The genetic basis of non-syndromic intellectual disability: a review

Definition and prevalence of intellectual disability

Intellectual disability (ID) is a common neurodevelopmental disorder that is characterized by an intelligence quotient (IQ) of 70 or below, and deficits in at least two behaviors related to adaptive functioning diagnosed by 18 years of age (American Psychiatric Association 2000). The prevalence of ID is between 1% and 3% (Roeleveld et al. 1997; Leonard and Wen 2002) and is present in every social class and culture (Leonard and Wen 2002). Despite its universal occurrence, there tends to be higher prevalence of ID in areas of lower socioeconomic status and developing countries, particularly for mild cases (Drews et al. 1995; Roeleveld et al. 1997; Durkin et al. 1998; Durkin 2002; Emerson 2007). It has been suggested that this discrepancy is likely due to environmental factors (Roeleveld et al. 1997; Durkin et al. 1998; Emerson 2007).

Approximately 30% more males are diagnosed with ID than females (American Psychiatric Association 2000; McLaren and Bryson 1987). However, despite a higher ratio of males to females among milder cases of ID, the ratio decreases as IQ decreases (American Psychiatric Association 2000; McLaren and Bryson 1987). Some studies suggest that severe ID may be more prevalent among females (Katusic et al. 1996; Bradley et al. 2002), however these studies were performed in quite specific communities, and may not necessarily be generalizable to other regions.

Classification of ID by IQ and syndromic vs. non-syndromic

ID is divided into 5 categories based on IQ: Mild, moderate, severe, profound and unable to classify (DSM IV). However, epidemiological studies often use a simplified classification, grouping their subjects into mild ID (IQ50-70) and severe ID (IQ<50) (Ropers and Hamel 2005). While the prevalence of severe ID is relatively stable, the prevalence of mild ID is variable and often depends heavily on external environmental factors such as level of maternal education, access to education/opportunity and access to healthcare (Leonard and Wen 2002; Drews et al. 1995; Roeleveld et al. 1997). Study design, age of subjects, and the catchment population for the studies may also contribute to the variability seen across mild ID prevalence studies (Leonard and Wen 2002; Drews et al. 1995; Roeleveld et al. 1997).

In addition to categorization by severity/IQ level, ID can also be grouped into syndromic intellectual disability (S-ID) and non-syndromic intellectual disability (NS-ID). In S-ID, patients present with one or multiple clinical features or co-morbidities in addition to ID. While S-ID has a clear definition, there is debate over the classification of NS-ID. Traditionally, NS-ID has been defined by the presence of intellectual disability as the sole clinical feature. However, it has been a challenge to rule out the presence of more subtle neurological anomalies and psychiatric disorders in these patients, as they may be less apparent, or difficult to diagnose due to the cognitive impairment. Additionally, symptoms of some syndromes may be so subtle that they are extremely difficult to diagnose unless the features are looked for specifically in the context of a known genetic defect previously associated with these features (Ropers 2006). Thus the distinction between S-ID and NS-ID is often blurred.

Causes of ID

ID can be caused by environmental and/or genetic factors. However, for up to 60% of cases, there is no identifiable cause (Rauch et al. 2006). Environmental exposure to certain teratogens, viruses or radiation can cause ID, as can severe head trauma or injury causing lack of oxygen to the brain. While these factors explain some cases of NS-ID, it is also important to consider genetic etiology.

Genetic causes of ID are thought to be present in 25–50% of cases, although this number increases proportionally with severity (McLaren and Bryson 1987). Chromosomal abnormalities have been reported in ID, with a broad range of prevalence, and many different types of aberrations have been identified (Rauch et al. 2006). Autosomal trisomies that are compatible with human viability and aneuploidies of the X-chromosome almost always result in some degree of ID as part of a syndrome, as illustrated by trisomy 21, or Down Syndrome—the most common genetic form of ID (Rauch et al. 2006). Additionally, pathogenic copy number variants (CNV) have been found to be associated with ID in a large number of studies, and will likely contribute to the discovery of many ID causing genes in the future (Ropers and Hamel 2005; Zahir and Friedman 2007).

Over the past 15 years many single gene causes of NS-ID have been identified. Many of these NS-ID genes may also cause S-ID, autism or other neurodevelopmental phenotypes, making it likely that other genetic modifiers or environmental factors may be involved in disease etiology. It also stresses the importance of detailed genotype/phenotype comparisons, which are often difficult to elucidate. There is also a possibility that some instances of NS-ID are multifactorial, with more than one gene contributing to disease in an individual, however this has not been well studied. Most known NS-ID genes are on the X-chromosome, however the number of autosomal genes associated with NS-ID is growing rapidly (Chelly et al. 2006).

Relevance of studying NS-ID genetics

Lionel Penrose was quoted as wishing to see each ID individual “as an integral part of the human race in its struggle for evolution and survival, unwittingly yielding information of the greatest value in the progressive understanding of the biological structure of the whole group” (from Berg 1998). As Penrose implied, the identification of the biological causes of NS-ID is necessary to our understanding of human cognition and intellect. Because NS-ID presents with intellectual impairment as the only feature, genes that cause it are likely related to the processes of learning and memory. These processes are fundamental to our understanding of the formation of normal intellectual capabilities, and in particular how intellect develops from a neurological perspective. Additionally, finding genes that cause NS-ID might help us to decipher relevant pathways that are involved in neurological development. Understanding these pathways may aid us in treating or relieving symptoms of NS-ID in certain cases. Knowledge of pathways involved in NS-ID will also make it easier to select candidate genes to analyze in research and clinically based studies. Understanding the genetics of a complex disease like NS-ID is also relevant to genetic counseling in families with affected individuals, particularly where consanguinity is involved (Modell and Darr 2002).

