Trans-genetic effects of circular RNA expression quantitative trait loci and potential causal mechanisms in autism

Trans-genetic effects of circular RNA expression quantitative trait loci and potential causal mechanisms in autism | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Trans -genetic effects of circular RNA expression quantitative trait loci and potential causal mechanisms in autism Te-Lun Mai, Chia-Ying Chen, Tai-Wei Chiang, Trees-Juen Chuang This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-155043/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Genetic risk variants and transcriptional expression changes in autism spectrum disorder (ASD) have been widely identified, but their causal relationship is largely unknown. Circular RNAs (circRNAs) are abundant in brain and often serve as upstream regulators of mRNAs. By integrating RNA-sequencing with genotyping data from autistic brains, we accessed expression quantitative trait loci (eQTL) of circRNAs (circQTLs) that influenced expression of distant genes ( trans -eGenes) and constructed 43,372 circQTL- trans -eGene pairs. Mediation test suggested that 19,393 pairs were significantly cis -mediated by expression of circRNAs near the circQTLs; meanwhile, causal inference test (CIT) suggested 1,000 pairs influencing 708 trans -eGenes, wherein trans -eGene expression mediated trans -eQTL effects on ASD diagnosis. The 708 trans -eGenes were enriched for ASD risk genes and overrepresented in neurons in the upper neocortical layer. Integration of mediation test- and CIT-passing pairs further constructed 257 circQTL-circRNA- trans -eGene-ASD propagation paths. These findings increase our understanding of causative biology and cryptic regulatory mechanisms underlying ASD. Cognitive Neuroscience Medical Genetics Autism spectrum disorder Circular RNA circQTL trans-eQTL Regulatory network Causative biology Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Introduction Autism spectrum disorder (ASD) is a highly pervasive neurodevelopmental and heritable complex disorder, which are characterized by limited social communication, restricted and repetitive interests or behaviors 1 . Previous large-scale genomic studies have identified a variety of genomic variants associated with the etiology of ASD 2,3 , providing valuable biological insights into this disorder. However, the contribution of genetic factors to this complex disease is highly heterogeneous and not yet well understood. On the other hand, through high-throughput RNA sequencing (RNA-seq) data of postmortem brains from people with idiopathic ASD and non-ASD controls, numerous studies identified valuable differential expression patterns in mRNAs 4,5 , long non-coding RNAs 5 , microRNAs (miRNAs) 6,7 , RNA editing 8 , and circular RNAs (circRNAs) 9,10 . While these results increased our understanding of the underlying molecular mechanisms in ASD pathophysiology, the causal relationships between genetic sequence variants and transcriptional expression changes in ASD remains mostly unknown. By integrating genome-wide genotyping with transcriptome profiling, expression quantitative trait loci (eQTL) analysis identifies single nucleotide polymorphisms (SNPs) that affect the expression levels of local ( cis -eQTL) or distant ( trans -eQTL) genes, providing a connection between genetics and regulatory mechanisms of gene expression 11 . cis -eQTL SNPs often reside within or close to promoter regions of their nearby genes ( cis -eGenes) and mediate gene expression through directly affecting the corresponding transcription factor binding process 12 . In contrast to frequent investigation of cis -eQTLs, the regulatory mechanism underlying trans -eQTLs are less known because of the relatively weaker effects on gene expression for trans- actions than for cis- actions and the multiple-testing burden 13,14 . However, compared with cis -effects on gene expression, trans -genetic effects were reported to explain more than two times the variability in gene expression 14 and be more cell type-specific 15 . Some disease-causing variants involved in trans -effect analyses may miss in cis -effect analyses 16 . A recent study further indicated that at least 70% of complex trait heritability is driven by trans -eQTL effects 17 . A possible regulatory mechanism for trans -eQTLs is that trans -eQTLs indirectly affect the expression of distant genes ( trans -eGenes) through regulating the expression of genes near the trans -eQTLs ( cis -eGenes) 13,18,19 . In such cases, a trans -eQTL is also a cis -eQTLs, wherein the corresponding cis -eGene acts as a mediator of the trans -eQTL. Investigation of trans -genetic effects and the causal inference for the corresponding eQTL-associated networks will allow us to better understand complex trait genetics. CircRNAs are an emerging class of RNAs, which are formed by pre-mRNA back-splicing with a structure of covalently closed loops and endogenously expressed as single-strand, non-polyadenylated circular molecules 20,21 . Although circRNAs are generally expressed at a much lower level compared with their corresponding co-linear mRNA isoforms, they are more stable than other types of RNAs 20,22 . CircRNAs were observed to be especially abundant in the brain 23,24 and often play important roles in development of nervous system 24,25 . As for the layer of gene regulatory networks, circRNAs often serve as an upstream regulator of mRNAs. For example, they may regulate genes in trans through mediating the activities of miRNAs or RNA binding proteins (RBPs), with the common miRNA/RBP target site(s) of circRNAs and the target genes 20,21 . Hence, we are curious about whether SNPs that influence expression of nearby circRNAs (referred to as circQTLs) may also affect expression of remote genes, whereby circRNAs serve as an intermediate regulator bridging circQTLs and trans -eGenes. If this is possible, we can generate circQTL-circRNA- trans -eGene propagation paths in autistic brains and shed light on the functional consequences of the genetic loci and the etiology of this complex disease. There were two challenges to achieve the abovementioned goals. First, eQTL analysis requires transcriptomic and genomic data from the same samples with a large sample size. Second, large-scale circRNA study is often hampered by the limitation of the intrinsic circRNA characteristic of lacking polyadenylated tails. The newly released Synapse database 5,26 comprises a large human brain sample size of both genotyping data and RNA-seq data (rRNA-depleted RNAs from total RNAs without poly(A)-selection) from ASD cases and controls, offering an unprecedented opportunity for us to decipher the effects of genetic variation on circRNA expression in ASD brain. The data from the Synapse database thus allowed us to identify a new reference of circQTLs/ trans -eQTLs and ASD-associated trans -eQTL hotspots in the brains and infer cis -mediators of the identified circQTLs (i.e., such circQTLs are also trans -eQTLs) and causal effects between circQTLs and ASD diagnosis. Our findings help to unveil the gap between genetic variation and phenotypic changes in ASD and provides a framework for future analyses of eQTLs in neuropsychiatric disease or other complex traits. Results Identification of circQTLs in ASD brain We first collected rRNA-depleted paired-end RNA-seq data 5 and genotyping data 26 of postmortem samples from frontal cortex (FC) (Brodmann area 9), temporal cortex (TC) (Brodmann area 22, 41, and 42), and cerebellar vermis (CV) from individuals with ASD and non-ASD controls (the Synapse database) (Table 1). To identify local genetic effects that regulate circRNA expression, we considered the RNA-seq data and the corresponding genotyping data, both of which were derived from the same individuals (Fig. 1a, Table 1, and Supplementary Table 1). The examined circRNAs (1,060 circRNAs; Supplementary Table S1) were extracted from our previous study 10 based on the RNA-seq data used in this study. The expression levels of circRNAs were calculated using the number of supporting circRNA junction reads per million uniquely mapped reads (RPM) 27 and adjusted for covariates such as the corresponding host gene expression, brain region, diagnosis, and other biological/technical confounding factors (see Methods and Supplementary Table 1). For accuracy, the genotyped and imputed SNPs should be satisfied imputation score ³ 0.8 and estimated minor allele frequency ³ 0.01. To focus on cis -effects of circQTLs, we limited our analysis to SNPs in ± 200 kb nucleotides of each back-splice site. CircQTLs were evaluated by testing the correlations between the imputed genotype dosages and covariate-adjusted circRNA expression using Matrix eQTL 28 . P values were adjusted across all circQTLs using false positive rate (FDR) correction (Methods). Significant circQTLs were determined if FDR values were less than 0.05. We thus identified 989 circQTLs associated with 81 circRNAs (Supplementary Table 2). Recent transcriptomics studies of ASD showed that expression profiles of transcripts (including circRNAs 10 , miRNAs 6 , and mRNAs 29,30 ) and allele-specific genes 26 were highly similar between the two cortical regions (FC and TC) but were quite distinct in the CV. We calculated the circQTL effects for the 989 circQTLs in the three brain regions, respectively. We found that the circQTL effect size and direction were highly concordant between the two cortical regions and relatively weakly concordant (or even not concordant) between the cortex and CV (Fig. 1b). The differences in circQTL effect size between the cortex and CV were more significant than that between the two cortical regions (Fig. 1c). We also performed principal component analysis (PCA) and observed the similar results for the expression profiles of the 81 circRNAs (Fig. 1d). Previous studies have reported that the ASD pathophysiology was present in the cortex 31,32 and transcriptional dysregulation in ASD were shown to be consistently stronger in the cortex than in the cerebellum 5 . These observations suggest that the cortical regions are more selectively vulnerable to transcriptomic changes than the cerebellum. Therefore, we focused on the RNA-seq data and the corresponding genotyping data from the cortex (FC/TC) samples (105 samples from 53 ASD cases and 52 controls; Table 1) for the subsequent circQTL analyses. For accuracy, we only considered the identified circQTL SNPs (666 circQTLs associated with 48 circRNAs; Supplementary Table 2) in which the number of individuals with the minor allele (heterozygotes and minor allele homozygotes) was larger than three. We then studied the relationship between circQTLs and circRNA formation. We first showed that the identified circQTLs were preferentially located in the flanking sequences close to the back-splice donor/acceptor sites (Fig. 1e). The majority of circQTLs (90.5%; 603 out of 666; Supplementary Table 2) were located in the flanking sequences of back-splice sites. These observations were consistent with previous studies for distance distribution analysis of circQTLs in other human samples 33,34 . Next, back-splicing was reported to be facilitated by reverse complementary sequences (RCSs) residing in the introns flanking circularized exons 27,35,36 . We found that circQTLs residing in the flanking sequences of back-splice sites were significantly enriched in RCSs compared with background (empirical P <10 -4 ; see Methods) and non-circQTL SNPs ( P =3.2´10 -7 by two-sided Fisher’s exact test; Fig. 1f). Of note, the examined non-circQTL SNPs were also located in the flanking sequences of the examined back-splice sites and not determined as circQTLs. Moreover, it was shown that RNA binding proteins (RBPs) can regulate back-splicing by binding to the flanking sequences 20,21 . Indeed, the circQTLs located in the flanking sequences were significantly enriched in RBP binding sites compared with background (empirical P <10 -4 ) and non-circQTL SNPs ( P =0.01; Fig. 1g). These observations support that circQTLs located in the flanking sequences of back-splice sites may contribute to circRNA formation. Mediation effects of trans -eQTLs (circQTLs) via circRNA expression We next examined whether the identified circQTLs were also trans -eQTLs affecting the expression of distant genes ( trans -eGenes) using Matrix eQTL. By definition, such circQTL SNPs and trans -eGenes should be located on different chromosomes or the same chromosome separated by a distance greater than 5Mb (from the SNP site to the transcription start site of the trans -eGene). Since some SNPs in high linkage disequilibrium (LD) were associated with different trans -eGenes, non-pruned circQTLs were considered in this study. With controlling for potential biological/technical confounding factors and multiple testing correction (Methods), 546 circQTLs (associated with 47 circRNAs) were identified to be also trans -eQTLs, which affected the expression of 7,165 trans -eGenes and constructed 43,372 circQTL- trans -eGene pairs (Fig. 2a; Supplementary Table 3). Meanwhile, these circQTL- trans -eGene pairs also resulted in 13,379 circRNA- trans -eGene pairs, where the expression levels of circRNAs and trans -eGenes were both regulated through the same circQTLs (Fig. 2a). We then explored the causes for the trans -effect of circQTLs on the expression of trans -eGenes. In addition to the direct effect of circQTLs on trans -eGene expression, we hypothesized that in some cases the expression of trans -eGenes were indirectly regulated through the expression of circRNAs near the circQTLs (i.e., the “mediation effect”; Fig. 2a). For the scenario of mediation effects, the circRNAs served as cis -mediators of trans -eQTLs. We first showed that the expression levels of circRNAs and trans -eGenes in the 13,379 circRNA- trans -eGene pairs were more likely to be correlated with each other than expected (empirical P <10 -4 ; see Supplementary Fig. 1 and Methods). To test the mediation effects of trans -eQTLs, we performed mediation analyses 37 for all 43,372 circQTL- trans -eGene pairs to identify the proportion of association between a circQTL and the trans -eGene that was caused by the effect of the circQTL on the corresponding circRNA expression. We found that 19,393 out of the 43,372 (45%) circQTL- trans -eGene pairs passed the mediation test (Supplementary Table 3), which were significantly mediated ( P and FDR<0.05 with adjusting for related covariates; see Methods) by expression of circRNAs near the circQTLs. The majority (67%, 13,023 pairs) of the 19,393 pairs exhibited that more than 10% of the circQTL- trans -eGene association were mediated by the expression of circRNAs; in some cases (2,099 pairs), the proportion of mediation of the circQTL- trans -eGene association by the circRNAs was even more than 50% (Fig. 2b). Regarding the 19,393 circQTL- trans -eGene pairs (or 19,393 circQTL-circRNA- trans -eGene axes), we observed that the percentages of circQTL- trans -eGene pairs passing the mediation test markedly increased with increasing significance levels of trans -eQTL effects for circQTL- trans -eGene pairs (Fig. 2c). Such a percentage even reached more than 70% if FDR values of trans -eQTL effects were smaller than 10 -8 . We further examined the relationship between mediation effects and the significance levels of circRNA- trans -eGene correlation of expression profile. We found 55% of the circRNA- trans -eGene pairs passing the mediation test if the circRNA expression and trans -eGene expression were significantly (Spearman’s P <0.05) correlated with each other (Fig. 2d), whereas such a percentage was significantly reduced from 55% to 40% (odd ratio=0.56 and P <2.2´10 -16 by Fisher’s exact test) if the significant correlation between circRNA and trans -eGene expression disappeared (Fig. 2d). The percentages of the circQTL- trans -eGene pairs passing the mediation test markedly increased with increasing significance levels of the correlation of expression profile between circRNAs and trans -eGenes (Fig. 2d). Such a percentage even reached 90% at Spearman’s P value <0.0005. Similar results were observed in the analyses based on the CV samples (Supplementary Fig. 2 and Supplementary Table 3). These results reveal that the mediation effects of the circQTLs ( trans -eQTLs) on circRNA expression are correlated with the magnitudes of trans -eQTL effect and circRNA- trans -eGene correlation of expression profile. Causal effects between circQTLs and ASD diagnosis Regarding the identified 43,372 circQTL- trans -eGene pairs, we then conducted causal inference test 38,39 (CIT; see Methods) to examine whether the trans -effects of circQTLs on trans -eGene expression might explain the association between the circQTLs and diagnosis status (ASD vs. non-ASD) and infer the direction of association between circQTLs, trans -eGene expression, and ASD diagnosis (Fig. 3a, top). At P and FDR<0.05, we identified 1,000 circQTL- trans -eGene pairs with a propagation path from circQTL SNP to ASD diagnosis via trans -eGene expression (Fig. 3a, bottom; Supplementary Table 4). We speculated that the trans -eGenes (708 genes; CIT-passing trans -eGenes) involved in the 1,000 circQTL- trans -eGene pairs may be implicated in ASD. We examined enrichment analysis for genes previously implicated in ASD from Simons Foundation Autism Research Institutive (SFARI) 40 . We observed that the CIT-passing trans -eGenes were significantly enriched for the SFARI genes, but not for genes implicated in monogenetic forms of other brain disorders (Fig. 3b). These CIT-passing genes were also enriched for other classes of ASD-relevant genes, including genes encoding postsynaptic density (PSD) proteins 41 , genes whose transcripts were bound by the RBPs of FMR1 42 , RBFOX1 43 , and ELAVL1 44 , and differentially expressed genes (DEGs) in ASD 5 that were derived from the same cortex samples used in this study (Fig. 3b). In contrast, these trends were not observed in the trans -eGenes not passing CIT (Fig. 3b). Moreover, a previous study has presented a genome-wide prediction of ASD risk genes and provided an estimated probability of ASD-association for each gene 45 . On the basis of the probabilities of ASD risk, we observed the CIT-passing genes indeed had a significantly higher probability of ASD risk compared with background gene set (empirical P <0.001; Supplementary Fig. 3) and the trans -eGenes not passing CIT (all P < 10 -8 by two-sided Fisher’s exact test; Fig. 3c). We further found that the CIT-passing trans -eGenes were more intolerant of a loss function mutation (measured by gene variant intolerance (pLI) scores 46 ) compared with the background gene set (Supplementary Fig. 3) and the trans -eGenes not passing CIT (Fig. 3d), reflecting previous observations that genes implicated in ASD tended to be subject to stronger selective constraints than the other genes 45,47 . Regarding DEGs in a cell type-specific manner, the CIT-passing genes were most overrepresented in vasoactive intestinal polypeptide (VIP)-expressing interneurons and L2/3 and L4 excitatory neurons. This trend was generally consistent with the previous observation for SFARI genes 48 . Gene Ontology (GO) analysis revealed that the CIT-passing genes were enriched in GO terms related to chemical synaptic transmission, synaptic signaling, neurogenesis, synapse, neuron projection, axon, and so on (Fig. 3f and Supplementary Table 4), also reflecting their enrichment for PSD. These observations support the relevance of the CIT-passing trans -eGenes to ASD pathophysiology. Furthermore, the above mediation analysis has identified 19,393 circQTL-circRNA- trans -eGene axes with significant mediation effects of circQTLs on trans -eGene expression through circRNA expression (Fig. 2b and Fig. 4a, middle). By integrating the CIT-passing circQTL- trans -eGene pairs with the 19,393 circQTL-circRNA- trans -eGene axes, we determined 257 circQTL-circRNA- trans -eGene-ASD diagnosis propagation paths, where the circRNA- trans -eGene axes may act as causal mediators for the circQTL-ASD diagnosis associations (Fig. 4a, left; Supplementary Table 4). The 257 ASD-associated circQTL-circRNA- trans -eGene regulatory axes involved 33 circQTLs ( trans -eQTLs), 8 circRNAs, and 124 trans -eGenes (Figs. 4a and 4b). We found that 97 of 124 (78%) trans -eGenes were ASD-relevant genes and 217 of 257 (84%) circQTL-circRNA- trans -eGene axes were associated with the 97 ASD-relevant genes (Fig. 4b). Particularly, 10 of the 97 genes were SFARI genes and five were top SFARI genes with score = 1 or 2, suggesting the regulatory role of the corresponding circQTL-circRNA- trans -eGene axes in ASD brain. Potential mediators for the circRNA- trans -eGene interactions Regarding the 19,393 circQTL-circRNA- trans -eGene axes (associated with 5,669 circRNA- trans -eGene axes) passing the mediation test, we proceeded to identify potential mediators for the circRNA- trans -eGene interactions (Fig. 5a, left). Since circRNAs and mRNAs may carry the common miRNA/RBP target sites, numerous cases of circRNAs were demonstrated to act as an upstream regulator of mRNAs through mediating miRNA/RBP activities 20,21 . To this end, we utilized crosslinking immunoprecipitation (CLIP)-seq data-supported RNA interactomes (extracted from ENCORI 49 ) to search for the common miRNA/RBP target sites of the 5,669 circRNA- trans -eGene axes (see Methods). We thus identified 3,141 circRNA-miRNA- trans -eGene axes (associated with 958 circRNA- trans -eGene axes; Fig. 5a, middle) and 54 circRNA-RBP- trans -eGene axes (associated with 30 circRNA- trans -eGene axes; Fig. 5a, right), respectively (Supplementary Table 5). Regarding the 257 abovementioned circQTL-circRNA- trans -eGene axes that passed both the mediation test and CIT (see Fig. 4a, right and Fig. 5b, left), we further found that 90 axes (Fig. 5b, right) overlapped with the 958 circRNA- trans -eGene axes with common miRNA target sites of the circRNAs and trans -eGenes and no axis overlapped with the 30 circRNA- trans -eGene axes with common RBP target sites of the circRNAs and trans -eGenes. Accordingly, 158 circQTL-circRNA-miRNA- trans -eQTL axes that might be causally related to ASD diagnosis were generated (Figs. 5b and 5c; Supplementary Table 5). The 158 axes were associated with 28 circQTLs, 6 circRNA, 17 miRNAs, and 34 trans -eGenes (Fig. 5b, right). Of note, 28 of the 34 trans -eGenes were ASD-relevant genes and one ( ANK3 ) was a top SFARI gene with score = 1 (Fig. 5c). A previous study reported that genetic variations in ANK3 were associated with autism susceptibility 50 . ANK3 can regulate the structure and function of glutamatergic synapses 51 . Alteration of ANK3 expression can regulate neuronal microtubule dynamics functions and affect the corresponding role in psychiatric illness 52 . Taken together, we provided a framework for detecting potentially causal relationships between genetics and trait and generating the corresponding regulatory axes in ASD. Discussion The majority of eQTL studies to date have focused on investigation of cis -genetic effects on gene expression, although trans -eQTL effects may contribute most trait heritability 17 . One challenge for genome-wide scanning for trans -eQTLs is the heavy burden of multiple testing 13,14 . This study focused on detecting the circQTLs that cis -regulated expression of nearby circRNAs and also trans -regulated expression of distant genes simultaneously. We first identified circQTLs and then searched for trans -eGenes based on these circQTL SNPs (not all SNPs extracted), greatly decreasing the burden of multiple trans -eQTL testing for the associations between SNPs and gene expression. We thereby identified 546 circQTLs that were also trans -eQTLs and generated 43,372 circQTL- trans -eGene associations (Fig. 2a). Since circRNAs often serve as an upstream regulator of mRNAs 20,21 , we asked whether in some cases the effects of trans -eQTLs on trans -eGene expression were mediated by the circRNA expression. The mediation analysis suggested that 45% (19,393) of the 43,372 circQTL- trans -eGene associations were mediated by the expression levels of the corresponding circRNAs. Some mediation test-passing pairs exhibited that more than 50% of the circQTL- trans -eGene association were mediated by the expression of circRNAs, supporting the importance of the mediation effects for trans -eQTLs (Fig. 2b). In addition, we observed that circQTLs residing in the flanking sequences of back-splice sites were enriched in RCSs and RBP binding sites (Figs. 1f and 1g), both of which are important for circularization. These observations imply a possible regulatory mechanism for the circQTLs ( trans -eQTLs) passing the mediation test, wherein circQTLs affect the formation of their nearby circRNAs and the circRNAs, as an upstream regulator of the corresponding trans -eGenes, consequently regulate the trans -eGene expression. This is worth further investigation. To increase our understanding of the association between genetic variants and complex traits/diseases such as ASD, previous genome-wide association studies (GWASs) have provided valuable insights into the role of genetic etiology of ASD 2,53 . Recent transcriptome-wide association studies (TWASs) further integrated GWAS loci with gene expression data to investigate cis -effects on gene expression associated with risk loci for ASD 54,55,56 . However, TWASs do not clarify the major cause of the association between eQTLs and the expression of the corresponding eGenes because the direction of association between eGene expression and trait is unclear. This leads to three possible models (i.e., eGene mediation model, reverse causality model, and independent model) to explain the relationships between an eQTL, an eGene, and a trait 19,57,58 . For the first model, the eGene acts as a causal mediator for the eQTL-trait associations, wherein the eQTL regulates the trait through mediating the eGene expression. For the second model the trait acts as a causal mediator, wherein alteration of the eGene expression is the consequence of the trait. For the third model, the effects of eQTL on eGene expression and trait are independent. This study focuses on identification of eGenes that are causally related to idiopathic ASD (i.e., the first model). To this end, we utilized CIT 38,39 to examine the 43,372 circQTL- trans -eGene pairs and identified 1,000 pairs with trans -eGenes as a potentially causal mediator for ASD diagnosis (ASD vs. non-ASD) (Fig. 3a). By definition of CIT, a CIT-passing pair should simultaneously satisfy four conditions (see Methods), especially for the two conditions: circQTLs should be independent of ASD diagnosis after adjusting for trans -eGene expression and trans -eGene expression should be associated with ASD diagnosis after adjusting for circQTLs. Therefore, upon passing the CIT screening, both reverse causality and independent models are less probable for the explanation of the CIT-passing circQTL- trans -eGene associations. We proceeded to examine the trans -eGenes (708 genes) involved in the CIT-passing circQTL- trans -eGene pairs and found significant enrichment for ASD-relevant genes such as SFARI genes, genes encoding PSD proteins, targets of FMR1, RBFOX1, and ELAVL1, and DEGs in ASD, but not for genes implicated in monogenetic forms of other brain disorders (Fig. 3b). Like SFARI genes, the CIT-passing trans -eGenes had high probabilities of ASD risk (Fig. 3c), high pLI scores (Fig. 3d) and similar patterns of DEG distribution among cell types (Fig. 3e). Particularly, we showed that the CIT-passing trans -eGenes exhibited significant enrichment for DEGs in neurons in the upper neocortical layer such as IN-VIP, L4, and L2/3 (Fig. 3e). A previous mouse study had demonstrated that excess upper-layer pyramidal neurons can result in an excitatory-inhibitory neuron imbalance, leading to behavioral abnormality similar to the features in ASD 59 . L2/3 neurons are known to be important for underpinning high-level cognitive functions and implicated in cognitive syndromes of high-level associative dysfunction such as ASD 60 . These observations support the important role of these CIT-passing trans -eGenes in ASD susceptibility. Of the 43,372 circQTL- trans -eGene pairs (associated with 546 circQTLs), although the circQTL SNPs did not reached genome-wide significance in the largest ASD GWAS 2 currently, 217 of 546 (40%) circQTLs were associated with the expression of at least one ASD risk genes (SFARI genes; see Supplementary Fig. 4 and Supplementary Table 6). The 217 circQTLs resided at 40 distinct LD blocks, 133 of which (residing at 30 LD blocks) affected the expression of even more than 50 trans -eGenes. The 30 LD blocks can be regarded as trans -eQTL hotspots associated with SFARI genes. Of the 133 circQTLs, 25 circQTLs were particularly connected with ASD risk genes, in which the affected trans -eGenes were significantly enriched for SFARI genes (all P < 0.05 by hypergeometric test; Supplementary Table 6). This result provides a useful resource for further investigating regulatory mechanisms of the corresponding ASD risk genes. Intriguingly, we found that five trans -eGenes ( CRHR1 , XKR6 , NKX2-2 , NSF , and ARL17B ) were previously prioritized as TWAS-significant genes/transcripts in ASD in two recent studies 55,56 (Table 2). This revealed that these TWAS-significant genes/transcripts might be regulated by not only cis -acting SNPs but also trans -acting ones. Of note, NSF was also a trans -eGene involved in the CIT-passing rs182920567- NSF -ASD diagnosis propagation path, supporting that NSF may be causally related to ASD diagnosis. This reflects a previous report that NSF may play a role in ASD pathophysiology through mediating serotonin transporter trafficking 61 . In addition, some circQTL- trans -eGene associations passed the mediation test, wherein the effects on trans -eGene expression ( CRHR1 , XKR6 , and ARL17B ) were mediated by the circRNA expression (circERC1 and circSPHKAP). Two phenomenon were observed. First, the expression of a TWAS-significant gene may be regulated by multiple circRNAs. We found circERC1-1 and circSPHKAP with more than 50% (or even more than 70%) of mediation effects on the relations of trans -eQTLs to CRHR1 , suggesting multiple upstream regulators of CRHR1 . Second, two TWAS-significant genes/transcripts ( XKR6 and ARL17B ) may be under the same regulatory control (rs572758241-circSPHKAP association). Of note, rs572758241 was located at a trans -eQTL hotspot and affected the expression of 724 trans -eGenes. The 724 genes significantly overlap with SFARI genes ( P = 0.01 by hypergeometric test; Supplementary Table 6), indicating the relevance of the trans -eQTL to ASD. These results thus provided potential upstream regulators of these TWAS-significant genes/transcripts. By integrating the circQTL- trans -eGene pairs that passed CIT with the circQTL-circRNA- trans -eGene axes that passed mediation test, we further detected 257 potential circQTL-circRNA- trans -eGene-ASD diagnosis propagation paths (Fig. 4a). Meanwhile, we searched for the common miRNA target sites of the corresponding circRNAs and trans -eGenes and generated potential regulatory associations between the circRNAs and the trans -eGenes. Combined the predicted circRNA-miRNA-trans-eGene axes with the 257 potential circQTL-circRNA- trans -eGene-trait propagation paths, we finally identified 158 circQTL-circRNA-miRNA- trans -eQTL-trait propagation paths (Fig. 5b). Of the 158 axes, it is possible to further utilize the mediation analysis to examine whether the circRNA- trans -eGene associations were mediated by the expression of miRNAs, if both mRNA and miRNA expressing data derived from the same samples are available. For example, a circQTL-circRNA-miRNA- trans -eGene-trait propagation path that passes CIT and two layers of mediation test was detected according to miRNA expressing data 6 derived from the same ASD and control samples used in this study (Supplementary Fig. 5). On the basis of a large brain sample size of both RNA-seq data and the corresponding genotyping data from ASD cases and non-ASD controls, this study is the first report, to the best of our knowledge, for systematically investigating trans -genetic effects of circQTLs and inferring the corresponding causal relations in diseases. The identified ASD-associated regulatory interactions, particularly the CIT-passing axes involving trans -eGenes that were previously implicated in idiopathic ASD (e.g., SFARI genes), may serve as a useful resource for further investigations of more complex regulatory mechanisms underlying ASD. Future studies with larger patient cohorts, including genome-wide transcriptomic and genomic sequencing data, will allow for more precise causal relationships between genetic variants and transcriptional expression changes in ASD. The provided framework can be applied to future analysis of eQTL-trait relations in other complex diseases. Methods Identification of circQTLs/ trans -eQTLs. The RNA-seq data of ASD and non-ASD brain samples 5 and the corresponding genotyping data 26 were obtained from Synapse (http://www.synapse.org) with permission under the accession number syn4587609. The circRNAs were extracted from our previous study 10 , which were identified by NCLscan 62 according to the same Synapse RNA-seq data used in this study and the Ensembl annotation (version 90) based on the human reference genome (GRCh38). A sample was excluded in our analysis if the number of identified circRNAs of the sample was below one standard deviation from the mean of the sample set 10 . A total of 212 samples, including 73 FC, 61 TC, and 78 CV samples, were retained (Table 1). For accuracy, we only considered the 1,060 circRNAs that were detected in more than 50% of the samples examined. The genotyping data were derived from 79 individuals, including 40 ASD patients and 39 non-ASD-affected controls 26 . Since eQTL analysis required both RNA-seq and genotyping data from the same individuals, RNA-seq data from 165 samples were examined in the following analyses (Fig. 1a, Table 1, and Supplementary Table 1). The expression levels of circRNAs were calculated using the number of supporting circRNA junction reads per million uniquely mapped reads (RPM) 27 and adjusted for biological/technical covariates including the corresponding host gene expression, brain region, diagnosis (ASD/non-ASD), age, sex, RNA quality (RNA integrity number; RIN), post-mortem interval (PMI), sequencing batch, and brain bank (Supplementary Table 1). The expression levels of genes were downloaded from Parikshak et al.’s study 5 , which were measured by log 2 normalized fragments per kilobase million mapped reads (FPKM) values. Imputation genotyping data (39,166,259 SNPs) were downloaded using the Synapse command line client. We excluded SNPs with imputed quality score (R 2 ) < 0.8 and minor allele frequency (MAF) < 0.01 and thus retained 7,714,475 SNPs for the following eQTL analysis. We limited our analysis to SNPs in ± 200 kb nucleotides of each back-splice site and evaluated circQTLs by testing the correlations between the imputed genotype dosages and covariate-adjusted circRNA expression using Matrix eQTL with an additive linear model 28 for adjusting for covariates including sex, age, diagnosis, brain region, RIN, PMI, sequencing batch, brain bank batch. To control for multiple testing, P values will be adjusted across all circQTLs using false positive rate (FDR) correction (i.e., Benjamini-Hochberg correction). At FDR<0.05, we thus identified 989 circQTLs based on the 165 samples from all three brain regions and 666 circQTLs based on the 105 cortex samples, respectively (Supplementary Table 2). To compare the effect sizes for the 989 circQTL-circRNA pairs from different brain regions (TC, FC, or CV), we conducted a Z-test and calculated Z-scores as Here the equation showed an example for TC-FC comparison Z-test. SE represents standard error. We examined whether the 666 circQTLs were also trans -eQTLs affecting the expression of trans -eGenes using Matrix eQTL with an additive linear model for adjusting for the covariates abovementioned. Such circQTL SNPs and trans -eGenes should be located on different chromosomes or the same chromosome separated by a distance greater than 5Mb (from the SNP site to the transcription start site of the trans -eGene). At FDR<0.05, 546 circQTLs were identified to be also trans -eQTLs (Fig. 2a; Supplementary Table 3). Detection of reverse complementary sequences (RCSs). The criteria for detecting RCSs were similar to those presented in a previous study 63 . For a circRNA, both flanking sequences (± 200 kb nucleotides of the back-splice site) were aligned each other using BLAST 64 with parameters –task blastn –word_size 11 –strand minus. The potential RCSs should be simultaneously satisfied the following rules: bitscore> 100, alignment length>50 bp, and identity>80%. Determination of miRNA and RBP binding sites. CLIP-supported RBP binding sites on hg19 assembly were downloaded from ENCORI 49 at http://starbase.sysu.edu.cn/. The liftOver tool 65 was employed to obtain the genomic coordinates of RBP binding sites on the GRCh38 assembly. The miRNA-circRNA, miRNA-mRNA, RBP-circRNA, and RBP-mRNA interactions supported by CLIP-seq data were also extracted from ENCORI. These data were retrieved through ENCORI-provided Web APIs. The used command lines were listed in Supplementary Table 7. Of note, a considered miRNA (or RBP) binding site should be supported by at least one AGO CLIP-seq (or CLIP-seq) experiments. CircRNA-miRNA-mRNA and circRNA-RBP-mRNA interactions were determined according to the common miRNA and RBP target sites of the circRNAs and mRNAs, respectively. A developed package, CircMiMi, for constructing CLIP-seq data-supported circRNA-miRNA-mRNA and circRNA-RBP-mRNA interactions was also provided (https://github.com/TreesLab/CircMiMi). In CircMiMi, users can input their identified circRNAs, not limited to ENCORI-provided circRNAs. Enrichment analyses based on empirical sampling. A sampling procedure was conducted to test whether the 666 circQTL SNPs (associated with 48 circRNAs) had a higher proportion ( proportion obs ) of SNPs located in RCSs residing in the introns flanking circularized exons compared to an empirical distribution of the proportion observed in the 10,000 times of random sampling (see Fig. 1f). For each time, the equivalent number (666) of SNPs were randomly selected from the background set (823,378 SNPs) and the proportion ( proportion i ) of SNPs located in RCSs was calculated. The 823,378 SNPs were located in the flanking sequences (± 200 kb nucleotides of the back-splice site) for the 48 circRNAs examined. The empirical P value was calculated as A similar procedure was designed to test whether the 666 circQTL SNPs were predominantly located in RBP binding sites residing in the introns flanking circularized exons for the 48 circRNAs examined (Fig. 1g). Regarding the 43,372 circQTL- trans -eGene pairs (associated with 47 circRNAs, 7,165 trans -eGenes, and 13,379 circRNA- trans -eGene pairs; see Fig. 2a), we conducted a sampling procedure to test whether the 13,379 circRNA- trans -eGene pairs had a higher proportion of pairs with a significant correlation ( P <0.05 by Spearman’s correlation test) between expression of the circRNAs and the trans -eGenes compared to an empirical distribution of the proportion observed in the 10,000 times of random sampling. For each time, the equivalent number (13,379) of circRNA-gene pairs was randomly selected from the 47 circRNAs and the 7,165 trans -eGenes. To test whether the CIT-passing trans -eGenes (708 genes) had a significantly higher probability of ASD risk 45 than background gene set (20,070 protein-coding genes), an empirical gene enrichment analysis was conducted on the basis of an empirical distribution of the proportion observed in the 10,000 times of random sampling (Supplementary Fig. 3a). The 20,070 genes were selected because they were expressed in the cortex on the basis of the RNA-seq data examined in this study. The equivalent number (708) of genes was randomly selected from the 20,070 genes for each time. A similar procedure was conducted to test whether the CIT-passing trans -eGenes were more intolerant of a loss function mutation (measured by pLI scores 46 ) compared with the background gene set (Supplementary Fig. 3b). The probability of ASD risk and pLI score for each gene were downloaded from http://asd.princeton.edu/ and https://gnomad.broadinstitute.org/ (loss-of-function curation results; v2.1.1), respectively. Mediation and causal testing. We conducted mediations analysis using the mediation package 37 downloaded from the Comprehensive R Archive Network (CRAN) at https://cran.r-project.org/web/packages/mediation/index.html. To test the mediation effects of trans -eQTLs, the mediation package was performed with circQTL ( trans -eQTL) as the “exposure,” circRNA expression as the “mediator,” trans -eGene expression as the “outcome,” and adjusting for covariates including sex, age, diagnosis, brain region, RIN, PMI, sequencing batch, and brain bank batch. The proportion of mediation for the circQTL was the proportion of causal mediation effect in the total effect (i.e., causal mediation effect + direct effect) 37 . To test whether the circRNA- trans -eGene associations were mediated by the expression of miRNAs, we first extracted miRNA expressing data based on the 63 cortex samples (30 ASD and 33 non-ASD samples) overlapped with the samples examined in our eQTL analysis. The miRNA expressing data were obtained by the request from the authors of the Wu et al.’s study 6 , which were measured by log 2 normalized read counts with adjusting for library size, GC content, RIN, PMI, and batch bank. The mediation package was then performed with circRNA expression as the “exposure,” miRNA expression as the “mediator,” trans -eGene expression as the “outcome,” and adjusting for covariates abovementioned. Significant mediation effects were defined at a simulation threshold of P < 0.05 (sims=1,000) and FDR (Storey and Tibshirani method 66 ) < 0.05. The CIT package 38,39 was downloaded from CRAN at https://cran.r-project.org/web/packages/cit/index.html. A CIT-passing circQTL- trans -eGene pair (CIT P <0.05 and FDR ( q ) <0.05) (Supplementary Table 4) represents that the circQTL has a causal effect on the diagnosis status (ASD vs. non-ASD) through the trans -eGene expression, which should simultaneously satisfy the following conditions: (1) the circQTL is associated with the diagnosis status, (2) the circQTL is associated with the trans -eGene mediator after adjusting for the diagnosis status, (3) the trans -eGene mediator is associated with the diagnosis status after adjusting for the circQTL, and (4) the circQTL is independent of the diagnosis