FGF9 variant in 46,XY DSD patient suggests a role for dimerization in sex determination

Abstract 46,XY gonadal dysgenesis (GD) is a Disorder/Difference of Sex Development (DSD) that can present with phenotypes ranging from ambiguous genitalia to complete male‐to‐female sex reversal. Around 50% of 46,XY DSD cases receive a molecular diagnosis. In mice, Fibroblast growth factor 9 (FGF9) is an important component of the male sex‐determining pathway. Two FGF9 variants reported to date disrupt testis development in mice, but not in humans. Here, we describe a female patient with 46,XY GD harbouring the rare FGF9 variant (missense mutation), NM_002010.2:c.583G > A;p.(Asp195Asn) (D195N). By biochemical and cell‐based approaches, the D195N variant disrupts FGF9 protein homodimerisation and FGF9‐heparin‐binding, and reduces both Sertoli cell proliferation and Wnt4 repression. XY Fgf9 D195N/D195N foetal mice show a transient disruption of testicular cord development, while XY Fgf9 D195N/− foetal mice show partial male‐to‐female gonadal sex reversal. In the general population, the D195N variant occurs at an allele frequency of 2.4 × 10−5, suggesting an oligogenic basis for the patient's DSD. Exome analysis of the patient reveals several known and novel variants in genes expressed in human foetal Sertoli cells at the time of sex determination. Taken together, our results indicate that disruption of FGF9 homodimerization impairs testis determination in mice and, potentially, also in humans in combination with other variants.

disruption of FGF9 homodimerization impairs testis determination in mice and, potentially, also in humans in combination with other variants. In most mammals, testis development is initiated by SRYmediated upregulation of SOX9 expression in somatic progenitor cells of the developing XY gonad, which directs Sertoli cell differentiation. 4 In mice, Fibroblast Growth Factor 9 (FGF9) helps to maintain Sox9 expression by repressing the ovarian RSPO1-WNT4 and FOXL2 pathways, [5][6][7] and XY foetal gonads lacking either Fgf9 or its cognate receptor Fgfr2c show partial or complete gonadal sex reversal. 5,[8][9][10] In humans, FGF9 is a candidate for a role in human testicular development and 46,XY GD. The finding of a 46,XX testicular DSD patient with a 30 kb duplication encompassing FGF9 suggests that increased FGF9 dosage can initiate testis development in chromosomally female gonads. 11 In support, FGF9 can stimulate Sertoli cell differentiation ex vivo in human testis cultures. 12 Moreover, the identification of an FGFR2c loss-of-function mutation and an FGFR2 deletion in 46,XY GD suggests the involvement of FGF ligands, which bind FGFR2c in human testis development. 13,14 Of the FGF ligands capable of binding FGFR2c, FGF9 is one of the few showing expression in the human foetal testis. 15,16 However, no loss-of-function variants in FGF ligands including FGF9 had been identified in genetic screens of 46,XY GD patients. 2,3,17 Clinically, the only phenotype associated with FGF9 variants is multiple synostoses syndrome (SYNS3 OMIM# 612961) with or without craniosynostosis, an autosomal dominant skeletal disorder caused by an overgrowth of bone at the joints or the cranial sutures. 18 (R190T). [18][19][20][21][22] The mouse models Fgf9 N143T and Fgf9 S99N show partial XY sex reversal 23 , suggesting that pathogenic FGF9 variants in humans might also contribute to gonadal defects.
Here we describe a patient with 46,XY GD with a kidney anomaly and mild scoliosis with an FGF9 missense variant (c.583G > A;p. (Asp195Asn) (D195N)). Analysis of this variant biochemically, in cultured Sertoli cells, and CRISPR-generated mice indicates that FGF9 homodimerization is important for testis determination in mice. A contribution of FGF9 to human testicular development is suggested.

| Human ethical approval and patient recruitment
Patient recruitment, consent and DNA extraction were carried out, as described previously. 2 Approval for this study was obtained from the Human Ethics Committee at the Royal Children's Hospital, Melbourne, Victoria, Australia (HREC22073).

| Patient DNA sequence analysis
The patient DNA was initially analysed using a DSD targeted massively parallel sequencing (MPS) screen for research purposes, as previously described. 2 Whole-exome library preparation and sequencing was performed by the Australian Genomics Research Facility using Agilent SureSelect Human All Exon V6 (Agilent Technologies, Santa Clara, California, USA) and a NovaSeq 6000 Sequencing System (Illumina, San Diego, California, USA). Sequencing data were analysed using the Cpipe pipeline. 24 The mean coverage obtained for the three samples was 79.5 reads (± 4.5 SD), and the variants were deposited into SeqR for further analysis (https://seqr.broadinstitute.org/). The patient and parental genotypes were confirmed by the Australian

