De novo rare EMX2 variants lead to idiopathic hypogonadotropic hypogonadism

Abstract

Purpose:

The genetic etiology of infertility remains unknown. To identify genes for human infertility, we applied a de novo variant analysis in 142 parent-proband trios with idiopathic hypogonadotropic hypogonadism (IHH), an infertility disorder caused by Gonadotropin Releasing Hormone (GnRH) deficiency.

Methods:

Rare de novo copy number and single nucleotide variants (CNVs & SNVs) were called from exome sequencing data of the IHH trios. An association study of common EMX2 variants and disease outcomes was performed in the Massachusetts General Brigham Biobank (N=65,253). GnRH neuronal development and migration was studied in organotypic explants with knocked down of Emx2 and in a mouse model lacking Emx2.

Results:

We identified that the gene EMX2 harbored both rare de novo CNVs and SNVs. Rare de novo EMX2 variants led to IHH, developmental delay and hearing loss. Common EMX2 variants were linked to infertility, Parkinson's disease and hearing loss. Knockdown of Emx2 in nasal explants resulted in attenuated GnRH cell migration and GnRH cells were confined to nasal regions of Emx2 knock out (KO) mice, consistent with IHH pathogenesis.

Conclusion:

By utilizing a de novo variant analysis and cellular assays, EMX2 was uncovered as a gene for human infertility.

Introduction

Mammalian reproduction is governed by hypothalamic Gonadotropin Releasing Hormone (GnRH) neurons that sit at the top of the hypothalamic-pituitary-gonadal cascade under the tight regulation of the Kisspeptin Neurokinin Dynorphin (KNDy) neurons(1, 2). During early development, the GnRH neurons originate outside the central nervous system, arising from neural precursors within the olfactory placode, migrating along olfactory sensory/vomeronasal/terminal nerve axons to reach their destination within the forebrain.

Defects in GnRH neurons lead to reproductive disorders, including the Mendelian disease of Idiopathic Hypogonadotropic Hypogonadism (IHH), also known as congenital hypogonadotropic hypogonadism (CHH) (OMIM IDs 619755, 614842, 614880, 615267, 610628, 146110, 614840, 618841, 615266, 614897, 616030, 615270, 614841, 147950, 612370, 614838, 612702, 614858, 229070, 614839, 615269, 619718, 614837, 228300, 244200, 615271 and 308700). Nearly half of IHH patients display olfactory abnormalities [Kallmann syndrome (KS)] attesting to a joint failure of olfactory and GnRH development, and 40% of IHH cases display additional non-reproductive defects, including bone abnormalities (23%), eye defects (13%), hearing loss (6%), kidney abnormalities (2%), etc.(3, 4). Studies in IHH individuals have yielded significant insights into the hypothalamic control of human reproduction. Despite these advances, the genetic etiology in 56% of IHH patients remains unknown(5). While rare de novo variants account for a large proportion of the genetic etiology underlying neurodevelopmental disorders in large scale sequencing studies, such approaches have not been broadly applied in IHH, where studies have focused on pedigree-based analyses, which are also limited by the subfertility of the affected probands(6-9).

Here, we analyzed de novo rare copy number and single nucleotide variants (CNVs and SNVs) in a large cohort of IHH trios (N=142) to identify genes for human reproduction. Candidate genes, harboring both rare de novo CNVs and SNVs, were validated by a rare variant association testing (RVAT) in a larger cohort of IHH probands (N=1,322) compared to gnomAD controls (N=125,784)(10). We identified empty spiracles homeobox 2 (EMX2, HGNC ID 3341), a gene previously linked to olfactory neuronal development, as a causative gene of IHH. We investigated neuroendocrine GnRH (nGnRH) neurons in Emx2 null mice and changes in GnRH cells after knockdown of Emx2 in nasal explants. The total number of nGnRH cells in Emx2 knockouts was increased in nasal regions from E12.5-E16.5. However, the number of nGnRH cells found in the mutant forebrain did not parallel this increase. In nasal explants, knockdown of Emx2 decreased the distance GnRH cells migrated. Together, these results indicate that Emx2 is required for nGnRH neurons to reach the forebrain in mice, consistent with EMX2 contributing to the IHH etiopathogenesis.

Methods:

Phenotypic evaluation for the IHH cohort:

A total of 1,322 IHH probands [902 males and 420 females with both normosmic IHH (nIHH) and IHH with anosmia (i.e. KS)] and 1,902 family members were recruited. IHH was defined as pubertal failure (>18 years old [yo]), low sex steroids, low/normal gonadotropin levels, normal other pituitary hormones and no pituitary abnormalities in brain MRI(11).