Homozygosity mapping

Given the innate heterogeneity of the NS-ID phenotypes, it is of little surprise that its genetics are equally as complex. It is thought that rare variants are likely responsible for most cases of NS-ID, and this has been the focus of NS-ID genetics for some time, although association studies are still used to identify polymorphisms that potentially contribute to the phenotype (e.g. SNAP25; Gosso et al. 2008). It is likely that many rare variants across many genes result in the same phenotype. Under the rare variants model of causation, large consanguineous families are particularly useful. Rare mutations, which may be identified in such families, can provide us with information about the types of etiological aberrations in NS-ID as well as the genes and relevant pathways that may be essential for normal neuronal functioning.

Our increasing ability to perform high throughput analyses of genotype using microarrays has been a significant contributor to the increased discovery rate of NS-ID genes over the past 10 years (Lugtenberg et al. 2007). All of the genes identified for autosomal recessive NS-ID thus far have been identified through microarray technology combined with homozygosity mapping using large consanguineous families. In these studies, large multiplex families are obtained and detailed family histories are taken. Affected family members are typically assessed for additional clinical phenotypes that may suggest S-ID, and are screened for more commonly known causes such as fragile-X mutations and gross chromosomal anomalies. DNA from the blood of affected family members, along with one or two unaffected family members, is analyzed using microarray technology and analysis software.

This method allows for a fast screen of individuals genotypes, and allows the researcher to identify regions of the genome that are homozygous for the same alleles among all affected individuals. Typically (or at least ideally), a single long stretch of DNA (often over 3 Mb), is identified, and the disease causing mutation would lie within this region. Use of consanguineous multiplex families allows the identification of these linkage regions with relative ease. The regions are then confirmed by microsatellite analysis, and sequencing of genes within the region will often lead to the identification of a disease gene.

Consanguinity is marriage between closely related individuals, and is common in many countries around the world, particularly in the Middle East and Asia. In Pakistan—a country of 173 million people—consanguineous unions make up 62.7% of marriages, and ~80% of which are between first cousins, according to a national census (Hussain and Bittles 1998). Other countries that also have high prevalence of consanguinity include Iran, where the overall prevalence is 40%, and India where prevalence ranges from 16% to 33% depending on the region (Hussain and Bittles 2000; Najmabadi et al. 2007).

Multiply affected consanguineous families have been integral in determining autosomal recessive causes of disease. The children of consanguineous individuals will have more homozygous DNA than the offspring of an outbred marriage. This leads to an increased likelihood of rare, recessive disease-causing variants being inherited from both parents. This is known as autozygosity or homozygosity-by-descent (HBD), which occurs when a rare allelic variant is passed down to offspring from a common ancestor via both maternal and paternal lineages. In populations where consanguinity is prevalent, there is a significant increase in infant morbidity and mortality (Modell and Darr 2002; Gustavson 2005; Khlat and Khoury 1991). One study showed that, in consanguineous populations, there were 10 times more recessively inherited congenital conditions, many of which resulted in early death (reported from the Birmingham Birth Study, as reviewed in Modell and Darr 2002).

Advantages and disadvantages of homozygosity mapping

There are several reasons why homozygosity mapping has been a successful approach for identifying rare genetic variants. Intuitively, the co-segregation of large stretches of homozygous alleles only in affected family members in a consanguineous family increases the likelihood that the region contains a gene that is relevant to the phenotype, and statistics in the literature support this concept (Lander and Botstein 1987). The method has been successfully utilized to identify recessive causes of many diseases such as Joubert syndrome, Charcot-Marie-Tooth syndrome, syndromic deafness, and oligodontia (Noor et al. 2008; Bolino et al. 2000; Senderek et al. 2003; Verpy et al. 2001; Noor et al. 2009).

A major drawback is the scarcity of families in outbred western populations that are suitable for this method. Also, a lack of appropriate clinical assessment tools and infrastructure may pose problems in some of the countries where consanguineous families are more common. Aside from the lack of large pedigrees, there are further obstacles in identifying the gene of interest. Homozygosity mapping studies, just like linkage studies, often identify very large regions in which the disease gene may be found. Some of these regions contain hundreds of genes, and selecting relevant candidate genes can be problematic. Often, there is no clear candidate gene, or there are many genes in the region with no known function and thus selection of suitable candidate genes is tricky. Using Next Generation Sequencing (NGS) might be an effective way to overcome this, however innate variation within the genome will likely result in the identification of multiple genetic aberrations in several genes that need to be ruled out as disease genes. For instance, in one of the few reports so far on identification of a recessive gene (DHODH, for Miller syndrome) through exome sequencing, many variants were identified, however, after filtering for known SNPs, just 9 candidate genes remained (Ng et al. 2010). For application in consanguineous families, this approach should be simpler, as one would expect that the exome sequence data would also be able to indicate regions of the genome containing large stretches of homozygosity, so the candidate genes could be narrowed down further to such a region.

