Linking the genome-wide associated genes of ASD and SCZ
Using the GWAS Catalog27, we searched the database for "Autism Spectrum Disorders" and identified the highly associated genes when collecting the data from 17 studies.We included the studies from diverse/different ethnic groups/ancestry as plotted in Supplementary Fig. 1A. A substantial number of studies were from European ancestry followed by African and East Asian ancestry.
Figure 1A shows the genes of the significantly associated variants with ASD that appeared in publications with a p-value of < 0.01; The genes are presented in a word cloud. The frequency of genes (the number of publications in which the gene was reported) was denoted by the size of the font. We then checked which of the genes that showed associations with ASD are also associated with SCZ using the GWAS Catalog as a reference database as analyzed in our previous study6. These genes have been marked in green in Fig. 1A.
Out of all 305 reported ASD genes in GWAS, 239 genes (78%) were also associated with SCZ, and 78 were associated with the 105 (74%) most frequently reported SCZ genes in GWAS (Fig. 1A). Remarkably, 74% of the most associated genes in ASD are also associated with SCZ (Supplementary Fig. 2A). The eight most frequently reported genes (MAD1L1, CACNA1C, SORCS3, TCF4, ITIH3, STAG1, TSNARE1, MPHOSPH9) were reported in more than three publications and are also associated with SCZ (Supplementary Fig. 2A). Some of the most associated genes are linked to neuron function. These include subunits of calcium channels (CACNA1C, CACNB2, CACNA1l), TSNARE1 (a part of the SNARE complex), GRIN2A (a subunit of the NMDA receptor), MAD1L1 (Mitotic Arrest Deficient 1 Like 1), CACNA1I, SORCS3, SPRK2, ITIH3, and more.
We then examined the functional linkage of the 23 genes with the highest prevalence in ASD GWAS and visualized a gene network as shown in Fig. 1B. For this purpose, weused the STRING functional enrichment analysis.16 genes exhibited significant enrichment for "Self-reported educational attainment" (p = 4.3e-17), and 15 genes demonstrated analogous enrichment for "Behavior" (p = 3.9e-16). CACNA1C was found to be directly connected to other five genes (CACNB2, GRIN2A, NRGN, RBFOX1,ITIH3). Other genes like TCF4 and NT5C2 also formed a part of the network. Most of these genes are known to be associated with calcium-mediated activity of neural cells42. The common significant terms ascribed to these genes were “Intelligence”,“Mental or behavioral disorder biomarker”,“Risk-taking behavior” etc. The genes MAD1L1, TSNARE1, and FURIN formed another distinct network as they are part of secretory pathways including synaptic vesicle release and fusion. Some of the terms associated with these genes were “Risk-taking behavior”, "Intelligence”, and “Mental or behavioral disorder biomarker”.
Specific SNPs that are common between SCZ and ASD
Subsequently, we aimed to investigate whether the type of mutation determines the specific brain disorder (ASD vs. SCZ) in the 17 associated genes that were common between ASD and SCZ (Fig. 1C). For example, for the TCF4 gene (a transcription factor), nine SNPS are associated with SCZ only, one with ASD only, and five SNPs are common. Similarly, for the CACNA1C gene, seven SNPs are associated with SCZ only, two with ASD only, and four with both ASD and SCZ. Overall, more SNPs are associated with SCZ and common SNPs. There are also a few SNPs associated exclusively with ASD, possibly attributed to the comparatively fewer ASD GWAS conducted compared to SCZ. As mentioned earlier, ASD patients are known to have an increased risk of SCZ9 and Fig. 1C provides additional support for this assertion, as the majority of SNPs linked to ASD are also associated with SCZ.
We also mapped the regions in the different chromosomes that are associated with ASD-specific genetic variants (Suppl Fig. 3A) and both with ASD and SCZ (Suppl Fig. 3B). Chromosome 3 (10 SNPs), 7(9 SNPs), and 10 (9 SNPs) had the highest number of specific SNPs reported for ASD. Chromosome 7 and 10(9 SNPs each) had the highest number of common SNPs shared in ASD and SCZ. Overall 63 SNPs from 17 genes were common between ASD and SCZ out of total 82 SNPs from 23 genes reported in ASD (Suppl Fig. 3).
