Research ArticleAutism

A Noncoding RNA Antisense to Moesin at 5p14.1 in Autism

See allHide authors and affiliations

Science Translational Medicine  04 Apr 2012:
Vol. 4, Issue 128, pp. 128ra40
DOI: 10.1126/scitranslmed.3003479

Abstract

People with autism spectrum disorder (ASD) are characterized by deficits in social interaction, language, and behavioral flexibility. Rare mutations and copy number variations have been identified in individuals with ASD, but in most patients, the causal variants remain unknown. A genome-wide association study (GWAS), designed to identify genes and pathways that contribute to ASD, indicated a genome-wide significant association of ASD with the single-nucleotide polymorphism (SNP) rs4307059 (P = 10−10), which is located in a gene-poor region of chromosome 5p14.1. We describe here a 3.9-kb noncoding RNA that is transcribed from the region of the chromosome 5p14.1 ASD GWAS association SNP. The noncoding RNA was encoded by the opposite (antisense) strand of moesin pseudogene 1 (MSNP1), and we therefore designated it as MSNP1AS (moesin pseudogene 1, antisense). Chromosome 5p14.1 MSNP1AS was 94% identical and antisense to the X chromosome transcript of MSN, which encodes a protein (moesin) that regulates neuronal architecture. Individuals who carry the ASD-associated rs4307059 T allele showed increased expression of MSNP1AS. The MSNP1AS noncoding RNA bound to MSN, was highly overexpressed (12.7-fold) in postmortem cerebral cortex of individuals with ASD, and could regulate levels of moesin protein in human cell lines. These data reveal a biologically functional element that may contribute to ASD risk.

Introduction

Autism spectrum disorder (ASD) is a lifelong neurodevelopmental disorder with childhood onset. Males are diagnosed four times more often than females, and the prevalence of ASD has increased to 1 in 110. Although environmental factors contribute to the risk of ASD (13), studies of twin and sibling recurrence risk suggest that ASD is among the more heritable of neuropsychiatric disorders (49), a fact that has encouraged research into the genetic basis of ASD risk (1014). Genome-wide association studies (GWASs) have proven successful in identifying causal genetic variants in other complex genetic disorders (1522), but application of GWAS to ASD has yielded mixed results (2325). Wang et al. (23) described genome-wide significant association (P = 10−10) in a total of four cohorts including 3115 cases and 7031 controls. The most statistically significant finding was association at chromosome 5p14.1 (23). The association peaked at a single marker (rs4307059), which was flanked by five other markers that all achieved genome-wide significance (P < 10−8) (23), indicating that the association at rs4307059 was not due to a technical artifact. Further, the ASD-associated rs4307059 genetic variant was identified as a predictor of stereotyped conversation and poorer communication skills such as maintaining appropriate conversational distance and eye contact in a population-based sample of >7000 individuals (26), suggesting that rs4307059 may be a quantitative trait locus for social communication phenotypes. The biological evidence for this conclusion is far less compelling. The significant association signal lies between the genes encoding cadherin 9 (CDH9) and cadherin 10 (CDH10), but the GWAS peak maps ~1 million base pairs (bp) from either cadherin gene. Further, brain expression of CDH9 and CDH10 did not correlate with the genotype at rs4307059 (23), suggesting that the GWAS-associated signal does not control expression of either CDH9 or CDH10.

We hypothesized that the chromosome 5p14.1 GWAS peak indicated the presence of a functional genetic element. Applying a strategy that has proven successful in identifying functional elements in non–protein-coding risk regions (27), we tested 5p14.1 for functional elements relevant to ASD.

Results

MSNP1 is present within the ASD GWAS peak

Bioinformatic analysis of expressed sequence tags (ESTs) and tiling arrays in the area around the chromosome 5p14.1 ASD GWAS signal suggested that a single ~4-kb RNA was expressed directly under the peak (Fig. 1). The ~4-kb RNA corresponded to moesin pseudogene 1 (MSNP1), which does not appear to be transcribed and which has 94% sequence identity to the mature mRNA of the protein-coding gene MSN, located on the X chromosome. The MSN gene contains 13 exons spanning 74 kb of chromosome Xq11.2, produces a 4-kb mRNA, and encodes the 577–amino acid protein moesin. The MSNP1 pseudogene lacked the intronic sequences of MSN.

