Identification of Four Highly Conserved Genes between Breakpoint Hotspots BP1 and BP2 of the Prader-Willi/Angelman Syndromes Deletion Region That Have Undergone Evolutionary Transposition Mediated by Flanking Duplicons

Isolation of BAC Clones and STS Mapping of BAC and YAC Clones

The CYFIP1 (SRA1) gene was identified by an in silico search of National Center for Biotechnology Information (NCBI) GeneMap ‘98. Of all EST clusters found in proximal 15q, the cluster containing the 4.4-kb cDNA KIAA0068 was the most robust, with numerous ESTs in different tissues, and was subsequently identified as SRA1/CYFIP1 (Kobayashi et al. 1998; Schenck et al. 2001). GeneMap ’98 presented compiled mapping data from two radiation hybrid panels, GB4 and G3; however, because the relative order of markers was not consistent between these panels, genes were further mapped, by PCR, to BAC and YAC contigs (see below). CITD BAC 3242E18 was identified by a BLAST search of the GenBank Genome Survey Sequence (GSS) database using CYFIP1 cDNA sequence (GenBank accession number D38549). Exons 7 and 8 are present in the 3242E18 end sequence (AQ203319) in a 5′→3′ orientation compared with the end sequence. Similarly, partially sequenced BAC clones RP11-26F2 (GenBank accession number AC011767) and RP11-289D12 (GenBank accession number AC090764) were identified by BLAST with CYFIP1 cDNA sequence and STS markers (Christian et al. 1999) D15S18, located in CYFIP1 intron 1 close to exon 2, and microsatellite markers D15S1035, D15S541, and D15S542 that are clustered within CYFIP1 intron 1 ∼15 kb from exon 1.

For murine Cyfip1, which was originally identified as Shyc (Köster et al. 1998), primers RN1102 and RN1103 were used to screen a Genome Systems mouse BAC library (129/Sv) by PCR, using 35 cycles at 94°C for 2 min, 55°C for 30 s, and 72°C for 1 min, with a 10-min final extension at 72°C. Mouse RPCI-22 BAC 252P22, which was identified by the PCR screen, was end sequenced, allowing a PCR product (amplified by RN1235 and RN1236) from one end of this BAC to be used to further screen the RPCI-22 BAC library by hybridization. This approach identified BACs 508F17, 402D17, 373J21, and 526E9, and BACs 466F3 and 304H15 were identified with a PCR probe from the 195C6 T7 BAC end (table 1). BLAST searches of GSS, with the use of cDNA sequences of Cyfip1, Nipa1, Gcp5, Herc2, and p, were conducted, identifying BACs RPCI-23 (C57BL/6) 88I18, 452K16, 363F10, 471F18, 335K21, 49E2, 195C6, 396N13, 256L9, and 181P23 and RPCI-24 247O4, 150F2, 354P8, and 265E12. DNA from all BACs was isolated according to the minipreparation protocol recommended by Research Genetics. To establish BAC contigs and to determine gene order for the mouse loci, several pairs of primers were designed from gene exons or BAC ends (table 1). PCR was conducted under conditions of 94°C for 2 min, 31 cycles of 94°C for 30 s, 60°C for 30 s, and 72°C for 30 s and a 10-min final extension at 72°C.

DNA from YAC clones spanning the BP1–BP2 region was isolated as described elsewhere (Amos-Landgraf et al. 1999). STS primers from single exons of the NIPA1, GCP5, and CYFIP1 genes (table 1) were used for PCR typing of these YACs, using 10 ng DNA for each PCR. PCR used the following cycling conditions: 94°C for 2 min, then 35 cycles of 94°C for 30 s, 55°C for 30 s, 72°C for 30 s, and a final extension for 10 min at 72°C.

