Genome-wide changes in lncRNA, splicing, and regional gene expression patterns in autism

ASD cortex shows dysregulation of primate-specific long noncoding RNAs.

“Our analysis reveals ASD-associated dysregulation of primate-specific long noncoding RNAs (lncRNAs), downregulation of the alternative splicing of activity-dependent neuron-specific exons, and attenuation of normal differences in gene expression between the frontal and temporal lobes.”

ASD cortex shows downregulation of alternative splicing of activity-dependent neuron-specific exons.

“Our analysis reveals ASD-associated dysregulation of primate-specific long noncoding RNAs (lncRNAs), downregulation of the alternative splicing of activity-dependent neuron-specific exons, and attenuation of normal differences in gene expression between the frontal and temporal lobes.”

ASD cortex shows attenuation of the normal gene expression differences between the frontal and temporal lobes.

“Our analysis reveals ASD-associated dysregulation of primate-specific long noncoding RNAs (lncRNAs), downregulation of the alternative splicing of activity-dependent neuron-specific exons, and attenuation of normal differences in gene expression between the frontal and temporal lobes.”

Gene expression in ASD can separate samples by disease status.

“We performed a classification analysis to confirm that gene expression in ASD could separate samples by disease status (Extended Data Fig. 2a) and confirmed the technical quality of our data with qRT–PCR (Extended Data Fig. 2b, c).”

There are high correlations between splicing factor expression and differential splicing in ASD cortex.

“Therefore, we next investigated whether the shared splicing signature in ASD could be explained by perturbations in splicing factors known to be important in nervous system function8,14 (Extended Data Fig. 4j), and found high correlations between splicing factor expression and differential splicing in the cortex (Fig. 1f) but not the cerebellum (Fig. 1g).”

The downregulated gene set in ASD cortex is enriched in neuron-expressed genes and neuronal pathways, including PVALB and SYT2.

“We next evaluated enrichment of the gene sets for pathways and cell types (Extended Data Fig. 2d, e), and found that the downregulated set was enriched in genes expressed in neurons and involved in neuronal pathways, including PVALB and SYT2, which are highly expressed in interneurons; by contrast, the upregulated gene set was enriched in genes expressed in microglia and astrocytes8.”

The ASD-associated lncRNAs as a group are enriched in brain relative to other tissues.

“As a group, these lncRNAs are enriched in the brain relative to other tissues (Extended Data Fig. 2l, m) and most that have been evaluated across species exhibit primate-specific expression patterns in the brain11, which we confirm for several transcripts (Supplementary Information, Extended Data Fig. 3a–h).”

ASD-related expression changes are consistently stronger in cortex than in cerebellum.

“The lack of significant DGE in the cerebellum is explained by the fact that changes in expression were consistently stronger in the cortex than in the cerebellum (Extended Data Fig. 2h), which suggests that the cortex is more selectively vulnerable to these transcriptomic alterations.”

There is no significant differential splicing signal in the cerebellum.

“We found a significant differential splicing signal over background in the cortex (1,127 differential splicing events in 833 genes; Methods), but not in the cerebellum (P distributions in Extended Data Fig. 4b, c).”

There is a significant differential splicing signal over background in ASD cortex, with 1,127 differential splicing events in 833 genes.

“We found a significant differential splicing signal over background in the cortex (1,127 differential splicing events in 833 genes; Methods), but not in the cerebellum (P distributions in Extended Data Fig. 4b, c).”

The ACP set is enriched in Wnt signalling, calcium binding, and neuronal genes.

“Pathway and cell-type analysis showed that the ACP set is enriched in Wnt signalling, calcium binding, and neuronal genes (Extended Data Fig. 6i, j, Supplementary Information).”

The genetically defined ASD subtype dup15q shares the core transcriptomic signature observed in idiopathic ASD.

“We further demonstrate that a genetically defined subtype of ASD, chromosome 15q11.2–13.1 duplication syndrome (dup15q), shares the core transcriptomic signature observed in idiopathic ASD.”

