Limits...
Integrated systems analysis reveals a molecular network underlying autism spectrum disorders.

Li J, Shi M, Ma Z, Zhao S, Euskirchen G, Ziskin J, Urban A, Hallmayer J, Snyder M - Mol. Syst. Biol. (2014)

Bottom Line: Expression of this module was dichotomized with a ubiquitously expressed subcomponent and another subcomponent preferentially expressed in the corpus callosum, which was significantly affected by our identified mutations in the network center.RNA-sequencing of the corpus callosum from patients with autism exhibited extensive gene mis-expression in this module, and our immunochemical analysis showed that the human corpus callosum is predominantly populated by oligodendrocyte cells.Our analysis delineates a natural network involved in autism, helps uncover novel candidate genes for this disease and improves our understanding of its molecular pathology.

View Article: PubMed Central - PubMed

Affiliation: Department of Genetics, Stanford Center for Genomics and Personalized Medicine Stanford University School of Medicine, Stanford, CA, USA.

No MeSH data available.


Related in: MedlinePlus

Integrative analysis of the genetic alteration in this studyEnrichment of the differentially expressed genes in module #13. RNA-sequencing of the corpus callosum of autism patients and their matched controls. Enrichment was not observed for the genes in the human synaptome or the collection of known autism genes (excluding genes in this module). Statistical significance was determined by hypergeometric test.The mutation pattern of the genes from the innermost layers of the interaction network (K ≥ 10) to the periphery layer (K = 1). Genes in the central and periphery layers in this module are more likely to be affected, while the trend cannot be observed in 10,000 random simulations. For individual bins, significant enrichment and depletion were observed in the central layers (K ≥ 10) and the intermediate layers (3 ≤ K < 6), respectively. Statistical significance of the enrichment was determined by hypergeometric test. 10,000 random permutations were performed to determine the statistical significance of the curve.Compositional bias of the mutated genes in central layers. The mutated genes in central layers are more biased toward the corpus callosum-specific subcomponent; this trend is not observed in background or other mutated genes with varying degree of K. Statistical significance of the enrichment was determined by hypergeometric test.Positive correlation between network coreness and gene expression in the corpus callosum. RNA-sequencing of the corpus callosum of six non-autistic individuals revealed a positive correlation, suggesting the central layers may play critical roles in the corpus callosum. Two outlier genes, DYNLL1 and BCAS1, are separately labeled due to their extreme expression in this tissue. The correlation coefficient r and its statistical significance were computed using Spearman's correlation.Predicted sub-complexes within this module. Genes in this module are topologically clustered to form sub-complexes, with the significantly mutated genes labeled in blue, mis-expressed genes in the corpus callosum labeled in green, and both in red. Two clusters, #6 for SHANK-DLGAP complexes and #6 for LRP2, and its binding partners, are enriched for the mis-expressed or mutated genes, respectively. Statistical significance of the enrichment was determined by hypergeometric test.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4300495&req=5

fig05: Integrative analysis of the genetic alteration in this studyEnrichment of the differentially expressed genes in module #13. RNA-sequencing of the corpus callosum of autism patients and their matched controls. Enrichment was not observed for the genes in the human synaptome or the collection of known autism genes (excluding genes in this module). Statistical significance was determined by hypergeometric test.The mutation pattern of the genes from the innermost layers of the interaction network (K ≥ 10) to the periphery layer (K = 1). Genes in the central and periphery layers in this module are more likely to be affected, while the trend cannot be observed in 10,000 random simulations. For individual bins, significant enrichment and depletion were observed in the central layers (K ≥ 10) and the intermediate layers (3 ≤ K < 6), respectively. Statistical significance of the enrichment was determined by hypergeometric test. 10,000 random permutations were performed to determine the statistical significance of the curve.Compositional bias of the mutated genes in central layers. The mutated genes in central layers are more biased toward the corpus callosum-specific subcomponent; this trend is not observed in background or other mutated genes with varying degree of K. Statistical significance of the enrichment was determined by hypergeometric test.Positive correlation between network coreness and gene expression in the corpus callosum. RNA-sequencing of the corpus callosum of six non-autistic individuals revealed a positive correlation, suggesting the central layers may play critical roles in the corpus callosum. Two outlier genes, DYNLL1 and BCAS1, are separately labeled due to their extreme expression in this tissue. The correlation coefficient r and its statistical significance were computed using Spearman's correlation.Predicted sub-complexes within this module. Genes in this module are topologically clustered to form sub-complexes, with the significantly mutated genes labeled in blue, mis-expressed genes in the corpus callosum labeled in green, and both in red. Two clusters, #6 for SHANK-DLGAP complexes and #6 for LRP2, and its binding partners, are enriched for the mis-expressed or mutated genes, respectively. Statistical significance of the enrichment was determined by hypergeometric test.

