Autism spectrum disorder (ASD), also known as pervasive developmental disorder (PDD), is defined as severe and pervasive impairments in the development of reciprocal social interaction and verbal and nonverbal communication skills. These disorders are also characterized by stereotypical behavior, interests and activities. The lifetime morbidity rate of ASD is 0.2% to 1.0% across studies [1]. In addition, twin and family studies of ASD have demonstrated a high heritability of approximately 90% [2], indicating that ASD is a heterogeneous condition that is likely to result from the combined effects of multiple genetic factors interacting with environmental factors. Recent genetic studies have identified several vulnerability loci and genetic mutations that cause ASD. One of the most striking revelations is the important role of genes that encode proteins at the neuronal synapse [3].

Rare mutations in the neuroligin 3 (NLGN3) and neuroligin 4 (NLGN4) genes, which map to chromosomes Xq13 and Xp22.3, have been reported in some patients with ASD and other neurodevelopmental impairments [4–8]. A particular mutation of NLGN3 (Arg451Cys) is known to cause a defect in protein processing of NLGN3 [9]. In addition, a particular mutation of NLGN4 (1186insT) causes a frameshift mutation that leads to premature termination of NLGN4 (D396X), resulting in a loss of 421 amino acids (51% of the protein) [4]. Neuroligins, which are postsynaptically localized cell adhesion molecules, play a crucial role in organizing excitatory glutamatergic and inhibitory GABAergic synapses in the mammalian brain by interacting with presynaptic β-neurexins (NRXN), thereby triggering the formation of functional presynaptic structures in contacting axons [10]. Mutations of the neurexin 1 (NRXN1) gene, at the chromosome locus 2q32, have been found in individuals with ASD [11–14]. Furthermore, de novo copy number variation analysis revealed deletion of the NRXN1-containing gene region in ASD [15]. The binding of NRXN1 and NLGN genes mediates synaptic development [16]. Interestingly, a mutation of NLGN3 results in a disruption of the ability to bind to NRXN [9]. In addition, neuroligins interact with a postsynaptic scaffolding protein, SHANK3, which is also implicated in ASD [17] and is located on the telomeric terminal of chromosome 22q13.3. Shank proteins couple neurotransmitter receptors, ion channels and other membrane proteins to the actin cytoskeleton and G protein-coupled signaling pathways, and they also play a role in synapse formation and dendritic spine maturation [18]. Deletion or translocation of the genomic locus, which includes the SHANK3 gene, and de novo mutations of the SHANK3 gene result in premature stop codons and have been found in ASD [17, 19, 20].

In a study of consanguineous autism families, Morrow et al. [21] observed a relationship between ASD and alterations in the sodium/hydrogen exchanger 9 (NHE9) gene. Specifically, they found a nonsense mutation in patients with ASD that is a heterozygous CGA-to-TGA transition, changing arginine 423 to a stop codon [21]. The NHE9 gene is located on chromosome 3q24 and is one of the families of Na⁺/H⁺ exchangers that regulate ion flux across membranes [22]. Rett syndrome is another PDD, and the methyl-CpG-binding protein 2 (MeCP2) gene is a causal gene for Rett syndrome. MeCP2 is a transcriptional repressor that binds to methylated CpG dinucleotides generally located at gene promoters and recruits histone deacetylase 1 and other proteins involved in chromatin repression [23]. De novo mutations of the MeCP2 gene located on chromosome Xp28 occur in 80% of female patients with Rett syndrome [24]. Some evidence of dysregulation of the phosphatidylinositol 3-kinase (PI3K)/AKT pathway is implicated in ASD, despite the fact that no mutation which causes ASD has been reported in association with the AKT1 gene. The expression and phosphorylation and/or activation of AKT were found to be decreased in the autistic brain [25]. The PTEN gene (phosphatase and tensin homolog deleted on chromosome 10) is a major negative regulator of the PI3K/AKT pathway, and PTEN mutations have been linked to ASD [26].

Recently, several studies have suggested that lymphoblastoid cells can be used to detect biologically plausible correlations between candidate genes and neuropsychiatric diseases, including Rett syndrome [27], nonspecific X-linked mental retardation [28], bipolar disorder [29], fragile X syndrome [30, 31] and dup(15q) [32]. In the present study, we compared mRNA expression levels of various genes in blood-derived lymphoblastoid cells from individuals with ASD and healthy controls.

In this study, we found that the mRNA expression levels of NLGN3 and SHANK3 were significantly lower in individuals with ASD than in healthy controls. Mutations of causal genes are rare, and they have been found to be associated with specific types of ASD. Our findings suggest that not only rare mutations of the causal genes but also functional alterations in the transcriptional activity of these genes might be associated with the pathophysiology of ASD. The NLGN3 and SHANK3 genes are synapse-related genes and were found to be affected in ASD, whereas other genes, including NHE9 and MeCP2, do not play major roles at the synapse and were not found to be affected in ASD. These findings suggest that impairments in synaptic function might be associated with the pathophysiology of ASD.