Individuals with ASD show age-related changes in the trajectory of microglial and synaptic function over the first two decades.

“Co-expression network analysis reveals that individuals with ASD show age-related changes in the trajectory of microglial and synaptic function over the first two decades, and suggests that genetic risk for ASD may influence changes in regional cortical gene expression.”

Differential splicing events in ASD cortex are significantly enriched in events previously shown to be regulated by neuronal activity.

“Indeed, differential splicing events were significantly enriched in those previously shown to be regulated by neuronal activity21 (Fig. 1h).”

• cites [21] Maze, I. et al. Critical role of histone turnover in neuronal transcription and plasticity. Neuron 87, 77–94 (2015) sourceNot checked yet

  Not checked in the free run.

Microglial and synaptic modules exhibit significant anticorrelation.

“It also confirms that downregulated modules are enriched in synaptic function and neuronal genes, that upregulated modules are enriched in genes associated with inflammatory pathways and glial function4,8, and that microglial and synaptic modules exhibit significant anticorrelation (Fig. 4c).”

Differential gene expression between control cortex and ASD cortex replicates with strong concordance (R2 = 0.60) with prior microarray profiles.

“We first validated differential gene expression (DGE) between samples of cortex from control individuals and those with ASD (ASD cortex) by comparing gene expression with that of different individuals from those previously profiled by microarray8, and found strong concordance (R2 = 0.60; Fig. 1a, Extended Data Fig. 1i).”

• cites [8] Voineagu, I. et al. Transcriptomic analysis of autistic brain reveals convergent molecular pathology. Nature 474, 380–384 (2011) sourceNot checked yet

  Not checked in the free run.

Most differential splicing events in ASD cortex involve exclusion of neuron-specific exons.

“Cell-type specific enrichment and pathway analysis of alternative splicing demonstrated that most differential splicing events involve exclusion of neuron-specific exons17 (Fig. 1e, Extended Data Fig. 4i).”

• cites [17] Zhang, Y. et al. An RNA-sequencing transcriptome and splicing database of glia, neurons, and vascular cells of the cerebral cortex. J. Neurosci. 34, 11929–11947 (2014) sourceNot checked yet

  Not checked in the free run.

The downregulated modules CTX.M10 and CTX.M16 are enriched in genes previously related to neuronal firing rate.

“Furthermore, the downregulated modules CTX.M10 and CTX.M16 are enriched in genes previously related to neuronal firing rate, consistent with the overlap of dysregulated splicing with events regulated by neuronal activity (Extended Data Fig. 9b and Fig. 1h).”

There is no significant differential gene expression in the cerebellum at FDR < 0.05.

“Although there was no significant DGE in the cerebellum (FDR < 0.05, P distributions in Fig. 1b), similar to observations in a smaller cohort8, there was a replication signal in the cerebellum and overall concordance between ASD-related fold changes in the cortex and cerebellum (Extended Data Fig. 2f–h).”

• cites [8] Voineagu, I. et al. Transcriptomic analysis of autistic brain reveals convergent molecular pathology. Nature 474, 380–384 (2011) sourceNot checked yet

  Not checked in the free run.

Most evaluated ASD-associated lncRNAs exhibit primate-specific expression patterns in brain.

“As a group, these lncRNAs are enriched in the brain relative to other tissues (Extended Data Fig. 2l, m) and most that have been evaluated across species exhibit primate-specific expression patterns in the brain11, which we confirm for several transcripts (Supplementary Information, Extended Data Fig. 3a–h).”

• cites [11] Necsulea, A. et al. The evolution of lncRNA repertoires and expression patterns in tetrapods. Nature 505, 635–640 (2014) sourceNot checked yet

  Not checked in the free run.

There is a remarkable difference in CTX.M19 and CTX.M20 during the first two decades of life.

“We identified a remarkable difference in CTX.M19 and CTX.M20 during the first two decades of life (Fig. 4g, additional age trajectories in Extended Data Fig. 9f) that is most consistent with an evolving process during early brain development that stabilizes starting in late childhood and early adolescence.”