Sequencing candidate genes

The sequencing of candidate genes has previously been employed in the identification of many XLMR genes. The X-chromosome has been a target in many studies looking for causes of NS-ID because of the high male to female ratio in the NS-ID population. The result is that most of the known NS-ID genes are X-linked. Often, large cohorts of individuals with ID are collected and target genes that have a suspected role in development or brain function are sequenced across the cohort. Recently, a re-sequencing of the X-chromosome exome in a large cohort has led to the identification of several novel NS-ID genes (Tarpey et al. 2009).

Typical genetic mapping strategies for autosomal dominant disorders have been unsuccessful for NS-ID, due to genetic heterogeneity and lack of suitable multiplex families, as procreation of affected individuals is unlikely. Thus, autosomal dominant NS-ID is likely to be sporadic, resulting from de novo mutations. Sequencing candidate genes may be the best approach for identifying autosomal dominant causes of NS-ID. SYNGAP1, STXBP1, and SHANK3 were all identified as autosomal dominant causes of NS-ID using candidate gene sequencing (Hamdan et al. 2009a, b; Michaud et al. 2009 Abstract).

While this approach can be effective, it may also require much work, and will frequently be unsuccessful if the level of genetic heterogeneity is as high as anticipated. However, as our knowledge of biological pathways involved in ID grows, our ability to select probable candidates will increase, and this strategy may become more plausible. Additionally, with improved technology such as NGS, techniques, sequencing entire exomes for causes of ID will become a less laborious and more productive screening method. NGS is an emerging method that produces large amounts of sequencing data quickly, at relatively low cost. With this technology, it is possible to sequence through large linkage regions, which can contain hundreds of genes, with much less labor and in shorter time than traditional sequencing. As well as application to critical linkage regions, NGS can be applied to HBD regions, or whole genome sequence analysis. It can also be applied in a more directed (and thus more economical) approach, by whole exome sequencing. This involves sequencing just the coding portion of the genome—approximately 1% of our genetic material. This method may be useful for identifying genes when no linkage or susceptibility region is known, as most disease causing mutations are likely to occur in exons.

Although NGS and exome sequencing present us with many opportunities for high-throughput analysis of whole genomes, some issues with the method have arisen. One major concern is that there are likely to be many false positives—there are thousands of common variants in the human genome, many of which have not been well documented. This is exemplified in the Tarpey et al. (Tarpey et al. 2009) re-sequencing of the X-chromosome exome. In this particular case, although standard rather than next-generation sequencing was used, Tarpey et al. (Tarpey et al. 2009) screened virtually the entire X-chromosome exome in a large cohort for genes causing ID. It was found that truncation of 1% or more of X-chromosome genes may still be compatible with normal phenotype (Tarpey et al. 2009). There are many examples of nonsense mutations that occur in the healthy population, where the resulting haploinsufficiency is presumably compensated for by other means (Tarpey et al. 2009). This demonstrates the need for caution when interpreting results of large quantities of sequencing data, such as that obtained by NGS and exome sequencing studies. For instance, in the whole genome sequencing of a patient with Charcot-Marie-Tooth neuropathy, ~9,000 synonymous and ~9,000 non-synonymous coding changes were identified, including 121 nonsense mutations (Lupski et al. 2010). (Also see comment in previous section on identification of the DHODH gene for Miller syndrome through exome sequencing; Ng et al. 2010).

Characterization of chromosomal aberrations

Characterization of chromosomal aberrations by breakpoint analysis has long been used as a method to identify autosomal dominant disease causing genes. Determining the exact location of the breakpoints and study of disrupted genes has led to the discovery of several candidate genes for non-syndromic autosomal dominant ID (NS-ADID). DOCK8, MBD5, CDH15 and KIRREL3 were all identified at chromosomal breakpoints of either gross chromosomal abnormalities or CNVs (Griggs et al. 2008; Bhalla et al. 2008; Wagenstaller et al. 2007). For MBD5, CDH15 and KIRREL3, further screening of these genes in cohorts of ID patients led to the identification of several mutations, indicating the utility of this method in identifying disease genes (Bhalla et al. 2008; Wagenstaller et al. 2007). X-linked NS-ID genes have also been found using this method. TSPAN7 was identified at the breakpoint of a translocation in one individual, and then screening of an NS-ID cohort led to the identification of mutations in TSPAN7 in 2/33 families (Zemni et al. 2000).

