Results Homozygous missense mutation in MSH5 identified in POI pedigree Two sisters (III4 and III5) from the non-consanguineous Han Chinese family (Fig. 1A), aged 31 and 29 years, experienced oligomenorrhea since menarche (14 and 13 years old), and amenorrhea occurred approximately 10 years later (Table 1). Both of them have elevated serum FSH, infantile uteri, and atrophic ovaries devoid of follicles. Chromosomal abnormalities, FMR1 premutation, autoimmune disorders, previous ovarian surgery or chemo-/radiotherapy were absent in any of the family members. Figure 1 Pedigree of a family with two daughters afflicted by POI and homozygous MSH5 p.D487Y variant. (A) The pedigree of the index family, ascertained through III5. WES was performed on the family members labeled with asterisk, and those labeled with genotypes were available for Sanger sequencing. “T” denotes the mutant MSH5 allele, and “G” wild type. Arrow indicates the proband. (B) The location of p.D487Y variant is in the DNA-binding domain of MSH5, and the residue is conserved from saccharomyces to human (C). (D) shows the mRNA level of MSH5 in fetal tissues, which is significantly higher in ovary than others. (E) The RT-PCR in fetal ovary, human granulosa cells (hGCs, obtained from one patient receiving in vitro fertilization treatment) and COV434 cells, shows that MSH5 is also highly expressed in adult granulosa cells. MT, mutant; and WT, wild type. Table 1 Clinical features of familial and sporadic POI patients with mutations in MSH5 a1–2 mg daily dose of oral estradiol valerate tablet for 21 days plus oral micronized progesterone (200 mg/days) for 12 days each month. In this POI pedigree, two affected siblings (III4, III5), both parents (II3, II4), and one unaffected daughter (III6) had been chosen to perform WES (Fig. 1A). In an autosomal-recessive inheritance model, 2 nonsynonymous homozygous variants in MSH5 (MIM 603382, chromosome 6p21.33) and ZNF391 (Zinc Finger Protein 391, chromosome 6p22.1) were revealed. Sanger sequencing confirmed that neither of the two homozygous variants were present in unaffected family members and both of the variants were absent in 400 fertile women. The ZNF391 variant (ENST00000244576: c.187A > G, p.S63G) was predicted to be benign by Polyphen2, and the Serine residue mutated was not conserved among species (Supplementary Material, Fig. S1). Furthermore, ZNF391 has not been related to any human disease and no mutation was found in 200 sporadic patients with POI. Therefore, the MSH5 variant (ENST00000375755: c.1459G > T, p.D487Y) remained as the only potential candidate for this POI family. Sanger sequencing for MSH5 in sporadic cases with POI identified 3 additional heterozygous mutations (ENST000 00375755: c.1057C > A, p.L353M; c.1459G > T, p.D487Y and c.2107 A > G, p.I703V), which had not been reported in either the Exome Variant Server or 1000 Genomes database. Among them, p.L353M and p.D487Y located in the DNA-binding domain and the original residues were highly conserved among species from yeast to human, while p.I703V occurred at the less conserved residue located at the ATPase domain (Supplementary Material, Fig. S2). Expression of MSH5 in the primate ovary Through RT-PCR in various tissues of human fetuses, which were induced abortion at 21 weeks, we found MSH5 was highly expressed in fetal ovary and adrenal gland (Fig. 1D). We also found MSH5 was highly expressed in adult human granulosa cells (hGCs), including the hGCs obtained from one patient receiving in vitro fertilization treatment and COV434 (human ovarian granulosa tumor cell line) cells (Fig. 1E). Homozygous mutant mouse model carrying Msh5 p.D486Y point mutation displayed POI phenotype The homologous residue for human MSH5 c.G1459 (ENST00 000375755: p.D487) in mouse is Msh5 c.G1456 (EN SM UST00000007250: p.D486), which is highly conserved (Fig. 1C). To examine the functional effect of p.D487Y identified in human POI, the mouse model carrying Msh5 p.D486Y point mutation was generated using a CRISPR/Cas9 system (Supplementary Material, Fig. S3A and B). The homozygous mutant mice Msh5D486Y/D486Y were obtained by intercrossing of Msh5+/D486Y mice (Supplementary Material, Fig. S3C). The homozygous mutant mice were viable at birth and no obvious developmental defects were observed in adults (Fig. 2A). However, all the Msh5D486Y/D486Yfemales were infertile. The size of the ovary was dramatically reduced compared to that of control females (Fig. 2B). The results of histological studies showed that the follicles at different developmental stages (black arrows) were observed in control ovaries (Fig. 2C), whereas no developing follicle was noted in Msh5D486Y/D486Y ovaries (Fig. 2D) at 2 month of age. Numerous Ddx4-positive germ cells (black arrows) were observed in control ovaries (Fig. 2E), but no germ cells were noted in Msh5D486Y/D486Y ovaries (Fig. 2F). Figure 2 Germ cell loss in ovaries from Msh5D486Y/D486Y mice. (A) The homozygous mutant mice were viable at birth and no obvious developmental defects were observed in adults. (B) The size of ovaries from homozygous mutant females was dramatically reduced compared to the control ovaries. The follicles at different developmental stages (arrows) were observed in control ovaries (C), whereas no developing follicle was noted in Msh5D486Y/D486Y ovaries (D). Numerous Ddx4-positive germ cells (arrows) were observed in control ovaries (E), but no germ cells were noted in Msh5D486Y/D486Y ovaries (F). Scale bars: 200 μm. Adverse effect of MSH5 p.D487Y on DNA repair for DSBs DSBs were induced by ETO treatment, and γH2AX foci were observed at the DSBs sites, which gradually disappeared along with the repair processing (Fig. 3A). The result showed that γH2AX level was much higher in the U2OS cells overexpressing mutant MSH5 than wild type (Fig. 3B). Furthermore, HeLa cells overexpressing mutant MSH5 showed a lower clonogenic survival rate than wild type, indicating the adverse effect of p.D487Y on DNA repair capacity or cellular resistance for DSBs. (Fig. 3C and D). Figure 3 MSH5 p.D487Y impaired DNA repair. (A) Immunofluorescence showed the γH2AX foci formation in U2OS cells overexpressing wild type (WT) or mutant (D487Y) MSH5-GFP protein when suffering from ETO treatment. Scale bars: 5μm. (B) The γH2AX concentration among U2OS cells overexpressing blank vector (Control), wild type (WT) or mutant (D487Y) MSH5-Flag protein was detected by western blot. The expression of wild type and mutant MSH5 was detected by Flag antibody and β-actin was used as the loading control. (C) and (D) showed the clonogenic survival rate of WT- and D487Y-overexpressing cells in response to ETO treatment. The experiments on U2OS and HeLa cells both had 3 replicates independently. Data in the figure are shown as mean ± SD. M, siRNA targeting at MSH5; NT, non-targeting siRNA; and WT, wild type.