Genomic DNA extraction and exome sequencing (ES):

Genomic DNA of all participants was extracted from peripheral blood samples and underwent ES at the Broad Institute (Cambridge, Massachusetts, USA) and the Yale Center for Mendelian Genomics (Orange, Connecticut, USA). Alignment of the ES data against the reference genome (hg19) and initial quality control were applied using GATK best practices (Broad Institute)(12).

SNV and CNV calling/confirmation:

SNV calling was performed using GATK (HaplotypeCaller and GenotypeGVCFs). VCF files were annotated using SnpSift 4.3k and Ensembl VEP release 93. Seqr was used for the de novo SNV analysis(13). Sanger sequencing and reviewing of the binary alignment/map (BAM) that are used to store aligned DNA sequencing reads allowing for review of total counts and affected alleles (minimum reads=10 and allele balance >25%) were used for SNV confirmation. CNV calls were also conducted using the GATK-gCNV pipeline(5). Samples with ten or more de novo events were classified as “outlier samples” (N=4 samples) (14).

RVAT:

The total number of alternate and reference alleles across each gene were aggregated into an alternate allele count (AAC) and reference allele count (RAC). The AACs and RACs were used in a rare variant burden test between the IHH and gnomAD participants(10) [Fisher’s exact test, p value <0.05/4 (0.0125), the number of genes tested].

Novel gene discovery pipeline:

Genes harboring both rare (MAF <1%) de novo CNVs and SNVs were prioritized. To confirm parent-child relationships, genotyping was performed at the Broad Institute using the Infinium^® PsychArray v1.1 and B. PLINK1.9 was used to assess for a non-paternal or non-maternal event by examining Mendelian errors in inheritance(15). Candidate genes were prioritized with an RVAT for protein truncating variants (PTVs: deletions, nonsense, essential splice-site, and frameshift) in the IHH (N=1,322) compared to the gnomAD cohort (N=125,784)(10).

Human evidence from additional cohorts/databases:

A dedicated search was performed in: (i) The Mass General Brigham Biobank (MGBB), which maintains genomic data from 65,253 consented participants. Genotyping data was searched for common EMX2 variants previously linked to common traits(16), and a phenotyping analysis was conducted based on: ICD10 codes (Supplementary Table 1) and health questionnaires completed by the MGBB females (N=34,956). A Fisher’s exact test (p value <0.05/number of common variants tested) was used; (ii) A Boston Children’s Hospital (BCH) cohort of participants with chromosomal defects detected by karyotypes/chromosomal microarrays; (iii) Publicly available variant databases(17-22) (Supplementary Table 2); (iv) Cases identified through the GeneMatcher dataset(23).

Ex Vivo Model:

Organotypic explants and the nGnRH cell isolation/microarray dataset from single migrating nGnRH neurons were generated as previously described(24, 25). Gene expression was calculated using the robust multiarray average (RMA) procedure. GnRH cells expressing EMX2 (N=4) were quantified after staining for EMX2 (rabbit, 5ug/ml, ThermoFisher PA5-103315, directly conjugated 555 secondary antibody) and GnRH (monoclonal mixture SMI-41 [Abcam, 1:6000] + F1D3C5 [Dr. A Karande, 1:4000, directly conjugated 488 secondary antibody). Explants were treated with two antisense oligos (ASOs; ACGTGGCACGTACCTTGGAAGC and GGAAACTCTCGGGACTTGTGTC) to knockdown Emx2 expression, or a randomer oligo (CACAGAGGGTGTGGACGCCAAA) with the same length and GC content(26). Post hoc analysis included the total number of GnRH cells, the distance the cells migrated, and the olfactory axon outgrowth(27). A Chi-square test was used to analyze cell migration.

Mice:

Animal experiments were conducted under NIH protocols (#1221-23 and 1212-17) using Emx2^Cre(28), Emx2 ^+/− (29) and Rosa26 ^tdTomato (Jackson Laboratory) mouse lines. Emx2 knockout embryos were generated by crossing Emx2^+/− males or Emx2^Cre; Rosa^tdT/tdT males with Emx2^+/− females, generating Emx2^Cre/−; Rosa^tdT/+ null mutants. Embryos were collected at E12.5, E14.5, and E16.5. Three consecutive series were generated for E12.5 and four consecutive series were generated for E14.5/E16.5 embryos. Cut sections were stored at −80°C prior to immunostaining.

