus, precise classifying of the genetic variants is complex.

The prevalence of autoimmune POI in our study (3%) is in the lower range of what has earlier been reported, possibly because we used more specific autoantibody assays than earlier studies (Novosad et al., 2003; Jiao et al., 2017; Kirshenbaum and Orvieto, 2019; Panay et al., 2020; Vogt et al., 2022). All three women had autoantibodies against both 21OH and SCC, and both therefore appear to be the markers with the highest diagnostic sensitivity for autoimmune POI (Winqvist et al., 1995; Hoek et al., 1997; Bakalov et al., 2005; Falorni et al., 2012). Autoantibodies against 17OH and NALP5 have previously been shown to be present in women with autoimmune POI, but we could not replicate these findings (Dal Pra et al., 2003; Brozzetti et al., 2015). Although autoantibodies against TPO were prevalent, these are common among postmenopausal women in the background population (15%) and therefore not useful as markers of ovary-specific autoimmunity (McLeod and Cooper, 2012).

After thyroid disease, adrenal autoimmunity is the most common autoimmune condition associated with POI (Kirshenbaum and Orvieto, 2019; Panay et al., 2020). While ∼10% of women with Addison’s disease have POI, 2–3% of women with POI develop adrenal autoimmunity (Bakalov et al., 2005; Reato et al., 2011; Kirshenbaum and Orvieto, 2019). Even though none of the women with autoantibodies against 21OH and SCC reported here had Addison’s disease, they are at risk of developing adrenal autoimmunity and must be followed up with functional adrenal testing (Webber et al., 2016; Vogt et al., 2021).

Some limitations apply to our study. First, there will always be a question of whether this cohort of POI women is representative. To our knowledge, we included all women with suspected POI who were referred to our hospital in the study period. It is, however, possible that women with extreme phenotypes were not included because they had received the POI diagnosis in childhood. The use of candidate genetic and autoimmune markers allows us only to identify already known POI markers of interest, limiting the possibility to discover new markers. The number of genes used in our NGS panel is therefore limited by current knowledge of genes of clinical relevance. Indeed, increasing the number of genes in gene panels, or ultimately expanding the analyses to WES strategies, may help identify previously undetected pathogenic variants, including those in genes not previously surveyed in smaller gene panels. However, expanding to larger gene panels or WES will also increase the number of variants of uncertain significance and incidental findings will place a considerable burden on medical genetic laboratories in analyzing and interpreting variants within a clinical setting (Molina-Ramírez et al., 2022). The workload of interpretation is an important consideration when implementing high-throughput sequencing in the clinic. We therefore argue that the limitations regarding the use of targeted NGS panels do not necessarily include the number of genes per se. From our perspective, the challenge is to focus the analyses on the highest possible number of genes consistent with the clinical phenotype of each individual patient. Furthermore, it would have been useful to evaluate familial cases of POI in order to further investigate any co-segregation with the genetic variants found in these women. We found a relatively high proportion POI-associated genetic variants in women from sub-Saharan Africa, and although previous studies have observed a higher prevalence of POI in non-European populations, lack of genetic diversity in the public genomic databases such as gnomAD can impact diagnostic accuracy (Luborsky et al. 2003; Kessler et al. 2016; Mishra et al. 2019; Eskenazi et al. 2021; Marwaha et al., 2022).

The strength of this study is that it encompasses a relatively large and well-characterized cohort. We also used a large gene panel and focused on rare variants using strict criteria that are useful in a clinical setting. Another strength is the use of specific autoantibodies instead of indirect immunofluorescence.

The introduction of NGS technology has dramatically altered the availability and effectiveness of genetic testing (Tucker et al., 2016; França and Mendonca, 2022). However, as no specific clinical or biochemical markers predicted the underlying cause of POI, the question of Who should be offered genetic testing remains controversial, and cost-effectiveness analyses are required to aid sustainable diagnostics in clinical guidelines. In this study, adding an NGS POI-related gene panel increased the determination of an etiological diagnosis. More research is needed to aid interpretative approaches and classification of genetic variants.

In conclusion, we suggest that women with newly diagnosed POI go through standard recommended diagnostic investigations including screening for chromosomal anomalies, FMR1 premutations and testing for 21OH autoantibodies. Adding NGS gene panels will increase the diagnostic yield.

Data availability

The data underlying this article will be shared on reasonable request to the corresponding author.

Acknowledgements

We thank Dr Kristin Viste and the Hormone Laboratory at Haukeland University Hospital for analyzing the blood tests as well as help in distributing information about the study. We also thank the National Registry of Organ-Specific Autoimmune Diseases (ROAS), Department of Medicine, Haukeland University Hospital, Bergen, Norway, for analyzing autoantibodies.

Authors’ roles

The study was designed and directed by M.Ø., E.H., A.L., S.B., and E.V. Clinical work and statistical analysis were performed by E.V. The genetic analysis and classifications were performed by E.B., S.B., and R.B. All authors contributed to interpretation of findings. The manuscript was drafted by E.V. with contributions to revision and final version by all authors.

Funding

The study was supported by grants and fellowships from Stiftelsen Kristian Gerhard Jebsen, the Novonordisk Foundation, the Norwegian Research Council, University of Bergen, and the Regional Health Authorities of Western Norway.

Conflict of interest

The authors decleare no conflict of interest.