Compared with a prior RNA-seq study of occipital cortex, there is a weak but significant correlation in fold changes, mostly due to upregulated genes (P = 0.038).

“Despite significant technical differences that reduce power to detect DGE, and profiling of different brain regions in that study, there was a weak but significant correlation in fold changes, which was due mostly to upregulated genes in both studies (P = 0.038, Extended Data Fig. 2i, j).”

• cites [4] 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) sourceNot checked yet

  Not checked in the free run.

Other phenotypes with larger, better-powered GWAS also demonstrate enrichment.

“However, other phenotypes with larger, better-powered genome-wide association studies (GWAS) also demonstrate enrichment (Extended Data Fig. 9h–i).”

There is no such correlation between splicing factor expression and differential splicing in the cerebellum.

“Therefore, we next investigated whether the shared splicing signature in ASD could be explained by perturbations in splicing factors known to be important in nervous system function8,14 (Extended Data Fig. 4j), and found high correlations between splicing factor expression and differential splicing in the cortex (Fig. 1f) but not the cerebellum (Fig. 1g).”

The first principal components for protein coding DGE, lncRNA DGE, and differential splicing are highly correlated (R2 > 0.8).

“When we compared the first principal component across samples for protein coding DGE, lncRNA DGE and differential splicing, we found remarkably high correlations (R2 > 0.8), indicating that molecular convergence is likely to be a unitary phenomenon across multiple levels of transcriptome regulation in ASD (Fig. 1i).”

Sixty lncRNAs are in the differential gene expression set in ASD cortex.

“We next explored lncRNAs, most of which have little functional annotation, and identified 60 lncRNAs in the DGE set (FDR < 0.05, Extended Data Fig. 2k).”

Predicted SOX5 targets exhibited net down-regulation, consistent with the repressive function of SOX5.

“We confirmed that a significant proportion of ACP genes are regulated by SOX5 by overexpressing it in human neural progenitors. SOX5 induced synaptic genes and repressed cell proliferation (Fig. 2e), and predicted SOX5 targets exhibited net down-regulation, consistent with the repressive function of SOX5 (Fig. 2f, Extended Data Fig. 6l, m).”

Downregulated modules are enriched in synaptic function and neuronal genes.

“It also confirms that downregulated modules are enriched in synaptic function and neuronal genes, that upregulated modules are enriched in genes associated with inflammatory pathways and glial function4,8, and that microglial and synaptic modules exhibit significant anticorrelation (Fig. 4c).”

Nine of the ASD-associated lncRNAs interact with FMRP.

“Moreover, 20 of these lncRNAs have been shown to interact with microRNA (miRNA)–protein complexes, and 9 with the fragile X mental retardation protein (FMRP), whose mRNA targets are enriched in ASD risk genes9,10.”

In ASD cortex compared with control cortex, 558 genes show decreased expression.

“Compared to control cortex, 584 genes showed increased expression and 558 showed decreased expression in ASD cortex (Fig. 1b; Benjamini–Hochberg FDR < 0.05, linear mixed effects model; see Methods).”

The upregulated module CTX.M20 was not found in previous analyses.

“Remarkably, the upregulated module CTX.M20 was not found in previous analyses, overlaps significantly with the ACP set (FDR < 0.05, Extended Data Fig. 9a), and contains genes implicated in development and regulation of cell differentiation (Fig. 4f).”

CTX.M24 exhibits significant enrichment for rare mutations found in ASD.

“One module, CTX.M24, exhibited significant enrichment for rare mutations found in ASD, while rare de novo mutations associated with intellectual disability were most strongly enriched in CTX.M22 (FDR < 0.05, Extended Data Fig. 9a).”

Twenty of the ASD-associated lncRNAs interact with miRNA-protein complexes.

