Feeding difficulty is the dominant feature in 12 Chinese newborns with CHD7 pathogenic variants

We pooled data from the Neonatal Birth Defects Cohort (NBDC, ClinicalTrials.​gov Identifier: NCT02551081) in Children’s Hospital of Fudan University during 2016–01 to 2018–11. This data included 11,572 whole exome sequencing, 13,636 clinical exome sequencing and 1284 whole genome copy number microarray analyses. In this retrospective study, enrolled newborns met one of the following four criteria: 1) a previously established heterozygous pathogenic variant in the CHD7 gene, with data obtained from both the public database (HGMD and ClinVar) and internal database; 2) the same amino acid change as a previously established pathogenic variant; however, different nucleotide changes were accepted; 3) a novel (both public database and internal database) heterozygous null variant (nonsense, frameshift, canonical +/− 1 or 2 splice sites, initiation codon, single or multiexon deletion) in CHD7 gene; and 4) a novel and de novo heterozygous variant with negative family history or inherited from the affected parents. All variants were classified according to ACMG guideline [12]. Patients were excluded if pathogenic copy number variants were identified using array-based comparative genomic hybridization (arry-CGH). Patients’ information was obtained from clinical records. A clinical diagnosis of CHARGE syndrome was made based on the Bergmann criteria [10], Blake criteria [2], Verloes criteria [3] and Hale criteria [11].

The criteria for genetic testing were approved by ethics committees of Children’s Hospital, Fudan University (2014–107). Pretest counseling was performed by physicians. Informed consent was obtained from the patient’s parents. High-throughput sequencing was performed according to standard protocols in Clinical Laboratory Improvement Amendments (CLIA: 99D2064856) compliant sequencing laboratory in Wuxi NEXTCODE (China). Sequences were generated using the Agilent ClearSeq Inherited Disease Kit, Illumina Cluster and SBS Kit. Next generation sequencing was performed on an Illumina HiSeq 2000/2500 platform. The detected variants were confirmed using PCR and PCR-amplified DNA products were subjected to direct automated sequencing (3500XL Genetic Analyzer, Applied Biosystems) according to the manufacturer’s specifications. De novo variants were detected by parental evaluating via Sanger sequencing.

Among the 12 affected neonates, 8 had T1- and T2-weighted MRI head scans from the picture archiving and communication system (PACS) in Children’s hospital of Fudan University. We matched these affected neonates with 92 control neonates of the same corrected gestational age and excluded 4 patients because of image registration failure caused by poor image quality. All digital imaging and communications in medicine (DICOMs) were concatenated into a NIfTi volume format by dcm2niix software [13]. Then, the brain images of the neonates were extracted through FSL BET based on the T2-weighted modality followed by N4 bias correction [14]. The skull-striped T2 images were normalized to the neonate-specific T2 weighted image template at 44 weeks corrected gestational age [15] using the SyN registration method in ANTS [16]. The T2 brain template included a parcellation atlas with 87 regions of interest (ROI). The volumes of each ROI in the brain were measured by the sum of the log relative Jacobian determinant from the nonlinear deformable field of the registration. The total brain volume was used as a nuisance variable for regression, and the t statistic map was constructed from the linear regression as follows: ROI_volume ~α variants+ β corrected_age + θ gender+ 1.

A total of 12 pathogenic/likely pathogenic variants in the CHD7 gene were identified by next generation sequencing in 12 unrelated newborns. All enrolled patients were from nonconsanguineous couples from the Chinese Han population. All patients’ family histories were negative. The clinical manifestations of each patient are shown in Table 1. The Bergmann criteria focuses on patients with suspected features of CHARGE syndrome with CHD7 analysis. According to Bergmann’s criteria, Patient 2, 6, 7, 8, 9, 10, 11, and 12 matched the guidelines and needed further CHD7 analysis. Patients 7, 8 and 11 matched two cardinal and one supportive criteria and need CHD7 analysis including MLPA. Patient 2, 6, 9 and 10 matched one cardinal criteria and one or more than one supportive criteria. According to Bergmann’s criteria, the four patients needed a temporal bone CT first to detect typical semicircular canal abnormalities. The Blake criteria has four major criteria (coloboma, choanal atresia/stenosis, characteristic ear anomalies and cranial nerve dysfunction) and seven minor criteria. Definitive CHARGE syndrome should match 4 major or 3 major features and 3 minor features. Probable or possible CHARGE is defined as 1 or 2 major features and several minor features. Based on the Blake criteria, 8 patients (2, 3, 6, 7, 8, 10, 11 and 12) were diagnosed with probable or possible CHARGE. The major criteria of Hale proposed in 2016 were coloboma, choanal atresia or cleft palate, ear abnormalities and pathogenic CHD7 variants. Patients matching two major criteria and any number of minor criteria are diagnosed with CHARGE syndrome. In this study, four patients (2, 3, 7 and 11) met Hale’s criteria for CHARGE syndrome. Major Verlos criteria include coloboma, choanal atresia/stenosis and hypoplasia/aplasia of the semicircular canals. Patient matching as least two major criteria can be considered to have CHARGE syndrome. As no patient in our cohort was reported with choanal atresia/stenosis and abnormalities of the semicircular canals, no patient could be diagnosed with CHARGE syndrome based on the Verloes criteria. In this population, some typical features of CHARGE syndrome were observed, including aplasia/dysplasia of the semicircular canals, cleft lip/palate and choanal atresia.

Twelve pathogenic/likely pathogenic variants in the CHD7 gene were identified. We detected 6 reported pathogenic variants and 6 novel variants, including 4 frameshifts, 4 stop-gain, 2 splice-donor region, 1 intron variant and 1 missense variant (Table 2). Stop-gain and frameshift variants accounted for 67% of variants in this study. None of these variants are included in the gnomeAD database (http:// gnomad.​broadinstitute.​org/), the 1000 gnommeAD database (http://​gnomad-old.​broadinstitute.​org/​) or our internal database (1833 probands and 6893 families). Among all six novel variants, Glu2408* and Lys651* (NM_017780) were identified as de novo variants by Sanger sequencing. We mapped the 12 variants into the structure of the CHD7 protein (http://​www.​ebi.​ac.​uk/​interpro/​). Two variants were located in the chromodomain, two were located in the SANT domain, one in the ATP-binding domain belongs to the Helicase superfamily (Helicase N domain) and one was in the BRK domain. We did not detect CHD7 copy number variants in our study.

MRI analysis revealed significant volume loss in the cingulate gyrus, occipital lobe, and cerebellum and volume gain in the left medial and inferior temporal gyri anterior white matter parts among neonates with the CHD7 variant (Fig. 1). In addition, different subregions showed different levels of volume loss among these three regions (Table 3).