Exome Sequencing for the Diagnosis of 46,XY Disorders of Sex Development

Abstract

Sex determination initiates when the bipotential gonad's genetic program determines the formation of either an ovary or testis. Subsequent differentiation of the internal and external genitalia is controlled by locally secreted and circulating sex hormones. Disruption of either determination or differentiation can lead to a disorder of sex development (DSD), ie, a discrepancy between an individual's chromosomal sex and phenotypic sex (1). Although an accurate genetic diagnosis and better understanding of genotype-phenotype correlations will offer a clearer prognosis to families, many DSD patients still do not receive a genetic diagnosis.

In 46,XY individuals, defects in testis determination often result in gonadal dysgenesis and can be caused by the loss of function of SRY (2) or NR5A1 (3). However, variants in these genes only account for 10–15% of cases each, leaving most 46,XY gonadal dysgenesis cases undiagnosed at the genetic level. 46,XY DSDs caused by defects of differentiation are most often due to disruption of sex hormone synthesis or receptors, such as variants in the androgen receptor (AR) (4). They are often diagnosed clinically by detection of alterations in circulating hormone levels (5) but are not always explained by variants in known genes.

Previously, we developed a targeted capture approach for 35 known DSD genes (6). This approach confirmed the genetic diagnosis in a known group of samples and identified a genetic cause in two of five previously undiagnosed patients. Here we have expanded this approach by using exome sequencing to capture almost all coding exons, followed by bioinformatic filtering using a comprehensive DSD gene list. The exome covers approximately 95% of RefSeq genes, thus covering most protein-coding sequence, which currently harbors 80–90% of known disease-causing variants (7). Therefore, all genes with any involvement in sex development can be analyzed concurrently, and new genes can be included in the analysis without having to reconfigure the sequencing pipeline or resequence the samples. We present data from a group of 40 46,XY DSD patients sequenced at the University of California-Los Angeles (UCLA) Clinical Genomic Center and analyzed using a gene list.

Materials and Methods

Samples were submitted to the UCLA research laboratory under an Institutional Review Board-approved protocol (no. 11-001775-AM-00007; Principal Investigator, E. Vilain) or to the UCLA Clinical Genomics Center. Exomes were captured using SureSelect All Exon 50 Mb capture kit (Agilent Technologies) and sequenced on a HiSeq2000 or HiSeq2500 (Illumina, Inc) as 50-bp or 100-bp paired-end runs. Base-calling was performed using Illumina's real-time analysis software. Sequence reads (QSEQ or FASTQ files) were aligned to the human reference genome (hg_g1k_b37 assembly) using Novoalign V2.07.13 (http://www.novocraft.com/main/page.php?s=novoalign). The output BAM file was sorted and merged, and PCR duplicates were removed using Picard (http://picard.sourceforge.net/). INDEL (insertion and deletion) realignment and recalibration were performed using the Genome Analysis Tool Kit (GATK) (http://www.broadinstitute.org/gatk/). Mean coverage was over 80 × for each sample, and approximately 93% of the RefSeq gene coding regions ± 2 bp was covered at 10 × or greater (individual gene coverage is indicated in Table 1). Single-nucleotide variants and small INDELs were called using GATK's Unified Genotyper, then recalibrated and filtered using GATK variant-quality score recalibration and variant filtration tools. Consanguinity analysis was performed by identifying regions >1 Mb of homozygosity using Linkdatagen (http://bioinf.wehi.edu.au/software/linkdatagen/) (8) and Plink software (http://pngu.mgh.harvard.edu/∼purcell/plink/) (9). All high-quality variants were annotated using Variant Annotator X, a custom-designed variant effect predictor (10).

Variants with a minor allele frequency of <1% in the Exome Sequencing Project (ESP) of more than 6500 individuals were intersected with a DSD gene list to identify mutations in known DSD genes. The gene list (Table 1) was generated by combining the genes included in our capture panel (6) with a search of online databases such as OMIM, HGMD professional, and GeneTests using the key word “sex.” HGMD contains information about genes and variants that have been identified in human disease, and findings from HGMD are considered “clinical genes.” OMIM contains information from both human disease and animal models. Thus, these two databases are overlapping, but each contains information not in the other, so using both generates the longest list. This list was curated so that all genes included were published in at least one human case of DSD, and it is dynamic so it can be updated as soon as new findings are published. When parental samples were available, sequencing results were filtered to identify all de novo, homozygous, and compound heterozygous variants, even if the variants were not within the known DSD gene list.