Fig. 1

MSNP1AS maps within the chromosome 5p14.1 GWAS significant ASD-associated peak. (A) ASD-associated markers on chromosome 5p14.1 between CDH10 and CDH9 [adapted from (23)]. (B) A ~4-kb RNA transcribed from 5p14.1, as indicated by ESTs and RNA localization (from the genome-wide ENCODE tiling array project). (C) The + strand of the 4-kb 5p14.1 region is the MSNP1. The − strand produces a noncoding 3.9-kb RNA, designated MSNP1AS.

The MSNP1 locus expresses a ~4-kb transcript antisense to MSNP1 and MSN

To determine whether a transcript is expressed from the MSNP1 locus, we performed Northern hybridization on total RNA from three human cell lines [HEK (human embryonic kidney), SK-N-SH, and SH-SY5Y] with 1.2-kb single-stranded RNA probes. The probe in antisense orientation to MSNP1 hybridized to a single RNA band consistent in size with the 4-kb MSN mRNA (Fig. 2). Excision of the band followed by reverse transcription–polymerase chain reaction (RT-PCR) and direct resequencing confirmed that the band represents expression of the chromosome Xq11.2 MSN gene. There was no evidence of RNA transcribed from the chromosome 5p14.1 pseudogene MSNP1 in this band. The probe in sense orientation to MSNP1 also hybridized to a single RNA band of ~4 kb (Fig. 2). Excision of the band followed by RT-PCR and direct resequencing indicated that this RNA was identical in sequence to the chromosome 5p14.1 genomic sequence and that the opposite (−) strand of MSNP1 was transcribed to produce MSNP1AS (moesin pseudogene 1, antisense) (Fig. 1). These results indicated that the site of transcription of MSNP1AS, a ~4-kb RNA that can bind specifically the mature mRNA of MSN, lies within the chromosome 5p14.1 genome-wide significant ASD GWAS signal.

Fig. 2

MSNP1AS is expressed and hybridizes to MSN. (A) Ethidium bromide–stained agarose formaldehyde gel with abundant 18S and 28S ribosomal RNA, indicating high-quality total RNA isolation from each of the three cell lines. (B) Northern hybridization with 1.2-kb probe antisense to MSNP1, showing a single transcript of ~4 kb. Excision of the band, RT-PCR, and direct resequencing confirmed that the transcript was the 3981-bp mature MSN mRNA transcribed from the X chromosome. (C) Northern hybridization with a 1.2-kb probe sense to MSNP1, showing a single transcript of ~4 kb. Excision of the band, RT-PCR, and direct resequencing confirmed that the transcript is a 3938-bp MSNP1AS transcript from chromosome 5p14.1. Size, DNA/RNA size marker ladder; HEK, human embryonic kidney cell line; SK, SK-N-SH human neuronal cell line; 5Y, SH-SY5Y human neuronal cell line.

MSNP1AS is 94% identical to MSN

Because our 1.2-kb Northern probe hybridized specifically to a ~4-kb RNA species, we did not know the extent of the chromosome 5p14.1 genomic sequence from which MSNP1AS is transcribed. Therefore, we performed a series of 5′ and 3′ rapid amplification of complementary DNA (cDNA) ends (RACE) experiments on total RNA isolated from the three human cell lines to define the length of the MSNP1AS transcript. MSNP1AS was a 3938-bp transcript with no open reading frame larger than 30 amino acids. MSNP1AS had 94% sequence identity in the reverse complement to MSN mRNA (NM_002444.2). The sequence of the RNA precisely matched the chromosome 5p14.1 genomic sequence; there were no introns in MSNP1AS. MSNP1AS corresponded to NT_006576.16 nucleotides 25903357 to 25899418. The 1.2-kb probe used for Northern hybridization corresponded to MSNP1AS nucleotides 2529 to 3695.