“Moreover, 20 of these lncRNAs have been shown to interact with microRNA (miRNA)–protein complexes, and 9 with the fragile X mental retardation protein (FMRP), whose mRNA targets are enriched in ASD risk genes9,10.”

In ASD cortex compared with control cortex, 584 genes show increased expression.

“Compared to control cortex, 584 genes showed increased expression and 558 showed decreased expression in ASD cortex (Fig. 1b; Benjamini–Hochberg FDR < 0.05, linear mixed effects model; see Methods).”

Upregulated modules are enriched in genes associated with inflammatory pathways and glial function.

“It also confirms that downregulated modules are enriched in synaptic function and neuronal genes, that upregulated modules are enriched in genes associated with inflammatory pathways and glial function4,8, and that microglial and synaptic modules exhibit significant anticorrelation (Fig. 4c).”

• cites [4] 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) sourceNot checked yet

  Not checked in the free run.

• cites [8] Voineagu, I. et al. Transcriptomic analysis of autistic brain reveals convergent molecular pathology. Nature 474, 380–384 (2011) sourceNot checked yet

  Not checked in the free run.

LINC00693 and LINC00689 interact with miRNA processing complexes.

“We highlight two primate-specific lncRNAs, LINC00693 and LINC00689. Both interact with miRNA processing complexes and are typically downregulated during development12, but are upregulated in ASD cortex (Fig. 1c, d, Extended Data Fig. 2n).”

The upregulated gene set in ASD cortex is enriched in microglia- and astrocyte-expressed genes.

“We next evaluated enrichment of the gene sets for pathways and cell types (Extended Data Fig. 2d, e), and found that the downregulated set was enriched in genes expressed in neurons and involved in neuronal pathways, including PVALB and SYT2, which are highly expressed in interneurons; by contrast, the upregulated gene set was enriched in genes expressed in microglia and astrocytes8.”

• cites [8] Voineagu, I. et al. Transcriptomic analysis of autistic brain reveals convergent molecular pathology. Nature 474, 380–384 (2011) sourceNot checked yet

  Not checked in the free run.

The differential splicing molecular signature is not driven by differential gene expression.

“Notably, the differential splicing molecular signature is not driven by DGE (Extended Data Fig. 4h), consistent with the observation that splicing alterations are related to common disease risk independently of gene expression changes16.”

CTX.M20 contains genes implicated in development and regulation of cell differentiation.

“Remarkably, the upregulated module CTX.M20 was not found in previous analyses, overlaps significantly with the ACP set (FDR < 0.05, Extended Data Fig. 9a), and contains genes implicated in development and regulation of cell differentiation (Fig. 4f).”

Enrichment for ASD-associated common variation is observed in CTX.M20 at FDR < 0.1.

“By contrast, enrichment for ASD-associated common variation was observed in CTX. M20 (FDR < 0.1, Extended Data Fig. 9h–1, Methods).”

Genome-wide (trans) DGE and differential splicing patterns are highly concordant between dup15q and ASD.

“Remarkably, genome-wide (trans) DGE and differential splicing patterns were highly concordant between dup15q and ASD (Fig. 3b, c, Extended Data Fig. 7c–e).”

CTX.M20 overlaps significantly with the ACP set.

“Remarkably, the upregulated module CTX.M20 was not found in previous analyses, overlaps significantly with the ACP set (FDR < 0.05, Extended Data Fig. 9a), and contains genes implicated in development and regulation of cell differentiation (Fig. 4f).”

Rare de novo mutations associated with intellectual disability are most strongly enriched in CTX.M22.

“One module, CTX.M24, exhibited significant enrichment for rare mutations found in ASD, while rare de novo mutations associated with intellectual disability were most strongly enriched in CTX.M22 (FDR < 0.05, Extended Data Fig. 9a).”

Specific lncRNA transcripts show primate-specific expression, being present in human and macaque or only in human but not in mouse.

“Examples of specific lncRNA transcripts that show primate-specific (in human and macaque, or only in human, but not in mouse) expression.”

