|Home | About | Journals | Submit | Contact Us | Français|
Recent studies are beginning to paint a clear and consistent picture of the impairments in psychological and cognitive competencies that are associated with microdeletions in chromosome 22q11.2. These studies have highlighted a strong link between this genetic lesion and schizophrenia. Parallel studies in humans and animal models are starting to uncover the complex genetic and neural substrates altered by the microdeletion. In addition to offering a deeper understanding of the effects of this genetic lesion, these findings may guide analysis of other copy-number variants associated with cognitive dysfunction and psychiatric disorders.
The 22q11.2 deletion syndrome (22q11.2DS; also known as velocardiofacial syndrome1 or DiGeorge syndrome2) is caused by a microdeletion in chromosome 22 and has an incidence of 1 in 2,000–4,000 live births3-5. Although atypical microdeletions have been described6, most of the microdeletions are either 3 megabases (Mb) in size7,8 (including approximately 60 known genes) or 1.5 Mb in size7,8 (including approximately 35 known genes), and most of the genes affected are expressed in the brain9 (Supplementary information S1 (table)). The phenotype of the disorder is highly variable10 and can affect multiple organs and tissues, but the severity is unrelated to the size of the deletion11, which suggests that genes within the 1.5 Mb region are crucial for the etiology of the syndrome. Phenotypic variability may be due to breakpoint heterogeneity as well as other genetic, environmental and stochastic factors. Common physical manifestations of the disorder include craniofacial and cardiovascular anomalies, immunodeficiency, short stature and hypocalcaemia3,10,12.
Individuals with 22q11.2DS also have cognitive and behavioural impairments and a high risk for developing schizophrenia. Long-term medical care and prenatal screening are increasingly being directed towards the treatment and recognition of these symptoms. Recent studies in humans and animal models have shed light on the neuroanatomical and cognitive manifestations of 22q11.2DS. In addition, there is growing evidence for a widespread role of copy-number variants (CNVs) — which include chromosomal microdeletions and microduplications — in determining susceptibility to cognitive disorders and schizophrenia13.
In this Review, we describe the cognitive, psychiatric and neuroanatomical phenotypes associated with 22q11.2DS. We summarize the emerging principles that underlie the genetic architecture of the syndrome and outline recent advances in animal model studies. Finally, we attempt to integrate existing knowledge on the disease and define future goals for the identification of the genetic and neural elements of the cognitive and psychiatric phenotypes. A detailed discussion of human genetic association studies in this locus is not within the scope of this Review.
Most school-aged children with 22q11.2DS have lower than typical full scale IQ (FSIQ). Borderline intellectual function (FSIQ 70–75) is most common, mild intellectual disability (FSIQ 55–75) is slightly less frequent and a small percentage of children fall into the low average intelligence range (FSIQ >85)14. Most children with 22q11.2DS achieve higher scores in verbal tasks than in nonverbal tasks, although this pattern of dysfunction is not universal14-17. Whether this profile of impairments changes as children age is unclear.
Not all functional domains have been studied equally, and a clear pattern of cognitive strengths and weaknesses in children with 22q11.2DS has not yet been established. However, standardized neuropsychological tests have revealed some consistencies. Although receptive language (the ability to understand what is being said) is generally strong in preschool children, by school age general language delays are apparent17-21. Nevertheless, during school years, reading, spelling and rote verbal memory scores are in the low average to superior range16,22,23 and verbal memory is consistently stronger than visual–spatial memory24-26. The domains of attention and executive function have not received detailed psychometric study, although impairments have been reported in both23,27. Overall, the profile of stronger verbal abilities and weaker nonverbal, spatiotemporal and numerical competence in most individuals is similar to that of other neurogenetic disorders, such as fragile X, Williams and Turner syndromes28.
The analytical power of standardized tests is limited because each one generally encompasses several cognitive functions. Analyses of more specific aspects of cognitive function in 22q11.2DS have been initiated (TABLE 1); however, this effort remains far from comprehensive. Much of the work attempts to explain the nonverbal cognitive impairments in 22q11.2DS in terms of changes in lower-level cognitive functions and the associated brain systems. Its hypothesis-driven nature has focused many studies on the attention system, which functions to select task- or goal-relevant information from the environment and to inhibit the processing of irrelevant information29-31. Attention seems to be essential for many nonverbal cognitive functions, which suggests that impairments in attention may be a major cause of the cognitive profile in 22q11.2DS. Studies suggest that children with 22q11.2DS have difficulty in finding and interpreting salient spatial and temporal information32-34, which indicates that their attention is engaged less efficiently in terms of navigating space, guiding vision and selecting and integrating goal-relevant information. Although the developing neural circuitry of children with 22q11.2DS has not been examined, these impairments indicate dysfunction in parietal and frontal cortical circuitry31,35.
Basic numerical cognition also depends on spatial attention32,36. Counting more than three or four items requires spatial attention, whereas subitizing — the rapid enumeration of fewer than three or four items — does not37-39. As predicted, counting but not subitizing performance is impaired in children with 22q11.2DS32. Children and adults with 22q11.2DS demonstrate impairments in ‘magnitude comparison’ and ‘time duration comparison’32,36,40, further strengthening the connections among spatial, temporal and numerical impairments and implicating dysfunction in the parietal and frontal circuitry that underlies attentional and numerical cognition41-44.