MSNP1AS is differentially expressed in human tissue

To determine the relative levels of MSN and MSNP1AS transcripts in a panel of human tissues, we designed custom TaqMan Gene Expression assays to target the region with the highest degree of sequence diversity between chromosome X MSN and chromosome 5 MSNP1AS. The probe was an 18-mer with six divergent nucleotides that distinguish MSN from MSNP1AS (fig. S1). Application of these quantitative PCR (qPCR) assays to cDNA derived from 15 normal human tissue samples revealed that MSN is expressed abundantly in all of the tissues tested, including multiple regions of cerebral cortex, heart, lung, and kidney in both adult and fetal tissue [expression relative to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) ranged from 0.3 to 87%] (Fig. 3). In contrast, the expression of MSNP1AS in the same tissue samples was lower and highly variable (Fig. 3). MSNP1AS was most abundant in adult temporal cerebral cortex, adult peripheral blood, fetal heart, and three immortalized cell lines (expression relative to GAPDH ≈ 0.1%). MSNP1AS was expressed at lower levels (expression relative to GAPDH ≈ 0.0001%) in seven other tissues, including adult frontal and occipital cerebral cortices, adult cerebellum, and adult heart (Fig. 3). MSNP1AS transcript was undetectable in five tissues, including fetal frontal cerebral cortex, adult lung, and adult skin (Fig. 3). These data indicate that MSN is expressed at consistently high levels, but that MSNP1AS is expressed at variable levels among human tissues.

Fig. 3

Expression of MSN and MSNP1AS in a panel of human tissues. To determine the patterns of MSN and MSNP1AS expression in human adult and fetal tissues, we performed qPCR on a panel of human tissue sources. Each data point is based on a single RNA sample, and data are calculated relative to GAPDH expression. ND, not detected (Ct > 50 cycles); FC, frontal cerebral cortex; TC, temporal cerebral cortex; OC, occipital cerebral cortex; Cb, cerebellum; SC, spinal cord; PB, peripheral blood; He, heart; Lu, lung; Ki, kidney; Sk, skin; HEK, human embryonic kidney cell line; SK, SK-N-SH human neuronal cell line; 5Y, SH-SY5Y human neuronal cell line.

MSNP1AS is increased in postmortem brain tissue of individuals with ASD

To determine whether MSN and MSNP1AS expression is altered in the brains of individuals with ASD, we performed qPCR on total RNA prepared from fresh-frozen, superior temporal gyri of 10 adult ASD-control pairs (table S1). MSNP1AS was increased 12.7-fold in the temporal cortex of individuals with ASD compared to controls (P = 0.004; Fig. 4A). Consistent with published microarray results in postmortem temporal cortex of individuals with ASD (28, 29), MSN was significantly increased 2.4-fold in ASD compared to controls (P = 0.029; Fig. 4B), but there was no significant change in the expression of either of the two protein-coding genes flanking the ASD GWAS region: CDH9 (P = 0.17; Fig. 4C) or CDH10 (P = 0.83; Fig. 4D).

Fig. 4

MSNP1AS expression is increased in postmortem brains of individuals with ASD. (A) MSNP1AS expression in temporal cortex of individuals with ASD and controls (Con) (P = 0.004). Relative expression was 0.288 ± 1.528 for controls and 3.649 ± 1.860 for individuals with ASD. (B) MSN expression in temporal cortex of individuals with ASD and controls (P = 0.029). Relative expression was 3.165 ± 1.113 for controls and 7.672 ± 1.478 for individuals with ASD. (C) CDH9 expression in temporal cortex of controls and individuals with ASD (P = 0.172). Relative expression was 0.939 ± 1.108 for controls and 0.530 ± 1.440 for individuals with ASD. (D) CDH10 expression in temporal cortex of controls and individuals with ASD (P = 0.831). Relative expression was 1.197 ± 1.109 for controls and 1.279 ± 1.319 for individuals with ASD. *, significant difference by unpaired Student’s t test.

MSNP1AS is correlated with ASD-associated genotype in postmortem brain samples

To determine whether genotype at the ASD-associated GWAS markers correlated with expression of MSNP1AS, CDH9, or CDH10, we isolated genomic DNA from each of the postmortem samples of temporal cerebral cortex, and we determined the genotype at rs7704909, rs12518194, and rs4307059 by direct resequencing. These genetic markers are in high linkage disequilibrium (LD) with each other (r2 > 0.98) (30), and each achieved genome-wide significant association with ASD in the Wang et al. (23) GWAS. Consistent with it having a genetic influence on ASD risk, MSNP1AS RNA was significantly increased in individuals with the ASD risk genotypes (Fig. 5, B to D). MSNP1AS expression was increased: 23.3-fold in individuals with the ASD-associated rs7704909 T/T genotype compared to the C/C genotype (Fig. 5B), 10.8-fold in individuals with the ASD-associated rs12518194 A/A genotype compared to the G/G genotype (Fig. 5C), and 22.0-fold in individuals with the ASD-associated rs4307059 T/T genotype compared to the C/C genotype (Fig. 5D). Individuals with heterozygous genotypes showed an intermediate level of MSNP1AS that was not significantly different from either of the homozygous genotypes (Fig. 5, B to D). Stratification by ASD diagnosis revealed significant main effects for both ASD diagnosis and genotype, with the highest expression of MSNP1AS observed in individuals with ASD and the risk genotype (fig. S2). In contrast to these positive results for MSNP1AS, there was no correlation between the ASD-associated GWAS marker genotypes and expression of either CDH9 (Fig. 5, E to G) or CDH10 (Fig. 5, H to J).

