Most cases of Angelman syndrome (AS) result from loss or inactivation of ubiquitin protein ligase 3A (UBE3A), a gene displaying maternal-specific expression in brain. Epigenetic silencing of the paternal UBE3A allele in brain appears to be mediated by a non-coding UBE3A antisense ( UBE3A-ATS ). In human, UBE3A-ATS extends ∼450 kb to UBE3A from the small nuclear ribonucleoprotein N ( SNURF/SNRPN ) promoter region that contains a cis -acting imprinting center (IC). The concept of a single large antisense transcript is difficult to reconcile with the observation that SNURF/SNRPN shows a ubiquitous pattern of expression while the more distal part of UBE3A-ATS , which overlaps UBE3A , is brain specific. To address this problem, we examined murine transcripts initiating from several alternative exons dispersed within a 500 kb region upstream of Snurf / Snrpn . Similar to Ube3a-ATS, these upstream (U) exon-containing transcripts are expressed at neuronal stages of differentiation in a cell culture model of neurogenesis. These findings suggest the novel hypothesis that brain-specific transcription of Ube3a-ATS is regulated by the U exons rather than Snurf / Snrpn exon 1 as previously suggested from human studies. In support of this hypothesis, we describe U- Ube3a-ATS transcripts where U exons are spliced to Ube3a-ATS with the exclusion of Snurf-Snrpn . We also show that the murine U exons have arisen by genomic duplication of segments that include elements of the IC, suggesting that the brain specific silencing of Ube3a is due to multiple alternatively spliced IC- Ube3a-ATS transcripts.
Received April 6, 2004; Revised May 19, 2004; Accepted June 7, 2004
Prader–Willi syndrome (PWS) and Angelman syndrome (AS) are behavioral/neurological disorders of genomic imprinting that map to the chromosome 15q11–13 and chromosome 7C regions in human and mouse, respectively. PWS is caused by the absence of a normal paternal contribution to chromosome 15q11–q13 and is characterized by hypotonia and failure to thrive in infancy, small hands and feet, hypogonadism, variable mental retardation and hyperphagia that leads to a marked obesity ( 1 ). Paternally expressed transcripts in the PWS/AS region include both protein-encoding mRNAs and non-coding RNAs. The genes encoding small ribonucleoprotein N ( SNURF/SNRPN ), the neuronal growth suppressor NECDIN ( NDN ), makorin ring finger protein 3 ( MKRN3 ) and the MAGE-like 2 protein ( MAGEL2 ) are the best-characterized protein-encoding mRNAs in the PWS/AS region. Expression of the paternal-specific genes in the PWS region is controlled by an imprinting center (PWS-IC), a region of 4.3 kb that contains a differentially methylated region (PWS-DMR) within the SNURF/SNRPN promoter ( 2 , 3 ). The PWS-DMR is characterized by a maternal-specific pattern of DNA methylation; the proposed function of the PWS-IC is to control the switch from maternal to paternal imprint during spermatogenesis ( 4 ). Small paternally inherited deletions encompassing the PWS-IC cause PWS with extensive methylation of the PWS-DMR ( 5 , 6 ).
The clinical features of AS include severe mental retardation, ‘puppet-like’ ataxic gait with jerky arm movements, seizures, EEG abnormalities, hyperactivity and bouts of inappropriate laughter ( 7 ). The AS candidate gene, UBE3A ( 8 , 9 ), displays bi-allelic expression in most tissues but is maternal-specific in brain ( 10 – 12 ). The mechanism of epigenetic silencing of UBE3A has been associated with a brain specific paternal antisense ( UBE3A-ATS ) transcript in human and mouse ( 13 – 15 ). The opposite patterns of imprinted expression of UBE3A and UBE3A-ATS led us to propose that the antisense transcript functions in cis to repress the paternal allele of UBE3A in brain ( 16 ). Such a model would be consistent with the sense–antisense epigenetic mechanisms that regulate Xist as well other imprinted genes, including Igf2r ( 17 , 18 ).
