Human MAMLD1 Gene Variations Seem Not Sufficient to Explain a 46,XY DSD Phenotype

MAMLD1 is thought to cause disordered sex development in 46,XY patients. But its role is controversial because some MAMLD1 variants are also detected in normal individuals, several MAMLD1 mutations have wild-type activity in functional tests, and the male Mamld1-knockout mouse has normal genitalia and reproduction. Our aim was to search for MAMLD1 variations in 108 46,XY patients with disordered sex development, and to test them functionally. We detected MAMDL1 variations and compared SNP frequencies in controls and patients. We tested MAMLD1 transcriptional activity on promoters involved in sex development and assessed the effect of MAMLD1 on androgen production. MAMLD1 expression in normal steroid-producing tissues and mutant MAMLD1 protein expression were also assessed. Nine MAMLD1 mutations (7 novel) were characterized. In vitro, most MAMLD1 variants acted similarly to wild type. Only the L210X mutation showed loss of function in all tests. We detected no effect of wild-type or MAMLD1 variants on CYP17A1 enzyme activity in our cell experiments, and Western blots revealed no significant differences for MAMLD1 protein expression. MAMLD1 was expressed in human adult testes and adrenals. In conclusion, our data support the notion that MAMLD1 sequence variations may not suffice to explain the phenotype in carriers and that MAMLD1 may also have a role in adult life.

Presently, there is some controversy about the causative role of MAMLD1 gene variations and the associated DSD phenotype in carriers for the following reasons. First, some MAMLD1 variants have also been detected in normal individuals (P359S, V505A, N662S) [3,5,7,12] and others are not present in all affected DSD individuals of the same family (P359S, Q580R) [3]. Second, the male Mamld1 knockout mouse has a normal genital phenotype and has normal reproduction [13]. Third, functional studies in vitro show normal results for several MAMLD1 mutations compared to wild type (WT) when studying their effect as suggested regulators of genes involved in sex development [5,14]. Finally, MAMLD1 variants are also found in other species such as dogs, cats and horses with or without DSD [15][16][17].
Moreover, the V505A MAMLD1 variant has been found in the WT genome of the homo Neanderthal and chimpanzee [18]. This variant is regarded as an ancestral, potentially compensated mutation which is only disease-causing in humans [18]. Nevertheless, analyses of large case-control studies revealed the double S-S haplotype including MAMLD1 P359S and N662S as a risk factor for hypospadias [5,7].
Little is known about the exact role of MAMLD1 in sexual development. A role in sex differentiation through supporting testosterone production in critical periods of male development has been suggested [14]. Studies in mice revealed increasing Mamld1 expression from E12.5 to E14.5 in fetal Leydig and Sertoli cells [13]. MAMLD1 is controlled by SF-1 which is a key transcription factor for numerous genes involved in sex development and steroidogenesis [14]. MAMLD1 transactivates also the non-canonical Notch targeted Hes3 promoter [14]. Hes3 regulates cell differentiation and proliferation during embryonic development [19]. MAMLD1 seems related to the production of testosterone as its knock-down reduces testosterone production and gene expression of CYP17A1 [20]. Studies in Mamld1-KO mice showed significantly reduced testicular expression of Leydig-specific genes such as Star, Cyp11a1, Cyp17a1, Hsd3b1 and Insl3, but normal expression of other genes related to steroidogenesis and sex-development [13]. However, Mamld1-KO mice have normal external genitalia and are able to reproduce similar to WT animals [13]. Taken together, there is justified doubt whether MAMLD1 gene variations are sufficient to explain the DSD phenotype in carriers warranting further studies.
In this study we searched for MAMLD1 sequence variations in a cohort of 108 46,XY DSD individuals in whom mutations in other candidate genes (AR, SRD5A2, NR5A1) were previously ruled out. We found 9 MAMLD1 mutations (7 of them novel) in 108 46,XY DSD patients (8.3%). Patients' characteristics were compared to reported cases. In vitro functional studies revealed negative results for most MAMLD1 variants. Comparative alignments showed that the original amino acids are mostly not conserved through evolution, yet V505 exists only in human. Overall, our data support the notion that MAMLD1 sequence variations may not suffice to explain the DSD phenotype in carriers.

