Discovering Genes Essential to the Hypothalamic Regulation of Human Reproduction Using a Human Disease Model: Adjusting to Life in the “-Omics” Era

Abstract

The neuroendocrine regulation of reproduction is an intricate process requiring the exquisite coordination of an assortment of cellular networks, all converging on the GnRH neurons. These neurons have a complex life history, migrating mainly from the olfactory placode into the hypothalamus, where GnRH is secreted and acts as the master regulator of the hypothalamic-pituitary-gonadal axis. Much of what we know about the biology of the GnRH neurons has been aided by discoveries made using the human disease model of isolated GnRH deficiency (IGD), a family of rare Mendelian disorders that share a common failure of secretion and/or action of GnRH causing hypogonadotropic hypogonadism. Over the last 30 years, research groups around the world have been investigating the genetic basis of IGD using different strategies based on complex cases that harbor structural abnormalities or single pleiotropic genes, endogamous pedigrees, candidate gene approaches as well as pathway gene analyses. Although such traditional approaches, based on well-validated tools, have been critical to establish the field, new strategies, such as next-generation sequencing, are now providing speed and robustness, but also revealing a surprising number of variants in known IGD genes in both patients and healthy controls. Thus, before the field moves forward with new genetic tools and continues discovery efforts, we must reassess what we know about IGD genetics and prepare to hold our work to a different standard. The purpose of this review is to: 1) look back at the strategies used to discover the “known” genes implicated in the rare forms of IGD; 2) examine the strengths and weaknesses of the methodologies used to validate genetic variation; 3) substantiate the role of known genes in the pathophysiology of the disease; and 4) project forward as we embark upon a widening use of these new and powerful technologies for gene discovery.

1. Introduction

2. The Past: Previous Strategies Used to Discover Genes in IGD

   1. Complex phenotypes, contiguous gene deletion syndromes, and karyotypic abnormalities point to IGD genes

   2. Phenotypic complexity identifies pleiotropic genes in syndromic cases

   3. Endogamous pedigrees lead to the discovery of recessive IGD genes

   4. Candidate gene approaches reveal novel genetic causes of IGD

   5. Summary

3. The Present

   1. Identifying IGD genes in the “-omics” era

   2. Limitations of current methodology

   3. Establishing criteria for novel gene validation

   4. A critical reappraisal of known IGD genes

   5. The complex biology of the GnRH genetic network

   6. The challenge of functional testing

   7. Testing of genetic burden in IGD

4. The Future

   1. New tools and approaches for the discovery of IGD genes

5. Conclusions

I. Introduction

The hypothalamic regulation of reproduction is an intricate developmental process requiring exquisite coordination of an assortment of cellular networks. During embryonic development, precursors of GnRH neurons originate from the neuroectoderm and neural crest, populate the olfactory placode, then rapidly expand their population and migrate along the vomeronasal and olfactory nerves to their final anatomic home within the medio-basal hypothalamus where they fully differentiate (1–6). Within the hypothalamus, these neurons integrate various physiologic signals and coordinately secrete GnRH in a distinctive pulsatile pattern into the hypophyseal-portal blood supply (7). This release, in turn, initiates and maintains the biosynthesis and release of LH and FSH from the anterior pituitary gonadotropes.

Figure 1 shows a timeline of scientific findings that have contributed to our understanding of this complex GnRH neuronal biology. These contributions were initiated and substantiated in animal models and, once the human genome was sequenced, continued to expand via genetic discoveries using the human disease model of isolated GnRH deficiency (IGD) (Online Mendelian Inheritance in Man phenotypic series: PS147950) (8–13). IGD is a family of rare Mendelian disorders (14) that have also been called isolated or idiopathic/congenital hypogonadotropic hypogonadism. The term congenital, which is often used, might well apply where evidence of the disorder exists at birth such as microphallus, cryptorchidism or hypogonadotropism. However, in the vast majority of cases, such definitive evidence is typically lacking making the use of this term somewhat inaccurate. Thus, the term IGD is preferable to these other terms because this disease occurs in the absence of any anatomical hypothalamic or pituitary lesion or other anterior pituitary hormone deficiency (ie, is isolated) and is caused by a failure of secretion and/or action of GnRH. The phenotypic result of this GnRH deficiency is hypogonadotropic hypogonadism and “downstream failure” of sexual maturation that typically manifests itself with a lack of pubertal development (8, 12, 15).

Figure 1.

Major contributions to our understanding of the ontogeny of GnRH neurons: from prehuman genome to human genetics and NGS.

The intimate topographical and developmental intertwining of the olfactory system and GnRH neurons explains the etiology of the “neurodevelopmental” form of IGD known as Kallmann syndrome (KS), which is also characterized by complete or partial loss of the sense of smell (anosmia/hyposmia) (16, 17) that accompanies the hypogonadotropism. However, despite the common feature of a reduction in (or absence of) GnRH function, IGD is clinically heterogeneous encompassing not only KS, but also normosmic idiopathic hypogonadotropic hypogonadism (nIHH) (18), the adult onset form of acquired hypogonadotropic hypogonadism (19), the more common transient loss of reproductive function seen in hypothalamic amenorrhea, and delayed gain of reproductive function seen in constitutional delay of puberty (20–22). Although these reproductive phenotypes typically dominate an IGD patient's clinical presentation, in addition to anosmia there is also a wide variety of associated nonreproductive phenotypes, including unilateral renal agenesis and other kidney malformations, cranio-facial defects (eg, cleft lip/palate), skeletal and digital defects, sensorineural hearing loss, eye and oculomotor defects, cardiac defects, and dental agenesis, among others (13, 23, 24).

Mirroring this phenotypic diversity, the underlying genetic architecture of IGD is equally heterogeneous with mutations in several genes causing the developmental loss of GnRH neurons and/or their abnormal neuroendocrine secretory function. Since the initial description of the anosmic form of this disease by Franz Kallmann in 1944, the genetic complexity underlying this disorder has puzzled physicians caring for IGD patients. In 1944, Kallmann observed that, “there is no reason to assume that it must always be the same mechanism …. or the same gene that causes the disease … ” (17). As the catalogue of candidate genes for IGD continues to expand, this prescience of Franz Kallmann is confirmed. The genetic complexity of IGD is also reflected in its different inheritance patterns. Whereas most IGD patients initially present as isolated or sporadic cases, closer reviews of their pedigrees often reveal families with X-linked, autosomal recessive, and autosomal dominant inheritance (12, 15, 23, 25). In addition to these Mendelian modes of inheritance, an even more complex genetic architecture for IGD (occurring in 10%–15% of IGD cases) has been documented wherein mutations in 2 or more IGD genes are found in a single case. These coexisting mutations could therefore play the role of “second hits” or act as “modifier” genes involved in this syndrome, perhaps accounting for some of the variable expressivity and incomplete penetrance often observed in IGD (26–30).