Fig. 5

MSNP1AS expression, but not CDH9 or CDH10 expression, is correlated with ASD risk genotypes at 5p14.1. (A) Map of chromosome 5p14.1 showing the nucleotide locations of MSNP1AS in relation to the markers with genome-wide significant association with ASD (rs7704909, rs12518194, and rs4307059). (B to D) Relative expression of MSNP1AS in postmortem temporal cortex after stratification by genotype at the rs7704909 (B), rs12518194 (C), and rs4307059 (D) loci. (E to G) Relative expression of CDH9 in postmortem temporal cortex after stratification by genotype at the rs7704909 (E), rs12518194 (F), and rs4307059 (G) loci. (H to J) Relative expression of CDH10 in postmortem temporal cortex after stratification by genotype at the rs7704909 (H), rs12518194 (I), and rs4307059 (J) loci. Asterisk, significant difference from opposite homozygous genotype by ANOVA followed by Fisher’s PLSD test.

Moesin protein is unchanged in postmortem brain of individuals with ASD

To determine whether increased expression of MSN and MSNP1AS altered the amount of moesin protein in the cerebral cortex of adult ASD subjects (table S1), we performed Western blot analyses with anti-moesin antibodies (Fig. 6A). We saw no significant difference in moesin protein between ASD and control temporal cerebral cortex samples (Fig. 6B). To determine potential mechanisms of moesin protein regulation, we performed quantitative correlation analyses on the RNA and protein levels of the postmortem temporal cortex samples. As expected, MSN RNA levels were positively associated with moesin protein levels (r = +0.45; P = 0.0009; Fig. 7A), consistent with MSN transcript levels being a major determinant of moesin protein levels. MSNP1AS transcript levels were also positively correlated with MSN transcript levels (r = +0.77; P = 0.0001; Fig. 7B), as would be expected if they were co-regulated. Moesin protein levels were negatively correlated with the difference between MSN and MSNP1AS expression levels (r = −0.44; P = 0.039; Fig. 7C), consistent with the possibility that MSNP1AS contributes to the regulation of moesin protein.

Fig. 6

Moesin protein levels are unchanged in postmortem brains of individuals with ASD compared to controls. (A) Examples of Western blots for moesin and GAPDH loading control in postmortem temporal cortex samples from two ASD patients and two controls. (B) Quantitation of moesin protein in individuals with ASD and controls (P = 0.120). For controls, average moesin/GAPDH density = 0.339 ± 0.038; for individuals with ASD, average moesin/GAPDH density = 0.524 ± 0.098.

Fig. 7

Expression levels of MSN RNA, MSNP1AS RNA, and moesin protein are correlated in postmortem cerebral cortex samples. (A) Association of MSN transcripts with moesin protein levels (r = +0.45; P = 0.0009). (B) Association of MSNP1AS transcript levels with MSN transcript levels (r = +0.77; P = 0.0001). (C) Negative correlation of moesin protein levels and the difference between MSN and MSNP1AS transcript levels (r = −0.44; P = 0.039). RNA data are shown relative to GAPDH expression.

Overexpression of MSNP1AS in human cell lines decreased moesin protein levels

To determine whether MSNP1AS overexpression could alter moesin protein levels, we transfected HEK and SK-N-SH cell lines with overexpressing constructs of MSNP1AS, MSN, and MSNP1. SH-SY5Y cells do not express endogenous moesin protein (fig. S3) and thus were not used in these experiments. The HEK and SK-N-SH cells were harvested 24 and 72 hours after transfection, and moesin protein was measured by Western blotting.