Patients and genetic analyses
A cohort of 108 46,XY DSD individuals in whom mutations in other candidate genes (AR, SRD5A2, NR5A1) were previously ruled out was analyzed for MAMLD1 sequence variation. Written informed consent was obtained from all individuals participants/legal guardians included in the study after full explanation of the purpose and nature of all the procedures used. Each patient's pediatric endocrinologist provided the clinical and biochemical data. The genetic analyses were performed at the Vall d'Hebron Research Institute in Barcelona and the in vitro and in silico functional studies were done at the Pediatric Endocrinology Research laboratory in Bern. The molecular studies were approved by the ethic committees of the Vall d'Hebron Research Institute, Barcelona, Spain and the Ethic Commission of the Kanton Bern, Switzerland. Data entering the study were provided by the clinicians and the genetic lab in coded forms and are stored and accessible to the scientific community as follows: a) requests for clinical and biochemical data may be addressed to specific clinicians by contacting the corresponding author, b) genetic data are accessible through the Biobank system of Vall d'Hebron (biobanc@vhir.org), c) relevant experimental data are provided as a Supporting file (S1 File). The methods were carried out in accordance with the approved guidelines. Genomic DNA was isolated from peripheral blood leukocytes. MAMLD1 coding regions, their flanking intronic sequences and part of the 5'UTR were amplified by PCR using specific primers [S1 Table, [4]]. The PCR products were sequenced using the BigDye Terminator v3.1 Cycle Sequencing Kit on an automated ABI PRISM 3100 Genetic Analyzer (Applied Biosystems, Foster City, CA, USA). Obtained sequences were analyzed against GenBank entries NG_017093.2 (genomic DNA), NM_005491.4 (mRNA) and NP_005482.2 (protein) (http://www.ncbi.nlm.nih.gov/). Genetic analyses of the AR, SRD5A2 and NR5A1 genes were performed as described [21][22][23]. All patients were checked for these 3 genes except patient 5, which was analyzed for NR5A1 only.
A MAMLD1 SNP genotyping (rs62641609: H347Q; rs41313406: P359S; rs61740566: V505A and rs2073043: N662S) of 155 normal adult male controls was performed by TaqMan assays for allelic discrimination using the Applied Biosystem Prism 7900 HT instrument and the allelic discrimination end-point analysis mode of the Sequence Detection software package, Version 2.3 (SDS 2.3). The following custom TaqMan SNP genotyping assays were used according to the protocols supplied by Applied Biosystems: C__64647092_10; C__26000187_10; C__25995288_20; C__15950293_10. Differences between controls and patients for SNP genotype and allele frequencies were analyzed with a contingency table analysis using the JMP 1 7 program (SAS Institute, Inc., Cary, NC, USA).

Tissue expression studies
MAMLD1 expression was studied for normal human adrenal and testicular tissues using cDNA samples purchased from amsbio (AMS Biotechnology (Europe) Limited, Abingdon, UK). According to amsbio information, cDNAs originate from two adult subjects aged 50 years for adrenals and aged 23 years for testes. Fetal material was from two samples 20 weeks gestation for adrenals and 30 weeks gestation for testes. Semiquantitative PCRs (35 cycles) for MAMLD1 expression were carried out using the recommended concentrations of the purchased cDNAs. Primers are listed in S1 Table. Agarose-gel electrophoresis was performed for the PCR products, which were detected by ethidium bromide on a UV transilluminator (Alphaimager, Proteinsimple, Santa Clara, CA, USA).
Expression vectors. The promoter luciferase reporter vector Hes3_luc (-2715~+261) was kindly gifted by Dr. Maki Fukami (National Research Institute for Child Health and Development, Tokyo, Japan) [14,24]. The human -3.7CYP17A1_Δluc vector was available from previous work [25,26]. For the MAMLD1 expression vectors, we used the WT cMyc-MAMDL1_pCMV and the minor transcript variant lacking exon 5 ΔE5 (previous ΔE4), both kindly gifted by Dr. Maki Fukami. In addition, we modified the WT cMyc-MAMLD1_pCMV vector according to the revised coding sequence NM_005491.4 (GenBank MAMLD1, isoform 2, new transcription start site in r.64, exon 2). This new WT construct was custom made by GenScript (Piscataway, NJ, USA). Mutant MAMLD1 expression vectors (c.605C>T, c.626delT, c.631G>A, c.1041C>A, c.1075C>T, c.1508C>A, c.1514T>C, c.1985A>G, c.2170C>G and c.2190G>A) were generated by PCR-based site-directed mutagenesis using specific primers (S1 Table) and the QuickChange protocol by Stratagene (Agilent Technologies Inc., Santa Clara, CA, USA) using the new WT expression vector as template. The MAMLD1 mutant c.1503_1504dupCAGCAG was also custom made by GenScript. All new constructs were verified by direct sequencing.
