MicroRNAs (miRNAs) are an abundant class of short, non-coding, endogenous RNAs that regulate gene expression at the post-transcriptional level [1, 2]. The mature miRNA is a ~20-23 nucleotides long single stranded sequence, which derives from primary transcript miRNA (pri-miRNAs) [3]. One miRNA can bind to hundreds of target genes at the 3’-UTR of mRNAs (miRNA target genes), and a single miRNA target gene can be targeted by multiple miRNAs [4, 5]. Based on bioinformatic predictions, up to 90% of human genes are believed to be regulated by miRNA [5]. MiRNAs have primarily been demonstrated to mediate translational repression or target gene degradation and silencing. Recently, it has been shown that they can also upregulate gene expression by targeting gene regulatory sequences [6].

Genetic polymorphisms in miRNA or their targets can add to the complexity of miRNA regulation and function. Single nucleotide polymorphisms (SNPs) in miRNA binding sequences have indeed been shown to affect miRNA-mediated gene regulation and alter the expression of target genes [7]. A low SNP density at miRNA genes exists compared to the reference human genome, suggesting a negative selection for miRNA with SNPs [8–10]. CNVs, as a major class of genomic variations, also have an effect on miRNA, as demonstrated by under-representation of miRNA in highly polymorphic CNVs compared to the reference genome [10]. In contrast, the number of miRNA target genes in polymorphic CNVs is higher than in non-CNV regions, suggesting that genes integral to polymorphic CNVs are more likely to be regulated by miRNAs, in order to counteract their expression changes due to copy number variability of the region in which they reside [9]. Multiple cancer studies show that miRNAs integral to CNVs demonstrate gain or loss at the genomic level, and are associated with expression changes for ~10% -20% miRNAs [11, 12].

In addition to cancer, miRNAs are involved in other human diseases, for example, cardiovascular diseases and neurological/neurodevelopmental disorders [13, 14]. Evidence for the role of miRNA in these diseases is based on identifying mutations or differential expression of specific miRNAs and/or their global expression [15–17], and the effect of genomic copy number change on miRNA function is largely unknown. In rare instances association of CNVs with miRNA expression was studied in subject with cognitive delay. For example in Down syndrome an extra copy of chromosome 21 was associated with upregulation of several miRNAs from this chromosome, while their targets are downregulated [18]. In addition, a deletion of 1p21.3 containing a miRNA MIR137 has recently been reported by Willemsen et al. [19], and resulted in downregulation of the miRNA, and upregulation of three targets in subjects with ID and congenital abnormalities. Considering the role of MIR137 and its targets in brain function (synapse maturation, morphogenesis of young neurons, and axon growth), their dysfunction is considered causative of the patients’ phenotype [19]. Finally, microdeletion of a miRNA cluster MIR17HG on chromosome 13 is associated with Feingold syndrome which includes microcephaly, skeletal abnormalities and variable levels of ID [20]. The role of this miRNA in the human phenotype is further supported by a mouse knock-out model [20].

Although the above examples demonstrate the relevance of copy number change on miRNA function and its role in human disease, these studies are focusing on one type of CNV exclusively; for example, on common CNVs from normal populations [8–10], lump of CNVs from single disease cohorts such as cancer [11, 12] or autism [21], or a few individual pathogenic CNVs associated with neurodevelopmental disorders [19, 20]. Considering that ~ 70% of miRNAs are expressed in brain [22], and function in neurodevelopment, neurotransmission, synaptic plasticity, neurite outgrowth and dysregulation [18, 23], we aimed to characterize and understand the total presence and functions of miRNAs in CNVs detected in idiopathic ID in comparison to neurotypical controls. Five different CNV subgroups were studied for their miRNA and miRNA target gene contents: (i) de novo, (ii) familial, (iii) common CNVs detected in subjects with idiopathic ID, (iv) common CNVs in cognitively normal subjects, and (v) CNVs associated with syndromic ID selected from DECIPHER (Database of Chromosomal Imbalance and Phenotype in Humans using Ensembl Resources) database (http://decipher.sanger.ac.uk/). CNVs from DECIPHER were considered as positive controls. Our study is unique in serving as the first report comparing the miRNA content and function in different CNV subgroups from ID subjects and cognitively typical controls.

MiRNA-mediated gene regulation is complex and genomic variations, such as CNVs and SNPs, which can modulate miRNA expression, add to this complexity. The miRNA-CNV relationship has been rarely studied and predominantly involved polymorphic CNVs found in control cohorts [9, 10].

Our study is novel in its comparison of miRNA content in different classes of CNVs detected in a cohort of subjects with ID relative to cognitively typical subjects. We found a significant increase in the number of miRNAs in de novo and DECIPHER CNVs versus common CNVs (weighted median 0.6 and 0.8 versus 0.0, P < 0.05). In addition, the miRNAs in de novo CNVs were more likely to have expression in brain-related tissues or cell lines compared to the miRNAs from common CNV groups. Our collective findings suggest that miRNAs from de novo and putatively pathogenic CNVs could contribute to ID etiopathogenesis, in addition to coding genes integral to CNVs.

