The p190 RhoGAPs, ARHGAP35 and ARHGAP5, are Implicated in GnRH Neuronal Development: Evidence from Patients with Idiopathic Hypogonadotropic Hypogonadism, Zebrafish and in vitro GAP Activity Assay

Abstract

Purpose:

To identify novel genes for Isolated Hypogonadotropic Hypogonadism (IHH).

Methods:

A cohort of 1,387 probands with IHH underwent exome sequencing and de novo, familial and cohort-wide investigations. Functional studies were performed on two p190 Rho-GTPase Activating Proteins (p190 RhoGAP), ARHGAP35 and ARHGAP5, which involved in vivo modeling in larval zebrafish and an in vitro p190A-GAP activity assay.

Results:

Rare protein-truncating variants (PTV; n=5) and missense variants in the RhoGAP domain (n=7) in ARHGAP35 were identified in IHH cases [rare variant enrichment: PTV (unadjusted p=3.1E-06) and missense (adjusted p=4.9E-03) vs. controls]. Zebrafish modeling using gnrh3:egfp phenotype assessment demonstrated that mutant larvae with deficient arhgap35a, the predominant ARHGAP35 paralog in the zebrafish brain, display decreased GnRH3-GFP+ neuronal area, a read-out for IHH. In vitro GAP activity studies demonstrated that one rare missense variant (ARHGAP35 p.(Arg1284Trp)) had decreased GAP activity. Rare PTVs (n=2) also were discovered in ARHGAP5, a paralog of ARHGAP35; however, arhgap5 zebrafish mutants did not display significant GnRH3-GFP+ abnormalities.

Conclusion:

This study identified ARHGAP35 as a new autosomal dominant genetic driver for IHH and ARHGAP5 as a candidate gene for IHH. These observations suggest a novel role for the p190 RhoGAP proteins in GnRH neuronal development and integrity.

INTRODUCTION

Individuals with isolated hypogonadotropic hypogonadism (IHH) present with a failure to enter or progress through puberty resulting in sexual infantilism and infertility in the absence of medication¹. IHH is a rare, heterogeneous, predominantly Mendelian disorder whose pathomechanism is defects in the biosynthesis, secretion, or action of gonadotropin-releasing hormone (GnRH).^1,2 These genetic defects can result in impairment of embryonic development and can prevent GnRH neuronal differentiation from stem cells within the developing olfactory epithelium, or the subsequent migration of GnRH neurons into the hypothalamus of the central nervous system (CNS). GnRH neuronal ontology is part of a shared embryonic developmental program intersecting olfactory, craniofacial and CNS development. As a result, IHH has served as a valuable human disease model to discover genes that yield insights into not only GnRH neuron development, migration, and action but also their role in developmental pleiotropy across organ systems.^3–6

The current report identifies an excess burden of rare variants in two p190-Rho GTPase-activating proteins (RhoGAPs), ARHGAP35, and ARHGAP5 through familial, de novo and cohort-wide investigations in a large ensemble of individuals with IHH. The p190 RhoGAPs, ARHGAP35 and ARHGAP5, are evolutionarily conserved, intolerant to genetic variation, and critically important for brain development.^7–11 p190 RhoGAPs are key negative regulators of RhoGTPase (e.g., RhoA) functions: neuronal differentiation, axonal guidance, and adhesion-mediated signaling.^7–9 As such, p190RhoGAPs, regulate pathways where protein-truncating variants (PTVs) result in haploinsufficiency and IHH.^12,13

To understand the human genetic evidence, GnRH development was assessed in stable ARHGAP35 and ARHGAP5 zebrafish mutants. A p190A-GAP assay was employed to measure GAP activity in ARHGAP35 missense variants. Further, using GeneMatcher, we aggregated additional individuals bearing heterozygous PTVs in ARHGAP35, providing human genetic and phenotypic evidence that ARHGAP35 plays a pleiotropic role during development resulting in autosomal dominant IHH and neurodevelopmental disorders (DD).^14,15

METHODS

Patient cohorts

Patients with IHH were enrolled in a genetic study at Massachusetts General Hospital as previously described (n =1387 probands, n=350 parents; Supplementary Methods).¹⁶ GeneMatcher facilitated the identification of patients in other cohorts with PTVs in ARHGAP35.¹⁵

Exome Sequencing (ES) was performed using The Broad Institute Genomics Platform in 1,387 patients with IHH and 350 parents, generating 175 trios. Variant calling was performed as previously described⁴. Rare sequence variants (RSVs) were required to: be non-synonymous variants (stop gain, frameshift, or missense) or canonical splice altering variants (± 2 base-pairs from exon-intron boundaries); have a minor allele frequency (MAF) of <0.1% in the genome Aggregation Database (gnomAD) based on maximum MAF across all population sub-groups; a quality depth >5; a call rate >0.95; MEDIUM or HIGH impact as annotated by Ensembl Variant Effect Predictor; and predicted as damaging in ≥ 2 prediction programs.^17–19 Copy number variant (CNV) calls were conducted from ES using the GATK-gCNV pipeline as previously described.²⁰ CNVs present in known IHH genes (MAF <1%) were curated²⁰ (Supplementary Methods and Supplementary Table 1).

