We conducted a study of 12 children with CS from four hospitals in different regions of Southern China. All patients met the clinical diagnostic criteria of CS updated by Hale et al. (2016). We required all physicians to collect as much clinical information as possible, including gestation period inspection, history of prenatal care, and other data, plus the survival rate during follow-up. All specimens were collected with the informed consent of the family. This study was reviewed and approved by the ethics committee of Guangdong Provincial People’s Hospital.

Genomic DNA was extracted from peripheral blood using the Solpure Blood DNA kit (Magen Biotechnology) according to the manufacturer’s instructions. The genomic DNA of the patients was then fragmented using a Q800R sonicator (Qsonica) to generate 300–500 bp insert fragments. The paired-end libraries were prepared following the Illumina^® library preparation protocol. Custom-designed NimbleGen SeqCap^® probes (Roche NimbleGen, Madison, Wisconsin, United States) were used for in-solution hybridization to enrich target sequences. Enriched DNA samples were indexed and sequenced on a NextSeq^® 500 sequencer (Illumina^®, San Diego, California, United States) with 100–150 cycles of single-end reads according to the manufacturer’s protocols.

Primary data came in FASTQ format after the image analysis, and base calling was conducted using the Illumina^® pipeline. The data were filtered to generate “clean reads” by removing adapters and low-quality reads (Q20). The average coverage depth of sequencing was >200 ×, and 96% was >20 ×. Sequencing reads were mapped to the reference human genome version hg19 (2009–02 release, http://genome.ucsc.edu/). Nucleotide changes observed in aligned reads were called and reviewed using the NextGENe^® software (SoftGenetics, State College, Pennsylvania, United States). In addition to the detection of deleterious mutations and novel single nucleotide variants, the coverage-based algorithm eCNVscan, developed in-house, was used to detect large exonic deletions and duplications. The normalized coverage depth of each exon of a test sample was compared with the mean coverage of the same exon in the reference file to detect copy number variants.

Sequence variants were annotated using population and literature databases, including 1000 Genomes, the Single Nucleotide Polymorphism database, the Genome Aggregation Database, ClinVar, the Human Gene Mutation Database, and OMIM. Online software was used to analyze the structure of the protein, predict the conservation and function domains, and perform the multiple sequence alignment. The variant interpretation was manipulated according to the American College of Medical Genetics (ACMG) guidelines (Allyse and Wick, 2017). Molecular data on databases (https://www.chd7.org) for ClinVar and CHD7 were also collected and collated. Pearson’s correlation analysis was used to analyze the correlation of the exons, phenotypes, and the Acute Physiology and Chronic Health Evaluation (APACHE II) score, where p < 0.05 indicated that the result was statistically significant.

The results of our study are detailed in Table 1 and summarized in Table 2. A total of 12 patients conformed to the diagnostic criteria for CS with CHD7 gene variants, comprising six male and six female patients, including ten infants. The maximum age was 16 years, and the minimum age was 2 days.

Of the 12 patients, there were 11 cases with respiratory malformations and manifestations, including laryngeal or tracheal–bronchial malacia (8/12), neonatal pneumonia (4/12), nostril atresia (1/12), and respiratory distress (8/12). The patients were admitted for respiratory distress syndrome. The main cardiovascular malformations (10/12) were patent ductus arteriosus (5/12), atrioventricular septal defect (4/12), persistent arterial trunk (1/12), and pulmonary hypertension (2/12). There were 9 cases of central nervous system malformations, 1 of which was found on a prenatal MRI. Patient 11 had congenital polydactyly malformations of the left hand, and Patient 7 had a persistent arterial trunk phenotype, which may be a rare phenotype of the CHD7 gene. Of the 12 patients, 7 patients with developmental delay; 6 patients with an abnormal appearance, manifested as a short neck, right webbed neck, low hairline, micromaxillary deformity, and a skewed mouth angle; 12 patients with internal (1/12) and external (8/12) ear and hearing abnormalities (5/12); and 3 patients with digestive tract abnormalities, including 2 patients with abnormal pharyngeal and palatine arches and another patient with esophageal atresia. There were 3 patients with abnormal vision, including two diseases of the retina and choroid; 2 patients had abnormalities in the urogenital system that showed cryptorchidism and hydronephrosis (Table 1).

Abnormal antenatal examinations were present in ten of the 12 patients. In addition to the four premature infants, there were six low-birth-weight babies, with the lowest being 1.700 kg. The mean birth weight of the 12 patients was 2.62 kg. Abnormalities were found in the low birth weight and premature infants, but most were not examined further. A head MRI of Patient 2 at 28 + 5 weeks’ gestation suggested a slight widening of the left ventricle. Patient 4 had low progesterone in the first trimester and gestational diabetes in the third trimester. Polyhydramnios was detected in Patient 5 in the third trimester. Patient 7 was born prematurely at 35 + 6 weeks after her fetal heart malformation was detected by ultrasound during her mother’s pregnancy without intervention or further examination. In the second trimester, the mother of Patient 8 caught a cold, was not on medication, and no abnormality was found in the late pregnancy examination. Patient 10’s mother tested for prenatal thalassemia. Patient 11’s mother presented with prenatal oligohydramnios. The mothers of patients 6, 9, and 12 had a history of convulsions of hands and feet during pregnancy.