status after adjusting for the trans -eGene mediator. The P value of CIT is defined as the maximum of the above four-component test P values. The FDR ( q ) values were further estimated using CIT-provided permutation test with default parameters. Illustration of networks was plotted by the Cytoscape package (https://cytoscape.org/). Functional analysis of the CIT-passing trans -eGenes. The SFARI 40 gene list was downloaded from https://gene.sfari.org/ (SFARI 08-07-2020 release). The lists of genes encoding PSD proteins and targets of FMRP, RBFOX1, and ELAVL1 were downloaded from Lee et al.’s study 26 . The list of DEGs in ASD was downloaded from Parikshak et al.’s study 5 . The lists of genes associated with Alzheimer’s disease, amyotrophic lateral sclerosis, Huntington’s disease, intellectual disability, and Parkinson’s disease were downloaded from Wang et al.’s study 67 . The schizophrenia gene list was downloaded from a GWAS report 68 . The list of DEGs in a cell type-specific manner was downloaded from Velmeshev et al.’s study 15 . For gene set enrichment analyses in Figure 3b, we used the abovementioned 20,070 protein-coding genes as the background set. We took the analysis of SFARI gene enrichment for the CIT-passing trans -eGenes as an example. We created a two-way contingency table with rows containing numbers of SFARI and non-SFARI genes and columns containing numbers of CIT-passing trans -eGenes and genes that were not CIT-passing trans -eGenes. The statistical significance and odds ratio were determined using a one-tailed Fisher’s exact test with the fisher.test R function. P values were FDR adjusted across 13 target groups for each gene list using Bonferroni correction. GO analysis was performed using the ToppFun module of ToppGene Suite software 69 at https://toppgene.cchmc.org/enrichment.jsp (version 2020-Sep-08 01:39). False discovery rate (FDR) adjusted P values were calculated using Bonferroni correction. Hypergeometric testing. For a circQTL ( trans -eQTL), we conducted a hypergeometric test to evaluate the significance of overlap of the trans -eGenes regulated by the circQTL and SFARI genes. Only the circQTLs that affected the expression of more than 50 trans -eGenes were considered for the test (Supplementary Table 6). We used 20,070 protein-coding genes as the background set, which involved 951 SFARI genes. The statistical significance was determined using the phyper R function. The coordinates of LD blocks were downloaded from the DistiLD database 70 at http://distild.jensenlab.org. Data availability. The CircMiMi package for constructing circRNA-miRNA-mRNA and circRNA-RBP-mRNA interactions is available at https://github.com/TreesLab/CircMiMi. The identified circQTLs/ trans -eQTLs and the related data/codes were deposited in Supplementary Tables 1-7 or GitHub at https://github.com/TreesLab/circQTL_ASD. Declarations Acknowledgements This work was supported by Genomics Research Center (GRC), Academia Sinica, Taiwan; the Ministry of Science and Technology (MOST), Taiwan, under the contract MOST 108-2311-B-001-020-MY3 (TJC); and the Taipei, Taichung, Kaohsiung Veterans General Hospital, Tri-Service General Hospital, Academia Sinica Joint Research Program under the contract AS-VTA-109-03 (TJC). Author contributions T.-L.M. performed the bioinformatics assays, analyzed the data, and helped writing the manuscript. C-Y.C. performed the statistical assays and analyzed the data. T.-W.C. developed and maintained the CircMiMi package. T.-J.C. conceived the study, analyzed the data, and wrote the article. Competing interest: The authors declare no competing financial interests. References Ansel A, Rosenzweig JP, Zisman PD, Melamed M, Gesundheit B. Variation in Gene Expression in Autism Spectrum Disorders: An Extensive Review of Transcriptomic Studies. Frontiers in Neuroscience 10 , 601 (2016). Grove J , et al. Identification of common genetic risk variants for autism spectrum disorder. Nat Genet 51 , 431-444 (2019). Satterstrom FK , et al. Large-Scale Exome Sequencing Study Implicates Both Developmental and Functional Changes in the Neurobiology of Autism. Cell 180 , 568-584 e523 (2020). Gupta S , et al. Transcriptome analysis reveals dysregulation of innate immune response genes and neuronal activity-dependent genes in autism. Nat Commun 5 , 5748 (2014). Parikshak NN , et al. Genome-wide changes in lncRNA, splicing, and regional gene expression patterns in autism. Nature 540 , 423-427 (2016). Wu YE, Parikshak NN, Belgard TG, Geschwind DH. Genome-wide, integrative analysis implicates microRNA dysregulation in autism spectrum disorder. Nat Neurosci 19 , 1463-1476 (2016). Mor M, Nardone S, Sams DS, Elliott E. Hypomethylation of miR-142 promoter and upregulation of microRNAs that target the oxytocin receptor gene in the autism prefrontal cortex. Mol Autism 6 , 46 (2015). Tran SS , et al. Widespread RNA editing dysregulation in brains from autistic individuals. Nat Neurosci 22 , 25-36 (2019). Gokool A, Anwar F, Voineagu I. The Landscape of Circular RNA Expression in the Human Brain. Biol Psychiatry 87 , 294-304 (2020). Chen YJ , et al. Genome-wide, integrative analysis of circular RNA dysregulation and the corresponding circular RNA-microRNA-mRNA regulatory axes in autism. Genome Res 30 , 375-391 (2020). Rockman MV, Kruglyak L. Genetics of global gene expression. Nat Rev Genet 7 , 862-872 (2006). Nica AC, Dermitzakis ET. Expression quantitative trait loci: present and future. Philos Trans R Soc Lond B Biol Sci 368 , 20120362 (2013). Pierce BL , et al. Mediation analysis demonstrates that trans-eQTLs are often explained by cis-mediation: a genome-wide analysis among 1,800 South Asians. PLoS Genet 10 , e1004818 (2014). Grundberg E , et al. Mapping cis- and trans-regulatory effects across multiple tissues in twins. Nat Genet 44 , 1084-1089 (2012). Liu X , et al. Functional Architectures of Local and Distal Regulation of Gene Expression in Multiple Human Tissues. Am J Hum Genet 100 , 605-616 (2017). Sun BB , et al. Genomic atlas of the human plasma proteome. Nature 558 , 73-79 (2018). Liu X, Li YI, Pritchard JK. Trans Effects on Gene Expression Can Drive Omnigenic Inheritance. Cell 177 , 1022-1034 e1026 (2019). Yang F, Wang J, Consortium GT, Pierce BL, Chen LS. Identifying cis-mediators for trans-eQTLs across many human tissues using genomic mediation analysis. Genome Res 27 , 1859-1871 (2017). Yao C , et al. Dynamic Role of trans Regulation of Gene Expression in Relation to Complex Traits. Am J Hum Genet 100 , 571-580 (2017). Chen I, Chen CY, Chuang TJ. Biogenesis, identification, and function of exonic circular RNAs. Wiley Interdiscip Rev RNA 6 , 563-579 (2015). Chen LL. The expanding regulatory mechanisms and cellular functions of circular RNAs. Nat Rev Mol Cell Biol , (2020). Chen LL, Yang L. Regulation of circRNA biogenesis. RNA Biol 12 , 381-388 (2015). You X , et al. Neural circular RNAs are derived from synaptic genes and regulated by development and plasticity. Nat Neurosci 18 , 603-610 (2015). Rybak-Wolf A , et al. Circular RNAs in the Mammalian Brain Are Highly Abundant, Conserved, and Dynamically Expressed. Mol Cell 58 , 870-885 (2015). Szabo L , et al. Statistically based splicing detection reveals neural enrichment and tissue-specific induction of circular RNA during human fetal development. Genome Biol 16 , 126 (2015). Lee C, Kang EY, Gandal MJ, Eskin E, Geschwind DH. Profiling allele-specific gene expression in brains from individuals with autism spectrum disorder reveals preferential minor allele usage. Nat Neurosci 22 , 1521-1532 (2019). Zhang XO, Wang HB, Zhang Y, Lu X, Chen LL, Yang L. Complementary sequence-mediated exon circularization. Cell 159 , 134-147 (2014). Shabalin AA. Matrix eQTL: ultra fast eQTL analysis via large matrix operations. Bioinformatics 28 , 1353-1358 (2012). Oldham MC , et al. Functional organization of the transcriptome in human brain. Nat Neurosci 11 , 1271-1282 (2008). Voineagu I , et al. Transcriptomic analysis of autistic brain reveals convergent molecular pathology. Nature 474 , 380-384 (2011). Willsey AJ , et al. Coexpression networks implicate human midfetal deep cortical projection neurons in the pathogenesis of autism. Cell 155 , 997-1007 (2013). Parikshak NN , et al. Integrative functional genomic analyses implicate specific molecular pathways and circuits in autism. Cell 155 , 1008-1021 (2013). Liu Z, Ran Y, Tao C, Li S, Chen J, Yang E. Detection of circular RNA expression and related quantitative trait loci in the human dorsolateral prefrontal cortex. Genome Biol 20 , 99 (2019). Ahmed I, Karedath T, Al-Dasim FM, Malek JA. Identification of human genetic variants controlling circular RNA expression. RNA 25 , 1765-1778 (2019). Jeck WR , et al. Circular RNAs are abundant, conserved, and associated with ALU repeats. RNA 19 , 141-157 (2013). Chuang TJ , et al. Integrative transcriptome sequencing reveals extensive alternative trans-splicing and cis-backsplicing in human cells. Nucleic Acids Res 46 , 3671-3691 (2018). Tingley D YT, Hirose K, Keele L, Imai K. mediation: R Package for Causal Mediation Analysis. Journal of Statistical Software 59 , 1-38 (2014). Millstein J, Zhang B, Zhu J, Schadt EE. Disentangling molecular relationships with a causal inference test. BMC Genet 10 , 23 (2009). Millstein J, Chen GK, Breton CV. cit: hypothesis testing software for mediation analysis in genomic applications. Bioinformatics 32 , 2364-2365 (2016). Abrahams BS , et al. SFARI Gene 2.0: a community-driven knowledgebase for the autism spectrum disorders (ASDs). Mol Autism 4 , 36 (2013). Iossifov I , et al. The contribution of de novo coding mutations to autism spectrum disorder. Nature 515 , 216-221 (2014). Darnell JC , et al. FMRP stalls ribosomal translocation on mRNAs linked to synaptic function and autism. Cell 146 , 247-261 (2011). Weyn-Vanhentenryck SM , et al. HITS-CLIP and integrative modeling define the Rbfox splicing-regulatory network linked to brain development and autism. Cell Rep 6 , 1139-1152 (2014). Mukherjee N , et al. Integrative regulatory mapping indicates that the RNA-binding protein HuR couples pre-mRNA processing and mRNA stability. Mol Cell 43 , 327-339 (2011). Krishnan A , et al. Genome-wide prediction and functional characterization of the genetic basis of autism spectrum disorder. Nat Neurosci 19 , 1454-1462 (2016). Lek M , et al. Analysis of protein-coding genetic variation in 60,706 humans. Nature 536 , 285-291 (2016). Iossifov I , et al. Low load for disruptive mutations in autism genes and their biased transmission. Proc Natl Acad Sci U S A 112 , E5600-5607 (2015). Velmeshev D , et al. Single-cell genomics identifies cell type-specific molecular changes in autism. Science 364 , 685-689 (2019). Li JH, Liu S, Zhou H, Qu LH, Yang JH. starBase v2.0: decoding miRNA-ceRNA, miRNA-ncRNA and protein-RNA interaction networks from large-scale CLIP-Seq data. Nucleic Acids Res 42 , D92-97 (2014). Bi C , et al. Mutations of ANK3 identified by exome sequencing are associated with autism susceptibility. Hum Mutat 33 , 1635-1638 (2012). Smith KR , et al. Psychiatric risk factor ANK3/ankyrin-G nanodomains regulate the structure and function of glutamatergic synapses. Neuron 84 , 399-415 (2014). Garza JC , et al. Disruption of the psychiatric risk gene Ankyrin 3 enhances microtubule dynamics through GSK3/CRMP2 signaling. Transl Psychiatry 8 , 135 (2018). Buniello A , et al. The NHGRI-EBI GWAS Catalog of published genome-wide association studies, targeted arrays and summary statistics 2019. Nucleic Acids Res 47 , D1005-D1012 (2019). Pain O , et al. Novel Insight Into the Etiology of Autism Spectrum Disorder Gained by Integrating Expression Data With Genome-wide Association Statistics. Biol Psychiatry 86 , 265-273 (2019). Hall LS , et al. Cis-effects on gene expression in the human prenatal brain associated with genetic risk for neuropsychiatric disorders. Mol Psychiatry , (2020). Gandal MJ , et al. Transcriptome-wide isoform-level dysregulation in ASD, schizophrenia, and bipolar disorder. Science 362 , (2018). Gusev A , et al. Integrative approaches for large-scale transcriptome-wide association studies. Nat Genet 48 , 245-252 (2016). Wainberg M , et al. Opportunities and challenges for transcriptome-wide association studies. Nat Genet 51 , 592-599 (2019). Fang WQ , et al. Overproduction of upper-layer neurons in the neocortex leads to autism-like features in mice. Cell Rep 9 , 1635-1643 (2014). Fame RM, MacDonald JL, Macklis JD. Development, specification, and diversity of callosal projection neurons. Trends Neurosci 34 , 41-50 (2011). Iwata K , et al. N-ethylmaleimide-sensitive factor interacts with the serotonin transporter and modulates its trafficking: implications for pathophysiology in autism. Mol Autism 5 , 33 (2014). Chuang TJ, Wu CS, Chen CY, Hung LY, Chiang TW, Yang MY. NCLscan: accurate identification of non-co-linear transcripts (fusion, trans-splicing and circular RNA) with a good balance between sensitivity and precision. Nucleic Acids Res 44 , e29 (2016). Ivanov A , et al. Analysis of intron sequences reveals hallmarks of circular RNA biogenesis in animals. Cell reports 10 , 170-177 (2015). Altschul SF, Gish W, Miller W, Myers EW, Lipman DJ. Basic local alignment search tool. J Mol Biol 215 , 403-410 (1990). Hinrichs AS , et al. The UCSC Genome Browser Database: update 2006. Nucleic Acids Res 34 , D590-598 (2006). Storey JD, Tibshirani R. Statistical significance for genomewide studies. Proc Natl Acad Sci U S A 100 , 9440-9445 (2003). Wang P, Zhao D, Lachman HM, Zheng D. Enriched expression of genes associated with autism spectrum disorders in human inhibitory neurons. Transl Psychiatry 8 , 13 (2018). Schizophrenia Working Group of the Psychiatric Genomics C. Biological insights from 108 schizophrenia-associated genetic loci. Nature 511 , 421-427 (2014). Chen J, Bardes EE, Aronow BJ, Jegga AG. ToppGene Suite for gene list enrichment analysis and candidate gene prioritization. Nucleic Acids Res 37 , W305-311 (2009). Palleja A, Horn H, Eliasson S, Jensen LJ. DistiLD Database: diseases and traits in linkage disequilibrium blocks. Nucleic Acids Res 40 , D1036-1040 (2012). Tables Table 1. Number of samples of the RNA-seq and genotyping data used in this project. Data type Number of samples ASD Non-ASD Sum RNA-seq data 5 111 (48) 101 (49) 212 (97) FC samples 37 36 73 TC samples 36 25 61 CV samples 38 40 78 Genotyping data 26 40 39 79 RNA-seq data with the corresponding genotyping data FC+TC+CV samples 81 (38) 84 (37) 165 (75) FC+TC samples 53 (35) 52 (34) 105 (69) Note. The numbers of the examined individuals are shown in parentheses. FC, frontal cortex (Brodmann Area 9). TC, temporal cortex (Brodmann Area 22, 41, and 42). CV, cerebellar vermis. The data were downloaded from the PsychENCODE Consortium Knowledge Portal at https://www.synapse.org/#!Synapse:syn4587609 under accession number syn458609. Table 2. The observed trans -eGenes that were also previously identified as TWAS-prioritized genes/transcripts in ASD. This study TWAS reference Trans -eGene circQTL/ trans -eQTL SNP circRNA Passing CIT? * Passing mediation test? * CRHR1 rs79543065, rs80018118, rs117601704 circERC1-1 No Yes (52-54%) Gandal et al. 56 rs572758241, rs145151802, rs188227047, rs146232238, rs148453911, rs147867900, rs141400333, rs150595674, rs182196454 circSPHKAP No Yes (51-72%) rs118122108, rs79177473, rs117315428 circERC1-2 No No rs182290762 circZNF236 No No XKR6 rs572758241 circSPHKAP No Yes (58%) NKX2-2 rs182920567 circACVR2A No No NSF rs182920567 circACVR2A Yes No Hall et al. 55 ARL17B rs572758241 circSPHKAP No Yes (33%) * “Passing CIT” and “passing mediation test” represent that the circQTL- trans -eGene-ASD diagnosis axes and the circQTL-circRNA- trans -eGene axes pass CIT (see also Supplementary Tables 4) and mediation test (see also Supplementary Tables 3), respectively. The proportions of mediation of the circQTL- trans -eGene associations mediated by the corresponding circRNA expression were denoted in parentheses. Additional Declarations There is NO Competing Interest. Supplementary Files SupplementaryFigTable.docx Supplementary Figures and Table SupplementaryTable1Dataused.xlsx Supplementary Table 1 SupplementaryTable2circQTLs.xlsx Supplementary Table 2 SupplementaryTable3transmediation.xlsx Supplementary Table 3 SupplementaryTable4CITnew20210115.xlsx Supplementary Table 4 SupplementaryTable5miRNARBP.xlsx Supplementary Table 5 SupplementaryTable6LDblock.xlsx Supplementary Table 6 Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-155043","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":10184042,"identity":"89c254a1-6b71-4d1d-9c9e-9aa4229a9806","order_by":0,"name":"Te-Lun Mai","email":"","orcid":"https://orcid.org/0000-0002-9356-6198","institution":"Genomics Research Center, Academia Sinica","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Te-Lun","middleName":"","lastName":"Mai","suffix":""},{"id":10184043,"identity":"0baa3560-aa87-4f83-85d6-1387afb7ca18","order_by":1,"name":"Chia-Ying Chen","email":"","orcid":"","institution":"Genomics Research Center, Academia Sinica, Taipei","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Chia-Ying","middleName":"","lastName":"Chen","suffix":""},{"id":10184044,"identity":"d0c326be-5df2-4cae-98b5-d88e8b5e0f25","order_by":2,"name":"Tai-Wei Chiang","email":"","orcid":"","institution":"Academia Sinica","correspondingAuthor":false,"submittingAuthor":false,"prefix":"","firstName":"Tai-Wei","middleName":"","lastName":"Chiang","suffix":""},{"id":10184045,"identity":"e9358ed5-cc6c-402b-b99d-45762560a191","order_by":3,"name":"Trees-Juen Chuang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA/0lEQVRIiWNgGAWjYNCCCiidwCABpiXwKeYBk2egDCQtBvi1MLbB9EIBXi320ocPPvw5r1benr2B+cPDHRby/A3MB2/zMPxJbMBlC19asjHvtuOGPTwH2CQSz0gYzjjAlmzNw2CAWwsPj5k047ZjjD0SCWwMiW0SCQwHgCJALbm4tfB///lzzjH7HvkHzB9AWuQP8H8joIWHjYG3oSaxB+hlCZAWgwM8bPi1nGEzluY5diC55wxQPRAZbjzMZmw5x8C4HpcW9h7mhx9/1NTZtrcfPvzxZ1udvNzx5oc33lTIGePQAQOHgZgRaiwziMAdLTBQR1DFKBgFo2AUjGAAAOP5TPiM3wz+AAAAAElFTkSuQmCC","orcid":"","institution":"Genomics Research Center, Academia Sinica, Taipei","correspondingAuthor":true,"submittingAuthor":false,"prefix":"","firstName":"Trees-Juen","middleName":"","lastName":"Chuang","suffix":""}],"badges":[],"createdAt":"2021-01-25 19:59:55","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-155043/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-155043/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":5664907,"identity":"302b5d98-b6ba-4456-890b-647a810ea7b6","added_by":"auto","created_at":"2021-02-05 15:49:46","extension":"jpg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":134427,"visible":true,"origin":"","legend":"Identification and analysis of circQTLs based on the brain samples from people with ASD and controls. (a) Flowchart of the overall approach. (b, c) Comparisons of effect sizes (b) for circQTL-circRNA associations observed in samples from frontal cortex (FC), temporal cortex (TC), and cerebellar vermis (CV) using Pearson correlation analysis (b) and Z-test (c). For c, P values of differences in absolute values of Z-scores were determined using two-tailed Wilcoxon signed-rank test. (d) Principal component plots of circRNA expression profiles of the 81 circRNAs associated with the identified circQTL-circRNA pairs in samples from TC, FC, and CV. PC1/PC2, the first and second principal components. (e) Distance distribution of the circQTL SNPs to their nearest back-spliced sites. For internal circQTL SNPs, the distance was calculated by the ratio of the distance between the SNP and the donor-to-acceptor back-spliced site. (f, g) Comparison of proportion of circQTL and non-circQTL SNPs located in (f) RCS and (g) predicted RBP binding sites. The non-circQTL SNPs were also located in the flanking sequences (± 200 kb nucleotides) of the examined back-splice sites. P values were determined using two-tailed Fisher’s exact test.","description":"","filename":"1.jpg","url":"https://assets-eu.researchsquare.com/files/rs-155043/v1/589e1fbcba7d9e6093a0ece7.jpg"},{"id":5664802,"identity":"74e253f3-3ab6-427e-a03a-cbd45af7a3ce","added_by":"auto","created_at":"2021-02-05 15:46:46","extension":"jpg","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":88337,"visible":true,"origin":"","legend":"Identification and analysis of circQTLs based on the brain samples from people with ASD and controls. (a) Flowchart of the overall approach. (b, c) Comparisons of effect sizes (b) for circQTL-circRNA associations observed in samples from frontal cortex (FC), temporal cortex (TC), and cerebellar vermis (CV) using Pearson correlation analysis (b) and Z-test (c). For c, P values of differences in absolute values of Z-scores were determined using two-tailed Wilcoxon signed-rank test. (d) Principal component plots of circRNA expression profiles of the 81 circRNAs associated with the identified circQTL-circRNA pairs in samples from TC, FC, and CV. PC1/PC2, the first and second principal components. (e) Distance distribution of the circQTL SNPs to their nearest back-spliced sites. For internal circQTL SNPs, the distance was calculated by the ratio of the distance between the SNP and the donor-to-acceptor back-spliced site. (f, g) Comparison of proportion of circQTL and non-circQTL SNPs located in (f) RCS and (g) predicted RBP binding sites. The non-circQTL SNPs were also located in the flanking sequences (± 200 kb nucleotides) of the examined back-splice sites. P values were determined using two-tailed Fisher’s