In HEK cells 24 hours after transfection, overexpression of MSNP1AS caused a significant 38% reduction in endogenous moesin protein levels (P = 0.019; Fig. 8), confirming that MSNP1AS can regulate moesin protein accumulation. In contrast, overexpression of MSNP1, which is not expressed endogenously, did not alter moesin protein levels (P = 0.51; Fig. 8), suggesting that this transcript would not regulate moesin protein if expressed. Overexpression of a full-length MSN construct caused a significant 18-fold increase in moesin protein (P = 0.001; Fig. 8). Cotransfection of MSNP1AS with MSN resulted in levels of moesin protein that were significantly lower than transfection of MSN alone (P = 0.04) but were also significantly higher than the negative control (P = 0.0009) (Fig. 8), indicating that a balance of MSN and MSNP1AS is critical to the expression of moesin protein.

Fig. 8

Overexpression of MSNP1AS in human cell lines causes decreased expression of moesin protein. (A) Representative Western blot of moesin and GAPDH after transfection of MSN alone, MSN plus MSNP1AS, MSNP1 alone, or MSNP1AS alone. (B) Quantitative summary of moesin protein expression in HEK and SK-N-SH cells at 24 and 72 hours after transfection. Note that the y axis is a log scale. Error bars represent SEM. n = 4 to 6. *P < 0.05, significant difference from the control by ANOVA followed by Scheffe F test.

In HEK cells 72 hours after transfection, a time-dependent relationship among the moesin-related species was revealed (Fig. 8). Overexpression of MSNP1AS caused a significant 39% reduction in moesin protein (P = 0.002). Overexpression of MSNP1 failed to alter moesin protein expression (P = 0.37). Overexpression of MSN caused a fourfold increase in moesin protein (P = 0.002). Cotransfection of MSNP1AS with MSN caused a significant reduction in moesin protein compared to transfection of MSN alone (P = 0.003). In contrast to the 24-hour time point, however, cotransfection of MSNP1AS with MSN resulted in levels of moesin protein that were indistinguishable from those of the control (P = 0.22) (Fig. 8), suggesting a time dependence of the regulation of moesin protein by MSNP1AS that is consistent with our observation that moesin protein levels were unchanged in postmortem temporal cortex tissue.

In human neuronal SK-N-SH cells, the results were similar to those observed in HEK cells (Fig. 8). Overexpression of MSNP1AS caused a significant reduction in endogenous moesin protein at 24 hours (45%; P = 0.0004) and 72 hours (40%; P = 0.0007). Overexpression of MSNP1 failed to alter moesin protein levels at either time point (24 hours: P = 0.45; 72 hours: P = 0.98). Overexpression of MSN caused a significant increase in moesin protein (24 hours: 4-fold; 72-hours: 2.5-fold, P = 0.002). Cotransfection of MSNP1AS with MSN resulted in levels of moesin protein that were intermediate between the negative control and transfection of MSN alone, with the levels of moesin protein at 72 hours being significantly different from transfection of MSN alone (P = 0.002) but indistinguishable from control (P = 0.27) (Fig. 8).

Discussion

Here, we identified a previously unknown, functional noncoding RNA, MSNP1AS, which is transcribed within the site of a genome-wide significant association signal for ASD at chromosome 5p14.1. Expression of MSNP1AS was correlated with the ASD-associated alleles, with homozygous carriers of the ASD risk alleles showing higher levels of MSNP1AS expression. Expression of MSNP1AS was increased 12.7-fold in the postmortem samples of temporal cortex of individuals with ASD. Previous studies with microarray analysis of >20,000 known transcripts and confirmatory qPCR have shown no protein-coding RNA with expression that was increased more than 10-fold in ASD (28, 29). Being previously unreported, MSNP1AS was not represented on any of these previously described microarray platforms (28, 29). Thus, MSNP1AS represents an ASD candidate gene with genome-wide significant genetic association, functional correlation with ASD-associated alleles, and a large change in expression in adult postmortem brain. As we have shown, MSNP1AS is antisense to and can bind MSN transcript, and overexpression of MSNP1AS causes a decrease in moesin protein.

Moesin (membrane-organizing extension spike protein) is a member of the ERM (ezrin/radixin/moesin) family of proteins that link the cellular membrane to the actin cytoskeleton. In the developing brain, moesin protein is localized to cellular regions of high motility and growth, including growth cones and filopodia that emanate from neurite shafts. In rat cerebral cortex, expression of ERM proteins peaks near birth and slowly declines through postnatal development (31), consistent temporally with their playing a role in synaptic maturation. At embryonic day 17, ERM proteins are prominently expressed in extending neuronal processes in the intermediate zone of developing rat cerebral cortex (32), as would be expected if they contributed to synapse formation. Knockdown of moesin with antisense RNA techniques in cultured cortical and hippocampal neurons results in marked phenotypes, including growth cone collapse (31), suppressed neurite formation (33), a 10-fold reduction in neurite advancement rate (31), suppression of glutamate-induced increase in active presynaptic boutons (34), and suppression of estrogen-induced increase in the formation of dendritic spines (35). These data indicate that moesin likely functions both presynaptically to maintain axonal growth cone development and postsynaptically to induce dendritic spine formation. Moesin-deficient mice exhibit altered responses to injuries of the liver and lung, but neither behavior nor brain development has been examined in moesin knockout mice (3638).