In human, UBE3A-ATS is a large (∼460 kb) transcript that initiates in the PWS-IC and extends distally through SNURF/SNRPN , IPW and overlaps UBE3A ( 19 ). The non-coding UBE3A-ATS is alternatively spliced and serves as a host for several types of small nucleolar RNA (snoRNA) of the box C/D class that are contained within the introns and are expressed upon processing of the paternal copy of the host transcript ( 20 – 22 ). Here, we present evidence that the murine Ube3a-ATS is also a large (∼1000 kb) transcript that is alternatively processed. By following the in vitro neuronal differentiation of P19 embryonic carcinoma (EC) cells, we confirm that the distal region of Ube3a-ATS , that is host to the class MBII-52 snoRNAs and overlaps Ube3a , is detected during the initial stages of neurogenesis. In contrast, Snurf / Snrpn is expressed in undifferentiated precursor cells and throughout the course of differentiation. In order to understand the difference in temporal expression between the proximal and distal parts of Ube3a-ATS , we examined the transcription and organization of exons upstream (U) of Snurf / Snrpn that display paternal-specific expression in mouse ( 23 ). The U exon-containing transcripts display the same time course of expression as Ube3a-ATS during P19 neuronal differentiation and are brain specific and paternally expressed in mouse. There are at least 9 U exons in mouse and more than 30 highly homologous copies in rat. As is the case in human, the U exons appear to have arisen from genomic duplication of segments that include elements of the IC. These findings suggest that the U exons serve as initiation sites for IC- Ipw / Ube3a-ATS transcripts in brain.
MATERIALS AND METHODS
U-Snurf/Snrpn-Ipw-Ube3a-ATS genomic organization and alternative transcript expression
For RT–PCR analysis of Ube3a-ATS , total RNA was isolated from brain of adult wild-type C57 Bl/6J mice, of the PWS-IC deletion mouse ( 24 ) and wild-type littermate by using RNAzol (Tel-Test, Inc. Friendswood, TX) or TRIzol (Life Technologies, Gaithersburg, MD) according to the manufacturers' instructions. Total RNA was isolated from P19 cells at different time points of RA-induced neuronal differentiation by the cesium chloride gradient method. Reverse transcription reactions (+RT) were performed with 2.5 μg of RNA and Superscript reverse transcriptase (Life Technologies) using either random hexamer primers or the following antisense strand-specific primers: M-1F: 5′-TTTCACTTGATCCTTTTACAAACAAC-3′; M-2F: 5′-GCCTTCATAGTGTAATAGCAG-3′; M-3F: 5′-ATAGACAATCCAAAGTCTCCCC-3′; M-4F: 5′-GCACCTGTTGGAGGACTAGG-3′ and E-1F: 5′-CTGTTATTTCCTTTTGTGCACTG-3′. A 2.5 μg aliquot of total RNA was incubated in a similar manner but without reverse transcriptase (−RT) as a control. One-fifteenth of the +RT or −RT reaction was used in PCR reactions using either Advantage 2 Taq polymerase (Clontech) or Taq polymerase (Invitrogen) under conditions suggested by the manufacturer. Specific annealing temperatures and extension times are available upon request. Oligonucleotides for PCR were G1-1R (5′-TTCATGCTCACTAGGATTACGC-3′); G1-2R (5′-CTGGTGTTTTGCAATCATCTGAC-3′); L-1R (5′-ATCGAAAAAACAAGCTATCCAATC-3′); B-1R (5′-CATGGGCCATGAGTGACATCC-3′); M-5F (5′-TTTGGATTACTCATTGCTACATC-3′); M-6F (5′-TGGAATTACTATCCTTTGAGAGG-3′); E-2F (5′-CACTGTGCAATTTCTTTACAGTG-3′); 36B4-F (5′-TCTTCCAGGCTTTGGGCATC-3′); 36B4-R (5′-CCTCCGACTCTTCCTTTGCTTC-3′); 11-C (5′-AACTGAGGGTCAGTTTACTCT-3′); Ipw A2 (5′-TCTGCATGTTTTCCATGGACC-3′); 12-B (5′-AGATCATACATCATTGGGTTA-3′); Ipw BC-F (5′-TCACCACAACACTGGACAAAAG-3′); Ipw BC-R (5′-TGCTGCTACACAGGAAAGAGG-3′); U7a (5′-GGTCCTGCTCAGAGAAGAGC-3′); U7b (5′-CAGCAAGGCAATACAGCAAAT-3′); Snrpn Ex1-F (5′-AGAGGGAGCCGGAGATGCC-3′); Delta (5′-GCCTGTCGGCCGTGGAGCA-3′); Atp10a-F (5′-AAAGTATCAAGTCCCACCGGC-3′); Atp10a-R (5′-AATCAAAAACATCAGACAGCTCG-3′); U5 (5′-GAGGCAGGCAAACATGCCTCT-3′). The oligonucleotide primers and PCR conditions for Nscl-2 , NeuroD ( 25 ), Snrpn exon 7 ( 23 ), Snrpn exons 1–3 ( 26 ), Snrpn exons U1-3 and exons U2-3 ( 23 ), Gapdh ( 24 ) and Hprt ( 27 ) have been described previously.
For expression of MBII52 and MBII85 during P19 EC cell RA-induced neuronal differentiation, 5 μg of total RNA was size-fractionated on a 1.0% agarose/formaldehyde gel and transferred to a BioTrans nylon membrane (ICN Biochemicals, Irvine, CA). Hybridization was performed in ExpressHyb Hybridization Solution (Clontech, Palo Alto, CA) at 55°C with previously described MBII52 and MBII85 radiolabeled probes ( 20 ) and an Actin control probe furnished by the manufacturer. Washes were carried out according to the manufacturer's instructions. Membranes were exposed to Kodak BIOMAX film and autoradiographed for 3 days.
Sequencing and sequence analysis
Nucleotide sequencing was performed by the University of Connecticut Health Center Molecular Core Facility. Sequence data were processed and assembled using the Sequencher (Gene Codes, Ann Arbor, MI) program, version 4.1.2 and MacDNASIS (Hitachi Software, San Francisco, CA) version 3.7. Nucleotide sequence comparisons were performed using the BLAST ( 28 ) engines available at www.ncbi.nlm.nih.gov/blast/ . Sequence alignments were performed using the Clustal algorithm in the MacVector 7.2 (Accelrys Inc, San Diego, CA) analysis package.
Differentiation of P19 cells
P19 EC cells were obtained from ATCC (Manassas, VA). P19 cells were maintained in Dulbecco's modified Eagle medium (DMEM, Invitrogen, Carlsbad, CA) supplemented with 10% fetal bovine serum (Hyclone, Logan, UT). For induction of neuronal differentiation ( 29 , 30 ), the cells were grown at a density of 10 6 per 100 mm bacteriological petri dish in 10 ml medium containing 5 × 10 −7 M all- trans retinoic acid (T-RA. Calbiochem, La Jolla, CA) to promote aggregate formation. After 48 h, the aggregates were collected, partially disrupted by trypsinization, diluted 1:2 and plated again on bacteriological dishes in medium with RA. The aggregates were again trypsinized 48 h later and plated in medium without RA on tissue culture plates pre-coated with gelatin. Cytosine-B- d -arabinofuranoside at a concentration of 15 μg/ml (ara-C, Sigma Chemical Co., St Louis, MO) was added 24 h later to eliminate the cells that were still dividing and had not differentiated. At 5 days after addition of RA, the culture medium was changed to 50% Neurobasal media with B27 supplement, 50% DMEM ( 29 ).