This clinical and genetic heterogeneity characterizing IGD has led to some of the most important contributions in the field of genetic discovery in reproductive endocrinology since the description of Turner and Klinefelter's syndromes (31, 32). Over the last 30 years, several research groups around the world have used a combination of clinical investigational strategies and genetic approaches to unravel the growing genetic complexity underlying IGD and link its phenotypic variation to specific genetic and biological causes (Figures 1 and 2). These approaches have been based on well-validated genetic and molecular tools and been critical to establishing this field. However, they have often lacked 2 major characteristics: speed and robustness of scale. New strategies, such as next-generation sequencing (NGS), are now empowering a global revolution in genetic discovery.

Figure 2.

Different strategies for gene discovery in IGD.

The ability of NGS to query the entire genome with increasing speed and accuracy is surfacing a surprising number of rare sequence variants (RSVs) of unknown significance in both known and novel IGD genes. Perplexingly, these RSVs are sometimes shared by both affected individuals and controls. Moreover, because many of the known IGD genes seem to act synergistically while individually displaying features of incomplete penetrance or variable expressivity, it is becoming ever more difficult to label a single genetic change in any individual IGD patient as “causative.” Therefore, before this field moves forward with new genetic tools and itsever-expanding genetic framework, it is important to pause and reassess what we know about the GnRH network and IGD genetics; how has this gene discovery evolved, and what does it mean for genes implicated in IGD? Therefore, this review will: 1) look back at the strategies used to discover the “known” IGD genes implicated in the rare forms of KS and nIHH; 2) examine the strengths and weaknesses of the methodologies used to validate their genetic variation; 3) substantiate the role of these now known genes in the pathophysiology of the disease according to the evolving standards in the field; and 4) project forward as we embark upon a widening use of these new and powerful technologies for gene discovery.

II. The Past: Previous Strategies Used to Discover Genes in IGD

A. Complex phenotypes, contiguous gene deletion syndromes, and karyotypic abnormalities point to IGD genes

The first gene associated with IGD, Kallmann 1 (KAL1), was surfaced during the investigation of brothers with complex clinical phenotypes suggestive of a large contiguous gene deletion syndrome on the X chromosome (33–35). In such syndromes, the multiorgan system defects seen in these patients represented a combination of phenotypes associated with each of the individual genes deleted within the contiguous region. As a case in point, KAL1 was identified in a male infant who displayed abnormal genitalia, hypogonadotropic hypogonadism, agenesis of the olfactory bulbs and tracts (ie, KS) associated with chondrodysplasia punctata and ichthyosis (34). Karyotype analysis and positional cloning revealed a large deletion in the Xp22.31 region that included the genes KAL1, ARSE, and STS (that together caused KS, chondrodysplasia punctata, and ichthyosis, respectively) and thus established KAL1 as an X-linked KS gene (34–38). Several years later, overlapping interstitial deletions in chromosome 8p11 were discovered in patients suffering from KS and hereditary spherocytosis and were mapped by fluorescence in situ hybridization and bacterial artificial chromosome cloning (39). Another contiguous gene deletion syndrome was verified and fibroblast growth factor receptor 1 (FGFR1) surfaced as an autosomal dominant KS gene (40).

In addition to such contiguous gene syndromes, chromosomal rearrangements have also provided powerful insights into the genetic underpinnings of IGD. Just a few years after the initial discovery of FGFR1, another analysis of a balanced chromosomal translocation (using fluorescence in situ hybridization and bacterial artificial chromosome cloning) in a patient with nIHH and cleft lip/palate led to the identification of FGFR1 as the genetic cause of that patient's phenotype, and established that FGFR1 mutations could also cause the normosmic form of IGD, nIHH (41, 42). Additional IGD cases harboring structural variations have uncovered novel genes such as the WD repeat domain 11 (WDR11) gene, which was found in a KS patient near a breakpoint of a balanced translocation between chromosomes 10 and 12 (43). Similarly, the semaphorin 3A (SEMA3A) gene was truncated in 2 seconds KS siblings harboring a large heterozygous deletion of the gene that removed 7 out of 11 of its exons and was inherited from their affected father (44). Subsequently, another group found pathogenic point mutations in SEMA3A in a cohort of KS patients (45). Importantly, the SEMA3A deletion was missed by traditional karyotyping whereas an oligonucleotide array detected it thus highlighting the importance of the constantly improving resolution of genetic technologies and their ability to contribute new insights.

Each of these genetic findings illustrate how both low-resolution karyotypes and high-resolution gene arrays have played key roles in the detection of causative candidate genes in IGD patients with complex phenotypes, implying the presence of an underlying structural event. In addition, as will be discussed, these discoveries have provided fundamental information about the underlying developmental biology of GnRH neurons. Thus, contiguous gene syndromes and complex chromosomal rearrangements represent a rich source of new gene discovery both in the field of human genetics generally, and, specifically, in the identification of genes controlling human reproduction.

B. Phenotypic complexity identifies pleiotropic genes in syndromic cases

It is important to note that the phenotypic richness found in “syndromic” cases does not always point to a contiguous gene interruption. In fact, Sanger sequencing and NGS in patients with equally complex phenotypes have yielded another large crop of novel IGD genes. These patients underscore another genetic feature of importance: the role of the pleiotropy. This phenomenon occurs when a single genetic cause results in the development of both a reproductive defect as well as additional clinical features, ie, when IGD occurs alone or accompanied by these syndromic associations.

The nuclear receptor subfamily 0, group B, member 1 (NR0B1) gene is an early example of a pleiotropic IGD gene. The discovery of this gene explained the paradox of the well-established association between the X-linked form of congenital adrenal hypoplasia and hypogonadotropic hypogonadism via its previously undescribed role in the development of the hypothalamic-pituitary axis (46, 47). Similarly, although IGD is only occasionally associated with morbid obesity, genetic defects in leptin (LEP) (48) and the leptin receptor (LEPR) (49–51) genes, as well as mutations in the proprotein convertase subtilisin/kexin type 1 (PCSK1) gene (52), have been found in IGD patients with an obesity phenotype.