Subtelomeric rearrangements and copy number variants

Copy number variants (CNV), which in the past few years have been found to contribute significantly to common genetic variation, also confer susceptibility to numerous diseases. Although CNVs cover approximately 12% of the entire genome in the normal population, rare CNVs have been associated with various neuropsychiatric disorders such as schizophrenia and autism (Redon et al. 2006; Marshall et al. 2008; Stefansson et al. 2008), and have also been implicated in ID (Zahir and Friedman 2007; Knight et al. 1999)

Submicroscopic subtelomeric rearrangements have been long implicated in the etiology of NS-ID. These rearrangements include deletions as well as balanced translocations and other chromosomal aberrations that are unable to be seen under the microscope. Subtelomeric regions are frequently a focus for analysis of chromosomal rearrangements, as the density of genes in this region is greater than in the rest of the genome (Saccone et al. 1992), and about half of all segmental aneusomies involve subtelomeric and terminal regions of chromosomes (Biesecker 2002). Subtelomeric rearrangements have been shown to be a significant cause of ID in many chromosomal studies and are thought to be responsible for 3–6% of cases (Knight et al. 1999; Ledbetter and Martin 2007). In a 2002 meta-analysis of studies of subtelomeric abnormalities in ID cohorts, it was confirmed that this type of genetic aberration is present in ~6% of cases, and that ~50% of them are inherited (Biesecker 2002). These finding were quite exciting because of the large number of cases with subtelomeric abnormalities, and indicated that analysis of subtelomeric regions could identify the molecular cause of ID in many idiopathic individuals, as well as leading to more effective clinical testing. With the advent of CGH microarrays, identification of submicroscopic subtelomeric rearrangements became more robust and more clinically feasible (Stankiewicz and Beaudet 2007).

Currently, deletions in subtelomeric regions are also picked up by CNV analysis—a type of analysis that has become essential for determining the etiology of NS-ID. With the increasing availability and sensitivity of microarray technology, pathogenic CNVs have been consistently detected in 10–15% of individuals with ID across many studies (Fan et al. 2007; Koolen et al. 2009; McMullan et al. 2009) (as reviewed by Zahir and Friedman 2007). These CNVs include both rare de novo, as well as rare inherited mutations, which are still of unknown significance (but may yet be important, since rare inherited subtelomeric rearrangements have also been seen frequently; McMullan et al. 2009).

As a demonstration of the successful application of CNVs for identifying novel NS-ID genes, very recently, SHANK2 was identified as an NS-ID and autism disease gene, through the identification of CNVs that deleted SHANK2 coding regions in two affected individuals. Subsequent screening by sequencing in large cohorts identified further NS-ID and autism patients with SHANK2 mutations (Berkel et al. 2010). Additionally, in the largest study of its kind to date, SHANK2, SYNGAP1 and ILRAPL1 CNVs were identified in autism probands, illustrating the importance of CNV analysis and the identification of genetic overlap between autism and NS-ID (Pinto et al. 2010). CNV analysis will clearly play an important role in the identification of further ID-related genes. A great number of pathogenic CNVs have been identified in NS-ID in recent years, and mapping these breakpoints will likely be a successful method in determining autosomal dominant as well as X-linked causes of ID, just as mapping breakpoints of cytogenetic abnormalities has in the past (Ropers and Hamel 2005).

Ionotropic glutamate receptors and excitatory synapses

Ionotropic glutamate receptors have long been suspected to be involved in the etiology of neuropsychiatric disease. Several examples of mutations in glutamate activated receptors and their downstream effectors are present in NS-ID (see Fig. 1). The MRT6 gene GRIK2 encodes GLuR6, which is a subunit of a Kainate receptor (KAR). Likewise, the MRD5 gene SYNGAP1 encodes SynGAP—a GTPase activating protein that is part of the NMDA receptor (NMDAR) complex, binding to the NR2B subunit (Kim et al. 2005). The NMDAR is a well-characterized ionotropic glutamate receptor. SynGAP is a negative regulator of NMDAR mediated ERK activation and causes inhibition of the Ras/ERK pathway (Kim et al. 2005). Over-expression of SynGAP has also been shown to down regulate GLuR1, a subunit of AMPA receptors (AMPAR), a class of ionotropic glutamate receptors which are regulated by the Ras/ERK pathway (Kim et al. 2005; Rumbaugh et al. 2006). Likewise, Syngap knockout mice implicate SynGAP in the regulation of LTP and AMPAR expression (Komiyama et al. 2002).

SAP102 (MRX90), which is part of the membrane-associated guanylate kinase (MAGUK) protein family and the product of DLG3, is also part of the NMDAR complex (Tarpey et al. 2004). MAGUKs are scaffolding proteins involved in the clustering, targeting and anchoring of ionotropic glutamate receptors in the excitatory postsynaptic density (Gardoni 2008). SAP102 directly interacts with the NR2B and NR2A subunits of the NMDAR and is likely to have a role in the clustering and targeting of these receptors (Muller et al. 1996). The protein is expressed at excitatory synapses, particularly during early brain development (Sans et al. 2000). When knocked down, SAP102 has been found to decrease AMPAR and NMDAR excitatory postsynaptic currents (EPSC), while over expression increases EPSC (Elias et al. 2008). Another MAGUK family protein gene, CASK, is frequently mutated in NS-ID as well (Tarpey et al. 2009; Hackett et al. 2009). It is a synaptic scaffolding protein, and acts as an Mg²⁺-independent neurexin kinase, and interacts with many other families of proteins at cellular junctions (Hata et al. 1996; Mukherjee et al. 2008). It also directly interacts with GLuR6 (Coussen et al. 2002).