MAMLD1 promoter transactivation experiments. Cells were cultured on 24-well plates and transiently transfected (Lipofectamine 2000TM, Invitrogen) with WT or mutant MAMLD1 together with the promoter luciferase reporter constructs Hes3_luc or -3.7CYP17A1_Δluc for 6 hours. The transfection mixture contained 1.25 μg of plasmid DNA and 50 ng Renilla of luciferase reporter (pRL-TK) control per well. Forty-eight hours after transfection, cells were washed with PBS, lysed and assayed for luciferase activity with the Dual-Luciferase Reporter (DLR™) Assay System (Promega AG, Wallisellen, Switzerland) on a Veritasmicroplate Luminometer reader (Turner BioSystems Luminometer and Software by Promega). Firefly luciferase readings were standardized against Renilla control readings and results expressed as relative luciferase units (RLU). Experiments were performed in duplicates and repeated 3 times. Data are given as mean±SEM. Statistical significance was examined using Student's t-test (Microsoft Excel). P-value for significance was set at 0.05.
CYP17A1 enzyme conversion experiments. To measure a possible effect of MAMLD1 on CYP17A1 activity, we assessed the conversion of progesterone to 17-hydroxyprogesterone and androstenedione in steroidogenic NCI-H295R and MA-10, and non-steroidogenic HEK293 cells. For this, all cells were transfected with WT or mutant MAMLD1 and non-steroidogenic HEK293 cells in addition with human CYP17A1 in pcDNA3. Experiments were performed on 12-well plates; transient transfection was for 6 hours (1.25 μg plasmid DNA/well) and experiments were closed after 48 hours. Steroid conversion was labeled by adding 20,000 cpm 14 Cprogesterone per well for 60 min before extracting all steroids from the cell medium. Steroids were then separated by thin layer chromatography (TLC; Macherey-Nagel, Düren, Germany) using the chloroform:ethylacetate (3:1) solvent system and steroid standards. TLC plates were exposed on imaging screens to visualize the radioactive steroids. The screens were read on a Fuji PhosphoImager Fla-7000 (Fujifilm, Dielsdorf, Germany). Steroids were identified according to known stardards and densitometrically quantified as % of total radioactivity per sample using Multi Gauge software (Fujifilm). Experiments were performed 2 times in 3 different cell line backgrounds.
MAMLD1 protein expression studies. Leydig MA-10 cells were transiently transfected with either WT (WT, original WT of ΔE5) or mutant MAMLD1 expression vectors, which all carried a Myc-tag. Cells were lysed and a Western blot was performed using an antibody against c-Myc (C6594, Sigma, Saint Louis, USA). Experiments were repeated twice. Expression of β-actin was the control.

MAMLD1 polymorphism testing
We chose 4 MAMLD1 variants (H347Q, P359S, V505A and N662S), previously detected in controls and 46,XY DSD patients [3,[5][6][7] and/or referenced in dbSNP (http://www.ncbi.nlm. study are shown in red and the novel ones are marked with an asterisk. C. Assessment of MAMLD1 expression in human fetal and adult adrenal and testis. Semiquantitative RT-PCRs were performed using specific primers (S1 Table). GAPDH was used as the internal control. A representative gel picture is shown (n = 3). For MAMLD1 the band at 581 bp corresponds to isoform 2 and the band at 506 bp to isoforms 1 and 3. Dashed red lines in A indicate the location of the PCR fragments amplified for the expression studies. nih.gov/snp/) to compare their genotype and allele frequency in a cohort of 155 normal adult male controls (Table 2) and our cohort of 108 46,XY patients (Table 2) (with phenotypes varying from penoscrotal hypospadias to female external genitalia and including the 9 in whom MAMLD1 sequence variations were detected, Table 1). The MAMLD1 variants H347Q and V505A were not found in our 155 normal adult male controls suggesting a causative role for these genetic variations; the H347Q variant was not present in two exome pools of the European population (http://evs.gs.washington.edu/EVS/ and.http://exac.broadinstitute.org/about), but the V505A was detected in 0.05% and 0.1% of these European subjects, and in 17.8% of African people (http://exac.broadinstitute.org/about). The variants P359S and N662S were also

MAMLD1 tissue expression
We studied MAMLD1 expression for human fetal and adult adrenal and testis tissues. We found that MAMLD1 is expressed in the adult adrenal and testis, and the fetal testis. No expression was found in the fetal adrenal ( Fig 1C). We detected all 3 isoforms in similar quantities.