Identification of Variants of Interest:

We conducted two family-based analyses. In the first, we started with three affected family members in two generations and examined ES data for segregating: a) RSVs in known IHH genes; b) heterozygous RSVs and c) homozygous RSVs, including the rare possibility of uniparental isodisomy. Segregation was defined as being present in the affected individual(s) in the family. RSVs were considered “RSVs of interest” if they were PTVs, absent from public databases (i.e. gnomAD), and under genic constraint as defined by gnomAD. In the second family, we analyzed a trio for “RSVs of interest” that were de novo. We then evaluated these discoveries in an IHH cohort (n=1,387) using gene-based burden testing to determine if the gene was enriched for variation in our IHH cohort compared with gnomAD.^10,16 All RSVs were confirmed using bi-directional Sanger sequencing in all probands and available family members.

Pathway analysis was completed using STRING v.11.5.²¹ All active, high confidence interaction scores in the whole STRING network were analyzed, and up to 50 first shell interactors were graphed for networks seeded with ARHGAP35 alone, ARHGAP5 alone, or known IHH genes, ARHGAP35, and ARHGAP5.

Zebrafish husbandry:

Transgenic gnrh3:gfp²² adult fish were maintained in standard conditions with a 14h light-10h dark cycle, and embryos were collected from natural matings of adult fish. Embryos and larvae were reared at 28.5°C in embryo medium (0.3 g/L NaLc, 75 mg/L CaSO₄, 37.5 mg/L NaHCO₃, and 0.003% methylene blue) until experimental endpoint.

CRISPR/Cas9 genome editing and stable zebrafish mutant lines:

To identify zebrafish orthologs for genome editing, we performed reciprocal BLAST searches between Homo sapiens ARHGAP35 (GenBank IDs: NM_004491.5; NP_004482.4) or ARHGAP5 (GenBank IDs: NM_001030055.2; NP_001025226.1) and the translated Danio rerio genome (GRCz11). Two zebrafish paralogs corresponding to ARHGAP35: arhgap35a (ENSDART00000154171.3) and arhgap35b (ENSDART00000084819.5); and ARHGAP5 (ENSDART00000086696.6) were targeted by single guide (sg) RNA sequences identified by ChopChop for CRISPR/Cas9 genome editing.²³ We synthesized sgRNAs using GeneArt Precision gRNA Synthesis Kit (ThermoFisher) according to the manufacturer’s instructions. Mutant zebrafish lines for arhgap35a, arhgap35b, or arhgap5 were established as previously described³ (Supplementary Methods; Supplementary Table 2).

Quantitative reverse transcription (qRT)-PCR of endogenous zebrafish gene expression:

We evaluated the endogenous arhgap35a, arhgap35b, or arhgap5 transcript amount in stable homozygous mutants for comparison with wild type (WT) siblings. At 2 days post-fertilization (dpf), we decapitated larvae in biological triplicates with micro-scissors; individual tails were subjected to DNA extraction for genotyping, and same-genotype heads (30/pool) were combined for extraction of total RNA using Trizol (ThermoFisher) according to manufacturer’s instructions. We reverse-transcribed cDNA using the QuantiTect Reverse Transcription Kit (Qiagen) per kit manual instructions using 1 μg total RNA. The resulting cDNA served as a template for each qRT-PCR reaction (30 ng cDNA template/reaction). All samples were analyzed in technical triplicates on a QuantStudio 3 real-time PCR system (ThermoFisher) using the Power SYBR Green PCR Master Mix (ThermoFisher) with three biological triplicates. The housekeeping gene gapdh was used to normalize gene-level expression using the 2^−ΔΔC_T method.

Automated in vivo image acquisition and analysis of zebrafish larvae:

To quantify GnRH3 neuronal patterning in zebrafish larvae at 5 dpf, we performed automated imaging essentially as described.⁴ (Supplementary Methods)

In vitro p190A GAP assay:

pGEX-6P1 plasmid encoding GST-tagged GAP domain of p190A was previously described.²⁴ Point variants were introduced into the WT form of p190A using the QuikChange II Site-Directed Mutagenesis Kit (Agilent) according to the manufacturer’s instructions using custom-designed primers (Supplementary Table 3). All plasmids were verified by Sanger sequencing. Recombinant GST-tagged p190A-GAP WT and mutant proteins were expressed in BL21 (DE3) bacteria and purified using glutathione-agarose beads (Thermo Scientific) as described.²⁴ PreScission protease (GenScript) was used to cleave the GST tag. Eluted purified p190A-GAP proteins were concentrated using Amicon Ultra-4 centrifugal filters (Millipore), resolved by SDS-PAGE, and stained with Coomassie blue. Proteins were quantified using bovine serum albumin (BSA) as a standard. The GAP activity of p190A-GAP WT and mutant proteins was assessed using the RhoGAP assay biochem kit (BK105, Cytoskeleton) according to the manufacturer’s instructions.

Statistical analysis:

Burden testing for RSVs seen in ARHGAP5 or ARHGAP35 individuals as compared to gnomAD was performed using Fisher’s exact test.¹⁶ Given the assessment of multiple regions of ARHGAP35 for regional enrichment for missense variants, Bonferroni correction was applied to the p-value for RhoGAP missense variants. We used a student’s t-test to detect statistical differences between zebrafish gene expression levels measured by qRT-PCR (biological triplicates). We used a one-way ANOVA with multiple comparison testing to compare GFP+ area and head size measurements across genotypes using GraphPad Prism 9 (biological triplicates). In vitro analysis of p190A-GAP assay WT compared to mutant using an unpaired student’s t-test.