exact test.","description":"","filename":"2.jpg","url":"https://assets-eu.researchsquare.com/files/rs-155043/v1/78191171685c860b783e223e.jpg"},{"id":5664808,"identity":"ed7ce3cd-a430-4849-af89-a753e6f6a203","added_by":"auto","created_at":"2021-02-05 15:46:47","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":124921,"visible":true,"origin":"","legend":"Inference of causal relations between circQTLs (trans-eQTLs), trans-eGenes, and ASD diagnosis by the causal inference test (CIT)38,39. (a) The circQTL-trans-eGene pairs (1,000 pairs) passing CIT with the propagation path from circQTL SNPs (50 circQTLs) to ASD diagnosis via trans-eGene (708 genes) expression. (b) Heat map showing enrichment of ASD-relevant genes and genes implicated in monogenetic forms of other brain disorders for trans-eGenes passing CIT and trans-eGenes not passing CIT. ASD-relevant genes included ASD risk genes from SFARI, genes encoding postsynaptic density (PSD) proteins, genes encoding transcripts bound by FMR1 (FMR1 target), RBFOX1 (RBFOX1 target), and ELAVL1 (ELAVL1 target), and differentially expressed genes (DEGs) in ASD. Top SFARI genes represented SFARI genes with score = 1 or 2. The enrichment odd ratios and FDR Bonferroni-corrected P values (in parentheses) were shown for enrichment with FDR \u003c 0.05. (c, d) Comparisons of the proportions of trans-eGenes passing CIT and trans-eGenes not passing CIT with (c) different probabilities of ASD risk and (d) different gene variant intolerance (pLI) scores. P values were determined using two-tailed Fisher’s exact test. Background, the background gene set of 20,070 protein-coding genes. (e) Enrichment analysis of DEGs in a cell type-specific manner for trans-eGenes passing CIT. (f) GO analysis for trans-eGenes passing CIT. For b and e, P values were determined using one-tailed Fisher’s exact test. For e and f, the red dashed lines represented FDR Bonferroni-corrected P values for enrichment with FDR = 0.05.","description":"","filename":"3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-155043/v1/d8e396757746f8b330546063.jpg"},{"id":5664908,"identity":"b8c75514-6791-46f4-9f45-cbbb315e18d3","added_by":"auto","created_at":"2021-02-05 15:49:46","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":118608,"visible":true,"origin":"","legend":"Potential circQTL-circRNA-trans-eGene-trait (ASD diagnosis) propagation paths. (a) Identification of potential circQTL-circRNA-trans-eGene-trait propagation paths (right) by integrating CIT-passing circQTL-trans-eGene pairs (left) with mediation-passing circQTL-circRNA-trans-eGene axes (middle). (b) The 257 identified circQTL-circRNA-trans-eGene-trait propagation networks plotted by the Cytoscape package. The ASD-relevant trans-eGenes were shown, in which the top SFARI genes were highlighted in red.","description":"","filename":"4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-155043/v1/5e2b80533e209a9e92d380a4.jpg"},{"id":5664910,"identity":"f42cf177-cf70-4209-8ea1-6b7f5d9dfce0","added_by":"auto","created_at":"2021-02-05 15:49:47","extension":"jpg","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":124022,"visible":true,"origin":"","legend":"Potential mediators for the circRNA-trans-eGene interactions. (a) Identification of potential circRNA-miRNA-trans-eGene and circRNA-RBP-trans-eGene axes according to the common miRNA and RBP target sites of the circRNAs and trans-eGenes, respectively. (b) Identification of potential circQTL-circRNA-miRNA-trans-eGene-trait propagation paths (right) by integrating the 257 identified circQTL-circRNA-trans-eGene-trait propagation paths (left; see also Fig. 4) with the identified circRNA-miRNA-trans-eGene axes in (a) (middle). (c) The 158 identified circQTL-circRNA-miRNA-trans-eGene-trait propagation networks plotted by the Cytoscape package. The ASD-relevant trans-eGenes were shown, in which the top SFARI genes were highlighted in red.","description":"","filename":"5.jpg","url":"https://assets-eu.researchsquare.com/files/rs-155043/v1/8e946a697b5f739d9f62b303.jpg"},{"id":17179091,"identity":"f335e02c-098c-420a-a55a-5469143c8457","added_by":"auto","created_at":"2022-01-10 20:40:49","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":873162,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-155043/v1/bd9d3028-80f0-428e-8602-4b1b40e1af00.pdf"},{"id":5664909,"identity":"761525f3-1d06-40f7-b247-805a925d7796","added_by":"auto","created_at":"2021-02-05 15:49:47","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":967573,"visible":true,"origin":"","legend":"Supplementary Figures and Table","description":"","filename":"SupplementaryFigTable.docx","url":"https://assets-eu.researchsquare.com/files/rs-155043/v1/db30afe861d0fb776f5df12f.docx"},{"id":5664806,"identity":"92d0528b-6179-4b3d-adc3-2c2abb7bd06d","added_by":"auto","created_at":"2021-02-05 15:46:47","extension":"xlsx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":1981344,"visible":true,"origin":"","legend":"Supplementary Table 1","description":"","filename":"SupplementaryTable1Dataused.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-155043/v1/bf07ecfb8eec0cb35e383533.xlsx"},{"id":5664914,"identity":"93e7dc04-da93-4731-9bbb-40e22e4e508d","added_by":"auto","created_at":"2021-02-05 15:49:47","extension":"xlsx","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":303694,"visible":true,"origin":"","legend":"Supplementary Table 2","description":"","filename":"SupplementaryTable2circQTLs.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-155043/v1/73ae685065b738d8840efc1d.xlsx"},{"id":5664915,"identity":"3da5d592-3f72-4ffd-9bbf-35d4b877bf22","added_by":"auto","created_at":"2021-02-05 15:49:47","extension":"xlsx","order_by":4,"title":"","display":"","copyAsset":false,"role":"supplement","size":29941169,"visible":true,"origin":"","legend":"Supplementary Table 3","description":"","filename":"SupplementaryTable3transmediation.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-155043/v1/c465424552ba5c66fbed0bc1.xlsx"},{"id":5665189,"identity":"367978ae-8e92-4dba-80c3-aac1b1d144fc","added_by":"auto","created_at":"2021-02-05 15:52:47","extension":"xlsx","order_by":5,"title":"","display":"","copyAsset":false,"role":"supplement","size":8139438,"visible":true,"origin":"","legend":"Supplementary Table 4","description":"","filename":"SupplementaryTable4CITnew20210115.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-155043/v1/ada10699c3bb0a86e3cb2e34.xlsx"},{"id":5665188,"identity":"1caa57dc-dc6e-4326-afee-fcea9d2121f4","added_by":"auto","created_at":"2021-02-05 15:52:47","extension":"xlsx","order_by":6,"title":"","display":"","copyAsset":false,"role":"supplement","size":555121,"visible":true,"origin":"","legend":"Supplementary Table 5","description":"","filename":"SupplementaryTable5miRNARBP.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-155043/v1/cb94b08553168f8e9b33f570.xlsx"},{"id":5664911,"identity":"12359d58-77c6-40a1-8689-ea1ea6694b1f","added_by":"auto","created_at":"2021-02-05 15:49:47","extension":"xlsx","order_by":7,"title":"","display":"","copyAsset":false,"role":"supplement","size":44491,"visible":true,"origin":"","legend":"Supplementary Table 6","description":"","filename":"SupplementaryTable6LDblock.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-155043/v1/e8bcf22feeccdbb45d1ec3fe.xlsx"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"\u003ci\u003eTrans\u003c/i\u003e-genetic effects of circular RNA expression quantitative trait loci and potential causal mechanisms in autism","fulltext":[{"header":"Introduction","content":"\u003cp\u003eAutism spectrum disorder (ASD) is a highly pervasive neurodevelopmental and heritable complex disorder, which are characterized by limited social communication, restricted and repetitive interests or behaviors \u003csup\u003e1\u003c/sup\u003e. Previous large-scale genomic studies have identified a variety of genomic variants associated with the etiology of ASD \u003csup\u003e2,3\u003c/sup\u003e, providing valuable biological insights into this disorder. However, the contribution of genetic factors to this complex disease is highly heterogeneous and not yet well understood. On the other hand, through high-throughput RNA sequencing (RNA-seq) data of postmortem brains from people with idiopathic ASD and non-ASD controls, numerous studies identified valuable differential expression patterns in mRNAs \u003csup\u003e4,5\u003c/sup\u003e, long non-coding RNAs \u003csup\u003e5\u003c/sup\u003e, microRNAs (miRNAs) \u003csup\u003e6,7\u003c/sup\u003e, RNA editing \u003csup\u003e8\u003c/sup\u003e, and circular RNAs (circRNAs) \u003csup\u003e9,10\u003c/sup\u003e. While these results increased our understanding of the underlying molecular mechanisms in ASD pathophysiology, the causal relationships between genetic sequence variants and transcriptional expression changes in ASD remains mostly unknown.\u003c/p\u003e\n\u003cp\u003eBy integrating genome-wide genotyping with transcriptome profiling, expression quantitative trait loci (eQTL) analysis identifies single nucleotide polymorphisms (SNPs) that affect the expression levels of local (\u003cem\u003ecis\u003c/em\u003e-eQTL) or distant (\u003cem\u003etrans\u003c/em\u003e-eQTL) genes, providing a connection between genetics and regulatory mechanisms of gene expression\u003csup\u003e11\u003c/sup\u003e. \u003cem\u003ecis\u003c/em\u003e-eQTL SNPs often reside within or close to promoter regions of their nearby genes (\u003cem\u003ecis\u003c/em\u003e-eGenes) and mediate gene expression through directly affecting the corresponding transcription factor binding process \u003csup\u003e12\u003c/sup\u003e. In contrast to frequent investigation of \u003cem\u003ecis\u003c/em\u003e-eQTLs, the regulatory mechanism underlying \u003cem\u003etrans\u003c/em\u003e-eQTLs are less known because of the relatively weaker effects on gene expression for \u003cem\u003etrans-\u003c/em\u003eactions than for \u003cem\u003ecis-\u003c/em\u003eactions and the multiple-testing burden \u003csup\u003e13,14\u003c/sup\u003e. However, compared with \u003cem\u003ecis\u003c/em\u003e-effects on gene expression, \u003cem\u003etrans\u003c/em\u003e-genetic effects were reported to explain more than two times the variability in gene expression \u003csup\u003e14\u003c/sup\u003e and be more cell type-specific \u003csup\u003e15\u003c/sup\u003e. Some disease-causing variants involved in \u003cem\u003etrans\u003c/em\u003e-effect analyses may miss in \u003cem\u003ecis\u003c/em\u003e-effect analyses \u003csup\u003e16\u003c/sup\u003e. A recent study further indicated that at least 70% of complex trait heritability is driven by \u003cem\u003etrans\u003c/em\u003e-eQTL effects \u003csup\u003e17\u003c/sup\u003e. A possible regulatory mechanism for \u003cem\u003etrans\u003c/em\u003e-eQTLs is that \u003cem\u003etrans\u003c/em\u003e-eQTLs indirectly affect the expression of distant genes (\u003cem\u003etrans\u003c/em\u003e-eGenes) through regulating the expression of genes near the \u003cem\u003etrans\u003c/em\u003e-eQTLs (\u003cem\u003ecis\u003c/em\u003e-eGenes) \u003csup\u003e13,18,19\u003c/sup\u003e. In such cases, a \u003cem\u003etrans\u003c/em\u003e-eQTL is also a \u003cem\u003ecis\u003c/em\u003e-eQTLs, wherein the corresponding \u003cem\u003ecis\u003c/em\u003e-eGene acts as a mediator of the \u003cem\u003etrans\u003c/em\u003e-eQTL. Investigation of \u003cem\u003etrans\u003c/em\u003e-genetic effects and the causal inference for the corresponding eQTL-associated networks will allow us to better understand complex trait genetics.\u003c/p\u003e\n\u003cp\u003eCircRNAs are an emerging class of RNAs, which are formed by pre-mRNA back-splicing with a structure of covalently closed loops and endogenously expressed as single-strand, non-polyadenylated circular molecules \u003csup\u003e20,21\u003c/sup\u003e. Although circRNAs are generally expressed at a much lower level compared with their corresponding co-linear mRNA isoforms, they are more stable than other types of RNAs \u003csup\u003e20,22\u003c/sup\u003e. CircRNAs were observed to be especially abundant in the brain \u003csup\u003e23,24\u003c/sup\u003e and often play important roles in development of nervous system \u003csup\u003e24,25\u003c/sup\u003e. As for the layer of gene regulatory networks, circRNAs often serve as an upstream regulator of mRNAs. For example, they may regulate genes in \u003cem\u003etrans\u003c/em\u003e through mediating the activities of miRNAs or RNA binding proteins (RBPs), with the common miRNA/RBP target site(s) of circRNAs and the target genes \u003csup\u003e20,21\u003c/sup\u003e. Hence, we are curious about whether SNPs that influence expression of nearby circRNAs (referred to as circQTLs) may also affect expression of remote genes, whereby circRNAs serve as an intermediate regulator bridging circQTLs and \u003cem\u003etrans\u003c/em\u003e-eGenes. If this is possible, we can generate circQTL-circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene propagation paths in autistic brains and shed light on the functional consequences of the genetic loci and the etiology of this complex disease.\u003c/p\u003e\n\u003cp\u003eThere were two challenges to achieve the abovementioned goals. First, eQTL analysis requires transcriptomic and genomic data from the same samples with a large sample size. Second, large-scale circRNA study is often hampered by the limitation of the intrinsic circRNA characteristic of lacking polyadenylated tails. The newly released Synapse database \u003csup\u003e5,26\u003c/sup\u003e comprises a large human brain sample size of both genotyping data and RNA-seq data (rRNA-depleted RNAs from total RNAs without poly(A)-selection) from ASD cases and controls, offering an unprecedented opportunity for us to decipher the effects of genetic variation on circRNA expression in ASD brain. The data from the Synapse database thus allowed us to identify a new reference of circQTLs/\u003cem\u003etrans\u003c/em\u003e-eQTLs and ASD-associated \u003cem\u003etrans\u003c/em\u003e-eQTL hotspots in the brains and infer \u003cem\u003ecis\u003c/em\u003e-mediators of the identified circQTLs (i.e., such circQTLs are also \u003cem\u003etrans\u003c/em\u003e-eQTLs) and causal effects between circQTLs and ASD diagnosis. Our findings help to unveil the gap between genetic variation and phenotypic changes in ASD and provides a framework for future analyses of eQTLs in neuropsychiatric disease or other complex traits.\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003eIdentification of circQTLs in ASD brain\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe first collected rRNA-depleted paired-end RNA-seq data \u003csup\u003e5\u003c/sup\u003e and genotyping data \u003csup\u003e26\u003c/sup\u003e of postmortem samples from frontal cortex (FC) (Brodmann area 9), temporal cortex (TC) (Brodmann area 22, 41, and 42), and cerebellar vermis (CV) from individuals with ASD and non-ASD controls (the Synapse database) (Table 1). To identify local genetic effects that regulate circRNA expression, we considered the RNA-seq data and the corresponding genotyping data, both of which were derived from the same individuals (Fig. 1a, Table 1, and Supplementary Table 1). The examined circRNAs (1,060 circRNAs; Supplementary Table S1) were extracted from our previous study \u003csup\u003e10\u003c/sup\u003e based on the RNA-seq data used in this study. The expression levels of circRNAs were calculated using the number of supporting circRNA junction reads per million uniquely mapped reads (RPM) \u003csup\u003e27\u003c/sup\u003e and adjusted for covariates such as the corresponding host gene expression, brain region, diagnosis, and other biological/technical confounding factors (see Methods and Supplementary Table 1). For accuracy, the genotyped and imputed SNPs should be satisfied imputation score \u0026sup3; 0.8 and estimated minor allele frequency \u0026sup3; 0.01. To focus on \u003cem\u003ecis\u003c/em\u003e-effects of circQTLs, we limited our analysis to SNPs in \u0026plusmn; 200 kb nucleotides of each back-splice site. CircQTLs were evaluated by testing the correlations between the imputed genotype dosages and covariate-adjusted circRNA expression using Matrix eQTL \u003csup\u003e28\u003c/sup\u003e. \u003cem\u003eP\u003c/em\u003e values were adjusted across all circQTLs using false positive rate (FDR) correction (Methods). Significant circQTLs were determined if FDR values were less than 0.05. We thus identified 989 circQTLs associated with 81 circRNAs (Supplementary Table 2).\u003c/p\u003e\n\u003cp\u003eRecent transcriptomics studies of ASD showed that expression profiles of transcripts (including circRNAs \u003csup\u003e10\u003c/sup\u003e, miRNAs\u003csup\u003e6\u003c/sup\u003e, and mRNAs\u003csup\u003e29,30\u003c/sup\u003e) and allele-specific genes \u003csup\u003e26\u003c/sup\u003e were highly similar between the two cortical regions (FC and TC) but were quite distinct in the CV. We calculated the circQTL effects for the 989 circQTLs in the three brain regions, respectively. We found that the circQTL effect size and direction were highly concordant between the two cortical regions and relatively weakly concordant (or even not concordant) between the cortex and CV (Fig. 1b). The differences in circQTL effect size between the cortex and CV were more significant than that between the two cortical regions (Fig. 1c). We also performed principal component analysis (PCA) and observed the similar results for the expression profiles of the 81 circRNAs (Fig. 1d). Previous studies have reported that the ASD pathophysiology was present in the cortex \u003csup\u003e31,32\u003c/sup\u003e and transcriptional dysregulation in ASD were shown to be consistently stronger in the cortex than in the cerebellum \u003csup\u003e5\u003c/sup\u003e. These observations suggest that the cortical regions are more selectively vulnerable to transcriptomic changes than the cerebellum. Therefore, we focused on the RNA-seq data and the corresponding genotyping data from the cortex (FC/TC) samples (105 samples from 53 ASD cases and 52 controls; Table 1) for the subsequent circQTL analyses. For accuracy, we only considered the identified circQTL SNPs (666 circQTLs associated with 48 circRNAs; Supplementary Table 2) in which the number of individuals with the minor allele (heterozygotes and minor allele homozygotes) was larger than three.\u003c/p\u003e\n\u003cp\u003eWe then studied the relationship between circQTLs and circRNA formation. We first showed that the identified circQTLs were preferentially located in the flanking sequences close to the back-splice donor/acceptor sites (Fig. 1e). The majority of circQTLs (90.5%; 603 out of 666; Supplementary Table 2) were located in the flanking sequences of back-splice sites. These observations were consistent with previous studies for distance distribution analysis of circQTLs in other human samples \u003csup\u003e33,34\u003c/sup\u003e. Next, back-splicing was reported to be facilitated by reverse complementary sequences (RCSs) residing in the introns flanking circularized exons \u003csup\u003e27,35,36\u003c/sup\u003e. We found that circQTLs residing in the flanking sequences of back-splice sites were significantly enriched in RCSs compared with background (empirical \u003cem\u003eP\u003c/em\u003e\u0026lt;10\u003csup\u003e-4\u003c/sup\u003e; see Methods) and non-circQTL SNPs (\u003cem\u003eP\u003c/em\u003e=3.2\u0026acute;10\u003csup\u003e-7\u003c/sup\u003e by two-sided Fisher\u0026rsquo;s exact test; Fig. 1f). Of note, the examined non-circQTL SNPs were also located in the flanking sequences of the examined back-splice sites and not determined as circQTLs. Moreover, it was shown that RNA binding proteins (RBPs) can regulate back-splicing by binding to the flanking sequences \u003csup\u003e20,21\u003c/sup\u003e. Indeed, the circQTLs located in the flanking sequences were significantly enriched in RBP binding sites compared with background (empirical \u003cem\u003eP\u003c/em\u003e\u0026lt;10\u003csup\u003e-4\u003c/sup\u003e) and non-circQTL SNPs (\u003cem\u003eP\u003c/em\u003e=0.01; Fig. 1g). These observations support that circQTLs located in the flanking sequences of back-splice sites may contribute to circRNA formation.