There are additional pleiotropic genes that cause IGD associated with neurological phenotypes. For example, whole-exome sequencing analysis in patients with Gordon Holmes syndrome (IGD and cerebellar ataxia) led to the discovery of ovarian tumor deubiquitinase 4 gene (OTUD4) and ring finger protein 216 (RNF216) (53), as well as mutations in stress-induced phosphoprotein 1 homology and U-box containing protein 1, E3 ubiquitin protein ligase (STUB1) and patatin-like phospholipase domain-containing protein 6 (PNPLA6) (54) that were found in IGD patients who also displayed ataxia (55). There are also several other genes causing 4H syndrome (hypomyelination, hypodontia and hypogonadotropic hypogonadism), including polymerase III, RNA subunit A and B (POLR3A and POLR3B) (56, 57) Additionally, a novel clinical syndrome defined by a specific genetic mutation in the neuronal tubulin-β class 3 (TUBB3) gene was recently described having the characteristics of Moebius syndrome (facial hypoplasia/paralysis and oculomotor defects) accompanied by KS, intellectual disabilities, and in some cases, vocal cord paralysis, tracheomalacia, and cyclical vomiting (58–60). In the future, it is likely that such pleiotropic genes will be associated with the nonsyndromic form of IGD as well.

So far, 2 pleiotropic IGD genes have been found in patients with IGD alone. Coloboma of the eye, heart defects, choanal atresia, growth retardation, genital anomalies, and ear defects (CHARGE) syndrome is a disorder with multiple affected organ systems. Hypogonadotropic hypogonadism is typically a minor feature of most CHARGE cases, but this reproductive feature can be entirely absent or missed, especially if these cases are reported in childhood before the pubertal defects become manifest (61). Conversely, upon closer phenotypic examination, many IGD cases with CHARGE features (for example, hearing loss) harbor mutations in the CHARGE syndrome gene, a chromodomain helicase DNA binding protein 7 (CHD7) (61, 62). We now know that patients with KS and hearing loss are more likely to present with mutations in the CHD7 gene than any other IGD gene (63), a phenotypic association extremely useful for the prioritization of the genetic screening (24). Interestingly, rare deleterious mutations in CHD7 can also be associated with IGD in the absence of any of the other phenotypic hallmarks of CHARGE (63, 64). This finding indicates a potential differential sensitivity of developing GnRH neurons to milder deleterious mutations in CHD7. Following the same lesson of pleiotropy as seen in CHD7, mutations in the sex-determining region of Y-box 10 (SOX10) gene, previously associated with Waardenburg syndrome, were recently identified in patients with IGD alone as well as IGD patients with milder Waardenburg features, including hearing loss, pigmentation defects, intellectual disability, or psychomotor delay (65). These examples show that cases of many syndromic IGD genes may result from mutations in single genes that have pleiotropic roles with differential effect(s) on various organ systems resulting in a wide range of syndromic and nonsyndromic phenotypes. This phenomenon may be particularly true of other genes which, like CHD7 and SOX10, regulate the expression of many downstream genetic targets involved in GnRH maturation, migration, and hypothalamic function (66–68).

Taken together, these findings in patients with complex phenotypic syndromes that include hypogonadotropic hypogonadism suggest the existence of a panoply of genes whose genetic pleiotropy is a clue to their contribution to the development or regulation of GnRH neurons. In the future, there will likely be much to gain with the pursuit of other similar genetic and biological leads in piecing together the complex biology of GnRH neuronal ontogeny. In addition, syndromic cases in other nonreproductive fields may well yield a similar crop of new genes involved in isolated disease phenotypes.

C. Endogamous pedigrees lead to the discovery of recessive IGD genes

The study of IGD in multiplex families from geographically delineated populations with high rates of endogamy has been another highly fruitful method of IGD gene discovery (69). The gene coding for kisspeptin receptor 1 (KISS1R) was discovered using linkage analysis in an endogamous pedigrees with multiple affected Bedouin family members displaying nIHH (70, 71). Similarly, the discovery of tachykinin 3 (TAC3) and its receptor (TACR3), was based on haplotype analysis of endogamous nIHH Turkish and Kurdish pedigrees (72). Homozygosity mapping (73), candidate gene screening (74), and whole-exome sequencing prompted the discovery of kisspeptin (KISS1) mutations in a consanguineous nIHH family (73), whereas mutations in the fasciculation and elongation protein zeta family zinc finger 1 (FEZF1) were newly identified in consanguineous KS families of Kurdish origin (75). Most recently, linkage analysis revealed a homozygous deletion in Drosophila melanogaster X-gene like-2 (DMXL2) gene carried by 3 brothers of a Senegalese consanguineous pedigree affected with a syndromic form of IGD that includes hypothyroidism, diabetes mellitus, and various neurological phenotypes (76). These examples suggest that analysis of other endogamous pedigrees will allow researchers to uncover novel recessive genes, giving additional insights into IGD disease pathology.

D. Candidate gene approaches reveal novel genetic causes of IGD

During the more than 25 years of successful genetic studies on IGD, there have been many attempts to screen candidate genes in patients suffering from this disease. Most were based either on data from animal studies or our previously limited knowledge of the biology of GnRH neurons. Examples of excellent candidate genes surfaced based on mouse model phenotypes include the GnRH receptor (GNRHR) (77) and its ligand, GnRH (GNRH1) (78), N-Methyl-D-aspartate receptor synaptonuclear signaling and neuronal migration factor (NSMF) (79, 80), and prokineticin 2 (PROK2) and its receptor (PROKR2) (81, 82). Previous findings of FGFR1 mutations in IGD patients led to the discovery of mutations in the candidate gene FGF8 (83), a ligand of FGFR1 in the central nervous system, whereas interactions of anosmin with heparin sulfate proteoglycans led to the validation of heparan sulfate 6-O-sulfotransferase 1 (HS6ST1) as an IGD gene (84). More recently, analysis of differential gene expression in 2 different SV40 transformed mouse GnRH cell lines has led to the discovery of mutations in AXL receptor tyrosine kinase (AXL) (85, 86).

However, many other candidate gene efforts have not been similarly successful (10), due to various limiting factors. First, small numbers of patients are generally screened meaning rare mutations are often missed. Second, incomplete penetrance of the candidate genes being investigated can make segregation of variants within and across pedigrees challenging. Finally, we still have a limited knowledge about the full systems biology involved in the disease, including the roles (and redundancy) of other protein family members and/or potential modifier genes/mutations that contribute to the defective phenotypes of loss of function (LoF) in man and mice (ie, the effect of background and strain genetics) (26).

Bioinformatic tools are rapidly expanding our knowledge of the underlying biology of GnRH neurons that was previously limited to in vitro studies using mouse GnRH cell lines (87, 88), in vivo and in vitro slice preparations (89, 90), and physiological studies in knockout models (91–93). These types of bioinformatically informed approaches have made substantial progress in revealing the role of protein interactions and coordinate signaling pathways in GnRH neuronal development and are thus leading to new and better informed candidate gene discoveries. For example, a combination of proteomic gene-set enrichment analysis and more selective candidate screening has led to the identification of several additional FGF signaling pathway genes causing IGD, including an additional ligand encoded by FGF17, IL-17 receptor D (IL17RD), dual specificity phosphatase 6 (DUSP6), sprouty homolog 4 (SPRY4), and fibronectin leucine-rich transmembrane protein 3 (FLRT3) (29). Recently, a combination of whole-exome sequencing with bioinformatics, in vitro functional studies, and studies of genetically modified mice led to the description of SEMA3E as the second semaphorin family member to be implicated in KS (94).

