Prader-Willi syndrome (PWS) and Angelman syndrome (AS) are two distinct neurological disorders that map to human chromosome 15q11–q13 and involve perturbations of imprinted gene expression. PWS is caused by a deficiency of paternal gene expression and AS is caused by a deficiency of maternal gene expression. Experiments in the last year have focused on molecular analysis of the human chromosomal region as well as the homologous region on central mouse chromosome 7. New transcripts and exons have been identified and the epigenetic status of the PWS/AS region in mice and humans has been examined. The imprinting center that is hypothesized to control the switch between the maternal and paternal epigenotypes has also been characterized in greater detail and a mouse model that deletes the homologous element demonstrates a conservation in imprinting center function between mice and humans. In addition, analysis of non-deletion AS patients has revealed that UBE3A intragenic mutations are found in a significant number of cases. However, both human patients and mouse model systems indicate that other genes may also contribute to the AS phenotype. Thus, although much has been learned in the last year, considerable information is still required before these complex syndromes are fully understood.
In mammals, both parents contribute equal genetic information to their offspring. This normal diploid complement means that most autosomal genes will be expressed from both the maternal and paternal alleles. A small group of genes defies this normal Mendelian mode of inheritance. Instead, imprinted genes are expressed from only one of the two alleles in a parent-of-origin-dependent manner. These genes are designated as imprinted since they retain the parental identity they acquired during gametogenesis. The regulation of imprinted genes is orchestrated by an epigenetic modification to DNA. As such, imprinted genes are not only susceptible to changes in the genetic sequence but also to disruptions in the epigenetic program that controls these genes.
The mechanisms controlling genomic imprinting are likely to be complex and at present are poorly understood ( 1–4 ). What is clear is that deviation from appropriate parent-of-origin-dependent expression may have dire consequences for the organism. Aberrant imprinted gene expression has now been determined to be the cause of a number of human diseases, including Prader-Willi syndrome (PWS) and Angelman syndrome (AS), emphasizing the importance of correct parental-specific expression of imprinted genes.
PWS and AS are two classic examples of imprinting in humans ( 5 , 6 ). PWS and AS result from molecular defects in 15q11–q13 that cause loss of expression of paternally transcribed and maternally transcribed genes, respectively. This review will focus on the intensive efforts of the last year to elucidate the mechanism that controls chromosome 15q11–q13 imprinted gene expression.
Genetic Etiology of PWS and AS
PWS and AS are clinically distinct neurological disorders. The major features associated with PWS are decreased fetal activity, neonatal hypotonia and feeding difficulties, hyperphagia with obesity and psychomotor and mental retardation (MIM 176270). The clinical manifestations of AS include severe cognitive impairment, seizures, ataxia and inappropriate laughter (MIM 105830). Both syndromes occur with an approximate frequency of 1 in 15 000 ( 7 , 8 ). Several molecular mechanisms have been identified that lead to PWS and AS, including large deletions, uniparental disomy, intragenic mutations, imprinting mutations and rare balanced translocations ( Table 1 ) ( 5 ). In all cases, loss of expression of at least one paternally expressed or one maternally expressed gene, respectively, at 15q11–q13 is the causative event in the development of these syndromes.
The most common molecular defect giving rise to these syndromes is a large chromosomal deletion (∼4 Mb) that includes a large cluster of imprinted genes (2–3 Mb) and a non-imprinted domain (1–2 Mb) ( 9 , 10 ). Paternal inheritance of the deletion results in PWS while maternal inheritance produces AS. In addition to the same rate of occurrence (∼70%), deletions in PWS and AS occupy the same cytogenetic position, 15q11–q13. Molecular analysis of the breakpoint ends indicates that the vast majority of deletions cluster at distinct sites ( 9 , 11 ). Two breakpoint clusters have been mapped centromeric to ZNF127 , with the more proximal breakpoint accounting for ∼65% of deletions ( 12 ). The distal breakpoint cluster has been mapped telomeric to the P locus. The inherent instability of 15q11–q13 may be attributed to an expressed low copy repeat sequence which is located in the vicinity of the breakpoint clusters ( 12 , 13 ). This sequence appears to be housed within the HERC2 gene, which encodes a gigantic HECT (homologous to E6-AP C-terminus) and RCC1 (regulator of chromatin condensation) domain protein and is located telomeric to P ( 13 ). At least seven expressed pseudogenes arising from genomic duplication of HERC2 are present in the human genome, including two copies that are located adjacent to the HERC2 locus at 15q13, three pseudogenes that were translocated to 15q11 and two pseudogenes that reside at 16p11.2 ( Fig. 1 ). Unequal crossover events between repeat sequences at 15q11 and 15q13 likely generate the large deletions observed in PWS and AS ( 14 , 15 ). Unlike the human, the mouse genome has only one copy of the Herc2 gene and no pseudogenes ( 13 , 16 ). The HERC2/Herc2 gene is unlikely to be imprinted given that it is expressed in somatic cell hybrids with a single maternal or a single paternal chromosome 15, it is located in a non-imprinted region in humans and mice and mutations in murine Herc2 are inherited as a recessive trait ( 13 , 16 ).