Mutations in IL1RAPL1, a gene with several known mutations in NS-ID and autism, result in the incorrect localization of the MAGUK family protein PSD-95 (DLG4), which is important for organization and function of NMDA receptors, ion channels and other signaling proteins (Carrie et al. 1999; Gardoni 2008; Kim and Sheng 2004; Pavlowsky et al. 2010). IL1RAPL1 has been shown to interact with PSD-95, and knockout of this gene decreased the post-synaptic density (PSD) and the localization of PSD-95 at excitatory synapses. Loss of IL1RAPL1 also results in a decrease of activity in the JNK pathway, which led to decreased phosphorylation of PSD-95 (Pavlowsky et al. 2010). It has also been shown to be important for the formation of excitatory synapses in vivo (Pavlowsky et al. 2010). PSD-95 directly interacts with several known NS-ID associated proteins including: CASK, SynGAP, GLuR6 and neuroligins (Kim and Sheng 2004).

Three other synaptic proteins, NLGN4, SHANK2 and SHANK3, have also been identified as causes of NS-ID and/or autism. NLGN4 is a protein that acts in neuronal cell adhesion. It is an important element in postsynaptic differentiation, forming complexes with β-neurexins and PSD-95 (Ichtchenko et al. 1995; Irie et al. 1997; Scheiffele et al. 2000). NLGN4 is linked to glutamatergic postsynaptic proteins and neuroligin/neurexin complexes appear to be sufficient for synaptogenesis (Graf et al. 2009). Interestingly, heterozygous translocations and CNVs disrupting neurexin 1 (NRXN1), which interacts with NLGN4 at the synapse, have been associated with autism (Kim et al. 2008; Autism Genome Project Consortium et al. 2007), and homozygous mutation of NRXN1 causes the S-ID disorder Pitt-Hopkins-like syndrome-2 (PTHSL2; Zweier et al. 2009). NRXN1 also interacts with the NS-ID causing scaffolding protein CASK, as discussed earlier in this review (Hata et al. 1996). SHANK2 and SHANK3 encode scaffolding proteins present at the post-synaptic density and in dendrites. They are important for scaffolding in the post-synaptic density—connecting ion channels, neurotransmitter receptors and other membrane proteins to the actin cytoskeleton—and act as a structural framework at this site (Boeckers et al. 2002; Hayashi et al. 2009). They are also likely to play a role in neuronal plasticity (Boeckers et al. 2002; Hayashi et al. 2009).

OPHN1 encodes an activity-dependant protein that interacts with AMPARs and is essential for their stabilization, thus playing a significant role in synaptic maturation and plasticity (Nadif Kasri et al. 2009). Interestingly, GRIA3, which causes S-ID is a subunit of AMPAR. The S-ID phenotype caused by GRIA3 is variable, but one case presented with only ID and aesthenic body habitus (Wu et al. 2007; Bonnet et al. 2009). Due to the small number of mutations found in this gene, it is difficult to discern if the ID in certain individuals is actually syndromic. The finding that ID can be caused by mutations in all three classes of ionotropic glutamate receptors suggests that the activity and regulation of glutamatergic synapses is essential for normal cognition. This is not surprising, as LTP and LTD often originate from these synapses, and are processes that are thought to be important for synaptic plasticity, memory and learning (Malenka and Bear 2004).

Other synaptic and neuronal proteins in NS-ID

Activity downstream of ionotropic glutamate receptors is also important in understanding the etiology of NS-ID. PRSS12 encodes the protein neurotrypsin, a synaptic protease that is activated by the NMDA receptor (Molinari et al. 2002; Matsumoto-Miyai et al. 2009). It acts in the synapse to cleave agrin, which is present at neuromuscular junctions (NMJ) and in the CNS, but appears to have distinct functions at each location (Stephan et al. 2008; Bezakova and Ruegg 2003). Neurotrypsin-dependent agrin cleavage appears to be an essential process for the formation of dendritic filopodia in the hippocampus, and is dependent on LTP generated by NMDA and AMPA receptors (Matsumoto-Miyai et al. 2009). Dendritic filopodia are the precursors for new dendritic spines in activity dependant synaptogensis (Jontes and Smith 2000). With so many NS-ID genes coding for proteins that are involved in organizing, signaling and downstream effects of ionotropic glutamate receptors, it is likely that excitatory synapses are very important in the normal development of intellectual function. CRBN is another NS-ID gene that codes for a synaptic protease (Higgins et al. 2004). Although its function is not as well characterized as PRSS12, it is known to contain a LON-Protease domain. It, however, is better known to exist in neurons and assist in coordinating the expression function of Ca²⁺-dependant K⁺ channels (Higgins et al. 2008).

Cell adhesion

Earlier in this review, NLGN4 was discussed as a neuronal cell adhesion molecule involved in NS-ID etiology. Cell adhesion molecules are critical for the maintenance of synaptic structure and neuronal plasticity (Sudhof 2008). Not surprisingly, several ID genes are involved in cell adhesion. CDH15 is a cadherin gene localized in the brain and skeletal muscle (Bhalla et al. 2008). Mutations of this gene in individuals with ID were found to decrease cell adhesion by greater than 80% (Bhalla et al. 2008). PCDH19, a protocadherin, has also been implicated in epilepsy with mental retardation limited to females (EFMR) (Dibbens et al. 2008; Hynes et al. 2009). Additionally, it has been postulated that TSPAN7 is involved in a complex of ß-integrins, which are involved in cell–cell and cell–matrix interactions (Zemni et al. 2000).

