Three hundred and fifty-seven probands were admitted to Peking Union Medical College Hospital (Beijing, China) between January 2005 and December 2012 and diagnosed as CHH. This study included 327 unrelated CHH Chinese probands: 148 nCHH (141 males; Seven females) and 179 KS (167 males; 12 females), excluding 30 patients diagnosed as carrying FGFR1 variants (Nie et al., 2021). The study was approved by the Ethics Committee for Human Research of Peking Union Medical College Hospital; all patients and guardians of children provided written informed consent.

The inclusion and exclusion criteria were laid down based on CHH standard consensus (Boehm et al., 2015). Inclusion criteria: (1) absent or incomplete puberty by the age of 18 years for male and 16 years for female; (Herbison, 2016) cryptorchidism without a definite diagnosis; (2) serum testosterone≤100 ng/dl for males or estradiol≤20 pg/ml for females, with low or normal serum gonadotropin level. The participants were classified as having KS; (3) when the inquiry information showed the presence of anosmia or hyposmia, otherwise they were classified as having nCHH. Exclusion criteria: diagnosis of secondary hypogonadotropic hypogonadism induced by tumors, trauma, drugs, or other systemic diseases.

We collected blood samples from all participants. Focusing on the CHD7 gene, the Sanger sequencing technique was used to detect causative variants. Genomic DNA extraction was performed using the QIAGEN Midi Blood kit (QIAGEN, Germany) from leukocytes in peripheral blood. Polymerase chain reaction (PCR) was used to amplify the specific sequence of CHD7 gene. All experiments followed the manufacturer’s instructions.

We read the sequencing results by 4 Peaks (Nucleobytes, Netherlands), then blasted each in NCBI (https://blast.ncbi.nlm.nih.gov/Blast.cgi), against the following reference sequences: NG_007009 (g.DNA); NM_017780 (c.DNA); NP_060250 (p.Protein). Considering the digenic or oligogenic pathogenic possibility, which may explain why individuals with the same CHD7 variant presented with different phenotypes, we screened and annotated the other CHH-related genes by Sanger sequencing, including ANSO1, FGF8, PROK2, PROKR2, GNRH1, GNRHR, KISS1, KISS1R, TAC3, TACR3, LEP, LEPR, NELF, WDR11, HS6ST1, and SEMA3A. All variants were reported according to Human Genome Variation Society (HGVS) nomenclature rules (Dunnen and Antonarakis, 2000).

Based on the ACMG standards and guidelines³, we filtered all detected variants step by step. The core workflow includes general population database search, case-control statistical comparison based on calculation of odds ratio (OR) and the 95% confidence interval (95%CI) (referring to the statistical analysis section for details), disease database search, literature query, variant-type-specific analysis, and computational prediction (Figure 1). We used the allelic frequency from gnomAD database to calculate OR and define a rare sequencing variant (RSV) [a maximum allele frequency (MAF) < 0.0001]. Variants databases from patients included ClinVar and CHD7 databases. Literature query was performed in PubMed, Google Scholar and China National Knowledge Infrastructure (CNKI). The search key word included “CHD7 gene,” “hypogonadotropic hypogonadism” and related variants without any language, timeframe or article type restriction.

Computational prediction included three parts: pathogenic prediction by in silico tools, conservation analysis by alignment, and three-dimensional visualization by three-dimensional modeling software. In silico tools included five [SIFT (Sim et al., 2012), Polyphen-2 (Adzhubei et al., 2010), SNP&GO (Schwarz et al., 2010), Mutation Assessor (Reva et al., 2011), and Mutpred (Li et al., 2009)] for missense pathogenicity prediction (when more than three predicted the variants as pathogenic, the evidence pointed to PP3, otherwise it was BP4) and two [Splice Site Score Calculation and SpliceAI (Jaganathan et al., 2019)] for splicing sites (Table 2). We judged the evolutionary conservation of each amino acid site by aligning the human CHD7 sequence with orthologs of other 25 related species in UniProt database with Clustal W; the more conservative the site is, the more important it may be for protein function, which indirectly reveals its pathogenicity. Amino acid change was visualized by ChimeraX, a three-dimensional modeling software. By comparing the change in the identity of wild and mutant amino acids and their contact with other amino acids in a spatial model, we can predict the potential pathogenic effects the specific variant brings to the protein.