In addition to large chromosomal deletions, smaller micro-deletions situated upstream of the SNRPN gene have been identified ( 17 , 18 , 19 , 20 , 21 ). These localized deletions appear to disrupt the epigenetic program that regulates imprinted gene expression across 15q11–q13, defining a putative cis -acting imprinting control center (IC) ( 20 ). This means that while chromosome 15 exhibits a normal biparental mode of inheritance, AS patients have two chromosomes with a paternal identity (hypomethylation and biallelic expression of paternally expressed genes) and PWS patients have two chromosomes with a maternal identity (hypermethylated and silent paternal genes). Molecular characterization of the IC has established that this region covers ∼100 kb of genomic sequence and consists of a bipartite structure ( 17 , 19 , 21 ). Deletion of the proximal portion of the IC (25–30 kb upstream of the SNRPN promoter) results in AS ( Fig. 1 ). Recently, the AS imprinting control element has been narrowed to a region of 1.15 kb [AS shortest region of overlap (AS-SRO)] ( 22 ). This element is hypothesized to be involved in the imprinting process that establishes the maternal epigenotype of 15q11–q13 ( 23 , 24 ). In PWS, it is the distal portion of the IC that is deleted. The PWS imprinting control element spans the SNRPN promoter and exon 1 and is estimated to be >4.3 kb in size, as determined by the shortest region of overlap (PWS-SRO) of microdeletions in PWS individuals ( 25 ). This element is hypothesized to function in the germline to establish the paternal identity of 15q11–q13 by switching the grandmaternal imprint to a paternal imprint ( 20 , 24 , 26 ). These imprinting elements act to regulate imprinted expression across a domain of 2–3 Mb. Although still poorly understood, several models have been proposed for the role that the IC may play in the imprinting process ( 4 , 20 , 22–28 ).
Candidate Genes for PWS
The PWS critical region extends over nearly half of 15q11–q13 and contains multiple paternally expressed genes. To date, three protein coding genes, small nuclear ribonucleoprotein N ( SNRPN ) and its bicistronic partner SNRPN upstream reading frame ( SNURF ), necdin ( NDN ) and zinc finger protein ( ZNF127 ), and five paternally expressed transcription units, ZNF127AS, PAR5, PARSN, IPW and PAR1 , have been mapped to this interval ( Fig. 1 ) ( 5 , 10 , 29 , 30 , 31 ). In addition, several paternally expressed SNRPN upstream exons have been localized to this region and are spliced in various combinations to produce the IC transcripts ( 5 , 10 , 20 , 29 ). Murine homologs of these genes transcripts [ Zfp127, Zfp127as, Ndn, Snurf-Snrpn and Ipw ] map to the syntenic region of mouse central chromosome 7 ( 5 , 31 , 32 , 33 ).
Since the PWS critical region is so large it is likely that more than one paternally expressed gene is involved in the patho-genesis of PWS. However, it is uncertain which of these genes is involved as no intragenic mutation affecting expression of only one PWS gene has been described ( Table 1 ) and loss of expression of a single specific candidate gene has not been correlated with PWS. In fact, in this latter case data are conflicting. Some PWS patients with rare balanced translocations show loss of expression of a subset of paternally expressed genes while others exhibit normal imprinted expression of these same genes ( 34 , 35 , 36 , 37 ). Recent identification of a novel protein contained within the 5′-portion of the SNRPN gene may help to explain these data ( 31 ). In two of the four translocation patients described above the novel protein-encoding exons, termed SNURF ( SNRPN upstream reading frame), are severed while the SNRPN exons remain intact ( 35 , 37 ). This locus has been renamed SNURF-SNRPN to depict the atypical, bicistronic nature of this gene (i.e. a single SNURF-SNRPN mRNA transcript from which the proteins SNURF and SmN are translated). SNURF-SNRPN is expressed from the paternal allele and, like SmN, SNURF protein is not detected in PWS patients. Interestingly, the SNURF exons are completely contained with the 4.3 kb PWS-SRO, perhaps suggesting an important role for SNURF in the genesis of PWS. This protein is not the sole factor responsible for PWS, however, since disruption of 15q11–q13 in the other translocation patients occurs downstream of the SNURF-SNRPN locus ( 34 , 36 ). Currently, at best PWS can be characterized as a contiguous gene syndrome involving multiple paternally expressed genes. Complex mutant mice carrying targeted deletion of the various mouse homologs (see below) may identify which genes play a role in PWS.