E. Summary

Figure 2 provides a summary of the various strategies used to identify IGD genes. Past approaches have mounted evidence from phenotypic presentations, gross chromosomal and structural abnormalities, and the underlying GnRH biology to support genetic discoveries. As will be discussed in the next section, NGS approaches will require researchers to use similar approaches while also adding further stringency to prove causality of genetic candidates.

III. The Present

A. Identifying IGD genes in the “-omics” era

Together, all of the aforementioned approaches have yielded a genetic cause in approximately 30%–35% of IGD patients (95), whereas the genetic etiology in the remaining 2/3 of patients is currently unknown. Genetic defects may have been missed due to the limited depth and range of existing DNA sequencing that, to date, has only screened coding sequences or short genome segments with low resolution techniques. Thus, mutations may still exist in noncoding areas of the “known IGD genes” or there may be unappreciated cryptic, structural defects in the genome, or epigenetic factors contributing to the disease. However, recent advances in NGS and its derivative analytic technologies (including genomics, epigenomics, proteomics, and metabolomics) have renewed efforts to unravel the genetic mystery of these “genetically unsolved” IGD cases. One key difference with these contemporary approaches is that, whereas most of the past gene discovery in IGD has been based upon limited Sanger sequencing of the genome due to cost restraints, NGS discovery tools have taken advantage of lowered costs and rapid turnaround times to base all their genetic inquiries in IGD patients on a genome-wide scale. These enhanced efforts allow for a more complete picture of the genetics underlying the disease in each patient to emerge. As high-throughput sequencing strategies become increasingly mainstream, the number of sequenced IGD patients and controls have increased, resulting in the generation of a burgeoning number of RSVs in “candidate IGD genes.” With this ever-growing list of unbiased candidates emerging from NGS, there is a pressing need for standardization of filtering based on sound statistical, biological, and genetic criteria to prioritize them and distinguish between the true genetic causes and false-positive associations.

B. Limitations of current methodology

Because IGD has been studied as a Mendelian disease, the segregation of variants within pedigrees and assessment of their penetrance have been key initial steps in the validation of putatively causative mutations. In addition to segregation, another important validation step is the replication of variants in other IGD subjects. This practice, assumes that: 1) sufficiently large cohorts of IGD patients are available for testing, and 2) large numbers of control individuals lack an association between the mutation and phenotype. However, these validation measures have been further complicated by incomplete penetrance of mutations and the fact that LoF in certain IGD candidate genes have been documented to occur in otherwise normal populations (96).

Consequently, once a rare variant is found in an affected IGD patient, variant-level experimental evidence is also required to address the functional impact of specific mutations on the level of gene/protein expression, splicing, and/or biochemical function. To address the question of how a candidate variant influences the gene function or the expression of the phenotype, several factors should be taken into consideration. First, the genetic nature of each identified variant must be assessed at the experimental level. The detection of severe, obviously “high impact” variation (intragenic deletions, frameshift deletions and insertions, stop-gain, and essential splice site mutations) may sufficiently explain the patient's phenotype without further need for experimental validation, because such mutations are a priori expected to cause LoF, particularly when occurring in the homozygous state. Conversely, although mutations in noncoding sequence and synonymous mutations are expected to be benign, this assumption has yet to be rigorously tested.

However, missense variation in a gene generally requires further experimental validation by examining its effect on mRNA expression and/or protein function. When assessing either LoF or gain of function in new candidate IGD genes, it is crucial to employ the appropriate experimental method for the functional validation of each candidate. The nature of the experimental system used to validate changes in function will vary according to the putative biology of the candidate under examination. For example, G protein-coupled receptors (97), transcription factors, and growth factors, etc, occupy special biologic niches that must be reflected in the careful choice of appropriate functional assay system selected for their validation. A corollary of this issue is that studying the gene and/or protein expression and its molecular activity in cell lines directly derived from patients may be more informative than performing such experiments in artificially constructed experimental cell lines. On the other hand, IGD is a complex disorder that ultimately affects a very small population of highly specialized neurons. Thus, simple cultured cell models may not be appropriate, whereas obtaining appropriate cell-lines, mammalian model systems, and/or human neuronal tissues can be challenging and expensive.

As an alternative to experimental validation, many researchers make use of bioinformatics tools that attempt to predict the functional effect of missense mutations (98). These programs make functional predictions based upon: 1) the biochemical role of the domain and specific amino acids within its protein; 2) the evolutionary conservation of the particular amino acid sequence across species; 3) the impact of the mutations in the different protein transcripts; and 4) the biochemical differences between the reference sequence and the proposed amino-acid substitution. One additional method that can be used to assess the causative nature of mutations is to examine the amount of observed functional variation in a gene as compared with what is expected based on evolutionary constraints (99). However, this type of analysis requires sophisticated statistical genetics.

Although the aforementioned recommendations are for assessing the variant-level implications of mutations, the remainder of this review will focus on analysis of the gene-level validation of the known IGD genes, trying to set the stage for the novel genes arising from NGS studies in IGD as well as in other rare Mendelian diseases and disorders. Lists of specific variants found in each of the IGD genes can be found in publically available databases, such as ClinVar (http://www.ncbi.nlm.nih.gov/clinvar/) along with supporting information on their pathogenicity. It is important to note that databases such as these are constantly changing and updating as new data are provided by researchers. Thus, at this time, there are relatively few RSVs that are “bona fide” pathogenic mutations (ie, have experimental validation of LoF), even in known IGD genes.

C. Establishing criteria for novel gene validation

NGS tools not only provide the opportunity for new gene discovery on a genome-wide basis but also create important statistical challenges in their interpretation and validation. The need for standards of validation for the growing number of candidate genes surfaced by NGS was thoroughly addressed in a recent publication by MacArthur et al (100). The establishment of such standards is critical, because false assignment of pathogenicity could have implications in genetic counseling and medical decision making (100).

