Introduction

Disorders of sex development (DSDs) cover a wide spectrum of phenotypes, ranging from complete sex reversal to ambiguous genitalia, the latter affecting 1 in 4500 births.1

Patients with 46,XY partial or complete gonadal dysgenesis have either underdeveloped testes, ovotestes (partial gonadal dysgenesis) or streak gonads (complete gonadal dysgenesis). Disruption of testis development leads to either undervirilization of the external genitalia, ambiguous genitalia or a female external phenotype. The clinical management of these complex conditions is made difficult because up to 70% of 46,XY gonadal dysgenesis cases lack a diagnosis and the underlying molecular cause remains unknown.2 However, point variants, deletions or duplications of genes such as sex-determining region Y (SRY), SRY-related HMG-box, gene 9 (SOX9), nuclear receptor subfamily 5, group A, member 1 (NRA51), and mitogen-activated protein kinase kinase kinase 1, E3 ubiquitin protein ligase (MAP3K1) have been shown to be responsible for 30% of 46,XY DSD cases.3

Premature ovarian insufficiency (POI) in 46,XX individuals is characterized by early failure of ovarian function, affecting 1% of women.4 In the majority of 46,XX POI cases, the underlying genetic cause remains unknown. Few genes such as autoimmune regulator (AIRE; causing autoimmune polyendocrinopathy-candidiasis-ectodermal dystrophy (APECED) syndrome) and forkhead box L2 (FOXL2) have been associated with 46,XX POI in a minority of patients.5

Studies in mouse have identified Nr5a1 as a key gene involved in gonadal sex determination, differentiation, and maintenance.3 Nr5a1 encodes steroidogenic factor 1 (SF1), an orphan nuclear receptor expressed early in the bipotential gonad, and later in the developing and mature testis and ovary. Nr5a1-null XY mice show streak gonads and feminization, which mimics the phenotype of patients affected by 46,XY complete gonadal dysgenesis. Genes downstream of NR5A1 in the testis and ovary have been identified by analysing different mouse strains and human fetal gonads. These include genes required for testis and ovary development, such as SOX9, anti-Müllerian hormone (AMH), and nuclear receptor subfamily 0, group b, member 1 (Nr0b1 or Dax1), as well as, in particular, gonadal steroidogenesis.6, 7, 8

Variants in and around NR5A1 are causative for 46,XY gonadal dysgenesis.9, 10 A recent study analysed four families with individuals exhibiting 46,XY gonadal dysgenesis and female relatives affected by 46,XX POI.11 NR5A1 was found to be the causative gene for both phenotypes in all four families, representing the first link of NR5A1 to 46,XX POI. The authors also found variants in NR5A1 in two sporadic 46,XX POI cases. Another study identified 10 novel variants in NR5A1 in a cohort of sporadic cases of 46,XY gonadal dysgenesis and 46,XX POI.12 Here we describe the analysis of a multigeneration family affected by 46,XY DSD and 46,XX POI, using low-coverage whole exome sequencing combined with linkage analysis. This approach was extremely powerful in quickly refining the search area within the exome down to the nucleotide level, allowing rapid identification of a novel 3-bp deletion in NR5A1 as the causative variant in this family. In addition, we analysed the function of the mutant SF1 protein to investigate the biologic consequences of the lesion.

Materials and methods

Clinical data: family members with 46,XY DSD

Case II:8: The 46,XY DSD patient II:8 (later referred to as ‘unclassified 46,XY DSD’) presented with ambiguous genitalia and gynaecomastia, severe penoscrotal hypospadias, bifid scrotum, micropenis, small bilateral testis (2–3 ml), and primary infertility. Owing to the mainly male external genitalia, he was raised as a male.

Case III:1: This 46,XY gonadal dysgeneis patient presented with atrophy of testis (bilateral small inguinal palpable testis; <1 ml), with labia major (fusion of labia major, which resembles a scrotum and is darker in colour) and small labia minor, blind ending vagina, and an enlarged clitoris. The child was raised as a girl. Abdominal testes were removed and the patient will have lifelong hormonal treatment. Pathology results from the removed testis revealed low number of Leydig cells and atrophy of seminiferous tubules. No malignant cells and spermatozoa, but few spermatid cells were detected. Low testosterone levels (8 ng/dl before and <20 ng/dl after gonadectomy) were measured.

