A homozygous donor splice-site mutation in the meiotic gene MSH4 causes primary ovarian insufficiency

Abstract

Premature ovarian insufficiency (POI) is a frequent pathology that affects women under 40 years of age, characterized by an early cessation of menses and high FSH levels. Despite recent progresses in molecular diagnosis, the etiology of POI remains idiopathic in most cases. Whole-exome sequencing of members of a Colombian family affected by POI allowed us to identify a novel homozygous donor splice-site mutation in the meiotic gene MSH4 (MutS Homolog 4). The variant followed a strict mendelian segregation within the family and was absent in a cohort of 135 women over 50 years of age without history of infertility, from the same geographical region as the affected family. Exon trapping experiments showed that the splice-site mutation induced skipping of exon 17. At the protein level, the mutation p.Ile743_Lys785del is predicted to lead to the ablation of the highly conserved Walker B motif of the ATP-binding domain, thus inactivating MSH4. Our study describes the first MSH4 mutation associated with POI and increases the number of meiotic/DNA repair genes formally implicated as being responsible for this condition.

Introduction

Primary ovarian insufficiency (POI) is a frequent pathology affecting women under 40 years of age (1). It is characterized by the cessation of menses (primary or secondary amenorrhea, according to the severity) and high serum FSH levels (>40 UI/l). Although the etiology of POI can be clearly defined in some women, a significant amount of cases remain idiopathic. Genetic abnormalities have been described in syndromic and non-syndromic POI but variants/mutations in only a few genes have been definitively implicated as responsible for the disease, despite numerous attempts to identify variants in candidate genes by Sanger sequencing (2,3). This may be because of the fact that female reproduction involves numerous steps, from ovary determination/differentiation to gametogenesis and ovulation, the perturbation of which may have an impact on oocyte health and fertility. This makes it difficult to select relevant candidates to be screened by Sanger sequencing. This constraint, as well as the logical scarcity of families affected by the disease has rendered the genetic studies of POI particularly challenging. Very recently, some studies based on next generation sequencing (NGS), undertaken in both familial and isolated POI cases, have led to the unequivocal implication in POI of variants/mutations in several new genes (4–6). Here, we describe the results of whole-exome sequencing (WES) of members of a family affected by POI, which led to the identification of a novel homozygous splice-site mutation in the meiotic gene MSH4, responsible for the production of a mutated MSH4 protein bearing a deletion within the ATP-binding domain.

Results

Individuals P1, P2 and C3, as shown in the pedigree of the family (Fig. 1A), underwent WES. Alignments of high quality reads produced more than 4 million on-target bases per genome (out of more than 6 million, >68%), mean target coverage was >80% and >75% target bases were covered at 20×. Although we could not establish any familial history of consanguinity, the patients came from an isolated region where recessive diseases have been observed (Boyacá, Colombia). Accordingly, we assumed a recessive mode of inheritance for variant filtering as a starting point (Fig. 1B). Only potentially pathogenic variants having minor allele frequencies below 0.05 were retained for subsequent analysis. They were expected to be homozygous in the affected sisters and heterozygous or absent in the non-affected one. We also considered the possibility of compound heterozygosity. This analysis produced five homozygous variants in five genes (Fig. 1B). All the variants and the relevant genes were analyzed in depth (Table 1). Only a novel homozygous variant in the meiotic gene MSH4 was found to be of interest (NM_002440.3: c.2355+1G>A, g.76356510: c.2355+1G>A). This variant affects the donor splice site on intron 17. Sanger sequencing confirmed the existence of the MSH4 c.2355+1G>A variant in a homozygous state in both P1 and P2. As expected, the parents (C1 and C2) were heterozygous carriers whereas C3 and her unaffected brothers displayed MSH4 wild-type sequences (Fig. 2A). The variant was absent in a cohort of 135 women over 50 years of age without history of infertility, from the same geographical region as the affected family (i.e. the C-POI group). It was also absent in the ExAC database (http://exac.broadinstitute.org). The mendelian segregation of the variant, its absence in the C-POI group as well as its potentially drastic consequences led us to further investigate it as the cause of the phenotype. In silico predictions of the potential effects of the c.2355+1G>A mutation led us to contemplate three potential mechanisms. Mech1 would involve skipping of exon 17 leading to the synthesis of the MSH4 p.Ile743_Lys785del protein. Mech2 would induce retention of intron 18 predicted to generate the truncated MSH4 p.Ala786Ilefs*34 protein, in the absence of non-sense mediated decay. A further, less likely mechanism, involving cryptic splice sites would lead to another truncated form (p.Gly772Tyrfs*13). Given the absence of suitable biological material to study the effect of the mutation at the mRNA level, we performed exon-trapping/minigene assays (Fig. 2B). These assays showed the presence of a small strong RT-PCR band in the mutated condition (Fig. 3A, 158 bp). This pointed toward skipping of exon 17 induced by the splice site mutation (the RT-PCR product corresponds to Ex16–Ex18). In the wild-type condition, we found a longer band corresponding to a properly spliced product (Ex16–Ex17–Ex18) along with the smaller fainter one (Ex16–Ex18) (Fig. 3A). Database searches showed that Exon 17 is present in all normal transcripts not only in humans but also in other eukaryotes, suggesting that the exon17-skipped variant in the wild-type condition is an artifact of minigene overexpression. Exon-skipping was confirmed by Sanger sequencing (Fig. 3B). After several attempts, we failed to amplify cDNA fragments using primers located on Ex16 and intron 17, which argues against retention of intron 17 (i.e. Mech2) (Fig. 3C).