Reduced expression of the NLGN3 and SHANK3 genes in lymphoblasts of individuals with ASD is consistent with previous reports indicating that mutations of these genes cause reduced expression or loss of function of the protein. Since the NLGN3 gene is located in chromosome X, there may be expressional difference between genders. However, no significant difference of NLGN3 gene expression normalized by β-actin or TBP was observed with regard to gender in healthy controls or individuals with ASD (P > 0.05). This might be due to inactivation of one X chromosome in females [41]. There are several possibilities that might explain the reduced expression of the NLGN3 and SHANK3 genes in ASD. First, our sporadic ASD cases might have mutations, polymorphisms or copy number variations in the NLGN3 or SHANK3 genes, which could result in reduced expression of these genes. Second, mutations or polymorphisms in genes that regulate the expression of NLGN3 or SHANK3 might contribute to the observed reduction in expression of the NLGN3 or SHANK3 genes. To our knowledge, although the regulation of NLGN3 by other genes has not been reported, there are some reports in the literature describing the regulation of SHANK3 gene expression. For example, SHANK3 expression is regulated by DNA methylation [42, 43]. In addition, SHANK3 is one of the predicted targets of dysregulated microRNA (miRNA), and altered miRNA expression levels were found in postmortem brain from autism patients [44]. Further epigenetic analyses might elucidate the mechanisms of reduced SHANK3 expression.

Some findings of gene expression in lymphoblastoid cell lines are in conflict with those of previous studies. For example, Beri et al. [42] reported that SHANK3 is not expressed in EBV-transformed human lymphoblastoid cell lines in an investigation of tissue-specific SHANK3 gene expression and DNA methylation. By using lymphoblastoid cells from autism patients, Talebizadeh et al. [8] detected novel splice isoforms of NLGN4. There are methodological differences between previous studies and our study. SHANK3 gene expression in the previous study [42] was analyzed by using a conventional RT-PCR method; however, we measured the expression levels of SHANK3 gene by using a real-time qRT-PCR method (the TaqMan method). Furthermore, the expression level of SHANK3 was relatively low, which is shown in the standard curve in Figure 1. It is possible that the sensitivity of our real-time qRT-PCR method to detect the SHANK3 gene expression level might be higher than that of a conventional RT-PCR method. On the other hand, we could not quantitatively measure the NLGN4 and NRXN1 genes by using the real-time qRT-PCR method. However, there were slight expressions of these genes in lymphoblastoid cell lines when we used a large quantity of cDNA for the real-time qRT-PCR (Figure 1). Unfortunately, the small expression levels of these genes made it impossible to quantitatively measure the gene expressions in our sample. This may explain possible discrepancies of the gene expression findings of previous studies and our results.

There are several limitations of this study. First, our positive results might have arisen from sample bias due to non-age-matched samples, although the Japanese are a relatively homogeneous population, so the use of non-age-matched samples is unlikely to explain our findings. Second, our sample size might not be small for type I errors but small for type II errors. There is a possibility of type II errors in mRNA expression differences of NHE9, MECP2 and AKT1 between individuals with ASD and healthy controls and expression differences of NLGN3 and SHANK3 between individuals with schizophrenia and healthy controls. In particular, NHE9 might be increased in individuals with ASD, as the expression level of NHE9 was marginally significant before correction for multiple testing. Thus, replication studies using a larger sample size are needed before a firm conclusion can be drawn. Third, we did not perform a mutation search for the examined genes in our sample to replicate the association between the examined genes and ASD and how the causal or risk variants of the genes regulate the gene expression. As the previous evidence for candidate genes of ASD are based on rare mutations and/or copy number variations of the genes, it might be difficult to find a mutation in our 35 individuals with ASD for analysis of the variant effects on the gene expression in this study. A mutation search study of these candidate genes should be done in future studies. Fourth, the IQ scores in the ASD group were lower than those in the healthy control group, so reduced gene expression could be related to lower IQ. However, lower expression of the NLGN3 or SHANK3 genes was not found in individuals with schizophrenia who had lower premorbid IQ scores, and no expression difference was observed in individuals with ASD and mental retardation versus individuals with ASD but without mental retardation (data not shown). Taken together, the reduced gene expression in ASD might be specific to ASD, although other neuropsychiatric diseases, such as attention-deficit/hyperactivity disorder, mental retardation, major depression and bipolar disorder, should be examined in future studies. The ASD cases in this study were consistent with idiopathic autism diagnosed on the basis of clinical features. We did not include individuals with Rett syndrome and the other syndromic autisms, such as multiple sclerosis, which could explain why we did not find altered expression of MeCP2 in this cohort. Our results suggest that the MeCP2 gene may not be associated with the common pathology of ASD, while NLGN3 and SHANK3 may be. Because lymphoblastoid cell lines are not neuronal cells, some of our findings might not reflect the pathophysiology in ASD brains. Further studies investigating these limitations are warranted.

Our study reveals reduced levels of NLGN3 and SHANK3 mRNA expression in lymphoblastoid cell lines derived from individuals with ASD, but not from those of individuals with schizophrenia. These results are consistent with findings that rare mutations of these genes in specific cases cause loss of function, suggesting that reduction of NLGN3 and SHANK3 mRNA expression could be related to the pathophysiology of ASD in a substantial population of patients. Although there are several limitations present in this study, lymphoblastoid cell lines may still allow investigation of the pathophysiology of ASD. Further analyses are required, such as a mutation analysis of the NLGN3 and SHANK3 genes and the genes regulating their expression, in addition to studies designed to elucidate the mechanisms of this reduced expression.