Candidate Genes for AS
In contrast to PWS, ∼20% of AS cases are predicted to have intragenic mutations in the putative AS gene, since deletions, paternal disomy and imprinting mutations have been excluded as possible molecular defects in these patients ( Table 1 ). The E6-AP ubiquitin-protein ligase ( UBE3A ) gene has been strongly implicated as the AS gene since genomic mutations and an inversion breakpoint have been identified in UBE3A in AS patients ( 38 , 39 , 40 ). In addition, the entire 120 kb genomic UBE3A sequence is contained within the 250 kb AS critical region ( Fig. 1 ) ( 41 ). UBE3A/Ube3a exhibits tissue-specific imprinting with preferential maternal expression in sub-regions of the brain in humans and mice ( 42 , 43 , 44 , 45 ). A fair number of AS patients have now been examined for the presence of mutations in UBE3A . Only 30% of AS patients in this class had loss-of-function mutations in UBE3A ( 46 , 47 ). The remaining 70% of patients had no identifiable defect in UBE3A . While this may be explained by misdiagnosis of AS, it is also possible that additional genes or silencing elements in the AS critical region are involved in the pathogenesis of AS. Recently, additional transcripts have been detected in this region, including a 3.5 kb sense transcript whose promoter is embedded in the 3′-UTR of the UBE3A gene ( 48 ). This transcript is preferentially expressed from the maternal allele in brain. Mutations in this candidate transcript/gene could account for the remaining patients not possessing mutations in UBE3A .
In addition to the sense transcript, an antisense transcript has also been identified ( 48 ). This transcript begins ∼6.5 kb from the UBE3A stop codon, includes sequences corresponding to the sense transcript and is coincident with the 3′-half of UBE3A . The size of this transcript has yet to be determined. In brain, the antisense transcript is expressed predominately from the paternal allele. A competition model has been proposed where transcription of the antisense gene would exclude paternal allele-specific UBE3A expression ( 4 , 48 ).
Epigenetic Modification of 15q11–q13
In addition to identifying the genes responsible for PWS and AS, much effort has focused on understanding how parental identity is established for this chromosomal region. It is widely believed that the mechanism governing the imprinting of the PWS/AS domain is likely to involve parent-of-origin-specific epigenetic modification of the DNA. Studies have focused on epigenetic modifications such as allele-specific DNA methylation, replication timing and chromatin structure. In fact, allele-specific methylationatD15S63 (PW71) and D15S9 ( ZNF127 ), which are centromeric to SNRPN , has been used extensively as a diagnostic indicator of PWS and AS ( 49 , 50 , 51 ). However, the paternal allele-specific methylation at PW71 is variable and it is unclear what role it may play in the designation of parental alleles ( 52 , 53 ). While ZNF127 and human and mouse NDN are differentially methylated ( 54 , 55 ), SNRPN methylation patterns, which have been studied in the most detail, are likely to be an important part of the mechanism that controls imprinting at this locus. Both the human and mouse genes are hypermeth-ylated on the inactive maternal allele at the promoter and first exon (corresponding to the PWS-SRO) and in the 3′-portion of the gene on the active paternal allele ( 18 , 32 , 52 , 56–59 ). There is also evidence to suggest that these methylation patterns are established in the gametes, thereby representing candidate sequences for conferring the allelic imprinting mark ( 57 , 58 ).
The well-established association between regulatory elements and nuclease hypersensitivity has led investigators to use chromatin analyses to search for regulatory elements. Consistent with the methylation patterns observed throughout the SNRPN locus, the promoter and exon 1 (PWS-SRO) are hypersensitive to nucleases on the paternal allele and the sequences further downstream are nuclease hypersensitive on the maternal allele ( 60 ). While the paternal allele-specific hypersensitivity could merely reflect the transcriptionally active state of the SNRPN gene, it is also possible that this part of the IC is controlling the paternal-specific epigenotype. Interestingly, the AS-SRO is hypersensitive to nucleases on the maternal allele ( 60 ). Thus, the nuclease hypersensitivity of the PWS-SRO and AS-SRO in the IC supports the proposal that these regions serve to mediate the switching between paternal and maternal epigenotypes.
SNRPN and the PWS/AS region display other properties that are characteristic of imprinted genes. Many genes in the region harbor repetitive elements. For example, the first intron of the mouse and human SNRPN genes contains structurally conserved G-rich repeats ( 32 , 59 ). As proposed for other imprinted genes, the repeats may be involved in establishing the imprinting or DNA methylation patterns of this gene ( 61 ). Additionally, the human chromosome 15 homologs replicate asynchronously and exhibit preferential association during late S phase of the cell cycle ( 62 , 63 ). Recently the region encompassing the PWS-SRO has been shown to silence gene activity when tested in a Drosophila system ( 64 ). Although this result would be inconsistent with the putative paternal-specific activation role of the PWS-SRO in humans and mice, it may reflect the link between crucial imprinting control sequences and evolutionarily controlled silencing mechanisms ( 64 ). Alternatively, it has been proposed that this region harbors a tissue-specific element to silence expression in non-neural tissues ( 25 ).