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMediation effects of \u003cem\u003etrans\u003c/em\u003e-eQTLs (circQTLs) via circRNA expression\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe next examined whether the identified circQTLs were also \u003cem\u003etrans\u003c/em\u003e-eQTLs affecting the expression of distant genes (\u003cem\u003etrans\u003c/em\u003e-eGenes) using Matrix eQTL. By definition, such circQTL SNPs and \u003cem\u003etrans\u003c/em\u003e-eGenes should be located on different chromosomes or the same chromosome separated by a distance greater than 5Mb (from the SNP site to the transcription start site of the \u003cem\u003etrans\u003c/em\u003e-eGene). Since some SNPs in high linkage disequilibrium (LD) were associated with different \u003cem\u003etrans\u003c/em\u003e-eGenes, non-pruned circQTLs were considered in this study. With controlling for potential biological/technical confounding factors and multiple testing correction (Methods), 546 circQTLs (associated with 47 circRNAs) were identified to be also \u003cem\u003etrans\u003c/em\u003e-eQTLs, which affected the expression of 7,165 \u003cem\u003etrans\u003c/em\u003e-eGenes and constructed 43,372 circQTL-\u003cem\u003etrans\u003c/em\u003e-eGene pairs (Fig. 2a; Supplementary Table 3). Meanwhile, these circQTL-\u003cem\u003etrans\u003c/em\u003e-eGene pairs also resulted in 13,379 circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene pairs, where the expression levels of circRNAs and \u003cem\u003etrans\u003c/em\u003e-eGenes were both regulated through the same circQTLs (Fig. 2a). We then explored the causes for the \u003cem\u003etrans\u003c/em\u003e-effect of circQTLs on the expression of \u003cem\u003etrans\u003c/em\u003e-eGenes. In addition to the direct effect of circQTLs on \u003cem\u003etrans\u003c/em\u003e-eGene expression, we hypothesized that in some cases the expression of \u003cem\u003etrans\u003c/em\u003e-eGenes were indirectly regulated through the expression of circRNAs near the circQTLs (i.e., the \u0026ldquo;mediation effect\u0026rdquo;; Fig. 2a). For the scenario of mediation effects, the circRNAs served as \u003cem\u003ecis\u003c/em\u003e-mediators of \u003cem\u003etrans\u003c/em\u003e-eQTLs. We first showed that the expression levels of circRNAs and \u003cem\u003etrans\u003c/em\u003e-eGenes in the 13,379 circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene pairs were more likely to be correlated with each other than expected (empirical \u003cem\u003eP\u003c/em\u003e\u0026lt;10\u003csup\u003e-4\u003c/sup\u003e; see Supplementary Fig. 1 and Methods). To test the mediation effects of \u003cem\u003etrans\u003c/em\u003e-eQTLs, we performed mediation analyses\u003csup\u003e37\u003c/sup\u003e for all 43,372 circQTL-\u003cem\u003etrans\u003c/em\u003e-eGene pairs to identify the proportion of association between a circQTL and the \u003cem\u003etrans\u003c/em\u003e-eGene that was caused by the effect of the circQTL on the corresponding circRNA expression. We found that 19,393 out of the 43,372 (45%) circQTL-\u003cem\u003etrans\u003c/em\u003e-eGene pairs passed the mediation test (Supplementary Table 3), which were significantly mediated (\u003cem\u003eP\u003c/em\u003e and FDR\u0026lt;0.05 with adjusting for related covariates; see Methods) by expression of circRNAs near the circQTLs. The majority (67%, 13,023 pairs) of the 19,393 pairs exhibited that more than 10% of the circQTL-\u003cem\u003etrans\u003c/em\u003e-eGene association were mediated by the expression of circRNAs; in some cases (2,099 pairs), the proportion of mediation of the circQTL-\u003cem\u003etrans\u003c/em\u003e-eGene association by the circRNAs was even more than 50% (Fig. 2b).\u003c/p\u003e\n\u003cp\u003eRegarding the 19,393 circQTL-\u003cem\u003etrans\u003c/em\u003e-eGene pairs (or 19,393 circQTL-circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene axes), we observed that the percentages of circQTL-\u003cem\u003etrans\u003c/em\u003e-eGene pairs passing the mediation test markedly increased with increasing significance levels of \u003cem\u003etrans\u003c/em\u003e-eQTL effects for circQTL-\u003cem\u003etrans\u003c/em\u003e-eGene pairs (Fig. 2c). Such a percentage even reached more than 70% if FDR values of\u003cem\u003e trans\u003c/em\u003e-eQTL effects were smaller than 10\u003csup\u003e-8\u003c/sup\u003e. We further examined the relationship between mediation effects and the significance levels of circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene correlation of expression profile. We found 55% of the circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene pairs passing the mediation test if the circRNA expression and \u003cem\u003etrans\u003c/em\u003e-eGene expression were significantly (Spearman\u0026rsquo;s \u003cem\u003eP\u003c/em\u003e\u0026lt;0.05) correlated with each other (Fig. 2d), whereas such a percentage was significantly reduced from 55% to 40% (odd ratio=0.56 and \u003cem\u003eP\u003c/em\u003e\u0026lt;2.2\u0026acute;10\u003csup\u003e-16\u003c/sup\u003e by Fisher\u0026rsquo;s exact test) if the significant correlation between circRNA and \u003cem\u003etrans\u003c/em\u003e-eGene expression disappeared (Fig. 2d). The percentages of the circQTL-\u003cem\u003etrans\u003c/em\u003e-eGene pairs passing the mediation test markedly increased with increasing significance levels of the correlation of expression profile between circRNAs and \u003cem\u003etrans\u003c/em\u003e-eGenes (Fig. 2d). Such a percentage even reached 90% at Spearman\u0026rsquo;s \u003cem\u003eP\u003c/em\u003e value \u0026lt;0.0005. Similar results were observed in the analyses based on the CV samples (Supplementary Fig. 2 and Supplementary Table 3). These results reveal that the mediation effects of the circQTLs (\u003cem\u003etrans\u003c/em\u003e-eQTLs) on circRNA expression are correlated with the magnitudes of\u003cem\u003e trans\u003c/em\u003e-eQTL effect and circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene correlation of expression profile.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCausal effects between circQTLs and ASD diagnosis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eRegarding the identified 43,372 circQTL-\u003cem\u003etrans\u003c/em\u003e-eGene pairs, we then conducted causal inference test \u003csup\u003e38,39\u003c/sup\u003e (CIT; see Methods) to examine whether the \u003cem\u003etrans\u003c/em\u003e-effects of circQTLs on \u003cem\u003etrans\u003c/em\u003e-eGene expression might explain the association between the circQTLs and diagnosis status (ASD vs. non-ASD) and infer the direction of association between circQTLs, \u003cem\u003etrans\u003c/em\u003e-eGene expression, and ASD diagnosis (Fig. 3a, top). At \u003cem\u003eP\u003c/em\u003e and FDR\u0026lt;0.05, we identified 1,000 circQTL-\u003cem\u003etrans\u003c/em\u003e-eGene pairs with a propagation path from circQTL SNP to ASD diagnosis via \u003cem\u003etrans\u003c/em\u003e-eGene expression (Fig. 3a, bottom; Supplementary Table 4). We speculated that the \u003cem\u003etrans\u003c/em\u003e-eGenes (708 genes; CIT-passing \u003cem\u003etrans\u003c/em\u003e-eGenes) involved in the 1,000 circQTL-\u003cem\u003etrans\u003c/em\u003e-eGene pairs may be implicated in ASD. We examined enrichment analysis for genes previously implicated in ASD from Simons Foundation Autism Research Institutive (SFARI) \u003csup\u003e40\u003c/sup\u003e. We observed that the CIT-passing \u003cem\u003etrans\u003c/em\u003e-eGenes were significantly enriched for the SFARI genes, but not for genes implicated in monogenetic forms of other brain disorders (Fig. 3b). These CIT-passing genes were also enriched for other classes of ASD-relevant genes, including genes encoding postsynaptic density (PSD) proteins \u003csup\u003e41\u003c/sup\u003e, genes whose transcripts were bound by the RBPs of FMR1 \u003csup\u003e42\u003c/sup\u003e, RBFOX1 \u003csup\u003e43\u003c/sup\u003e, and ELAVL1 \u003csup\u003e44\u003c/sup\u003e, and differentially expressed genes (DEGs) in ASD \u003csup\u003e5\u003c/sup\u003e that were derived from the same cortex samples used in this study (Fig. 3b). In contrast, these trends were not observed in the \u003cem\u003etrans\u003c/em\u003e-eGenes not passing CIT (Fig. 3b). Moreover, a previous study has presented a genome-wide prediction of ASD risk genes and provided an estimated probability of ASD-association for each gene \u003csup\u003e45\u003c/sup\u003e. On the basis of the probabilities of ASD risk, we observed the CIT-passing genes indeed had a significantly higher probability of ASD risk compared with background gene set (empirical \u003cem\u003eP\u003c/em\u003e\u0026lt;0.001; Supplementary Fig. 3) and the \u003cem\u003etrans\u003c/em\u003e-eGenes not passing CIT (all \u003cem\u003eP\u003c/em\u003e \u0026lt; 10\u003csup\u003e-8\u003c/sup\u003e by two-sided Fisher\u0026rsquo;s exact test; Fig. 3c). We further found that the CIT-passing\u003cem\u003e trans\u003c/em\u003e-eGenes were more intolerant of a loss function mutation (measured by gene variant intolerance (pLI) scores \u003csup\u003e46\u003c/sup\u003e) compared with the background gene set (Supplementary Fig. 3) and the \u003cem\u003etrans\u003c/em\u003e-eGenes not passing CIT (Fig. 3d), reflecting previous observations that genes implicated in ASD tended to be subject to stronger selective constraints than the other genes \u003csup\u003e45,47\u003c/sup\u003e. Regarding DEGs in a cell type-specific manner, the CIT-passing genes were most overrepresented in vasoactive intestinal polypeptide (VIP)-expressing interneurons and L2/3 and L4 excitatory neurons. This trend was generally consistent with the previous observation for SFARI genes \u003csup\u003e48\u003c/sup\u003e. Gene Ontology (GO) analysis revealed that the CIT-passing genes were enriched in GO terms related to chemical synaptic transmission, synaptic signaling, neurogenesis, synapse, neuron projection, axon, and so on (Fig. 3f and Supplementary Table 4), also reflecting their enrichment for PSD. These observations support the relevance of the CIT-passing \u003cem\u003etrans\u003c/em\u003e-eGenes to ASD pathophysiology.\u003c/p\u003e\n\u003cp\u003eFurthermore, the above mediation analysis has identified 19,393 circQTL-circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene axes with significant mediation effects of circQTLs on \u003cem\u003etrans\u003c/em\u003e-eGene expression through circRNA expression (Fig. 2b and Fig. 4a, middle). By integrating the CIT-passing circQTL-\u003cem\u003etrans\u003c/em\u003e-eGene pairs with the 19,393 circQTL-circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene axes, we determined 257 circQTL-circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene-ASD diagnosis propagation paths, where the circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene axes may act as causal mediators for the circQTL-ASD diagnosis associations (Fig. 4a, left; Supplementary Table 4). The 257 ASD-associated circQTL-circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene regulatory axes involved 33 circQTLs (\u003cem\u003etrans\u003c/em\u003e-eQTLs), 8 circRNAs, and 124 \u003cem\u003etrans\u003c/em\u003e-eGenes (Figs. 4a and 4b). We found that 97 of 124 (78%) \u003cem\u003etrans\u003c/em\u003e-eGenes were ASD-relevant genes and 217 of 257 (84%) circQTL-circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene axes were associated with the 97 ASD-relevant genes (Fig. 4b). Particularly, 10 of the 97 genes were SFARI genes and five were top SFARI genes with score = 1 or 2, suggesting the regulatory role of the corresponding circQTL-circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene axes in ASD brain.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePotential mediators for the circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene interactions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eRegarding the 19,393 circQTL-circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene axes (associated with 5,669 circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene axes) passing the mediation test, we proceeded to identify potential mediators for the circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene interactions (Fig. 5a, left). Since circRNAs and mRNAs may carry the common miRNA/RBP target sites, numerous cases of circRNAs were demonstrated to act as an upstream regulator of mRNAs through mediating miRNA/RBP activities \u003csup\u003e20,21\u003c/sup\u003e. To this end, we utilized crosslinking immunoprecipitation (CLIP)-seq data-supported RNA interactomes (extracted from ENCORI \u003csup\u003e49\u003c/sup\u003e) to search for the common miRNA/RBP target sites of the 5,669 circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene axes (see Methods). We thus identified 3,141 circRNA-miRNA-\u003cem\u003etrans\u003c/em\u003e-eGene axes (associated with 958 circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene axes; Fig. 5a, middle) and 54 circRNA-RBP-\u003cem\u003etrans\u003c/em\u003e-eGene axes (associated with 30 circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene axes; Fig. 5a, right), respectively (Supplementary Table 5). Regarding the 257 abovementioned circQTL-circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene axes that passed both the mediation test and CIT (see Fig. 4a, right and Fig. 5b, left), we further found that 90 axes (Fig. 5b, right) overlapped with the 958 circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene axes with common miRNA target sites of the circRNAs and \u003cem\u003etrans\u003c/em\u003e-eGenes and no axis overlapped with the 30 circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene axes with common RBP target sites of the circRNAs and \u003cem\u003etrans\u003c/em\u003e-eGenes. Accordingly, 158 circQTL-circRNA-miRNA-\u003cem\u003etrans\u003c/em\u003e-eQTL axes that might be causally related to ASD diagnosis were generated (Figs. 5b and 5c; Supplementary Table 5). The 158 axes were associated with 28 circQTLs, 6 circRNA, 17 miRNAs, and 34 \u003cem\u003etrans\u003c/em\u003e-eGenes (Fig. 5b, right). Of note, 28 of the 34 \u003cem\u003etrans\u003c/em\u003e-eGenes were ASD-relevant genes and one (\u003cem\u003eANK3\u003c/em\u003e) was a top SFARI gene with score = 1 (Fig. 5c). A previous study reported that genetic variations in \u003cem\u003eANK3\u003c/em\u003e were associated with autism susceptibility \u003csup\u003e50\u003c/sup\u003e. \u003cem\u003eANK3\u003c/em\u003e can regulate the structure and function of glutamatergic synapses \u003csup\u003e51\u003c/sup\u003e. Alteration of \u003cem\u003eANK3\u003c/em\u003e expression can regulate neuronal microtubule dynamics functions and affect the corresponding role in psychiatric illness \u003csup\u003e52\u003c/sup\u003e. Taken together, we provided a framework for detecting potentially causal relationships between genetics and trait and generating the corresponding regulatory axes in ASD.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThe majority of eQTL studies to date have focused on investigation of \u003cem\u003ecis\u003c/em\u003e-genetic effects on gene expression, although \u003cem\u003etrans\u003c/em\u003e-eQTL effects may contribute most trait heritability \u003csup\u003e17\u003c/sup\u003e. One challenge for genome-wide scanning for \u003cem\u003etrans\u003c/em\u003e-eQTLs is the heavy burden of multiple testing \u003csup\u003e13,14\u003c/sup\u003e. This study focused on detecting the circQTLs that \u003cem\u003ecis\u003c/em\u003e-regulated expression of nearby circRNAs and also \u003cem\u003etrans\u003c/em\u003e-regulated expression of distant genes simultaneously. We first identified circQTLs and then searched for \u003cem\u003etrans\u003c/em\u003e-eGenes based on these circQTL SNPs (not all SNPs extracted), greatly decreasing the burden of multiple \u003cem\u003etrans\u003c/em\u003e-eQTL testing for the associations between SNPs and gene expression. We thereby identified 546 circQTLs that were also \u003cem\u003etrans\u003c/em\u003e-eQTLs and generated 43,372 circQTL-\u003cem\u003etrans\u003c/em\u003e-eGene associations (Fig. 2a).\u003c/p\u003e\n\u003cp\u003eSince circRNAs often serve as an upstream regulator of mRNAs \u003csup\u003e20,21\u003c/sup\u003e, we asked whether in some cases the effects of \u003cem\u003etrans\u003c/em\u003e-eQTLs on \u003cem\u003etrans\u003c/em\u003e-eGene expression were mediated by the circRNA expression. The mediation analysis suggested that 45% (19,393) of the 43,372 circQTL-\u003cem\u003etrans\u003c/em\u003e-eGene associations were mediated by the expression levels of the corresponding circRNAs. Some mediation test-passing pairs exhibited that more than 50% of the circQTL-\u003cem\u003etrans\u003c/em\u003e-eGene association were mediated by the expression of circRNAs, supporting the importance of the mediation effects for \u003cem\u003etrans\u003c/em\u003e-eQTLs (Fig. 2b). In addition, we observed that circQTLs residing in the flanking sequences of back-splice sites were enriched in RCSs and RBP binding sites (Figs. 1f and 1g), both of which are important for circularization. These observations imply a possible regulatory mechanism for the circQTLs (\u003cem\u003etrans\u003c/em\u003e-eQTLs) passing the mediation test, wherein circQTLs affect the formation of their nearby circRNAs and the circRNAs, as an upstream regulator of the corresponding \u003cem\u003etrans\u003c/em\u003e-eGenes, consequently regulate the \u003cem\u003etrans\u003c/em\u003e-eGene expression. This is worth further investigation.\u003c/p\u003e\n\u003cp\u003eTo increase our understanding of the association between genetic variants and complex traits/diseases such as ASD, previous genome-wide association studies (GWASs) have provided valuable insights into the role of genetic etiology of ASD \u003csup\u003e2,53\u003c/sup\u003e. Recent transcriptome-wide association studies (TWASs) further integrated GWAS loci with gene expression data to investigate \u003cem\u003ecis\u003c/em\u003e-effects on gene expression associated with risk loci for ASD \u003csup\u003e54,55,56\u003c/sup\u003e. However, TWASs do not clarify the major cause of the association between eQTLs and the expression of the corresponding eGenes because the direction of association between eGene expression and trait is unclear. This leads to three possible models (i.e., eGene mediation model, reverse causality model, and independent model) to explain the relationships between an eQTL, an eGene, and a trait \u003csup\u003e19,57,58\u003c/sup\u003e. For the first model, the eGene acts as a causal mediator for the eQTL-trait associations, wherein the eQTL regulates the trait through mediating the eGene expression. For the second model the trait acts as a causal mediator, wherein alteration of the eGene expression is the consequence of the trait. For the third model, the effects of eQTL on eGene expression and trait are independent. This study focuses on identification of eGenes that are causally related to idiopathic ASD (i.e., the first model). To this end, we utilized CIT \u003csup\u003e38,39\u003c/sup\u003e to examine the 43,372 circQTL-\u003cem\u003etrans\u003c/em\u003e-eGene pairs and identified 1,000 pairs with\u003cem\u003e trans\u003c/em\u003e-eGenes as a potentially causal mediator for ASD diagnosis (ASD vs. non-ASD) (Fig. 3a). By definition of CIT, a CIT-passing pair should simultaneously satisfy four conditions (see Methods), especially for the two conditions: circQTLs should be independent of ASD diagnosis after adjusting for \u003cem\u003etrans\u003c/em\u003e-eGene expression and \u003cem\u003etrans\u003c/em\u003e-eGene expression should be associated with ASD diagnosis after adjusting for circQTLs. Therefore, upon passing the CIT screening, both reverse causality and independent models are less probable for the explanation of the CIT-passing circQTL-\u003cem\u003etrans\u003c/em\u003e-eGene associations.\u003c/p\u003e\n\u003cp\u003eWe proceeded to examine the \u003cem\u003etrans\u003c/em\u003e-eGenes (708 genes) involved in the CIT-passing circQTL-\u003cem\u003etrans\u003c/em\u003e-eGene pairs and found significant enrichment for ASD-relevant genes such as SFARI genes, genes encoding PSD proteins, targets of FMR1, RBFOX1, and ELAVL1, and DEGs in ASD, but not for genes implicated in monogenetic forms of other brain disorders (Fig. 3b). Like SFARI genes, the CIT-passing \u003cem\u003etrans\u003c/em\u003e-eGenes had high probabilities of ASD risk (Fig. 3c), high pLI scores (Fig. 3d) and similar patterns of DEG distribution among cell types (Fig. 3e). Particularly, we showed that the CIT-passing \u003cem\u003etrans\u003c/em\u003e-eGenes exhibited significant enrichment for DEGs in neurons in the upper neocortical layer such as IN-VIP, L4, and L2/3 (Fig. 3e). A previous mouse study had demonstrated that excess upper-layer pyramidal neurons can result in an excitatory-inhibitory neuron imbalance, leading to behavioral abnormality similar to the features in ASD \u003csup\u003e59\u003c/sup\u003e. L2/3 neurons are known to be important for underpinning high-level cognitive functions and implicated in cognitive syndromes of high-level associative dysfunction such as ASD \u003csup\u003e60\u003c/sup\u003e. These observations support the important role of these CIT-passing \u003cem\u003etrans\u003c/em\u003e-eGenes in ASD susceptibility.