Genetic aetiology of primary adrenal insufficiency in Chinese children

Background Primary adrenal insufficiency (PAI) is life-threatening, and a definitive aetiological diagnosis is essential for management and prognostication. We conducted this study to investigate the genetic aetiologies of PAI in South China and explore their clinical features. Methods Seventy children were enrolled in this cross-sectional study. Clinical information was collected, and combined genetic tests were performed according to the children’s manifestations. Statistical analysis was performed among the different groups. In silico or in vitro experiments were applied to determine the pathogenicity of novel variants. Results Among the 70 children, 84.3% (59/70) were diagnosed with congenital adrenal hyperplasia (CAH), and 21-hydroxylase deficiency (21-OHD) was genetically confirmed in 91.5% of these cases. Salt wasting (SW), simple virilization (SV), and non-classic (NC) CAH accounted for 66.1% (39/59), 30.5% (18/59), and 3.4% (2/59) of the cases, respectively. The 17-hydroxyprogesterone (17-OHP) and testosterone (TES) levels were significantly higher in children with SW than with SV. The 17-OHP and cortisol levels in female SW patients were significantly higher than those in males. The 17-OHP, cortisol, dehydroepiandrosterone (DHEAS) and TES levels in female SW patients were significantly higher than those in female SV patients. Additionally, 72.7% (8/11) of uncharacterized PAI patients had positive genetic findings. Among all the patients, two novel variants in the CYP21A2 gene (c.833dupT and c.651 + 2T > G) were found. A microdeletion (Xp21.2–21.3) and five novel variants, including 2 in the NR0B1 gene (c.323–324CG > GA and c.1231_1234delCTCA), 2 in the AAAS gene (c.399 + 1G > A and c.250delT) and 1 in the NNT gene (c.2274delT), were detected. The novel variant c.399 + 1G > A in the AAAS gene was further confirmed to lead to exon 4 skipping during mRNA transcription and produce a truncated ALADIN protein. Conclusions We found ethnicity-based differences in the CYP21A2 gene variant spectrum among different study populations. Female 21-OHD patients tended to have higher 17-OHP and TES levels, which warrants caution in relation to the effects of virilization. Novel gene variants detected in the CYP21A2, NR0B1, AAAS and NNT genes expanded the genetic spectrum of PAI, however, further improvement of genetic testing tools beyond our protocol are still needed to uncover the complete aetiology of PAI in children. Supplementary Information The online version contains supplementary material available at 10.1186/s12920-021-01021-x.

as well as adrenal sex steroids [1]. PAI in children is very different from that in adults, as congenital disorders caused by genetic aberrations are much more frequent [2]. Based on the affected biological process, the aetiology of inherited PAI can be categorized into several groups [3]. The first is impaired steroidogenesis, such as congenital lipoid adrenal hyperplasia caused by mutations in STAR [4] and 21-hydroxylase deficiency caused by CYP21A2 [5]. This group includes different abnormalities in the steps of steroid biosynthesis. Congenital adrenal hyperplasia (CAH), which is caused by 21-hydroxylase deficiency, accounts for most paediatric PAI cases, ranging from approximately 70-80% of cases [2,3]. The second group is adrenal hypoplasia, including X-linked adrenal hypoplasia congenita (AHC) caused by NR0B1 [6], adrenal hypoplasia caused by steroidogenic factor-1/NR5A1 [6], IMAGe syndrome caused by CDKN1C [7] and MIRAGE syndrome caused by SAMD9 [8]. Patients in this group usually lack all adrenocortical hormones, and some of them may have accompanying symptoms such as intrauterine growth restriction and genital abnormalities. The third is adrenocorticotrophic hormone (ACTH) resistance, including familial glucocorticoid deficiency type 1 caused by MC2R [9] and familial glucocorticoid deficiency type 2 caused by MRAP [10]. Children in this group usually have a preserved reninaldosterone axis, and plasma ACTH is often markedly increased despite normal glucocorticoid treatment. Adrenal destruction constitutes another group, consisting of Allgrove syndrome caused by AAAS [11], nicotinamide nucleotide transhydrogenase deficiency caused by NNT [12]and thioredoxin reductase deficiency caused by TXNRD2 [13]. The next group is complex lipid metabolism, including X-linked adrenoleukodystrophy (X-ALD) caused by ABCD1 [14], neonatal adrenoleukodystrophy caused by PEX1 [15] and Wolman disease caused by LIPA [16]. The last group is autoimmune destruction, including isolated autoimmune adrenalitis associated with CLTA-4, HLA-DR3, HLA-DR4, HLA-B8 and BACH2 [3].
Among the remaining patients, the aetiology varies greatly based on the study population and test platform. In Asia, Amano et al. [17] applied target gene sequencing and discovered that when excluding 21-hydroxylase deficiency and 3β-hydroxysteroid dehydrogenase deficiency, genetic defects were detected in 84.7% of patients, with STAR (32.2%), NR0B1 (30.5%) and SAMD9 (11.9%) accounting for the top three. In North America, Tsai et al. [18] used the same method to determine the cause in 81.8% of patients, and the most common was STAR (45.5%). In the Middle East, Guran et al. [19] employed a gene panel plus next-generation sequencing and found 81% positive genetic diagnoses, with MC2R (26.3%), NR0B1 (12.6%) and STAR (11.6%) as the three most common causes. In Europe, Chan et al. [20] applied whole exome sequencing (WES) and made genetic diagnoses in 41.9% of patients, with CYP11A1 (14.0%), NR0B1 (9.3%) and AAAS (4.7%) as the most affected genes.
In clinical practice, the differential diagnosis and management of paediatric PAI are quite challenging [21]. Eyal et al. reviewed the epidemiology and risk factors for adrenal crises in children with adrenal insufficiency (AI) between 1990 and 2017 at four Israeli paediatric endocrinology units and found that diagnosis and long-term management of paediatric patients remained a challenge [22]. Adrenal crises are life-threatening emergencies, but studies on the rate and risk factors for adrenal crises in children with AI are scarce [22]. Currently, the exact cause of more than 5% of paediatric PAI cases is undetermined despite continual detection of novel genetic causes [23]. Considering the limited data for Chinese PAI children, we used multiple molecular testing strategies to explore genetic causes of PAI in Chinese children and to determine the correlation between genotype and phenotype.