Neuronal splicing factor expression levels across samples form three major clusters.

“demonstrating three major clusters”

Transcription factor motifs are significantly enriched in the ACP gene set.

“Enrichment statistics for transcription factor motifs found to be significantly enriched in the ACP set”

Differentially expressed lncRNAs show significantly different expression in brain compared with other tissues.

“Lines above the plot indicate pairwise significance with a one-sided Wilcoxon rank-sum test between brain and the other tissues.”

• cites [12] Jaffe, A. E. et al. Developmental regulation of human cortex transcription and its clinical relevance at single base resolution. Nat. Neurosci. 18, 154–161 (2015) sourceNot checked yet

  Not checked in the free run.

ASD versus control fold changes are similar in frontal cortex and temporal cortex.

“Comparison of fold changes between frontal cortex (FC) and temporal cortex (TC) for all samples, demonstrating similar changes in both regions.”

Differential splicing analysis identifies events that are independent of the differential gene expression signal.

“Differential splicing analysis identifies events independent of DGE signal.”

Four module eigengenes, CTX.M4, CTX.M9, CTX.M10, and CTX.M16, are associated with ASD.

“Temporal trajectories for four module eigengenes (CTX.M4, CTX.M9, CTX.M10, CTX.M16) associated with ASD”

Differentially expressed lncRNAs show significantly different expression in brain compared with embryonic stem cells and stem-cell-derived cell types.

“Similar to l, except for embryonic stem cells and stem-cell-derived cell types.”

ASD cortex differential gene expression principal component 1 differs significantly from controls for samples with dup15q mutations, ADI-R-confirmed diagnoses, and clinically supported diagnoses.

“Samples from individuals with diagnoses confirmed by dup15q mutations, confirmed by Autism Diagnostic Interview-Revised (ADI-R), and supported by clinical records are all significantly different from controls by two-sided pairwise Wilcoxon rank sum tests.”

The genes for these nine splicing events are not in the differential gene expression set but are detected in the differential alternative splicing set because of altered transcript structure.

“Notably, these genes are not in the DGE set, but are detected in the differential alternative splicing set owing to altered transcript structure.”

Three cerebellum modules are associated with ASD.

“Grey bars with labels signify three ASD-associated modules.”

ASD versus control differential gene expression fold changes replicate between previously evaluated and newly analyzed ASD samples in cortex when measured by RNA-seq.

“Independent replication of ASD versus control DGE fold changes between previously evaluated and new ASD samples in cortex by RNA-seq using samples from ref. 8”

• cites [8] Voineagu, I. et al. Transcriptomic analysis of autistic brain reveals convergent molecular pathology. Nature 474, 380–384 (2011) sourceNot checked yet

  Not checked in the free run.

ASD-associated modules are enriched with modules from ref. 4.

“Enrichment of ASD-associated modules with that from ref. 4.”

• cites [4] 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) sourceNot checked yet

  Not checked in the free run.

The first principal component of the cortex differential gene expression set is primarily associated with diagnosis and not with other factors.

“The first principal component of the cortex DGE set is primarily associated with diagnosis, and not with other factors.”

ASD versus control differential gene expression fold changes replicate between previously evaluated and newly analyzed ASD samples from cerebellum by microarray and RNA-seq.

“Independent replication analysis of ASD versus control DGE fold changes between previously evaluated and new ASD samples from cerebellum by microarray and RNA-seq using samples from ref. 8”

• cites [8] Voineagu, I. et al. Transcriptomic analysis of autistic brain reveals convergent molecular pathology. Nature 474, 380–384 (2011) sourceNot checked yet

  Not checked in the free run.

Autism spectrum disorder has a substantial genetic component.

“Autism spectrum disorder (ASD) involves substantial genetic contributions.”

• cites [1] Gaugler, T. et al. Most genetic risk for autism resides with common variation. Nat. Genet. 46, 881–885 (2014) sourceNot checked yet

  Not checked in the free run.