Common genes between ASD GWAS and Developmental Brain Disorder Database (DBD)
Moreover, we sought to investigate the overlap between our catalog of genes identified in ASD GWAS and those present in the DBD (Fig. 2A & Supplementary Fig. 2B).The DBD includes information obtained through exome and genome sequencing, chromosome microarray analyses, and copy number variation studies. It covers six different developmental brain disorders, including ASD and SCZ, and provides comprehensive phenotypic details that aid in the detection of pathogenic loss of function variants.We identified 30 genes that were shared between ASD GWAS genes and DBD genes. Notably, the top reported ASD GWAS genes CACNA1C, TCF4, STAG1, RBFOX1, and RERE were among this shared gene list.
We next used the RNA consensus tissue gene data to determine the brain regions that may be severely impacted by ASD from human protein atlas (HPA). With the overlapping gene list of genes shared between ASD GWAS and DBD, the average gene expression was computed for each tissue.The findings were represented using color-coded visualization, demonstrating the distinct brain regions influenced by ASD (Fig. 2B). The cerebellum exhibited the highest expression level, while the spinal cord displayed the lowest expression level, with a 37% decrease compared to the cerebellum.
Furthermore, we analyzed which of the genes were associated with autistic vs. schizophrenic behaviors as illustrated in Fig. 2C. 16 genes exhibited significant enrichment for "Autistic behavior" (p = 9.7e-16). In contrast, a more limited enrichment was found for "Schizophrenia symptom severity measurement" (p = 0.0003), with only three genes displaying significant associations. Thus, it appears that the resulting gene list is more likely to be associated with autistic behavior. In addition, network analysis revealed the presence of three distinct gene clusters. In particular, two smaller clusters, consisting of three and four genes, respectively, showed significant associations with autistic behavior. Conversely, the largest cluster, comprising 16 genes, showed pronounced associations with ASD and SCZ.
iPSC models and neuronal transcriptome in ASD
We further performed a meta-analysis of previously published work using iPSC models (see methods) (Suppl. Table 1) for ASD, as performed previously for SCZ6. Initially, we segmented the data to see the different ASD types (Fig. 3A). The majority of the data is classified as “ASD” (29.4%), while a large portion of the data is Rett Syndrome (27.5%), and the next largest group is fragile X (FXS 11.8%). Approximately 39.2% of the iPSC models had isogenic lines for controls (Fig. 3B). The cell type distribution from which the reprogramming was performed is shown in Fig. 3C. The vast majority of the studies started by reprogramming fibroblasts (52.9%). The reprogramming was usually performed with a retrovirus (50.9%), or a lentivirus (35.3%),or a Sendai virus (21.6%). In the context of ASD, most research is performed on neurons or neural progenitor cells (Fig. 3E). It is interesting to note that more than 25% of the ASD patients in the studies also have epilepsy (Fig. 3F). The distribution into neuronal types of the analyzed studies is shown in Fig. 3G.
Figure 4A provides a summary of the phenotypes that were found using the iPSC models. The studies report changes such as differentiation rate, neurites’ length, and differential gene expression. Out of 51 publications, 19.6% reported low synaptic and network activity in neurons derived through NPC from ASD patients. In addition, 11.8% reported shorter dendrites and neurites and 6% reported increasing associations with mitochondrial function in the neurons derived from the ASD patients. However, the direction is reversed when specifically examining cortical neurons. Five out of seven publications reported increasing synaptic and network activity and likewise, other publications could detect increasing spike frequency and increased activity in calcium imaging18,19,34,35. Differential gene expression analyses showed a trend toward increased regulation of genes18,35.
Figure 4B summarizes the functional phenotypes that are usually reported. The neurons derived from ASD patients usually start with an expedited maturation19,34 compared to the control neurons. This includes increased sodium and potassium currents, hyperexcitability, more arborized neurites, and even more synaptic connections initially. However, as the neurons mature, they lose their synaptic connections, have reduced currents, are less excitable, and are less arborized compared to the control neurons derived from individuals without the disorder19,35. Interestingly, the neurons derived from the SCZ patients have a different trajectory but in the later time points, they end up with a similar phenotype18. The SCZ neurons start less arborized, are less excitable, have decreased sodium and potassium currents, and have less synaptic activity compared to the control neurons. When the SCZ patient-iPSC-derived neurons develop, they always lag behind the control neurons and when they are mature18, many of their functional and morphological phenotypes are similar to the neurons derived from the ASD patients (Fig. 4B).