Pathogenic Variants in ZSWIM7 Cause Primary Ovarian Insufficiency

Abstract

Context

Primary ovarian insufficiency (POI) is a genetically heterogeneous condition associated with infertility and an increased risk of comorbidities. An increased number of genes implicated in DNA damage response pathways has been associated with POI as well as predisposition to cancers.

Objective

We sought to identify and characterize patients affected by POI caused by pathogenic variants in genes involved in DNA damage response during meiosis.

Setting

Study subjects were recruited at academic centers.

Patients or Other Participants

Individuals with a diagnosis of POI and their family members were enrolled for genetic analysis. Clinical findings, family history, and peripheral blood samples were collected.

Research design

Exome sequencing was performed on the study participants and their family members (when available). Protein conservation analysis and in silico modeling were used to obtain the structural model of the detected variants in the ZSWIM7 gene.

Main Outcome Measure(s)

Rare deleterious variants in known and candidate genes associated with POI.

Results

Homozygous deleterious variants in the ZSWIM7 gene were identified in 2 unrelated patients with amenorrhea, an absence of puberty, and prepubertal ovaries and uterus. Observed variants were shown to alter the ZSWIM7 DNA-binding region, possibly affecting its function.

Conclusions

Our study highlights the pivotal role of the ZSWIM7 gene involved in DNA damage response during meiosis on ovarian development and function. Characterization of patients with defects in DNA repair genes has important diagnostic and prognostic consequences for clinical management and reproductive decisions.

Primary ovarian insufficiency (POI) encompasses heterogeneous conditions characterized by primary or secondary amenorrhea from the depletion of ovarian follicles or ovarian dysfunction before age 40 years. Several chromosomal and monogenic causes have been identified, leading to a spectrum of POI manifestations ranging from milder forms of hypogonadism to a more severe presentation such as ovarian dysgenesis (1). POI-causing genes have diverse functions, including adrenal and gonadal steroidogenesis, germ cell recruitment and folliculogenesis, DNA repair, meiosis, and neuroendocrine signaling, all of which are pivotal in the development and function of gonads and other reproductive organs.

Recent applications of exome and genome sequencing on consanguineous families with multiple affected individuals have revealed a substantial contribution of DNA repair and damage response genes to gonadogenesis and etiology of POI (2-4). Multiple genome- and transcriptome- wide studies have demonstrated a global effect of DNA damage response and repair genes on the depletion of the ovarian reserve and the age of natural menopause onset (5-7). In fetal oocytes, hundreds of DNA double-strand breaks are generated throughout the genome to promote meiotic recombination. In both somatic and germ cells, the inability to efficiently repair DNA damage compromises genomic integrity inducing cell-cycle arrest and apoptosis. Biallelic pathogenic variants in several genes involved in responses to DNA damage during meiosis such as BRCA2 (MIM*600185), HFM1 (MIM*615684), MCM8 (MIM*608187), MCM9 (MIM*610098), MEIOB (MIM*617670), MSH4 (MIM*602105), MSH5 (MIM*603382), and XRCC4 (MIM*194363) have been established to cause POI in multiple unrelated patients with primary amenorrhea and ovarian dysgenesis (1-3, 8, 9). Many of these genes are also expressed in testis. Pathogenic variants in other DNA damage response genes such as DMC1 (MIM*602721), XRCC2 (MIM*600375), and ZSWIM7 (MIM*614535) have been reported to cause DNA replication errors and aberrant meiotic recombination, resulting in spermatogenic arrest and infertility in men (8, 10, 11). However, their role in POI is not yet established.

Materials and Methods

Human Subjects

The study was approved by the ethics committee of the Cukurova University Faculty of Medicine (KARAR#1; DA:KADB-13F-R.00), the institutional review boards at the University of Pittsburgh (PRO09080427), and the Center for Reproductive Medicine of Shandong University (institutional review board #77). Cohorts of POI patients and their family members were recruited by participating institutions. Informed written consent was obtained from each participating subject. Genomic DNA samples were isolated from peripheral blood samples of affected individuals (P20428 and TPOF-GII2) as well as from TPOF-G unaffected family members using a Puregene Blood Kit B (QIAGEN, Hilden, Germany).