• cites [2] Gratten, J., Visscher, P. M., Mowry, B. J. & Wray, N. R. Interpreting the role of de novo protein-coding mutations in neuropsychiatric disease. Nat. Genet. 45, 234–238 (2013) sourceNot checked yet

  Not checked in the free run.

• cites [3] de la Torre-Ubieta, L., Won, H., Stein, J. L. & Geschwind, D. H. Advancing the understanding of autism disease mechanisms through genetics. Nat. Med. 22, 345–361 (2016) sourceNot checked yet

  Not checked in the free run.

SOX5 contributes to the reduction in regional gene expression differences in ASD.

“Our data suggest that SOX5, a transcription factor involved in neuron fate specification, contributes to this reduction in regional differences.”

Dysregulation of lncRNAs is an integral component of the transcriptomic signature of ASD.

“These data show that dysregulation of lncRNAs, many of which are brain-enriched, primate-specific, and predicted to affect protein expression through miRNA or FMRP interactions, is an integral component of the transcriptomic signature of ASD.”

The cortex is more selectively vulnerable to these transcriptomic alterations in ASD.

“The lack of significant DGE in the cerebellum is explained by the fact that changes in expression were consistently stronger in the cortex than in the cerebellum (Extended Data Fig. 2h), which suggests that the cortex is more selectively vulnerable to these transcriptomic alterations.”

Previous experimental perturbation of Rbfox1, SRRM4, and PTBP1 shows strong overlap with the differential splicing changes found in ASD cortex.

“Previous experimental perturbation of three splicing factors, Rbfox1 (ref. 18), SRRM4 (ref. 19), and PTBP1 (ref. 20), shows strong overlap with the differential splicing changes found in ASD cortex, further supporting these predicted relationships (Fig. 1h, Extended Data Fig. 5b).”

• cites [18] Lovci, M. T. et al. Rbfox proteins regulate alternative mRNA splicing through evolutionarily conserved RNA bridges. Nat. Struct. Mol. Biol. 20, 1434–1442 (2013) sourceNot checked yet

  Not checked in the free run.

• cites [19] Raj, B. et al. A global regulatory mechanism for activating an exon network required for neurogenesis. Mol. Cell 56, 90–103 (2014) sourceNot checked yet

  Not checked in the free run.

• cites [20] Gueroussov, S. et al. An alternative splicing event amplifies evolutionary differences between vertebrates. Science 349, 868–873 (2015) sourceNot checked yet

  Not checked in the free run.

Molecular convergence is likely a unitary phenomenon across multiple levels of transcriptome regulation in ASD.

“When we compared the first principal component across samples for protein coding DGE, lncRNA DGE and differential splicing, we found remarkably high correlations (R2 > 0.8), indicating that molecular convergence is likely to be a unitary phenomenon across multiple levels of transcriptome regulation in ASD (Fig. 1i).”

The difference in CTX.M19 and CTX.M20 trajectories is most consistent with an evolving process during early brain development that stabilizes starting in late childhood and early adolescence.

“We identified a remarkable difference in CTX.M19 and CTX.M20 during the first two decades of life (Fig. 4g, additional age trajectories in Extended Data Fig. 9f) that is most consistent with an evolving process during early brain development that stabilizes starting in late childhood and early adolescence.”

Diverse genetic perturbations can lead to phenotypic convergence at multiple biological levels in a complex neuropsychiatric disorder.

“Our findings illustrate how diverse genetic perturbations can lead to phenotypic convergence at multiple biological levels in a complex neuropsychiatric disorder.”

Because CTX.M20 is enriched for the ACP gene set, there is a potential link between polygenic risk and regional attenuation of gene expression in ASD.

“As CTX.M20 is enriched for the ACP gene set, this suggests a potential link between polygenic risk and regional attenuation of gene expression in ASD.”

The highly overlapping expression pattern shared by individuals with dup15q and the majority of those with idiopathic ASD represents an evolving adaptive or maladaptive response to a primary insult rather than a secondary environmental hit.