FISH

For human studies, metaphase chromosome preparations were made from lymphoblastoid cell lines derived from class I and class II AS deletion patients (I: WJK29, WJK43; II: WJK18, WJK35, WJK48; Knoll et al. 1990), as well as normal individuals, according to standard protocols. BAC 3242E18 was used as the probe for the human chromosome FISH studies. BAC DNA was labeled with biotin or digoxigenin, by nick translation. Chromosomal hybridizations were detected as described elsewhere (Knoll and Lichter 1994; Amos-Landgraf et al. 1999). Ten to twenty metaphase cells were analyzed per cell line. To identify the chromosomal location of murine Cyfip1, BAC 252P22 was digoxigenin labeled and hybridized to murine metaphase chromosomes derived from Tg^PWS(del) splenocytes, along with a mouse chromosome 7 centromere probe (kindly provided by A. G. Matera), which also hybridizes to the telomere of chromosome 5, labeled with biotin, as described elsewhere (Gabriel et al. 1999). Probes were detected by fluoroscein-labeled antidigoxigenin and Texas red avidin, with DAPI (4′,6-diamino-2-phenylindole) counterstaining.

DNA-replication–timing studies were performed as described elsewhere (Greally et al. 1998). A mouse splenocyte culture was pulsed with bromodeoxyuridine (BrdU, 100 μM) 90 min prior to harvesting. The cells were then harvested, fixed with methanol–acetic acid and spread onto slides (Henegariu et al. 2001). The BAC probes (452K16 and 252P22) were labeled by nick translation using digoxigenin, Texas red, and biotin. FISH was performed by denaturing the slides, using 0.05N NaOH at room temperature for 4 min, followed by suppressive hybridization using probes preannealed with mouse Cot-1 DNA. The haptens were detected using antidigoxigenin-rhodamine and avidin-CY5, while simultaneously detecting S-phase cells with anti-BrdU-FITC (Pharmingen), followed by counterstaining with DAPI. Multiple image planes were acquired for each cell, to ensure that all signals were identified. Replication patterns in S-phase cells were assigned as prereplicative if a single signal was seen at a locus and were assigned as postreplicative if more than one signal was evident. The replication patterns for 50 nuclei were analyzed for each probe. For the parent of origin-specific patterns at the Nipa1-Nipa2-Cyfip1 locus, simultaneous probing with RPCI-22 BAC 434N7 was performed with a different fluorophore on c^32DSD mice with a Tyr deletion on their paternal chromosome (Rinchik et al. 1993). The 434N7 signal from the maternal chromosome colocalized with one of the Cyfip1 homologues in S-phase cells, allowing the confident assignment of parental origin in >90% of cells. The few cells in which the Tyr signal was not clearly associated with one chromosome were not included in our analyses.

Identification of NIPA1 and NIPA2 cDNAs

BLAST search of the GenBank GSS database, using Cyfip1 cDNA sequence (GenBank accession number AF072697), matched sequence at one end of a mouse 10-kb plasmid 1M0204M02F (GenBank accession number AZ424789). The other end of this plasmid had a segment 99% identical to the Mus musculus hypothetical protein MNCb-2146 (renamed Nipa2) mRNA (GenBank accession number NM_019997; 1,867-nt sequence), 85% to several human cDNA clones (GenBank accession numbers HSU90904, BC011775, and BC000957), and 83% to a Gallus gallus cDNA (GenBank accession number AL588904) in the nonredundant GenBank database. A full-length human NIPA2 cDNA (2,421 bp) spanning exons 1–7 was then assembled in silico on the basis of these cDNAs and partial genomic sequences from BACs 26F2 and 289D12 (see above). The cDNA sequences included one clone with exons 1 and 3–7 (2247-bp, BC011775), and one 5′ clone with exons 1, 2, 2b, and 3 (GenBank accession number BF203654). Exon 2b is a rare alternatively spliced exon in the 5′ UTR of human NIPA2, but it is not conserved and is likely nonfunctional. Subsequent BLAST searches confirmed numerous mouse and human ESTs with the predominant structure including sequence spanning seven exons. NIPA2 is oriented tail to tail with CYFIP1, with 1,747 nt between the two genes. A Gallus gallus Nipa2 cDNA clone (GenBank AL588904) from a chicken brain library was obtained from the Roslin Institute, United Kingdom, and was sequenced completely through the DNA Sequence Facility at the University of Pennsylvania School of Medicine. The 5′ end was from a chicken EST (GenBank accession number AJ452290). BLAST searches using human and mouse Nipa2 coding sequence in the Fugu Project Database identified five exons spanning the conserved ORF sequence of Fugu Nipa2 (FT:T000059 Scaffold 59; these data were generously provided by the Fugu Genome Consortium for use in this publication only).