Data for all pathogenic and benign variants considering the above aspects were collected when available, and then based on the weight of the evidence, we classified the variants into six groups: pathogenic (P), likely pathogenic (LP), uncertain significance with paradoxical evidence (US1), uncertain significance without enough pathogenic evidence (US2), likely benign (LB), and benign (B). All specific criteria above were based on the ACMG guideline, except that we further divided the US ones into US1 and US2 based on the type of evidence (paradoxical evidence or insufficient pathogenic evidence).

We recorded basic information of all CHH patients during their first visit, including sex, age, height, weight, testicular size, penis length, hormone levels, and olfactory magnetic resonance imaging (MRI) results, medical history and family history. Prader orchidometer was used to measure the testicular size, and the mean volume was calculated for statistical analysis. Hormone tests results included levels of serum follicle-stimulating hormone (FSH), luteinizing hormone (LH), and testosterone (T), which were measured by chemiluminescent immunoassays (Bayer Diagnostics Corporation, United States).

For CHH patients with CHD7 RSVs, detailed phenotyping was performed when possible. Clinical evaluation items were based on the Human Phenotype Ontology database and were further classified into three types: CHH-related, CHARGE syndrome-related (including the function and appearance of eyes, nose, ears, mouth, and face; growth status, intelligence, and heart condition), and others (including the condition of the skeleton, muscle, palatal arch, skin, digestive system, and urinary system and synkinesia). The CHARGE-related symptoms were based on Verloes’ criteria¹³. Information about bone deformity, limb development, skin texture, condition of the digestive and urinary systems, function and appearance of the eyes and ears, and medication history was collected by inquiry and physical examination. MRI or temporal bone computed tomography (CT) was used to evaluate the middle and inner ear conditions whenever possible along with an auditory test. The condition of the eyes was evaluated by ophthalmologist with fundoscopy. Cardiac ultrasound were performed in cases with clinical suspicion of cardiac abnormalities. Results of genotypic and phenotypic analyses of family members were also collected if available.

Continuous variables were checked for normal distribution by SPSS version 21 software package (IBM, China). Normally distributed variables are described by the mean and SD, and nonnormally distributed variables are described by the median and interquartile range. Categorical variables are presented as a percentage. The OR and 95% CI were calculated by using an online OR calculator (http://www.hutchon.net/ConfidOR.htm/) (Bland and Altman, 2000). When OR > 5.0 and the 95% CI did not include 1.0, it was assumed that the prevalence of the variant in patients is significantly increased compared with the prevalence of controls.

A total of 36 CHH (20 KS; 15 nCHH) out of 327 (11.01%) carried heterozygous CHD7 gene RSVs (MAF <0.0001) (Supplementary Figure S1). All these patients complained of puberty absence and small testicular size [1.5 (1, 2.75)], and the percentage of cryptorchidism and concealed penis were 16.67% (6/36) and 5.56% (2/36), respectively. The mean diagnosis age was 19.75 ± 4.88 years. Among 13 patients who accepted the olfactory MRI, all patients had absence or hypoplasia of the olfactory bulb or (and) tract to different degrees. The results of hormone tests, the mean basal LH, FSH, and testosterone levels were consistent with hypogonadotropic hypogonadism, and the specific values are shown in Table 1.

Thirty CHD7 gene variants were detected, including three splice site (c.2835+1G > A, c.5210+3A > G, and c.7972-1G > C) and 27 missense variants (D3N, Q40H, S137A, D618G, D728N, E738D, D740G, P783S, I872T, R886W, M888V, R897P, R944H, C1251F, I1311T, G1506S, K1699E, R1743C, N2118S, S2123C, K2235Q, R2284Q, V2327G, R2361S, D2390E, S2453N, and N2808K) (Supplementary Figure S1). All variants were distributed throughout the CHD7 gene and showed a certain cluster in exons 2, 4, and 8. Six variants were located in known functional domains of CHD7 protein: three (D740G, M888V, and R886W) in Chromo2, one (C1251F) in SNF2, one (I1311T) in Helicase, and one (c.7972-1G > C) in BRK2, and 12 variants clustered in a protein region that was not conserved and its function was unknown (Figure 2).

Particularly, each variant existed only in one proband, except for Q40H, R1743C and E738D presenting in two patients and G1506S in three patients. Three patients harbored other heterozygous missense gene variants besides CHD7: one patient with GNRH1 (G34R in Patient 16) and two patients with WDR11 (V336F in Patient 22 and L891W in Patient 12) (Table 1).