“Parsimony suggests that the highly overlapping expression pattern shared by individuals with dup15q and the majority of those with idiopathic ASD represents an evolving adaptive or maladaptive response to a primary insult rather than a secondary environmental hit.”

Attenuated patterning of SOX5 between cortical regions contributes to direct alterations in patterning of SOX5 targets.

“These findings support the prediction that attenuated patterning of the transcription factor SOX5 between cortical regions contributes to direct alterations in patterning of SOX5 targets.”

qRT-PCR confirmed the technical quality of the RNA-seq data.

“We performed a classification analysis to confirm that gene expression in ASD could separate samples by disease status (Extended Data Fig. 2a) and confirmed the technical quality of our data with qRT–PCR (Extended Data Fig. 2b, c).”

Genetic risk for ASD may influence changes in regional cortical gene expression.

“Co-expression network analysis reveals that individuals with ASD show age-related changes in the trajectory of microglial and synaptic function over the first two decades, and suggests that genetic risk for ASD may influence changes in regional cortical gene expression.”

The absence of neuronal splicing factor differential expression or correlation with splicing changes in cerebellum suggests these splicing factors contribute to cortex-biased differential splicing.

“The absence of neuronal splicing factor DGE or correlation with splicing changes in the cerebellum is consistent with the absence of a differential splicing signal in the cerebellum and suggests that these splicing factors contribute to cortex-biased differential splicing.”

Dup15q cortex differential gene expression findings are similar when different control samples are used.

“Comparison of DGE fold changes demonstrating that using different control samples (control samples used in the idiopathic analysis, column 2 of Extended Data Fig. 7b) for the dup15q cortex analysis yields similar findings.”

The period of synapse elimination and stabilization after birth in humans corresponds to late childhood.

“This corresponds to the period of synapse elimination and stabilization after birth in humans28,29, which may have significant implications for intervention.”

• cites [28] Huttenlocher, P. R. Morphometric study of human cerebral cortex development. Neuropsychologia 28, 517–527 (1990) sourceNot checked yet

  Not checked in the free run.

• cites [29] Khundrakpam, B. S., Lewis, J. D., Zhao, L., Chouinard-Decorte, F. & Evans, A. C. Brain connectivity in normally developing children and adolescents. Neuroimage 134, 192–203 (2016) sourceNot checked yet

  Not checked in the free run.

Primate-specific lncRNAs are probably relevant to understanding human higher cognition.

“Our analyses also reveal primate-specific lncRNAs that are probably relevant to understanding human higher cognition11,30.”

• cites [11] Necsulea, A. et al. The evolution of lncRNA repertoires and expression patterns in tetrapods. Nature 505, 635–640 (2014) sourceNot checked yet

  Not checked in the free run.

• cites [30] Zhang, Y. E., Landback, P., Vibranovski, M. D. & Long, M. Accelerated recruitment of new brain development genes into the human genome. PLoS Biol. 9, e1001179 (2011) sourceNot checked yet

  Not checked in the free run.

Aberrant microglia-neuron interactions reflect an early alteration in developmental trajectory that becomes more evident in late childhood.

“We interpret these data to suggest that aberrant microglia–neuron interactions reflect an early alteration in developmental trajectory that becomes more evident in late childhood.”

The moderate agreement between ref. 4 and this study is largely driven by concordant upregulation of microglial genes in both studies.

“Co-expression network analysis demonstrated that the moderate agreement is largely driven by concordance in upregulation of microglial genes in both studies”

• cites [4] 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) sourceNot checked yet

  Not checked in the free run.

Confounders do not drive the clustering of about two-thirds of ASD cortex samples.

“demonstrating that confounders do not drive clustering of about two-thirds of samples.”

Some of the transcriptome changes in ASD are likely consequences or compensatory responses rather than causal factors.

“Although we observe no significant association of the ASD-associated transcriptome signature with either clinical or technical confounders, some of the changes are likely to represent consequences or compensatory responses, rather than causal factors.”