Immunocytochemistry:

Primary Antibodies used were Rabbit polyclonal anti-GnRH-1 (SW-1; RRID:AB_2629221), rat anti-phospho-Histone-3 (pHis3, Novusbio. NB600-1168, mitotic cell marker), mouse anti-peripherin (Stem Cell 7C5, olfactory sensory marker), and rabbit anti-peripherin (Chemicon AB1530). Staining for GnRH, phospho-histone3 or peripherin-positive olfactory cells/fibers was performed as previously described (30).

Analysis of GnRH cell and pHis3 cell counts.

GnRH positive cells were counted in all series, in four regions (olfactory pit/vomeronasal organ (OP), olfactory tracks, nasal forebrain junction (NFJ), and brain), from E12.5, E14.5 and E16.5 (N=3/genotype). The number of pHis3 positive cells was counted in OP at E12.5 (N=4 KO and 3 WT). Statistical analysis was conducted using Statview 5 (SAS Institute, Inc) or Prism 7 (GraphPad Inc). Two-way ANOVA, followed by Fisher’s PLSD test was performed on total GnRH cell counts, pHis3 cell counts, and the distribution of nGnRH cells.

EMX2 lineage expression in nGnRH cells.

Sections from Emx2^Cre/+; Rosa^tdT/+ embryos (N=2) were stained using anti-GnRH-1 and a directly conjugated 488 secondary antibody and imaged with a spinning disk confocal system CSU10 (Yokogawa), an Eclipse TE200 microscope (Nikon), an EMCCD ImageM digital camera (Hamamatsu), and iVision software (BioVision). Images were processed for viewing using ImageJ (W Rasband, NIH, Bethesda, MD, United States) and Photoshop (Adobe Systems Inc., Salinas, CA) software.

Results:

Rare de novo variant analysis identifies EMX2 as a novel candidate gene for IHH:

The 142 IHH trios harbored 43 rare de novo CNVs/deletions and 213 rare de novo SNVs, spanning 263 and 191 genes (Figure 1). Four genes were affected by both de novo CNVs and SNVs: FGFR1 (fibroblast growth factor receptor, HGNC ID 3688), ANOS1 (anosmin 1, HGNC ID 6211), EMX2 (empty spiracles homeobox 2, HGNC ID 3341), and VCX (variably charged, X chromosome, HGNC ID 12667) (Supplementary Table 3). An RVAT for PTVs in the IHH cohort (N=1,322) compared to gnomAD controls (N=125,784)(10) revealed that only 3/4 genes were enriched: two known IHH genes [FGFR1(6) and ANOS1(9)] and one candidate gene EMX2 (Supplementary Table 4). The identification of well-validated IHH genes serves as both a positive control proxy for the de novo-based approach employed in this report and strengthens the candidacy of EMX2 as a causative for IHH gene.

Figure 1: Discovery pipeline for detection of novel genes for IHH.

ES from 142 IHH trios was analyzed for rare de novo CNVs/deletions and SNVs. Candidate genes were subjected to a rare protein truncating variant association test between the IHH cohort and gnomAD controls(10). The discovery pipeline identified 2 known IHH genes (ANOS1 and FGFR1) and 1 gene novel for IHH (EMX2). Created in BioRender. Stamou, M. (2025) https://BioRender.com/ii6c7cn.

Rare de novo EMX2 variants lead to IHH, developmental delay and hearing loss:

(i). Rare de novo CNVs (Table 1):

Table 1:

Characteristics of probands with pathogenic and likely-pathogenic EMX2 rare variants.

Case	EMX2 variant	Segregation	Age	Gender	Reproductive phenotypes	Non-reproductive phenotypes
Copy number variants
1	9 MB deletion (NC_000010.10:g.115312639_124097792del)	De novo	17 (followed until the age of 21)	Female	KS	Developmental delay, bilateral hearing loss, microcephaly, hypotonia, bilateral clinodactyly
2	5 MB deletion (NC_000010.10:g.117938033_122568487del)	De novo	8	Male	Cryptorchidism	Developmental delay, hypotonia, clinodactyly
3	2.5 MB deletion (NC_000010.10:g.119100470_121640030del)	De novo	9	Female	The ultrasound could not visualize the uterus/gonads	Developmental delay, bilateral hearing loss, microcephaly
Single nucleotide variants
4	NP_004089.1:p.(Lys199Gln) (NM_004098.4):c.595A>C)	De novo	18	Male	nIHH, cryptorchidism, microphallus	Developmental delay, bilateral hearing loss
5	NP_004089.1:p.(Lys199Glu) (NM_004098.4:c.595A>G)	De novo	9.5	Male	Cryptorchidism	Learning disabilities, ADHD, bilateral hearing loss
6	NP_004089.1:p.(Ser111Ter) (NM_004098.4):c.332C>A)	Unknown	25	Male	KS, cryptorchidism	None