The RHO pathway

Rho GTPases are also a common pathway in NS-XLMR (see Tables 2 and 3). OPHN1, PAK3, ARHGEF6 and FGD1 all encode proteins that are involved in cellular signaling through Rho GTPases or downstream effects (Ramakers 2002). OPHN1 encodes oligophrenin 1, which has been found to stimulate the GTPase activity of RhoA, Rac1 and Cdc42 (Billuart et al. 1998). PAK3, a serine/threonine protease that plays a role in regulating the actin cytoskeleton, is induced by Rac and Cdc42 (Allen et al. 1998; Manser et al. 1995; Daniels and Bokoch 1999). ARHGEF6 encodes a guanine nucleotide exchange factor (GEF) for Rac1 and Cdc42 that interacts with PAK family proteins (Manser et al. 1995; Daniels and Bokoch 1999). FGD1 is also a GEF for Cdc42 (Zheng et al. 1996). All of these gene products interact in the Rho signaling pathway, suggesting that events triggered by this pathway, such as vesicular trafficking in dendritic spines and cytoskeleton organization, may be important processes in cognition and that, when disturbed, cause ID.

Synaptic vesicle trafficking and exocytosis

Several NS-ID genes are involved in neurotransmitter release by exocytosis (see Fig. 1). This process has been extensively studied over the years and has a number of key components. The SNARE complex is essential for neurotransmitter release. Reconstitution experiments have shown that SNARE alone is sufficient for membrane fusion (Weber et al. 1998). SNAP25 (MIM: 600322), a subunit of plasma membrane SNARE complexes (t-SNARE) has been associated with ID through SNP analysis (Gosso et al. 2008). Studies of common variants in SNAP25 have shown significant association with the extremes of IQ (Gosso et al. 2008). This is one of the few examples of common variants playing a role in intellectual disability.

STXBP1 (Munc18-1) (MIM: 602926) binds to syntaxin-1, the other subunit of the t-SNARE complex, and there is evidence that it has a variety of functions in exocytosis including vesicle priming, SNARE assembly, localization of syntaxin-1 in the plasma membrane, as well as regulatory effects (As reviewed by Rizo and Rosenmund 2008). Truncating mutations in STXBP1 have been reported in two cases of NS-ID, and in ID with non-syndromic epilepsy (Hamdan et al. 2009a, b).

Additionally, SYP, which encodes synaptophysin, an integral membrane protein found in transport vesicles in the brain, also causes NS-ID (Tarpey et al. 2009). It interacts with synaptobrevin (VAMP2) and is involved in the regulation, sorting and distribution of synaptobrevin in neurons, but the molecular mechanism by which this occurs is unknown (Bonanomi et al. 2007). Synaptobrevin is the essential component of the vesicle SNARE complex (v-SNARE) (Rizo and Rosenmund 2008). Four mutations have been found in SYP in individuals with NS-ID and ID with epilepsy (Tarpey et al. 2009).

The Rab3 family of small GTPases is also known to be important in membrane trafficking and the release of neurotransmitters. GDI1 encodes a GDP-dissociation inhibitor, which causes GTPases Rab3a and Rab3b to remain inactive by maintaining it in their GDP-bound form (D’adamo et al. 1998; Bienvenu et al. 1998). Rab3 GTPase isoforms A, B and C are localized to synaptic vesicles, and function to regulate neurotransmitter release by regulating the SNARE complex (Fischer von Mollard et al. 1994; Geppert et al. 1997; Gonzalez and Scheller 1999). Interestingly, the scaffolding protein CASK interacts with rabphilin3a, an upstream effector of Rab3a (Zhang et al. 2001). Rabphilin3a maintains Rab3a in its active, GTP-bound state, so disruptions in the rabphilin3a complex with CASK may have a deleterious impact on its interaction with Rab3a (Geppert et al. 1997). It is possible that this leads to the disruption of synaptic vesicle exocytosis, but requires further investigation. Another Rab family gene, Rab39, was found to be mutated in two families with ID and variable other phenotypes including autistic features, macrocephaly and epilepsy (Giannandrea et al. 2010). This suggests that Rab family proteins are important for normal cognitive development and other Rab genes may be considered as candidates for ID.

The ERK/MAPK pathway

The ERK/MAPK pathway is a signaling cascade that responds to growth factors. This pathway is particularly interesting because it is required for certain types of synaptic plasticity (Zhu et al. 2002; Thomas and Huganir 2004). Genes coding for ERK/MAPK pathway proteins and regulators have also been found to cause NS-ID. SynGAP, discussed earlier in this review, has been shown to negatively regulate the ERK/MAPK pathway. When SynGAP activity was depleted by RNAi, ERK activation by NMDAR became sustained (Kim et al. 2005). Conversely, over-expression of SynGAP in cultured neurons diminishes ERK activation (Rumbaugh et al. 2006). These results suggest that SynGAP is necessary for terminating ERK activation thus negatively regulates excitatory synaptic transmission and AMPAR cell surface expression (Kim et al. 2005; Rumbaugh et al. 2006). RPS6KA3 is another NS-ID gene that is involved in ERK/MAPK signaling (Merienne et al. 1999). RPS6KA3 encodes Ribosomal S6 Kinase (RSK) 2, which is a downstream effector of the ERK signaling pathway. ERK activation of RSK2 causes phosphorylation of SHANK3, and appears to be required for normal functioning of AMPARs (Thomas and Huganir 2004). These findings suggest that the ERK/MAPK pathway is an important pathway for normal cognitive development.