Our results indicate that the genome-wide significant signal on chromosome 5p14.1 may arise from a pathway that contributes to ASD risk involving regulation of moesin, which may be altered in ASD. Further, network analysis of microarray data from postmortem ASD brain placed MSN as a central node (29). Although it has not been possible to examine directly in people with ASD (due to a lack of postmortem brain samples during development), a decrease of moesin protein at critical developmental stages could potentially contribute to altered short- and long-range connectivity in the brains of individuals with ASD (3944) or to the early brain overgrowth and later reduction in brain size beginning at 2 to 3 years in ASD (4547). Although our results indicate that moesin protein levels were not changed in postmortem temporal cortex of individuals with ASD, alterations in moesin protein could occur only at developmentally relevant time points in individuals with ASD. The fact that moesin protein is reduced markedly and specifically in the cerebral cortex of fetuses with Down syndrome compared to controls (48) supports the idea that moesin dysregulation may be involved with abnormal brain development. Whole-genome analyses have not yet identified copy number variations associated with ASD that disrupt either the MSN or the MSNP1AS locus.

Our results show that a noncoding RNA transcribed at the ASD GWAS site on chromosome 5 (MSNP1AS) may regulate protein expression of a gene transcribed on chromosome X (MSN). The antisense RNA MSNP1AS directly binds MSN in human cells. A similar, but functionally distinct, mechanism was recently reported in cancer, in which expression of the noncoding pseudogene PTENP1 acts as a functional decoy and disrupts microRNA regulation of the oncogene PTEN (49). Deletions of a complex noncoding RNA locus (PTCHD1AS1/PTCHD1AS2) 5′ to the X chromosome patched-related (PTCHD1) locus are associated with ASD (50). Although how these RNAs contribute to disease has not been explored, they are presumed to regulate expression of the protein-coding gene on the opposite strand, PTCHD1, which is itself associated with ASD by microdeletions and mutations (50, 51). Many disease-associated genetic signals are at genomic locations outside annotated protein-coding genes (22, 52). Consequently, future analyses of GWAS associations, copy number variations, and rare noncoding mutations in ASD should consider the possibility that the functional element may not be the nearest protein-coding gene, but rather may be a noncoding RNA.

The limitations of this study include our focus on a single noncoding RNA (MSNP1AS) and a single brain region (superior temporal gyrus). It is possible that other functional noncoding elements exist under the chromosome 5p14.1 GWAS peak. We have shown that MSNP1AS is variably expressed among brain regions (Fig. 3), but the prediction that moesin protein levels will be in higher abundance in frontal and occipital cortices compared to temporal cortex needs to be tested directly. A further limitation of the study is that we hypothesize, on the basis of correlated expression, that MSN and MSNP1AS are co-regulated in temporal cortex. The mechanism of this co-regulation, as well as the cell types that express MSNP1AS and MSN during development, has not been identified. Investigations in animal models will be necessary to determine the impact of MSNP1AS expression and moesin deficiency on brain development.

Materials and Methods

Bioinformatics

The University of California Santa Cruz (UCSC) Genome Browser was used to identify human ESTs and RNA identified by the ENCODE Affymetrix/CSHL Subcellular RNA Localization by Tiling Array within the 100-kb LD block defined by the Wang et al. (23) GWAS peak. No potential functional elements other than a single noncoding RNA, MSNP1AS, were identified by this bioinformatics analysis (Fig. 1). Our finding of a single noncoding RNA in this region encouraged us to focus exclusively on MSNP1AS, although it is possible that functional elements are yet to be discovered. Since we completed our original bioinformatics analysis, several genome-wide RNA-Seq tracks have been added to the UCSC Genome Browser that confirm transcript expression from the MSNP1AS locus in several tissue sources, often as the only significantly expressed transcript within 500 kb of rs4307059. If we were beginning our bioinformatics analysis again today, we would start with examination of RNA-Seq data.