Clinical characteristics of the probands with EMX2 rare variants (CNVs and SNVs) from the adult and the pediatric populations. EMX2 variants carried by the probands, their sex, diagnosis and additional non-reproductive features are shown. nIHH = normosmic Isolated Hypogonadotropic Hypogonadism, KS = Kallmann syndrome, i.e., IHH with anosmia, ADHD = Attention Deficit/Hyperactivity Disorder (OMIM ID 143465). The ACMG classification of all variants is provided in the supplemental information (Supplementary Table 6).

Case 1 harbored a novel de novo 9 MB chr10q25-26 deletion (NC_000010.10:g.115312639_12409779del) spanning 49 genes, including EMX2. The female proband presented with IHH (LH <1.6 IU/L, FSH <1.6 IU/L and estradiol <20 pg/mL), anosmia, developmental delay, microcephaly, hypotonia, and bilateral hearing loss. Other pituitary hormones and brain MRI were normal. She did not carry any rare variants in other causative IHH genes. Review of the BCH cohort revealed 2 additional participants harboring de novo 10q25 deletions spanning EMX2: (i) Case 2, an 8 yo male harbored a 5MB deletion (NC_000010.10:g.117938033_122568487del) and presented with developmental delay, hypotonia, bilateral cryptorchidism, and clinodactyly; and (ii) Case 3, a 9 yo female with a 2.5MB deletion (NC_000010.10:g.119100470_121640030del) presented with developmental delay, bilateral hearing loss, and microcephaly. She underwent an abdominal ultrasound without visualization of the uterus or gonads. The minimal critical region, i.e., the overlapping genomic region in all CNVs, spanned 18 genes, including EMX2 (Figure 2A). Apart from EMX2, no other gene within this region exhibited significant enrichment for rare variants (both PTVs and missense) in the large IHH cohort (N=1,322) compared to gnomAD controls (N=125,784)(10) (Supplementary Table 5), further supporting the role of EMX2 in the etiopathogenesis of IHH.

Figure 2: Rare human genetic variants identified in EMX2.

2A) The minimal critical region, i.e. the overlapping genomic region, for chr10-q25-26 CNVs identified. 2B) Rare EMX2 variants were identified across all protein domains. Rare deletions (CNVs, given that they span the whole gene: NC_000010.10:g.115312639_124097792del, NC_000010.10:g.117938033_122568487del and NC_000010.10:g.119100470_121640030del), the stop gain variant [upstream of the DNA binding domain, [NM_004098.4:c.332C>A, NP_004089.1:p.(Ser111Ter)], and the 2 de novo missense EMX2 variants (shown in red: [NM_004098.4:c.595A>C, NP_004089.1:p.(Lys199Gln) and NM_004098.4:c.595A>G,NP_004089.1:p.(Lys199Glu)] affected the DNA-binding domain. Created in BioRender. Stamou, M. (2025) https://BioRender.com/03yqy25.

(ii). Rare de novo SNVs (Table 1):