Case III:2: Presented with 46,XY gonadal dysgenesis, micropenis (phallus length 5.4 cm at the age of 10 years), Quigley stage 2, Tanner stage V, severe scrotal hypospadias (meatus at glans), bifid scrotum, severe chordee, and small bilateral testis (both in scrotum and both 5 ml volume). The patient underwent plastic surgery three times to repair the hypospadias and was raised as a male. No uterus or ovary could be detected.

Clinical data: family members with 46,XX POI

The four female relatives who presented with 46,XX POI (I:2, II:2, II:4, and II:7) were all diagnosed at different ages (II:4 onset at 35, and II:7 at 24 years of age; for I:2 and II:2 the age of onset is unknown), by onset of secondary amenorrhoea.

Patient DNA samples

DNA from the 15 family members was extracted from peripheral EDTA blood samples according to a standard protocol (desalting method). Approval for this study was obtained by the Human Ethics Committee of the Faculty of Medicine, Diponegoro University, Semarang, Indonesia. All patients and family members gave their informed consent before taking part in this study. This study also received approval from the Human Ethics Committee at the Royal Children’s Hospital, Melbourne, Victoria, Australia.

Exome capture, sequencing, and bioinformatic analysis

Exome capture (TruSeq Exome Enrichment Kit; Illumina Inc., San Diego, CA, USA) and sequencing (Illumina HiSeq 2000; Illumina) were performed by the Australian Genome Research Facility (Melbourne, VIC, Australia). The 100 bp paired-end reads were aligned to UCSC hg19 using Novoalign version 2.07.09 in paired-end mode (http://www.novocraft.com). Reads mapping to multiple locations were discarded; presumed PCR duplicates were discarded using the MarkDuplicates utility from Picard (http://picard.sourceforge.net/). Variants were detected using the mpileup and bcftools view commands from SAMtools version 0.1.17,13, 14 specifying parameters −C50 and −q13. Low confidence variants were discarded using the vcfutils.pl varFilter script from the same program. Variants were annotated against the UCSC KnownGene annotation, dbSNP132, and the May 2011 release of the 1000 Genomes Project15 using ANNOVAR.16 In silico analysis of the functional effects of coding non-synonymous variants was performed with SIFT17 and HumVar-trained PolyPhen-2 v.2.1.0.18 We used SAMtools to infer, from the exome sequence data, genotypes at the location of HapMap Phase II and III SNPs, specifying parameters −cg and −t0.5.19 Parametric multipoint linkage analysis was performed using MERLIN20 under a fully penetrant autosomal dominant model with a 0% phenocopy rate and a disease allele frequency of 0.00001. SNP allele frequencies were obtained from HapMap CEU (Utah residents with ancestry from northern and western Europe from the CEPH collection) genotypes.

In silico prediction on the wild-type and mutant protein

To predict changes between the wild-type and mutant protein structure, an in silico prediction of the two protein structures was performed using the protein comparative modelling server 3D-JIGSAW (http://bmm.cancerresearchuk.org/~3djigsaw/).

Plasmids

The mammalian expression vector pCMV6-Entry-hNR5A1 (RC207577; OriGene Technologies Inc., Rockville, MD, USA) containing the human cDNA ORF of NR5A1 was used to obtain the expression vector containing the mutant NR5A1 cDNA (pCMV6-Entry-hNR5A1-p.(Lys372del)). Site-directed mutagenesis was performed using the QuickChange II XL Site-directed Mutagenesis Kit (Agilent Technologies Inc., Santa Clara, CA, USA) according to the manufacturer’s instructions (for primer sequences see Supplementary Table S5). To create the pGL4.10-hNR0B1 construct for dual-luciferase reporter assays, the human NR0B1 (hNR0B1) promoter was PCR amplified (for PCR primer sequences see Supplementary Table S5) and cloned into the pGL4.10 vector (Promega Corporation, Madison, WI, USA) using the XhoI and HindIII restriction sites within the multiple cloning site.