Mouse Models of PWS and AS
The complex phenotypes of PWS and AS together with the unusual and novel nature of the IC lend importance to the development of mouse models that recapitulate the phenotypes/genotypes of human patients. Such models will allow the identification of genes responsible for each syndrome and will also help to elucidate the mechanism of imprinting at this locus. The first candidate mouse models for PWS and AS were described by Cattanach et al. ( 65 , 66 ). They used intercrosses between mice harboring translocations to derive progeny with uniparental disomy of the PWS/AS homologous region in mice. While these mice display phenotypic characteristics indicative of the two syndromes, the large region of uniparental disomy makes it difficult to assign the phenotype to individual genes.
Yang et al. have generated a mouse model for PWS and IC mutations by using homologous recombination in embryonic stem (ES) cells to engineer a deletion of the Snprn gene and the region that corresponds to the distal portion of the IC (including the PWS-SRO) ( 67 ). Chimeric males with a mutation in their maternally derived allele were not capable of reversing the maternal epigenotype of the mutant allele in their germline and, as such, the progeny inheriting this allele from the chimeric male failed to express the genes normally transcribed exclusively from the paternal allele (i.e. Snrpn, Zfp127, Ndn and Ipw ). These mice displayed some of the phenotypes characteristic of PWS. Thus, in addition to generating a mouse model of PWS, this mutation mimics human IC mutations, indicating that the position and hypothesized role of the IC are conserved between mice and humans. A second mouse model for PWS has also been described by R.D. Nicholls et al. (personal communication). In this model, insertion of a transgene resulted in a large deletion of the PWS/AS homologous region and mice that inherit the deletion from the father display characteristics of PWS.
The above two mouse models and data from PWS patients provide compelling evidence that perturbations in multiple paternally expressed genes are involved in the pathogenesis of PWS. In agreement with this proposal, mice that harbor an intragenic deletion of the Snrpn gene are phenotypically normal ( 67 ). Thus, perturbations in Snrpn gene expression alone are not sufficient to cause PWS symptoms in the mouse. To prove that more than one gene is involved, mice with mutations in multiple paternally expressed genes will have to be derived.
ES cell technology has also been used to generate mutations in two candidate genes for AS, Ube3a and the β 3 subunit of the GABA A receptor ( Gabrb3 ). Mice with maternal deficiency of the imprinted Ube3a gene display a phenotype that mimics AS, including motor dysfunction, inducible seizures and a context-dependent learning deficit ( 45 ). Although absence of an imprinted gene that is expressed exclusively from the maternal allele (i.e. UBE3A ) is the most likely AS candidate gene, the non-imprinted β 3 subunit of the GABA A receptor ( GABRB3 ) gene is located in the large deletion region of the majority of AS patients and may also contribute to the phenotype. Mice lacking the Gabrb3 gene exhibit seizures, learning and memory deficits, poor motor skills and hyperactivity, features that are common to AS ( 68 , 69 ). Additionally, heterozygous Gabrb3 mutant mice exhibit a partial phenotype, suggesting that haploinsufficiency of the GABRB3 gene could be a contributing factor in AS deletion patients. As is the case with PWS mouse models, additional mice with mutations in both Ube3a and Gabrb3 are required to determine the relative contribution of these genes to the full expression of AS.
During the last year, many research contributions have advanced our understanding of the pathogenesis of PWS and AS. Identification of the HERC2 gene and pseudogenes provided a molecular explanation for 15q11–q13 being a hotspot for recombination and thereby generating one of the most common interstitial deletions in humans. The lack of pseudogenes also explains why the same chromosomal instability is not observed in the mouse. A cursory glance at the genetic maps reveals a paucity of genes/transcripts in mouse when compared with humans. Will the novel sense and antisense Ube3a transcripts be found in mouse? Will homologs of the Parsn, Par5 and Par1 transcription units also be identified? More to the point, if the IC contains information that confers the primary imprinting mark, will sequence identity between humans and mice be revealed at the genomic level? While sequence conservation is certainly an attractive idea, no conserved cis -acting elements have yet been identified in other imprinted genes. The generation of murine mutations that mimic imprinting mutations in humans suggests functional conservation of the IC. Will the upstream Snrpn exons that compose the IC transcripts be conserved in the mouse? Alternatively, does anonymous transcription play a role in determining parental identity? Nuclease hypersensitivity studies may help to determine whether a similar underlying chromatin structure exists in the mouse. With the advent of murine PWS and AS models, these questions may be addressed, providing greater understanding of the complex molecular mechanism that governs PWS and AS imprinted gene expression.