RESULTS

Identification of heterozygous ARHGAP35 Protein-Truncating Variants in five IHH Pedigrees

In a single pedigree, Family 1, ES of three affected family members in two generations uncovered one heterozygous RSV of interest that was shared by all affected family members: ARHGAP35 ([GenBank ID: NM_004491.4] c.3569_3570delAG, p.(Glu1190Glyfs*9); Figures 1–2; Table 1; Supplementary Tables 4-5; clinical phenotype of proband published as case 4²⁵). Consistent with filtering criteria, this RSV was absent from gnomAD (v2.2.1; 4/11/2022), and ARHGAP35 is highly constrained for both PTVs (LOEUF=0.057) and missense variants (Z-score>3.0).¹⁰ There were no additional RSVs of interest segregating with IHH in Family 1. There were no segregating RSVs or CNVs in known IHH genes. Segregation analysis confirmed that the variant was present in all affected individuals but also present in one unaffected child, compatible with variable penetrance, a common phenomenon documented in IHH.¹

Figure 1. Pedigrees with rare ARHGAP35 variants.

a. Loss of function variants; enrichment compared to controls, unadjusted p= 3.1E-06; b. RhoGAP domain missense variants; enrichment compared to controls, adjusted p=4.9E-3. c. GeneMatcher loss of function variants. Parents are shown when their phenotype is known. Abbreviations: “+”, wild-type allele; “M”, variant allele; nIHH, normosmic idiopathic hypogonadotropic hypogonadism; KS, Kallmann syndrome; CDP, constitutional delay of growth and puberty

Figure 2. Rare variants identified in ARHGAP35.

a. Schematic of ARHGAP35 protein (GenBankID: NP_004482.4) and position of loss of function variants and RhoGAP domain missense variants (phenotypes: black – idiopathic hypogonadotropic hypogonadism (IHH), green – developmental disorder (DD) previously published, green underlined DD this report, red - neonatal hypogonadism and DD). FF, phenyalanine domains (blue); pG1, pseudo-GTPase domain 1 (salmon); pG2, pseudoGTPase domain 2 (salmon); PBD, phospholipid binding domain (gray); RhoGAP, Rho GTPase-activating proteins domain (yellow); P, proline rich domain (gray); amino acid position is labeled at the bottom. b. Conservation of missense variants identified in IHH in the RhoGAP domain across multiple vertebrate species.

Table 1:

Phenotype of Individuals in this Study

						Reproductive phenotypes				Non-reproductive phenotypes
Family	Individual	Sex	Ethnicity	Age at Presentation	Diagnosis	Micropenis	Cryptorchidism		Reversal	Face/Head	Bone	Anosmia	Neuro	Renal	Cardiac	Brain Imaging
ARHGAP35
1	1	M	White	14a	KS	−	+	e	−	−	−	+	−	−	−	Asymetric olfactory sulci
	2	F	White	32	IHH	NA	NA		−	−	−	Mild microsmia	Dyslexia, Dyspraxia	−	−	NL olfactory bulbs
	3	M	White	17.7	KS	−	−		+	+	−	+	Dyslexia	−	−	NL olfactory bulbs
2	1	F	Asian	16a	KS	NA	NA		−	Deviated septum; Brachydactyly (type A3, R)	Osteopenia	+	Peripheral Neuropathy	−	−	Absent olfactory bulbs
3	1	M	White	21	IHH	+	+		+	−	Osteoporosis	ND	−	−	Coarctation of Aorta	NL
4	1	M	White	14	IHH	+	−		−	NR	NR	−	NR	NR	NR	NL
5b	1	M	White	NR	IHH	NR	NR		NR	NR	NR	−	NR	NR	NR	NR
6b	1	M	Brazilian	14a	IHH	−	−		−	−	−	ND	−	−	−	NR
7b	1	M	White	NR	IHH	NR	+		NR	NR	NR	ND	NR	NR	NR	NR
8b	1	M	White	NR	KS	NR	NR		NR	NR	NR	+	Synkinesia	Agenesis	NR	NR
	2	M	White	NR	KS	NR	NR		NR	NR	NR	+	Synkinesia	Agenesis	NR	NR
9b	1	M	White	NR	KS	NR	+	e	NR	NR	NR	+	NR	NR	NR	NR
10b	1	M	White	NR	KS	NR	−		NR	NR	NR	+	NR		NR	NR
11	1	M	White	18.5	IHH	−	−		−	−	−	−	−	−	−	NL
12b	1	F	White Hispanic	17a	IHH	NA	NA		−	High arched palate	Scoliosis	Hyposmia	Synkinesia, Math Learning Disability	−	−	NR
13c	1	M	White	Adolescencea	IHH	+	+	e	−	NR	Osteopenia, +2.5 SD	NR	Repeated grade in school, no formal testing	−	−	Pituitary stalk is thin, embryonic cyst of the 2nd pharyngeal cleft
14c	1	M	White	Neonatal periodd	Neonatal hypogonadism, DD, CS	+	+	e	NA	Left coronal CS, brachycephaly, epicanthic folds, nevus flammeus & diffuse haemangioma on scalp, uplifted earlobes, downward slanting palpebral fissures	ND	NA	Autism & ID	ND	ND	Distortion of intracranial structures due to CS, thin CC posteriorly
15c	1	M	White	5d	Tall stature, DD	−	−		NA	Supernumerary incisor, Swallowing difficulties	Advanced by 2yrs at age 5, Weight +3 SD, Height +4SD	NA	Immature behaviour with slight adaptative defect, occipitalfrontal circumference +2SD	−	−	ND
ARHGAP5
16	1	M	White	14a	KS	+	−		−	−	Osteoporosis	+	−	−	−	NL
17	1	M	White	15a	KS	−	−		−	Crowded teeth	−	+	−	−	−	NR