Cell culture, transfections, and preparation of nuclear extracts

COS7 (monkey kidney cell line), KK-1 (mouse Granulosa cell line), 15P-1 (mouse Sertoli cell line), and Hek293 (human embryonic kidney) cells were maintained in D-MEM supplemented with 10% FBS and 1.8 g/l NaHCO3 and incubated at 37 °C. For nuclear extracts, cells were transfected with 30 μg of either the wild-type or mutant (pCMV6-Entry-hNR5A1 or pCMV6-Entry-hNR5A1-p.(Lys372del)) construct. The transfection reagent Lipofectamine 2000 (Invitrogen, Thermo Fisher Scientific Inc., Waltham, MA, USA) was used for transfections. Cells were harvested 48 h posttransfection and nuclear extracts were prepared using the NE-PER Nuclear and Cytoplasmic Extraction Reagents (Thermo Fisher Scientific Inc.). Protein concentrations were determined using the Pierce BCA Protein Assay Kit (Thermo Fisher Scientific Inc.). Absorbance at 562 nm was measured using the Infinite M200 Pro plate reader (Tecan, Männedorf, Switzerland).

SDS-PAGE and western blot

Fifteen micrograms of nuclear extract was diluted to 30 μl with H2O, 10 μl of 4 × loading buffer (50% (v/v) glycerol, 0.16 M Tris-HCl (pH 6.8), 2% (w/v) SDS, 0.01% (w/v) bromophenol blue, and 350 mM DTT) was added, and the samples were denatured at 95 °C for 15 min. Total protein of 7.5 μg and 5 μl of Precision Plus Protein Dual Color Standard (Bio-Rad Laboratories, Hercules, CA, USA) were loaded onto a NuPAGE 10% Bis-Tris gel. SDS-PAGE was performed in 1 × NuPAGE MOPS running buffer (Invitrogen) at 100 V for 2 h. After electrophoresis, proteins were blotted onto a PVDF membrane (GE Healthcare Bio-Sciences, Pittsburgh, PA, USA) using 1 × NuPAGE Transfer buffer (Invitrogen) with 20% methanol overnight at 4 °C at 100 mA. Blocking and incubation with primary (1:1000 dilutions) and secondary antibodies (donkey anti-rabbit AlexaFlour 488 and donkey anti-goat AlexaFluor 488; Invitrogen) was performed as described previously. Fluorescence signals were visualized using a Typhoon Trio Variable Mode Imager (GE, Fairfield, CT, USA).

Electrophoretic-mobility shift assays

Fluorescently labelled oligonucleotide (0.5 nmol each) and the corresponding unlabelled synthetic complementary strand, containing the mouse Nr0b1 SF1-responsive element (other SF1-responsive elements were also tested – data not shown; for oligosequences see Supplementary Table S5), were annealed and diluted to 10 ng/μl in 1 × TE buffer. Fifteen micrograms of either the wild-type or mutant SF1 overexpression nuclear extracts were incubated in 2.5 × binding buffer21 for 10 min at RT. One microgram of poly(dI–dC) (Roche Diagnostics Pty Ltd, Rotkreuz, Switzerland), 5 μl of a rabbit anti-FLAG antibody (for supershifts) (Sigma-Aldrich, Castle Hill, NSW, Australia), and water were added to a final volume of 48 μl and incubated for 30 min. Two microliters (20 ng) of double-stranded probes were added to a final volume of 50 μl. The samples were incubated for 30 min and the reaction stopped by adding 5 μl of 6 × loading buffer (250 mM Tris-HCl, 40% (v/v) glycerol, and 0.01% (w/v) Orange G). Twenty microliters of each reaction was loaded onto a 7.5% Mini-Protean TGX gel (Bio-Rad) and DNA–protein complexes were separated by electrophoresis at 100 V for 90 min. Fluorescence signals were visualized using a Typhoon Trio Variable Mode Imager (GE).