Consequently, before the field fully embraces these novel technologies for new gene discovery, it seems reasonable to pause and examine the genes already reported in the literature that have already been associated with IGD. Our analysis will focus on 25 genes that have been found to be disrupted in human cases with pure IGD and thus are generally thought to be causative (see Table 1), even though, as previously mentioned, other genes derived from syndromic cases are also implicated in GnRH biology. We will apply the 7 criteria proposed by MacArthur et al (100) as a retrospective validation of these causative IGD genes. These criteria include:

• Criterion 1: proof that the gene and/or gene product function is demonstrably altered in patients carrying these candidate mutations;

• Criterion 2: evidence that there is tissue-specific expression of the protein that the gene encodes;

• Criterion 3: demonstration that the biochemical function of the gene product is shared with other known genes in the disease or consistent with the phenotype;

• Criterion 4: confirmation that the gene product interacts with other proteins already implicated in the disease;

• Criterion 5: generation of nonhuman animal or cell-culture models with a similarly disrupted copy of the affected gene shows a phenotype consistent with human disease state;

• Criterion 6: proof that the cellular phenotype in patient-derived cells or engineered equivalents can be rescued by addition of the wild-type gene product;

• Criterion 7: statistical evidence that there is an increased genetic burden of variation in the gene in disease vs normal populations.

Table 1.

Assessment of the 25 IGD Genes Based on the MacArthur Criteria

Gene	Criterion 1: Gene Disruption	Criterion 2: Expression	Criterion 3: Biochemical Function	Criterion 4: Protein Interactions	Criterion 5: Model Systems	Criterion 6: Rescue	Criterion 7: Gene Burden	Number of Criteria Met
KAL1	+	+	+	The first gene discovered	+	+		6
GNRHR	+	+	+	+	+	+		6
GNRH1	+	+	+	+	+	+		6
KISS1	+	+	+	+	+	+		6
FGFR1	+	+	+	+	+			5
FGF8	+	+	+	+	+			5
KISS1R	+	+	+	+	+			5
TAC3	+	+	+	+	+			5
TACR3	+	+	+	+	+			5
PROKR2	+	+	+		+		+	5
PROK2	+	+	+		+		+	5
SEMA3A	+	+	+	+	+			5
SOX10	+	+	+	+	+			5
AXL	+	+	+	+	+			5
SEMA3E	+	+	+	+	+			5
NSMF (NELF)	+	+	+		+			4
WDR11	+	+		+	+			4
CHD7	+	+	+		+			4
FGF17	+	+	+	+				4
IL17RD	+	+	+	+				4
HS6ST1	+	+		+	+			4
FEZF1	+	+	+		+			4
SPRY4	+	+		+				3
FLRT3	+	+		+				3
DUSP6	+	+		+				3

The 25 IGD genes were assessed for the next criteria. Gene disruption: the gene and/or gene product function is demonstrably altered in patients carrying candidate mutations. Expression: the gene is expressed in tissues relevant to the disease of interest and/or is altered in expression in patients who have the disease. Biochemical function: the gene product performs a biochemical function shared with other known genes in the disease of interest, or consistent with the phenotype. Protein interactions: the gene product interacts with proteins previously implicated (genetically or biochemically) in the disease of interest. Model systems: nonhuman animal or cell-culture models with a similarly disrupted copy of the affected gene show a phenotype consistent with human disease state. Rescue: the cellular phenotype in patient-derived cells or engineered equivalents can be rescued by addition of the wild-type gene product. Gene burden: the affected gene shows statistical excess of rare (or de novo) probably damaging variants segregating in cases compared with control cohorts or null models.

D. A critical reappraisal of known IGD genes

Importantly, as shown in Table 1, all of the “established” IGD genes meet criteria 1 and 2 (gene disruption and expression) reflecting the careful research by different scientific groups and the dedicated efforts each has put into validation of their newly discovered genes. Most of these genes also pass at least one of the criteria assessing the biological impact of the gene(s) to the disease, ie, criteria 3, 4, and/or 5 (biochemical function, protein interactions, and model systems), because after establishing a gene disruption in human subjects, considerable efforts were made to assess these genes for an association with GnRH biology. Table 2 shows publically available gene expression data in various human tissue (101–107). Although each of the public databases have their limitations, according to existing data and current methodology, all of the genes do pass criterion 2 even when held to a standard of relevant expression in human tissues.

Table 2.

IGD Gene Expression in Human Tissues

	Tissue Expression Database
	TiGER	BioGPS	Human Protein Atlas (n = 3, 1 Child, 2 Adults)	Riken FANTOM5 (Adult n = 1)	Riken FANTOM5 (Fetal n = 1)	GTEx (n = 22–30 Adults)
Total # Tissue Types	30	78	45	56	20	46
Relevant tissue samples	Whole brain	Whole brain fetal brain olfactory bulb prefrontal cortex pituitary hypothalamus	Cerebral cortex lateral ventricle	Diencephalons olfactory apparatus pituitary	Occipital lobe parietal lobe temporal lobe	Frontal cortex hypothalamus pituitary
Gene		Expression pattern
KAL1	+	All	Cerebral cortex (RNA)	All	All	Frontal cortex, hypothalamus
GNRHR	+	All	−	Pituitary	−	Pituitary
GNRH1	+	All	All	Diencephalon (low)	Parietal (low)	All
KISS1	−	−	−	−	Temporal (low)	−
FGFR1	+	All	All	All low	All low	All
FGF8	Not found	All	−	−	−	Frontal cortex, hypothalamus
KISS1R	Not found	All	−	Pituitary (low)	−	Hypothalamus, pituitary
TAC3	+	−	All	All (pituitary low)	All	Frontal cortex, hypothalamus
TACR3	Not found	All	All	Olfactory apparatus (low)	All low	Hypothalamus
PROKR2	Not found	−	All	−	−	All
PROK2	Not found	−	Lateral ventricle (protein)	Diencephalon (low)	Temporal (low)	−
SEMA3A	+	All	All	Pituitary (low)	−	All
SOX10	+	Whole brain, olfactory bulb, prefrontal cortex, hypothalamus	All	All low	All	Frontal cortex, hypothalamus
AXL	+	All	Cerebral cortex (RNA and protein)	All low	Occipital (low), Temporal	All (low)
SEMA3E	+	All	All	All low	All low	Frontal cortex, hypothalamus
NSMF (NELF)	Not found	Whole brain, prefrontal cortex	All	All (pituitary low)	All	All
WDR11	Not found	Prefrontal cortex, hypothalamus	All	All low	Occipital, parietal, temporal (low)	All
CHD7	+	All	All	All low	All	All
FGF17	Not found	All	Cerebral cortex (RNA)	−	−	All
IL17RD	+	Prefrontal cortex, hypothalamus, olfactory bulb	All	All	All	All
HS6ST1	+	All	Cerebral cortex (RNA)	−	−	All
FEZF1	Not found	−	−	−	−	Hypothalamus
SPRY4	+	All	All	−	Occipital (low), temporal (low)	All
FLRT3	+	−	Cerebral cortex (RNA and protein)	All	All	Frontal cortex, hypothalamus
DUSP6	+	Olfactory bulb	All	All low	Occipital (low), temporal (low)	All