Sanger sequencing using custom-designed primers was used to confirm the exome sequencing results for all research samples (except case RDSD005, where insufficient sample remained). As of May 2013, a UCLA Clinical Genomics Center retrospective data analysis showed that Sanger sequencing confirmation was unnecessary for single-nucleotide variants with high exome sequencing quality score (11). All INDELs are validated by Sanger sequencing.

Using American College of Medical Genetics and Genomics guidelines, we classified variants into five main categories: pathogenic, likely pathogenic, variants of uncertain clinical significance (VUS), likely benign, and benign (12). Premature termination codons and splice site variants are considered mutations by definition (12). Variants are also called pathogenic if previously reported in a similar clinical phenotype. Novel variants in genes related to the clinical phenotype that are predicted to be damaging are classified likely pathogenic. To assess the possible impact on protein structure and function, we used in silico algorithms SIFT (13), PolyPhen2 (14), and Condel (15). Unless otherwise stated, all novel variants discussed here were predicted pathogenic by all three algorithms. Single variants in apparently dominant conditions were only considered if present in the ESP at less than 0.1%.

Results from the Clinical Genomics Center are evaluated by the UCLA Genomic Data Board, a team of experts that meets weekly to analyze exome findings. The Board consists of the center's three directors, laboratory professionals, American Board of Medical Genetics and Genomics board-certified geneticists, genetic counselors, and clinicians including, if possible, the referring physician for each case. All variant calling is discussed and ultimately decided by this interdisciplinary group, a great strength of the UCLA Clinical Genomics Center.

Results

We report the results of exome sequencing in individuals with a 46,XY karyotype and a range of DSD phenotypes. This data set contains all samples of 46,XY DSD submitted to our research lab that did not have a genetic diagnosis after all other testing methods had been exhausted and the first 13 sequential samples submitted to the UCLA Clinical Genomics Center for testing of 46,XY DSD. Material of X and Y origin was confirmed by exome sequencing. When known, phenotypic characteristics and results of previous genetic testing history are described in Table 2. The range of presenting phenotypes was wide, as is typical of DSD, with external genitalia classified as typical female with or without clitoromegaly (21 cases), ambiguous (12 cases), or typical male with or without micropenis (7 cases). Seven patients had associated nongenital malformations, not representing an easily recognizable syndrome.

Variants in MAP3K1

One of the most striking findings in our study was the identification of MAP3K1 variants in a total of four cases. Variants in this gene have recently been associated with complete gonadal dysgenesis (16), and two of our cases had the same previously reported variant p.Gly616Arg. Patient RDSD014 was a female who presented in adolescence with complete gonadal dysgenesis. In contrast, patient RDSD023 was a male with ovotesticular DSD, ascertained at birth due to the presence of ambiguous genitalia, a finding not previously associated with this variant. We also identified novel, likely pathogenic variants in two additional patients. Patient RDSD017 had complete gonadal dysgenesis and a de novo p.Arg339Gln missense variant. Patient CDSD040 had a p.Pro257Leu missense variant predicted damaging by two of the three in silico algorithms (SIFT and PolyPhen) and presented as a male with complex ambiguous genitalia but no gonadal dysgenesis (Table 2). Pearlman et al (16) examined only patients with complete gonadal dysgenesis, whereas our study included a wider range of 46,XY DSD phenotypes. Most MAP3K1 variants so far identified cluster in exons 2–4, and the p.Gly616Arg is in exon 10; thus, there is no obvious genotype-phenotype correlation.

Variants in WT1

We found two variants in WT1, a gene associated with 46,XY gonadal dysgenesis in several conditions including Denys-Drash syndrome (17). Patient RDSD024 presented with end-stage renal failure and Denys-Drash syndrome in the differential diagnosis. We identified a novel likely pathogenic p.His469Gln missense variant located in exon 9 of WT1, the location and type of variants most often associated with Denys-Drash syndrome (18). RDSD019, a patient with similar genital features, also had a novel missense variant (p.Arg458Gln) in exon 9 of WT1. Subsequent testing of parental samples showed that the variant was inherited from the unaffected father, making it less likely to be pathogenic. However, a new publication reported a familial case of Denys-Drash syndrome with the well-established exon 9 p.Arg394Trp variant identified in both the proband and his unaffected father (19), suggesting incomplete penetrance. With this report of incomplete penetrance in a case of an established WT1 disease-causing variant, we decided that the p.Arg458Gln was in fact likely causative of the proband's phenotype and exhibits reduced penetrance in the apparently unaffected father. In consequence, both of the novel WT1 variants identified in our study are likely the cause of the observed phenotype.