Dual-luciferase reporter assays

Cells were split onto 96-well plates (Nunc) 12–24 h before transfection and transfected with 75 ng of the reporter construct, pGL4.10-hNR0B1, and 15 ng of the Renilla construct (pRL-TK) for normalization of transfection efficiency. For the transactivation assay, cells were also co-transfected with 50 ng of either wild-type or mutant pCMV6-Entry-hNR5A1 construct. To adjust for differences in the amount of DNA added to each reaction, an independent, non-active plasmid (pSLAX) was used to adjust total DNA amounts to 240 ng for each reaction. Lipofectamine 2000 (Invitrogen) was used as a transfection reagent. Cells were harvested 48 h posttransfection and luciferase activity was measured using the Dual-Luciferase Reporter 1000 Assay System Kit (Promega) according to the manufacturer’s instructions. Chemiluminescence was measured on an Infinite M200 Pro plate reader (Tecan). Each experiment was performed in biologic triplicate.

Results

Large families are rare in DSD research, because most DSDs result in infertility. As large pedigrees are an extremely powerful tool for genetic linkage analysis, we decided to sequence the entire exome of all 15 individuals from a multigeneration family from Indonesia (Figure 1a).

Figure 1
figure 1

A novel 3-bp deletion in exon 6 of NR5A1 (exons are numbered as in NG_008176.1) was identified in a multigeneration Indonesian family to be associated with 46,XY DSD and 46,XX POI. (a) Pedigree of the family, showing all 15 individuals who were available for this study. Squares and circles represent males and females, respectively. Open symbols indicate unaffected individuals, and filled squares and circles highlight affected individuals.The 46,XY gonadal dysgenesis (46,XY GD) patients were assigned male or female symbols according to their phenotype and assigned gender. The 46,XX POI patients are labelled with 46,XX POI below the symbol. The pedigree has been reduced to the individuals who were available for the study for simplicity and for the sake of confidentiality. (b) Schematic view of the NR5A1 genomic region on chr9:g.127 243 515–127 269 699 (GRCh37/hg19) including the exon–intron structure (NR5A1 is located on the reverse strand, but shown here in its 5′–3′ orientation from left to right). Exon 6 is highlighted in red. (c) Sanger sequencing of an unaffected and affected individual, showing the wild-type exon 6 sequence (unaffected) and the heterozygous, 3-bp deletion (c.1114_1116del) in one affected individual.

Exome capture, sequencing, and bioinformatic analysis

Picard MarkDuplicates software estimated rates of PCR duplication to be very high, ranging from 83.3 to 96.7% of uniquely aligned reads (Supplementary Table S1). As a result, coverage of targeted bases following duplicate removal was very low (Supplementary Table S2), with median coverage ranging between 2 and 9 (median 3).

We attempted to infer genotypes at the locations of 4 071 900 HapMap Phase II and III SNPs. Genotypes were set to missing, unless an SNP was covered by ≥5 reads with base and mapping qualities ≥13. We discarded SNPs where genotypes were missing for more than 10 of the 15 exomes, resulting in 57 666 SNPs being retained. We then discarded 414 SNPs where genotypes were inconsistent with Mendelian inheritance and 37 685 SNPs where non-missing genotypes did not vary between individuals. Finally, we chose a subset of 4814 SNPs in approximate linkage equilibrium for linkage analysis;22 on average, CEU individuals are heterozygous for 38% of these markers. Individual genotyping rates for these SNPs ranged between 0.8 and 96.7% (median 21.5%) (Supplementary Table S3).

Linkage analysis identified two peaks on chromosomes 9 and 18 achieving maximum LOD scores of 2.69 and 2.42 (Figure 2). Positive LOD scores were obtained between 128.4 and 159.6 cM (rs1930713–rs4842247) on chromosome 9 and between 29.1 and 48.0 cM (rs767300–rs17448282) on chromosome 18. Given the high number of genotypes missing, we examined the robustness of our results by using MERLIN’s – sample option to sample 10 sets of haplotypes according to their likelihood (eg Supplementary Figure S1). This allows assessment of the uncertainty of the haplotype inference; the less an individual’s inferred haplotypes vary across samples, the more confidence we have in the haplotyping. For the chromosome 9 peak, all 10 sets of haplotypes were consistent with the autosomal dominant model for 14 of the family members (eg Supplementary Figure S1), indicating high confidence in the results. Inference could not be performed for the remaining individual, affected II:7, for whom all genotypes within the linkage region were missing. For the chromosome 18 peak, only two haplotype sets accorded with the genetic model for 14 family members, indicating that we cannot be very confident of linkage. Unaffected II-3 was missing all genotypes within the linkage region.

