This study provides the first evidence of alterations in expression of specific small regulatory non-coding RNA in the temporal cortex of autism spectrum disorder (ASD) brains. The major findings include the following: 1) There are differences in miRNA, snoRNA, and precursor miRNA in both the association cortex of the superior temporal sulcus (STS) and the primary auditory cortex (PAC). While the STS is involved in functions associated with the core features of ASD, the many changes in sncRNAs in primary auditory cortex were not expected as it is not functionally associated with core deficits of ASD. 2) Functional pathways for the gene targets of dysregulated miRNA across both regions were associated with nervous system, cell cycle, cell signaling, and other processes. 3) miRNAs that regulate immune functions were only altered in STS, supporting a role for immune dysfunction in STS associated with the core social deficits of autism. Even a few dysregulated miRNA and other sncRNAs have the potential to affect transcription and/or translation of hundreds of target genes and affect brain signaling, connectivity, and behavior.
Differentially expressed sncRNA in both primary and association cortex in ASD
As hypothesized, based on functional and structural differences known to be present in ASD STS, we found that the STS contained differentially expressed regulatory sncRNA. One mature miRNA, miR-4753-5p, exhibited increased expression levels and another, miR-1, had decreased expression in ASD relative to control tissue. Five immature, stem-loop precursor miRNA showed increased expression, and three had decreased expression in the STS. Given their regulatory role, miRNA can impact hundreds of gene targets and can greatly influence expression levels of mRNA and protein translation [
8], potentially making a substantial contribution to differences in overall gene expression levels in ASD.
There were also differences in a number of small nucleolar RNAs (snoRNAs) in STS in ASD. snoRNAs are small RNA molecules that act to guide chemical modifications of other RNAs including ribosomal RNAs, transfer RNAs, and small nuclear RNAs. Three snoRNAs, including ACA39, SNORA11C, SNORA27, exhibited increased expression, and one, U13 (Chr1), had decreased levels. Functionally, sequences within snoRNAs target them to specific mRNA targets, with the C/D box snoRNAs methylating target bases and H/ACA box snoRNAs being associated with pseudouridylation of target bases. These modifications likely affect translation as well as RNA silencing, telomerase maintenance, and alternative splicing [
38]. For example, U13 is involved in proper formation of the 3′ end of the 18S ribosomal subunit [
39], potentially impacting protein translation and alternative splicing. Aberrant expression of U13 could have widespread cellular effects in the brain. We find that non-coding snoRNAs along with non-coding miRNA are dysregulated in ASD brain, while other studies have also implicated long non-coding RNAs in ASD neuropathology [
40,
41].
To our knowledge, there have not been any previous studies targeting molecular abnormalities specifically in primary sensory cortex of ASD brains, making this the first report suggesting regulatory abnormalities in a primary sensory region. In PAC, two mature miRNAs exhibit increased expression, miR-664-3p and miR-4709-3p, and two decreased expression, miR-4742-3p and miR-297. Five snoRNAs showed decreased expression, including SNORA22, SCARNA6, and three paralogs of U13. Four stem-loop precursors were additionally up-regulated, while three had decreased expression.
It is notable that three of the identified precursor miRNA have been previously associated with ASD. miR-4436 has been described as a candidate susceptibility miRNA gene in autism [
42]. miR-548 is dysregulated in ASD lymphoblastoid cell lines [
43] and targets genes associated with several pathways implicated in ASD including ubiquitin, Wnt, axon guidance, LTP, and natural killer cells [
30]. Additionally, it can regulate host antiviral responses via direct targeting of IFN-λ1 [
44], which may have relevance for maternal immune activation models of ASD and other neurodevelopmental disorders [
45,
46]. miR-146 is reported to have altered expression in ASD cerebellar cortex [
47] and in ASD lymphoblastoid cell lines. In the nervous system, miR-146 (a,b) modulates AMPA receptor endocytosis [
48] and glial proliferation via Notch signaling [
49]. miR-146 also influences the activation state of brain microglia [
50]. Microglia are reported to increase in number, show evidence of activation, and show increased gene expression in ASD post-mortem brain tissue [
51‐
55]. Moreover, deficiencies in neuron-microglia signaling result in impaired functional brain connectivity and can influence social behavior [
56]. Although the functional importance of immature miRNA is not clear, the mature forms of the three stem-loop precursors found in the PAC have previously been associated with ASD.
A number of the predicted gene targets of the miRNA we found different between ASD and typically developing STS have previously been identified in studies looking at gene expression in the temporal cortex [
4,
33]. These target genes (INHBA, WDR1, TAGLN2, FNDC3B, ZFP36L1, CNN3, CPLX2, ZNF365, and GLS2) have important roles including those related to “cell communication and motility” and “cell fate and differentiation.” As we observed decreased miR-1 expression and increased miR-4753-5p expression in our study, we would expect to find higher and lower expression of the respective gene transcripts they target in STS. The regulation of these genes and others by altered patterns of miRNA may be at least in part responsible for regional and cellular differences that shape the social and behavioral phenotypes associated with ASD.