Attention is also important for inhibiting the processing of irrelevant information. Impairments in this ability have been reported in children with 22q11.2DS27,45-47. Other inhibitory impairments, such as reduced sensorimotor gating (usually in the form of inhibition of a startle response using the ‘prepulse inhibition’ (PPI) paradigm), have been reported and are associated with impairments in tests of attention and with high-risk schizophrenia symptoms27,48. Related functional impairments have been reported in children and young adults with 22q11.2DS49,50, using the ‘mismatch negativity’ (MMN) paradigm. This task, which is often associated with executive dysfunction and psychiatric symptoms, measures the ability to detect a novel stimulus within the context of many similar items.
Despite the importance of working memory (WM) as a core component of executive function and its association with the dorsolateral prefrontal cortex (PFC), there has been relatively little study of WM in 22q11.2DS. However, a functional MRI (fMRI) study of non-spatial WM51 in children and psychometric studies of children25 and adults52 have reported WM impairments in patients with 22q11.2DS.
A strong and specific relationship exists between the presence of the 22q11.2 microdeletion and schizophrenia53,54. Individuals with the 22q11.2 deletion are sometimes given other diagnoses early in life, including attention-deficit hyperactivity disorder (ADHD), generalized anxiety disorder, obsessive compulsive disorder and autism spectrum disorders55-59. However, with the exception of schizophrenia, most of these diagnoses may not meet the criteria set forth in the literature60,61 for a behavioural phenotype that is specifically associated with a syndrome. Specifically, most of these conditions are not found at a higher frequency among individuals with the 22q11.2 microdeletion than in cohorts with other developmental disorders associated with learning disabilities. In addition, no significant enrichment of 22q11.2 deletions has been identified in cohorts of children diagnosed with some of these conditions62-64. This indicates that the behavioural diagnoses other than schizophrenia represent nonspecific expressions of factors that affect brain development and function or, alternatively, that their diagnostic criteria are not designed for children with developmental disabilities and so may lead to the reporting of erroneously high rates of these phenotypes. Additional complications may arise owing to the association with schizophrenia. For example, diagnoses of autism spectrum disorders may reflect misdiagnosis of social impairments associated with premorbidity to schizophrenia65.
In late adolescence and early adulthood, up to one-third of all individuals carrying the 22q11.2 deletion develop schizophrenia or schizoaffective disorder as defined strictly in the Diagnostic and Statistical Manual of Mental Disorders66-70. This inordinately high risk of developing schizophrenia is not associated with any other neurogenetic syndrome. 22q11.2 microdeletions account for up to 1–2% of schizophrenia cases53,54,71,72 and are the only confirmed recurrent structural mutations that are responsible for introducing sporadic cases of schizophrenia into the population. The 22q11.2 microdeletion is therefore one of the highest known risk factors for schizophrenia (BOX 1). Importantly, there are no major clinical differences in the core schizophrenia pheno type between individuals with schizophrenia who are 22q11.2 microdeletion carriers and those who are not73,74. Moreover, due to high phenotypic variability, many individuals with 22q11.2DS who develop schizophrenia have no serious intellectual disability and their congenital abnormalities (such as facial dysmorphologies) can be so subtle that they may be missed on clinical evaluation, making these individuals indistinguishable from other schizophrenia patients. Whether any neuroanatomical abnormalities distinguish those patients carrying the 22q11.2 deletion who develop schizophrenia from those who do not is unclear (see below). The few completed studies aimed at defining the risk factors in 22q11.2DS that predict the development of psychosis have hinted at the relationships to focus on68 (BOX 2), but further exploration is needed.
Psychiatric disorders, like other common disorders, are multifactorial in nature with complex genetic etiologies. The genetic architecture underlying disease susceptibility is characterized by both the frequency and penetrance of risk alleles162. Frequency is defined here as the proportion of the chromosomes carrying the risk allele, and penetrance is defined as the proportion of individuals carrying a risk allele that actually exhibit the disease symptoms. The common disease–common allele (CDCA) hypothesis emphasizes the importance of relatively common alleles, each of low penetrance, acting together to increase disease risk. Conversely, the common disease–rare allele (CDRA) hypothesis emphasizes the impact of individually rare yet highly penetrant alleles. The reproducible observation of rare and highly penetrant de novo structural mutations at the 22q11.2 locus in sporadic (non-familial) cases of schizophrenia provided the first evidence to support the importance of rare recurrent mutations in schizophrenia vulnerability53. This association seems to be specific, as 22q11.2 microdeletions are not enriched in, for example, cases with autism64. A number of recent studies, made possible by technological advances, have extended and complemented the initial observation in the 22q11.2 locus54,71,72. Although the pattern of bidirectional association described for the 22q11.2 microdeletion has yet to be demonstrated for any of the recently discovered structural mutations, it is now widely recognized that rare structural mutations (both inherited and de novo) collectively have a substantial etiological role and account for a considerable portion of sporadic and familial cases of the disease. Notably, de novo mutations, such as the 22q11.2 microdeletions, can at least in part explain how schizophrenia persists in the population despite the low fecundity of affected individuals54.