Figure 2
figure 2

Genome-wide parametric LOD scores under a fully penetrant autosomal dominant inheritance model for the Indonesian family.

Of the 300 591 variants detected in at least one of the 15 exomes, 4453 variants were located within one of the two linkage peaks; 1148 of these were rare (not present or present at a frequency ≤0.01 in the 1000 Genomes Project). Of these, 45 variants were predicted to affect protein sequence and/or splicing. Given the low coverage, we did not apply a strict genotype filter; instead, we retained variants that were not detected in the eight unaffected exomes, but were detected in at least two affected exomes. This reduced the list of candidate variants to five (Supplementary Table S4). One of these variants, a 3-bp deletion removing a single amino acid, is located within exon 6 of NR5A1 (NM_004959.4:c.1114_1116del [p.(Lys372del)]; data have been submitted to the NR5A1 gene variant database at http://www.LOVD.nl/NR5A1, patient ID 0063642.) (see Figure 1b).

Supplementary Figure S2 shows read alignments at the location of the NR5A1 variant for affected and unaffected family members with coverage at this location. Sanger sequencing indicated that all seven affected family members were heterozygous for the deletion, whereas the six unaffected family members were all wild type, as were the two girls considered at risk of developing 46,XX POI (III:3 and III:4; see Figure 1b and c).

In silico prediction on the wild-type and mutant SF1 protein

The heterozygous, 3-bp in-frame c.1114_1116del in exon 6 of NR5A1 results in deletion of the lysine at position 372 (p.(Lys372del)) of SF1, within the predicted ligand-binding domain (LBD; see Figure 3a). Although recent publications have described phosphatidylinositols, sphingosine, and phospholipids as potential ligands for SF1, their biologic role remains to be elucidated.23, 24, 25 To investigate the potential biologic relevance of the p.(Lys372del) variant, we performed an in silico prediction on the wild-type and mutant proteins using the online protein comparative modelling server 3D-JIGSAW (http://bmm.cancerresearchuk.org/~3djigsaw/). Direct comparison of the two predicted protein structures (Figure 3a–c) revealed truncation of helix 8 of the LBD at position 372 in the mutant protein. Furthermore, the loss of the β-pleated sheets in the LBD and the DNA-binding domain (DBD) indicated a potential loss of the DNA-binding ability of the mutant protein compared with the wild-type protein. Also, the tertiary structure of the whole protein was predicted to change slightly, indicating a crucial role of Lys372 in the wild-type SF1 protein structure.

Figure 3
figure 3

Protein structure of SF1. (a) Schematic view of the different domains of the SF1 protein. The approximate locations of previously published variants within NR5A1 in families affected by 46,XY DSD and 46,XX POI and sporadic cases of 46,XX POI are shown below the protein structure as resulting protein changes.11, 12 The novel 3-bp deletion in exon 6 of NR5A1 (p.(Lys372del)), located within the predicted LBD, is shown above the protein structure. (b) In silico predictions of the wild-type and (c) the p.(Lys372del) mutant SF1 proteins. The LBD is highlighted by green boxes, whereas the DBD is highlighted by blue boxes. Comparison of the wild-type and mutant SF1 structures indicates that the deletion of Lys372 results in the truncation of helix 8 (highlighted with white boxes) of the LBD at this position. Further changes in the secondary structure of the protein include the loss of the β-pleated sheets in both the LBD and the DBD, indicating a potential loss of the DNA-binding ability of the mutant compared with the wild-type SF1 protein.

Functional assays of SF1 activity

We assessed whether the mutant SF1 protein was correctly translocated into the nucleus. Using immunofluorescence, we confirmed that import of the mutant SF1 protein into the nucleus was not affected by the p.(Lys372del) variant (data not shown).

