Nexus between genome-wide copy number variations and autism spectrum disorder in Northeast Han Chinese population

We enrolled 16 individuals with ASD aged 2 to 7 years from the Chunguang Rehabilitation hospital in Jilin Province, after cases with fragile X syndrome, Rett syndrome, chromosomal abnormalities, or any neurological or psychiatric disorders were excluded. The individuals with ASD were diagnosed by Pediatric Neurology and Neurorehabilitation doctors using the Diagnostic and Statistical Manual of Mental Disorders (5^th edition) [32]. All the individuals with ASD were northeast Han Chinese.

Genomic DNA was extracted from peripheral blood samples using DNA extraction kits, according to the manufacturer’s instructions (DP319 TIANamp Blood DNA Kit, TIANGEN BiotechCo. Ltd, Beijing, China) [33]. We used Nano Drop (Cat#ND-1000, ThermoFisher, Waltham, MA, US) and 1% agarose gel electrophoresis to check the quantity and quality of the isolated DNA. We used aCGH for genome-wide CNVs screening (Agilent SurePrint G3 Human CGH 60 K). Male and female DNA samples were hybridized with male and female reference DNA samples (G1471, G1521, Promega), respectively.

We converted the raw data using FEATURE EXTRACTION software 10.7 and analyzed CNVs using Agilent CytoGenomics software 4.0.3.12 (Agilent technologies, Santa Clara, CA, US). The human genome assembly NCBI36/hg18 was used as a reference. The analysis settings for CNVs calling were Aberration Detection Method 2 algorithm, centralization threshold 6, bin size 10, and minimum number of adjacent probes 3. Thresholds were set via log2-ratio (log₂ ^R) (for detecting duplications, log₂ ^R ≥ 0.25; for detecting deletions, log₂ ^R ≤ -0.25).

We calculated the frequency of each overlapping or non-overlapping CNV in DNA samples from our subjects. CNVs with same overlapping sequence were defined as one kind of CNV, and a non-overlapping CNV was also sorted as one kind of CNV. The circular plot of CNVs distribution in chromosome was visualized using circlize package in R3.6.2 software [34]. We converted our bed file from exon coordinates for human build NCBI36 (hg18) into GRCh37 (hg19) using UCSC LiftOver tool (http://​genome.​ucsc.​edu/​cgi-bin/​hgLiftOver). The classification of CNVs was based on Database of Genomic Variants (DGV, http://​dgv.​tcag.​ca/​dgv/​app/​home), Database of Genomic Structural Variation (dbVar, https://​www.​ncbi.​nlm.​nih.​gov/​dbvar), Clinical Genome Resources (ClinGen, https://​clinicalgenome.​org/​), and Online Mendelian Inheritance in Man (OMIM, https://​www.​ncbi.​nlm.​nih.​gov/​omim). CNVs were classified as benign, likely benign, a variant of unknown significance (VOUS), likely pathogenic, and pathogenic using AnnotSV program (https://​lbgi.​fr/​AnnotSV/​) according to American College of Medical Genetics guideline [35].

CNVs were considered of strong putative interest when they reached the following criteria: (1) they were classified as likely pathogenic or pathogenic; (2) they were of large size (> 100 kb); (3) they had been found in the knowledgebases for the genetic evidence of ASD (Simons Foundation Autism Research Initiative [SFARI, https://​www.​sfari.​org/​resource/​sfari-gene/​], or AutismKB [http://​www.​autismkb.​com]); (4) they had been found in the Database of genomic variation and phenotype in Humans using Ensembl Resources (DECIPHER, https://​decipher.​sanger.​ac.​uk/​about#overview); and (5) they contained previously reported ASD-relative genes.

We selected potential pathogenic genes within potential pathogenic CNVs on the basis of the following criteria: (1) genes enriched in ASD-related pathways; and (2) same genes shared with 363 genes in SFARI classified as high-confidence or strong-candidate, or with 228 genes in AutismKB classified as high-confidence.

MicroRNAs (miRNAs) are involved in the pathogenesis of ASD [30, 36]. Because genes implicated in CNVs that we found encode miRNAs, we further selected potential-pathogenic-CNVs-encoded miRNAs by retrieving PubMed according to experimental evidence documenting nervous system dysfunction.

