Blood transcriptomic comparison of individuals with and without autism spectrum disorder: A combined-samples mega-analysis

Abstract

Blood-based microarray studies comparing individuals affected with autism spectrum disorder (ASD) and typically developing individuals help characterize differences in circulating immune cell functions and offer potential biomarker signal. We sought to combine the subject-level data from previously published studies by mega-analysis to increase the statistical power. We identified studies that compared ex vivo blood or lymphocytes from ASD-affected individuals and unrelated comparison subjects using Affymetrix or Illumina array platforms. Raw microarray data and clinical meta-data were obtained from seven studies, totaling 626 affected and 447 comparison subjects. Microarray data were processed using uniform methods. Covariate-controlled mixed-effect linear models were used to identify gene transcripts and co-expression network modules that were significantly associated with diagnostic status. Permutation-based gene-set analysis was used to identify functionally related sets of genes that were over- and under-expressed among ASD samples. Our results were consistent with diminished interferon-, EGF-, PDGF-, PI3K-AKT-mTOR-, and RAS-MAPK-signaling cascades, and increased ribosomal translation and NK-cell related activity in ASD. We explored evidence for sex-differences in the ASD-related transcriptomic signature. We also demonstrated that machine-learning classifiers using blood transcriptome data perform with moderate accuracy when data are combined across studies. Comparing our results with those from blood-based studies of protein biomarkers (e.g., cytokines and trophic factors), we propose that ASD may feature decoupling between certain circulating signaling proteins (higher in ASD samples) and the transcriptional cascades which they typically elicit within circulating immune cells (lower in ASD samples). These findings provide insight into ASD-related transcriptional differences in circulating immune cells. © 2016 Wiley Periodicals, Inc.

Keywords: gene expression; immune system; machine learning; microarray.

Figure 1

Results of Permutation-Based Gene-Set Analysis. Test statistics from the covariate-controlled single-gene mega-analysis (differences in diagnostic group means after adjustment for covariate effects) were supplied for permutation-based gene-set analysis. As described in the Methods section, this approach assesses whether a given a priori-defined set of genes, on average, shows more evidence of an ASD-associated expression difference than randomly selected gene-sets of equal size. Here we show results that reaches a Bonferroni-corrected p < 0.05. The functional themes of over-expressed gene-sets are shown within the red-colored upward-pointing arrow, and the functional themes of under-expressed gene-sets are shown within the green-colored downward-pointing arrow. P-values (Bonferroni-corrected for the number of sets per database) are displayed in parenthesis, and p-values < 1×10⁻⁶ (reflecting the minimum possible p-value based on the number of permutations) are denoted with **. Full results of this analysis are available in Supplementary Table 6 and gene-set names can be examined within the Molecular Signature Database for additional context.

Figure 2

ASD-Associated Gene Co-expression Network Modules. Gene co-expression network analysis performed on non-SVA-corrected data identified thirty network modules when ASD cases and comparison subjects were analyzed together. ASD-associated modules were identified using linear mixed models (as described in the single-gene analysis) to predict module eigengenes; five modules showed a significant association with ASD (Benjamini-Hochberg-corrected p < 0.05). Modules with higher eigengene values among ASD cases are described as over-expressed (left side) and those with lower eigengene values among ASD cases are described as under-expressed (right side). Within each colored panel, we indicate the module color name, the threshold of its significance for association with ASD, the number of positively and negatively loading genes, and functional and cell-type enrichments based on hypergeometric test statistics (Benjamini-Hochberg p < 0.05; Full Results in Supplementary Table 9). We also include the names of genes that showed ASD GWAS signal at an uncorrected gene-level p < 0.05 in a recent meta-analysis. The symbol * beside the number of GWAS signal genes denotes whether the network module showed significant GWAS signal enrichment based on a quantitative permutation testing (see Methods). We also indicate the identities of SFARIgene candidate-genes contained within each module. Beside each colored panel is a plot depicting the 25 most highly inter-correlated genes for that network module; the 5 most highly correlated “hub” genes are depicted in the center. Each gene is depicted as a small colored circle, with the top right quarter indicating the relative over- or under-expression of that gene in ASD cases based on the single-gene covariate-controlled mega-analysis analysis; dark red indicates highly significant over-expression (FDR q < 0.05), while light red indicates nominal over-expression (uncorrected p < 0.05). The same relationships can be understood for dark green and light green, with respect to under-expressed genes in ASD. For each gene, the bottom left quarter of its circle was colored pink to indicate a gene showing nominal GWAS signal.

Figure 3

Summary of The Present Study’s Findings in Comparison with Previous Studies of Human Blood and Brain Tissues with Respect to ASDs. We conducted a literature search and attempted to compare our blood transcriptomic findings for various biological functions against previous studies that reported on the expression of either RNA or protein markers relevant to those functions in human blood or brain tissue, comparing ASD samples (including genetic syndromes with high rates of ASD) with unaffected comparison samples. All enumerated references supporting this table can be found in Supplementary Table 12. With respect to protein markers, we considered studies that examined the measured levels of circulating protein (e.g., blood cytokine or growth factor studies) or the relative quantity of a protein in cell or tissue, or the relative activation of protein signaling (e.g., quantity or proportion of phosphorylated protein). When clear directional findings were apparent from the reviewed literature, we denoted the conclusions with up or down arrows (reflecting ASD samples relative to controls) and provide brief descriptions of what was show, with supporting citations. When the literature clearly supports the identification of both increased and decreased activity of a biological function in ASD, we denoted this by including both up and down arrows, and attempt to provide more information on factors that might account for these findings. When insufficient evidence was found to draw a conclusion, we denoted this with a question mark. All supporting citations are discussed more thoroughly in the Supplementary Materials.