STAR variant and adrenal insufficiency

In a phenotypically female patient with suspected adrenal insufficiency and absent uterus (RDSD008), we found a homozygous variant at c.64+1G>A in the STAR gene. Splice site variants are considered mutations by definition because they generally result in a truncated protein (12). Homozygous mutations in STAR are associated with 46,XY sex reversal as part of lipoid congenital adrenal hyperplasia (20). This patient had four large regions of homozygosity greater than 10 Kb in size, equivalent to 1.97% of the genome being homozygous. The STAR variant was located within the largest homozygous region, spanning more than 20.6 Kb on chromosome 8. This sample had been subjected to microarray analysis for detection of large deletions and duplications, and none were detected (Table 2). The finding of this splice site variant in a homozygous interval of the patient's genome is probably a true homozygous finding, and the match with the reported phenotype of adrenal insufficiency makes this a likely genetic diagnosis.

Leydig cell hypoplasia

Patient RDSDO15 has typical female external genitalia and no response to T treatment, but no variant in the AR. Exome sequencing identified a homozygous c.562G>T nonsense variant in the LHCGR gene, predicted to lead to a truncated protein p.Glu188*, a likely null allele (12). We also sequenced the parents, who were known to be related, and several short homozygous interval(s) (5–10 Mb) were observed in the patient, encompassing 5.95% of the genome. The LHCGR variant identified here occurs in a region of homozygosity on chromosome 2. Inactivating variants of the LHCGR gene result in failure of Leydig cells to develop in the testis (21), leading to an extremely rare condition known as Leydig cell hypoplasia. Although these genetic findings are likely diagnostic for the Leydig cell hypoplasia, unfortunately they do not explain the lack of response to exogenous T in this patient.

Variants in the anti-Müllerian hormone (AMH) receptor/persistent Müllerian duct syndrome (PMDS)

AMH causes regression of the paramesonephric ducts through the AMH receptor, AMHR2 (22). In PMDS, Müllerian-derived structures remain in 46,XY individuals who are otherwise normal males (5). PMDS is a recessive condition generally caused by variants in the AMH or AMHR2 genes. Unfortunately, testing for these genes is not available on a clinical basis in the United States. In two unrelated patients clinically diagnosed with PMDS, we identified a known 27-nucleotide deletion (23) and an additional variant in the AMHR2 gene. In case RDSD026, the second variant was within the deleted region and therefore must be present on the other allele. In case RDSD027, the second variant was located in a different region of the gene, and without parental samples, we could not ascertain phase. However, given the strong association of AMHR2 variants with the diagnosis of PMDS, we believe a genetic diagnosis has been achieved in both cases.

Likely pathogenic variants in partial androgen insensitivity syndrome (AIS)

Variants in the AR are well known causes of AIS (4, 24). We found a previously unreported missense variant in AR in RDSD009 that is likely causative of the DSD features in this patient. This patient had previously undergone deletion analysis for AR with no findings (25). To the best of our knowledge, there are no additional reports of such a constellation of clinical features as seen in this patient, and it seems unlikely that this single missense variant in AR would be responsible for them (Table 2).

NR5A1/SF1 is associated with 46,XY gonadal dysgenesis and adrenal insufficiency (3). In patient RDSD016, we identified an NR5A1 variant previously reported in a patient with isolated distal hypospadias, generally considered a mild form of DSD (26). Because NR5A1 variants are associated with a range of phenotypes, from severe gonadal dysgenesis to isolated hypospadias or even male infertility (27, 28), this variant is likely causative of the phenotype.

Variants in the HSD17B3 gene were identified in two patients. In CDSD028, exon 1 of the gene was deleted in a region of homozygosity. Subsequent deletion/duplication analysis of the gene confirmed a 461-bp homozygous deletion of the gene. The deletion includes the initiating ATG and thus likely results in a complete lack of protein. In CDSD033, two different missense variants previously reported as damaging (29–31) were identified. HSD17B3 deficiency is a classic differential diagnosis for AIS (4). Our results show that it might be less rare than previously thought.

