DiGeorge syndrome, velocardiofacial syndrome and various other malformations have been described in association with deletions and translocations involving human chromosome 22q11. Many of the structural malformations observed are also seen in animal models of neural crest disruption suggesting that the haplo-insufficiency resulting from the deletion somehow affects this group of cells or their interactions. Over the past few years it has been shown that the deletion predisposes to a range of psychotic conditions prompting the hypothesis that the deleted region may contain a predisposition locus for psychotic illness. The DiGeorge chromosomal region has been entirely sequenced and many of the genes mapping to the deletion interval have been studied in some detail. Despite these efforts, no gene has yet been proved to play a defined role in the pathogenesis of the syndrome. Current efforts are directed at the study of engineered chromosome mouse models which offer the potential to dissect at least some of the developmental pathways disrupted in this intriguing group of malformation syndromes.
Received 14 June 2000; Revised and Accepted 14 July 2000.
Deletion of chromosome 22q11 is the most frequent known interstitial deletion found in man with an incidence of 1 in 4000 live births. Although published estimates vary, it is likely that 5–10% of deletions are inherited. The deletion has been reported in association with >80 different birth defects and malformations occurring in many combinations and with widely differing severity. This has resulted in the deletion being linked with several diagnostic labels including the DiGeorge syndrome (DGS), velocardiofacial (or Shprintzen) syndrome (VCFS), conotruncal anomaly face, Cayler syndrome and Opitz GBBB syndrome (1). As different diagnoses have been made in different members of the same family segregating the same deletion, it is reasonable to view these diagnoses as reflecting various outcomes of the same underlying genetic defect. For the purposes of this review the collective term 22q11 deletion syndrome (22q11DS) will be used.
The main cause of morbidity and mortality in the disorder is congenital heart defect (CHD) (2). Some form of CHD is found in 75% of patients and, in one series, 43 of 44 patients who died before 6 months of age did so primarily because of heart problems (2). The type of CHD observed is variable, but frequently there is involvement of the outflow tract of the heart and the derivatives of the branchial arch arteries. Malformations include interrupted aortic arch type B, persistence of the truncus arteriosus and tetralogy of Fallot. Associated birth defects include a hypoplastic or aplastic thymus gland and parathyroid gland, which can result in cell-mediated immunity and hypoparathyroidism, respectively. There is facial dysmorphism which is most pronounced during childhood. These, and the other structural malformations of 22q11DS, have been well described elsewhere (2,3). Learning difficulty and behavioural problems are common and have been a focus of several studies in recent times.
NEUROPSYCHIATRIC PROBLEMS IN 22q11DS
Many studies have reported an increased incidence of psychotic illness (schizophrenia or bipolar disorder) in adolescents and adults with the 22q11 deletion (4–6). Although most studies have been retrospective in the sense that they assessed the psychological status of patients known to have a 22q11 deletion, two studies in adults screened for 22q11.2 deletions in patients ascertained for psychiatric illness. The first series found a deletion in 2 of 100 schizophrenic patients (7) and the second a deletion in 1 of 326 psychiatric inpatients (8). IQ in the schizophrenic group was not significantly different from the non-schizophrenic deletion group. Usiskin et al. (9) recruited 47 patients with childhood onset schizophrenia and found three with a deletion; all patients had premorbid impairments of language, motor and social development. A separate study identified a deletion in 1 of 91 patients meeting DSM-IIIR criteria for pervasive developmental disorder (10).
Several studies present preliminary data on the cognitive and behaviour profile of children, of varying ages, with the 22q11 deletion (11–14). It is clear that deficits are widespread and not dependent on the presence of associated anomalies such as congenital heart defect or cleft palate. Median IQ is ∼75 and usually in the range of 50–100 (figures significantly lower than for cleft palate or heart defect controls). Children have a receptive–expressive language impairment from the onset of language, with speech and expressive language development delayed beyond what might be expected given their developmental or receptive language performance. Children with a deletion have a greater incidence of behaviour problems and personality problems than matched controls with a distinct craniofacial anomaly, internalizing problems and attention problems being prevalent. Intervention strategies are currently being assessed (15).