Children with 22q11.2 microdeletions exhibit impairments in a range of cognitive functions that are also seen in those diagnosed with schizophrenia49,163,164. The same is true for some of the impairments in temporal processing165,166. It is unlikely that a simple relationship exists between individual cognitive dysfunctions or sets of cognitive dysfunctions and high risk for schizophrenia. Similarly, the predictive relevance for future psychosis of many structural, functional and connective brain findings remains equally unclear. Several studies49,56,57,68 have identified patterns of psychotic, prodromal and associated symptoms in young people with 22q11.2 deletion syndrome. The most stable symptoms seem to be attention deficit, mood and anxiety disorders and impaired social adaptive skills. The only known prospective study68 found that in children, the development of psychotic symptoms between being studied at age 12 and at age 18 was best predicted in part by the presence of psychotic symptoms at the time of the baseline study and in part by anxiety and depression scores. When predicting follow-up psychosis rating-scale scores, lower IQ at baseline was a further predictor. Attention-deficit hyperactivity disorder was not a predictor of psychotic outcomes. A related study167 found that intracranial and cerebellar white matter and superior temporal gyrus and caudate volumes increased with age but that amygdalar volumes decreased. Declining verbal IQ was associated with reductions in left cerebral grey matter.
Given the well-established links between specific brain regions and circuits, and the cognitive functions that are impaired in 22q11.2DS, it is reasonable to suggest that there may be specific neural changes in individuals with 22q11.2DS75. Given that many physical anomalies associated with 22q11.2DS affect midline structures, there is good reason to expect that this is also true for the brain.
As in other neurodevelopmental disorders, reduced brain volumes relative to typical developing (TD) controls are well documented in children with 22q11.2DS, with slightly greater reductions in posterior compared with anterior regional volumes and in total white matter compared with grey matter76-80. Despite an extensive literature on neuroimaging81, studies have been less consistent regarding changes in specific neural regions or structures. Therefore, to be conservative, we discuss here (BOX 3) only the differences that have been reported by at least two independent groups in separate studies involving mostly or exclusively children. Direct functional correlates of these structural changes have not yet been firmly established, largely due to the use of different measurement protocols in few studies with relatively small but differing samples and with diverse covariates.
Some brain structures are either volumetrically enlarged or reduced in children with 22q11.2 deletion syndrome relative to typical developing (TD) controls. Enlarged structures include the lateral ventricles, the caudate and the insula. However, both the hippocampus and medial cerebellum – including the vermis, anterior lobes and neocerebellum – are reduced relative to controls. The figure shows the location of some of these structures in the human brain. The table summarizes the findings of several different studies for each brain region.
Cerebellar differences are summarized in the table but were omitted from the figure because the findings were generated by taking into account the area derived from a single, mid-sagittal slice. Therefore, the extent of the difference in cerebellar size is not known. The figure was created by mapping regional parcellations onto a brain template using the Harvard–Oxford atlas (for the caudate, lateral ventricles and hippocampi) and the Montreal Neurological Institute (MNI) label atlas (for the cerebellum and insula). These atlases are provided in the brain atlas tools distributed with the FMRIB Software Library (FSL). The regions were rendered by in-house scripts created by S. Srivastava at the University of California, Davis, USA, using Matlab and OpenGL.
|Caudate||22q11.2 > TD control (volume)||76,79|
|Medial cerebellum (including vermis,|
anterior lobes and neocerebellum)
|22q11.2 < TD control (area)||170,171|
|Insula||22q11.2 > TD control (volume)||32,76|
|Hippocampus||22q11.2 < TD control (volume)||32,76-79|
|Lateral ventricles||22q11.2 > TD control (volume)||32,76,88|
Recently, more advanced methods have replicated and extended these findings. Children with 22q11.2DS and TD children have similar overall cortical thickness. However, lateral thinning in parieto-occipital, occipital pole and inferior prefrontal regions, and medial thinning in anterior cingulate, medial frontal gyrus, subgenual prefrontal, posterior cingulate gyrus, cuneus and lingual gyrus regions is found in 22q11.2DS. Furthermore, the correlation between cortical thinning and increasing age was greater in the 22q11.2DS group82,83. Reduced cortical gyral complexity was found in the frontal and parietal cortices of those with 22q11.2DS, with localized reductions overlapping many of the thinner cortical regions84,85. These differences in gyral complexity, might be associated with changes in neural connectivity, as it has been suggested that tensile influences created by developing connective tracts underlie cortical folding86.
Diffusion tensor imaging (DTI) has enabled the connective patterns of large axonal fibre tracts to be examined more directly, revealing changes in the posterior corpus callosum80,87 (FIG. 1 Supplementary information S2 (table)). Functional relationships have been reported between counting ability88, spatial attention36 and arithmetic89 and frontal and parietal DTI measurements or callosal area, which suggests that this altered connectivity might contribute to these phenotypes in the 22q11.2DS.
There have been a few fMRI studies on children with 22q11.2DS, which revealed atypical parietal lobe activation during an arithmetic task90, and atypical parietal activation (rather than prefrontal activation) in adolescents with 22q11.2DS in a non-spatial WM task51 and an inhibitory Go–NoGo task91.
This pattern of cortical and subcortical changes is generally consistent with the cognitive impairments observed in 22q11.2DS and with well-established links between particular neuroanatomical structures and cognitive functions in typical individuals. These differences thus provide evidence for atypical development of particular brain circuits in 22q11.2DS. Although the mechanisms that drive this atypical development are not well understood, neuroimaging studies in adolescent and adult carriers of the 22q11.2 microdeletion suggest the presence of a progressive, abnormal maturation process. In individuals with 22q11.2DS, the rostrocaudal gradient of structural alterations described in childhood seems to disappear with age, giving way to more widespread loss of brain tissue, particularly in the frontal and temporal lobes.