CHD7 variants in atypical CHARGE syndrome presentations

Mutations in the CHD7 gene can cause CHARGE syndrome, a complex multiorgan disorder including genital abnormality (32), but not all variants in CHD7 lead to the full CHARGE syndrome phenotype (33). In two patients with very different presentations, we identified novel missense CHD7 variants. Patient CDSD037 presented with a fairly typical genital presentation in CHARGE syndrome; however, no other anomalies were found that would warrant a clinical diagnosis of CHARGE even after follow-up by the clinician after the exome report. In contrast, the genital presentation of patient RDSD001 was less typical of CHARGE syndrome, but she had associated anomalies in organs typically affected in CHARGE syndrome (Table 2). However, follow-up by the referring physician showed that they were not typical of CHARGE syndrome; thus, this variant also remains a VUS. In both of these cases, we cannot determine whether the variants are benign and unrelated to the patient's phenotype or instead add to the growing body of evidence expanding the spectrum of phenotypes associated with CHD7 variants (33).

Other VUS

VUS are potentially clinically actionable, and further clinical tests in patients in whom they are identified may assist in refining their categorization. We identified several VUS in our study. When these were in clinical samples, they were included on the report. Loss-of-function alleles of Desert Hedgehog (DHH) cause recessive 46,XY gonadal dysgenesis (OMIM no. 233420). Patient RDSD005 with complete gonadal dysgenesis had a heterozygous missense p.Glu348Val variant in DHH. The only previously reported heterozygous variant is a frameshift mutation on a 46XY, 45X mosaic background. The genital phenotype of this patient is an excellent fit with previous reports, and we feel the variant is potentially causative of this patient's phenotype; but, with the current evidence, it remains a VUS.

Hypospadias is a common malformation associated with DSDs, but the genetic etiology remains unclear. In two patients, we identified VUS in genes previously associated with hypospadias. A hemizygous variant in MAMLD1 (OMIM no. 300758) was identified in a patient raised as male with undescended testes and “abnormal genitalia,” associated with neuropathy and hypotonia (CDSD035). The variant cannot explain the patient's nongenital clinical symptoms but may be involved in the DSD part of the phenotype. CDSD038, a patient with penoscrotal hypospadias and a complex genital phenotype, harbored a missense BNC2 variant, another gene associated with hypospadias (34, 35), and a variant in FGFR1, a gene associated with hypogonadotropic hypogonadism, especially in association with mutations FGF8 or GNRHR, a case of oligogenic etiology in DSD (36). The BNC2 variant, or the combination of the FGFR1 and BNC2 variants, may cause the genital phenotype in this patient.

When parental samples are available, we further mine the data outside of the primary gene list for de novo heterozygous variants and inherited compound heterozygous variants. In CDSD032, a de novo variant was found in the Neuropilin 1 (NRP1) gene. Neuropilin 1 interacts with Sema3A, and variants in SEMA3A are associated with hypogonadotropic hypogonadism (MIM no. 614897). A mouse model preventing the interaction of Nrp1 and Sema3a proteins results in a Kallmann-like phenotype (37). This variant was reported as a VUS; however, endocrine testing in the child at age 12 did not identify hypogonadotropic hypogonadism. There are reports of spontaneous reversal of hypogonadotropic hypogonadism (38), but in this case endocrine testing did not support the diagnosis, and the genital phenotype of this patient remains unexplained genetically.

Discussion

Exome analysis of 46,XY DSD cases generated a genetic diagnosis in a total of 35% (14 of 40) of cases, with an additional six VUS that may be reclassified as literature evolves. Exome sequencing allowed an unprecedented level of genetic diagnostic success in this cohort, especially considering that, for most patients, other endocrine and genetic testing had been exhausted.

Historically, DSD patients have been diagnosed through a combination of endocrinology and phenotypic examination, with a genetic diagnosis being secondary. However, an early genetic diagnosis can guide future endocrine and imaging tests and help limit potentially unnecessary invasive testing and costs. For example, the variants in HSD17B3 uncover a risk for virilization at puberty. Because 46,XY DSD can be associated with multiple genetic findings and variable clinical features, exome sequencing is also useful to identify a genetic cause without preconceived phenotype ideas. Patients with variants in the same gene can present very differently, as exemplified in our study by the MAP3K1 variants we found in four patients with highly variable phenotypes. Conversely, patients with a clinical diagnosis of AIS were found to have likely pathogenic variants in the NR5A1 or HSD17B3 genes.