Human NIPA1 was identified by in silico methods, beginning with identification, within partial BAC 26F2 genome sequence, of a second strong CpG island not associated with the 5′ end of NIPA2. This unique sequence matched with three human ESTs (GenBank accession numbers AW962805, AA355571, and D81972), which identify an exon-intron structure by comparison with BAC 26F2 sequence, suggesting the presence of a unique gene that further analyses identified as NIPA1. The AW962805 EST sequence covers part of exons 1–4 of NIPA1, as well as an exon termed 3b. Exon 3b is alternatively spliced (see below), but it has an in-frame stop codon, is not present in other EST clones, occurs at lower levels in RT-PCR products than does product with exons 2–4 (see below), and is not conserved in the mouse. These facts led us to conclude that exon 3b is nonfunctional. BLAST search of the GenBank EST database also identified an initial fourth human EST (GenBank accession number BE003132), which spans part of exons 4 and 5 and the 3′ UTR of NIPA1, as well as three putatively orthologous mouse ESTs (GenBank accession numbers BF118592, BE654018, and BF539166). Genome sequence within BAC 26F2 that lay 3′ of the NIPA1 ORF was then used, combined with BLAST searches of the EST database, to identify numerous cDNA clones containing the two 3′ ends of NIPA1 predicted on the basis of the mRNA size from northern blots (2.2 and 7.5 kb) and the presence of a polyadenylation signal and polyA tract at one end of the EST sequences.

The mouse BF539166 cDNA clone was purchased from Research Genetics and was sequenced to completion; it represents a full-length 1.9-kb mouse Nipa1 cDNA. Several mouse Nipa1 gene exon-intron boundary sequences were initially predicted on the basis of sequences identified by BLAST search of the mmtrace database and were subsequently confirmed by direct sequencing (data not shown) and by in silico analyses of BAC and shotgun sequence from the public genome sequence, as well as Celera databases. BLAST search of the Fugu Project Database with mouse NIPA1 amino acid sequence identified all five exons for the ORF sequence of the orthologous Fugu Nipa1 (as above for Nipa2).

Analyses of NIPA1 and NIPA2 Polypeptide Sequences

In addition to the NIPA1 and NIPA2 genes identified as above for human, mouse, Fugu, and chicken (NIPA2 only), using translated BLAST searches of cDNA and genome project databases, we identified an NIPA2 ortholog in Xenopus (encoding a putative polypeptide of 362 amino acids), as well as a single ancestral gene in Drosophila (385 amino acids), Anopheles gambiae (351 amino acids), and Caenorhabditis elegans (358 amino acids), and a representative paralog in Arabidopsis (343 amino acids). CLUSTAL W (Thompson et al. 1994) was used to align NIPA1, NIPA2, and ancestral gene amino acid sequences, with comparison of NIPA1/2 orthologs and paralogs by an identity/similarity matrix program, while phylogenetic analyses used drawtree to plot an unrooted tree diagram. Putative transmembrane helices were predicted based on hydrophobicity plots using the TMHMM (v. 2.0) program. It may be noted that a WW protein-protein interaction domain was predicted to occur in mouse and human NIPA1 proteins; however, we assume this was a chance similarity, since it spans a strong transmembrane (TM) helix (H) domain and is not conserved.