Exome Sequencing and Data Analysis

Genomic DNA (~1.5 μg) underwent exome capture using an Agilent SureSelectXT Human All Exon V6 kit (Santa Clara, CA), and sequenced on a HiSeq X Ten platform (Illumina, San Diego, CA) with an average depth coverage of 100×. The Burrows–Wheeler Alignment tool was used to align raw data to the human reference genome sequence (UCSC Genome Browser hg19). Variant calling was performed using the Genome Analysis Toolkit (12), and ANNOVAR software was further used to annotate all variants. Common variants with minor allele frequencies (MAF) > 0.001 present in the 1000 Genomes Project (http://www.1000genomes.org/), the Genome Aggregation Database (gnomADv2, http://gnomad.broadinstitute.org/), ExAC Browser (http://exac.broadinstitute.org/, and ClinVar (https://www.ncbi.nlm.nih.gov/clinvar/) were filtered out. Rare (MAF ≤ 0.001) protein-altering variants were retained for analysis. Functional effect was evaluated by SIFT (http://sift.jcvi.org/), PolyPhen-2 (http://genetics.bwh.harvard.edu/pph2), and MutationTaster (http://www.mutationtaster.org) bioinformatic tools.

Genome Sequencing and Data Analysis

Raw data produced from sequencing machine were converted to fastqs using Cassava 2.20 software. The produced fastqs were uploaded to Illumina basespace using basespace API commands. After the upload was complete, DRAGEN 3.8.4 pipeline (13) was triggered through the same API to produce the vcfs, Grch38 was used to align the reads to reference genome using bwa-mem aligner. Structural variant and copy number variant callers were used to generate their respective vcfs. Single nucleotide variant (SNV), structural variant, and copy number variant VCFs were uploaded in Moon (Diploid/Invitae). Only coding variants with a read depth > 10, genotype quality > 40, and MAF < 2% in GnomAD and that segregated with the disease were further considered for interpretation. DNA samples of family members were used in segregation analysis.

In Silico Modeling

We used Alphafold (14) to obtain the structural model of ZSWIM7. No postmodeling refinement was performed. Mutations were modeled in the Pymol Molecular Graphics System (Schrödinger, Inc.) with no refinement. Images of the structures were generated in Pymol. Eighty-nine percent of the residues were modeled with very high confidence; the C-terminal 7 residues were poorly modeled.

Protein Conservation Analysis

Protein sequences of human ZSWIM7 and its homologs in Mus musculus, Danio reiro, Canis lupus, Pan troglodytes, Bos taurus, and Gallus gallus were aligned using Clustal Omega (15).

Results

The first patient, P20428 (Fig. 1A), was a 28-year-old woman from China diagnosed with POI due to elevated FSH and LH levels (45.73 and 16.65 mIU/mL, respectively) (Table 1) and amenorrhea. No parental consanguinity was reported, and family history was unremarkable. Patient had menarche at 16 years of age followed by 3 more light bleeding episodes and subsequent amenorrhea at age 18 years. Neither ovaries nor uterus was visualized on pelvic ultrasound. Prepubertal uterus was visualized by magnetic resonance imaging. On the physical examination, external genitalia had normal anatomic appearance.

Figure 1.