To investigate the potential effect of the p.(Lys372del) variant on the DNA-binding ability of SF1, as predicted by 3D-JIGSAW, we performed electrophoretic mobility shift assays (EMSAs) and dual-luciferase reporter assays using known SF1-responsive elements. Figure 4b shows the results for the wild-type and mutant SF1 proteins, using the mNr0b1 (mDax1) SF1-responsive element.21 Western blot analysis was used to ensure similar levels of wild-type and mutant SF1 protein, nuclear extracts were loaded (see Figure 4a). The binding of wild-type SF1 to the mNr0b1 SF1-responsive element yielded two signals as the probe contains two overlapping SF1-binding sites, one on each of the DNA strands. The lower band represents one molecule of SF1 and the second higher band represents two SF1 molecules bound to the probe (see Figure 4b). To confirm that both signals were caused by SF1 binding, we performed supershifts, using an anti-FLAG antibody, which recognizes both the wild-type and mutant SF1 proteins (expressed via the pCMV6-Entry-hNR5A1 vector, which contains C-terminal DDK and c-Myc tags). We obtained two supershift signals, caused by one or two SF1 molecules bound to the responsive element. The shift and supershift experiments with the mutant SF1 overexpression nuclear extracts produced one very weak shift and supershift signal (representing one SF1 molecule bound), and loss of the second signal (two SF1 molecules bound) (see Figure 4b). This indicates that the p.(Lys372del) mutant protein bound with a lower efficiency than the wild-type protein to SF1-responsive elements in vitro.

Figure 4
figure 4

Functional assays on SF1 activity using the mNr0b1 and hNR0B1 SF1-responsive element and promoter. (a) Western blot of the wild-type and mutant SF1 overexpression nuclear extracts (NE mut and NE wt, respectively) showing similar expression levels of the two proteins. Proliferating cell nuclear antigen (PCNA) was used as loading control. (b) EMSA using the mouse Nr0b1 (mDax1) SF1-responsive element. Lane one shows the signal for the probe, and lane two shows the signals obtained by adding the wild-type nuclear extract (wt NE). Two signals are detected owing to the composite SF1-binding site within this element. Lane three shows the supershifts, using an anti-FLAG antibody. Lanes 4 and 5 show the results for the mutant nuclear extract (mut NE). Both signals and supershifts, owing to SF1 binding to the probe, which were observed with the wild-type nuclear extract, are almost completely lost when the mutant nuclear extract is added. (c) Dual-luciferase reporter assays using the hNR0B1promoter. In all three cell lines (15P-1, KK-1, and Hek293), an increase of luciferase expression can be detected when the wild-type SF1 expression vector is added to the reporter vector (pGL4.10-hNR0B1), whereas addition of the mutant construct results in little or no change in luciferase activity.

To confirm these results, we performed dual-luciferase reporter assays using the hNR0B1 (hDAX1) promoter (Figure 4c). When co-transfecting the pGL4.10-hNR0B1 promoter vector with the wild-type pCMV6-Entry-hNR5A1 overexpression vector, we found a 2- to 7.5-fold increase in luciferase activity compared with the reporter construct-only transfections. The fold increase in the luciferase activity varied between the three cell lines (15P-1, KK-1, and Hek293) tested. When transfecting with the mutant overexpression construct (pCMV6-Entry-hNR5A1-p.(Lys372del)), luciferase activity remained at similar levels to the promoter construct-only transfections. These data support the results from the EMSAs, where the mutant SF1 protein did not bind SF1-responsive elements in vitro. Furthermore, the luciferase data showed that the mutant SF1 protein failed to activate SF1-responsive promoters in vitro.

Discussion

In this study, we identified a novel, 3-bp, in-frame deletion located in exon 6 of NR5A1 in all seven affected family members (46,XY DSD (unknown and gonadal dysgenesis) and 46,XX POI). As NR5A1 had been previously described to cause both phenotypes11, 12 and was inherited in the proposed autosomal dominant manner, we focussed on this variant as the most likely explanation for these clinical phenotypes.

In silico prediction of the wild-type and mutant predicted SF1 protein structures suggested that the mutated lysine 372 has an important structural role. Deletion of this single amino acid is predicted to result in the truncation of helix 8 of the LBD and the loss of all β-pleated sheets, in the LBD, as well as in the DBD. It is also predicted to affect the entire tertiary protein structure, potentially explaining the observed loss of the DNA-binding capability of the mutant SF1 protein. The EMSAs and dual-luciferase reporter assays with known SF1-responsive elements of downstream targets of SF1, such as mNr0b1/hNR0B1 (mDax1/hDAX1), confirmed the in silico prediction, showing a marked reduction in the DNA-binding and transactivation ability of the mutant SF1 protein compared with the wild-type SF1 protein in vitro. This indicates that the mutant SF1 protein present in the affected family members fails to bind and activate downstream targets of SF1.