In vitro and functional studies
In 2012 the reference sequence of MAMLD1 has been revised (GenBank NM_005491.4, [6]). Therefore, we modified the human MAMLD1 expression plasmid which has been originally used for functional studies in all published studies accordingly. We then first performed promoter transactivation studies (Fig 2) with the Hes3 promoter reporter in non-steroidogenic HEK293 cells to compare the activity of the revised MAMLD1 WT vector with the original WT (Fig 2A). We found similar transactivation activity on the Hes3 promoter construct for both WT isoforms (revised and original) as well as for the revised shorter isoform without exon 5. We then created mutant MAMLD1 expression vectors by site-directed mutagenesis according to identified sequence variations in our 46,XY DSD patients. For unclear test results in the literature [3,5,7,14], we added MAMLD1 P359S and N662S to our test series. The ability of these 11 MAMLD1 variants to transactivate the Hes3 promoter in HEK293 cells was then assessed (Fig 2B). Surprisingly, only the L210X mutant lost transcriptional activity compared to WT.
Given these results, we performed similar studies with the human CYP17A1 promoter (Fig 3), as studies in mice revealed an effect of MAMLD1 on CYP17A1 expression, activity and testosterone production [13,20]. Again, most MAMLD1 variants showed WT effect on the CYP17A1 promoter reporter (Fig 3A). Similar to the Hes3 promoter, mutant L210X showed a loss of activity on the CYP17A1 promoter, while variants L724V and S730S showed a decrease in promoter activation compared to WT (p = 0.052 for S730S).
Next, we also tested the effect of WT and MAMLD1 variants on CYP17A1 enzymatic activity. For that we transfected MAMLD1 expression vectors into human adrenal NCI-H295R and into mouse Leydig MA-10 cells. We also co-transfected MAMLD1 expression vectors together with a CYP17A1 expression vector into non-steroidogenic HEK293 cells. After transfection, activity of CYP17A1 was assessed by measuring the conversion of radiolabeled progesterone to 17-hydroxyprogesterone (Fig 3B, S2 Fig). We found no difference for CYP17A1 activity for WT MAMLD1 and variants in all 3 cell systems (data for NCI-H295R cells in Fig 3B; data for MA-10 and HEK293 in S2 Fig). Remarkably, there was also no difference between the control vector and the WT MAMLD1 indicating that MAMLD1 does not regulate human CYP17A1 activity or that the effect is indirect through essential co-factors that were not present in the 3 cell systems used for our experiments. Some MAMLD1 nonsense mutants have been suggested to affect protein expression by nonsense mediated RNA decay [E197X (old E124X), Q270X (old Q197X) and R726X (old R653X) [3]]. Therefore, we analyzed protein expression of WT and mutant MAMLD1 in MA-10 cells. Cells were transfected with the MAMLD1 expression vectors containing a Myc-tag, then Western blots were performed using an antibody against c-Myc (Fig 4). We found no significant difference for MAMLD1 protein expression for almost all of the tested variants compared to WT. The L210X mutant presented a lower band on the Western blot according to its shorter length and its amount may therefore not be compared to the full-length missense variants.