Variants in the ZSWIM7 gene. (A, B) Family pedigrees. Affected females are shown in gray-filled circles. Both patients had normal 46,XX karyotype and no history of autoimmune or iatrogenic conditions as a cause of primary ovarian insufficiency. Double horizontal lines indicate consanguinity. The ZSWIM7 genotype is provided for each individual. The wild-type allele is indicated as WT. (C) Gene structure with known protein domains of the gene product. The ZSWIM7 (zinc finger SWIM domain-containing protein 7) gene is located on chromosome 17p12 (chr17:15,879,875_ 15,903,006; hg19) and composed of 5 exons. Exons are specified as green-colored boxes. In patient P20428, the ZSWIM7 homozygous missense c.38T > C variant occurs in exon 1. In TPOF-GII2, a homozygous frameshift c.231_232delAT variant is located in exon 4, encoding the zinc finger SWIM domain (purple-shaded box). ZSWIM7, a 140-amino acid protein, contains 2 β-strands, a zinc finger SWIM (ZF SWIM) domain, and an α-helix. (D) Protein sequence alignment of human ZSWIM7 and its homologs in Mus musculus, Danio reiro, Canis lupus, Pan troglodytes, Bos taurus, and Gallus gallus. The affected Leu13 residue (marked in red) is highly conserved among all shown organisms. Residues are shaded based on similarity. Symbols denote: (*) conserved residue; (;) conservation between groups of strongly similar properties (>0.5 in the Gonnet PAM 250 matrix); (.) conservation between groups of weakly similar properties (≤0.5 in the Gonnet PAM 250 matrix). (E) In silico modeling of amino acid substitutions in ZSWIM7. The WT and 2 altered protein structures are shown. The leucine to proline change at residue 13 (p.Leu13Pro) is predicted to “break” the N-terminal helix, dividing this helix into 2, whose central axes no longer align (red lines). This could reorganize the packing of the entire molecule, including the DNA-binding region. Furthermore, if the regions surrounding position 13 still were able to form helices, the smaller proline sidechain would destabilize its interaction with neighboring amino acids. The p.Cys78fs*21 frameshift change at residue 78 that truncates the protein after 21 residues would impact the β-sheet and helix integral to supporting the zinc finger fold.

Table 1.

Clinical laboratory investigations in affected individuals with the ZSWIM7 variants

Findings	Normal range	P20428	TPOF-GII2	Patient 1a	Patient 2a
Ethnicity	-	Han Chinese	Turkish	Turkish	Turkish
cDNA change: NM_001042697	-	c.38T>C	c.231_232delAT	c.173C>G	c.173C>G
Protein change: NP_001036162.1	-	p.(Leu13Pro)	p.(Cys78Phefs*21)	p.(Ser58*)	p.(Ser58*)
Age at menarche (y)	-	16	PA	PA	PA
Age at menopause (y)	-	18	-	-	-
FSH (mIU/mL)	1.8-22.5	45.73	30.99	94.8	77.8
LH (mIU/mL)	1.2-100	16.65	13.88	17.2	14.3
Estradiol (pg/mL)	30 to 300	<10	4	Undetectable	Undetectable
Prolactin (ng/mL)	3.3-26.7	11.30	11.93	NR	NR
Uterus size	-	Prepubertal	Prepubertal	Prepubertal	Prepubertal
Ovarian volume (cm3)	6.6	Not visualized	Not visualized	streak ovaries	Not visualized
Tanner Stage (pubic hair)	-	3	3	NR	NR
Tanner Stage (breast)	-	2	1	NR	NR

Abbreviations: NR, not reported; PA, primary amenorrhea.

^aTwo affected sisters previously described in the literature (4).

Patient TPOF-GII2 (Fig. 1B) was a teenage girl of Kurdish descent from Turkey who presented with absent breast development, primary amenorrhea, and short stature. The parents were healthy first-degree cousins. Physical examination revealed cubitus valgus and a broad chest with widely spaced nipples. Both her pubic hair and axillary hair were at Tanner Stage 3. Her basal estradiol level was low with elevated LH and FSH levels (Table 1). Pelvic ultrasonography and magnetic resonance imaging showed no visible uterus or ovaries. At age 14 years, she had regular menstrual bleeding with combined estrogen-progestin replacement therapy and achieved appropriate pubic hair and breast development. Her older sister (subject TPOF-GII1) had normal breast development (Tanner Stage 4) and had regular menstrual bleeding. The patient’s 13-year-old brother (TPOF-GII3) had normal development and normal (8 mL) testicular volumes.

To identify the genetic cause of primary amenorrhea, we performed exome sequencing on patient P20428 (parental samples were unavailable) and genome sequencing on the TPOF-G family members. SNVs were filtered to retain novel and rare protein-coding changes with minor allelic frequency MAF ≤ 0.1% (GnomADv2). We prioritized variants in genes that are known to cause POI (1), followed by genes not yet reported but considered as candidates based on their involvement in gonadogenesis and oogenesis, hormonal synthesis, signaling and ovulation processes, meiotic chromosome segregation, and DNA repair in animal models (1, 8). In TPOF-GII2, we also evaluated homozygous SNVs that were absent in the patient’s sister because an autosomal-recessive inheritance was most plausible in this consanguineous family.