Determination of Gene Expression by RT-PCR and Northern Blot Analyses

Total RNA samples were extracted using TRIzol (Invitrogen), according to the manufacturer's protocol, from lymphoblast cell lines derived from normal individuals, unaffected and affected patients with an imprinting defect from the PWS-U (Buiting et al. 1995) and AS-J families (Saitoh et al. 1996), and rodent-human somatic cell hybrids with a single maternal (20L-28) or paternal (A59-3az) human chromosome 15 (Gabriel et al. 1998). For the hybrids analysis, PCR primers were human specific. In addition, RNA was extracted from homogenized mouse brain from three genotypes of mice: wild-type, transgenic PWS, and AS (Tg^PWS(del) and Tg^AS(del)), the last two of which have an ∼4-Mb deletion of mouse chromosome 7C (Gabriel et al. 1999). From each sample, 1–10 μg of total RNA was pretreated with DNase I (GIBCO-BRL), with half of this used to synthesize first-strand cDNA, using SuperscriptII reverse transcriptase (GIBCO BRL) and an oligo dT primer plus random hexamers (RT+), and the other half was reverse transcriptase free (RT−). One twenty-fifth of each reaction was used subsequently for PCR amplification, with 10 mM Tris-HCl (pH 8.4), 50 mM KCl, 2 mM MgCl₂, 0.2 mM dNTPs, 2 U Taq DNA polymerase (GIBCO BRL), and 200 nM each of primers, in a 25-μl reaction volume. To assess imprinting, we utilized 35 cycles of 94°C for 2 min, 55°C–60°C for 30 s, and 72°C for 1 min, with a 10-min final extension at 72°C. Primers used for each locus (table 1) were as follows: human and mouse NIPA1 exons 2–4; human and mouse NIPA2 exons 3–6; human and mouse GCP5 exons 10 and 11; human CYFIP1 3′ UTR, as well as exons 17–25; mouse Cyfip1 exons 4–7; human SNURF-SNRPN exons 2 and 3; mouse Snurf-Snrpn exons 3–9; and mouse Mkrn1.

RT-PCR products obtained from human and mouse NIPA1 (167 bp), NIPA2 (339 bp), and GCP5 (399 bp), human CYFIP1 3′ UTR (505 bp), and mouse Cyfip1 3′ UTR (258 bp, RN1102, and RN1103), as described above, were subcloned into a TA vector and used for preparation of probes. The probe for SNURF-SNRPN was similarly prepared from TA-cloned PCR product of exons 2 and 3 by amplification with RN420 and RN423. Probes were labeled with α-³²P-dCTP and were hybridized to human multiple tissue and brain multiple tissue northern blots (Clontech), human 12-lane MTN blot (BD Biosciences), or mouse multiple tissue northern blot-2 (Sigma), according to the manufacturer’s protocol. β-ACTIN or SNURF-SNRPN and Gapdh were used as controls for human and mouse northern blots, respectively. Hybridizations were performed using ExpressHyb (Clontech) at 65°C and were washed at high stringency in 0.1× SSC + 0.1% SDS at 60°C–65°C. Membranes were exposed to autoradiographic film at −80°C.

Identification and Fine Mapping of Novel Genes in the BP1–BP2 Region of Human 15q11.2

To identify novel genes in the BP1–BP2 region, we first used BLAST searches with the chromosome 15q11.2 STS marker D15S18 and microsatellite markers D15S1035, D15S541, and D15S542 (Christian et al. 1999) to identify partially sequenced (subsequently completed) BACs RP11-26F2 and RP11-289D12, which span this region (fig. 1a). From GeneMap ’98, we identified the KIAA0068 EST, subsequently recognized as SRA1/CYFIP1 (Kobayashi et al. 1998; Schenck et al. 2001), with exons in each of the two BP1–BP2 region BACs, a location confirmed by PCR (see below). Subsequently, we identified a second gene, GCP5 (Murphy et al. 2001), adjacent to CYFIP1, by BLAST searches with genomic sequence from BAC 289D12. Gene structures for these two genes have not been previously determined. The human CYFIP1 gene spans 111.4 kb and has 31 exons (table 2) overlapping BACs 26F2 (exons 23–31) and 289D12 (exons 1–24) (fig. 1a), with short 5′ and 3′ UTRs of 52 nt and 562 nt, respectively. Exon 1 is noncoding, with the AUG start codon present at the 5′ end of exon 2. Mouse Cyfip1 has a similar 31-exon gene structure. Human GCP5 (also known as TUBGCP5) spans 46.8 kb and 23 exons with a 54-nt 5′ UTR and a 643-nt 3′ UTR (table 3), with a similar structure in the mouse (data not shown). Although 45% of the human GCP5 3′ UTR and 17% of the mouse 583-nt 3′ UTR are an Alu and a truncated B2 repetitive element, respectively, these short interspersed nuclear elements (SINEs) have different evolutionary origins and are inserted at different positions and orientations, which suggests coincidental insertion and a lack of selection on the 3′ UTR in each species. The CYFIP1 and GCP5 genes are arranged in a head-to-tail orientation in human and mouse (fig. 1a), with the 5′ ends of both genes associated with strong CpG islands. In addition, two novel CpG island sequences not associated with either known gene were identified from genomic sequence of BAC RP11-26F2, and these are characterized in detail below as being located at the 5′ ends of the NIPA1 and NIPA2 genes.