Variants T202M, D211N, H347Q, P359S and A503E are predicted to impact on protein function because the amino acid change comprises a change in the physico-chemical property. Proline is a bending amino acid, thus it may cause a change in the conformation of the protein and it is therefore considered important. According to this, P359 could have some effect on the protein function, although it is not a conserved position in mammals (S1 Fig). Furthermore, we aligned the human MAMLD1 isoform 2 (774 amino acids) with homologous sequences (ranging 720-820 amino acids) from 40 mammalian species with the CLC Sequence Viewer software (S1 Fig). To study all detected species available in the NCBI database (e.g. beyond mammals) was not possible because of high variability in length. L210 and L724 (1 change each) are conserved along mammalian evolution, followed by S730 (deleted amino acid (-) in 2 species). They might therefore be of importance for protein function. The remaining variants' positions T202, D211, P359, Q501Q502, A503, V505 and N662 are not conserved and therefore not considered important for protein function. Variants in positions 202, 347, Role of MAMLD1 in 46,XY DSD 505 and 662 are WT in other mammals indicating that these amino acid changes may not be of importance. T202 and S202 are similarly found: T202 is present in one third of the species (primates and bovidae), while the other mammals present S202. For position 347, H is mainly present, but 2 other changes, including our variant Q347 (in 10 species), are seen. V505 is only present in human, whereas most other mammals harbor A505 (27 species), suggesting that human V505A may just be a reversion to the ancestral state. In position N662 4 variants are present including the variant S662. Finally, the duplication Q501Q502 lies in a Q sequence, which has variable length in different species. This indicates that the length change may not harm the protein function.

Discussion
We detected 7 novel and 2 previously published MAMLD1 sequence variations in 9 46,XY DSD patients presenting with a broad phenotype (Table 1). Similar to other studies, we found that for MAMLD1 sequence variations genotype-phenotype correlation and functional studies reveal ambiguous results. In our series only the truncated MAMLD1 L210X mutation, found in a severely affected 46,XY DSD patient, showed loss of function in transactivation assays using the Hes3 and the CYP17A1 promoters as interacting partner. L201X can therefore be considered a deleterious mutation. Remarkably, this patient also presented with hypogonadotropic hypogonadism at pubertal age suggesting that MAMLD1 may also be involved in the HPG axis, what has not been described so far, or that the patient may harbor additional genetic defects. By contrast, L724V, carried by a 46,XY DSD patient with a typical MAMLD1 phenotype (46,XY male with penoscrotal hypospadias and small penis), and the synonymous S730 mutation, found in 46,XY female patient (also with penoscrotal hypospadias and small penis), had only impaired CYP17A1 but normal Hes3 transactivation activity. Both positions are conserved in mammals. Although variant H347Q showed no functional impairment in our in vitro tests, its presence in 2 of our patients with a severe 46,XY DSD phenotype (female and ambiguous genitalia) and in a published 46,XY DSD patient with hypospadias and microphallus [6], as well as its absence in 155 control males suggests a disease causing role. Variant V505A, which we found in one patient, has been previously described in 2 46,XY DSD subjects [4,7] and in a 46,XX dysgenetic woman [11]. Interestingly, this human variant is the WT found in the genome of the Neanderthal, the chimpanzee and many other species and has been discussed to be an ancestral, potentially compensated mutation [18].
According to published literature of microdeletions and rearrangements in the Xq28 region containing the MAMLD1-MTM1-MTMR1 genes [see Table 3 and [1,2,[28][29][30]], which causes myopathy for the involvement of the MTM1 gene, it appears for the DSD phenotype that the C-terminal region of the MAMLD1 gene is critical for abnormal sex development [29]. By contrast, MAMLD1 variants which have been associated with DSD are found throughout the gene (Fig 1B).
To test novel MAMLD1 sequence variations for their disease causing role, we used Hes3 promoter activation assays as previously performed for other mutations in other laboratories. In addition, we tested for CYP17A1 promoter activation and an effect on steroid enzyme activity as well as for variable expression of WT and mutant MAMLD1 proteins. Overall, all these studies were not giving conclusive results to explain the role of MAMLD1 gene variants for abnormal sex development. Similar confounding results for functional studies are found in the published literature [Table 3 and [5,11,14]]. Although studies in mice revealed a role of MAMLD1 for Cyp17 gene expression and activity [13], our WT MAMLD1 was only able to modulate the human CYP17A1 promoter activity, but had no effect on steroid conversion and thus enzyme activity. Human MAMLD1 does not bind to the Hes3 promoter or to the human HES3 upstream region directly [14]. Therefore, it has been proposed that MAMLD1 may act through other partners such as the Hes3-DNA-binding transcription factor [14], but to date such partners remain obscure. Our in silico search for possible interacting partners of human MAMLD1 was negative when using tools to search for functional partners (string-db.org) or for physical or genetic interactions (http://thebiogrid.org).
