Four members of a five-generation consanguineous Han Chinese family with POI, namely, two affected sisters (V-1; V-2, Figure 1A ) and two unaffected second-cousin parents (IV-1; IV-2, Figure 1A ) participated in this study. Peripheral blood samples from the four individuals were collected, and genomic DNA (gDNA) was extracted using a QIAamp DNA Blood Midi Kit (Qiagen, Hilden, Germany). The four family members provided their written informed consent to participate in this study and agreed with the publication of this case report. This study was reviewed and approved by the ethics committee of the Reproductive & Genetic Hospital of CITIC-Xiangya of Central South University, China.

The gDNA from the proband was subjected to WES following protocols described previously (He et al., 2018b). Whole exomes were captured using Agilent SureSelect version 4 (Agilent Technologies, Santa Clara, CA) and sequenced on a HighSeq2000 sequencing platform (Illumina, San Diego, California, USA). WES data analysis was performed using the Genome Analysis Toolkit (GATK) and consisted of removing adaptors, mapping WES raw reads to a reference human genome sequence (NCBI GRCh37, reference genome Hg19) using the Burrows–Wheeler Aligner (BWA), eliminating PCR duplicates, and sorting sequences using Picard (http://broadinstitute.github.io/picard/). Variant identification was performed using the GATK package according to the recommended best practices, including base recalibration variant calling with Haplotype Caller and variant quality score recalibration and annotation using ANNOVAR software.

All candidate variants needed to meet the following criteria: (i) absent or occurring at a frequency less than 0.01 in any of the following databases: 1,000 Genomes Variant Database, Exome Aggregation Consortium (ExAC), or NHLBI-GO Exome Sequencing Project (GO-ESP); (ii) homozygous variants were considered with priority; (iii) predicted to be deleterious; and (iv) relevant to the phenotype shared by the patients based on comprehensive expression data (expression in ovary), and model organism data (a female-lacking-oocytes phenotype presented in animal models).

Specific PCR primers targeting two variants of the STAG3 gene (NM_001282716) were designed and validated. The sequences were as follows: 5′-AACCACATGCAGAGGGGTTAT-3′ and 5′-TCCAGCTGCATTAATTCTGGGA-3.′

The full-length coding sequence of STAG3 was amplified from control human blood cDNA, and the STAG3 product was cloned into pDsRed2-N1, resulting in STAG3 followed by DsRed2 protein, as STAG3-WT-DsRed2. Additionally, the cDNA encoding FLAG₃-STAG3 was cloned into pcDNA3.1, namely, FLAG₃-STAG3-WT. Variants c.877_885del and c.891_893dupTGA were separately or simultaneously introduced into the STAG3-WT-DsRed2 and FLAG₃-STAG3-WT plasmid vectors using a Mut Express II Fast Mutagenesis Kit V2 (Vazyme, Guangzhou, China) to achieve site-directed mutagenesis. Altogether, six mutants were generated (Mut 1 indicates STAG3-p.293_295del-DsRed2 or FLAG₃-STAG3-p.293_295del; Mut 2 indicates STAG3-p.297_298insAsp-DsRed2 or FLAG₃-STAG3-p.297_​298insAsp; Mut 3 indicates STAG3-p.293_295del and p.297_298insAsp-DsRed2 or FLAG₃-STAG3-p.293_295del and p.297_298insAsp). Moreover, cDNAs encoding REC8 and SMC1A were separately cloned into pEGFP-N1 and pcDNA3.1-HA. All expression constructs were sequenced to exclude unintended PCR-generated variants and confirm the presence of the desired variants.

Chinese hamster ovary (CHO) cells and human embryonic kidney (HEK) 293 cells were cultured in Dulbecco’s modified Eagle medium nutrient mixture F-12 (Ham) (DMEM/F12) (Gibco, Gaithersburg, MD, USA) with 10% fetal bovine serum (FBS) (Gibco, Gaithersburg, MD, USA) at 37°C in a humidified 5% CO₂ incubator.

According to the manufacturer’s instructions for Lipofectamine 3000 (Thermo Fisher Scientific, Carlsbad, CA, USA), CHO cells were transfected with expression vectors containing wild-type STAG3-DsRed2 and REC8-EGFP plasmids or mutated STAG3-DsRed2 and REC8-EGFP plasmids. CHO cells transfected with empty vector (EV; without STAG3) and REC8-EGFP plasmids were used as negative controls. CHO cells were fixed for fluorescence localization after culture for 48 h. 4’,6-Diamidino-2-phenylindole (DAPI) (Beyotime) was used to visualize DNA. All fluorescent images were captured on an orthostatic fluorescence microscope (Olympus, Japan).

Co-IP was performed as described previously (Wolf et al., 2018). Briefly, HEK293 cells were transfected with expression vectors containing wild-type FLAG ₃ -STAG3 and REC8-EGFP plasmids (or SMC1A-HA plasmids) or mutated FLAG ₃ -STAG3 and REC8-EGFP plasmids (or SMC1A-HA plasmids), according to the manufacturer’s instructions for Lipofectamine 3000 use. Untransfected HEK293 cells were used as a negative control. After incubation for 36 h, the culture medium was supplemented with 0.2 µg/ml nocodazole (Sigma, Louis, MO, USA) for prometaphase arrest of the cells. Total protein was extracted using RIPA lysis buffer (Thermo Fisher Scientific) after 12 h, and then, western blotting was performed to detect the expression of STAG3, REC8, and SMC1A using anti-FLAG, anti-GFP, and anti-HA (Proteintech) antibodies, respectively, at a 1:1,000 dilution. Next, immunoprecipitation was performed using Dynabeads Protein G (Invitrogen) following the manufacturer’s instructions.