The structural anomalies that distinguish those 22q11.2 microdeletion carriers who develop schizophrenia from those who do not are unknown. Preliminary results point to some distinguishing features that are consistent with changes reported for general schizophrenia. These include larger lateral and third ventricles and a decrease in grey and especially white matter volumes in frontal and temporal lobes92,93. In addition, carriers who develop schizophrenia exhibit cognitive impairments that may reflect differences in the development and function of frontal brain regions52,94. One longitudinal study also suggests that decreases in frontal grey matter volume contribute to the susceptibility of adolescents with 22q11.2DS to schizophrenia95.
The findings described indicate a pattern of cognitive dysfunction characterized by an impairment of attentional functioning, accompanied by a very high risk of developing schizophrenia. These symptoms seem to be correlated with aberrant structure, connectivity patterns or activity in several cortical and subcortical areas. Given that a large number of genes are affected by the 22q11.2 microdeletion, these aberrations are likely to be due to combined anomalies in neuronal development, connectivity and synaptic plasticity.
There are at least three ways in which the removal of genes from the region deleted in 22q11.2DS could cause the cognitive and psychiatric phenotypes of the disorder: firstly, the microdeletion may unmask one or more recessive mutations; secondly, a single dosage-sensitive gene may exert a major effect; and lastly, underexpression of a number of dosage-sensitive genes may have a cumulative effect. According to the cumulative under expression model, although some genes may have a greater impact on phenotype, the imbalance of several deleted genes determines the overall phenotype. The presence of additional trans- or cis-acting genetic modifiers (including DNA variation in genes in the intact chromosome) may also contribute to the variability of the cognitive and psychiatric phenotypes95,96-98.
As outlined below, existing data from animal models also seem to support the underexpression model, and this suggests that there is an oligogenic basis for the neural, behavioural and cognitive 22q11.2DS phenotypes and that the sets of genes responsible for specific deficits may only partially overlap. Indeed, systematic and candidate genetic association approaches also seem to implicate more than one 22q11.2 gene in the psychiatric and cognitive phenotypes of 22q11.2DS99,100. Studies that have linked specific 22q11.2 genes to schizophrenia are indicated in Supplementary information S1 (table), and a more comprehensive account can be found on the Schizophrenia Research Forum website. A detailed discussion of each study is beyond the scope of this Review. These studies typically assume that common variants in the 22q11.2 genes contribute to the disease risk in karyo-typically normal individuals, and they have provided important initial insights into the neurobiology of the 22q11.2DS101-107. Given the well-established difficulties in determining genetic causality in psychiatric disorders using genetic association studies of common variants108, these findings remain for the most part only suggestive due to the small sample sizes used, especially in the absence of functional data. Until these issues are resolved using novel next-generation sequencing approaches, genetically engineered animal models offer a complementary and powerful approach for dissecting the genetic and neural influences that shape the cognitive and behavioural manifestations of the 22q11.2DS.
Several questions arise from the human studies described above. These studies suggest that function and connectivity of several brain areas are altered in 22q11.2DS. However, the studies could not address how abnormal function and connectivity are manifested in neural circuits at the molecular, cellular and synaptic levels. Moreover, given the aforementioned limitations of current human genetic approaches, it remains unclear how human findings relate to individual genes from this locus. Model organisms that mimic the human condition are likely to have an important role in clarifying these issues because they allow direct assessment of the impact of genetic factors on neural activity across multiple brain areas.
Cognitive and psychiatric disorders can be deconstructed into individual phenotypic components, which in turn can be analysed in animal models109. The mouse equivalent of the human 1.5 Mb 22q11.2 locus harbours orthologues of all the human functional genes except clathrin, heavy polypeptide-like 1 (CLTCL1). Another difference is that humans (but not mice) carry two functional genes (DiGeorge syndrome critical region 6 (DGCR6) and DGCR6-like (DGCR6L)) in the deleted locus, as a result of intralocus duplication. Overall, there is a high degree of conservation with the 22q11.2 region, which provides an opportunity to create etiologically valid mutant animal models. Indeed, several studies have reported the presence of neuronal and behavioural abnormalities resulting from deletions of this locus (FIG. 2). In addition, several groups have inactivated the mouse orthologues of individual genes from the 22q11.2 region and tested the mutant mice for neuronal and behavioural abnormalities.
Analysis of mouse models of 22q11.2DS has focused primarily on hippocampal and frontal circuitry — for example, the mouse frontal cortex region that is functionally analogous to the human dorsolateral PFC based on pharmacological and lesions studies110 — because many cognitive impairments described in individuals with 22q11.2DS probably depend on these brain areas. Furthermore, several studies suggest that functional connectivity between the PFC and temporal lobe is abnormal in schizophrenia111-113. Specific genetic pathways have been identified that affect neuronal connectivity and neuromodulation and may be crucial for the cognitive dysfunction and high risk of schizophrenia seen in 22q11.2DS.