Because of the repeated nature of HERC2- and other duplicons in BP1 and BP2 and because of poor BAC and DNA sequence coverage in these regions, such analyses could not determine the orientation of BACs 26F2 and 289D12 or of the four BP1–BP2 region genes with respect to the centromere. Using primers derived from the 5′ end of NIPA1 (fig. 1b), both 5′ and 3′ ends of GCP5 (fig. 1b) and the 3′ UTR (exon 31) of the CYFIP1 gene (fig. 1c), we performed PCR on a panel of YACs known to span the BP1–BP2 region (fig. 1d). This YAC panel was characterized previously by STS content and HERC2-duplicon mapping (Amos-Landgraf et al. 1999). Both NIPA1 exon 1 and the 3′ UTR of CYFIP1 were detected on YACs 931C4 and A124A3 (fig. 1b, c, and d), confirming that these genes map proximal to BP2 of MKRN3 and were therefore proximal to the known PWS imprinted domain. Furthermore, both ends of GCP5 are within YAC 931C4 but not A124A3 or any BP2 region YACs (fig. 1b and 1d), unambiguously defining the order as cen–NIPA1–CYFIP1–3′-GCP5-5′–BP2–tel.

FISH Mapping of BP1- and BP2-Associated Rearrangements

BAC clone 3242E18, containing exons 7–31 of CYFIP1 from the BP1–BP2 region (fig. 1e), was utilized for FISH studies on metaphase chromosomes from three normal human lymphoblastoid cell lines, with hybridization to a single region of proximal chromosome 15 in band 15q11.2 (fig. 2a). No other chromosomal hybridizations were detected, including in interphase cells, confirming that BAC 3242E18 does not contain duplicon sequences. Therefore, we investigated the pattern of hybridization of BAC 3242E18 against a panel of cell lines from patients with AS deletions. Deletions in patients with AS and PWS (fig. 1e) are either class I, which extends from BP1 (15q11.2) to BP3 (15q13), or class II, which extends from BP2 (15q11.2) to BP3 (Knoll et al. 1990; Amos-Landgraf et al. 1999). As expected, the 3242E18 FISH assay distinguishes class I and class II deletions in patients with AS, because the BAC is present on both chromosomes 15 in cells of three class II patients and is present only on the normal chromosome 15 in cells from two class I patients (fig. 1e; data not shown).

Physical Map Spanning Six Genes at the Centromeric End of Mouse Chromosome 7C

The human CYFIP1 gene is orthologous to the mouse Shyc (Cyfip1) gene (Köster et al. 1998; Schenck et al. 2001). We isolated the Cyfip1 129/Sv RPCI-22 252P22 BAC clone (fig. 1f) by PCR screening with a 3′ UTR STS. FISH was then performed on metaphase chromosomes from a mouse model of PWS with a chromosome 7C deletion (Gabriel et al. 1999). BAC 252P22 (Cyfip1) localizes to the normal chromosome 7C region and within the mouse transgene-induced deletion (fig. 2b), which expands the known syntenic relationship between human 15q11-q13 and mouse 7C (fig. 1a).

To generate a physical map, seven unsequenced 129/Sv and 14 partially sequenced C57BL/6 BACs were identified (see “Material and Methods” section) and typed by PCR with STSs derived from the p, Herc2, Nipa1, Nipa2, Cyfip1, and Gcp5 genes and with BAC end sequences created in this study, allowing generation of an overlapping BAC contig map (fig. 1f). These data indicate that the genes in this region are arranged in the order cen–Gcp5–Cyfip1–Nipa2–Nipa1–Herc2-p–tel (fig. 1f). The map is anchored at one end by an STS marker, D7Mit70, which maps close to the l71Rl locus involved in peri-implantation survival (Wu et al. 2000), that is ∼22 kb centromeric to Gcp5.