Zinc finger proteins

A number of zinc finger proteins, autosomal and X-linked, have been implicated in NS-ID. Both missense and nonsense mutations cause NS-ID in these genes. ZNF41, ZNF81 and ZNF674 are members of a cluster of 7 highly related zinc-finger protein genes on the X-chromosome, and are thought to be involved in transcriptional regulation (Lugtenberg et al. 2006). These zinc finger proteins are all part of the Kruppel-associated box (KRAB) family, which is the largest class of zinc finger proteins (Urrutia 2003). KRAB family zinc finger genes tend to occur in clusters in the genome. The largest such cluster is present on 19q13 (Rousseau-Merck et al. 2002), and contains more than half of the known KRAB domain zinc finger proteins. KRAB family zinc finger proteins have been shown to contribute to transcriptional repression by binding to co-repressors (Witzgall et al. 1994; Margolin et al. 1994) They have been found to function in the formation of hp-1 heterochromatin via complex formation with Kap-1 and SetDB1, a histone 3 lysine 9 (H3-K9) methylase (Schultz et al. 2002). During mouse embryogenesis, KRAB domain proteins also cause irreversible gene silencing through methylation of promoter sequences (Wiznerowicz et al. 2007). It is also notable that KRAB family zinc finger proteins are only present in higher vertebrates such as mammals, birds and amphibians, but not in fish or lower eukaryotes, and it has been postulated that they may have evolved to play an important role in functions such as the nervous or immune systems (Urrutia 2003).

The gene ZNF711 is also implicated in NS-ID and is located on the X-chromosome (Tarpey et al. 2009). Currently the function is unknown but its sequence and domain structure bears similarity to other zinc fingers involved in transcriptional activation (Tarpey et al. 2009). ZC3H14, another zinc finger gene putatively involved in NS-ID (Garshasbi et al. 2009 Abstract), contains a polyadenosine RNA binding domain, and co-localizes with the splicing factor SC35, suggesting a potential role for it in mRNA processing, but further investigation is required to support this claim (Leung et al. 2009).

Since these genes are involved in NS-ID, it is possible that their protein products target the regulation of specific neuronal genes that are involved in cognitive development, learning or memory formation, resulting in an NS-ID phenotype. Future studies that detail the genes or chromosomal regions that are influenced by zinc finger proteins will result in a clearer picture of the mechanism by which mutations in zinc finger proteins cause NS-ID. Numerous other NS-ID genes also appear to be involved in transcriptional regulation (Gecz et al. 2009).

Transcriptional regulation and chromatin remodeling

In addition to the zinc finger proteins, several other transcriptional regulators are implicated in NS-ID. Regulation of transcription appears to be a common theme among NS-ID genes. CC2D1A, an autosomal gene, is responsible for repression of transcription of several serotonin and dopamine receptor genes, and is a putative activator of the NF-κB transcription factor (Basel-Vanagaite et al. 2006). Likewise, TRAPPC9 is an activator of the NF-κB pathway (Hu et al. 2005). NF-κB up-regulates the expression of many neuronal genes, as well as genes of the immune system. Multiple studies have implicated NF-κB in long-term memory formation (Albensi and Mattson 2000; Kassed et al. 2002; Meffert et al. 2003). Additionally, evidence suggests that NF-κB is induced in the hippocampus by group I metabotropic glutamate receptors (GpI-mGLuRs; O’Riordan et al. 2006).

ARX is a homeobox-containing gene that is part of the Aristaless-related gene family. This is a family of transcription factors that are required for various essential events during vertebrate embryogenesis, including CNS development (Meijlink et al. 1999). Based on the experimental data and gene structure it has been speculated that ARX regulates transcription by both gene activation and suppression (as reviewed by Friocourt et al. 2006). Many mutations have been identified in ARX causing intellectual disability. Diseases caused by mutations in this gene include: West syndrome (Stromme et al. 2002a), Partington syndrome (Stromme et al. 2002b), XLAG (Kitamura et al. 2002), XLMR (Bienvenu et al. 2002), Proud Syndrome (Kato et al. 2004), and various forms of epilepsy (Stromme et al. 2002a, b; Scheffer et al. 2002). There seems to be a genotype/phenotype correlation between the location and nature of the mutation and the severity of the phenotype. Loss of function mutations such as truncations and mutations occurring in the highly conserved homeobox domain tend to result in more severe disease including brain malformations as is seen in Proud Syndrome and XLAG (Sherr 2003; Gecz et al. 2006). This gene is one of the most frequently mutated in XLMR. Strikingly, one mutation, a 24-bp in frame duplication that expands the 12-polyalanine tract to 20, is found in >5% of individuals with XLMR (Bienvenu et al. 2002; Stromme et al. 2002a, b; Friocourt et al. 2006). This mutation is also found in individuals with other ARX related S-ID, emphasizing the pleiotropy observed with ARX mutations (Gecz et al. 2006; Friocourt et al. 2006). The activity of this transcription factor is clearly essential for normal cognitive development, and further study of the genes regulated by ARX will be important for our understanding of NS-ID.