Abbreviations: IHH = idiopathic hypogonadotropic hypogonadism, KS = Kallmann Syndrome (IHH + anosmia), NR = not reported, ND = not done, NA = not applicable, NL= normal, M= male, F=female, DD = developmental disorder, CS = craniosynostosis, ID = intellectual disability, CC = corpus callosum

^adiagnosis confirmed after age 18;

^bassessment by outside provider;

^cidentified through GeneMatcher;

^dunable to assess for IHH as children are pre-pubertal;

^eunilateral

Next, we examined ES data from 1,386 unrelated IHH probands in our cohort for additional RSVs in ARHGAP35. Four protein-truncating RSVs in ARHGAP35 ([NM_004491.4] c.345delC, p.(Tyr116Ilefs*55); c.352A>T, p.(Lys118*); c.516del, p.(Asn173Thrfs*23); c.1800_1803del, p.(Val601Tyrfs*26)) were identified in Families 2–5 (Figures 1- 2; Table 1; Supplementary Figure 1; Supplementary Table 4). Similarly, the probands in Families 2–4 did not carry any rare PTVs in known IHH genes. Family 5’s proband harbored a TUBB3 [NM_006086.3] PTV (c.815G>A, p.Trp272*) of unknown significance (VUS): TUBB3 PTVs have not been associated with human disease, and mouse modeling suggests no brain or reproductive defect from complete loss of Tubb3.^26,27 There were rare missense VUS in known IHH genes in Families 2 and 5 (Figure 1, Supplementary Tables 4-5). Family 2’s proband harbors a heterozygous VUS in FGFR1 c.2182G>A p.E728K in the tyrosine kinase domain which is predicted to be deleterious. Family 5’s proband harbors a heterozygous missense VUS in SEMA3A ([NM_006080.2] c.2189G>A, p.Arg730Gln) that has been shown to have no effect on secretion or signaling activity of Sema3A in prior in vitro modeling and a novel heterozygous missense variant in KLB c.3086A>T, p.K1029I which according to SIFT is a low confidence, deleterious call.^28,29 To assess the pathogenic likelihood of protein-truncating RSVs in ARHGAP35 contributing to IHH, we performed gene-based burden testing for PTVs. This analysis revealed a significant enrichment unadjusted p=3.1E-06 in IHH probands (n=1,387) compared to gnomAD controls (n=70,133).

Data Sharing Identified ARHGAP35 PTVs in IHH, Neonatal Hypogonadism, and DD

Three additional individuals with rare PTVs in ARHGAP35 were identified using GeneMatcher¹⁵ (Figures 1–2; Table 1; Supplementary Table 4). The proband from Family 13 had IHH and neonatal hypogonadism (unilateral cryptorchidism, microphallus) with an unreported PTV in ARHGAP35 (c.2565C>A, p.(Tyr855*)). Family 14’s proband, who is currently prepubertal, had evidence for neonatal hypogonadism (unilateral cryptorchidism, microphallus), left coronal craniosynostosis, and DD; he carries a previously unreported de novo PTV in ARHGAP35 (c.3283_3286delinsT, p.(Val1095*)). Family 15’s proband is also prepubertal, has tall stature with advanced bone age, and has DD with a rare PTV in ARHGAP35 (c.325C>T, p.(Arg109*)).

p190 RhoGAP zebrafish mutants display GnRH neuron defects.

Zebrafish (D. rerio) are a tractable model to study genes implicated in IHH.^3,4,30 To explore whether p190 RhoGAP influences the establishment of the GnRH neuronal network during early development, we generated and characterized zebrafish mutant models with a transgenic reporter marking the relevant neuronal population, gnrh3-expressing neurons [tg(gnrh3:egfp)].²²

We identified the zebrafish orthologs of ARHGAP35 through reciprocal BLAST searches of the human protein with the translated zebrafish genome (Supplementary Figure 2a-b). The zebrafish genome harbors two paralogs of ARHGAP35, likely resulting from the major teleost genome duplication event³¹ (arhgap35a: 80% similarity, 69% identity; arhgap35b: 78% similarity, 67% identity for human vs. zebrafish protein). Mining of publicly available transcriptomic datasets revealed that both transcripts are present from the zygote stage through larval day 5 (EMBL-EBI Expression Atlas). Notably, arhgap35b transcript has been reported at substantially lower levels in the zebrafish head than arhgap35a at 5 dpf (arhgap35b level 35% of arhgap35a)³²; however, we could not exclude the possibility that both paralogs perform redundant and/or GnRH-relevant cellular roles and modeled both independently.