We classified the 30 CHD7 variants according to HGVS, then analyzed their pathogenicity based on ACMG criteria (Table 2). Besides R944H, which had OR < 5.0 and 95% CI included 1.0, the other variants were either absent in the gnomAD database or the corresponding OR ≥ 5.0 and 95% CI did not include 1.0. Among these 30 variants, 11 were recurrent and 19 were reported for the first time. Six variants existed in the ClinVar database: S137A, P783S and D2390E were reported in CHARGE syndrome patients; R944H both in CHARGE syndrome and HH patients; R2284Q had no specific disease description. The amino acid sites of five variants were reported in the CHD7 variants database: three (c.5210+3 A > G, D728N, and R944H) had the same amino acid change and two (C1251F and R2284Q) mutated into a different one. Besides, four [D728N (Bartels et al., 2010), R886W (Marcos et al., 2014), D740G (Bilan et al., 2012), R944H (Bartels et al., 2010)] were reported by other researchers in CHH or CHARGE syndrome cohorts, and the detection rates were 1/642, 1/313, 1/50, and 1/642, respectively.

In computational pathogenicity prediction, firstly, the results of in silico analyses showed that, except for E738D and K2235Q (less than three tools defined these as pathogenic) with BP4 evidence, the remaining missense variants and three splicing site variants all were pathogenic and showed PP3 evidence. Secondly, focusing on 17 novel missense variants, alignment results indicated that except for N2808K, the other 16 wild type residues at a specific site were highly conserved across 25 different species (Figure 3). Thirdly, as three-dimensional models of R897P, I1311T, C1251F, R1743C, and K1699E (built by SwissModel) are available, we can see the direct harmful effect of a single amino acid change in the tertiary structure of the protein. The size, charge, and hydrophobicity of all five residues changed at the specific site: for R897P, the acidic amino acid changed into a nonpolar aliphatic one. Seven old contacts and 1 hydrogen bond (H-bond) broke, which accompanied with 6 new contacts and 1 clashes formed; for C1251F, the neutral polar amino acid changed to an aromatic one, besides, 1 old contacts, 1 clashes (unfavorable interactions where atoms are too close together) and 1 H-bond lost; for I1311T, a neutral polar one replaced the nonpolar aliphatic amino acid, and 3 old contacts broke; for K1699E, 8 new contracts formed; for R1743C, 7 contacts lost (Figure 4). Based on these findings, 3/30 (10.00%) variants were classified as P variants, 4/30 (13.33%) were LP, 3/30 (10.00%) were US1, and 20/30 (66.67%) were US2.

None of the second variant could be classified as the digenic pathogenic one according to the existing evidence. GNRH1 variants were inherited in an autosomal recessive manner; therefore, a direct pathogenic effect was impossible. WDR11 variants, which showed an autosomal dominant pattern of inheritance, had V336F classified as US1 and L891W as US2. Owing to the lack of experimental evidence, it is necessary to study whether these variants contribute to disease pathogenesis in a synergistic manner.

Twenty-three patients (16 KS and seven nCHH; 21 male and two female) harboring RSVs accepted detailed clinical evaluation, including two with P variant, two with LP variant, two harboring US1, and 17 with US2 (Table 3). Most of them (65.22%; 15/23) had extra-gonad phenotypes; CHARGE or CHARGE-like symptoms accounted for 80.0% (13/15) cases and other CHD7-related symptoms were seen in 53.3% (8/15) cases. All CHARGE or CHARGE-like features, except for rhombencephalic anomalies and malformation of mediastinal viscera, occurred in our patients; hearing loss and high myopia were the most frequent ones (5/10; 50%). Patient 21(P21 in Table 3) was re-diagnosed as having typical CHARGE syndrome based on three major and six minor diagnostic items. Other CHD7-related symptoms included high arched palate (50%; 4/8), spinal malformation (25%; 2/8), digestive system dysfunction (25%; 2/8), secondary hypothyroidism (12.5%; 1/8), six fingers in the left hand (12.5%; 1/8), inguinal hernia (12.5%; 1/8), micrognathism (12.5%; 1/8), widened palpebral fissure (12.5%; 1/8), epicanthus (12.5%; 1/8), and short philtrum fish mouth (12.5%; 1/8).

Occurrence of extragonadal symptoms correlated with the pathogenicity of variants. All CHH patients with P or LP variant manifested CHARGE-related symptoms or (and) other CHD7-related manifestations besides CHH. For patients carrying US1 and US2 variants, the frequency of CHARGE-related symptoms or others was 0% (2/2) and 64.71% (11/17), respectively.