A number of conclusions can be drawn from these studies. Individuals with 22q11 deletions have moderate mental retardation and a variety of cognitive deficits. They may progress to frank psychiatric illness, with this progression relatively independent of IQ or coexisting malformations. Like other manifestations of this deletion syndrome, the neuropsychiatric phenotype can be highly variable between individuals with the same deletion.
Many groups have attempted to map schizophrenia and bipolar depression predisposition loci in the general psychiatric population. Although recently proposed criteria for significance in such studies have not been reached (16), it is interesting that several analyses have found positive lods in the 22q11–q12 region (17–20). Investigators are therefore intrigued by the possibility that a gene in the VCFS deletion region acts as a schizophrenia predisposition locus.
EMBRYOLOGY OF THE STRUCTURES AFFECTED IN 22q11DS
The structures primarily affected, thymus gland, parathyroid gland, branchial arch arteries and face, are all derivatives of the branchial arch/pharyngeal pouch system. Furthermore, each of the tissues involved receives a contribution from the rostral neural crest during embryogenesis (Fig. 1). Experiments which ablate the premigratory crest or otherwise perturb crest function can produce a good phenocopy of the main features of 22q11DS (21–26). Natural or induced mutations of genes thought to have a role in controlling neural crest development can similarly give malformations reminiscent of 22q11DS (27–30). These observations stimulated the hypothesis that the deletion of 22q11 disrupts the rostral neural crest cells, or the cells with which the neural crest interacts, at a critical phase of organogenesis.
Mapping the deletions
In common with strategies pursued for other chromosome deletion syndromes, investigators compared deletions from different patients in order to establish a shortest region of deletion overlap (SRDO) map. Most patients were found to have an interstitial deletion of ∼3 Mb, the typically deleted region (TDR) (1). Comparison of terminal deletions and interstitial deletions gave a deletion overlap of ∼750 kb and rapid progress seemed assured when a chromosome 2:22 balanced translocation breakpoint known by the acronym ADU was shown to map to this SRDO. The SRDO has been called the DiGeorge syndrome chromosomal region (DGCR) and this acronym will be used here, for the sake of simplicity. It was confidently predicted that the main malformations found in 22q11 DS were the consequence of the haplo-insufficiency of one gene mapping within the DGCR and being disrupted by the ADU balanced translocation. The balanced translocation breakpoint was cloned (31), but the transcripts disrupted by the rearrangement seemed to belong to a pseudogene (32). Moreover, the 2:22 breakpoint was subsequently found to map 100 kb proximal to a DGS deletion reported by Levy et al. (33) (the so-called G deletion). This raised the possibility that the gene or genes were haplo-insufficient in 22q11DS could be influenced by long-range effects of chromosomal rearrangement and, therefore, that SRDO and breakpoint mapping would prove of limited use in precisely mapping any critical gene. To emphasize this point, a region referred to as the minimal DGCR (MDGCR) (Fig. 2) is defined by the overlap of the TDR with a 15:22 translocation (34). There is minimal overlap between the MDGCR and the G deletion, encompassing the CTP and CTLD genes, neither of which is likely to have a role in the syndrome.
Over the last 5 years there have been a number of reports describing DGS- or VCFS-like features in patients with atypical deletions (35–38). Some of these do not overlap with the SRDO and one has no overlap with the TDR (39), making it difficult to come to any overall conclusion concerning the position of critical genes. Long-range effects of balanced chromosome translocations have been observed at distances of up to 800 kb from the SOX9 gene in campomelic dysplasia (40). Therefore, if 22q11DS can be ascribed to haplo-insufficiency of a single gene, one might expect it to be within the 1 Mb of DNA distal to the ADU breakpoint.
Mechanism of deletion
A study of chromosome 22 recombination in parents of children with a 22q11 deletion detected an excess of meiotic events within 22q11, suggesting that unequal crossing over may underlie deletion origin. The early stages of the physical mapping of 22q11 were hampered by the presence of low-copy number repeat (LCNR) elements which in many cases were unique to this region of the genome (41). More detailed analyses of the breakpoints of interstitial deletions have identified clusters of different LCNR classes at the breakpoints (42–44). The repeats at the typical deletion breakpoints extend over 200 kb and include GGT, BCRL, V7-rel, POM121-like and GGT-rel sequences. As the repeats are virtually identical in these two blocks it is likely that the deletion is mediated by homologous recombination, probably intrachromosomal, at these points. This hypothesis is supported by the description of one family segregating the reciprocal duplication (43). Interestingly, the breakpoint of the only known recurrent, non-Robertsonian constitutional translocation in man, the t11:22, occurs in one of the 22q11 LCNRs (45,46).