Mentions: Given the apparent importance of oligodendrocytes in the corpus callosum, we further hypothesized that gene expression in this module is likely to be perturbed in the corpus callosum of ASD patients. We obtained frozen postmortem samples from six young Caucasian males with a diagnosis of autism together with their respective matched controls from the NICHD Brain and Tissue Bank (Materials and Methods and Supplementary Table S5). Total RNAs were prepared and subjected to high-coverage (180M reads/sample) deep RNA-sequencing. Biological replicates (with the same sequencing depth) were performed on half of the samples, using different sections of the same tissue block. The biological replicates produced highly reproducible results with a median Pearson's coefficient equal to 0.95 (range 0.9–0.96; Supplementary Fig S12), whereas the correlations among samples from different individuals were substantially lower (median correlation coefficient 0.89, P = 4.4e-3, Wilcoxon rank-sum test), demonstrating the high intra-individual reproducibility of our platform. Because gene expression in the brain is age dependent in patients with autism (Chow et al, 2012), we compared gene expression in each case–control pair with identical age, ethnicity, sex and comparable postmortem intervals (PMIs). We then identified genes showing the most extreme expression changes in at least one case–control pair (fold change > 2, above the 97.5% upper bound for up-regulation and below 2.5% for down-regulation across the entire transcriptomes, Supplementary Table S6). Genes encoding components of the module #13 showed significant enrichment for the differentially expressed genes relative to the genes encoding the entire protein interaction network (P = 5e-4, hypergeometric test, Fig5A). We conducted comparisons against two control gene sets: a complete list of 1,886 known synapse-related genes (the synaptome in Fig5A) from SynaptomeDB (Pirooznia et al, 2012) and the other control included a list of known 383 autism candidate genes represented on the network. In each case, the gene set contained a similar fraction of differentially expressed genes as the entire transcriptome background (P = 0.39 and 0.14, hypergeometric tests, respectively). Thus, expression of module #13, but not synaptic genes in general or known ASD candidate genes, was significantly altered in the corpus callosum of the ASD patients relative to the matched controls.


Integrated systems analysis reveals a molecular network underlying autism spectrum disorders.

Li J, Shi M, Ma Z, Zhao S, Euskirchen G, Ziskin J, Urban A, Hallmayer J, Snyder M - Mol. Syst. Biol. (2014)

Integrative analysis of the genetic alteration in this studyEnrichment of the differentially expressed genes in module #13. RNA-sequencing of the corpus callosum of autism patients and their matched controls. Enrichment was not observed for the genes in the human synaptome or the collection of known autism genes (excluding genes in this module). Statistical significance was determined by hypergeometric test.The mutation pattern of the genes from the innermost layers of the interaction network (K ≥ 10) to the periphery layer (K = 1). Genes in the central and periphery layers in this module are more likely to be affected, while the trend cannot be observed in 10,000 random simulations. For individual bins, significant enrichment and depletion were observed in the central layers (K ≥ 10) and the intermediate layers (3 ≤ K < 6), respectively. Statistical significance of the enrichment was determined by hypergeometric test. 10,000 random permutations were performed to determine the statistical significance of the curve.Compositional bias of the mutated genes in central layers. The mutated genes in central layers are more biased toward the corpus callosum-specific subcomponent; this trend is not observed in background or other mutated genes with varying degree of K. Statistical significance of the enrichment was determined by hypergeometric test.Positive correlation between network coreness and gene expression in the corpus callosum. RNA-sequencing of the corpus callosum of six non-autistic individuals revealed a positive correlation, suggesting the central layers may play critical roles in the corpus callosum. Two outlier genes, DYNLL1 and BCAS1, are separately labeled due to their extreme expression in this tissue. The correlation coefficient r and its statistical significance were computed using Spearman's correlation.Predicted sub-complexes within this module. Genes in this module are topologically clustered to form sub-complexes, with the significantly mutated genes labeled in blue, mis-expressed genes in the corpus callosum labeled in green, and both in red. Two clusters, #6 for SHANK-DLGAP complexes and #6 for LRP2, and its binding partners, are enriched for the mis-expressed or mutated genes, respectively. Statistical significance of the enrichment was determined by hypergeometric test.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4300495&req=5