Expression of the 25 IGD genes according to publically available databases. Only relevant tissue types were examined. TiGER (Tissue-specific Gene Expression and Regulation) (http://bioinfo.wilmer.jhu.edu/tiger/) (107) and BioGPS (http://biogps.org/#goto=welcome) (101) databases integrate data from multiple sources. The Human Protein Atlas (http://www.proteinatlas.org/) (106), FANTOM5 (Riken consortium Functional Annotation of the Mammalian Genome) (103), and GTEx (Genotype-Tissue Expression) (http://www.gtexportal.org/home/) (105) databases contain expression data from multiple tissues taken from human subjects. Data from the FANTOM5 project were obtained indirectly from the EMBL-EBI expression atlas, which compiles expression data from many of these databases (https://www.ebi.ac.uk/gxa/home) (102). +, expressed; −, not expressed; not found, expression not examined; (low), expression was just detectable.

E. The complex biology of the GnRH genetic network

We now know that most IGD genes can either disrupt the development and/or migration of the GnRH neurons and cause KS (neurodevelopmental pathway) or alter the secretion and/or action of GnRH neurons after this migratory journey has been completed, in which case they cause nIHH (neuroendocrine pathway). In addition, several IGD genes are considered “overlap genes,” because they are found to be disrupted in both KS and nIHH carriers. Presumably, these overlap genes have multiple roles in GnRH biology although for many genes this overlap remains to be examined. Figure 3 portrays the summary of all of these genes and their roles in each of these pathways. Although it is clear that the complete biology of GnRH neurons will ultimately be far more complex than this relatively simplistic formulation, this schema represents at least a temporary map onto which each new IGD candidate gene can be placed.

Figure 3.

Complex neurodevelopmental and neuroendocrine regulation of GnRH neurons. Genes to the left are critical to GnRH neuronal migration, whereas those on the right are involved in proper GnRH function. A number of IGD genes (center) are involved in both processes.

Several KS genes affect the formation of the olfactory bulbs and the migration of the GnRH neurons from the olfactory placode (8, 12). Significant protein interactions between some of these gene products, such as FGFR1 and its ligands with the heparin sulfate proteoglycans and anosmin (the protein encoded by KAL1), explain the effects of mutations in these proteins on the expression of the KS phenotype (84, 108–110). In contrast, some more recently discovered genes (eg, SOX10 and CHD7), are not associated with olfactory development, but rather cause a disruption of neural crest cell migration indicating their potential role in a possible alternate GnRH neuronal migratory pathway. These observations suggest and that the primary origin of some GnRH neurons might exist in other central nervous system sites distinct from the olfactory placode within the brain (5). Importantly, IGD patients carrying mutations in these genes display other phenotypic characteristics that could be attributed to the defective development of neural crest cells in other organ systems such as craniofacial nerve defects, midline and retinal abnormalities, cardiac defects, and inner ear maldevelopment (61, 65).

Genes that control the neuroendocrine regulation of GnRH cause the nIHH phenotype. Apart from the obvious biological significance of GNRH1 and its receptor GNRHR, a more complex neuronal network forms a multilevel neuroendocrine regulatory system that controls GnRH secretion. This complex neuroendocrine network appears to be “supervised” by the kisspeptin-neurokinin B-dynorphin neurons that reside in the arcuate nucleus and regulate GnRH neuronal function (111–113). Although the kisspeptin signaling system is highly sensitive to the feedback of sex steroid hormones and acts directly on a large proportion of the GnRH neurons (114–117), there are findings suggesting a more modulatory role of the neurokin B/neurokinin B receptor 3 feedback (118–120). However, kisspeptin-neurokinin B-dynorphin neurons project their axons directly to GnRH neuron terminals that express the neurokinin B receptor 3 (TACR3) in the median eminence, suggesting a direct interaction of neurokin B with kisspeptin neurons (121, 122).

F. The challenge of functional testing

One of the most compelling ways to substantiate the causative nature of a novel gene is to model the genetic mutation in an animal or cell system to recapitulate the disease phenotype. Although there are many examples highlighting the successful use of animal models in IGD gene validation (Table 1, criterion 5) (92), knockout and mutant mouse models are slow to develop and expensive to maintain. In addition, phenotypic readouts from animal models are often difficult to interpret and, due to different genetic background strains, do not always produce a human disease phenocopy nor provide an accurate readout of the reproductive defects caused by specific RSVs in these genes. Even though recessive LoF mutations in genes can be validated by the use of knockout mice and other knockdown models, missense mutations require further individual evaluation because of unpredictable effects on protein function. Most studies have used in vitro modeling to assess individual missense mutations. However, such experiments are technically demanding, time consuming, and costly. Importantly, because most IGD mutations are either heterozygous, compound heterozygous, or arise in a digenic or oligogenic manner, accurate modeling necessitates the development of a model system, such as zebrafish, worm, or modified stem cells, that can tolerate the simultaneous insertion of multiple hits and use the emerging CRISPR/Cas9 system of gene perturbation (123). Alternatively, it may be possible to use patient-derived samples to generate IPS cell lines for in vitro studies. These may be particularly useful for fulfilling criterion 7, the rescue of specific mutations. Currently, with the exception of a few mutations in 4 IGD genes, KAL1, KISS1, GNRH1, and GNRHR, most IGD mutations have not been tested in rescue experiments to prove that they are a causative (Table 1, criterion 7) (124–128). Although rescue of specific mutations is currently not the “gold-standard” for validating a gene candidate, it may well be critical to assigning pathogenicity to missense mutations in known genes.

Moving forward, it will be crucial to consider how to select the right functional assays for gene validation, particularly as some mutations may only alter specific second-messenger pathways or imbue neomorphic protein functions. The development of alternative reproductive model systems that are more scalable, cost effective, robust, and high throughput is a highly desirable goal and may, in fact, be critical to the future of the field.

G. Testing of genetic burden in IGD

Although a few previous studies have used case-control association tests to establish the potential burden of pathogenicity of individual IGD gene variants based on their frequency in IGD vs control populations (Table 3) (14, 28–30, 41, 44, 45, 61, 62, 71–75, 77, 78, 80, 83–85, 94, 129–146), few have addressed the overall genetic burden for each individual IGD gene. Past pursuit of this particular line of investigation has been limited for several reasons including: 1) the rarity of most genes causing IGD, which argues for large numbers of control individuals to be screened to obtain a proper grasp of any potential genetic burden in the disease cohort; 2) the lack of such data from appropriate ethnically matched controls; 3) the economic burden of screening full coding sequences of large genes for all allelic RSVs when compared with the more limited costs of targeted screening of specific mutations; and 4) the known limitations of bioinformatic programs based upon protein structure, evolutionary conservation, etc, to predict the functional impact of RSVs accurately.