Exome sequencing identified genetic diagnoses of extremely rare conditions and identified variants in genes not currently available for clinical testing. Although MAP3K1 and AMHR2 mutations are known causes of DSD, clinical testing for these genes is not yet available in the United States, and the limited clinical testing of LHCGR available would not have detected the variant we identified here. Most samples here had already been tested for variants in SRY and NR5A1; thus we cannot directly address whether the proportion of DSD cases accounted for by them will change as new genes such as MAP3K1 are identified. However, from our study, we anticipate that in an unbiased cohort of 46,XY DSD individuals, these three genes will each account for 10–15% of cases.

Genetic diagnoses are useful for patients and clinicians, contribute to clinical knowledge of DSD, and are invaluable for genetic counseling of couples contemplating future pregnancies. A genetic diagnosis can also bring reassurance to patients and their families. The patient with the NR5A1 variant was raised female but did not feel comfortable in that role. The diagnosis of AIS meant that she would be unlikely to respond to T treatment, but having self-administered T, she felt she had responded to it. The finding of an NR5A1 variant previously reported in a male with isolated hypospadias was very reassuring for this patient. It supported her feeling that she should be male, validated her suspicion that she responded to T, and ultimately supported her transition to a male body habitus. Anecdotally, we have found that many families are relieved to receive a genetic diagnosis, even when prognosis and treatment options are not impacted.

In this study, we identified a number of VUS such as in CHD7 or DHH (Table 3). Parental samples would be instructive in determining the pathogenicity of these variants, but were not available. For most of the 20 individuals for whom no interpretable variant was found, we found at least one variant in the DSD gene list, but they did not reach the level of clinical significance. Reasons included: a single variant in a gene associated with a recessive condition (in cases where copy-number variations had been ruled out by microarray analysis); published reports of a phenotypic spectrum that did not extend to the patient's findings; a potentially dominant variant present in the ESP with a minor allele frequency greater than 0.1%; and the variant was predicted to be benign or not affecting the canonical transcript. These variants may be reclassified as our understanding of DSD genetics evolves, but they currently cannot be interpreted or reported clinically. Ultimately, the combination of genetics with endocrine and imaging will validate the functionality of variants, thus advancing our understanding of DSD and treatment options for future patients. Additional reasons for not identifying a genetic cause include mechanisms that clinical exome sequencing cannot identify such as: nonexonic mutations; mutations in yet undiscovered DSD genes; oligogenic etiologies of DSD, as demonstrated in the case of FGFR1 mutations (36); and epigenetic/environmental influences.

In summary, our data show that exome sequencing is an effective test for genetic diagnosis in DSDs. For a comparable cost of full sequencing of a single gene such as the AR or a limited-capture panel of genes, exome sequencing can examine all genes with known or suspected involvement in DSD. Exome sequencing should therefore be considered a good first-tier diagnostic or rule-out test by clinicians (39, 40). Recent advances in the sequencing technologies are leading to substantial decreases in turnaround time, and soon it will be possible to obtain results in under 2 weeks. This will allow the test to be useful even in urgent cases.

Acknowledgments

The authors thank all the patients and families who contributed samples to this project and consented to have their data shared. We thank all the participants in the UCLA Genomic Data Board meeting; Traci Toy and Thien Huynh at the UCLA Clinical Genomics Center, for exome sequencing both clinical and research samples; Bret Harry, for assisting in bioinformatics analysis; and Jean Reiss, for assisting with Sanger sequencing validation. We thank the National Heart Lung and Blood Institute Grand Opportunity Exome Sequencing Project and its ongoing studies that produced and provided exome variant calls for comparison: the Lung GO Sequencing Project (HL-102923), the WHI Sequencing Project (HL-102924), the Broad GO Sequencing Project (HL-102925), the Seattle GO Sequencing Project (HL-102926), and the Heart GO Sequencing Project (HL-103010).

Funding for this project was from the Doris Duke Foundation, a National Institutes of Health T032 training grant (5T32GM008243-25), and National Institute of Child Health and Human Development Grant RO1HD06138 DSD-Translational Research Network.

Disclosure Summary: The authors have nothing to disclose.

Abbreviations

 
• AIS

  androgen insensitivity syndrome

 
• AMH

  anti-Müllerian hormone

 
• AR

  androgen receptor

 
• DSD

  disorder of sex development

 
• INDEL

  insertion and deletion

 
• PMDS

  persistent Müllerian duct syndrome

 
• VUS

  variants of uncertain clinical significance.

References