Organization, processing and allele-specific expression of Ube3a-ATS
Multiple brain specific Ube3a-ATS transcripts are detected in the Ube3a to Snrpn interval ( http://genome.ucsc.edu ). Other antisense mRNAs are also known to map to this interval but the presence of a large gap in the sequence contig precludes a precise localization of the additional transcripts. The gap between Ube3a and Snrpn also prevents a display of the complete organization of Ipw . The putative Ube3a-ATS transcripts are all derived from brain cDNAs except for some of those covering the 5′ end of Snrpn . The Ube3a-ATS cDNAs span the entire Ube3a coding region with several mapping to beyond the 3′-untranslated region (3′-UTR) of Ube3a . In order to determine whether these transcripts are part of the antisense, we have analyzed the region of Ube3a-ATS that overlaps the Ube3a 3′-UTR and extends further downstream by strand-specific RT–PCR using a number of primer pair combinations. This analysis revealed that Ube3a-ATS transcription occurs throughout the ∼10 kb Ube3a 3′ end ( Figure 1 and Table 1 ).
| Exon || Nucleotide position a|| 3′ splice || Exon size (bp) || 5′ splice || Intron size (kb) |
| Exon || Nucleotide position a|| 3′ splice || Exon size (bp) || 5′ splice || Intron size (kb) |
The nucleotide sequences of novel Ube3a-ATS RT–PCR products were compared to genomic database sequence to determine the splice consensus sites, exons and introns.
GenBank accession no. AC027299.25.
EST accession no. AK049008.
The RT–PCR analysis resulted in the identification of 4.3 and 5.6 kb amplification products ( Figure 1 , splice forms I and II) that contained 8 exons, distributed over 35 kb of genomic DNA ( Figure 1 , Table 1 ). Two exons (designated G1a and G1b) shared >80% nucleotide homology with the bipartite Ipw exon G1 ( 20 , 31 ) ( Table 1 , data not shown) while a third showed a similar level of identity to exon G2 ( Table 1 , data not shown). The difference in size between the 4.3 and 5.6 kb cDNAs obtained by RT–PCR (splice forms I and II, Figure 1 ) was due to the presence of an unprocessed MBII-52G1a cassette in the 5.6 kb splice form II cDNA. The other exons in the 4.3 and 5.6 kb cDNAs were designated Ipw exons I, J, K and L ( Figure 1 ). To confirm that these exons are part of Ube3a-ATS , additional RT–PCRs were performed using different primer pair combinations ( Figure 1 , splice forms III–V).
We were also able to connect exon L to the antisense that overlaps the Ube3a 3′ end ( Figure 1 , 1.3 kb cDNA, splice form VI). A brain specific Ube3a-ATS splice form that contains exon L was also identified by a database search (GenBank accession no. AK049008). Ipw was also connected to Ube3a-ATS by performing RT–PCR using an RT oligo, M-2F, from the region overlapping the Ube3a 3′ end ( Figure 1 ) and PCR primers specific for the upstream exons B and E ( Figure 1 , splice form VII) indicating that an upstream fragment of Ipw can be amplified by priming cDNA from the Ube3a 3′ end. We also attempted to link the region containing the novel Ipw exons to upstream Ipw exons B, C and E but we could not do so. The most likely explanation for this difficulty is that such transcripts potentially contain multiple copies of the bipartite exon G and cannot be amplified by RT–PCR due to their size and/or complex secondary structure. While the total number of G2-MBII-52-G1 cassettes in mouse cannot be determined because of the gaps in the Ipw to Ube3a contig, there are 47 HBII-52 containing cassettes in the human UBE3A-ATS ( 32 ).
Additional support for the hypothesis that Ipw and Ube3a-ATS are part of the same transcript comes from the database retrieval of a cDNA clone (GenBank accession no. CB616315) from rat hypothalamus ( Figure 1 ). The 5′ end of CB616315 contains six copies of Ipw G-like exons. Comparison of the genomic rat sequence to that of CB616315 reveals that the genomic structure is similar to that in mouse with G-like exons flanking introns that contain RBII-52 snoRNAs ( Figure 2A ). As observed between mouse and human ( 20 ), the rat and mouse BII-52 snoRNAs are almost 100% identical ( Figure 2A ). Although the size and organization of the rat RBII-52-containing cassettes ( Figure 2A ) are strikingly similar to that in mouse ( 20 ), the mouse and rat G exons are only ∼60% identical at nucleotide sequence level (data not shown). Sequence comparisons failed to identify Ipw exons B and E in the rat genome although exon C is highly conserved between mouse and rat ( Figure 2B ).