Replication Analyses of the Nipa1-Nipa2-Cyfip1 Locus

Replication-timing studies of imprinted regions have demonstrated a parent-of-origin–dependent asynchrony, using FISH and quantitative PCR techniques (Kitsberg et al. 1993; Knoll et al. 1994; Simon et al. 1999). For two BACs of different sizes that collectively span the Nipa1, Nipa2, and Cyfip1 genes, we see a pattern of asynchrony that is similar to that at the imprinted Mkrn3 and Snurf-Snrpn loci (fig. 2c). To test whether the pattern at Nipa1-Nipa2-Cyfip1 was determined by parent of origin, we used mice heterozygous for a deletion at the mouse Tyr (c) locus (c^32DSD) (Rinchik et al. 1993) to examine which chromosome was replicating earlier in each cell. The relatively close linkage of Tyr to the imprinted 7C region (∼14 cM) makes this a useful marker for in situ studies of this region while being sufficiently distant that the likelihood that cis effects of the deletion cause disruption of imprinting is very small. We found a pattern of replication in these cells that was asynchronous but was not due to parent-of-origin influences. In other words, the proportion of cells that showed the paternal chromosome replicating early was equal to the proportion that showed the maternal chromosome replicating early (fig. 2c).

Characterization of the NIPA1 Gene in Vertebrates

Human NIPA1 maps within BAC 26F2 and was identified, on the basis of a strong CpG island, by visual inspection of genomic DNA sequence, followed by assembly of a cDNA sequence contig, using BLAST searches extending 3′ from the CpG island (see “Material and Methods” section for details). This led to an NIPA1 cDNA sequence of 2,247-bp that spanned five exons and 38.8 kb in genomic DNA (fig. 3a) and encoded a putative polypeptide of 329 amino acids (fig. 4a; see below). The five NIPA1 exon-intron boundaries conform to consensus splice acceptor and donor sequences (table 4), with the translational AUG start codon present in exon 1, the stop (TGA) codon in exon 5, and in the 3′ UTR an ATTAAA variant polyadenylation signal at cDNA position 2221–2226 (fig. 3a; table 4). Although single base changes of the hexanucleotide (AATAAA) poly(A) signal sequence greatly reduce polyadenylation, the A→T substitution at position +2 of the noncanonical ATTAAA hexamer is the mildest mutation and corresponds to the most common variant (Sheets et al. 1990). Nevertheless, multiple NIPA1 ESTs have polyA tails beginning 13–17 nt 3′ of this polyA signal, suggesting this as a bona fide 3′ end of low-abundance NIPA1 transcripts, which was supported by northern analysis of gene expression, although these studies also identify a larger neuronal transcript (see below). Therefore, we examined genomic sequence further 3′ both visually and by BLAST against the EST database, which allowed identification of the major human NIPA1 3′ UTR polyadenylation signal (AATAAA) at cDNA positions 6540–6545. The latter is embedded in a region that is highly conserved in five eutherian mammals and that is diagnostic for the NIPA1 3′ end (fig. 3b).

Within the GC-rich (75%) exon 1 of NIPA1, we identified a (GCG)_n repeat, encoding polyalanine (fig. 4a), which we examined for polymorphism by one or both of following methods: direct sequencing of exon 1 PCR products or detection of PCR products by use of a ³²P–dCTP labeled primer. This analysis showed polymorphism in 135 chromosomes of European and Asian origin with the number of repeats ranging from 6 to 10 (data not shown). The two most frequent alleles, (GCG)₈ and (GCG)₇, have allele frequencies of 0.78 and 0.2, respectively. Additional studies are needed to determine whether expansions occur in the NIPA1 (GCG)_n repeat and are disease associated.

The mouse Nipa1 cDNA was identified by in silico methods, using the human coding sequence as well as sequencing of a 1.9-kb cDNA clone. By comparison with mouse genome sequence, the Nipa1 locus was found to span 41.1 kb of genomic DNA and to include five exons that are highly conserved with the human NIPA1 gene (fig. 3a; table 4). Mouse Nipa1 encodes a putative polypeptide of 323 amino acids (fig. 4a), with an N-terminal polyalanine sequence encoded by a (GCG)₃ GCT(GCG)₃ sequence in exon 1. In a manner similar to that in humans, two polyadenylation signals occur in the mouse Nipa1 3′ UTR at cDNA positions 1866 nt and 6840 nt of the cDNA (fig. 3a), both of which are AATAAA signals and correspond to the two Nipa1 transcripts (see below). We also identified in silico the complete orthologous Nipa1 gene from the Fugu Project Database by translated BLAST analyses, using human and mouse amino acid sequences. This allowed us to predict a virtual cDNA sequence of 932 bp, with five exons of similar sizes and conserved consensus splice acceptor and donor sequences, as for the human and mouse orthologous genes (fig. 3a; table 4), and encoding a putative polypeptide of 304 amino acids (fig. 4a). As expected, the genomic locus of Fugu Nipa1 is much smaller than that of mammalian NIPA1 loci and spans only ∼1.4 kb, not including the UTRs.