Epigenetic regulation is also an important mechanism by which transcription is regulated. MECP2 and ATRX, two genes that function in epigenetic regulation, have been linked to NS-ID etiology. MECP2 encodes the methyl CpG binding protein 2 (MeCP2), which is believed to act as a transcriptional modulator, capable of repressing or activating genes through long-range chromatin re-organization through binding to methylated CpG DNA (see Gonzales and LaSalle 2010 for review). MeCP2 also interacts with ATRX, and in Mecp2 null mice ATRX localization in neurons is disturbed (Nan et al. 2007). Mutations in ATRX cause changes to DNA methylation at several specific highly repeated sequences during development, and MeCP2 mutations cause a loss of transcriptional regulation through binding methylated CpG-containing DNA (Dragich et al. 2000; Gibbons et al. 2000). Interestingly, these are both genes in which mutations cause variable phenotypes. MECP2 mutations cause Rett syndrome, MRXS13, LUBS X-linked ID syndrome, autism and NS-ID, and ATRX mutations cause alpha-thalassemia/MR syndrome, MR-hypotonic facies syndrome, alpha-thalassemia myelodysplasia syndrome and NS-ID.

Chromatin remodeling, a mechanism that is essential for both gene expression and the maintenance of chromatin structure, also appears to play a role in NS-ID etiology. JARIDIC, also known as KMD5C, is one of the more common genes related to X-linked intellectual disability with over twenty mutations known in XLMR patients (Tzschach et al. 2006; Tahiliani et al. 2007). It is a histone demethylase containing a PHD-finger domain that is characteristic of zinc finger proteins and specifically demethylates di- and tri-methylated histone 3 lysine 4 residues (H3K4me2/me3) (Cloos et al. 2008; Iwase et al. 2007; Tahiliani et al. 2007; Christensen et al. 2007). Trimethylation at this residue is extremely important for transcriptional regulation and chromatin structure. It is likely involved in REST-mediated transcriptional repression. It has been shown to regulate the expression of several REST-mediated genes, as well as regulate the H3K4me2/H3K4me3 levels at their promoters (Tahiliani et al. 2007). JARID1C has many allelic variants causing disease and has been linked with NS-ID, S-ID and autism (Jensen et al. 2005; Abidi et al. 2008; Adegbola et al. 2008).

Two other PHD-domain genes are mutated in different forms of ID. Mutations in PHF6 (MIM: 300414) on Xq26.3 result in Borjeson-Forssman-Lehmann syndrome (BFLS). This is a form of S-ID presenting with severe ID, epilepsy, obesity, endocrine defects and several dysmorphic features (Brun et al. 1974). Mutations in PHF8 (MIM: 300560) on Xp11.2 cause ID with characteristic facial features and cleft lip/palette—a form of S-ID also known as Siderius X-linked mental retardation syndrome. The gene functions as an H3K9me1/me2 demethylase and contains a PHD finger that binds to H3K9me2/me3 (Kleine-Kohlbrecher et al. 2010). It has recently been shown that PHF8 interacts with ZNF711, and that these two protein products co-localize at PHF8 target genes in the nucleus, resulting in transcriptional activation (Kleine-Kohlbrecher et al. 2010). Interestingly, JARID1C is one of these target genes (Kleine-Kohlbrecher et al. 2010). This study shows that PHF8, ZNF711, and JARID1C are involved in a common pathway of transcriptional regulation and chromatin remodeling, providing further evidence that these functions are important for normal cognitive development. It also illustrates the value of identifying pathways involved in transcriptional regulation, for the study of ID.

MBD5 (methyl binding domain 5) has been proposed as an autosomal dominant candidate for ID. Mutations in this gene result in variable phenotypes, ranging from syndromic to apparently non-syndromic ID (Wagenstaller et al. 2007). Although not much is known about the function of this gene in humans, it is a methyl-CpG binding protein and may, like other methyl-binding proteins, function in regulation of transcription. It is also notable that BRWD3, a gene found in both NS-ID and S-ID contains a bromodomain—a conserved protein motif that recognizes acetylated lysine residues, and is found in chromatin associated proteins and histone acetyltransferases (Field et al. 2007; Sanchez and Zhou 2009). Although there is currently no experimental data to confirm that BRWD3 is a chromatin modifying protein, it is a possibility based on its motif structure (Field et al. 2007). Further work on both of these gene products will provide evidence of whether or not they are important for chromatin remodeling and regulation of transcription.

Because of the variable phenotypes resulting from mutations in this class of genes, it seems likely that chromatin maintenance and remodeling play a crucial role in normal cognitive development. It is possible that different mutations might lead to varying levels of protein function, and that function might be retained or disturbed differentially by different mutations. In all of these genes, multiple allelic variants are involved and several forms of S-ID as well as NS-ID may result from mutation.

These are only some examples of potential mechanisms by which NS-ID is developed. As we learn more about genes involved in NS-ID, more common biological pathways will become apparent, and will lead us to an understanding of how it is acquired, and what can be done to prevent it or alleviate it in genetic cases.