Genes within the deletion
Several teams have contributed the positional cloning of genes from the 22q11 deletion interval and the region was one of the first to be extensively sequenced as part of the effort to sequence the first human chromosome (47). Moreover, much of the homologous region of the mouse genome has also been sequenced (48). It is therefore particularly frustrating that no one gene has been definitively demonstrated to play any specific role in 22q11DS. Non-deletion cases of DGS and VCFS have been extensively screened for point mutations within candidate genes, but none have been found. It is theoretically possible that the causative gene remains to be discovered, but other explanations are more likely. There are several genetic and teratogenic models which to a greater or lesser extent phenocopy the malformations seen in 22q11DS. Therefore, this non-deletion patient population is likely to be aetiologically heterogeneous and it is worth remarking that no families with multiple affected members demonstrating linkage to 22q11 loci have been described. Another possibility is that the 22q11DS is caused by the combined haplo-insufficiency of more than one gene. Mutations of just one of these genes would be insufficient to recapitulate the entire syndrome and point mutation in two (or more) 22q11 genes in the same individual would, on statistical grounds, be a vanishingly rare event. In this eventuality, the best hope for human genetic approaches would be to concentrate efforts on genes implicated in mouse models and to extend the range of patients screened to those with single malformations within the 22q11DS spectrum.
In order to prioritize genes for mutation screening and further functional analysis investigators sought genes for which sequence data might give a clue as to function, or for which expression data suggested a role in branchial arch or neural crest development. Several of the genes encode proteins with similarity to proteins of known function, or contain domains giving a clue as to potential function. Space precludes a description of all genes and expressed sequence tags, but some of those that have received particular attention are discussed below and depicted on the map in Figure 2. IDD/SEZ1/LAN is the protein-encoding gene closest to the ADU breakpoint. It is predicted to encode a transmembrane protein with extracellular domains similar to LDL-binding and c-type lectin domains (49–52). GSCL encodes a homeodomain transcription factor similar to goosecoid (53). However, whereas Gsc–/– mice have craniofacial abnormalities, Gscl–/– mice are apparently normal (54–56). Both these genes, though, lie proximal to the breakpoint in patient G and therefore outside the SRDO. HIRA is predicted to be a transcriptional regulator by virtue of its sequence similarity to two yeast proteins Hir1 and Hir2. It is expressed in the developing neural tube of mouse embryos and interacts biochemically with the homeodomain of the transcription factor PAX3 (57). Mice heterozygous for Hira mutations are apparently normal and homozygotes die at embryonic day 9.5 with a wide range of malformations (P.J. Scambler, unpublished data). Antisense attenuation of Hira expression in the cardiac neural crest during chick embryogenesis results in an increased incidence of persistent truncus arteriosus (58), although this could be secondary to a non-specific effect of Hira down-regulation on cellular differentiation rather than a specific effect within neural crest cells. UFD1L, encoding a protein with sequence similarity to a yeast protein involved in ubiquitin-mediated protein degradation, was recently proposed as a candidate for 22q11DS. Although it had been identified during the positional cloning of the deletion interval (59), attention was redirected towards UFD1L when it was identified as a downstream target of the Hand2 transcription factor in Hand2–/– embryos (60). Hand2 mutants have defects of heart development including outflow tract malformations (61). Whole mount in situ hybridization indicated that Ufd1l was expressed in the pharyngeal pouches and the fourth branchial arch artery, making UFD1L an excellent candidate gene. Although one patient was suggested to have a deletion that disrupted UFD1L and the neighbouring gene CDC45L, extensive mutation screening of these two genes proved negative (62).
The difficulty in identifying which of the many genes within 22q11 were important for 22q11DS using human genetic techniques prompted some investigators to switch attention to the mouse. Comparative genetic, molecular cytogenetic and physical mapping indicated that, although gene order was not conserved between mouse and man, almost all the genes present within the 22q11 deletion region mapped to the same region of MMU16 (63–65). This region is termed the murine DiGeorge chromosomal region (musdgcr) and is depicted alongside the human map in Figure 2. Recent developments in gene targeting technology have allowed investigators to establish deletions within the musdgcr in attempt to provide an animal model for the disorder.