MAMLD1 gene variations are also reported in other species, both male and female, and with and without abnormalities in sex development [15][16][17]. Yet none of them has been clearly related to the DSD phenotype. In dogs, both DSD and controls were carriers [15], in cats only an intronic change has been described [16], and in a male horse with hypospadias a benign mutation was detected, conserved in all MAMLD1 sequences available at that time [17].
MAMLD1 seems involved in sexual development during fetal life. It is expressed in human fetal testes and ovaries [3,32], and in mice testes [3,13]. Mamld1 is present in mice testes at least from E11.5 which overlaps with the start of androgen biosynthesis (13 dpc) and with the formation of the male external genitalia (16.5 dpc) [13]. Mamld1 is expressed at low levels in postnatal mice testes until 1 week of age [3,33]. In human fetal testes, MAMLD1 is expressed at high levels in the second trimester of gestation [3,32]. In addition, our study shows that it is also expressed in human adult testes and adrenals, but not in fetal adrenal tissue. These results suggest that MAMLD1 may not only have a role during fetal development but also in adult life.  In mice, Mamld1 regulates the expression of steroidogenic enzymes Cyp17a1, Cyp11a1, Star, Hsd3b1, Hsd17b3, and of genes involved in testes descent (Amh and Insl3) [13,20]. It seems involved in testosterone production by regulating Cyp17a1 activity and expression [20]. However, Mamld1-KO mice develop normal external genitalia and have normal reproduction, which both depend on testosterone production and Cyp17 activity [13]. Overall, these results challenge the role of MAMLD1 in sex development.
In conclusion, the importance of MAMLD1 in the sexual development becomes less and less clear. Its exact role remains unknown. Studies of MAMLD1 variations in humans suggest that there is a genetic correlation between MAMLD1 sequence variations and DSD. However, the wide range in phenotype and poor genotype-phenotype correlation indicate that MAMLD1 gene variations may not suffice to explain the DSD pathology. Therefore, in DSD patients harboring MAMLD1 sequence variations further genetic studies should be performed searching for additional genetic hits explaining the immense variability. This may be done by specific DSD chips containing a larger array of known genes involved in DSD, which are currently being used in several DSD research projects and should become available soon. Alternatively, an untargeted approach like exome sequencing might be employed that also allows to find novel genes so far unrevealed in DSD [34]. In addition, to find copy number variations array-based comparative genomic hybridization (aCGH) or multiplex ligation-dependent probe amplification (MLPA) may be used [35]. Although so far no multiple hits in DSD genes were described involving MAMLD1, it would not be surprising to find them in the near future with the above mentioned next generation methods.
Supporting Information S2 Fig. Effect of WT and mutant MAMLD1 on CYP17A1 enzyme activity in HEK293 and MA-10 cells. Cells were transiently transfected with MAMLD1 WT and mutant expression vectors. HEK293 cells were also co-transfected with the CYP17A1 expression vector as they do not express it endogenously. The effect of WT and mutant MAMLD1 on CYP17A1 enzyme activity was assessed by measuring the conversion of progesterone (P) to 17-hydroxyprogesterone (17OHP) in non-steroidogenic HEK293 cells, and conversion of P to 17OHP and then to androstenedione (Δ4A) in steroidogenic mouse Leydig MA-10 cells. Steroid production was labeled with [ 14 C]progesterone for 60 min. Steroids were extracted and resolved by thin-layer chromatography, then quantified as % conversion. A representative steroid profile obtained from HEK293 (A) and MA-10 (B) cells is shown (n = 2). Similar to experiments performed in NCI-H295R cells (Fig 3B), no effect of MAMLD1 on CYP17-hydroxylase activity was detected. Ve: empty vector; WT: wild type; Ã : co-transfected with empty vector; NT: non-transfected.