fig05: Integrative analysis of the genetic alteration in this studyEnrichment of the differentially expressed genes in module #13. RNA-sequencing of the corpus callosum of autism patients and their matched controls. Enrichment was not observed for the genes in the human synaptome or the collection of known autism genes (excluding genes in this module). Statistical significance was determined by hypergeometric test.The mutation pattern of the genes from the innermost layers of the interaction network (K ≥ 10) to the periphery layer (K = 1). Genes in the central and periphery layers in this module are more likely to be affected, while the trend cannot be observed in 10,000 random simulations. For individual bins, significant enrichment and depletion were observed in the central layers (K ≥ 10) and the intermediate layers (3 ≤ K < 6), respectively. Statistical significance of the enrichment was determined by hypergeometric test. 10,000 random permutations were performed to determine the statistical significance of the curve.Compositional bias of the mutated genes in central layers. The mutated genes in central layers are more biased toward the corpus callosum-specific subcomponent; this trend is not observed in background or other mutated genes with varying degree of K. Statistical significance of the enrichment was determined by hypergeometric test.Positive correlation between network coreness and gene expression in the corpus callosum. RNA-sequencing of the corpus callosum of six non-autistic individuals revealed a positive correlation, suggesting the central layers may play critical roles in the corpus callosum. Two outlier genes, DYNLL1 and BCAS1, are separately labeled due to their extreme expression in this tissue. The correlation coefficient r and its statistical significance were computed using Spearman's correlation.Predicted sub-complexes within this module. Genes in this module are topologically clustered to form sub-complexes, with the significantly mutated genes labeled in blue, mis-expressed genes in the corpus callosum labeled in green, and both in red. Two clusters, #6 for SHANK-DLGAP complexes and #6 for LRP2, and its binding partners, are enriched for the mis-expressed or mutated genes, respectively. Statistical significance of the enrichment was determined by hypergeometric test.
Mentions: Given the apparent importance of oligodendrocytes in the corpus callosum, we further hypothesized that gene expression in this module is likely to be perturbed in the corpus callosum of ASD patients. We obtained frozen postmortem samples from six young Caucasian males with a diagnosis of autism together with their respective matched controls from the NICHD Brain and Tissue Bank (Materials and Methods and Supplementary Table S5). Total RNAs were prepared and subjected to high-coverage (180M reads/sample) deep RNA-sequencing. Biological replicates (with the same sequencing depth) were performed on half of the samples, using different sections of the same tissue block. The biological replicates produced highly reproducible results with a median Pearson's coefficient equal to 0.95 (range 0.9–0.96; Supplementary Fig S12), whereas the correlations among samples from different individuals were substantially lower (median correlation coefficient 0.89, P = 4.4e-3, Wilcoxon rank-sum test), demonstrating the high intra-individual reproducibility of our platform. Because gene expression in the brain is age dependent in patients with autism (Chow et al, 2012), we compared gene expression in each case–control pair with identical age, ethnicity, sex and comparable postmortem intervals (PMIs). We then identified genes showing the most extreme expression changes in at least one case–control pair (fold change > 2, above the 97.5% upper bound for up-regulation and below 2.5% for down-regulation across the entire transcriptomes, Supplementary Table S6). Genes encoding components of the module #13 showed significant enrichment for the differentially expressed genes relative to the genes encoding the entire protein interaction network (P = 5e-4, hypergeometric test, Fig5A). We conducted comparisons against two control gene sets: a complete list of 1,886 known synapse-related genes (the synaptome in Fig5A) from SynaptomeDB (Pirooznia et al, 2012) and the other control included a list of known 383 autism candidate genes represented on the network. In each case, the gene set contained a similar fraction of differentially expressed genes as the entire transcriptome background (P = 0.39 and 0.14, hypergeometric tests, respectively). Thus, expression of module #13, but not synaptic genes in general or known ASD candidate genes, was significantly altered in the corpus callosum of the ASD patients relative to the matched controls.

Bottom Line: Expression of this module was dichotomized with a ubiquitously expressed subcomponent and another subcomponent preferentially expressed in the corpus callosum, which was significantly affected by our identified mutations in the network center.RNA-sequencing of the corpus callosum from patients with autism exhibited extensive gene mis-expression in this module, and our immunochemical analysis showed that the human corpus callosum is predominantly populated by oligodendrocyte cells.Our analysis delineates a natural network involved in autism, helps uncover novel candidate genes for this disease and improves our understanding of its molecular pathology.

View Article: PubMed Central - PubMed

Affiliation: Department of Genetics, Stanford Center for Genomics and Personalized Medicine Stanford University School of Medicine, Stanford, CA, USA.

No MeSH data available.


Related in: MedlinePlus