A de novo likely pathogenic EMX2 variant NP_004089.1:p.(Lys199Gln) (NM_004098.4:c.595A>C) was found in a proband (Case 4) with IHH (LH 0.3 IU/L, FSH 2.2 IU/L, and total testosterone 30 ng/dL), developmental delay and bilateral hearing loss. In addition, through the GeneMatcher platform, we identified a pediatric case presenting with a de novo likely pathogenic SNV within the same residue NP_004089.1:p.(Lys199Glu) (NM_004098.4:c.595A>G) (Case 5), who was diagnosed with bilateral cryptorchidism, learning disabilities, attention deficit disorder and bilateral hearing loss. The Lys199 residue resides within the EMX2 DNA-binding domain, a critical regulatory domain for its transcription factor (Figure 2B). In follow up to the identification of de novo variants in the trio cohort (N=142), we then searched for pathogenic EMX2 SNVs in probands of IHH families in whom parental DNA was not available (i.e., within the large IHH cohort of 1,322 probands). A novel nonsense EMX2 variant NP_004089.1:p.(Ser111Ter) (NM_004098.4:c.332C>A) was identified in an IHH proband with anosmia (Case 6). Notably, all 3 SNVs identified were deemed pathogenic and likely-pathogenic by the ACMG classification(31) (Supplementary Table 6), were novel (i.e., not found in the gnomAD dataset(10)), and none of the IHH probands harboring these SNVs carried variants in other IHH genes that were deemed causative based on their known inheritance mechanisms (Supplementary Table 6). Further, phenotypic analyses demonstrated overlapping non-reproductive phenotypes, including hearing loss and neurodevelopmental defects in cases with de novo EMX2 variants, suggestive of the pleiotropic nature of EMX2. Six additional missense variants of uncertain significance (VUS) (n=4) or likely benign SNVs (n=2) were also identified in IHH probands (Supplementary Table 6). While the 4/6 IHH probands with EMX2 VUS were not genetically solved by pre-existing variants in known IHH genes (Supplementary Table 6), the 2 likely benign EMX2 variants occurred in cases who each harbored a causal known IHH variant, suggesting that these 2 EMX2 variants were likely non-contributory(5, 32) (Supplementary Table 6). All variants were confirmed by detailed review of the binary alignment/map (BAM) files that are used to store aligned DNA sequencing reads allowing for review of total counts and affected alleles and/or Sanger sequencing. All confirmations are shown in Supplementary Figure 1. All variants have been reported in ClinVar(19) under SUB15295896, SUB15295887, SUB15295885, SUB15295876, SUB15295865, SUB15295863, SUB15295817 and SUB15295673.

In summary, this human genetic data shows that de novo EMX2 CNVs and SNVs lead to reproductive deficits (IHH in adults and genitourinary defects in children), hearing loss, and developmental delay, in both adult and pediatric populations. Nine out of 1,322 IHH participants (0.7% of the large IHH cohort) harbored rare variants in EMX2: 3 participants harbored pathogenic and likely pathogenic EMX2 variants (0.25% of the IHH cohort), and 6 participants (0.45% of the IHH cohort) carried VUS and likely benign variants in EMX2 as outlined above (Supplementary Table 6). Additionally, 3 pathogenic variants were identified in pre-pubertal participants with genitourinary defects and additional non-reproductive phenotypes that were similar to the ones observed in the adult IHH participants (Table 1). The pubertal progression and subsequent transition to IHH in the pediatric population will require longitudinal follow-up.

EMX2 variants are linked to reproductive deficits in publicly available databases and the MGBB:

A dedicated search in publicly available databases showed that rare EMX2 variants were linked to urogenital deficits and other non-reproductive phenotypes, including developmental delay, schizencephaly, and microcephaly (Supplementary Table 2). Additionally, review of genomic data from 65,253 MGBB participants revealed that EMX2 common variants were significantly associated with delayed menarche and female infertility (Supplementary Table 7). Notably, the common variant NC_000010.10:g.119379656A>G (rs181654) was linked to an increased prevalence of delayed menarche, hearing loss, and Parkinson’s disease, further strengthening the pleiotropic nature of EMX2.

EMX2 is expressed in nGnRH cells during prenatal development:

(i). Ex vivo model:

The RMA value for EMX2 in migrating nGnRH neurons was 5.04, similar to other transcription factors that regulate GnRH expression (Supplementary Table 8); ~94% of the GnRH cells in the explants expressed EMX2 (Figure 3A & 3B). In explants, GnRH cells migrate out of the nasal pits after 2-3 days and into the periphery along olfactory axons (18). Explants treated with antisense oligonucleotides against Emx2 and randomers were stained for GnRH and peripherin (Figure 3C). While no differences were found in GnRH cell number [p=0.5203; randomer=138±14 (N=21); antisense=152±16 (N=20)] or olfactory axon extension [(p=0.3216; randomer=1.275 ±39um (N=17); antisense =1357±77um (N=13)], a significant difference was found in the location of GnRH cells (X² p= <0.0001; X²=112.734). Cells of Emx2 knockdowns remained closer to the explant compared to controls (Figure 3D). Thus, knockdown of Emx2 in explants attenuates GnRH cell migration.

Figure 3: EMX2 is expressed in GnRH neurons and knockdown of Emx2 attenuates GnRH neuronal migration.