Table 3.

Properties of Statistical Association Tests Performed on Identified Variants in IGD Genes

Gene	Total Number of IGD Probands Screened	Diagnosisa	Total Number of Controls	Citation
KAL1	1	KS	100	129
	1	KS	150	130
	28	KS	100	131
	30	KS	100 and dbSNP	14
GNRHR	2	nIHH	20	77
	46	nIHH	75	132
GNRH1	310	nIHH	192	78
FGFR1	28	KS	100	131
	80	nIHH and KS	100	133
	7	nIHH	235	41
	141	KS	150	134
	150	KS	250	135
	134	nIHH	270	136
	2	nIHH and KS	200	28
	30	KS	100 and dbSNP	14
FGF8	451	nIHH, KS and AHH	180	83
KISS1R	64a	nIHH	130	71
	30	nIHH and CDP	185	137
	166	nIHH	276	138
	69	nIHH	120	139
KISS1	103b	nIHH	100	73
	1025	nIHH, KS, CDP, HA, AHH	703 (178 and 1kG)	74c
TAC3/TACR3	56b	nIHH	100	72
	154b	nIHH	100	140
	345	nIHH	292	141
PROK2/PROKR2	192	KS	250	30c
	107	nIHH and KS	100	142
SEMA3A	386	KS	386 and EVS	45
	48 screened for deletions	KS	520	44
	34	KS	200 and EVS	143
SOX10	103	KS +− WS	50, 1kG and EVS	65
	18	KS	90, dbSNP, 1kG, EVS, and LOVD	144
AXL	104	nIHH and KS	150, NHLBI, 1kG, and EVS	85
NSMF (NELF)	33	KS	100	80
	168	nIHH and KS	372	145
	1	KS	384	28
WDR11	201	nIHH and KS	880	146
CHD7	101 (+96 exons 6–10)	nIHH and KS	>180	61
	36	KS	300	62
IL17RD	386	KS	155 and 1kG	29
DUSP6		KS
SPRY4		nIHH and KS
FLRT3		nIHH and KS
FGF17		nIHH and KS
HS6ST1	338	nIHH and KS	500 and EVS	84
FEZF1	30	KS	136, dbSNP, 1kG, and EVS	75
SEMA3E	121	nIHH and KS	1kG and EVS	94

Size of patient and control cohorts used for statistical association analyses for variants found in each of the 25 genes. dbSNP, The Single Nucleotide Polymorphism Database (http://www.ncbi.nlm.nih.gov/SNP/); 1kG, 1000 genomes database (http://www.1000genomes.org/); EVS, Exome variant server (http://evs.gs.washington.edu/EVS/); LOVD, Leiden Open Variation Database (http://www.lovd.nl/3.0/home); HA, hypothalamic amenorrhea.

^aDiagnosis of probands with mutations in the target gene.

^bThe studies included both endogamous pedigrees and sporadic cases.

^cAdditional genetic burden test performed.

The first attempt at establishing the genetic burden of an IGD gene was reported by Dode et al (30), who sequenced 192 unrelated KS patients and 250 ethnically matched control subjects for the genes PROK2 and PROKR2. Burden testing demonstrated that the total proportion of PROKR2 alleles was significantly higher in KS patients than in controls, strongly suggesting the involvement of this gene in IGD (30). This observation of an enhanced genetic burden in IGD vs controls was also supported by the validating presence of deleterious (nonsense) LoF mutations in their IGD cohort. Although this study was rigorous by the standards at the time, the number of healthy controls used is now considered inadequate. Employing similar statistical methods, Chan et al (74) found no statistically significant difference in the prevalence of the total number of KISS1 RSVs in 1025 patients (displaying both rare and common forms of IGD) vs 703 controls, mainly due to lack of controls.

Apart from these efforts, the cumulative burden of a subset of IGD genes was first tested by Sykiotis et al (26), who analyzed a cohort of 397 IGD patients and 179 healthy controls that underwent Sanger sequencing for 8 known IGD genes (KAL1, FGFR1, PROKR2, PROK2, NSMF, FGF8, GNRHR, and KISSR1). Even though detailed burden analysis for each separate gene was not described, the overall proportion of altered alleles in these 8 genes in the IGD cohort exceeded those found in the controls, showing that the IGD patients were significantly more likely to harbor rare protein-altering variants (26). This finding of the cumulative enrichment of RSVs in a subset of IGD genes remained consistent in a subsequent study by Miraoui et al, who investigated an expanded list of 17 genes (29), and in a study of GnRH-deficient females screened for 11 genes (147). Although these studies also had small numbers of controls available for comparisons, taken together, it is clear that there is a cumulative burden of variation in these genes in IGD patients.

More recently, a large control cohort became available to assert the causality of RNF216 in syndromic patients with Gordon Holmes syndrome (cerebellar ataxia and IGD). Although this gene is not yet a proven IGD gene, lessons can be learned from the approach taken to assess its burden. In comparison with the burden test done for PROKR2, this gene was screened in a patient population of 8 probands contrasted with more than 13 000 control chromosomes assembled from patients with apparently unrelated conditions assembled by the National Heart, Lung, and Blood Institute's Exome Sequencing Project as well as 672 chromosomes from ethnically matched control subjects. Using such large control populations, these investigators were able to conclude that the percentage of deleterious mutations in RNF216 in IGD patients exceeded what is expected by chance (53).

Importantly, although using such large cohorts robustly increases the number of control subjects, false positive findings may still occur due to the intrinsic imbalance of contrasting large-scale control datasets obtained from loosely phenotyped “normals” to smaller, more densely phenotyped and sequenced rare IGD patient cohorts. Additionally, to reduce the “noise” in these complex comparisons as much as possible, the screening of healthy individuals should ideally use identical sequencing pipelines and variant calling algorithms as the one that surfaced the candidate mutations in the disease population, a criterion that is often difficult to fulfill. Moreover, the depth of detailed phenotyping almost always varies considerably in subjects recruited for large-scale control cohorts, indicating the need for both careful clinical assessments and expansion of the number of individuals needed for recruitment to these datasets.

Another important consideration is that studies in isolated populations have now revealed significant enrichment for rare LoF mutations in known IGD genes (148). For example, the frequency of mutations in the PROKR2 gene is far higher in the geographically limited Maghrebian (North African) KS population than in European KS patients (149, 150). Additionally, 77% of familial nIHH cases of Turkish origin have been shown to be enriched for variants in a subset of neuroendocrine IGD genes, including GNRHR, TACR3, TAC3, KISS1R, and KISS1 (69). These examples suggest that different selection pressures operating within isolated populations may cause some variants to look common when in other contexts they are actually quite rare.