\u003c/p\u003e\n\u003cp\u003eOf the 43,372 circQTL-\u003cem\u003etrans\u003c/em\u003e-eGene pairs (associated with 546 circQTLs), although the circQTL SNPs did not reached genome-wide significance in the largest ASD GWAS \u003csup\u003e2\u003c/sup\u003e currently, 217 of 546 (40%) circQTLs were associated with the expression of at least one ASD risk genes (SFARI genes; see Supplementary Fig. 4 and Supplementary Table 6). The 217 circQTLs resided at 40 distinct LD blocks, 133 of which (residing at 30 LD blocks) affected the expression of even more than 50 \u003cem\u003etrans\u003c/em\u003e-eGenes. The 30 LD blocks can be regarded as \u003cem\u003etrans\u003c/em\u003e-eQTL hotspots associated with SFARI genes. Of the 133 circQTLs, 25 circQTLs were particularly connected with ASD risk genes, in which the affected \u003cem\u003etrans\u003c/em\u003e-eGenes were significantly enriched for SFARI genes (all \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05 by hypergeometric test; Supplementary Table 6). This result provides a useful resource for further investigating regulatory mechanisms of the corresponding ASD risk genes. Intriguingly, we found that five \u003cem\u003etrans\u003c/em\u003e-eGenes (\u003cem\u003eCRHR1\u003c/em\u003e,\u003cem\u003e XKR6\u003c/em\u003e,\u003cem\u003e NKX2-2\u003c/em\u003e,\u003cem\u003e NSF\u003c/em\u003e, and \u003cem\u003eARL17B\u003c/em\u003e) were previously prioritized as TWAS-significant genes/transcripts in ASD in two recent studies \u003csup\u003e55,56\u003c/sup\u003e (Table 2). This revealed that these TWAS-significant genes/transcripts might be regulated by not only \u003cem\u003ecis\u003c/em\u003e-acting SNPs but also \u003cem\u003etrans\u003c/em\u003e-acting ones. Of note, \u003cem\u003eNSF\u003c/em\u003e was also a \u003cem\u003etrans\u003c/em\u003e-eGene involved in the CIT-passing rs182920567-\u003cem\u003eNSF\u003c/em\u003e-ASD diagnosis propagation path, supporting that\u003cem\u003e NSF\u003c/em\u003e may be causally related to ASD diagnosis. This reflects a previous report that \u003cem\u003eNSF\u003c/em\u003e may play a role in ASD pathophysiology through mediating serotonin transporter trafficking \u003csup\u003e61\u003c/sup\u003e. In addition, some circQTL-\u003cem\u003etrans\u003c/em\u003e-eGene associations passed the mediation test, wherein the effects on \u003cem\u003etrans\u003c/em\u003e-eGene expression (\u003cem\u003eCRHR1\u003c/em\u003e, \u003cem\u003eXKR6\u003c/em\u003e, and \u003cem\u003eARL17B\u003c/em\u003e) were mediated by the circRNA expression (circERC1 and circSPHKAP). Two phenomenon were observed. First, the expression of a TWAS-significant gene may be regulated by multiple circRNAs. We found circERC1-1 and circSPHKAP with more than 50% (or even more than 70%) of mediation effects on the relations of \u003cem\u003etrans\u003c/em\u003e-eQTLs to \u003cem\u003eCRHR1\u003c/em\u003e, suggesting multiple upstream regulators of \u003cem\u003eCRHR1\u003c/em\u003e. Second, two TWAS-significant genes/transcripts (\u003cem\u003eXKR6\u003c/em\u003e and \u003cem\u003eARL17B\u003c/em\u003e) may be under the same regulatory control (rs572758241-circSPHKAP association). Of note, rs572758241 was located at a \u003cem\u003etrans\u003c/em\u003e-eQTL hotspot and affected the expression of 724 \u003cem\u003etrans\u003c/em\u003e-eGenes. The 724 genes significantly overlap with SFARI genes (\u003cem\u003eP\u003c/em\u003e = 0.01 by hypergeometric test; Supplementary Table 6), indicating the relevance of the \u003cem\u003etrans\u003c/em\u003e-eQTL to ASD. These results thus provided potential upstream regulators of these TWAS-significant genes/transcripts. \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eBy integrating the circQTL-\u003cem\u003etrans\u003c/em\u003e-eGene pairs that passed CIT with the circQTL-circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene axes that passed mediation test, we further detected 257 potential circQTL-circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene-ASD diagnosis propagation paths (Fig. 4a). Meanwhile, we searched for the common miRNA target sites of the corresponding circRNAs and \u003cem\u003etrans\u003c/em\u003e-eGenes and generated potential regulatory associations between the circRNAs and the \u003cem\u003etrans\u003c/em\u003e-eGenes. Combined the predicted circRNA-miRNA-trans-eGene axes with the 257 potential circQTL-circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene-trait propagation paths, we finally identified 158 circQTL-circRNA-miRNA-\u003cem\u003etrans\u003c/em\u003e-eQTL-trait propagation paths (Fig. 5b). Of the 158 axes, it is possible to further utilize the mediation analysis to examine whether the circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene associations were mediated by the expression of miRNAs, if both mRNA and miRNA expressing data derived from the same samples are available. For example, a circQTL-circRNA-miRNA-\u003cem\u003etrans\u003c/em\u003e-eGene-trait propagation path that passes CIT and two layers of mediation test was detected according to miRNA expressing data\u003csup\u003e6\u003c/sup\u003e derived from the same ASD and control samples used in this study (Supplementary Fig. 5).\u003c/p\u003e\n\u003cp\u003eOn the basis of a large brain sample size of both RNA-seq data and the corresponding genotyping data from ASD cases and non-ASD controls, this study is the first report, to the best of our knowledge, for systematically investigating \u003cem\u003etrans\u003c/em\u003e-genetic effects of circQTLs and inferring the corresponding causal relations in diseases. The identified ASD-associated regulatory interactions, particularly the CIT-passing axes involving \u003cem\u003etrans\u003c/em\u003e-eGenes that were previously implicated in idiopathic ASD (e.g., SFARI genes), may serve as a useful resource for further investigations of more complex regulatory mechanisms underlying ASD. Future studies with larger patient cohorts, including genome-wide transcriptomic and genomic sequencing data, will allow for more precise causal relationships between genetic variants and transcriptional expression changes in ASD. The provided framework can be applied to future analysis of eQTL-trait relations in other complex diseases.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003e\u003cstrong\u003eIdentification of circQTLs/\u003cem\u003etrans\u003c/em\u003e-eQTLs. \u003c/strong\u003eThe RNA-seq data of ASD and non-ASD brain samples\u003csup\u003e5\u003c/sup\u003e and the corresponding genotyping data\u003csup\u003e26\u003c/sup\u003e were obtained from Synapse (http://www.synapse.org) with permission under the accession number syn4587609. The circRNAs were extracted from our previous study \u003csup\u003e10\u003c/sup\u003e, which were identified by NCLscan \u003csup\u003e62\u003c/sup\u003e according to the same Synapse RNA-seq data used in this study and the Ensembl annotation (version 90) based on the human reference genome (GRCh38). A sample was excluded in our analysis if the number of identified circRNAs of the sample was below one standard deviation from the mean of the sample set \u003csup\u003e10\u003c/sup\u003e. A total of 212 samples, including 73 FC, 61 TC, and 78 CV samples, were retained (Table 1). For accuracy, we only considered the 1,060 circRNAs that were detected in more than 50% of the samples examined. The genotyping data were derived from 79 individuals, including 40 ASD patients and 39 non-ASD-affected controls \u003csup\u003e26\u003c/sup\u003e. Since eQTL analysis required both RNA-seq and genotyping data from the same individuals, RNA-seq data from 165 samples were examined in the following analyses (Fig. 1a, Table 1, and Supplementary Table 1). The expression levels of circRNAs were calculated using the number of supporting circRNA junction reads per million uniquely mapped reads (RPM) \u003csup\u003e27\u003c/sup\u003e and adjusted for biological/technical covariates including the corresponding host gene expression, brain region, diagnosis (ASD/non-ASD), age, sex, RNA quality (RNA integrity number; RIN), post-mortem interval (PMI), sequencing batch, and brain bank (Supplementary Table 1). The expression levels of genes were downloaded from Parikshak et al.\u0026rsquo;s study\u003csup\u003e5\u003c/sup\u003e, which were measured by log\u003csub\u003e2\u003c/sub\u003e normalized fragments per kilobase million mapped reads (FPKM) values. Imputation genotyping data (39,166,259 SNPs) were downloaded using the Synapse command line client. We excluded SNPs with imputed quality score (R\u003csup\u003e2\u003c/sup\u003e) \u0026lt; 0.8 and minor allele frequency (MAF) \u0026lt; 0.01 and thus retained 7,714,475 SNPs for the following eQTL analysis. We limited our analysis to SNPs in \u0026plusmn; 200 kb nucleotides of each back-splice site and evaluated circQTLs by testing the correlations between the imputed genotype dosages and covariate-adjusted circRNA expression using Matrix eQTL with an additive linear model \u003csup\u003e28\u003c/sup\u003e for adjusting for covariates including sex, age, diagnosis, brain region, RIN, PMI, sequencing batch, brain bank batch. To control for multiple testing, \u003cem\u003eP\u003c/em\u003e values will be adjusted across all circQTLs using false positive rate (FDR) correction (i.e., Benjamini-Hochberg correction). At FDR\u0026lt;0.05, we thus identified 989 circQTLs based on the 165 samples from all three brain regions and 666 circQTLs based on the 105 cortex samples, respectively (Supplementary Table 2). To compare the effect sizes for the 989 circQTL-circRNA pairs from different brain regions (TC, FC, or CV), we conducted a Z-test and calculated Z-scores as\u003c/p\u003e\u003cp\u003e\u003cimg src=\"https://myfiles.space/user_files/58653_1b1c6aeb34a62c68/58653_custom_files/img1612470283.jpg\"\u003e\u003c/p\u003e\u003cp\u003eHere the equation showed an example for TC-FC comparison Z-test. SE represents standard error. We examined whether the 666 circQTLs were also \u003cem\u003etrans\u003c/em\u003e-eQTLs affecting the expression of \u003cem\u003etrans\u003c/em\u003e-eGenes using Matrix eQTL with an additive linear model for adjusting for the covariates abovementioned. Such circQTL SNPs and \u003cem\u003etrans\u003c/em\u003e-eGenes should be located on different chromosomes or the same chromosome separated by a distance greater than 5Mb (from the SNP site to the transcription start site of the \u003cem\u003etrans\u003c/em\u003e-eGene). At FDR\u0026lt;0.05, 546 circQTLs were identified to be also \u003cem\u003etrans\u003c/em\u003e-eQTLs (Fig. 2a; Supplementary Table 3).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDetection of reverse complementary sequences (RCSs). \u003c/strong\u003eThe criteria for detecting RCSs were similar to those presented in a previous study \u003csup\u003e63\u003c/sup\u003e. For a circRNA, both flanking sequences (\u0026plusmn; 200 kb nucleotides of the back-splice site) were aligned each other using BLAST\u003csup\u003e64\u003c/sup\u003e with parameters \u0026ndash;task blastn \u0026ndash;word_size 11 \u0026ndash;strand minus. The potential RCSs should be simultaneously satisfied the following rules: bitscore\u0026gt; 100, alignment length\u0026gt;50 bp, and identity\u0026gt;80%.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDetermination of miRNA and RBP binding sites. \u003c/strong\u003eCLIP-supported RBP binding sites on hg19 assembly were downloaded from ENCORI\u003csup\u003e49\u003c/sup\u003e at http://starbase.sysu.edu.cn/. The liftOver tool \u003csup\u003e65\u003c/sup\u003e was employed to obtain the genomic coordinates of RBP binding sites on the GRCh38 assembly. The miRNA-circRNA, miRNA-mRNA, RBP-circRNA, and RBP-mRNA interactions supported by CLIP-seq data were also extracted from ENCORI. These data were retrieved through ENCORI-provided Web APIs. The used command lines were listed in Supplementary Table 7. Of note, a considered miRNA (or RBP) binding site should be supported by at least one AGO CLIP-seq (or CLIP-seq) experiments. CircRNA-miRNA-mRNA and circRNA-RBP-mRNA interactions were determined according to the common miRNA and RBP target sites of the circRNAs and mRNAs, respectively. A developed package, CircMiMi, for constructing CLIP-seq data-supported circRNA-miRNA-mRNA and circRNA-RBP-mRNA interactions was also provided (https://github.com/TreesLab/CircMiMi). In CircMiMi, users can input their identified circRNAs, not limited to ENCORI-provided circRNAs.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEnrichment analyses based on empirical sampling. \u003c/strong\u003eA sampling procedure was conducted to test whether the 666 circQTL SNPs (associated with 48 circRNAs) had a higher proportion (\u003cem\u003eproportion\u003csub\u003eobs\u003c/sub\u003e\u003c/em\u003e) of SNPs located in RCSs residing in the introns flanking circularized exons compared to an empirical distribution of the proportion observed in the 10,000 times of random sampling (see Fig. 1f). For each time, the equivalent number (666) of SNPs were randomly selected from the background set (823,378 SNPs) and the proportion (\u003cem\u003eproportion\u003csub\u003ei\u003c/sub\u003e\u003c/em\u003e) of SNPs located in RCSs was calculated. The 823,378 SNPs were located in the flanking sequences (\u0026plusmn; 200 kb nucleotides of the back-splice site) for the 48 circRNAs examined. The empirical \u003cem\u003eP\u003c/em\u003e value was calculated as\u003c/p\u003e\u003cp\u003e\u003cimg src=\"https://myfiles.space/user_files/58653_1b1c6aeb34a62c68/58653_custom_files/img1612470332.jpg\"\u003e\u003c/p\u003e\u003cp\u003eA similar procedure was designed to test whether the 666 circQTL SNPs were predominantly located in RBP binding sites residing in the introns flanking circularized exons for the 48 circRNAs examined (Fig. 1g). Regarding the 43,372 circQTL-\u003cem\u003etrans\u003c/em\u003e-eGene pairs (associated with 47 circRNAs, 7,165 \u003cem\u003etrans\u003c/em\u003e-eGenes, and 13,379 circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene pairs; see Fig. 2a), we conducted a sampling procedure to test whether the 13,379 circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene pairs had a higher proportion of pairs with a significant correlation (\u003cem\u003eP\u003c/em\u003e\u0026lt;0.05 by Spearman\u0026rsquo;s correlation test) between expression of the circRNAs and the \u003cem\u003etrans\u003c/em\u003e-eGenes compared to an empirical distribution of the proportion observed in the 10,000 times of random sampling. For each time, the equivalent number (13,379) of circRNA-gene pairs was randomly selected from the 47 circRNAs and the 7,165 \u003cem\u003etrans\u003c/em\u003e-eGenes. To test whether the CIT-passing \u003cem\u003etrans\u003c/em\u003e-eGenes (708 genes) had a significantly higher probability of ASD risk\u003csup\u003e45\u003c/sup\u003e than background gene set (20,070 protein-coding genes), an empirical gene enrichment analysis was conducted on the basis of an empirical distribution of the proportion observed in the 10,000 times of random sampling (Supplementary Fig. 3a). The 20,070 genes were selected because they were expressed in the cortex on the basis of the RNA-seq data examined in this study. The equivalent number (708) of genes was randomly selected from the 20,070 genes for each time. A similar procedure was conducted to test whether the CIT-passing \u003cem\u003etrans\u003c/em\u003e-eGenes were more intolerant of a loss function mutation (measured by pLI scores \u003csup\u003e46\u003c/sup\u003e) compared with the background gene set (Supplementary Fig. 3b). The probability of ASD risk and pLI score for each gene were downloaded from http://asd.princeton.edu/ and https://gnomad.broadinstitute.org/ (loss-of-function curation results; v2.1.1), respectively.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMediation and causal testing. \u003c/strong\u003eWe conducted mediations analysis using the mediation package\u003csup\u003e37\u003c/sup\u003e downloaded from the Comprehensive R Archive Network (CRAN) at https://cran.r-project.org/web/packages/mediation/index.html. To test the mediation effects of \u003cem\u003etrans\u003c/em\u003e-eQTLs, the mediation package was performed with circQTL (\u003cem\u003etrans\u003c/em\u003e-eQTL) as the \u0026ldquo;exposure,\u0026rdquo; circRNA expression as the \u0026ldquo;mediator,\u0026rdquo; \u003cem\u003etrans\u003c/em\u003e-eGene expression as the \u0026ldquo;outcome,\u0026rdquo; and adjusting for covariates including sex, age, diagnosis, brain region, RIN, PMI, sequencing batch, and brain bank batch. The proportion of mediation for the circQTL was the proportion of causal mediation effect in the total effect (i.e., causal mediation effect + direct effect) \u003csup\u003e37\u003c/sup\u003e. To test whether the circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene associations were mediated by the expression of miRNAs, we first extracted miRNA expressing data based on the 63 cortex samples (30 ASD and 33 non-ASD samples) overlapped with the samples examined in our eQTL analysis. The miRNA expressing data were obtained by the request from the authors of the Wu et al.\u0026rsquo;s study \u003csup\u003e6\u003c/sup\u003e, which were measured by log\u003csub\u003e2\u003c/sub\u003e normalized read counts with adjusting for library size, GC content, RIN, PMI, and batch bank. The mediation package was then performed with circRNA expression as the \u0026ldquo;exposure,\u0026rdquo; miRNA expression as the \u0026ldquo;mediator,\u0026rdquo; \u003cem\u003etrans\u003c/em\u003e-eGene expression as the \u0026ldquo;outcome,\u0026rdquo; and adjusting for covariates abovementioned. Significant mediation effects were defined at a simulation threshold of \u003cem\u003eP\u003c/em\u003e \u0026lt; 0.05 (sims=1,000) and FDR (Storey and Tibshirani method \u003csup\u003e66\u003c/sup\u003e) \u0026lt; 0.05.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe CIT package\u003csup\u003e38,39\u003c/sup\u003e was downloaded from CRAN at https://cran.r-project.org/web/packages/cit/index.html. A CIT-passing circQTL-\u003cem\u003etrans\u003c/em\u003e-eGene pair (CIT \u003cem\u003eP\u003c/em\u003e\u0026lt;0.05 and FDR (\u003cem\u003eq\u003c/em\u003e) \u0026lt;0.05) (Supplementary Table 4) represents that the circQTL has a causal effect on the diagnosis status (ASD vs. non-ASD) through the \u003cem\u003etrans\u003c/em\u003e-eGene expression, which should simultaneously satisfy the following conditions: (1) the circQTL is associated with the diagnosis status, (2) the circQTL is associated with the \u003cem\u003etrans\u003c/em\u003e-eGene mediator after adjusting for the diagnosis status, (3) the \u003cem\u003etrans\u003c/em\u003e-eGene mediator is associated with the diagnosis status after adjusting for the circQTL, and (4) the circQTL is independent of the diagnosis status after adjusting for the \u003cem\u003etrans\u003c/em\u003e-eGene mediator. The \u003cem\u003eP\u003c/em\u003e value of CIT is defined as the maximum of the above four-component test \u003cem\u003eP\u003c/em\u003e values. The FDR (\u003cem\u003eq\u003c/em\u003e) values were further estimated using CIT-provided permutation test with default parameters. Illustration of networks was plotted by the Cytoscape package (https://cytoscape.org/).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunctional analysis of the CIT-passing \u003cem\u003etrans\u003c/em\u003e-eGenes. \u003c/strong\u003eThe SFARI\u003csup\u003e40\u003c/sup\u003e gene list was downloaded from https://gene.sfari.org/ (SFARI 08-07-2020 release). The lists of genes encoding PSD proteins and targets of FMRP, RBFOX1, and ELAVL1 were downloaded from Lee et al.\u0026rsquo;s study \u003csup\u003e26\u003c/sup\u003e. The list of DEGs in ASD was downloaded from Parikshak et al.\u0026rsquo;s study \u003csup\u003e5\u003c/sup\u003e. The lists of genes associated with Alzheimer\u0026rsquo;s disease, amyotrophic lateral sclerosis, Huntington\u0026rsquo;s disease, intellectual disability, and Parkinson\u0026rsquo;s disease were downloaded from Wang et al.\u0026rsquo;s study \u003csup\u003e67\u003c/sup\u003e. The schizophrenia gene list was downloaded from a GWAS report\u003csup\u003e68\u003c/sup\u003e. The list of DEGs in a cell type-specific manner was downloaded from Velmeshev et al.\u0026rsquo;s study\u003csup\u003e15\u003c/sup\u003e. For gene set enrichment analyses in Figure 3b, we used the abovementioned 20,070 protein-coding genes as the background set. We took the analysis of SFARI gene enrichment for the CIT-passing \u003cem\u003etrans\u003c/em\u003e-eGenes as an example. We created a two-way contingency table with rows containing numbers of SFARI and non-SFARI genes and columns containing numbers of CIT-passing \u003cem\u003etrans\u003c/em\u003e-eGenes and genes that were not CIT-passing \u003cem\u003etrans\u003c/em\u003e-eGenes. The statistical significance and odds ratio were determined using a one-tailed Fisher\u0026rsquo;s exact test with the \u003cem\u003efisher.test\u003c/em\u003e R function. \u003cem\u003eP\u003c/em\u003e values were FDR adjusted across 13 target groups for each gene list using Bonferroni correction. GO analysis was performed using the ToppFun module of ToppGene Suite software \u003csup\u003e69\u003c/sup\u003e at https://toppgene.cchmc.org/enrichment.jsp (version 2020-Sep-08 01:39). False discovery rate (FDR) adjusted \u003cem\u003eP\u003c/em\u003e values were calculated using Bonferroni correction.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eHypergeometric testing. \u003c/strong\u003eFor a circQTL (\u003cem\u003etrans\u003c/em\u003e-eQTL), we conducted a hypergeometric test to evaluate the significance of overlap of the \u003cem\u003etrans\u003c/em\u003e-eGenes regulated by the circQTL and SFARI genes. Only the circQTLs that affected the expression of more than 50 \u003cem\u003etrans\u003c/em\u003e-eGenes were considered for the test (Supplementary Table 6). We used 20,070 protein-coding genes as the background set, which involved 951 SFARI genes. The statistical significance was determined using the \u003cem\u003ephyper\u003c/em\u003e R function. The coordinates of LD blocks were downloaded from the DistiLD database \u003csup\u003e70\u003c/sup\u003e at http://distild.jensenlab.org.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData availability. \u003c/strong\u003eThe CircMiMi package for constructing circRNA-miRNA-mRNA and circRNA-RBP-mRNA interactions is available at https://github.com/TreesLab/CircMiMi. The identified circQTLs/\u003cem\u003etrans\u003c/em\u003e-eQTLs and the related data/codes were deposited in Supplementary Tables 1-7 or GitHub at https://github.com/TreesLab/circQTL_ASD.