3A and 3B) Double-labeling GnRH cells for EMX2 revealed co-expression in ~94% of the GnRH neuronal population. 3C) Migration assays for GnRH cells in explants. Photomicrographs of an explant treated with a randomer (left panels) or antisense oligomer (right panels). Top panels show low power photomicrographs on which 100 μm concentric circles have been superimposed (White asterisk = tip of nasal cartilage). Boxed areas in top panels are shown at a higher magnification in bottom panels. GnRH cells are stained blue-black and olfactory axons are stained brown. 3D) A significant difference between randomer and antisense treated explants was found in the location of GnRH cells (X² p= <0.0001; X²=112.734). More GnRH cells in Emx2 knockdowns (hatched bars) remained closer to the explant compared to those treated with randomers (black bars). More GnRH cells in randomer treated groups migrated further from the explant compared to antisense treated (>900um). X² analysis creates an unbiased group (expected) based on treated and control groups (represented by Zero baseline in the graph). Plotted in 3D is the % difference (based on cell number in each zone) that the randomer vs antisense treated groups were relative to the expected group at three different migration distances: 1-400 μm, 500-800 μm and >900 μm from the main explant tissue.

(ii). In vivo model:

Staining for GnRH on Emx2^Cre/+; Rosa^tdT/+ showed EMX2 lineage expression in nGnRH cells throughout their migratory route (Figure 4A): in the OP, olfactory tracts, and at the NFJ. The number of GnRH cells co-expressing EMX2 was attenuated, with 40-50% lacking the Emx2 tdT reporter signal (Supplementary Figure 2). However, the microarray data showed a substantial decrease in expression of EMX2 in nGnRH cells at the oldeststage examined (3.09, logarithmic scale), consistent with a decrease in EMX2 expression in GnRH cells during development in vivo.

Figure 4. Total number of GnRH cells increases in Emx2 KO, but migration into the brain is attenuated.

4A) Schematic showing location of nGnRH cells as they migrate from the olfactory pit (OP) to the nasal forebrain junction (NFJ) before entering the forebrain (fb). Right panels) Immunofluorescent staining for GnRH (green) on tissue sections from Emx2^Cre/−; Rosa^tdT/+ and Emx2^Cre/+; Rosa^tdT/+ mice show double labeled GnRH cells (white arrows) in the OP (bottom), along olfactory axons crossing the nasal septum (middle), at the nasal forebrain junction (top). Bar in B = 20μm) B-F. nGnRH positive cells were counted in four regions (nasal pit, olfactory tracks, nasal forebrain junction, and brain). 4B) The total number of nGnRH cells was significantly greater in the Emx2 KO mice compared to WT littermates (p=0.0437; E12.5: WT=816±153, KO=1334±141; E14.5: WT= 1123±186, KO=1546±212; E16.5: WT=1249±31, KO=1641±476). 4C and 4D) At all ages, more nGnRH cells were counted in the nasal regions, both in the nasal pit (C, p=0.016) and along nasal tracks (D, p=0.0396). 4E and 4F) Between E14.5 and E16.5, nGnRH cells normally cross the nasal forebrain junction and migrate into the forebrain. KOs had more cells at the nasal-forebrain junction than WT (p= 0.0151). Notably, at all ages, there was no significant difference in the number of GnRH cells that reached the brain (p=0.4269), indicating that fewer nGnRH cells reached the forebrain in KO mice, remaining in nasal areas.

nGnRH cells increase in Emx2 KO mice, but fewer make it to the forebrain compared to WT:

In Emx2 KO mice and WT littermates, nGnRH cells were detected along the nasal migratory route from the OP to the NFJ and in the forebrain (Supplementary Figure 3). A few nGnRH fibers were located in the median eminence by E16.5 in KOs but less than that detected in WT. The total number of nGnRH cells was significantly greater in the Emx2 KO mice compared to WT littermates (Figure 4B, p=0.0173 N=9/genotype KO vs WT) at each age (p=0.0177, N=3/genotype/age). Distributional differences between KOs and WTs were also analyzed: the cell number in KOs was significantly greater in the OP (Figure 4C, p= 0.00134), tracks: (Figure 4D, p= 0.0396) and NFJ (Figure 4E, p=0.0055). Normally, nGnRH cells cross the NFJ and migrate into the forebrain between E14.5 and E16.5. Compared to WTs, KOs had more nGnRH cells at the NFJ at E16.5 than at earlier ages (WT = 405±57, KO = 682±78). However, no significant difference was detected between WT and KOs in brain (Figure 4F, p = 0.3471). Notably, At E16.5, 37% of the nGnRH cells resided in the forebrain in WT mice compared to 21% in Emx2 deficient mice, with most cells in KOs found more rostral than in WT. Thus, from E12.5-E14.5, the total number of nGnRH cells increased in nasal regions in Emx2 KO mice compared to WT mice while, in contrast, at E16.5, fewer cells than expected were found in the forebrain of Emx2 KO. This indicates that nGnRH cells failed to cross from the nasal regions into the forebrain in mice lacking Emx2, consistent with the large increase in GnRH cells at the NFJ detected in E16.5 KO (Figure 4E) and the attenuated GnRH neuronal migration detected in the ex vivo model. Staining for peripherin, an olfactory sensory axon marker, and GnRH indicated 1) that olfactory fibers grew to the developing olfactory bulb, which though present in these KO mice, was morphologically less layered than that detected in WT mice, and 2) that GnRH cells crossed into the forebrain, in both genotypes, at the base of the olfactory bulb (Supplementary Figure 4A & B). Examination of proliferating cells (pHis3 positive) in the OP at E12.5 revealed significantly more (2X) pHis3 cells in KO mice [(p=0.0488, KO=96±11 (N=4); WT=48±16 (N=3)] consistent with the increase in nGnRH cells counted in the OP of KOs at this stage (1.8X). In addition, a group of labeled ‘non-neuroendocrine’ GnRH cells was detected in the presumptive rhinencephalic subventricular zone in KOs at E14.5-E16.5 [(N=6, 83% (from ~80 at E14.5 to ~400 at E16.5; Supplementary Figure 4C-F)] that were not detected in the WT (N=6, 0%). These cells exhibited few processes/fibers compared to nGnRH cells.

Discussion

Utilizing an unbiased de novo approach in a large cohort of IHH trios, EMX2 was discovered as a causative gene for IHH and human infertility. Given that most gene-disrupting CNVs experience a strong negative selection, discovering rare and de novo CNVs that alter protein-coding sequences has a significant utility in human genetic research and highlights the pathogenicity of the genes affected by such events. We discovered that both rare and common EMX2 variants are linked to reproductive deficits, hearing loss, and neurodevelopmental phenotypes in diseased and hospital-based cohorts, respectively, highlighting the pleiotropic nature of this gene. In addition, our explant and mouse studies further lend credence for the role of EMX2 in GnRH biology. The level of EMX2 expression in primary nGnRH cells was comparable to that of other known transcription factors implicated in IHH. EMX2 was co-expressed in nGnRH cells and knockdown of Emx2 in an ex vivo model attenuated GnRH neuronal migration. Mice lacking Emx2 showed an increase in total GnRH cell number, but the majority of these cells failed to enter the forebrain. Together, these data are consistent with a failure of nGnRH neurons to migrate to their destination after perturbations of Emx2.

Prenatally, nGnRH cells migrate from the olfactory placode into the developing forebrain(33). Cell fate decision, differentiation, and migration of the neurogenic progenitors of both the olfactory and GnRH neurons are regulated by multiple transcription factors. EMX2 is a vertebrate homeobox transcription factor that is related to the empty spiracles (ems) Drosophila gene(34). In the developing mouse, its RNA expression has been described in the anterior CNS and the olfactory epithelium(35), with Emx2 KO animals displaying lack of olfactory projections(29). In mice, GnRH neurons are first detected in the OP between E10.5 and E11.5, differentiate in the developing VNO and continue differentiating as they migrate along olfactory axons to the NFJ before entering the brain(36). In Emx2 deficient mice we found that, although brain development is perturbed (17,19), olfactory axons do extend to the developing olfactory bulb, nGnRH cells migrate during prenatal development, and there is an increase in the number of nGnRH cells in nasal regions but not a corresponding increase in the forebrain. Our data indicate that two mechanisms are likely at action: 1) Absence of EMX2 increases the number of progenitors in the OP some of which give rise to GnRH cells thus increasing GnRH cell number and 2) EMX2 is normally expressed at higher levels earlier in development (higher expression levels on microarray) and would maintain low levels of GnRH expression. As such, KOs experience an increase in GnRH expression earlier than normal which leads to further differentiation and decreases migration. When the GnRH cells reach the NFJ (12.5-E14.5), the olfactory bulb is not developing normally, the olfactory sensory axons respond to this deficit, expression of GnRH is augmented, and these changes cause the GnRH cells to remain stuck at the NFJ at E16.5. The in vivo KO data are supported by knockdown of Emx2 in explant models where GnRH neurons exhibited reduced migration. The explant model does not exhibit a similar increase in GnRH cell number as found in vivo, but this is expected since knockdown of Emx2 occurred after GnRH cells had differentiated.