Subjects and method
Subjects Seventy children were diagnosed with PAI from July 2012 to August 2017 at Children's Hospital of Fudan University according to published criteria [24]: (i) initial symptoms suggesting low cortisol levels (e.g., hyperpigmentation, fatigue, failure to thrive, vomiting, electrolyte disturbances); (ii) laboratory examinations indicating high plasma ACTH levels accompanied by low or normal cortisol levels and a less than 500 nmol/L peak during the ACTH stimulation test; and (iii) availability for genetic testing. All the patients received detailed clinical evaluations, and all female CAH patients received Prader genital scale (PGS) scores for external genitalia. Patients with hyperkalaemia, hyponatremia, elevated serum 17-hydroxyprogesterone (17-OHP), and external genitalia virilization manifestations were initially suspected of having 21-hydroxylase deficiency (21-OHD), which was further verified through CYP21A2 gene testing [25,26]. The diagnostic algorithm for PAI patients is shown in Additional file 1: Fig. S1. The study was approved by the Ethics Committee of the Children's Hospital of Fudan University. Written informed consent for the study was obtained from all patients' parents.

Molecular analysis
Genomic DNA was extracted from peripheral lymphocytes, and four genetic tests were performed according to individual clinical presentations: (i) for 59 patients who presented with cortisol deficiency, androgen excess and/or elevated 17-OHP, the CYP21A2 gene was amplified using PCR (Additional file 4: Table S1) and double checked using multiplex ligation-dependent probe amplification (MLPA; SALSA MLPA Probemix, MRC-Holland, Netherlands); (ii) for 2 patients with developmental delay/intellectual disability or congenital anomalies, array-based comparative genomic hybridization (Array-CGH, Agilent Technologies, CA, USA) was performed; and (iii) the remaining patients underwent WES (Sure-SelectXT Human All Exon Kit V6, Agilent Technologies, CA, USA).

Determination of the AAAS mRNA structure with the c.399 + 1G > A variant
One patient harboured a mutant AAAS gene with c.399 + 1G > A. Subsequent mRNA analysis was performed to test its pathogenicity. Human Splicing Finder (HSF) 3.0 software (http:// www. umd. be/ HSF3/ index. html) was used to predict the potential donor or acceptor site alteration of the altered splice site. Total RNA was isolated from the peripheral blood of the proband, the parents and normal control individuals (Total RNA Kit, TIANGEN BIOTECH, Beijing, China). The forward primer 5'-TGG ATC AAT CTT CCT GTC CTA CAA C-3' , covering the first 25 bp of exon 2, and the reverse primer 5'-TAC AGT GAA CAG CAG TCG G-3, 46 bp upstream of the end of exon 11, were used to amplify the transcripts of both the wild-type and mutant AAAS genes. The amplification product of normal individuals was predicted to be 919 bp. The PCR products were separated by agarose electrophoresis and then extracted from the agarose gel (TIANgel Midi Purification Kit, TIANGEN BIO-TECH, Beijing, China). The diversity of transcripts was verified by Sanger sequencing.