Similar to the increase in the number of miRNAs in de novo and DECIPHER CNVs, the number of miRNA target genes integral to these CNV types was also higher in comparison to familial or common CNVs. Lower numbers of miRNAs and miRNA target genes in common CNVs compared to the de novo or pathogenic CNVs suggests that they participate in processes that can tolerate functional variation. Previous studies have shown that targets for miRNAs from polymorphic CNVs tend to participate in environment-orientated processes including stimulus responses and immune responses, while the targets from non-CNV regions of the genome are enriched for fundamental biological processes such as maintenance of chromatin, chromosome segregation and nucleic acid processes [26]. In our study, genes from common CNVs that are targets for miRNAs do not show enrichment in any pathway while those from familial CNVs show enrichment in ubiquitin mediated proteolysis. In contrast, brain function related pathways such as axon guidance, Huntington’s disease, neutrophin signaling pathway, are found to be among the top 10 pathways enriched for targets from de novo and pathogenic CNVs. Targets from these two CNV groups also showed enrichment in the MAPK (mitogen-activated protein kinase) signaling pathway. The MAPK signaling cascade is involved in a wide variety of cellular processes and has been recently reported to be involved in ID pathogenesis [27–29]. Seven genes from de novo and DECIPHER CNVs which are targets for miRNAs were found to be involved in the MAPK pathway (MAPK9, MAPT, MEF2C, MKNK2, PAK2, RPS6KA1, and TAOK2). Three of them are known to be related to ID (MAPT[30], MEF2C[31, 32], PAK2[33]). Copy number changes of these targets could affect their regulation by miRNA. Target genes for CNV miRNAs detected in subjects with autism were also found to be enriched in MAPK pathway [21].

For familial CNVs we noted that although the number of miRNAs is much lower than in de novo or DECIPHER pathogenic CNV groups, and not different than in common CNVs, the number of miRNA target genes was significantly higher than in common CNVs. This might suggest potential phenotypic impact of these familial CNVs. In our cohort we had two familial CNVs, predisposing to ID: 16p11.2 and 1q21.1. These CNVs are well-known for their heterogeneous phenotypes and familial or de novo occurrence. The 1q21.1 CNV contains a single miRNA of unknown role (miR-5087), while the 16p11.2 paternal duplication covers 2 miRNAs: miR-3680-3p and miR-3680-5p and the de novo 16p11.2 deletion has no miRNA content (Additional file 1: Table S2). Consequences of the variability in copy number and function of miRNA integral to these CNVs is yet unknown. Understanding the role of miRNAs and miRNA targets in familial CNVs is of interest since this type of CNV represents a significant and clinically relevant interpretational challenge that could be guided by their miRNA features.

The ultimate proof that miRNAs influence the pathogenicity of genomic changes will come from an empirical confirmation of their copy number and expression change, and influence on the expression of their targets. The limitation of our own and other studies is due to their dependence on accurate prediction of miRNA targets and the fact that miRNA numbers increase dramatically between different versions of miRBase database. Therefore, comparison between different studies in terms of miRNA content in CNVs is challenging. Furthermore, some miRNAs may still represent false positive discoveries [10]. Unfortunately, we do not have miRNA expression data from our CNVs as RNA was not available. However, a recent study by Garcia-Orti et al. [34], demonstrated 10% of 259 studied miRNAs from regions of gain and loss detected in AML had significant change in expression concordant with the type of copy number change.

A more recently published study highlighted the significance of specific CNV-miRNAs and their targets in autism [21]. It assessed the content and function of 378 autism-associated CNVs, and detected 71 miRNAs. Five miRNAS were previously reported in ASD and 3 were known to have neuronal function. In our study, among the 84 miRNAs in the de novo CNVs, 3 were found to be associated with neurodegenerative disorders (miRNA-7, miRNA-9, miRNA-106b) [35–37] and 1 with ID (miRNA-9) [37].

Our analysis of miRNAs and miRNA targets related to CNVs is a first attempt to evaluate their role in a patient cohort manifesting idiopathic intellectual disabilities. Additional studies of cohorts of subjects with ID would benefit further evaluation of the apparent increase in miRNA content in de novo CNVs demonstrated in this study. In clinical practice, the interpretation of miRNAs that occur in patient CNVs is frequently challenging, particularly if they are the only genes that are included in the CNV. With global investigations of miRNAs in subjects with ID, followed by their expression analysis, our understanding of the role of miRNA in ID pathogenesis will be further improved.

Our findings support a possible role of miRNA copy number change in cognition and/or CNV-mediated developmental delay based on increased number of miRNA and miRNA target genes in de novo versus common CNVs in subjects with ID as well as their expression profile and participation in pathways. Systematic analysis of expression/function of miRNAs in addition to coding genes integral to CNVs could uncover new causes of ID.