Our data indicate that Emx2 negatively regulates GnRH expression. This is consistent with the previously described role of Emx2 in the telencephalon(37), where it interferes with binding of positive Sox2 regulators (SOX2 HGNC ID11195). Emx2 deficiency leads to an increase of Sox2 cells, partially counteracting hippocampal neurogenesis defects observed in Sox2 deficient mice. Normally, GnRH transcription changes as the nGnRH cells mature during their migratory journey. The nGnRH cells pause at the NFJ, and as they enter the forebrain, there is a significant increase in GnRH transcription with concomitant changes in protein expression and neuronal activity. Prior studies in mouse showed that MSX, a non-Hox homeodomain transcription factor, represses GnRH expression(38). Msx mutant mice had more nGnRH expressing cells at E13.5 and that most cells were confined to nasal regions being distributed in both expected regions and ectopically in the olfactory epithelium(38). Similarly, the robust staining detected in Emx2 deficient mice compared to WT, is consistent with increased GnRH expression. Emx2 deficiency also resulted in an ectopic group of ‘non-neuroendocrine’ GnRH cells adjacent to the rostral region of the lateral ventricle, a region associated with developing olfactory structures(39), supporting the notion that Emx2 acts as a repressor of GnRH expression. As such, variants releasing Emx2 repression increase GnRH expression which, as observed in MSX(38) (HGNC IDs 7391 and 7392) and POU6F2(40) (HGNC ID 21694), likely results in early cell cessation. Thus, augmentation of GnRH expression early during development can disrupt normal nGnRH cell migration, consistent with EMX2 variants contributing to IHH pathogenesis.

While EMX2 was suspected to be involved in the pathogenesis of IHH before(28), in this study we report human genetic variants in IHH participants. EMX2 variants have been previously linked to other reproductive diseases, including Mullerian duct defects in females(41) and endometrioid adenocarcinomas(42), suggesting a significant role for EMX2 in the elongation process of Mullerian ducts. This is in line with Emx2 KO mice exhibiting infertility and genital tract defects(43) and the urogenital defects shared by probands harboring deleterious EMX2 variants in this study. Although prior reports also suggested a link of EMX2 to schizencephaly (OMIM ID 269160)(44), the role of EMX2 in neurodevelopmental phenotypes remains unclear. Within the IHH participants carrying EMX2 variants, 2 recurrent comorbidities were noted: developmental delay and hearing loss. Even though, the syndromic presentations could be attributed to the multigenic nature of the identified contiguous gene deletions, the identification of the same clinical features in the participants with the de novo deleterious missense variants in the DNA binding domain of the protein highlights a pleiotropic role of EMX2. Of note, our phenotypic analysis in the MGBB demonstrated a link between common EMX2 variants and Parkinson’s disease, in line with prior analyses linking EMX2 with a Parkinson associated gene, SNCA (HGNC ID 11138)(45). Furthermore, a significant association between both rare and common EMX2 variants with hearing impairment was detected. Studies in both zebrafish and mouse models show that Emx2 plays a key role in hair bundle polarity of the middle and inner ear(46). Our study provides human evidence of the critical role of EMX2 in auditory function. Further studies are required to define if EMX2 human variants cause non-syndromic forms of hearing loss.

This study has several strengths. We utilized an unbiased approach for novel gene discovery in one of the largest cohorts of IHH trios. We identified 6 pathogenic and likely pathogenic EMX2 variants. Given their heterozygous nature, we pursued in vitro studies showing that knockdown of Emx2 attenuated GnRH neuronal migration, consistent with the IHH phenotype. Their de novo and novel nature and the absence of additional causal IHH variants in probands harboring these variants suggests that these variants are causal for IHH and associated non-reproductive phenotypes, including hearing loss and neurodevelopmental phenotypes. Hence, our data will be of great benefit for genetic counseling to both screen for and adjudicate EMX2 variants in IHH pedigrees.

In conclusion, this study demonstrates the power of de novo trio-based approaches for Mendelian gene discovery and validates a novel role for EMX2 in human reproduction.

Supplementary Material

Supplementary Table 1

Supplementary Table 2

Supplementary Table 3

Supplementary Table 4

Supplementary Table 5

Supplementary Table 6

Supplementary Table 7

Supplementary Table 8

Supplementary Figure 1

Supplementary Figure 2

Supplementary Figure 3

Supplementary Figure 4

Data availability:

Data and materials will be made available by the authors individually upon request subject to the data sharing plan and consent provided by the study participants.