Figure 1.

(A) Pedigree of the family affected by POI. Squares and circles indicate males and females, respectively. Black circles symbolize females affected with POI. (B) Variant filtering strategy according to a recessive inheritance mode and the corresponding statistics.

Figure 2.

Analysis of the POI family. (A) Validation of the splice site variant by Sanger sequencing. The two affected sisters (P1 and P2) were confirmed to be homozygous for the MSH4 c.2355 + 1G>A variant, the parents (C1 and C2) were heterozygous whereas the unaffected sister and the two brothers were homozygous for the reference sequence. Examples of chromatogram are displayed. (B) Schematic representation of the minigenes used to test the effect of the candidate variant c.2355 + 1G>A in MSH4 on splicing along with the expected results (Mech 1: Ex17 deletion and Mech2: intron 17 retention).

Figure 3.

Results of the minigene assay and structural impact of the c.2355 + 1G>A on MSH4. (A) RT-PCR results showed the presence of a 158-bp band in the mutated condition compatible with exon-17 skipping whereas in the wild-type condition a 287-bp band was obtained corresponding to a properly spliced product (and a smaller fainter one corresponding the exon-skipped product). (B) This was confirmed by Sanger sequencing of the two amplicons. (C) Schematic representation of the alternative protein products potentially resulting from the mutated allele, namely, exon skipping, intron retention, cryptic splice site activation and alternative spliceosome usage. (D) 3D representation of the (paralogous) deleted region in MSH4 according to Mech 1 (in red) within the context of the human MSH2–MSH6 heterodimer (in brown and green respectively). PDB structure 2O8B (22).

Table 1.

Details of the candidate variants

Gene	Type	DNA	Protein	FUNCTION	EXPRESSION
LCK	Missense	cGg/cAg	p.R45Q	Transduction of TCR	Lymphoid cells
SYDE2	Missense	Gtt/Att	p.V545I	GTPase activator	Placenta, followed by uterus, retina, adrenal gland, ovary and skeletal muscle
DCP1A	Missense/ deletion	tCCAGTGCCA GTGCCAtc/ tCCAGTGCCAtc	–	mRNA-decapping complex	Heart, brain, placenta, lung, skeletal muscle, liver, kidney and pancreas
SLC28A1	Missense/ insertion	cTc/cTTGTc	p.L140LV	Sodium/nucleoside cotransporter 1,	Kidney and liver
MSH4	Donor splice site	c.2355 + 1G>A	–	Protein involved in meiotic recombination	Testis- and ovary-specific. KO sterile secondary to defective meiosis

Discussion

In the present study, we describe two sisters affected by secondary amenorrhoea, exhibiting classical clinical signs of POI. Segregation analysis of the MSH4 c.2355+1G>A variant in the family was consistent with a recessive mode of inheritance. The parents were heterozygous and the unaffected sister (C3) and two of the brothers (II:1 and II:5) with normal fertility displayed a wild-type genotype.

MSH4 belongs to the DNA mismatch repair (MMR) family of proteins, which have been linked to post-replicative MMR and is involved in meiotic recombination. Consistent with a meiotic function, MSH4 is highly expressed in the testis (and in germ cells) whereas very low levels have been detected in other organs. The human MSH4 gene consists of 20 exons encoding a 936-aminoacid protein containing conserved regions present in MutS proteins including the ATP-binding domain and the carboxyl terminal helix-turn-helix structural motif (7). Unlike other proteins of the family, a DNA MMR function has not been demonstrated for MSH4 and its close homologue MSH5 and their knock-out in mice lead to both female and male sterility secondary to defective chromosome synapsis during meiosis (8–11). MSH4 mutations are expected to be a rare cause of POI, as this gene has already been sequenced in more than 100 affected women and no pathogenic variants were found (12,13).