The Gene Ontology (GO) and KEGG pathway analyses of the genes from potential pathogenic CNVs were performed using clusterProfiler package in R3.6.2 software [37, 38]. P-value < 0.05 was considered statistically significant. miRWalk 2.0 database, which contained 12 miRNA-target-prediction database, was used to predict target genes of CNVs-encoded miRNAs [39]. We selected the target genes according to the criteria—target genes existed in at least seven of the 12 databases. Moreover, interactive relationship between CNVs-encoded miRNAs and target genes was presented using Cytoscape 3.8.0 (http://​www.​cytoscape.​org/​).

To detect CNVs, aCGH was performed in all DNA samples from the 16 subjects with ASD (13 males and 3 females). We identified 364 CNVs (153 deletions and 211 duplications) with an average genomic size of 211.982 kb (114.091 kb for deletions and 258.705 kb for duplications). The mean number of CNVs per subject was 22.750 (9.563 for deletions and 13.188 for duplications). The mean number of deletions in male (10.462) was greater than that in females (5.667) (Table 1).

A total of 20 CNVs from 364 CNVs failed to be converted to GRCh37 (hg19); thus, we obtained 72 benign, 65 likely benign, 9 VOUS, 167 likely pathogenic, and 31 pathogenic CNVs (Table 2). We found that more than half CNVs were likely pathogenic or pathogenic.

After we calculated the frequency of each overlapping or non-overlapping CNV in DNA samples from our subjects, 344 CNVs were converted into 115 kinds of CNVs (45 deletions and 70 duplications). All the 115 kinds of CNVs were further classified (benign: 13 kinds; likely benign: 18 kinds; VOUS: two kinds; likely pathogenic: 60 kinds; and pathogenic: 13 kinds) (Supplementary Table 1). The distribution of the 115 kinds of CNVs in chromosome is visualized by circular plot (Fig. 1).

We investigated SFARI, AutismKB, and DECIPHER database to identify potential pathogenic CNVs from the 115 kinds of CNVs, revealing that 22 kinds of CNVs (6 deletions and 16 duplications) were potentially pathogenic. The 22 kinds of CNVs were distributed in chromosome 1, 2, 5, 7, 10, 11, 12, 14, 15, 16, 22, and X. Among them, 19 kinds of CNVs were rare (Table 3).

A total of 511 genes from the 22 potential pathogenic CNVs were functionally annotated by GO. The annotated genes were classified into three GO domains (biological processes [BP], cellular component [CC], and molecular function [MF]). For BP, some gene sets were enriched in synaptic-related functions, including modulation of chemical synaptic transmission (GO: 0050804), regulation of trans-synaptic signaling (GO: 0099177), positive regulation of excitatory postsynaptic potential (GO: 2000463), positive regulation of synaptic transmission (GO: 0050806), chemical postsynaptic transmission, (GO: 0099565), modulation of excitatory postsynaptic potential (GO: 0098815), and regulation of postsynaptic membrane potential (GO: 0060078), and in central nervous system related functions (positive regulation of neurological system process [GO: 0031646]). For CC, the top five CC terms included keratin filament (GO: 0045095), myelin sheath (GO: 0043209), Golgi lumen (GO: 0005796), glutamatergic synapse (GO: 0098978), and neuron to neuron synapse (GO: 0098984). For MF, the top five MF terms encompassed catecholamine binding (GO: 1901338), dopamine binding (GO: 0035240), magnesium ion binding (GO: 0000287), insulin receptor binding (GO: 0005158), and lipase activity (GO: 0016298). The top 20 GO functions are presented in Fig. 2 and Supplementary Tables 2, 3, and 4.

KEGG pathway enrichment analysis showed enriched key pathways, such as dopaminergic synapse (hsa04728), mTOR signaling pathway (hsa04150), insulin signaling pathway (hsa04910), and cholinergic synapse (hsa04725). The top 20 pathways are presented in Fig. 2 and Supplementary Table 5.