Isolation of RNA and protein from human cell lines

Human neuronal cell lines, SH-SY5Y and SK-N-SH, and HEK cells were purchased from the American Type Culture Collection. Cells were grown on standard medium with 10% fetal bovine serum in 75-cm flasks at 37°C and 5% CO2. Cells were grown to 70% confluency and split 24 hours before harvest to six-well 10-cm plates. Cells were harvested by scraping in cold phosphate-buffered saline. RNA and protein were isolated with the triplePrep kit (GE Healthcare) according to the manufacturer’s protocol. Each cell line was harvested on four separate occasions, with each harvest representing an independent experiment.

Northern hybridization

A 1.2-kb genomic DNA fragment of MSNP1 from chromosome 5p14.1 was PCR-amplified and cloned into a vector (pSTBlue-1). The chromosome 5p14.1 genomic DNA fragment is 95% identical to nucleotides 208 to 1374 of the mature MSN mRNA (NM_002444.2). Single-stranded RNA probes were transcribed in each orientation from the T7 and SP6 RNA polymerase sites of pSTBlue-1 with the DIG Northern Starter kit (Roche). Total RNA was isolated from each of the cell lines with the Qiagen RNeasy kit according to the manufacturer’s protocol. Total RNA was loaded into two independent lanes (1 μg each lane) of a 2% formaldehyde/agarose gel, and RNA species were separated by electrophoresis. Size-separated RNA was transferred from the formaldehyde/agarose gel to nylon membrane by overnight capillary transfer. We then cut the membrane in half and hybridized each half separately with one of the complementary in vitro–transcribed RNA probes.

Rapid amplification of cDNA ends

5′ RACE was performed with the Invitrogen 5′ RACE kit (version 2.0) according to the manufacturer’s protocol. 3′ RACE was performed with the Invitrogen 3′ RACE kit according to the manufacturer’s protocol. Total RNA was isolated from HEK, SK-N-SH, and SH-SY5Y cells with the Qiagen RNeasy kit according to the manufacturer’s protocol.

Quantitative RT-PCR

Custom TaqMan Gene Expression assays were developed for MSNP1AS (assay ID AIX0ZZG) and MSN (assay ID AIWR1S8) in collaboration with the Life Technologies Research & Development team. The assays use primers that overlap between MSNP1AS and the mature MSN transcript (fig. S1). The probes are designed to bind to the region with the highest degree of sequence diversity between the two transcripts, an 18-mer with six divergent nucleotides (fig. S1). TaqMan Gene Expression array assays for CDH9 (assay ID Hs00940349_m1) and CDH10 (assay ID Hs00198772_m1) were also purchased from Life Technologies. qPCR assays were performed with the TaqMan Gene Expression Master Mix (Applied Biosystems). qPCR for each sample was run in triplicate for MSNP1AS and MSN, along with two “housekeeping” genes for normalization: GAPDH (assay ID Hs9999905_m1) and POL2RA (assay ID Hs00172187_m1). All results presented represent normalization to GAPDH; statistical conclusions were identical for normalizations to POL2RA. qPCR assays were run on a Bio-Rad CFX96 Real-Time System and analyzed with CFX Manager software (version 2.1).

Tissue panel qPCR

Total RNA was purchased from BioChain for 10 human adult tissues and five human fetal tissues. cDNA was generated with the Invitrogen SuperScript III First-Strand Synthesis kit according to the manufacturer’s protocol. qPCR was performed in triplicate for each tissue sample with four assays (MSN, MSNP1AS, GAPDH, and POL2RA) as described above.

Isolation of DNA, RNA, and protein from human postmortem brain samples

Fresh-frozen postmortem brain samples were obtained from the Autism Tissue Program (Autism Speaks). Ten ASD brain samples were matched to 10 control samples on the basis of sex, age, and postmortem interval (table S1). Each tissue sample was a 0.5-cm3 block of superior temporal gyrus (Brodmann area 22). Four independent dissections of each block were performed with a razor blade to excise a ~0.5-mg sliver from each frozen block. Genomic DNA, RNA, and protein were isolated with the triplePrep kit (GE Healthcare) according to the manufacturer’s protocol. Each dissection was considered a separate experiment.