We used CRISPR/Cas9 genome editing to target both zebrafish genes in tg(gnrh3:egfp) embryos. We identified sgRNA targets in the coding regions, in vitro transcribed sgRNA, and co-injected with Cas9 protein in single-cell stage embryos. Founder (F0) mosaic mutants were grown to adulthood and outcrossed to adults harboring the gnrh3:gfp transgene to establish stable F1 heterozygous mutants. We isolated frameshifting variants in each transcript (arhgap35a:16 bp insertion resulting in p.His826Glnfs*34; arhgap35b: 7 bp insertion resulting in p.Asn1147Serfs*11; Supplementary Figure 2a-b). To validate whether each variant resulted in gene ablation, we performed F1 heterozygous in-crosses and monitored endogenous transcript levels using qRT-PCR starting from total RNA isolated from heads of homozygous mutants compared to WT siblings at 2 dpf. For both mutants, we observed significant reduction of endogenous transcript (p<0.01, Student’s t-test; 28%, 40%of WT levels for arhgap35a and arhgap35b homozygous mutants, respectively; Supplementary Figure 2d).

After establishing significant gene disruption in each p190 RhoGAP zebrafish mutant, we quantified GnRH3-GFP+ cells with live automated imaging of fluorescent signal at 5 dpf in larvae resulting from heterozygous mutant adult in-crosses (Supplementary Table 7). Notably, we consistently observed expected Mendelian ratios for each of the mutant lines, arguing against the possibility of early lethality (Figure 3b, d; Supplementary Figure 3). Further, we observed no apparent developmental delay or other morphological defects in mutant animals (Supplementary Figure 3a, d). Measurement of GnRH3-GFP+ area on dorsal images revealed a significant reduction for arhgap35a homozygous mutants compared to WT (p=0.0019, one-way ANOVA; Figure 3a-b); with no detectable differences in head size area (Supplementary Figure 3a-c). However, we observed no differences between arhgap35b homozygotes and WT counterparts (GnRH3-GFP+ area or head size). Notably, heterozygous arhgap35a animals do not display detectable GnRH phenotypes (Figure 3a, b), however this disparity between zebrafish and humans is well-recognized for several human genes implicated in Mendelian disease^33–35. We speculate that genetic redundancy and/or genetic compensation in zebrafish likely results in genetic buffering that is different from humans³⁶.

Figure 3. Characterization of GnRH3-GFP neuronal patterning in arhgap35a, arhgap35b, and arhgap5 mutant zebrafish.

a, c, e. Representative dorsal images acquired from 5 dpf larvae to evaluate gnrh3:gfp signal. Anterior, left; posterior, right. Dotted lines indicate eyes and the most anterior region of the head. Scale bar, 100μm. b, d, f. Quantification of GnRH3-GFP+ signal in 5 dpf siblings compared across genotypes obtained from in-crosses of heterozygous mutant adults. Abbreviations: au, arbitrary units; WT, wild type; Het, heterozygous; Hom, homozygous; ns, not significant; significant differences detected with a one-way ANOVA; *p<0.05, **p<0.01; error bars indicate standard deviation.

Together, these data suggest that: (1) arhgap35a, the more highly conserved paralog that is more highly expressed in anterior structures, impacts the phenotype of a relevant GnRH neuronal population offering a mechanistic link between ARHGAP35 dysfunction and reproductive phenotype of our human cohort; and (2) arhgap35b does not appear to be necessary for GnRH3-GFP neuronal patterning in larval zebrafish.

Heterozygous ARHGAP35 missense variants are only enriched in the RhoGAP domain in IHH

In addition to the PTVs, IHH subjects also harbor ARHGAP35 missense variation, but significant enrichment was observed only for missense variants that affect the RhoGAP domain (adjusted p=4.9E-03; IHH cohort: n=6/2774 alleles; gnomAD: n=83/231350 alleles, RhoGAP domain missense in Figure 2; Table 1; Supplementary Tables 4-6). Expanding this analysis to include ARHGAP35 p.P1433A with a maximum MAF in the Ashkenazi Jewish population of 0.39% does not change the observed enrichment (adjusted p=1E-04; IHH cohort: n=7/2774 alleles; gnomAD: n=126/231350 alleles). Notably, a single de novo missense variant previously published in a DD study also occurs in the RhoGAP domain (p.P1407S; Figure 2)¹⁴, suggesting that RSVs affecting this domain may be pathogenic. Consistent with this finding, three of the seven ARHGAP35 missense RSVs identified in our cohort are hitherto unreported, predicted deleterious by in silico programs (Polyphen-2; SIFT), and found in families with no other known IHH RSVs or CNVs (Figures 1–2, Supplementary Table 4-5) suggesting they likely to contribute to disease. Two of the seven families with missense RSVs in ARHGAP35 harbored pathogenic RSVs or CNVs in ANOS1 that explained their phenotypes (Family 8 and 10, Figure 1, Supplementary Table 4).