Three other variants within the LBD of NR5A1 have been described in sporadic cases of 46,XX POI (c.691_699delCTGCAGCTG (p.(Leu231_Leu233del)) and c.704C>T (p.(Pro235Leu))) or familial cases of 46,XY DSD and 46,XX POI (c.877G>A (p.(Asp293Asn))).11, 12 A direct comparison of the three other published variants and our variant is difficult, as different SF1-responsive promoters and cell lines were used; however, it appears that single amino-acid exchanges within the LBD lead to a moderate (c.877G>A (p.(Asp293Asn)) (Cyp11a1 and Cyp19a1 promoters); c.704C>T (p.(Pro235Leu)) (Cyp17a1 promoter)) or no significant reduction (c.704C>T (p.(Pro235Leu)) (CYP11A1 and HSD3B2 promoters) of the activation of SF1-responsive promoters. By contrast, the effect on the activation of SF1-responsive promoters of the single amino-acid deletion c.1114_1115del (p.(Lys372del)) described here corresponds to the effect seen with larger deletions c.691_699delCTGCAGCTG (p.(Leu231_Leu233) (Cyp11a1 and Cyp19a1 promoters)) or variants found in the DBD. This further supports the in silico prediction and the proposed structural importance of lysine 372.

The two DSD conditions (46,XY DSD (unknown and gonadal dysgenesis) and 46,XX POI) caused by the same variant can be explained by the different roles of SF1 in the developing testis and ovary. In testes, SF1 is expressed in Leydig and Sertoli cells and is required to upregulate SRY and thus to establish the male pathway, resulting in the bipotential gonad developing into a testis.26 SF1 has also been shown to synergize with SRY to upregulate SOX9 and with SOX9 to upregulate Cyp26b1.6, 27 In addition, SF1 is required to upregulate AMH expression and causing the regression of the Müllerian ducts in XY male embryos.28 In the ovary, NR5A1 is expressed in theca and granulosa cells and its expression can be detected in the embryonic, postnatal, prepubertal and adult ovary.29, 30 SF1 has been shown to regulate key genes of ovarian steroidogenesis. SF1 clearly acts at multiple points during gonad development, differentiation, and in crucial steroidogenic pathways even into adulthood. Given this, it is not surprising that the same variant in NR5A1 can cause testis failure and 46,XY DSD (including gonadal dysgenesis) in male family members, whereas in female relatives it causes 46,XX POI.

We also observed that the same variant caused different external genitalia in the three individuals with 46,XY DSD. II:8 and III:2 had predominantly male genitalia and were raised as boys, whereas III:1 had predominantly female external genitalia at birth and was raised as a girl. This variation is most likely explained either by incomplete penetrance and/or by SNPs or variants in modifying genetic factors that have yet to be identified. MPS studies on larger numbers of DSD patients will help identify the missing genes and networks, which, when disrupted, cause DSDs. Our findings in this family are entirely consistent with earlier studies, which associated variants in NR5A1 with 46,XY gonadal dysgenesis and 46,XX POI in families. High variability in the phenotype of the external genitalia in familial cases of 46,XY DSD has also been reported.31

This study has shown the particular utility of using low-coverage MPS and linkage analysis to identify rapidly the causative variant. In an effort to reduce costs, many groups select a few individuals from large pedigrees for MPS at high coverage. As an alternative, the approach used here involved sequencing more individuals to ensure sufficient power for linkage analysis, combined with cheaper low-coverage MPS to identify the disease-causing genes at a similar cost. The identification of the molecular cause and the mode of inheritance of 46,XY DSD (including cases of 46,XY gonadal dysgenesis) and 46,XX POI in families such as this allows for the provision of genetic counselling to female family members under or at a reproductive age, as they may carry a dominant variant, but have not presented with secondary amenorrhoea. The two girls (III:3 and III:4) in this family are in this situation. Although their mother (II:4) carried the variant in NR5A1 and was affected by 46,XX POI, both her daughters tested negative for the variant and were not at risk for POI.