Genotyping of rs7704909, rs12518194, and rs4307059 loci

For each genetic marker, PCR primers were designed to uniquely amplify a 700- to 1000-bp fragment centered on the ASD-associated genetic variant. Genomic DNA was amplified with the KOD Xtreme Hot Start PCR kit (EMD Biosciences), as described in the manufacturer’s protocol. Each amplicon was submitted to direct resequencing (Eton Biosciences), and chromatograms were analyzed with Sequencher software.

Western blot

Proteins were separated by electrophoresis on a 4 to 20% gradient polyacrylamide gel (Bio-Rad). Proteins were transferred to nitrocellulose membrane by capillary transfer with a Bio-Rad Criterion blotter system. Primary antibody directed to moesin (Cell Signaling) was diluted 1:1000. Secondary antibody was anti-rabbit immunoglobulin G, horseradish peroxidase–linked (Cell Signaling) diluted 1:2000. Images were captured on a BioSpectrum Imaging System (UVP) and analyzed with VisionWorks LS version 6.8 software. Membranes were reprobed with antibody directed to GAPDH (Cell Signaling) diluted 1:2000.

Transfection of overexpression constructs

Full-length MSN and MSNP1AS were PCR-amplified from cDNA isolated from HEK cells with the KOD Xtreme PCR kit (Novagen). Each ~4-kb PCR was cloned into pSTBlue-1 (Novagen) according to the manufacturer’s protocol. Minipreps were sequenced to identify clones with inserts appropriate for directional subcloning into the mammalian transfection vector pIRES2-AcGFP1 (Clontech) using Kpn I–Hind III digests. MSNP1 represents MSNP1AS subcloned into pIRES2-AcGFP1 in the opposite orientation. The pIRES2-AcGFP1 constructs containing MSN, MSNP1AS, and MSNP1 were transfected (2 μg) with the Amaxa Nucleofector kit V, according to the manufacturer’s protocol, into HEK and SK-N-SH cells. At 24 hours after transfection (n = 4) and 72 hours after transfection (n = 6), cells were harvested and subjected to Western blot analysis, as described above.

Statistical analysis

Quantitative PCR. Triplicate Ct values for each sample were averaged to a single Ct value. ΔCt was calculated by subtracting the Ct value for GAPDH for each sample from the Ct values for MSN, MSNP1AS, CDH9, and CDH10. ΔCt values were analyzed by Student’s unpaired t test for qPCR analyses of ASD compared to control for postmortem temporal cortex samples (n = 4). ΔCt values were analyzed by analysis of variance (ANOVA) followed by Fisher’s protected least significant difference (PLSD) test for genotype-expression comparisons. Data are expressed as relative expression compared to GAPDH.

Western blot. Total density of each moesin and GAPDH protein band was determined with VisionWorks LS software. For each sample, the ratio of moesin to GAPDH total density was calculated. Moesin-to-GAPDH ratios were analyzed by Student’s unpaired t test for comparison of ASD to control postmortem brain samples (n = 4).

Correlation analysis. Correlation statistics were calculated with the correlation Z test of the StatView statistics package (SAS Institute; version 5.0).

Transfection analysis. Moesin-to-GAPDH ratios were determined for each transfection condition. Moesin-to-GAPDH ratios were analyzed by ANOVA followed by Scheffe post hoc tests to determine the means with statistically significant differences. Each set of transfections represents a separate experiment. n = 4 for 24 hours after transfection; n = 6 for 72 hours after transfection.

Supplementary Materials

www.sciencetranslationalmedicine.org/cgi/content/full/4/128/128ra40/DC1

Fig. S1. Map of custom TaqMan gene expression qPCR assays for MSNP1AS and MSN.

Fig. S2. Correlation of MSNP1AS expression with ASD-associated genotypes after stratification by ASD diagnosis.

Fig. S3. Expression of moesin protein in HEK, SK-N-SH, and SH-SY5Y cell lines.

Table S1. Description of postmortem brain samples (superior temporal gyrus, Brodmann area 22) obtained from the Autism Tissue Program.

References and Notes

  1. Acknowledgments: We thank the Autism Tissue Program (Autism Speaks) and participating families for their contributions. Funding: This work was supported entirely by start-up funds from the Zilkha Neurogenetic Institute (D.B.C.). Partial salary support for G.A.C. was provided by R01CA136924. Author contributions: All authors contributed to the experimental design and interpretation of the data. D.B.C. wrote the manuscript, and all authors contributed to editing. Competing interests: The authors declare that they have no competing interests.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article