| Microarray analysis
A custom 60 K oligo CGH microarray was designed on the SureDesign software and manufactured by Agilent Technologies (Santa Clara, CA), using GRCh37/hg19. Probes were selected to detect CNVs at the exon level for a defined number of genes known to cause DSD (van den Bergen, Unpublished). Some SOX9 regulatory regions were also targeted CHR17:69475000-69 560 000 (XXSR/RevSex) and CHR17:70102435-70 105 514 (TESCO), to detect CNVs at a functional resolution of 500 bp. An approximate whole-genome backbone resolution of 120 kb was maintained. Patient and reference DNA were processed as specified by manufacturer protocols. Computational analysis was performed using CytoGenomics (Agilent), CGH algorithms ADM2 used at a threshold of 6.0.

| Mice
Fgf9 D195N/+ mice were generated by the MAGEC laboratory (WEHI) using CRISPR/Cas9. 26 Fgf9-null mice 8,10 were obtained via Dr. Josephine Bowles (University of Queensland). Both mouse lines were maintained on C57Bl/6, a genetic background sensitive to XY sex reversal. 27 All animal experimentation was approved and performed according to procedures determined by the Monash Medical Centre Animal Ethics Committee.
See Supplementary files for additional information.

| Statistical analysis
All statistical analysis was conducted in GraphPad Prism v8.1 (GraphPad Inc., San Diego, CA). using appropriate tests, as stated in

| Other methods
See Supplementary files, for the following methods.
In silico variant prediction of pathogenicity.
In silico protein structure analysis.

Mice.
Plasmid Construction and Site-Directed Mutagenesis.

| Case report
A phenotypically female patient of Caucasian origin presented at 2-month-old with a right inguinal hernia. Subsequent surgery revealed the presence of a structure, thought macroscopically to be an ovary in the sac, which was replaced into the abdominal cavity. The left side also displayed a small empty hernial sac. At the age of 16 years, the patient attended for clinical evaluation of primary amenorrhea, less breast development than expected for her age, clitoromegaly, and a uterine 'nubbin'. Levels of testosterone and free androgen index were elevated, and 17α-hydroxyprogesterone was reduced relative to female clinical reference levels ( Table 1). Karyotyping of the patient revealed 46,XY, with the unaffected father and mother showing 46,XY and 46,XX, respectively. As such, the patient was diagnosed with 46,XY GD. At the age of 18 years, MRI revealed a small uterus.
Due to the increased risk of gonadal tumours, the patient underwent a second operation to remove the uterine appendages (adnexa). Gross and histological examination of the left adnexa revealed a vas deferens and a fallopian tube whose fimbriated end was attached to a cystic soft pink-tan tissue, thought to be a streak gonad measuring  To rule out copy-number variations (CNVs) in known XY DSD genes, the patient DNA was analysed on a custom CGH microarray. 28 No duplications or deletions were detected in known XY DSD genes, including SRY and the human SOX9 enhancers. 29 We undertook nextgeneration sequencing to identify potential DSD-causing gene variants. The patients' DNA was screened using a custom DSD-specific gene panel (64 diagnostic genes and 967 candidate genes) 2  D195N also increases the distance of the hydrogen bond between residue 195 and K58 (Figure 2A,B). However, in silico modelling showed that D195N could also lead to the formation of a novel hydrogen bond with R62 ( Figure 2B). On balance, the model suggests that the D195N substitution will potentially disrupt FGF9 homodimer stability in vivo.

| D195N disrupts FGF9 homodimerisation but not FGFR binding
FGF9 dimers exhibit a higher affinity for heparan sulphate proteoglycan (HSPG) than FGF9 monomers. 32,33 A reduction in FGF9 dimer formation due to the D195N substitution is therefore expected to reduce HSPG-binding. We used heparin-sepharose affinity chromatography to test whether D195N affected FGF9 affinity for heparin, a commonly used surrogate for HSPG. Recombinant FGF9-wildtype and FGF9-D195N proteins were expressed and purified using bacterial expression plasmids ( Figure S1) 34 then bound to a heparin column.
Heparin-binding strength was compared by application of an increasing salt gradient. While FGF9-wildtype eluted at 852 mM NaCl, FGF9-D195N eluted at the lower concentration of 823 mM, indicating a reduced binding affinity to heparin ( Figure 2C).
To determine whether D195N disrupts FGF9 homodimerisation, in situ proximity ligation assays (PLAs) were performed. FGF9-wildtype and variants S99N, R62G and N143T were used as controls. FGF9 dimer formation is reportedly disrupted by R62G and N143T, 19,33 but not S99N. 18 Plasmids expressing HA-and FLAG-tagged wildtype or mutant FGF9 were transiently transfected into HEK293-T cells. Previous structural analyses reported that D195 is required for FGF9 homodimerisation but not for FGFR binding. 32 On the other hand, FGF9-FGFR binding is disrupted by S99N, R62G and N143T. 18,19,33 To test whether D195N alters receptor binding, we again used PLAs. Since the "c" isoform of FGFR2 mediates FGF9-signalling during mouse sex determination, 5 we focused our analysis on FGF9-FGFR2c interactions using a MYC-tagged FGFR2c mammalian expression plasmid ( Figure 2F). Analysis of FGF9-D195N-FGFR2c interaction revealed no loss of receptor binding (94% when compared to wildtype-FGFR2c) ( Figure 2G). Taken together, these results indicate that the D195N substitution reduces FGF9 homodimerisation and heparin affinity, but not FGF9-FGFR2c binding.