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis work was supported by Genomics Research Center (GRC), Academia Sinica, Taiwan; the Ministry of Science and Technology (MOST), Taiwan, under the contract MOST 108-2311-B-001-020-MY3 (TJC); and the Taipei, Taichung, Kaohsiung Veterans General Hospital, Tri-Service General Hospital, Academia Sinica Joint Research Program under the contract AS-VTA-109-03 (TJC).\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eT.-L.M. performed the bioinformatics assays, analyzed the data, and helped writing the manuscript. C-Y.C. performed the statistical assays and analyzed the data. T.-W.C. developed and maintained the CircMiMi package. T.-J.C. conceived the study, analyzed the data, and wrote the article.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interest:\u003c/strong\u003e The authors declare no competing financial interests.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eAnsel A, Rosenzweig JP, Zisman PD, Melamed M, Gesundheit B. Variation in Gene Expression in Autism Spectrum Disorders: An Extensive Review of Transcriptomic Studies. \u003cem\u003eFrontiers in Neuroscience\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 601 (2016).\u003c/li\u003e\n\u003cli\u003eGrove J\u003cem\u003e, et al.\u003c/em\u003e Identification of common genetic risk variants for autism spectrum disorder. \u003cem\u003eNat Genet\u003c/em\u003e \u003cstrong\u003e51\u003c/strong\u003e, 431-444 (2019).\u003c/li\u003e\n\u003cli\u003eSatterstrom FK\u003cem\u003e, et al.\u003c/em\u003e Large-Scale Exome Sequencing Study Implicates Both Developmental and Functional Changes in the Neurobiology of Autism. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e180\u003c/strong\u003e, 568-584 e523 (2020).\u003c/li\u003e\n\u003cli\u003eGupta S\u003cem\u003e, et al.\u003c/em\u003e Transcriptome analysis reveals dysregulation of innate immune response genes and neuronal activity-dependent genes in autism. \u003cem\u003eNat Commun\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 5748 (2014).\u003c/li\u003e\n\u003cli\u003eParikshak NN\u003cem\u003e, et al.\u003c/em\u003e Genome-wide changes in lncRNA, splicing, and regional gene expression patterns in autism. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e540\u003c/strong\u003e, 423-427 (2016).\u003c/li\u003e\n\u003cli\u003eWu YE, Parikshak NN, Belgard TG, Geschwind DH. Genome-wide, integrative analysis implicates microRNA dysregulation in autism spectrum disorder. \u003cem\u003eNat Neurosci\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, 1463-1476 (2016).\u003c/li\u003e\n\u003cli\u003eMor M, Nardone S, Sams DS, Elliott E. Hypomethylation of miR-142 promoter and upregulation of microRNAs that target the oxytocin receptor gene in the autism prefrontal cortex. \u003cem\u003eMol Autism\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 46 (2015).\u003c/li\u003e\n\u003cli\u003eTran SS\u003cem\u003e, et al.\u003c/em\u003e Widespread RNA editing dysregulation in brains from autistic individuals. \u003cem\u003eNat Neurosci\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 25-36 (2019).\u003c/li\u003e\n\u003cli\u003eGokool A, Anwar F, Voineagu I. The Landscape of Circular RNA Expression in the Human Brain. \u003cem\u003eBiol Psychiatry\u003c/em\u003e \u003cstrong\u003e87\u003c/strong\u003e, 294-304 (2020).\u003c/li\u003e\n\u003cli\u003eChen YJ\u003cem\u003e, et al.\u003c/em\u003e Genome-wide, integrative analysis of circular RNA dysregulation and the corresponding circular RNA-microRNA-mRNA regulatory axes in autism. \u003cem\u003eGenome Res\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 375-391 (2020).\u003c/li\u003e\n\u003cli\u003eRockman MV, Kruglyak L. Genetics of global gene expression. \u003cem\u003eNat Rev Genet\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, 862-872 (2006).\u003c/li\u003e\n\u003cli\u003eNica AC, Dermitzakis ET. Expression quantitative trait loci: present and future. \u003cem\u003ePhilos Trans R Soc Lond B Biol Sci\u003c/em\u003e \u003cstrong\u003e368\u003c/strong\u003e, 20120362 (2013).\u003c/li\u003e\n\u003cli\u003ePierce BL\u003cem\u003e, et al.\u003c/em\u003e Mediation analysis demonstrates that trans-eQTLs are often explained by cis-mediation: a genome-wide analysis among 1,800 South Asians. \u003cem\u003ePLoS Genet\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, e1004818 (2014).\u003c/li\u003e\n\u003cli\u003eGrundberg E\u003cem\u003e, et al.\u003c/em\u003e Mapping cis- and trans-regulatory effects across multiple tissues in twins. \u003cem\u003eNat Genet\u003c/em\u003e \u003cstrong\u003e44\u003c/strong\u003e, 1084-1089 (2012).\u003c/li\u003e\n\u003cli\u003eLiu X\u003cem\u003e, et al.\u003c/em\u003e Functional Architectures of Local and Distal Regulation of Gene Expression in Multiple Human Tissues. \u003cem\u003eAm J Hum Genet\u003c/em\u003e \u003cstrong\u003e100\u003c/strong\u003e, 605-616 (2017).\u003c/li\u003e\n\u003cli\u003eSun BB\u003cem\u003e, et al.\u003c/em\u003e Genomic atlas of the human plasma proteome. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e558\u003c/strong\u003e, 73-79 (2018).\u003c/li\u003e\n\u003cli\u003eLiu X, Li YI, Pritchard JK. Trans Effects on Gene Expression Can Drive Omnigenic Inheritance. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e177\u003c/strong\u003e, 1022-1034 e1026 (2019).\u003c/li\u003e\n\u003cli\u003eYang F, Wang J, Consortium GT, Pierce BL, Chen LS. Identifying cis-mediators for trans-eQTLs across many human tissues using genomic mediation analysis. \u003cem\u003eGenome Res\u003c/em\u003e \u003cstrong\u003e27\u003c/strong\u003e, 1859-1871 (2017).\u003c/li\u003e\n\u003cli\u003eYao C\u003cem\u003e, et al.\u003c/em\u003e Dynamic Role of trans Regulation of Gene Expression in Relation to Complex Traits. \u003cem\u003eAm J Hum Genet\u003c/em\u003e \u003cstrong\u003e100\u003c/strong\u003e, 571-580 (2017).\u003c/li\u003e\n\u003cli\u003eChen I, Chen CY, Chuang TJ. Biogenesis, identification, and function of exonic circular RNAs. \u003cem\u003eWiley Interdiscip Rev RNA\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 563-579 (2015).\u003c/li\u003e\n\u003cli\u003eChen LL. The expanding regulatory mechanisms and cellular functions of circular RNAs. \u003cem\u003eNat Rev Mol Cell Biol\u003c/em\u003e, (2020).\u003c/li\u003e\n\u003cli\u003eChen LL, Yang L. Regulation of circRNA biogenesis. \u003cem\u003eRNA Biol\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 381-388 (2015).\u003c/li\u003e\n\u003cli\u003eYou X\u003cem\u003e, et al.\u003c/em\u003e Neural circular RNAs are derived from synaptic genes and regulated by development and plasticity. \u003cem\u003eNat Neurosci\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, 603-610 (2015).\u003c/li\u003e\n\u003cli\u003eRybak-Wolf A\u003cem\u003e, et al.\u003c/em\u003e Circular RNAs in the Mammalian Brain Are Highly Abundant, Conserved, and Dynamically Expressed. \u003cem\u003eMol Cell\u003c/em\u003e \u003cstrong\u003e58\u003c/strong\u003e, 870-885 (2015).\u003c/li\u003e\n\u003cli\u003eSzabo L\u003cem\u003e, et al.\u003c/em\u003e Statistically based splicing detection reveals neural enrichment and tissue-specific induction of circular RNA during human fetal development. \u003cem\u003eGenome Biol\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 126 (2015).\u003c/li\u003e\n\u003cli\u003eLee C, Kang EY, Gandal MJ, Eskin E, Geschwind DH. Profiling allele-specific gene expression in brains from individuals with autism spectrum disorder reveals preferential minor allele usage. \u003cem\u003eNat Neurosci\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 1521-1532 (2019).\u003c/li\u003e\n\u003cli\u003eZhang XO, Wang HB, Zhang Y, Lu X, Chen LL, Yang L. Complementary sequence-mediated exon circularization. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e159\u003c/strong\u003e, 134-147 (2014).\u003c/li\u003e\n\u003cli\u003eShabalin AA. Matrix eQTL: ultra fast eQTL analysis via large matrix operations. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e28\u003c/strong\u003e, 1353-1358 (2012).\u003c/li\u003e\n\u003cli\u003eOldham MC\u003cem\u003e, et al.\u003c/em\u003e Functional organization of the transcriptome in human brain. \u003cem\u003eNat Neurosci\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, 1271-1282 (2008).\u003c/li\u003e\n\u003cli\u003eVoineagu I\u003cem\u003e, et al.\u003c/em\u003e Transcriptomic analysis of autistic brain reveals convergent molecular pathology. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e474\u003c/strong\u003e, 380-384 (2011).\u003c/li\u003e\n\u003cli\u003eWillsey AJ\u003cem\u003e, et al.\u003c/em\u003e Coexpression networks implicate human midfetal deep cortical projection neurons in the pathogenesis of autism. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e155\u003c/strong\u003e, 997-1007 (2013).\u003c/li\u003e\n\u003cli\u003eParikshak NN\u003cem\u003e, et al.\u003c/em\u003e Integrative functional genomic analyses implicate specific molecular pathways and circuits in autism. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e155\u003c/strong\u003e, 1008-1021 (2013).\u003c/li\u003e\n\u003cli\u003eLiu Z, Ran Y, Tao C, Li S, Chen J, Yang E. Detection of circular RNA expression and related quantitative trait loci in the human dorsolateral prefrontal cortex. \u003cem\u003eGenome Biol\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, 99 (2019).\u003c/li\u003e\n\u003cli\u003eAhmed I, Karedath T, Al-Dasim FM, Malek JA. Identification of human genetic variants controlling circular RNA expression. \u003cem\u003eRNA\u003c/em\u003e \u003cstrong\u003e25\u003c/strong\u003e, 1765-1778 (2019).\u003c/li\u003e\n\u003cli\u003eJeck WR\u003cem\u003e, et al.\u003c/em\u003e Circular RNAs are abundant, conserved, and associated with ALU repeats. \u003cem\u003eRNA\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, 141-157 (2013).\u003c/li\u003e\n\u003cli\u003eChuang TJ\u003cem\u003e, et al.\u003c/em\u003e Integrative transcriptome sequencing reveals extensive alternative trans-splicing and cis-backsplicing in human cells. \u003cem\u003eNucleic Acids Res\u003c/em\u003e \u003cstrong\u003e46\u003c/strong\u003e, 3671-3691 (2018).\u003c/li\u003e\n\u003cli\u003eTingley D YT, Hirose K, Keele L, Imai K. mediation: R Package for Causal Mediation Analysis. \u003cem\u003eJournal of Statistical Software\u003c/em\u003e \u003cstrong\u003e59\u003c/strong\u003e, 1-38 (2014).\u003c/li\u003e\n\u003cli\u003eMillstein J, Zhang B, Zhu J, Schadt EE. Disentangling molecular relationships with a causal inference test. \u003cem\u003eBMC Genet\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 23 (2009).\u003c/li\u003e\n\u003cli\u003eMillstein J, Chen GK, Breton CV. cit: hypothesis testing software for mediation analysis in genomic applications. \u003cem\u003eBioinformatics\u003c/em\u003e \u003cstrong\u003e32\u003c/strong\u003e, 2364-2365 (2016).\u003c/li\u003e\n\u003cli\u003eAbrahams BS\u003cem\u003e, et al.\u003c/em\u003e SFARI Gene 2.0: a community-driven knowledgebase for the autism spectrum disorders (ASDs). \u003cem\u003eMol Autism\u003c/em\u003e \u003cstrong\u003e4\u003c/strong\u003e, 36 (2013).\u003c/li\u003e\n\u003cli\u003eIossifov I\u003cem\u003e, et al.\u003c/em\u003e The contribution of de novo coding mutations to autism spectrum disorder. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e515\u003c/strong\u003e, 216-221 (2014).\u003c/li\u003e\n\u003cli\u003eDarnell JC\u003cem\u003e, et al.\u003c/em\u003e FMRP stalls ribosomal translocation on mRNAs linked to synaptic function and autism. \u003cem\u003eCell\u003c/em\u003e \u003cstrong\u003e146\u003c/strong\u003e, 247-261 (2011).\u003c/li\u003e\n\u003cli\u003eWeyn-Vanhentenryck SM\u003cem\u003e, et al.\u003c/em\u003e HITS-CLIP and integrative modeling define the Rbfox splicing-regulatory network linked to brain development and autism. \u003cem\u003eCell Rep\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 1139-1152 (2014).\u003c/li\u003e\n\u003cli\u003eMukherjee N\u003cem\u003e, et al.\u003c/em\u003e Integrative regulatory mapping indicates that the RNA-binding protein HuR couples pre-mRNA processing and mRNA stability. \u003cem\u003eMol Cell\u003c/em\u003e \u003cstrong\u003e43\u003c/strong\u003e, 327-339 (2011).\u003c/li\u003e\n\u003cli\u003eKrishnan A\u003cem\u003e, et al.\u003c/em\u003e Genome-wide prediction and functional characterization of the genetic basis of autism spectrum disorder. \u003cem\u003eNat Neurosci\u003c/em\u003e \u003cstrong\u003e19\u003c/strong\u003e, 1454-1462 (2016).\u003c/li\u003e\n\u003cli\u003eLek M\u003cem\u003e, et al.\u003c/em\u003e Analysis of protein-coding genetic variation in 60,706 humans. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e536\u003c/strong\u003e, 285-291 (2016).\u003c/li\u003e\n\u003cli\u003eIossifov I\u003cem\u003e, et al.\u003c/em\u003e Low load for disruptive mutations in autism genes and their biased transmission. \u003cem\u003eProc Natl Acad Sci U S A\u003c/em\u003e \u003cstrong\u003e112\u003c/strong\u003e, E5600-5607 (2015).\u003c/li\u003e\n\u003cli\u003eVelmeshev D\u003cem\u003e, et al.\u003c/em\u003e Single-cell genomics identifies cell type-specific molecular changes in autism. \u003cem\u003eScience\u003c/em\u003e \u003cstrong\u003e364\u003c/strong\u003e, 685-689 (2019).\u003c/li\u003e\n\u003cli\u003eLi JH, Liu S, Zhou H, Qu LH, Yang JH. starBase v2.0: decoding miRNA-ceRNA, miRNA-ncRNA and protein-RNA interaction networks from large-scale CLIP-Seq data. \u003cem\u003eNucleic Acids Res\u003c/em\u003e \u003cstrong\u003e42\u003c/strong\u003e, D92-97 (2014).\u003c/li\u003e\n\u003cli\u003eBi C\u003cem\u003e, et al.\u003c/em\u003e Mutations of ANK3 identified by exome sequencing are associated with autism susceptibility. \u003cem\u003eHum Mutat\u003c/em\u003e \u003cstrong\u003e33\u003c/strong\u003e, 1635-1638 (2012).\u003c/li\u003e\n\u003cli\u003eSmith KR\u003cem\u003e, et al.\u003c/em\u003e Psychiatric risk factor ANK3/ankyrin-G nanodomains regulate the structure and function of glutamatergic synapses. \u003cem\u003eNeuron\u003c/em\u003e \u003cstrong\u003e84\u003c/strong\u003e, 399-415 (2014).\u003c/li\u003e\n\u003cli\u003eGarza JC\u003cem\u003e, et al.\u003c/em\u003e Disruption of the psychiatric risk gene Ankyrin 3 enhances microtubule dynamics through GSK3/CRMP2 signaling. \u003cem\u003eTransl Psychiatry\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 135 (2018).\u003c/li\u003e\n\u003cli\u003eBuniello A\u003cem\u003e, et al.\u003c/em\u003e The NHGRI-EBI GWAS Catalog of published genome-wide association studies, targeted arrays and summary statistics 2019. \u003cem\u003eNucleic Acids Res\u003c/em\u003e \u003cstrong\u003e47\u003c/strong\u003e, D1005-D1012 (2019).\u003c/li\u003e\n\u003cli\u003ePain O\u003cem\u003e, et al.\u003c/em\u003e Novel Insight Into the Etiology of Autism Spectrum Disorder Gained by Integrating Expression Data With Genome-wide Association Statistics. \u003cem\u003eBiol Psychiatry\u003c/em\u003e \u003cstrong\u003e86\u003c/strong\u003e, 265-273 (2019).\u003c/li\u003e\n\u003cli\u003eHall LS\u003cem\u003e, et al.\u003c/em\u003e Cis-effects on gene expression in the human prenatal brain associated with genetic risk for neuropsychiatric disorders. \u003cem\u003eMol Psychiatry\u003c/em\u003e, (2020).\u003c/li\u003e\n\u003cli\u003eGandal MJ\u003cem\u003e, et al.\u003c/em\u003e Transcriptome-wide isoform-level dysregulation in ASD, schizophrenia, and bipolar disorder. \u003cem\u003eScience\u003c/em\u003e \u003cstrong\u003e362\u003c/strong\u003e, (2018).\u003c/li\u003e\n\u003cli\u003eGusev A\u003cem\u003e, et al.\u003c/em\u003e Integrative approaches for large-scale transcriptome-wide association studies. \u003cem\u003eNat Genet\u003c/em\u003e \u003cstrong\u003e48\u003c/strong\u003e, 245-252 (2016).\u003c/li\u003e\n\u003cli\u003eWainberg M\u003cem\u003e, et al.\u003c/em\u003e Opportunities and challenges for transcriptome-wide association studies. \u003cem\u003eNat Genet\u003c/em\u003e \u003cstrong\u003e51\u003c/strong\u003e, 592-599 (2019).\u003c/li\u003e\n\u003cli\u003eFang WQ\u003cem\u003e, et al.\u003c/em\u003e Overproduction of upper-layer neurons in the neocortex leads to autism-like features in mice. \u003cem\u003eCell Rep\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 1635-1643 (2014).\u003c/li\u003e\n\u003cli\u003eFame RM, MacDonald JL, Macklis JD. Development, specification, and diversity of callosal projection neurons. \u003cem\u003eTrends Neurosci\u003c/em\u003e \u003cstrong\u003e34\u003c/strong\u003e, 41-50 (2011).\u003c/li\u003e\n\u003cli\u003eIwata K\u003cem\u003e, et al.\u003c/em\u003e N-ethylmaleimide-sensitive factor interacts with the serotonin transporter and modulates its trafficking: implications for pathophysiology in autism. \u003cem\u003eMol Autism\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 33 (2014).\u003c/li\u003e\n\u003cli\u003eChuang TJ, Wu CS, Chen CY, Hung LY, Chiang TW, Yang MY. NCLscan: accurate identification of non-co-linear transcripts (fusion, trans-splicing and circular RNA) with a good balance between sensitivity and precision. \u003cem\u003eNucleic Acids Res\u003c/em\u003e \u003cstrong\u003e44\u003c/strong\u003e, e29 (2016).\u003c/li\u003e\n\u003cli\u003eIvanov A\u003cem\u003e, et al.\u003c/em\u003e Analysis of intron sequences reveals hallmarks of circular RNA biogenesis in animals. \u003cem\u003eCell reports\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 170-177 (2015).\u003c/li\u003e\n\u003cli\u003eAltschul SF, Gish W, Miller W, Myers EW, Lipman DJ. Basic local alignment search tool. \u003cem\u003eJ Mol Biol\u003c/em\u003e \u003cstrong\u003e215\u003c/strong\u003e, 403-410 (1990).\u003c/li\u003e\n\u003cli\u003eHinrichs AS\u003cem\u003e, et al.\u003c/em\u003e The UCSC Genome Browser Database: update 2006. \u003cem\u003eNucleic Acids Res\u003c/em\u003e \u003cstrong\u003e34\u003c/strong\u003e, D590-598 (2006).\u003c/li\u003e\n\u003cli\u003eStorey JD, Tibshirani R. Statistical significance for genomewide studies. \u003cem\u003eProc Natl Acad Sci U S A\u003c/em\u003e \u003cstrong\u003e100\u003c/strong\u003e, 9440-9445 (2003).\u003c/li\u003e\n\u003cli\u003eWang P, Zhao D, Lachman HM, Zheng D. Enriched expression of genes associated with autism spectrum disorders in human inhibitory neurons. \u003cem\u003eTransl Psychiatry\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 13 (2018).\u003c/li\u003e\n\u003cli\u003eSchizophrenia Working Group of the Psychiatric Genomics C. Biological insights from 108 schizophrenia-associated genetic loci. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e511\u003c/strong\u003e, 421-427 (2014).\u003c/li\u003e\n\u003cli\u003eChen J, Bardes EE, Aronow BJ, Jegga AG. ToppGene Suite for gene list enrichment analysis and candidate gene prioritization. \u003cem\u003eNucleic Acids Res\u003c/em\u003e \u003cstrong\u003e37\u003c/strong\u003e, W305-311 (2009).\u003c/li\u003e\n\u003cli\u003ePalleja A, Horn H, Eliasson S, Jensen LJ. DistiLD Database: diseases and traits in linkage disequilibrium blocks. \u003cem\u003eNucleic Acids Res\u003c/em\u003e \u003cstrong\u003e40\u003c/strong\u003e, D1036-1040 (2012).\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Tables","content":"\u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;'\u003e\u003cstrong\u003e\u003cspan style=\"font-family: Calibri, sans-serif; font-size: 15px;\"\u003eTable 1.\u003c/span\u003e\u003c/strong\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u0026nbsp;Number of samples of the RNA-seq and genotyping data used in this project.