We constructed intersections among 511 genes that we found, 363 high-confidence or strong-candidate risk genes of ASD reported in SFARI database, and 228 high-confidence risk genes related to ASD reported in AutismKB database (Fig. 3). After investigating genes in the intersections, we found that cholinergic receptor nicotinic alpha 7 subunit gene (CHRNA7) was involved in the regulation of excitatory postsynaptic potential and cholinergic synapse; dopamine receptor D4 gene (DRD4) was involved in the regulation of synaptic transmission, dopamine binding, and glutamatergic synapse; HRas proto-oncogene (HRAS) played roles in the regulation of excitatory postsynaptic potential, glutamatergic synapse, and mTOR signal pathway; oligophrenin 1 gene (OPHN1) correlated with regulated synaptic signal, ionic glutamate receptor binding, and glutamatergic synapse; phosphatase and tensin homolog (PTEN) was implicated in the regulation of synaptic signal, neuron differentiation of central nervous system, ionic glutamate receptor binding, sphingolipid signaling, and mTOR signaling; SH3 and multiple ankyrin repeat domains 3 gene (SHANK3) was involved in the regulation of synaptic signal, ionic glutamate receptor binding, neuronal synapse, postsynaptic density, and asymmetric synapse; solute carrier family 6 member 3 gene (SLC6A3) played roles in dopamine binding, neurotransmitter: sodium cotransporter activity, and neurotransmitter transport activity; and TSC complex subunit 2 gene (TSC2) was involved in synapses, postsynaptic density, asymmetric synapses, and mTOR signaling pathways. Scores of all these genes (CHRNA7, DRD4, HRAS, OPHN1, PTEN, SHANK3, SLC6A3, and TSC2) in AustismKB and corresponding ranks in SFARI are listed in Table 4. DRD4, HRAS, OPHN1, SHANK3, SLC6A3, and TSC2 were in the regions of CNVs duplication. CHRNA7 and PTEN were in the regions of CNVs deletion.

We found 50 potential-pathogenic-CNVs-encoded miRNAs (45 encoded by duplication regions and 5 encoded by deletion regions). According to experimental evidence documenting nervous system dysfunction, we retrieved PubMed, identifying that 12 CNVs-encoded miRNAs were previously reported to be associated with brain or nervous system dysfunction (Table 5).

We intersected CNVs-encoded-miRNAs-targeted genes predicted using miRWalk 2.0 database with the union between SFARI and AutismKB (Supplementary Fig. 1). A total of 219 target genes were chosen for further study. We presented the interaction networks between CNVs-encoded miRNAs and 219 target genes (Figs. 4 and 5). The CNVs-encoded miRNAs and target genes are presented in Supplementary Tables 6 and 7.

We further investigated potential functions of the 219 target genes using GO analysis. For BP, some gene sets were enriched in synaptic-related functions, including synapse organization (GO: 0050808), modulation of chemical synaptic transmission (GO: 0050804), regulation of trans-synaptic signaling (GO: 0099177), synaptic transmission, glutamatergic (GO: 0035249), postsynaptic density organization (GO: 0097106), postsynaptic specialization organization (GO: 0099084) and regulation of glutamatergic synaptic transmission (GO: 0051966), and in central nervous system related functions, including learning or memory (GO: 0007611), cognition (GO: 0050890), and neurotransmitter transport (GO: 0006836). For CC, some gene sets were enriched in synaptic-related cellular components, including synapse membrane (GO: 0097060), postsynaptic specialization (GO: 0099572), and neuron to neuron synapse (GO: 0098984). For MF, some gene sets were enriched in ion-gated channel activity (GO: 0022839), gated channel activity (GO: 0022836), ion channel activity (GO: 0,005,216), ionotropic glutamate receptor activity (GO: 0004970), and transmitter gated channel activity (GO: 0022835). The top 20 GO functions are presented in Fig. 6 and Supplementary Tables 8, 9, and 10.

KEGG pathway enrichment analysis showed enriched key pathways, such as glutamatergic synapse (hsa04724), dopaminergic synapse (hsa04728), and Wnt signaling pathway (hsa04310). The top 20 pathways are presented in Fig. 6 and Supplementary Table 11.