Our experimental data suggest that exon skipping may take place because of the presence of the mutation and argues against intron retention or cryptic splice site activation. Exon skipping leads to the synthesis of a protein lacking 43 amino acids (p.Ile743_Lys785del), including the highly conserved Walker B motif (…LIDELGRGT…) (14). The interstitially deleted MSH4 protein may have retained some ability to heterodimerize with MSH5 and even to bind DNA. Indeed, the protein segment 786–936 located in the MSH4 C-terminal region, present in the mutated protein, mediates binding to MSH5 ensuring heterodimer formation (15), which in normal conditions interacts with recombination intermediates (16,17). However, it is known that in the Msh2–Msh6 heterodimer (Fig. 3D), point mutations affecting the ATPase sites severely impair the functionality of the whole complex and result in null alleles (14). Thus, a deletion of the Walker B motif in the mutated MSH4 protein argues against any functionality of the MSH4–MSH5 dimer. We cannot formally exclude the presence of a low level of intron retention or cryptic splice site activation generating truncated MSH4 forms containing an intact Walker B motif. However, they would be unable to bind MSH5 and it is known that a functional MSH4–MSH5 heterocomplex is important for female gamete production because the MSH5 p.Pro26Ser variant, reported in a POI patient, perturbs the interaction between MSH4 and MSH5 (12,15). Another homozygous mutation (p.D487Y) has recently been reported in a case of POI and the mouse knock-in model clearly shows meitoic defects (18).

Although the deleted MSH4 protein is probably underlying infertility in our patients, it is worth noting that they presented with secondary amenorrhea. This suggests that either some residual MSH4 activity may have been retained or that, unlike in the mouse, some degree of functional compensation by an MSH paralog is taking place in the human germ-cells. Another possibility is that the c.2355+1G>A mutation transformed a canonical splice site (i.e. processed by the major spliceosome) into one able to be processed, though at much lower efficiency, by the minor (U12) spliceosome. This system splices not only AT-AC but also AT-AG introns. In the cases reported here, the MSH4 mutated donor splice site is 5’ATATTCTTT3’, which is very close to the consensus U12-type donor splice site 5’ATATCCTTT3’ (19). According to this scenario, a small amount of normal mRNA and MSH4 protein that escaped detection by exon trapping would be produced. This would explain why the phenotype of our patients seems to be less drastic than that observed in Msh4^-/-mouse females, which display a complete loss of ovarian structures during early ovarian development (11). Our genetic and biochemical data suggest that the mutant protein owing to the MSH4 c.2355+1G>A mutation does affect meiosis. However, the generation of a mouse model simulating the human mutation will help understand the details of its impact on meiosis and oogenesis.

Six recent studies of families with recessive non-syndromic POI have led to the identification of causal variants/mutations in meiotic or DNA repair genes using formal genetics and/or NGS (e.g. MCM9, MCM8, SYCE1, STAG3, MSH5) (4–6,18,20). On the dominant side, exome sequencing, performed in four non-consanguineous women, led to the identification of a heterozygous causative mutation in the CSB-PGBD3 gene, whose product is involved in the response to DNA damage (21). Our study describes the first MSH4 mutation associated with POI and increases the number of meiotic/DNA repair genes formally implicated as being responsible for this condition. Thus, MSH4 becomes a new genetic biomarker of POI with potential utility in a clinical setting.

Materials and Methods

Patients and controls

The probands P1 and P2, corresponding, respectively, to individuals II:2 and II:4 in the pedigree shown in Figure 1A, were two Colombian sisters who attended the Center For Research in Genetics and Genomics (CIGGUR) in Bogotá (Colombia). P1 was a 42-year-old woman who experienced menarche at 12, followed by regular menstrual cycles. At age 34, she had irregular cycles for 1 year, which rapidly evolved to amenorrhea. She showed a normal development of secondary sexual characteristics (Tanner stage 5 for breast and pubic hair growth). FSH and LH plasma levels were 78.4 and 42.2 IU/l, respectively (normal ranges: FSH: 4–13 IU/l, LH: 2–15 IU/l). TSH plasma levels were 1.9 mUI/ml (normal ranges 0.2–4.2 mUI/ml). Transvaginal ultrasound revealed small uterine myomas. Ovaries displayed normal volume and gross morphology. Her karyotype was 46, XX. No other pathological conditions were recorded. P2 was a 35-year-old woman, with normal 46, XX karyotype, who experienced menarche at 14 years of age. She had irregular menstrual cycles when she was 25 and amenorrhea 5 years later. She showed a normal development of secondary sexual characteristics (Tanner stage 4 for breast and pubic hair growth). At the age of 30 she had high FSH (97.1 IU/l) and LH (51.2 UI/l) plasma levels. TSH plasma levels were 1.9 mUI/ml. Patients did not receive hormone replacement therapy. The mother C1 (I:1) experienced a normal menopause at the age of 59 and had no history of infertility. The sister (C3, 41 years old) has a normal fertility (and 2 children). The study of this family was approved by the Institutional Ethics Committee (Universidad del Rosario) and all individuals signed a written informed consent form. The control group (C-POI group) consisted of 135 women over 50 years old, originated from the same geographical region that the patients, who lack antecedents of infertility.