The findings summarized above establish that the murine Ipw and Ube3a-ATS are part of the same transcript. When combined with the report that Snurf/Snrpn and Ipw are also a continuous transcript ( 33 ), the results suggest that, as previously postulated for human ( 19 ), Ube3a-ATS is a large transcript that extends from Snurf / Snrpn to Ube3a .
Expression of Ipw/Ube3a-ATS and snoRNAs during in vitro neurogenesis
The expression levels of several transcripts from the AS/PWS region ( Figure 3 ) were examined in P19 EC cells during the course of in vitro neurogenesis ( 29 , 30 ). Consistent with previous results ( 25 , 34 ), the neurogenic basic helix–loop–helix (bHLH) transcription factors, Nscl2 and NeuroD , are expressed transiently during P19 differentiation, with the peak levels of steady-state RNA observed between days 3 and 5 ( Figure 3A ). These temporal expression patterns for Nscl2 and NeuroD are consistent with the expected course of P19 neuronal differentiation ( 25 , 34 ). Ube3a-ATS transcription was assayed by RT–PCR using primers from Ipw exon L and from the region of overlap with the Ube3a 3′ end ( Figure 1 ) and the transcript was detected at day 5 and with more marked expression at the later times of differentiation ( Figure 3B ). Similar results are obtained for a transcript containing Ipw exons B and E that correspond to the more proximal part of Ube3a-ATS ( Figure 3B ). Snurf , Snrpn , Ube3a and Atp10a , another protein-encoding transcript from the AS/PWS region ( 35 , 36 ), are expressed throughout the time course of differentiation ( Figure 3 ). Gapdh and the 36B4 ribosomal gene display a relatively constant level of expression during P19 neurogenesis ( 34 ) and serve as internal controls. The data summarized in Figure 3 are consistent with the brain specific expression of the Ipw / Ube3a-ATS since expression of Ipw and Ube3a-ATS is first detectable at day 5 and up-regulated by day 7 of RA-induction, the time where P19 cells display neuronal-specific β tubulin J and neurite extensions (data not shown) ( 30 , 37 ).
Processing of the Ipw / Ube3a-ATS host transcript in differentiating P19 cells should also result in the co-ordinate expression of the MBII-85 and MBII-52 classes of snoRNAs ( 19 – 22 ). In order to test expression of both these classes of snoRNA in P19 cells, northern blot hybridization was performed ( Figure 4 ) using RNA samples from the P19 time course experiment in Figure 3 . The results of Figure 4 show that, similar to the Ipw / Ube3a-ATS , MBII-52 snoRNA is first detectable at days 5–7 and increases markedly by day 9 of P19 neuronal induction, consistent with the previous report that MBII-52 displays brain-specific expression ( 20 ). Both classes of snoRNAs are highly expressed in total mouse brain ( Figure 4 ). The faint MBII-85 expression at the earlier stages as well as in the undifferentiated cells ( Figure 4 ) most likely represents a basal level of expression in P19 cells since we observe low-level expression of other neuronal-specific genes in undifferentiated P19 cells (D.L.B. and M.L., unpublished data).
The markedly increased levels of Ube3a-ATS during neuronal differentiation occur in the absence of any significant augmentation of Snrpn and/or Snurf expression ( Figure 3 ). While this observation is difficult to reconcile with the finding that Snrpn and Ube3a-ATS are part of the same transcript, it could suggest that there are neuronal-specific elements that regulate the expression of Snurf / Snrpn / Ipw / Ube3a-ATS . This led us to examine the organization and transcription of the exons upstream of Snrpn, including U2 and U7, that, similar to Ube3a-ATS , are up-regulated during the neuronal stages of P19 differentiation ( Figure 3B and C ).