\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n\u003ctable style=\"border-collapse:collapse;border:none;\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"2\" style=\"width: 177.2pt;border-top: 1pt solid windowtext;border-left: none;border-bottom: 1pt solid windowtext;border-right: none;padding: 0in 5.4pt;vertical-align: bottom;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:justify;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u003cstrong\u003eData type\u003c/strong\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"3\" style=\"width: 238.1pt;border-top: 1pt solid windowtext;border-left: none;border-bottom: 1pt solid windowtext;border-right: none;padding: 0in 5.4pt;vertical-align: bottom;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u003cstrong\u003eNumber of samples\u003c/strong\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 79.35pt;border-top: none;border-right: none;border-left: none;border-image: initial;border-bottom: 1pt solid windowtext;padding: 0in 5.4pt;vertical-align: bottom;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u003cstrong\u003eASD\u003c/strong\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79.35pt;border-top: none;border-right: none;border-left: none;border-image: initial;border-bottom: 1pt solid windowtext;padding: 0in 5.4pt;vertical-align: bottom;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u003cstrong\u003eNon-ASD\u003c/strong\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79.4pt;border-top: none;border-right: none;border-left: none;border-image: initial;border-bottom: 1pt solid windowtext;padding: 0in 5.4pt;vertical-align: bottom;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u003cstrong\u003eSum\u003c/strong\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 177.2pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:justify;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u003cstrong\u003eRNA-seq data\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e\u003csup\u003e5\u003c/sup\u003e\u003c/strong\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79.35pt;border: none;padding: 0in 5.4pt;vertical-align: bottom;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e111 (48)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79.35pt;border: none;padding: 0in 5.4pt;vertical-align: bottom;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e101 (49)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79.4pt;border: none;padding: 0in 5.4pt;vertical-align: bottom;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e212 (97)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 177.2pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:justify;text-indent:15.85pt;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003eFC samples\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79.35pt;border: none;padding: 0in 5.4pt;vertical-align: bottom;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e37\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79.35pt;border: none;padding: 0in 5.4pt;vertical-align: bottom;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e36\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79.4pt;border: none;padding: 0in 5.4pt;vertical-align: bottom;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e73\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 177.2pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:justify;text-indent:15.85pt;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003eTC samples\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79.35pt;border: none;padding: 0in 5.4pt;vertical-align: bottom;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e36\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79.35pt;border: none;padding: 0in 5.4pt;vertical-align: bottom;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e25\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79.4pt;border: none;padding: 0in 5.4pt;vertical-align: bottom;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e61\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 177.2pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:justify;text-indent:15.85pt;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003eCV samples\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79.35pt;border: none;padding: 0in 5.4pt;vertical-align: bottom;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e38\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79.35pt;border: none;padding: 0in 5.4pt;vertical-align: bottom;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e40\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79.4pt;border: none;padding: 0in 5.4pt;vertical-align: bottom;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e78\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 177.2pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:justify;text-indent:15.85pt;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u0026nbsp;\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79.35pt;border: none;padding: 0in 5.4pt;vertical-align: bottom;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u0026nbsp;\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79.35pt;border: none;padding: 0in 5.4pt;vertical-align: bottom;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u0026nbsp;\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79.4pt;border: none;padding: 0in 5.4pt;vertical-align: bottom;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u0026nbsp;\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 177.2pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:justify;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u003cstrong\u003eGenotyping data\u0026nbsp;\u003c/strong\u003e\u003cstrong\u003e\u003csup\u003e26\u003c/sup\u003e\u003c/strong\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79.35pt;border: none;padding: 0in 5.4pt;vertical-align: bottom;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e40\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79.35pt;border: none;padding: 0in 5.4pt;vertical-align: bottom;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e39\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79.4pt;border: none;padding: 0in 5.4pt;vertical-align: bottom;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e79\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 177.2pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:justify;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u003cstrong\u003eRNA-seq data with the corresponding genotyping data\u003c/strong\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79.35pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79.35pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79.4pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 177.2pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:justify;text-indent:15.85pt;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003eFC+TC+CV samples\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79.35pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e81 (38)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79.35pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e84 (37)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79.4pt;border: none;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e165 (75)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 177.2pt;border-top: none;border-right: none;border-left: none;border-image: initial;border-bottom: 1pt solid windowtext;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:justify;text-indent:15.85pt;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003eFC+TC samples\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79.35pt;border-top: none;border-right: none;border-left: none;border-image: initial;border-bottom: 1pt solid windowtext;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e53 (35)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79.35pt;border-top: none;border-right: none;border-left: none;border-image: initial;border-bottom: 1pt solid windowtext;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e52 (34)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 79.4pt;border-top: none;border-right: none;border-left: none;border-image: initial;border-bottom: 1pt solid windowtext;padding: 0in 5.4pt;vertical-align: top;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e105 (69)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:justify;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003eNote. The numbers of the examined individuals are shown in parentheses. FC, frontal cortex (Brodmann Area 9). TC, temporal cortex (Brodmann Area 22, 41, and 42). CV, cerebellar vermis. The data were downloaded from the PsychENCODE Consortium Knowledge Portal at https://www.synapse.org/#!Synapse:syn4587609 under accession number syn458609.\u0026nbsp;\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n\u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:justify;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u0026nbsp;\u003c/span\u003e\u003c/span\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u0026nbsp;\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n\u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:justify;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u003cstrong\u003eTable 2.\u003c/strong\u003e The observed \u003cem\u003etrans\u003c/em\u003e-eGenes that were also previously identified as TWAS-prioritized genes/transcripts in ASD.\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n\u003ctable style=\"width:418.2pt;border-collapse:collapse;border:none;\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"5\" style=\"width:340.2pt;border-top:solid windowtext 1.0pt;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:none;padding:0in 5.4pt 0in 5.4pt;height:20.0pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u003cstrong\u003eThis study\u003c/strong\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" style=\"width:78.0pt;border-top:solid windowtext 1.0pt;border-left:none;border-bottom:solid windowtext 1.0pt;border-right:none;padding:0in 5.4pt 0in 5.4pt;height:20.0pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u003cstrong\u003eTWAS reference\u003c/strong\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width:49.65pt;border:none;border-bottom:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.0pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:justify;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u003cstrong\u003e\u003cem\u003eTrans\u003c/em\u003e\u003c/strong\u003e\u003cstrong\u003e-eGene\u003c/strong\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:127.55pt;border:none;border-bottom:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.0pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:justify;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u003cstrong\u003ecircQTL/\u003cem\u003etrans\u003c/em\u003e-eQTL SNP\u003c/strong\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:65.8pt;border:none;border-bottom:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.0pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:justify;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u003cstrong\u003ecircRNA\u003c/strong\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:43.05pt;border:none;border-bottom:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.0pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:justify;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u003cstrong\u003ePassing CIT?\u003csup\u003e*\u003c/sup\u003e\u003c/strong\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:54.15pt;border:none;border-bottom:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:16.0pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:justify;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u003cstrong\u003ePassing mediation test?\u003csup\u003e*\u003c/sup\u003e\u003c/strong\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"4\" style=\"width:49.65pt;border:none;border-bottom:dashed windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:justify;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u003cem\u003eCRHR1\u003c/em\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:127.55pt;border:none;border-bottom:dashed windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003ers79543065, rs80018118, rs117601704\u0026nbsp;\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:65.8pt;border:none;border-bottom:dashed windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:justify;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003ecircERC1-1\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:43.05pt;border:none;border-bottom:dashed windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003eNo\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:54.15pt;border:none;border-bottom:dashed windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003eYes\u0026nbsp;\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e(52-54%)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"6\" style=\"width:78.0pt;border:none;border-bottom:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003eGandal et al.\u003csup\u003e56\u003c/sup\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width:127.55pt;border:none;border-bottom:dashed windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003ers572758241, rs145151802, rs188227047, rs146232238, rs148453911, rs147867900, rs141400333, rs150595674, rs182196454\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:65.8pt;border:none;border-bottom:dashed windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:justify;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003ecircSPHKAP\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:43.05pt;border:none;border-bottom:dashed windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003eNo\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:54.15pt;border:none;border-bottom:dashed windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003eYes\u0026nbsp;\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e(51-72%)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width:127.55pt;border:none;border-bottom:dashed windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003ers118122108, rs79177473, rs117315428\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:65.8pt;border:none;border-bottom:dashed windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:justify;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003ecircERC1-2\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:43.05pt;border:none;border-bottom:dashed windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003eNo\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:54.15pt;border:none;border-bottom:dashed windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003eNo\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width:127.55pt;border:none;border-bottom:dashed windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003ers182290762\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:65.8pt;border:none;border-bottom:dashed windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003ecircZNF236\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:43.05pt;border:none;border-bottom:dashed windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003eNo\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:54.15pt;border:none;border-bottom:dashed windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003eNo\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width:49.65pt;border:none;border-bottom:dashed windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:justify;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u003cem\u003eXKR6\u003c/em\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:127.55pt;border:none;border-bottom:dashed windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003ers572758241\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:65.8pt;border:none;border-bottom:dashed windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:justify;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003ecircSPHKAP\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:43.05pt;border:none;border-bottom:dashed windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003eNo\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:54.15pt;border:none;border-bottom:dashed windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003eYes\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e(58%)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width:49.65pt;border:none;border-bottom:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:justify;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u003cem\u003eNKX2-2\u003c/em\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:127.55pt;border:none;border-bottom:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003ers182920567\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:65.8pt;border:none;border-bottom:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:justify;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003ecircACVR2A\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:43.05pt;border:none;border-bottom:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003eNo\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:54.15pt;border:none;border-bottom:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003eNo\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width:49.65pt;border:none;border-bottom:dashed windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:justify;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u003cem\u003eNSF\u003c/em\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:127.55pt;border:none;border-bottom:dashed windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003ers182920567\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:65.8pt;border:none;border-bottom:dashed windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:justify;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003ecircACVR2A\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:43.05pt;border:none;border-bottom:dashed windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003eYes\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:54.15pt;border:none;border-bottom:dashed windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003eNo\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" style=\"width:78.0pt;border:none;border-bottom:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003eHall et al.\u003csup\u003e55\u003c/sup\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width:49.65pt;border:none;border-bottom:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:2.35pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:justify;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u003cem\u003eARL17B\u003c/em\u003e\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:127.55pt;border:none;border-bottom:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:2.35pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003ers572758241\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:65.8pt;border:none;border-bottom:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:2.35pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:justify;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003ecircSPHKAP\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:43.05pt;border:none;border-bottom:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:2.35pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003eNo\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width:54.15pt;border:none;border-bottom:solid windowtext 1.0pt;padding:0in 5.4pt 0in 5.4pt;height:2.35pt;\"\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003eYes\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:center;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e(33%)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;text-align:justify;'\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u003csup\u003e*\u003c/sup\u003e \u0026ldquo;Passing CIT\u0026rdquo; and \u0026ldquo;passing mediation test\u0026rdquo; represent that the circQTL-\u003cem\u003etrans\u003c/em\u003e-eGene-ASD diagnosis axes and the circQTL-circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene axes pass CIT (see also Supplementary Tables 4) and mediation test (see also Supplementary Tables 3), respectively. The proportions of mediation of the circQTL-\u003cem\u003etrans\u003c/em\u003e-eGene associations mediated by the corresponding circRNA expression were denoted in parentheses.\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n\u003cp\u003e\u003cspan style=\"font-size: 15px;\"\u003e\u003cspan style=\"font-family: Calibri, sans-serif;\"\u003e\u0026nbsp;\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n\u003cp style='margin:0in;font-size:16px;font-family:\"Calibri\",sans-serif;'\u003e\u003cspan style=\"font-family: Calibri, sans-serif; font-size: 15px;\"\u003e\u0026nbsp;\u003c/span\u003e\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Autism spectrum disorder, Circular RNA, circQTL, trans-eQTL, Regulatory network, Causative biology","lastPublishedDoi":"10.21203/rs.3.rs-155043/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-155043/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eGenetic risk variants and transcriptional expression changes in autism spectrum disorder (ASD) have been widely identified, but their causal relationship is largely unknown. Circular RNAs (circRNAs) are abundant in brain and often serve as upstream regulators of mRNAs. By integrating RNA-sequencing with genotyping data from autistic brains, we accessed expression quantitative trait loci (eQTL) of circRNAs (circQTLs) that influenced expression of distant genes (\u003cem\u003etrans\u003c/em\u003e-eGenes) and constructed 43,372 circQTL-\u003cem\u003etrans\u003c/em\u003e-eGene pairs. Mediation test suggested that 19,393 pairs were significantly \u003cem\u003ecis\u003c/em\u003e-mediated by expression of circRNAs near the circQTLs; meanwhile, causal inference test (CIT) suggested 1,000 pairs influencing 708 \u003cem\u003etrans\u003c/em\u003e-eGenes, wherein \u003cem\u003etrans\u003c/em\u003e-eGene expression mediated \u003cem\u003etrans\u003c/em\u003e-eQTL effects on ASD diagnosis. The 708 \u003cem\u003etrans\u003c/em\u003e-eGenes were enriched for ASD risk genes and overrepresented in neurons in the upper neocortical layer. Integration of mediation test- and CIT-passing pairs further constructed 257 circQTL-circRNA-\u003cem\u003etrans\u003c/em\u003e-eGene-ASD propagation paths. These findings increase our understanding of causative biology and cryptic regulatory mechanisms underlying ASD.\u003c/p\u003e","manuscriptTitle":"Trans-genetic effects of circular RNA expression quantitative trait loci and potential causal mechanisms in autism","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2021-02-05 15:46:45","doi":"10.21203/rs.3.rs-155043/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"ae9318ac-7cee-4a28-b010-2bf25132b234","owner":[],"postedDate":"February 5th, 2021","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":2254391,"name":"Cognitive Neuroscience"},{"id":2254392,"name":"Medical Genetics"}],"tags":[],"updatedAt":"2022-01-10T20:40:44+00:00","versionOfRecord":[],"versionCreatedAt":"2021-02-05 15:46:45","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-155043","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-155043","identity":"rs-155043","version":["v1"]},"buildId":"FbvkV6FR0MCFSLy54lSbu","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}