The proband (V-2, Figure 1A ) initially went to the clinic because of primary amenorrhea and was diagnosed with POI at 18 years old. Her breast development was at Tanner stage I, although she had a normal height (160 cm) and weight (50 kg). The results of hormonal studies showed that her serum FSH level was high (51 mIU/ml, normal range 2–22 mIU/ml), whereas her serum estradiol level was low (< 10 pg/ml, normal range 30–190 pg/ml), and her anti-Müllerian hormone (AMH) level was too low to be detected. She had a normal level of luteinizing hormone (LH; 11.08 mIU/ml, normal range 1–19 mIU/ml). Her uterus and ovaries were not visualized upon ultrasonographic examination, and she had an abnormal vulva without labia minora. The proband did not accept hormonal replacement therapy (HRT) because of the increased prevalence of invasive breast cancer after HRT (Biscup, 2003).

The proband’s affected elder sister (V-1, Figure 1A ) showed normal growth and development but presented with POI when she was 21 years old. Similar to the proband, she went to the hospital because of primary amenorrhea. Her serum FSH level was elevated (72.19 mIU/ml, normal range 2–22 mIU/ml), whereas her serum estradiol level was low (< 10 pg/ml, normal range 30–190 pg/ml), and her LH level was normal (18.58 mIU/ml, normal range 1–19 mIU/ml). Ultrasonographic examination showed that her uterus was anteflexed and small (23×10×20 mm). Both ovaries were also small: the left ovary was 14×10×11 mm, and the right one was 15×11×13 mm, consistent with the size of streak gonads. After receiving HRT with progesterone for 3 months, she showed normal menses, and her uterus grew (36×20×35 mm) to a size close to that of a normal adult-sized uterus.

Additionally, both sisters had a normal karyotype (46, XX) and FMR1 CGG repeats, and no associated adrenal or thyroid autoimmune disorders were found.

WES of the proband (V-2, Figure 1A ) resulted in approximately 139,631,574 raw reads with a mean 182.73-fold depth in the target regions, revealing high-quality sequencing ( Supplementary Table 1 ). For rare inherited diseases, the frequency of possible pathogenic variants should be very low in a healthy population. Therefore, the results were screened against minor allele frequency (MAF) > 1% in public SNP and indel databases for variants predicted to be deleterious or to result in loss of function. Homozygous variants were preferentially considered candidates because the patients were from a consanguineous family. Finally, we focused on the phenotypic relevance for POI in the proband. Only two homozygous variants of STAG3 (NM_001282716: c.877_885del, p.293_295del; c.891_893dupTGA, p.297_298insAsp) fulfilled these criteria ( Supplementary Table 2 ).

The two homozygous in-frame STAG3 variants were confirmed by Sanger sequencing and were also detected in her affected sister ( Figure 1B ). Their unaffected parents were heterozygous carriers of the two variants ( Figure 1B ). In addition, homozygosity mapping indicated that the two STAG3 variants were not located in heterozygous regions ( Figure 1C ) and that positions p.293_295 and p.297_298 are highly conserved across different species ( Figure 1D ).

According to the American College of Medical Genetics and Genomics (ACMG) standards and guidelines for the interpretation of variations (Richards et al., 2015), the two in-frame variants of STAG3 (c.877_885del, p.293_295del; c.891_893dupTGA, p.297_298insAsp) are both classified as likely pathogenic variants. Therefore, we speculated that both homozygous STAG3 variants were candidates as causes of POI.

Co-expression of wild-type STAG3 and REC8 is necessary for REC8 to enter nuclei. A fluorescence localization analysis was performed to evaluate the effects of the two variants. As expected, REC8 localized to nuclei, as did the will-type STAG3 protein, when CHO cells were transfected with expression vectors containing wild-type STAG3 and REC8 plasmids; ( Figure 2A ). In contrast, REC8 localized exclusively in the cytoplasm ( Figure 2A ) when CHO cells were co-transfected with the REC8 construct and empty vector or mutant STAG3 construct. In the latter scenario, nearly all of the mutant STAG3 proteins were excluded from nuclei ( Figure 2A ).

We then performed a Co-IP analysis to determine the pathogenicity of the two variants. As shown, wild-type STAG3 interacted with both REC8 and SMC1A as previously reported (Prieto et al., 2001; Wolf et al., 2018) ( Figure 2B ). Additionally, when we directly examined the ability of the three transiently expressed mutant STAG3 proteins (p.293_295del; p.297_298insAsp; p.293_295del, and p.297_298insAsp) to associate with REC8 or SMC1A, the results revealed that there was no interaction between the three mutant STAG3 proteins and REC8 or SMC1A ( Figure 2B ).