NGS, Sanger sequencing and in silico analysis

Total genomic DNA from all participants was purified from blood leucocytes using classic salting-out procedure. A Sure Select^XT (Agilent Technologies, Santa Clara, CA, USA) target enrichment system kit was used for exome capture. Raw reads (length ∼150 bp) passing quality filters were multiplexed using Casava software (Illumina, San Diego, CA, USA). The BWAMEM algorithm was used for aligning high quality reads with the human reference genome (HG19 version GRCh37). P1, P2 and C3 alignments of high quality reads gave the following results. P1 sample: 4 390 150 405 bases (out of 6 248 010 746, 70.26%) were on target; mean target coverage was 89.04%, 77.16% target bases being covered at 20× and 32.31% at 100×. P2 sample: 4 570 319 395 bases (out of 6 655 081 903, 68.67%) were on target; mean target coverage was 92.05%, 78.28% target bases being covered at 20× and 34.09% at 100×. C3 sample: 4 128 614 330 reads (out of 5 961 101 264, 69.29%) were on target; mean target coverage was 83.09%, 80.54% target bases being covered at 20× and 29.40% at 100×. Variant calling was performed using Samtools 0.1.18. Exome results (.vcf format) were uploaded to the PhenomeCentral database (accession codes: P1: P0002496, P2: P0002497 C1: P0002498).

Variant filtering was performed directly on all exome data according to a recessive inheritance model. Potentially pathogenic variants (missense, non-sense, splice site and frameshifts) having <0.05 minor allele frequencies (1000 genome project or Exome Variant Server-University of Washington-EVS) were considered for subsequent analysis. Thus, we considered either homozygous or compound heterozygous variants present in the affected sisters, being at a heterozygous state or absent in C3. The variant c.2355+1G>A in MSH4 fulfilled the above criteria and its presence in P1 and P2 was confirmed by PCR and direct sequencing. Primer sequences and PCR conditions are available upon request. The MSH4 sequence variant was also genotyped by Sanger sequencing in two healthy brothers (II:1 and II:5) and in the C-POI group. Netgene2 (v2.4) and Human Splicing Finder (v3) and Mutalyzer bioinformatics tools were used for predicting alternative transcripts generated by the MSH4 c.2355+1G>A splice site variant

Exon trapping/minigene assays

To test the effect of the candidate variant c.2355+1G>A in MSH4 on splicing, amplicons, generated by standard overlapping PCR and digestion/ligation procedures were cloned into the pCDNA 3.1/V5-His-TOPO TA vector (Life Technologies) and its sequence verified by Sanger sequencing. The cassette included exon 16 (with an artificial ATG to generate a potentially translatable open reading frame to stabilize the transcript), intron 16, exon 17, the first 183 bases of intron 17 containing (MUT) or not (WT) the c.2355+1G>A substitution, then 743 bases located at the 5′ region of exon 18 and exon 18 with an artificial stop codon. Constructs were transfected into HeLa cells using a standard calcium phosphate protocol. Cells were harvested 48 h after transfection and total RNAs were extracted with TRIzol (ThermoFisher) according to the manufacturer’s protocol. Reverse transcription was performed using SuperScript II (ThermoFisher). RT-PCR were performed using Herculase II Fusion DNA Polymerase (Agilent). Amplicon sequencing was performed at Eurofins Genomics according to their in-house procedures.

Acknowledgements

We sincerely thank Dr S. Caburet for her comments on this manuscript.

Conflict of Interest statement. None declared.

Funding

This work was supported by the Universidad del Rosario grant: [CS/ABN062/GENIUROS 017] to P.L. and from the Fondation pour la Recherche Médicale grant: [DEQ20150331757] to R.A.V.

Author notes

†

These authors contributed equally to this work.