Missense Variants in ARHGAP35 have variable impact on GAP Activity

The in vitro p190A-GAP assay evaluates the ability of the ARHGAP35 RhoGAP domain to enhance the rate of GTP hydrolysis by the small GTPase RhoA, a critical protein function²⁴. To examine this function, WT p190A-GAP or the indicated mutant proteins were expressed as GST fusion proteins in E. coli for in vitro GAP assays. WT p190A-GAP and mutant proteins were purified and resolved by SDS-PAGE, showing efficient protein stability and solubilization (Supplementary Figure 4). As expected, RhoA showed little intrinsic GTPase activity alone while the addition of WT p190A-GAP significantly enhances the hydrolysis of GTP (Figure 4). On the contrary, the ARHGAP35 p.(Arg1284Trp) variant, which occurs at the critical conserved “arginine finger,” demonstrated severely impaired phosphate release; thus confirming abolished GAP activity.³⁷ Control variants (p.(Arg1419Cys), p.(Asp1340Asn)) and case variants (p.(Met1412Thr), p.(Ala1414Thr), p.(Arg1350Gln), p.(Pro1433Ala)) in non-conserved residues³⁷ which were also seen in gnomAD, all showed normal GAP activity. The other previously unreported ARHGAP35 missense variants, p.(Pro1331Leu) and p.(His1369Asp), could not be tested because of insufficient purified protein amounts. However, since Pro1331 is a well conserved residue amongst the various RhoGAP proteins and His1369 is localized in the four-helix bundle of the RhoGAP domain³⁷, it is most likely that these variants cause protein misfolding and impaired GAP activity. This well-validated assay provides a read-out of ARHGAP35 GAP activity; however, because it utilizes just the RhoGAP domain in vitro, variants that result in conformational changes that alter the ability of the protein to bind targets or destabilize the protein could be inaccurately classified as benign.

Figure 4: RhoGAP modeling of ARHGAP35 missense variants in vitro.

In vitro GAP activity assay towards RhoA in the absence (intrinsic) or presence of p190A-GAP wild type (WT) or mutant proteins. Arg1284Trp substitution is a loss-of-function variant in p190A GAP domain. Data are presented as mean ± S.E.M of three independent experiments (n=3). **, p < 0.01; ***, p < 0.001; unpaired student’s t test.

ARHGAP35 RSVs are Inherited and Result in Variable Expressivity

Across all families in the IHH cohort, 50% of probands with IHH and ARHGAP35 RSVs were anosmic. Fifty percent (5/10) of male probands reported in this study displayed severe neonatal hypogonadism as evidenced by microphallus and/or cryptorchidism. In addition, 2 of 5 individuals with IHH and PTVs in ARHGAP35 underwent reversal of their hypogonadism in adulthood, resulting in normal gonadal function in the absence of medications. Regarding additional non-reproductive phenotypes, 3 of 14 individuals with IHH in 2 families reported learning disabilities (Table 1; Supplementary Table 4). For all individuals in the IHH cohort for which parental genotyping data is available, the ARHGAP35 RSV is inherited; although, the Family 14 proband from GeneMatcher with neonatal hypogonadism and DD had a de novo PTV.

Heterozygous ARHGAP5 PTVs are present in two Kallmann syndrome Families

De novo analysis of ES data in 175 IHH trios revealed a previously unreported heterozygous RSV in ARHGAP5 ([NM_001030055.1] c.2366dupC, p.(Phe790Ilefs*2)) in Family 16(Figure 5, Supplementary Table 4). ARHGAP5 is the most closely related member to ARHGAP35 (51% identity, 68% amino acid similarity in humans) and is highly constrained for PTV (LOEUF=0.263).¹⁰ No additional protein-truncating “RSVs of interest” were found in Family 16. Examination of the larger KS cohort (IHH + anosmia, n = 702) revealed one additional previously unreported heterozygous PTV in ARHGAP5 (c.1504delT, p.(Tyr502Metfs*3)) in Family 17. Identifying two KS probands with rare protein-truncating RSVs in ARHGAP5 motivated us to test for gene-based burden which revealed a moderate enrichment for PTVs (unadjusted p=0.003) in KS probands (alleles=2/1402) compared to gnomAD controls (alleles=13/233254).¹⁰

Figure 5: Kallmann Syndrome Pedigrees with rare ARHGAP5 variants.

a. Pedigrees of probands with previously unreported loss of function variants. ”+”, wild-type allele; “M”, variant allele. b. Schematic of human ARHGAP5 protein (GenBank ID: NP_001025226.1) and position of variants identified in IHH cases. FF, phenyalanine domains (blue); pG1, pseudo-GTPase domain 1 (salmon); pG2, pseudoGTPase domain 2 (salmon); K, lysine rich domain (gray); RhoGAP, Rho GTPase-activating proteins domain (yellow); P, proline rich domain (gray); amino acid position is labeled at the bottom.

Based on the modest enrichment for PTVs in ARHGAP5 and given the de novo nature of one of the identified variants, we examined the relevance of ARHGAP5 in our zebrafish model. We identified a single zebrafish ortholog of ARHGAP5 (arhgap5: 85% similarity, 73% identity for human vs. zebrafish amino acid sequence) through reciprocal BLAST searches of the human protein with the translated zebrafish genome that is present from zygote stage through larval day 5 (Supplementary Figure 2c; EMBL-EBI Expression Atlas). We used CRISPR/Cas9 genome editing to target arhgap5 in tg(gnrh3:egfp) embryos, as described above for ARHGAP35. We isolated a frameshifting variant in arhgap5: 86bp insertion resulting in p.Val1038Metfs*6For mutants. We observed significant reduction of endogenous transcript (p<0.01, Student’s t-test; 47% of WT levels for arhgap5 homozygous mutants; Supplementary Figure 2d).