Recently, insights have emerged from the Del(Dgcr2–Hira)2Aam mouse strain (in which Aam refers to the initials of the genetic engineer who created the strain) (FIG. 2). Studies in this strain have provided compelling evidence that the 22q11.2 microdeletion results in abnormal processing of brain microRNAs (miRNAs)114, a class of small, non-coding RNAs that regulate the stability and translation of mRNA115 (Supplementary information S3 (figure)). One gene disrupted by the 22q11.2 microdeletion is Dgcr8, a component of the ‘microprocessor’ complex that is essential for miRNA production116. Dgcr8 haploinsufficiency results in the downregulation (by ~20–70%) of a specific subset of mature miRNAs (Supplementary information S3 (figure)). Two additional miRNA genes (mir-185 and mir-649) are removed by the 22q11.2 microdeletion. A synchronous, modest decrease in several miRNAs may have considerable impact on target transcript stability114 but direct targets for these miRNAs have not been reported so far. Nevertheless, the potential of miRNAs to contribute to the epigenetic regulation of expression of multiple genes in the brain could be an important component of the genetically complex architecture of psychiatric and cognitive phenotypes associated with the 22q11.2DS and may also contribute to the phenotypic variability117. This discovery highlights emerging novel connections between miRNAs and psychiatric and neurodevelopmental disorders.
Evidence has been found for abnormalities in the formation of dendritic spines and dendritic complexity in cultured hippocampal neurons and in CA1 pyramidal neurons in Del(Dgcr2–Hira)2Aam mice118. This is consistent with the dysregulation of a number of neuronal and synaptic genes (see below). Cultured neurons showed a reduction in the density of mushroom spines (the larger, mature and most stable spines), as well as a decrease in spine head-width and length. Furthermore, electrophysiological recordings and quantitative immunocytochemistry indicated a reduction in the density of functional glutamatergic synapses. Finally, there was a decrease in the number of primary dendrites, the extent of dendrite branching and the total length of dendrites. The concordance of the analysis in dissociated cultures with that in the intact animal was extensive; however, several differences were observed. The reduction in the number of glutamatergic synapses and the density of mushroom spines was less pronounced in vivo than in vitro and the length of mushroom spines seemed to be normal in vivo118. This may be due to long-range network effects that cannot be recapitulated in culture and to compensatory changes in gene expression during development. Nevertheless, even modest alterations in dendritic and spine formation may result in a suboptimal number of synaptic connections, an increased number of inappropriate connections or changes in the integration of synaptic inputs, all of which will ultimately lead to considerable changes in information processing119,120.
It is unknown whether similar alterations in dendritic morphology are also present in the cortex. However, using mice from the Del(Dgcr2–Hira)1Rak strain (FIG. 2), it was shown that diminished dosage of 22q11.2 genes compromises neurogenesis and subsequent neuronal differentiation in the cerebral cortex121.
Future experiments need to address the developmental course of all of these changes. Unfortunately, it is unknown whether similar cytoarchitectural alterations are present in the brains of humans with 22q11.2DS. As such, it remains unclear whether the severity and extent of these alterations correlates with cognitive and behavioural dysfunction. Nevertheless, it is tempting to speculate that changes in dendritic complexity and neuronal density may account, at least in part, for the reduction in cortical thickness or regional decreases in grey matter volumes observed in some individuals with 22q11DS. Changes in the distribution of cortical inhibitory neurons may contribute to higher-order cognitive dysfunctions, whereas alterations in spine densities and size may contribute to the observed intellectual dysfunction.
To understand how the abnormalities in neural structure in mouse models of 22q11.2DS emerge, it is important to consider the contribution of individual genes and genetic pathways. Hemizygosity for DGCR8 and the consequent changes in miRNA abundance may contribute to deficits in synaptic development and maturation122-124. In addition, other genes in the 22q11.2 locus might contribute to these changes (Supplementary information S1 (table)) — for example, zinc finger, DHHC-type containing 8 (ZDHHC8) encodes a palmitoyl-transferase (PAT) and is a member of a multigene family of PATs101,118,125,126. Evidence from several studies suggests that ZDHHC8 is localized in the Golgi apparatus and vesicular compartment of neurons118,127 and that protein palmitoylation is its central function. One report has suggested that ZDHHC8 is localized to the mitochondria, where it has been proposed to have other functions128. However, this finding needs to be confirmed, as it is based on the overexpression of potentially mistargeted carboxy-terminal green fluorescent protein (GFP) fusion protein in non-neuronal cell lines and the use of antibodies of unknown specificity. Palmitoylated proteins can regulate the formation and function of dendritic spines and glutamatergic synapses129,130. ZDHHC8 seems to be involved in the palmitoylation of a specific subset of neuronal proteins, including the postsynaptic density protein PSD95 (also known as DLG4), an adaptor molecule that was previously shown to modulate the number of dendritic spines and possibly of dendritic branches118. Recent studies show that the complementary effects of impaired miRNA biogenesis and impaired palmitoylation contribute to the impaired dendritic growth and spine development observed in Del(Dgcr2–Hira)2Aam mice. Decreases in spine density could be accounted for by deficiency of Zdhhc8, whereas decreases in spine size could be accounted for by deficiency of Dgcr8. Similar complementary effects seem to influence dendritic growth. Hemizygosity of other genes in the 1.5 Mb region (Supplementary information S1 (table)) may also contribute to the observed neural abnormalities, although evidence from animals lacking some candidate genes has been mostly negative131,132.