Genotype-phenotype correlation in children with CAH
CAH patient genotypes were categorized into 5 groups as previously described [27,28]. Briefly, group 0 included null variants on both alleles; group A included homozygotes for the I2G variant or compound heterozygotes for the I2G variant and a group 0 variant; group B included patients who were either homozygous or compound heterozygous for the p.I173N variant with group 0 or A variants; group C included patients who were either homozygous or compound heterozygous for the p.P31L or p.V281L variants with group 0, A or B variants; and group D included unidentified variants and new variants. The genotype-phenotype correlation was assessed in the patients with both genetic and initial clinical information. The positive predictive value (PPV) was calculated as the percentage of patients with the predicted phenotype. The predicted phenotypes were salt wasting (SW) for groups 0 and A, simple virilization (SV) for group B, and nonclassic (NC) for group C.

Statistical analysis
SPSS Statistics (19.0) software was used for statistical analysis. The data are expressed as the median and inter quartile range (IQR) due to non-normal distributions. Comparisons between groups were performed using the Mann-Whitney U test without the missing data, and differences were regarded as significant if p < 0.05.

Mutation spectra in the CYP21A2 gene and the genotypephenotype correlation in 21-OHD patients
Combined Sanger sequencing and MLPA tests yielded 91.5% positive (54/59) variant findings. The most common variants were c.293-13C > G (31.3%), Del (18.6%), p.I173N (17.0%), E3 Δ8 (5.1%), p.R483PfsX58 (4.2%) and p.R357W (3.4%) ( Table 1). Two novel variants, namely, c.833dupT (NC_000006.1: g.32007876dupT; NM_000500.9: c.833dupT, reference protein NP_000491: p.279GfsX17) and c.651 + 2T > G (NC_000006.1: g.32007486T > G; NM_000500.9, c.651 + 2T > G, reference protein NP_000491: p.?), were found in two patients (Additional file 2: Fig. S2). The variant c.833dupT in exon 7 of the CYP21A2 gene was found in an SV patient with the p.I173N variant on the other allele. The splicing variant c.651 + 2T > G at the end of exon 5 of the CYP21A2 gene was found in an SW patient with the c.293-13C > G variant on the other allele ( Table 2). The protein was predicted to be p.(184D_217Rdel). Complete concordance was observed only in group 0, while the concordance values were 83.3% and 84.6% for groups A and B, respectively. Two SW patients in group D with only one pathogenic allele presented with elevated 17-OHP, growth retardation, hyperkalaemia, hyponatremia and adrenal hyperplasia. The other two SW patients in group D with the same presentations had no pathogenic variants in the CYP21A2 gene. All the above patients refused to undergo further genetic examination. One NC patient in group D who developed increased 17-OHP and dehydroepiandrosterone with mild pigmentation and bilateral adrenal hyperplasia had a monoallelic variant (Additional file 6: Table S3).

Mutation spectra and the genotype-phenotype correlation in children with uncharacterized PAI
Highly diverse genetic defects were detected in 72.7% (8/11) of probands from 7 families, and five novel variants were detected in five patients, including 2 in the NR0B1 gene (NC_000023.1:    Fig. S3). Case 1 (mutant ABCD1 gene) showed elevated very longchain fatty acid levels. Case 2 (IL1RAP-NR0B1-GK deletion, Xp21.2-21.3) presented with undetectable adrenal glands, increased triglyceride levels, hypothyroidism and cryptorchidism, in addition to typical AI and mineralocorticoid deficiency. Among the four patients with    (Table 3).