One approach allowed creation of a deletion in a single targeting event. Kimber et al. (66) used a targeting vector with long homology arms to create a 150 kb deletion encompassing seven genes and referred to as Δ150 (Fig. 2). Mice heterozygous for the deletion appear normally viable and fertile and have no evidence of any of the structural abnormalities associated with 22q11DS. Behavioural analysis indicates an increase in the prepulse inhibition of startle reflex. The relevance to the behavioural abnormalities seen in 22q11DS is debatable as schizophrenic patients tend to have a decreased prepulse inhibition of startle. Unsurprisingly, homozygosity for the deletion resulted in an early embryonic lethality.
A larger 1.2 Mb deletion was engineered by Lindsay et al. (67) using the Cre-loxP system. The genes Es2 (68) and Ufd1l were targeted intrans. Cre-induced recombination resulted in a deletion, Df1, on one chromosome and the corresponding duplication on the other (Fig. 2). Mice carrying either the deletion or the duplication could therefore be derived. Live-born Df1+/– mice were born at predicted Mendelian ratios, but 10% died shortly after birth. At 18.5 days of embryogenesis, 20–30% of embryos had congenital heart defects of a type often observed in 22q11DS, including interrupted aortic arch type B and tetralogy of Fallot. At 9.5 days 100% of Dfl+/– mice had a hypoplastic fourth branchial arch artery, but 1 day later only half of the mutant embryos were affected, suggesting that a capacity for recovery from a developmental delay is present in the mouse (69). The fourth arch artery remodels to form part of the arch of the aorta and therefore this apparently simple abnormality could underlie the more varied defects seen later in development. Importantly, the Dfl mutation can be complemented by the corresponding duplication demonstrating that the deficiency is secondary to hemizygosity of a gene or genes within the targeted deletion and not due to a long-range effect of the chromosome rearrangement.
The Df1 deletion overlaps proximally with the Δ150 deletion mentioned above (Fig. 2), meaning that genes deleted in common can be eliminated as being sufficient to cause the phenotype by haplo-insufficiency. As Ufd1l was the distal anchor gene of the Dfl deletion, mice heterozygous for a null mutation at this candidate were generated. As expected given the negative mutation screens mentioned above, these mice were apparently normal (67).
The Dfl1+/– embryos did not show any evidence of the non-cardiovascular defects seen in humans with 22q11 DS. This may be because mice, or at least the strain of mouse used in these experiments, are less susceptible to the effects of the haplo-insufficiency than humans. Alternatively, haplo-insufficiency of a gene lying outside the Df1 deletion might be necessary for full expression of the disorder.
The description of the partial 22q11DS phenotype in Df1+/– mice has been a major step for the field as it demonstrated that the disorder can be modelled and complemented in an experimental system. Engineering smaller deletions will allow SRDO mapping in the mouse. The recent development of a means of creating nested deletions using retroviral insertion will expedite research in this area (70). As an adjunct to these experiments, transgenic complementation has the potential to highlight regions of interest. This latter approach is feasible, if laborious, even should haplo-insufficiency of more than one gene be required to produce the Df1 phenotype. Evidently, the results of these investigations are going to have a major impact on the human genetics of 22q11DS. If the mouse genetics implicates one or two genes, mutation analysis can be concentrated on these genes and their control regions; concentration of effort should allow a much larger patient set to be screened. On the other hand, if the malformations transpire to be the result of a combined haplo-insufficiency of more than one gene then it is unlikely that point mutations would underlie non-deletion cases of DGS and VCFS. Either way, we are likely to develop a much deeper understanding of the embryological basis of 22q11DS over the next few years. In the clinical arena, the efforts to understand better the cognitive and behavioural problems could lead to some form of effective intervention techniques being available in the future. If these particular features of the syndrome are found to have a simple genetic cause (single gene haplo-insufficiency, for instance), this might have far-reaching implications for our understanding of non-syndromic psychosis.
The author’s work is supported by the British Heart Foundation and Wellcome Trust.
Tel: +44 207 905 2635; Fax: +44 207 835 0488; Email: email@example.com