Dysregulation of sncRNAs in the PAC of ASD brain is a novel molecular finding we did not predict. Although sensory dysfunction is not considered a core diagnostic feature, neurophysiological studies have reported auditory abnormalities in ASD patients. Abnormalities in cortical auditory-evoked potentials have been reported when sound is presented to the left ear [
57]. In contrast, speech stimulation does not produce fMRI abnormalities in PAC of individuals with ASD [
23]. Atypical vertical sound localization and sound-onset sensitivity evidenced in ASD cases likely involves primary auditory areas [
58]. In addition, a magnetoencephalography study examining differences in auditory cortex function in young ASD and TD children found that there is atypical brain function in the auditory cortex in young children with ASD, regardless of language development [
59]. Coupled with neurophysiological findings, the dysregulated sncRNA in PAC in this study provide additional evidence for PAC dysfunction in ASD. Increased attention to primary cortices may be warranted in future investigations of ASD, as such dysfunction could conceivably contribute to cognitive impairments in complex social and linguistic tasks.
miRNA targets incorporated into pathways associated with ASD pathogenesis
We postulated that miRNA play an important role in regulating the transcriptional changes that have been reported in ASD brain tissue. Several genomic and gene expression studies of ASD brain have shown distinct and reproducible differences from typically developing cases that involve neurons/synapses and the immune system. In one study, Voineagu et al. showed consistent differences in transcriptome organization between ASD and typically developing brain by gene co-expression analysis [
4]. Moreover, two main modules of co-expressed genes in ASD brain were identified including a neuronal/synapse module that was associated with genes identified by GWAS in ASD and an immune module that was not associated with ASD GWAS genes [
4,
60]. These data supported prior studies of neuron, glia, and immune abnormalities in ASD brain [
33,
51,
61‐
67]. Another recent whole genome study showed age-related differences of gene expression in ASD brain that also affected neuron and immune pathways including neurogenesis, cell cycle, cell differentiation, cell death, and immune response in young ASD brain [
5]. A recent study used weighted gene co-expression network analysis (WGCNA) combined with ASD candidate genes to suggest that cortical projection neurons in layers V and VI of prefrontal/frontal cortex during 10–24 weeks post-conception were key cells in ASD pathogenesis [
6]. In a similar WGCNA analysis using different ASD candidate genes, the superficial (pyramidal) glutamatergic cortical projection neurons were implicated in ASD pathogenesis [
7].
We found many pathways associated with differentially expressed miRNA in ASD temporal cortex were similarly associated with nervous system and immune function, as well as basic cellular processes previously reported to be dysregulated in ASD. Of the 34 predicted pathways and processes that were represented in the target genes of miRNA in STS and PAC (Fig.
4), a number have been implicated in previous ASD studies including glutamatergic, dopaminergic, and cholinergic synapse pathways, axon guidance, synaptic vesicle cycling, neurotrophins, immune pathways, hedgehog and Wnt signaling, spliceosome, and RNA degradation [
2‐
4,
27,
60]. Comparing our broader temporal cortex findings with previous analyses, it is notable that 44 of the 75 pathways regulated by differentially expressed miRNA in our study are shared with the 57 pathways reported to be affected by copy number variants in ASD [
34] (Fig.
5, Additional file
1: Table S4). Moreover, we found 7 and 15 pathways, respectively, are shared with the 16 and 24 pathways reported to demonstrate alterations in the two prior analyses of mRNA expression ASD temporal cortex [
4,
33] (Fig.
5). Common functions associated with pathways across all studies of ASD temporal cortex include cell cycle, metabolic, and neuron-related pathways and processes. Given the substantial number of convergent pathways between studies of temporal cortex mRNA and miRNA expression, it is highly plausible that miRNA are exerting a regulatory influence over mRNA associated with atypical cortical function in ASD.
Within the temporal lobe, two pathways were shared between PAC and STS, the “PI3K-Akt pathway” and “pathways in cancer.” The PI3K-Akt pathway is of interest as it has been implicated in a number of previous ASD studies. The PI3K-Akt/mTOR pathway is dysregulated in monogenic disorders associated with ASD including Fragile X, tuberous sclerosis, neurofibromatosis, and PTEN [
68]. Rare and even some common genetic variants associated with ASD affect PI3K-Akt signaling [
2,
69]. Studies of gene expression in blood of ASD subjects have found abnormal PI3K-Akt/mTOR signaling [
70]. Our studies of alternative splicing of mRNA expressed in blood leukocytes detected abnormalities of PI3K-Akt signaling in about half the 2–4 year old ASD children with idiopathic ASD [
27]. Thus, the current study suggests abnormal PI3K-Akt signaling in temporal cortex of ASD subjects in both primary auditory and STS association cortex, and previous studies suggest PI3K-Akt abnormalities in blood and differential alternative splicing of PI3K-Akt pathways in blood of many ASD subjects. These results would support this as a potential therapeutic target in at least a subgroup of ASD subjects as has been suggested by others [
71].
One of the major findings of this study is that few predicted miRNA-regulated pathways were shared between the PAC and STS, despite the fact that they are adjacent cortical structures. Because many pathways are specific for PAC or STS, there is evidence that differential regulation of miRNA, their mRNA targets, and associated pathways can be region specific in ASD.
Differentially expressed miRNA having a significant number of gene targets in immune and most of the cell cycle KEGG pathways (except “pathways in cancer”) were restricted only to STS. Cell cycle genes might be related to neuronal cell death and/or to glial proliferation and perhaps linked to increased microglial and neuron numbers in ASD brain [
51‐
53,
63,
72]. Observed immune pathways in this study, related to both viruses and bacteria, are consistent with many reports of immune pathway activation in ASD brain and blood [
4,
5,
33,
60,
73‐
75]. Since the immune regulated genes in ASD brain were not among the GWAS-associated ASD genes [
4], these immune genes might reflect the environmental factor in ASD pathogenesis. ASD is considered to be a complex genetic disorder in which several genes interact with an environmental precipitant. Activation of immune and cell cycle pathways in STS might suggest that this is a region of primary pathology, and lack of immune and cell cycle activation in PAC might suggest gene changes in PAC are secondary to those in STS association cortex because of interconnections between the two regions. At the very least, these findings suggest molecular and immune disruption pathologies present in the temporal association cortex that are not shared with primary sensory cortex. This further emphasizes the importance of assessing molecular abnormalities in distinct cortical regions.