In the patients we studied, there were some variations in the diagnostic criteria. Laryngeal or tracheal-bronchial malacia, lung infection, heart malformation, and central nervous system malformations were the most common phenotypes. The proportion of developmental delay was also higher than the clinical phenotypes in the minor criteria. Electrolyte disturbances were present in five patients, four of whom presented with hypocalcemia. Atopic disorders were present in two patients (Tables 1 and 2). There were four patients with endocrine abnormalities, including hypogonadotropism, hypothyroidism, hypoparathyroidism, and low gonadal hormone secretion. In addition, two patients had an abnormal immune function, one of whom was diagnosed with DiGeorge syndrome.

We assessed the severity of the disease by retrospectively analyzing the APACHE Ⅱ scores at admission. Scores ranged from 0 to 19, with a mean of 11.58. Although only 12 patients were included in the study, we found that the younger the age, the higher the critical illness score, indicating the severity of the disease. Patients 4–11 were in a critical condition with APACHE Ⅱ scores >12. They were treated with non-invasive or invasive ventilators at admission. However, Patients 8–10 did not respond well to the treatments, and the guardian finally gave up treatment and discharged them. The other patients are still alive, but most have abnormal respiratory tract development, leading to recurrent lung infections.

Sanger sequencing after genomic DNA was extracted from 12 patients, and these variants interpretation was manipulated according to the American College of Medical Genetics (ACMG) guidelines. All patients were found to be CHD7 heterozygous mutation positive and these variants were de novo. There were two cases of splicing mutation: c.6936+2T > C (Figure 1) and c.8077-2A > G. The two splicing mutations and c.5405-3C > G were located in introns (Table 3; Figure 2). There were five cases of nonsense mutation: c.253C > T (p.Q85*), c.7012C > T (p.Q2338*) (Figure 3), c.7252C > T (p.R2418*), c.6070C > T (p.R2024*), and c.5428C > T (p.R1810*). There were four cases of frameshift mutation: c.6018dup (p.S2007Ifs*2), c.7868delC (p.P2623Rfs*16) (Figure 4), c.4667dupC (p.R1557Kfs*16), and c.780del (p.S261Lfs*44). Except for c.253C > T (p.Q85*) (PubMed reference number [PMID] 21258681), c.6018dup (p.S2007Ifs*2) (PMID 20884005), c.7252C > T (p.R2418*) (PMID 21258681), c.6070C > T (p.R2024*) (PMID 15300250), c.4667dupC (p.R1557Kfs*16) (PMID:23024289), c.780del (p.S261Lfs*44) (PMID 21258681), and c.5428C > T (p.R1810*) (PMID 16400610), the others were confirmed as novel mutations of the CHD7 gene.

Alamut was used to predict the effect of each mutation on protein translation. The splicing and protein function predicted by Alamut may have been affected by c.6936 + 2T > C and c.8077-2A > G, while c.5405-3c > G was less likely to affect the splicing.

Figure 1 shows Sanger sequencing results for Patient 4. The chromatographs show that a heterozygous mutation of the CHD7 gene was detected: a mutant c.6936 + 2T > C was a splicing mutation of intron, and the patient’s parents did not have this mutation. We mapped the 12 variants into the structure of the CHD7 gene (Figure 5 and Table 4). On the transcriptional protein domain, one variant was located in the chromodomain, two were located in the SANT domain, and the others were distributed in the remaining exon region. Exons 33, 36 and Introns 25, 32, 37 had one new mutation each, and we found two mutation sites in exon 36.

The scatter plot showed that exon location was positively correlated with phenotypes and the APACHE Ⅱ score (exon site vs. phenotypes: R = 0.58, p = 0.0487; exon site vs. APACHE Ⅱ score: R = 0.65, p = 0.0233). There was also a positive correlation between the number of phenotypes and the APACHE Ⅱ score (R = 0.38, p = 0.219) (Table 4; Figure 6).

According to data analysis from the ClinVar and CHD7 gene databases, a total of 1,806 variants are CHD7 gene mutations, where 1,054/1,806 are pathogenic/likely pathogenic variants that mainly consisted of frameshift (462/1,054), nonsense (327/1,054), missense (155/1,054), and splicing (120/1,054) mutations, while missense (613/768) mutations had the highest probability of benign/likely benign or uncertain significance.