Characterization of the NIPA2 Gene

As with NIPA1, the human NIPA2 gene was identified by in silico methods using BLAST of BAC 26F2 genomic sequence, as well as identification of a second CpG island in this BAC that represents the 5′ end of NIPA2. We compiled a synthetic NIPA2 cDNA sequence of 2,450 bp, with an AATAAA poly(A) signal at positions 2444–2449 utilized for polyadenylation (fig. 3c). NIPA2 spans 29 kb in the human genome and has seven exons (fig. 3c), each with consensus exon-intron boundaries (table 4), and for which exons 3–7 encode a NIPA2 putative polypeptide of 360 amino acids (fig. 4b). The mouse Nipa2 cDNA (see the “Material and Methods” section) is very similar to human NIPA2, having seven conserved exons (fig. 3c; table 4), encoding a putative polypeptide of 359 amino acids in exons 3–7 (fig. 4b), and with this locus spanning 29.8 kb in the mouse genome. Two forms of the mouse Nipa2 cDNA (1,867 bp and 3,200 bp; fig. 3c) differ in the positions of alternative polyadenylation sites in the 3′ UTR, created through the use of poly(A) signals at cDNA positions 1846–1851 (ATTAAA) and 3131–3136 (AATAAA). Both transcripts are observed in different tissues (see below).

We also sequenced and assembled a 1,805-bp Gallus gallus Nipa2 cDNA, encoding a 361–amino acid putative polypeptide. Adjacent to the Fugu Nipa1 gene, we identified the Fugu Nipa2 ortholog, encoding a 366–amino acid putative polypeptide. There are five coding exons that correspond to exons 3–7 of the orthologous human and mouse genes (fig. 3c) with conserved exon-intron boundaries (table 4), although putative noncoding exons 1 and 2 have yet to be identified in Fugu. The Fugu Nipa1 and Nipa2 genes were found in conserved positions with respect to the mouse locus (fig. 1f), and Fugu orthologs for the p, Herc2, Cyfip1, and Gcp5 genes were also identified in conserved positions (data not shown), providing strong evidence that each represents the ortholog of the mammalian genes. Finally, the position of the splice donor and acceptor sequences for the fourth intron of NIPA2 is conserved in all homologs analyzed (fig. 3c), including the single intron present in the insect “ancestral” gene, intron 2 for the “ancestral” C. elegans gene (total of six exons), and intron 2 of an Arabidopsis paralog with nine total exons.

Comparison of human NIPA2 and mouse Nipa2 cDNAs demonstrates two peaks of homology, one that begins at the NIPA2 start codon in exon 3 and the other that is a short region (∼120 bp) in exon 1 (fig. 3d). The latter encodes a potential 36-to-40–amino acid upstream ORF (uORF) that is conserved in mammals (figs. 3d and 3e), and several additional uORFs occur in exons 1–3. Distantly related uORF sequences also occur in the NIPA2 5′ region in chicken (where it is the second uORF) and Xenopus (fig. 3e). A conserved uORF, such as that observed here in NIPA2 loci, is likely to regulate NIPA2 translation (Morris and Geballe 2000).

Polypeptide and Phylogenetic Analyses of the NIPA1 and NIPA2 Family

Analysis of the human, mouse, chicken, Xenopus, and Fugu NIPA1 and NIPA2 polypeptide sequences (figs. 4a and 4b), as well as the ancestral invertebrate and related plant members (fig. 4b), identified nine hydrophobic segments in each that likely encode putative transmembrane helices (TMH) (fig. 4c). Although the TMHMM program assigned a lower probability for TMH6 in human and mouse NIPA1, it is nevertheless clear that there is a hydrophobic domain of sufficient length to span the lipid bilayer and that TMH6 is clearly predicted in Fugu NIPA1, as well as in all related proteins (fig. 4c). We conclude that the NIPA1 and NIPA2 gene family encode polypeptides that have nine transmembrane-spanning helices.