Genes located outside the 1.5 Mb region, but within the 3 Mb region of the 22q11.2 deletion (Supplementary information S1 (table)), may also contribute to the neuronal manifestations of 22q11.2DS. A ubiquitously expressed syntaxin-binding protein (synaptosomal-associated protein 29 (SNAP29)) modulates the function of the machinery that mediates membrane fusion, a crucial step in synaptic vesicle recycling. Low SNAP29 levels in primary neurons affect synaptic transmission133, whereas homozygous truncating mutations of the gene cause cortical dysplasia134 and abnormalities in the corpus callosum. Deficiency of another gene (CRK-like (CRKL)) in the 3 Mb region may affect signalling mediated by reelin, an extracellular matrix protein that is crucial for cell positioning and neuronal migration during brain development135. Unfortunately, it remains unknown whether any one of these genes is haploinsufficient in 22q11.2DS.
In addition to the reduction in dosage of genes in the 22q11.2 region, microarray analysis of the Del(Dgcr2–Hira)2Aam mouse strain revealed more widespread, genome-wide alterations of transcriptional programs114. These reflect either downstream effects of gene deletions in the 22q11.2 region (for example, of DGCR8) or compensatory changes. Large-scale changes in expression of genes flanking the deletion have not been reported. In the hippocampus (HPC), the genes affected were primarily related to synaptic transmission. In the PFC, expression changes involved mitochondrial genes; it remains to be determined whether this was directly related to inadequate levels of known mitochondrial genes (such as proline dehydrogenase (Prodh), mitochondrial ribosomal protein L40 (Mrpl40), thioredoxin reductase 2 (Txnrd2) and solute carrier family 25, member 1 (Slc25a1)) or putative mitochondrial genes (such as T10 (also known as D16H22S680E))128 (Supplementary information S1 (table)) in the 22q11.2 region or to adaptive energetic changes caused by altered PFC function. Interestingly, changes in mitochondrial genes in the dentate gyrus of the Del(16Es2el–Ufd1l)217Bld strain (FIG. 2) were also reported136.
Dopamine modulates cognition and psychosis but its role in 22q11.2DS has not been assessed directly in mouse models. However, accumulating evidence shows that catechol-O-methyltransferase (COMT), a gene deleted in 22q11.2DS that encodes an intracellular, postsynaptic enzyme, modulates dopamine clearance in the PFC. Removal of dopamine from the extracellular space following evoked release was twofold slower in Comt (also known as Comt1)-homozygous knockout mice137. Notably, baseline dopamine levels seemed to be normal in both heterozygous and homozygous Comt-deficient mice138,139. This suggests that the partial reduction in CoMT levels that occurs in 22q11.2DS may have an effect primarily under conditions of increased dopamine release. Such an increase might result from deficiencies in other genes in or outside the microdeletion, from environmental factors or from both of these.
This scenario is supported by analysis of mice lacking Prodh, another gene located in the 22q11.2 microdeletion that has been linked by human genetic studies to schizophrenia103. Prodh encodes an enzyme that metabolizes l-proline140 — a putative neuromodulatory amino acid that may affect glutamatergic or GABA (γ-aminobutyric acid)-ergic transmission96,141. Approximately half of the individuals with 22q11.2DS have elevated serum levels of l-proline, which can predispose to learning impairments, epilepsy142 and schizoaffective disorder143,144. Mice carrying a missense mutation in Prodh have high serum levels of l-proline that are comparable to those observed in 22q11.2DS145. Although no anomalies in basal cortical dopamine levels were detected in these mice, in vivo microdialysis revealed increased cortical dopamine release in the PFC following acute, systemic administration of the psychostimulant amphetamine146. Prodh-deficient mice also show hypersensitivity to the locomotor effects of amphetamine. Notably, this effect is potentiated by pharmacological inhibition of COMT activity and levels of COMT transcript and protein are upregulated in the PFC of Prodh-deficient mice, probably indicating a negative-feedback loop that restrains increased local hyperdopaminergia. Thus, there is an epistatic interaction between the two genes, and the effect of the Prodh gene is modified by Comt146.
The neural circuitry and synaptic plasticity subserving perception, attention and memory can be readily characterized in mice through objective tests. The most important criterion in choosing behavioural tests to model the human deficit is the conservation of neural circuitry underlying a given behaviour109. However, the underlying neural circuitry for the behaviours and the psychological capacities that are impaired in individuals with 22q11.2DS has not yet been fully characterized. Nevertheless, studies suggesting functional and structural pathology in the hippocampal and frontal circuitry of individuals with 22q11.2DS have prompted assessment of behavioural and cognitive modalities that typically depend on these areas.
Behavioural analysis of animal models of the 22q11.2DS has provided evidence that one or more genes from this locus modulate sensorimotor gating, fear memory and WM (FIG. 3a–c). Although other types of memory (such as reference or recognition memory) have not been studied, future studies comparing performance across various tasks will help to disentangle the mechanisms underlying the observed learning impairments. Unfortunately, considering the central importance of attentional dysfunction in 22q11.2DS and schizophrenia, there is a dearth of studies measuring attention in mutant models of 22q11.2DS. However, several rodent paradigms for measuring attentional control exist147 and can be used in future studies designed to model and dissect attentional deficits.
Cognitive impairments are core features of schizophrenia and may be mediated by the same abnormal neural processes as other human symptoms, such as hallucinations (which are less amenable to study in animal models)147. Therefore, analysis of cognition in animal models of 22q11.2DS may also have important clinical implications for understanding the increased schizophrenia risk associated with this locus147.