After establishing significant gene disruption in arhgap5 mutants, we quantified GnRH3-GFP+ cells with live automated imaging of fluorescent signal at 5 dpf in larvae resulting from heterozygous mutant adult in-crosses (Supplementary Table 7). Notably, we consistently observed expected Mendelian ratios for arhgap5, arguing against the possibility of early lethality (Figure 3f; Supplementary Figure 3). Further, we observed no apparent developmental delay or other morphological defects in mutant animals (Supplementary Figure 3g). Comparison of GnRH neuronal area in arhgap5 homozygotes versus WT siblings did not reveal any significant differences with stringent testing (one-way ANOVA; Figure 3e-f; Supplementary Figure 3g-i); however, we noted a trend toward a reduced GFP+ area that was modestly significant using a pairwise comparison between these two genotypes only (Student’s t-test; p<0.05). These data indicate that arhgap5 might play a role in GnRH neuronal patterning but could not be captured using our current mutant model system and experimental parameters. Taken together with our human genetic data, these findings implicate ARHGAP5 as a potential candidate contributor to IHH pathogenesis, but this assertion requires further human genetic validation.

ACMG Classification of ARHGAP35 and ARHGAP5 Variants

To determine how variants in ARHGAP35 and ARHGAP5 should be evaluated clinically, we incorporated the above results into ACMG criteria for classifying variants. ³⁸ PTVs in ARHGAP35 are likely pathogenic based on the human, in vivo, and population and computational data (Supplementary Table 5). In contrast, missense RSVs in the RhoGAP domain have variable interpretations. ARHGAP35 p.(Arg1284Trp) is likely pathogenic, while p.(Ala1414Thr) and p.(Met1412Thr) are likely benign. The other four missense variants in ARHGAP35 are VUS: p.(Pro1331Leu), p.(Arg1350Gln), p.(His1369Asp), and p.(Met1412Thr). For ARHGAP5, the de novo PTV is likely pathogenic (p.(Phe790Ilefs*2)), but the other PTV is a VUS (p.(Tyr502Metfs*3)).

STRING pathway analysis

To explore further the p190 RhoGAP interactors, we assembled protein-protein interaction data. STRING pathway analysis revealed high confidence interaction scores connecting ARHGAP35 to ARHGAP5 and to the semaphorin network through PLXNC1, PLXNB1 and SEMA4D. (Supplementary Figure 5). Seeding with known IHH genes (Supplementary Table 1) confirmed these findings and established connections for ARHGAP35 and ARHGAP5 to known IHH genes (FGF17, KLB, FGFR1, FGF8, GNRH1, DUSP6) (Supplementary Figure 6).

DISCUSSION

This study implicates p190 RhoGAPs, ARHGAP35 and ARHGAP5, ⁷⁻¹¹in IHH. p190 RhoGAPs, mainly ARHGAP35, have been implicated previously in neurodegenerative diseases, DD, and human cancers.^14,39 Here, we report 14 individuals with IHH in 12 unrelated families who harbor different heterozygous protein-truncating or RhoGAP missense variants in ARHGAP35 in an autosomal dominant inheritance pattern who were identified from 1,387 families. We functionally validated the ARHGAP35 PTVs and one missense change (p.Arg1284Trp). Through GeneMatcher, we report three additional PTVs in ARHGAP35 in individuals with IHH, IHH-related phenotypes, and/or DD. We also report the discovery of 2 individuals with KS in 2 families with different PTVs in ARHGAP5. This report expands the biologic role of the p190 RhoGAPs into reproductive biology and implicates IHH as an ARHGAP35-related human disorder.

Among ARHGAP35-related human disorders described to date, germline RSVs similar to this study have only been reported once in humans with DD wherein de novo ARHGAP35 variants were identified.¹⁴ DD are clinically heterogeneous, common, and genetically characterized by an excess of de novo CNVs/RSVs.¹⁴ The genetic characteristics of DD and IHH differ with regards to the inheritance mode and the distribution of ARHGAP35 RSVs. In contrast to the de novo variation in the DD study, ARHGAP35 RSVs in IHH identified in this study were predominantly inherited with incomplete/variable penetrance, a pattern consistent with other known genetic forms of IHH.¹ While a similar percentage of RSVs were missense in the two cohorts (58% IHH, 53% DD), ninety percent (9/10) of identified missense RSVs in DD were outside the RhoGAP domain; whereas only missense variants within the RhoGAP domain were enriched in IHH.¹⁴ Precise determination of these seemingly overlapping ARHGAP35 allelic disorders (IHH and DD) is hindered by a lack of information about reproductive phenotypes in the DD cohort.¹⁴ In this regard, it is notable that in 2/12 families with ARHGAP35-related IHH, neurodevelopmental phenotypes such as learning disabilities (dyslexia, dyspraxia, math learning disability) were present. Furthermore, through GeneMatcher, we identified a de novo ARHGAP35 PTV in a child with signs of neonatal hypogonadism and autism with intellectual disability. Thus, our observations suggest a phenotypic overlap between the ARHGAP35-related disorders of DD and IHH and that DD and IHH may lie across a longitudinal phenotypic spectrum. Indeed ARHGAP35 is expressed across fetal development in the brain: the hypothalamus, thalamus, amygdala, ventricular zone, and marginal zone.⁴⁰ Additional deep, longitudinal phenotyping and genetic analyses for all individuals with RSVs in ARHGAP35 are needed to understand this pleiotropic phenomenon and to discern the extent to which IHH and DD represent ARHGAP35-related overlapping disorders across a phenotypic continuum.