In summary, although the biological relevance of known IGD genes is fairly well established, in the past, these genes have been indicted largely owing to the considerable weight of LoF mutations, chromosomal structural abnormalities, and evidence from human genetics (such as large pedigrees with endogamy). Although deleterious variants and other evidence may have been sufficient to substantiate a gene in the past, moving forward, as NGS technology continues to improve, we will soon be faced with such a large number of novel genes and an even larger number of missense RSVs, which cannot be eliminated by pedigree analyses or other traditional approaches. Hence, when sequencing is done on a cohort of patients, rather than a family, it will ultimately become crucial to calculate the enrichment for the RSVs in a single gene (ie, their “genetic burden”) in the disease population compared with an ethnically matched control population with much greater accuracy. Although there may be reasons (such as gene size, sequence coverage, domain-specific location of RSVs, etc) that a valid candidate gene may not pass a genetic burden test, when there is a sufficiently large, well-phenotyped, ethnic-specific control population to compare with, the assessment of genetic burden should really occur as an important step to reduce the number of genes/RSVs requiring detailed functional analyses. In addition, although this review is only a discussion of “gene-level” burden, it will become increasingly important for this analysis to be done at the “variant level,” even after a gene has been sufficiently indicted as causative in IGD.

IV. The Future

A. New tools and approaches for the discovery of IGD genes

Although the clinical presentation of IGD patients has remained constant, newly developed genetic discovery tools are continually improving and there is constant effort to optimize these new techniques for genetic discovery. Thus, even though the theory behind the genetic analysis of IGD cases remains the same, NGS is highlighting the complexity of the underlying human genetic architecture of this disease and the need for methods to analyze it further. Several of the most recent successes in gene discovery in IGD are now derived from genome-wide searches, whereas in the past they had been quite targeted to shorter DNA segments based on the available approaches. It seems clear that NGS methods will continue to improve and as the numbers of IGD patients tested increases, the yield in new gene discovery will similarly rise. Consequently, we anticipate the rapid accrual of several helpful new tools that should improve the current ambiguities of determining the pathogenicity of RSVs in candidate genes. These include: 1) improved accuracy and calling programs for NGS; 2) increasing size and robustness of control cohorts from appropriate ethnic population studies; 3) rapid improvements in the understanding of the structural biology of proteins and the potential functional implications of specific mutations from crystallographic structural modeling; 4) increased availability of gene expression data from both diseased and normal human tissues, with more tissue subtype specificity; and 5) improved bioinformatics programs using all of the known interactions between proteins, pathway analyses, evolutionary data, and genetic associations of related diseases, etc.

Clearly, whole-genome sequencing (WGS) will increasingly become the genetic standard for mutational screening as its costs decline and supporting bioinformatics capabilities become more widely available and standardized. The data generated from WGS will provide additional insights into the genomic complexity of IGD patients and help optimize the data already collected using already standardized tools. For example, even though traditional karyotyping can identify chromosomal abnormalities, novel techniques like jumping libraries-based WGS that focus on deeper genomic resolution allow for a more precise definition of breakpoints and the detection of cryptic structural events, balanced chromosomal rearrangements, and deletions/duplications of various sizes (151, 152).

In addition to WGS to identify rare variants, genome-wide association studies to assess common variation are increasingly large, available, and able to identify genetic loci associated with more quantitative traits such as the onset of puberty (153) and menopause (154–157), thereby offering new insights in the genetic complexity of reproductive disorders. Importantly, few loci associated with the age of menarche were mapped close to known IGD genes, and most genome-wide associations are either 1) localized at some distance from coding regions revealing possible long-range DNA interactions; 2) interrupt transcription factor binding with possible effects on downstream targets; or 3) interrupt noncoding RNA, which may play a role in the regulation of transcription and translation of target genes (158, 159). New approaches will need to be used to assist in the interpretation of such variation found in noncoding regions. This will be a particularly challenging endeavor, because there is likely exquisite tissue specificity of expression of many genes. To prioritize the gene candidates, publicly available online tools (such as GENets, the ENCODE database, etc) should be used to construct gene-regulatory networks that map the new candidate in gene ontology pathways, assess DNA methylation, histone marks, DNase-I hypersensitivity sites, as well as RNA sequencing data. Additional prioritization of the target genes will also be based on tissue-specific expression and their potential to play a role in the development of the GnRH network.

All of these new types of data generation and analyses will also introduce powerful new ways to investigate environmental impacts on known genes in the initiation of puberty and the biology of pubertal development. In fact, both DNA methylation and histone modifications are now being investigated for several of the IGD genes (160, 161). Thus, apart from the traditional approaches already used for detection of known genes, new tools are likely to provide greater insight in the genomic and epigenomic complexity of the both known IGD genes and future gene candidates, while pointing the field towards genomic regulation and the detection of rare cryptic events.

V. Conclusions

The discovery and validation of the known 25 IGD genes has led to a remarkable improvement in our basis of understanding of the development of the GnRH neuronal network, the neuroendocrine regulation of these unique neurons, and the role these genes play in several reproductive disorders. With the field moving forward using NGS, we assessed the known genes as held to the current standards in the field. Notably, most the genes lack rescue experiments that provide support for genetic causality and are missing well-constructed gene burden testing (criteria 6 and 7). Although it may not be necessary to meet these 2 criteria when the other (particularly genetic) support is overwhelming, in the future, we expect that the rescue experiments will be particularly critical for variant-level analyses, and that the statistical analysis will become an important proxy for genetic information in sporadic cases or small families. As we expected, most the criteria were met by most of the older and more established genes; however, further analysis is required for the determination of the causality of the most recently reported candidates (see Table 1). Such efforts should be focused on mapping the role of the new candidates in the known neurodevelopmental or neuroendocrine pathways of IGD, using expression experiments and bioinformatic tools that assess the interactions between candidate proteins, leveraging those already known to cause IGD. Further studies using the proper control cohorts should also be undertaken for a more precise assessment of the genetic burden for each of the associated genes. In addition, functional studies should be conducted for testing the pathogenicity of the detected variants both in vitro and in vivo. Although the focus of this review has been assessing the candidates on the gene level, variant level analysis will also be critical to ensure that reporting of “causal” genetic defects is confined to RSVs that have sufficient experimental validation. In addition, although we have applied criteria and analyzed genetic discoveries in a specific disease model, many of the same methodologies and considerations apply in the analyses of other Mendelian disorders.