Mice deficient for 18 of the orthologues of human genes in the 1.5 Mb region (Del(16Es2el–Ufd1l)217Bld mice) (FIG. 2) have impaired conditioned fear memory148, whereas those hemizygous for just 7 of the orthologues (Del(16Zpf520–Slc25a1)1Awb mice (FIG. 2) do not149. Similarly, mice with a hemizygous deletion spanning all of the orthologous genes in the 1.5 Mb region (Del(Dgcr2–Hira)2Aam mice) (FIG. 2) exhibit robust deficits in both cued and contextual fear memory114. One simple interpretation of these findings is that one or more of the genes between Slc25a1 and Hira (FIG. 2) are important for the formation of fear memory. However, as explained below, it may be difficult to infer such simple genotype–phenotype relationships in 22q11.2DS based on phenotypic characterization of variable size deficiencies.
Del(Dgcr2–Hira)2Aam mice also show deficits in spatial WM tasks, consistent with studies in children and adolescents with the 22q11.2DS27,48,51. These mice are impaired in the acquisition of a delayed non-match to place (DNMP) task114 that assesses spatial WM-dependent performance. Interestingly, this deficit arises in part from deficiency of Dgcr8 (and presumably miRNAs), as heterozygous deletion of Dgcr8 (Dgcr8Gt(XH158)Byg) alone also affects acquisition of this task114 without affecting associative memory. By contrast, hemizygosity of other genes in the 1.5 Mb deletion does not seem to cause robust WM deficits: mice deficient for the NOGO receptor (reticulon 4 receptor (Rtn4r)tm1Gogo), Comt150-152 or Prodh146 perform normally in DNMP. However, interfering pharmacologically with the epistatic interaction between Comt and Prodh unmasks an underlying dopamine dysfunction and reveals WM deficits in Prodh mutant mice146. Notably, inhibition of COMT also exacerbates other behaviours that are influenced by cortical dopamine (such as sensitivity to amphetamine and PPI)146.
These observations indicate that microdeletion carriers, who carry only one COMT copy, may be particularly disadvantaged by their inability to restrain emergent dopamine dysregulation. Indeed, subsequent studies of human carriers confirmed an epistatic interaction between l-proline levels and low COMT activity. These studies provided additional evidence that individuals with 22q11.2DS who carry a low-activity COMT variant (Met-COMT) in the intact chromosome, especially those with the most elevated serum levels of l-proline, are at higher risk for psychosis or related cognitive and neurophysiological impairments95-97. These studies collectively show the power of iterative human genetic and animal model studies for understanding the neural substrates and genetic influences of the 22q11.2DS.
Studies addressing individual genetic contributions to PPI have been particularly informative in addressing the genetic architecture of 22q11.2-linked neurobehavioural impairments. Like children with 22q11.2DS, animal models carrying large deficiencies at the 22q11.2 locus114,148,153 have consistently shown abnormalities in PPI, although the pattern of alterations differs among studies. A number of genes in the deleted region differentially affect PPI. Heterozygous deficiency of guanine-nucleotide-binding protein (G protein) β-polypeptide 1-like (Gnb1l), Dgcr8 and possibly T-box 1 (Tbx1), as well as elevations in l-proline levels (especially in conjunction with inhibition of COMT activity) decrease PPI114,146,153,154, whereas heterozygous deficiency of Rtn4r, Zdhhc8, Comt and septin 5 (Sept5) has no effect101,131,150,155. It is interesting that one set of 22q11.2 deletion mice (Del(16Zpf520–Slc25a1)1Awb mice) (FIG. 2) were shown to have increased PPI149. This suggests that alterations in PPI are the result of interplay between both positive and negative contributions from genes in this locus. Similar patterns of genetic contributions may underlie other behavioural impairments, underscoring the difficulty in using traditional deletion-mapping approaches to dissect genotype–phenotype relationships from models of large CNVs.
These behavioural studies provided new and unexpected insights into the genetic architecture of the human syndrome, including the contribution of distinct or partially overlapping sets of genes to specific behavioural deficits, the identification of positive and negative influences of local genes and the existence of local negative-genetic-feedback loops. Behavioural studies also greatly facilitated the identification of neural substrates — for example, observation of deficits in fear memory indicates a previously largely unappreciated amygdala circuit dysfunction in 22q11.2DS. Although fear memory has not been assessed in humans with 22q11.2DS, amygdala dysfunction may have a crucial role in the pathophysiology of 22q11.2-associated anxiety disorders. Finally, the fact that a number of behavioural alterations described in humans (such as deficits in PPI or WM) could be recapitulated in mice supports the view that 22q11.2 mouse models have strong validity and can be used to dissect the underlying abnormalities in genes and neural circuits.