Multiple animal models demonstrate neurodevelopmental phenotypes that support the observed pleiotropy in humans. In zebrafish, loss of arhgap35a, the predominant p190 RhoGAP in the zebrafish brain, resulted in reduced GnRH neuronal area. Supporting this observation, recent in vitro sequencing of human GnRH neurons identified ARHGAP35 as part of a network of top 50 upregulated genes across GnRH neuronal development.⁴¹ While GnRH neurons have not been examined in p190 RhoGAP mouse models, ARHGAP35 is expressed in mouse GnRH neurons, and GnRH-containing fibers form paths caudally and rostrally around the anterior commissure (AC).^42,43 Mice with a homozygous truncated variant of Arhgap35 die perinatally and demonstrate loss of AC and hippocampal commissures (HC), disorganized cerebral cortex, and agenesis of the corpus callosum (CC) due to neuronal misrouting as a result of impaired adhesion and adhesion-mediated signaling.^8,9 Defects in the AC, HC, and CC can result in DD phenotypes in humans, including the dyspraxia seen in this report in a patient with KS due to an ARHGAP35 PTV.^7,44,45 In addition, a mouse model with the variant ARHGAP35 p.L1396Q demonstrated hypoplastic and/or glomerulocystic kidneys,²⁴ which has not been observed in our IHH cohort or other animal models.⁷ Further examination of reproductive, renal, and neuronal phenotypes across animal models, humans, and ARHGAP35 variant type is needed to dissect the observed pleiotropy.

Our pathway analyses showed that ARHGAP35 interacts with three known IHH genes, SEMA3A, SEMA3F, and PLXNA1^28,46–48, and two candidate IHH genes, SEMA4D and PLXNB1.¹³ This raises the hypothesis that ARHGAP35 may result in IHH through regulation of semaphorin signaling. The SEMA3A-F/neuropilin/PLXNA1 complex provides signal guidance for the vomeronasal axons along which GnRH neurons migrate.^13,48 Loss of this signaling results in GnRH neurons that cannot migrate past the cribriform plate into the brain in mice.^13,46,47 In addition, SEMA4D/PLXNB1 signaling stimulates GnRH-1 neuronal migration and regulates GnRH-1 expression in the hypothalamus.¹³ Loss of this signaling results in reduced GnRH neuronal number in the hypothalamus along with decreased GnRH-1 expression resulting in impaired fertility in mice.¹³ ARHGAP35 physically associates with both PLXNA1 and PLXNB1, and the RhoGAP activity of ARHGAP35 is needed for plexin signaling.¹² Knock-down of ARHGAP35 transcripts using siRNA: 1) reduces semaphorin-mediated repulsion of endothelial cells, a marker of semaphorin impact on migration, for Sema3A, Sema3F, and Sema4D; and 2) results in the loss of Sema4D-induced neurite outgrowth in PC12 neural cells.¹² Haploinsufficiency of ARHGAP35 could result in decreased semaphorin/plexin signaling in SEMA3A, SEMA3F, and SEMA4D pathways resulting in impairment of GnRH neuronal migration and reduced neurite outgrowth. However, further studies will be needed to validate these putative downstream consequences relating to ARHGAP35 deficits.

Because of the pathogenicity of PTVs presented in this report, we would recommend screening of ARHGAP35 in patients with IHH, regardless of the presence or absence of associated phenotypes. We recognize that only one missense variant in this study was deemed to be pathogenic, pointing to the need for additional investigation of these types of nucleotide changes. We also note gnomAD may have limitations as a control arm for gene-based burden testing. However, because IHH genes are characterized by variable penetrance, it is likely that disease causing variants are in gnomAD biasing any gene-based burden testing towards the null.

In conclusion, this study provides robust human genetic evidence and functional data implicating heterozygous PTVs in ARHGAP35 and a single ARHGAP35 missense variant as causal genetic drivers for IHH. ARHGAP35’s role as a master regulator of multiple semaphorins may underlie the pathogenesis of the IHH phenotype. Phenotypic features of IHH patients with ARHGAP35 variants and their prior association with DD strongly imply that ARHGAP35 variants result in pleiotropic developmental phenotypes. Future research into the mechanisms of ARHGAP35 pleiotropy and detailed phenotyping across multiple organ systems of individuals with ARHGAP35 RSVs are needed to decipher the full extent of ARHGAP35-related human phenotypes.

Data Availability:

Deidentified individual level data has been shared in Figures, and Supplemental Tables and Figures. Sequencing data for variants in this manuscript are deposited into ClinVar (https://www.ncbi.nlm.nih.gov/clinvar/; SUB11458441).