Organization and expression of the U exons
Transcripts containing exons U1 and U2, that lie upstream of the PWS-IC, splice to Snurf / Snrpn exon 2 and are paternal- and brain-specific ( 23 ). The availability of genomic sequence from the region upstream of Snrpn has allowed the in silico identification of nine additional U-like sequences ( Figure 5A ). The sequences of exons U1–U9 are highly homologous, with a conserved splice donor site but with no apparent splice acceptor consensus sequence ( Figure 5A and Table 2 ). Two other U-like sequences were also identified that had lost the consensus splice donor sites, do not appear to be transcribed and are presumed to be pseudo U exons ( Table 2 ) suggesting that these could represent the 5′ end of individual transcripts. The U1–U9 consensus sequence was used to perform homology searches of the rat genome, resulting in the identification of 36 highly homologous candidate U exons ( Figure 5B ). The consensus sequence derived for the 36 rat sequences shows >90% identity to the mouse U exon consensus sequence ( Figure 5A ). The neighbor joining analysis of the 36 copies of the rat U-like exons suggests that the murine U exons have arisen from multiple duplication events ( Figure 5B ). The U exons also show significant (∼70%) homology to Snurf / Snrpn exon 1 ( Figure 5A and C ) suggesting that conserved elements of the PWS-IC are duplicated in the upstream region.
|Consensus sequence||Donor splice site|
| C 35|| A 62|| G 77|| g 100|| t 100|| a 60|| a 74|| g 84|
|Consensus sequence||Donor splice site|
| C 35|| A 62|| G 77|| g 100|| t 100|| a 60|| a 74|| g 84|
The numbers given for the consensus sequence nucleotides specify the frequency of occurrence (%). Upper and lower case letters represent exonic and intronic nucleotides, respectively. Highly conserved nucleotides are shown in bold. The exons designated pseudo-Ua and-Ub, despite their homology to U exons, do not have splice donor consensus sequences and are unlikely to serve as transcription initiation sites.
The organization of the mouse U exons is summarized in Figure 6A . As noted previously, transcripts containing the U1 and U2 exons splice to Snurf / Snrpn exon 2 ( 23 ). Transcripts containing exons U5 and U7 also splice to Snurf / Snrpn ( Figure 6A ). The additional exons designated gamma and delta are incorporated into the U- Snurf / Snrpn transcripts ( Figure 6A ). The expression of these novel U- Snurf / Snrpn transcripts was assayed in several different tissues and found to be exclusively present in brain ( Figure 6B ). We were able to determine the pattern of allele-specific expression of the novel U5- Snurf / Snrpn variant. A single nucleotide polymorphism in exon U5 was identified between Mus spretus and C57Bl6/J and used to design a C57Bl/6J allele-specific oligonucleotide. RT–PCR performed with this primer in a two-generation cross clearly shows that the U5 exon-containing transcripts are expressed exclusively from the paternal allele ( Figure 6C ).
Additional expression analyses using brain RNA and database searches allowed the identification of additional novel transcripts and ESTs that contain exons U2, U5, U7 and U8 ( Figure 7 ). These transcripts use additional alternative exons (beta, gamma, etc.) and splice to Snurf / Snrpn . U- Snurf / Snrpn transcripts from rat brain are also evident upon analysis of EST databases. These findings indicate that multiple brain specific U exons display a complex pattern of splicing to Snurf / Snrpn .