Macroscopic measurements of brain structure and activity in 22q11.2DS individuals suggest that structural and functional long-range connectivity between brain areas may be altered, an anomaly that may also be related to the onset of psychosis. Recent work capitalizing on the observation of impaired WM performance in 22q11.2 animal models has helped to provide important new insights into the nature of altered brain connectivity that emerges as a result of the 22q11.2 microdeletion. The coordinated, rhythmic activity of neuronal populations in the brain gives rise to oscillations at a broad range of frequencies156,157, which are often used to synchronize neural firing of interconnected cells in and between brain areas — for example, PFC neuron spiking is modulated by HPC θ-frequency (4–12 Hz) oscillations158. In a recent study159, HPC–PFC synchrony was recorded in Del(Dgcr2–Hira)2Aam mice while they performed a task requiring WM (FIG. 3d). In wild-type mice, this synchrony was increased during WM performance, whereas Del(Dgcr2–Hira)2Aam mice showed dramatically reduced HPC–PFC synchrony159 (FIG. 3d). Furthermore, the magnitude of HPC–PFC synchrony at the onset of training predicted the time it took the mutant mice to learn the WM task. Notably, synchrony in HPC or PFC was not impaired, suggesting that θ-oscillations can organize local neural activity. The results show that the 22q11.2 microdeletion can disrupt functional connectivity and synchrony among brain areas. Abnormal PFC–HPC coupling has been observed in schizophrenia patients111-113 as well as in healthy individuals carrying candidate schizophreniarisk variants160,161, and may affect not only cognition but also other aspects of the clinical diagnosis. Therefore, the results suggest that 22q11.2 deletion may contribute to the development of schizophrenia by altering prefrontal connectivity.
FIG. 4 attempts to conceptualize the pathogenesis and pathophysiology of 22q11.2DS and to guide future studies of this locus, as well as those of other pathogenic CNV loci. According to this framework, reduction in dosage of a subset of neighbouring genes has a cumulative effect on the properties of neural networks. This results in locus-specific neuronal dysconnectivity, aberrant synaptic plasticity and neuromodulation, as well as altered functional integration in and across brain regions. These effects lead to cognitive dysfunction and schizophrenia when a critical threshold is surpassed. Although one or a few genes may have a greater impact, it is the cumulative effect of the imbalance of several genes in the deletion that determines the overall phenotype. Genes responsible for specific deficits are only partially overlapping. The presence of additional genetic (trans- or cis-acting) and environmental modifiers may also contribute to the variability of the cognitive and psychiatric phenotypes of the syndrome.
What determines the specificity of the phenotype? Similar neuronal abnormalities in dendrites or spines, for instance, have been observed in other neurogenetic syndromes, such as fragile X syndrome, that do not share the pattern of cognitive and psychiatric impairments observed in 22q11.2DS. There is not yet a satisfactory answer to this question; however, as analyses become more detailed, differences at the cellular level as well as in the spatial distribution pattern across the whole brain may reveal distinct anomalies in local and long-range connectivity and synaptic plasticity that may account for the phenotypic differences among syndromes. In addition, abnormal neuromodulatory function in dopamine neurotransmission brought on by combined deficiency in genes specific to this locus could further alter network properties, modify the cognitive phenotypes and possibly predispose a subset of microdeletion carriers to psychiatric symptoms.
A number of details at the circuit and neuronal level remain to be elucidated. The relationship among brain structure, function, connectivity and cognitive impairments in children and adults with 22q11.2DS is an important but challenging75 area of future research and will be essential to compare anomalies in mouse models and humans. In addition, there is a need to undertake longitudinal prospective studies — including experimental and standardized cognitive assessments alongside neuroimaging — to identify one or more endophenotypes of schizophrenia, as well as a predictive prodrome that can be preventively treated during childhood and adolescence.
It will also be interesting to determine how the affected genetic pathways (such as miRNAs) modulate the input and output function of specific brain areas and what accounts for the variable degree of penetrance of the genetic effects. However, the complexities of the genetic architecture of the disease risk associated with this locus suggest that approaches that focus on the mutation as an entity rather than deconstructing it into individual genetic components may prove equally fruitful. A number of questions remain to be addressed. What is the effect of the microdeletion on inhibitory neurotransmission or on both short- and long-term synaptic plasticity? How do such deficits correlate with the cognitive impairment? How does brain dysconnectivity relate to the etiology of the psychiatric and cognitive symptoms and how does it manifest in the activity of neural circuits or at the level of single neurons? How widespread are the recently identified deficits in brain synchrony? And are anatomical connections affected in terms of axonal extension, branching, myelination or synaptic properties? Finally, there is also a need to identify which of the molecular, cellular and circuit level deficits can serve as the best targets for the development of treatments for the cognitive and psychiatric symptoms.
The authors wish to acknowledge grant support for their work from the US National Institutes of Health (grant R01MH67068 to M.K. and J.A.G. and grant R01HD42974 to T.J.S.). J.A.G. also gratefully acknowledges support from the Simons Foundation and M.K. from the McKnight and March of Dimes Foundations. The authors wish to thank A. Arguello for help with creating figures figures22,,33 and and4,4, B. Xu for help with Supplementary information S1 (table) and Supplementary information S3 (figure) and S. Srivastava for help with box 3 and figure 1. The authors also thank A. Arguello, L. Drew, B. Xu and other members of their laboratories for comments and critical feedback.
Competing interests statement The authors declare no competing financial interests.
UniProtKB: http://www.uniprot.org SNAP29
Maria Karayiorgou and Joseph A. Gogos’ homepage: http://schizophreniaresearch.columbia.edu
Tony J. Simon’s homepage: http://cabil.mindinstitute.org
Schizophrenia Research Forum: http://www.szgene.org
See online article: S1 (table)|S2 (table)|S3 (figure)
ALL LINKS ARE ACTIVE IN THE ONLINE PDF