Pathogenicity analysis of the novel splicing variant c.399 + 1G > A
Case 5 (Table 3) harboured a novel variant c.399 + 1G > A at the splicing juncture site at the end of exon 4 in the AAAS gene, which was predicted to affect splicing, as the wild-type motif "TCTgtaagt" was changed to "TCTataagt" without uncovering a cryptic splice site. Agarose electrophoresis of the cDNA from the proband, the parents and a healthy control demonstrated that the parents had two bands: one was a higher-molecular-weight band of the same size as that of the normal control individual, and the other was a lower-molecular-weight band of the same size as that of the patient. Sanger sequencing confirmed the exon 4 skipping in the lower-molecular-weight molecular band, and the protein was predicted to be truncated (Fig. 2).
Novel variants, namely, c.651 + 2T > G and c.833dupT, were found in the CYP21A2 gene in two children (Table 2), and these variants were predicted to be damaging. Variant c.651 + 2T > G located at the end of exon 5 was predicted to alter the wild-type donor site and most likely lead to whole exon 5 skipping, which would result in the loss of 184-217 amino acid residues. With 9 highly conserved residues [39], these residues are essential to construct important functional domains, such as steroid-binding sites (residues 203-207) and large hydrophobic areas (residues 211-218). This splicing site variant likely impairs enzyme activity, because the proband displayed the SW phenotype. The other novel variant, c.833dupT, most likely causes complete loss of enzyme activity, because another comparable frameshift variant, c.923dupT (p.L308FfsX6), which retains more amino acid residues, causes complete loss of enzyme activity [40]. However, more functional studies are required to confirm the exact impact of the novel variants on enzyme activity in the future.
In our research, SW females had higher PGS scores than SV females, as reported by previous studies [41,42], and the reason may be the higher 17-OHP and TES levels discovered in SW patients (Fig. 1). Interestingly, compared with males, females had higher serum 17-OHP and cortisol levels in both the SW and SV groups, which is consistent with another study [32]. Regarding the genotype-phenotype correlation, the PPV for group 0 was 100%, which is higher than those in most similar studies, while the PPVs in group A (82.6%) and group B (84.6%) were in accordance with those in previous studies [27,32,35,37,38]. Surprisingly, we found that 8.5% (5/59) of the patients clinically presented with the SW (n = 4) and NC (n = 1) phenotypes without biallelic variants (Additional file 6: Table S3). Studies have reported that monoallelic variants and absent variants accounted for 2.2% to 24% of enrolled children [27,33,[36][37][38]43]. The aetiology is unknown. We deduced that an amplification allele dropout effect may occur due to the high similarity between the CYP21A1P and CYP21A2 genes. Additionally, the CYP21A2 gene promoter region and other intronic variants that have not been analysed in studies may also reduce transcriptional activity [21].
In our uncharacterized PAI patients, we detected a total of 7 pathogenic variants, including one chromosome microdeletion and 5 novel variants in the NR0B1, AAAS, NNT and ABCD1 genes. The NR0B1 gene is the most frequent gene responsible for X-linked AHC in males ( Table 4). The absence of NR0B1 is proposed to cause progenitor cells to prematurely differentiate into steroidogenic cells without adequate maturity [44,45]. We found two patients with one novel variant (c.323-324 CG > GA and c.1231_1234delCTCA) and one infant with an NR0B1 deletion that manifested as PAI, mineralocorticoid deficiency and diminished adrenal glands; however, no hypogonadotropic hypogonadism was noted due to the patient's young age. We also found two novel homozygous pathologic variants in the AAAS gene (c.399 + 1G > A and c.250delT) in 3 patients (Table 3, cases 6, 7, and 8) from 2 families, and both parents were consanguineous. Alacrima and AI were found in all three patients; however, achalasia was observed in only an 8-year-old boy, manifesting as swallowing difficulties. The other two sisters shared the same genotype with different phenotypes. These results suggest that careful estimation should be performed in every child with PAI with alacrima or achalasia. The ABCD1 gene is another causative gene that is often observed in male patients and results in X-linked ALD [45]. The clinical presentation of X-linked ALD is variable, and no phenotype-genotype correlation has been observed [46]. These phenotypes include cerebral, adrenal, spinal cord and peripheral nerve involvement [46]. Fifty percent of affected children ultimately develop adrenomyeloneuropathy within 10 years [47]. Our patients showed clinical PAI, elevated serum long-chain fatty acid levels and normal neurological examination results but had white matter lesions detected by magnetic resonance imaging. We also found a novel pathogenic variant in the NNT gene (c.2274delT). Researchers reported that 53% of patients with NNT gene variants presented with hyperpigmentation, while 17% had mineralocorticoid deficiency [48]. Cardiac and thyroid involvement may also exist [49]. Therefore, a close long-term follow-up is still needed, as our patient presented with only hyperpigmentation. Evident genetic defects were not detected in three PAI patients, and we deduced that our current test methods may fail to identify deep intronic variants and epigenetic changes [17][18][19][20].

Conclusions
In conclusion, the genetic causes of PAI in children are very diverse, with a high prevalence of 21-OHD. A total of 7 novel gene variants were detected in the CYP21A2, NR0B1, AAAS and NNT genes in our cohort. SW 21-OHD females showed higher PGS scores than SV females, which was consistent with the higher levels of 17-OHP and TES. Ethnicity-based differences existed in the genetic variant spectrum of the CYP21A2 gene. Evident genetic variants were not detected in 11.43% of the PAI patients despite our combined genetic testing protocol, suggesting that testing tools must be improved in future studies.