Identification and regulation of U exon- Ube3a-ATS transcripts by the PWS-IC
It was previously reported that, upon paternal transmission, the 35 kb PWS-IC deletion (ΔPWS-IC) abrogates the transcription of a Ube3a-ATS fragment overlapping Ube3a intron 8 ( 14 ). As shown in Figure 9 , Ube3a-ATS spanning exon L and the Ube3a 3′-UTR is also under the control of the PWS-IC. We performed an RT–PCR using brain RNA and oligonucleotide primers in exon U1 and a reverse primer in Ipw exon A2 ( 31 ) ( Figures 8 and 9 ). A product of 829 bp was detected in the brain of the normal littermate but not in that of the ΔPWS-IC. By nucleotide sequencing, the 829 bp cDNA consists of exon U1 and 6 copies of Ipw exon A2. These results not only indicate that the U1- Ipw A2 transcript is under the control of the PWS-IC but, also, that Ube3a-ATS transcripts initiating in a U exon can splice directly to Ube3a-ATS and exclude Snurf / Snrpn . In this regard, an EST containing exon U8 (BB625787; Figure 7 ) corresponds to nucleotide sequence from the 5′ end of a cDNA clone derived from diencephalon (GenBank accession no. 9330122E19). The 3′ EST of this cDNA clone (BB076454) maps to the interval beyond the Snrpn 3′-UTR, indicating that exon U8-containing transcripts also extend to beyond Snurf / Snrpn .
Deciphering the organization of the murine Ube3a-ATS as well as the mechanism of its brain-specific regulation are crucial to developing mouse models of the imprinting regulation of Ube3a . Such mouse models are particularly needed to define the structure and function of the IC in both mouse and human. The previous characterization of Ube3a-ATS was limited to a small region that overlaps Ube3a ( 14 ). Here, we demonstrate that, similar to the case in human ( 19 ), the Ube3a-ATS is part of a larger transcript that includes Ipw . In contrast to what has been reported in human, we present evidence that, in brain, Ube3a-ATS initiates in exons that are distributed throughout a 500 kb region upstream of Snurf / Snrpn , resulting in an antisense transcript that extends over a 1000 kb genomic region.
Alternative transcripts extending from Ipw exons G1 and G2 to the region of overlap with the Ube3a 3′ end ( Figures 1 and 2 ) indicate that Ipw and Ube3a-ATS are part of the same transcript. The extensive nucleotide homology of the BII-52 snoRNAs and of Ipw exon C ( Figure 2 ) confirms the conservation of the Ipw transcript between mouse and rat. The available sequence from the 3′ extent of the rat CB616315 EST overlaps the putative Ube3a exon 1 region suggesting that Ipw / Ube3a-ATS transcripts can extend to the Ube3a promoter. Ube3a-ATS transcripts that overlap the Ube3a promoter (see also AK038522, Figure 1 ) could function to regulate Ube3a by altering promoter activity or by other mechanisms that involve extensive sense–antisense overlap ( 38 ).
Our data also convincingly show that there is extensive processing of the Ipw / Ube3a-ATS in brain ( Figure 1 and 2 ), a finding consistent with northern blot hybridization analysis of Ipw that yields a smear extending from 4 kb to at least 10 kb in size ( 39 ). Processing of the Ipw/Ube3a-ATS and expression of the MBII-52 and MBII-85 snoRNAs are detectable in the P19 in vitro model of neurogenesis. cDNAs containing exons B, C, and E and L of Ipw/Ube3a-ATS transcript as well as MBII-52 snoRNAs are first detectable at days 5–7 after induction of the P19 cells with retinoic acid ( Figures 3 and 4 ). This co-ordinate temporal pattern of expression during in vitro neurogenesis is also consistent with Ipw and Ube3a-ATS being part of the same transcript. The P19 model system is quite useful for comparing the expression pattern of different transcripts from the AS/PWS region as a function of time after induction of neurogenesis. In particular, it is evident that Snurf / Snrpn shows no significant change in temporal expression pattern while Ipw , MBII-52 snoRNAs and Ube3a-ATS are detectable only by days 5–7 of P19 induction ( Figures 4 and 5 ). These findings are consistent with the observation that expression of Ube3a-ATS is predominantly brain specific while Snurf / Snrpn is expressed in a wide range of tissues ( Figure 6B ).