At the time of the initial analysis, no variants in known POI-causing genes were observed in either P20428 or TPOF-GII2 patient. Thirteen homozygous variants that fulfilled our filtering and family segregation criteria were identified in the TPOF-GII2 patient (Table 2). Of those, 12 were ruled out as a cause after examination of gene expression profiles and their known function. Similarly, no variants in known POI-causing genes were initially identified in the P20428 patient.

Table 2.

Homozygous variants found in TPOF-GII2, but not in her unaffected sister

Chr	Position, hg19	Gene	Disorder	Effect	cDNA change	Protein effect	Ensembl transcript ID	RefSeq transcript ID	MAF gnomAD	Allele hmz	Allele htz
10	113921475	GPAM	No associated disorder	Missense	c.1444A>C	p.Ile482Leu	348367		0	0	0
11	46404352	MDK	No associated disorder	Frameshift	c.461_462insGC	p.Arg155fs	489525		0.00194525	0	280
11	114549328	NXPE2	No associated disorder	Splice_region	c.26 + 5G>A		389586	NM_182495.5	6.35E-06	0	1
13	24463420	MIPEP	Combined oxidative phosphorylation deficiency 31	Missense	c.40G>A	p.Ala14Thr	382172	NM_005932.3	0.000398142	0	6
13	52971362	THSD1	Nonimmune hydrops fetalis, THSD1-related	Splice_region	c.1021 + 5G>T		258613	NM_018676.3	0	0	0
14	81609751	TSHR	Hyperthyroidism	Missense	c.1349G>A	p.Arg450His	298171	NM_000369.2	0.000234637	0	59
17	7834015	TRAPPC1	No associated disorder	Missense	c.347C>T	p.Pro116Leu	303731	NM_021210.4	7.96E-06	0	2
17	8045782	PER1	No associated disorder	Splice_region	c.3260-6T>A		317276	NM_002616.2	0.000302402	0	65
17	15884427	ZSWIM7	Male infertility, ZSWIM7-related	Frameshift	c.231_232delAT	p.Cys78fs	399277	NM_001042698.1	0.000364808	0	91
17	17070643	MPRIP	No associated disorder	Missense	c.3842A>G	p.Asn1281Ser	313485		0	0	0
17	38320381	CASC3	No associated disorder	Missense	c.1433C>T	p.Pro478Leu	264645	NM_007359.4	0.00110912	0	269
20	25457678	NINL	No associated disorder	Missense	c.2249C>T	p.Ala750Val	278886	NM_025176.4	0.000235843	0	54
22	46653641	PKDREJ	No associated disorder	Missense	c.5579C>T	p.Pro1860Leu	253255	NM_006071.1	0	0	0

Abbreviations: Allele hmz, number of allele counts in a homozygous state as per GnomADv2; Allele htz, number of allele counts in a heterozygous state as per GnomADv2; MAF, minor allelic frequency as per GnomADv2.

Recently, ZSWIM7 pathogenic variants were identified in infertile males (https://preview.ncbi.nlm.nih.gov/clinvar/variation/1013608/), and mouse Zswim7 knockout model showed gonadal dysgenesis and infertility in both male and female mice (11, 16). Reanalysis of sequencing data in the TPOF-GII2 patient showed that the ZSWIM7 homozygous pathogenic c.231_232delAT (rs368517882) frameshift variant was present in TPOF-GII2, but not in her unaffected siblings (Fig. 1C). This specific variant is seen in heterozygous state in the 1000 Genomes, dbSNP, and gnomADv2.1 databases among multiple ethnicities with a global MAF = 0.0003774 (gnomADv2.1) and has been reported in the homozygous state in patients with nonobstructive azoospermia (11). In the P20428 patient, a novel homozygous ZSWIM7 c.38T > C, p.(Leu13Pro) variant (rs200271348) was identified and confirmed by Sanger sequencing. This variant has a low frequency (MAF = 0.00001464) in the human population and is present in a highly conserved residue among multiple species (Fig. 1D). The ZSWIM7 c.38T>C, p.(Leu13Pro) variant is predicted to be damaging by bioinformatic tools (CAAD score = 31). In silico modeling of the ZSWIM7 c.38T>C, p.(Leu13Pro) and c.231_232delAT (rs368517882) frameshift variants was performed to investigate the molecular impacts of these variants on the ZSWIM7 structure (Fig. 1E). Both variants were found to reorganize or truncate the DNA-binding region of ZSWIM protein, affecting the zinc finger function.