| D195N reduces FGF9-mediated cell proliferation and Wnt4 repression in vitro
To assess the ability of FGF9-D195N to induce Sertoli cell proliferation and repress Wnt4 expression, the 15P-1 mouse Sertoli cell-line was cultured in a reduced growth medium and treated with  gonads showed no differences to XY wildtype controls, which exhibited AMH-expressing cords and an absence of FOXL2 ( Figure S2). 34 In contrast, E12.5 XY homozygous   Figures S3 and S4). 34 In addition, FOXL2 and the meiotic germ cell marker γH2AX were absent in XY Fgf9 D195N/D195N testes, like in control testes ( Figure S4). 34 qRT-PCR analyses showed that expression of Sox9, Amh, FGF9 signalling target Dusp6, Foxl2 and the meiotic germ cell marker Stra8 was unchanged in XY Fgf9 D195N/D195N testes compared to wildtype testes.
However, increased expression was observed for the ovarian marker Wnt4 in both XY Fgf9 D195N/+ (3.2-fold) and XY Fgf9 D195N/D195N (3.7-fold) gonads ( Figure S5). 34 Taken together, these data indicate that during testis develop- IF for FOXL2 showed that two of seven E12.5 XY Fgf9 +/À control gonads expressed ectopic FOXL2-positive cells at the anterior pole ( Figure 5B), a phenotype not reported previously. 8 (n = 13), five at the anterior pole only ( Figure 5C) and eight at both poles ( Figure 5D). In addition, IF for SOX9/MVH showed a disruption to Sertoli cell arrangement, while germ cell numbers appeared reduced ( Figure 5, compare h and i with f). IF for Laminin revealed that compared to XY Fgf9 +/À gonads, which contained well-formed cord structures, XY Fgf9 D195N/À gonads exhibited fewer (1.7-fold reduction) and abnormal cords in the central region ( Figure S3b, d, k), 34 while at the poles, laminin staining was comparable to XX controls ( Figure S3d, e, k). 34 To investigate somatic cell proliferation in E12.5 XY  Figure 6F).
However, in the centre of XY Fgf9 D195N/À gonads, IF for Laminin showed that testis cords were well defined and comparable in numbers to Fgf9 +/À control gonads ( Figure S3g, i, k). 34

mRNA levels of
Foxl2 and Wnt4 were increased in XY Fgf9 D195N/À gonads compared to XY wildtype gonads, consistent with the partial gonadal sex reversal phenotype observed by IF. However, no changes in expression were detected for Sox9, Amh, Stra8 or Dusp6 ( Figure S7). 34 Taken together, our results show that the sex reversal phenotype in XY Fgf9 D195N/À gonads is more severe than in XY Fgf9 D195N/D195N gonads and completely rescued by wildtype FGF9 (in XY Fgf9 D195N/+ gonads), indicating that D195N leads to FGF9 loss-of-function in the developing testis.  (Table S2). 34 Six novel variants not reported in gnomAD were identified of which F11R and HIST1H1E are de novo to the patient.

| WES identifies potential exacerbating variants in Sertoli cell genes
STRING analysis of the 25 candidates ( Figure S8) 34  involving FGF9 include CYP27A1, SERPINA5 and TRIM55, all of which are highly upregulated specifically in human Sertoli cells 15 (Table S3). 34  The patient did present with mild scoliosis, a phenotype seen occasionally in patients with FGFR2c mutations. 43 The patient also presented with a small slightly dysplastic right kidney with normal renal function. Although no abnormal renal phenotypes have been reported in patients with FGF9 variants, [18][19][20][21][22] in mice FGF9 and FGF20 act redundantly and are crucial for kidney development. 44 One

| DISCUSSION
Fgf9 knockout allele on a Fgf20 null background leads to a reduction in nephron numbers and small dysplastic kidneys. Moreover, Fgf9; Fgf20 double knockout in mice, or a homozygous mutation in human FGF20 results in renal agenesis. 44 FGF9 could therefore also play a role in human kidney development, but heterozygous variants are likely to be rare due to functional redundancy with FGF20.
In summary, we have identified the first variant in FGF9,