The observation that the U2- and U7- Snurf / Snrpn transcripts show a time course of expression in P19 cells that is very similar to that of Ipw / Ube3a-ATS ( Figure 3 ) raises the possibility that the U exons could function in the initiation of brain-specific Snurf / Snrpn / Ipw / Ube3a-ATS transcripts. Consistent with a potential function in transcript initiation, the U exons have highly conserved consensus splice donor site but no consensus splice acceptor site ( Figure 6A , Table 2 ). In order to directly test our hypothesis, we performed an RT–PCR that connected exon U1 to Ipw exon A2 ( Figure 8 ). The identification of this transcript, that splices out Snurf / Snrpn , demonstrates that brain-specific Ube3a-ATS transcripts can initiate from the U exons rather than from exon 1 as previously shown for the human UBE3A-ATS . A detailed in situ hybridization analysis of the U exon-containing transcripts and Ube3a-ATS confirm that there is extensive overlap in their tissue-specific and developmental expression, further supporting our hypothesis that Ube3a-ATS initiates from the U exons in brain (E. Le Meur and F. Muscatelli., manuscript in preparation). Our results argue for the existence of two classes of antisense transcript ( Figure 9 ). Snurf/Snrpn is included in one class (Un- Snurf / Snrpn - Ube3a-ATS ) but excluded from the second (Un- Ube3a-ATS ).
We also demonstrate that deletion of the PWS-IC in brain abolishes expression of U1- Ipw A2 ( Figure 8 ). Since the U1-Ipw A2 transcript splices out the 35 kb region that is deleted in the ΔPWS-IC ( 14 , 24 ) ( Figure 7 ), this result demonstrates that U1– Ube3a-ATS transcripts are under the control of the IC. A smaller (4.8 kb) deletion encompassing the putative murine IC did not, however, affect the expression of a U1- Snurf/Snrpn transcript ( 23 ). The simplest interpretation of these results is that elements lying within the 35 kb but outside 4.8 kb PWS-IC deletions are required for establishing the paternal expression pattern of U exon-containing transcripts.
The organization and transcription of U- Snurf / Snrpn transcripts is also similar between mouse and rat ( Figure 7 ) and there is significant sequence similarity between Snurf / Snrpn exon 1 and the U exons ( Figure 5C ). The region of homology includes three sequence elements of Snurf / Snrpn exon 1 that have been implicated as being required for the function of the bipartite IC ( 40 ). This is the first evidence to suggest that, as is the case in human ( 41 ), the murine IC has undergone multiple duplication events. In human, alternative transcripts initiate from exons (u) that are dispersed throughout the region upstream of SNURF/SNRPN although no UBE3A-ATS transcripts containing u exons have been reported ( 41 , 42 ). Molecular analysis of the micro-deletion in a subset of AS imprinting defect patients has allowed the delineation of the AS smallest region of overlap (AS-SRO) to a single 880 bp exon located 35–40 kb upstream of the PWS-IC ( 41 ). It has been proposed that the PWS-IC and AS-SRO represent the functional elements of a bipartite IC ( 43 , 44 ) and that the maternal AS-SRO functions in the methylation of PWS-IC ( 45 ). We speculate that the duplicated U exons contain functional elements of the murine ortholog of the AS-SRO. The disruption of U exons should allow the production of mouse models that will be crucial for characterizing the structure and function of the IC and for studying the mechanism of imprinting of Ube3a .
The authors are grateful to Drs M. McBurney and B. Howell for providing reagents and advice on the P19 cells. This paper is dedicated to the memory of Dr Camilynn Brannan. This work was supported by NIH grants R01 NS30628 (M.L.) and F31 NS42490 (D.B.).
Department of Genetics and Developmental Biology, University of Connecticut Health Center, Farmington, CT 06030, USA, 1NMDA/IBDM, Campus de Luminy Case 907 13288 Marseille Cedex 09, France, 2Génétique moléculaire murine CNRS URA 1947, Institut Pasteur, 25 rue du Docteur Roux, Paris Cedex 15, 75724, France and 3Department of Molecular Genetics and Microbiology, University of Florida College of Medicine, Box 100266, Gainesville, FL 32610, USA