Discussion

Studies in mice show that ZSWIM7 is involved in meiotic homologous recombination repair through formation of a heterodimeric complex with SWIM-type zinc finger 7 associated protein 1 (MIM*614536) and plays a role in RAD51 and DMC1 driven repair of DNA double-strand breaks (16, 17). ZSWIM7 is a conserved protein, and the SWIM domain is required for its function. Based on our analysis, both variants are predicted to adversely affect ZSWIM7 by impacting the structural integrity of the DNA-binding domain. Inadequate repair of damaged DNA in mitotic cells results in genomic instability, growth defects, and ultimately germ cell loss. Germ cells are known to be sensitive to unrepaired double-stranded DNA breaks, as shown in knockout animal models. Male and female Zswim7^-/- and Swsap1^-/- mice show no obvious morphological anomalies, but have gonadal failure because of premature loss of postmeiotic germ cells (16). Mice homozygous for a knock-out allele exhibit female and male infertility with decreased testis and ovary weights, azoospermia, absent ovarian follicles, and impaired chromosomal synapsis.

Both of our patients, and recently published affected sisters from a single family with a loss-of-function ZSWIM7 variant (4), presented with early-onset amenorrhea, absence of puberty, and atrophic ovaries, a phenotype that complements one described in affected male individuals (11, 18) and mimic mouse knockout phenotype.

The presence of ZSWIM7 variants in diverse ethnic groups (11, 18) suggests that it might be a more frequent cause of gonadal dysgenesis and infertility than anticipated. Future and follow-up studies on POI and nonobstructive azoospermia individuals with pathogenic variants in ZSWIM7 and other DNA damage response genes will be necessary to define the spectrum of pathogenic variants for clinical diagnosis and human reproductive and other phenotypes attributed to mitotic and meiotic genomic instability. Reproductive problems or menstrual irregularities have not been reported in females heterozygous for the pathogenic variants in the ZSWIM7 gene; however, there are too few families identified to establish such an association. DNA repair genes have been shown to influence the reproductive lifespan regulating germ cells proliferation, the size of the initial pool of primordial follicles, and the rate of its depletion (2, 3, 5, 7). Although heterozygous variants are not associated with obvious gonadal pathology, they may confer risk for accelerated loss of ovarian reserves or infertility. The identification of genetic causes and characterization of phenotypic spectrum of patients with pathogenic variants in infertility have diagnostic and prognostic consequences for clinical management and reproductive decisions. Moreover, longitudinal studies are necessary to determine if these variants and genes contribute to additional morbidity and mortality (7) previously associated with loss of ovarian function before age 40 years.

Glossary

Abbreviations

MAF

minor allele frequency

POI

primary ovarian insufficiency

SNV

single nucleotide variant

Financial Support

This work was supported by the National Institute of Child Health and Human Development (R01HD070647 and R21HD074278 to A.R.) as well as National Science Fund for Distinguished Young Scholars (82125014) and the National Natural Science Foundation of China (82071609, 81771541 to Y.Q.).

Author Contributions

A.R., Y.Q., S.A..Y, F.G., and K.T. conceived the experiments and collected the clinical information; P.M.M., Y.Q., S.A.Y., and A.R. analyzed the next-generation sequencing data and identified the pathogenic variants; A.J.B. performed in silico modeling. The manuscript was written by S.A.Y., M.R.E., and A.R. and reviewed by P.M.M., Y.Q., A.J..B, F.G., K.T., and A.R. All authors read and approved the final manuscript.

Disclosures

All authors: No reported conflicts of interest. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.