Alignment of amino acid sequences of the NIPA1 polypeptide from human, mouse, and Fugu reveals these to be highly conserved (fig. 4a; table 5). Indeed, the human and mouse sequences display 98% identity, with a single amino acid change other than the length of the polyalanine tract, whereas NIPA1 in Fugu is ∼55% identical to that in mammals. Alignment of amino acid sequences of the NIPA2 polypeptide from human, mouse, chicken, Xenopus, and Fugu shows identities ranging between 78% and 96%, with the C-terminus being the most divergent (fig. 4b; table 5). Remarkably, NIPA1 and NIPA2 polypeptide sequences are related, with 32%–36% identity (and 47%–50% similarity) between any paralogous pair within these five species. Similarly, the ancestral Drosophila, C. elegans, and Anopheles polypeptides and a representative Arabidopsis member are highly related, with 40%–49% identity to vertebrate NIPA2 and lower identity (29%–34%) to vertebrate NIPA1 (table 5). The phylogenetic relationships are best seen in a tree diagram, which clearly illustrates the grouping of NIPA2 and NIPA1 homologs within vertebrates, and the distant ancestral members in invertebrates and Arabidopsis. The phylogenetic distance between NIPA1 and NIPA2 paralogs is as great as that between them and ancestral members, strongly suggesting that NIPA1 and NIPA2 have evolved related but different functions.

Analyses of Imprinting for BP1–BP2 Genes and Mouse Orthologs

The most proximal known imprinted gene in 15q is MKRN3, which is located just distal to BP2, although the presence of replication asynchrony in the pericentromeric region (see the “Introduction”) and at mouse Nipa1-Nipa2-Cyfip1 (fig. 2c) raises the possibility that the imprinted domain could extend further. Consequently, we tested whether the four BP1–BP2 region genes might be imprinted in human, using RNA from cells from patients with imprinting mutations who have two maternally (PWS) or two paternally (AS) imprinted chromosomes 15 (Buiting et al. 1995; Saitoh et al. 1996), as well as a somatic cell hybrid imprinting assay in which the rodent cell lines retain either a maternal or a paternal human chromosome 15 (Gabriel et al. 1998). RT-PCR shows that human NIPA1 (both the major mRNA detected by a 167-bp band and a larger form with the alternatively spliced exon 3b), NIPA2, GCP5, and CYFIP1 are each expressed in lymphocyte cells from a normal individual and PWS and AS imprinting mutation patients (figs. 5a and 5b), whereas SNURF-SNRPN, a paternally expressed imprinted gene in the PWS region as a control is only expressed in the normal and AS individuals (fig. 5a). In addition, CYFIP1 is expressed from both the maternal and paternal chromosome 15 in somatic cell hybrids (fig. 5b). Therefore, the four BP1–BP2 genes are nonimprinted in peripheral blood lymphocytes and in fibroblasts.

To test imprinting in the mouse, we used a transgenic PWS and AS mouse model that has a paternally or maternally derived chromosome 7C deletion, respectively, spanning at least from Frat3 through Herc2 (Gabriel et al. 1999; Chai et al. 2001). On the basis of the deletion of BAC 252P22 (fig. 2b), the 3′ UTR of Cyfip1 and all of Nipa1 and Nipa2 are included in the deletion (see fig. 1f), as are 5′ Cyfip1 and Gcp5 (J.M.G., J-H.C., and R.D.N., unpublished data), allowing this model to be used to test imprinting of these genes. All four genes (Nipa1, Nipa2, Cyfip1, and Gcp5) are expressed in wild-type, Tg^PWS(del), and Tg^AS(del) mice, resembling the control Mkrn1 profile (fig. 5c). In contrast, the imprinted Snurf-Snrpn gene is not expressed in the PWS mouse, as expected (fig. 5c). Therefore, mouse Nipa1, Nipa2, Cyfip1, and Gcp5 genes are expressed from both the paternal and maternal alleles and are nonimprinted in the brain.

Expression Analyses of BP1–BP2 Genes and Mouse Orthologs

The tissue patterns of expression for the four BP1–BP2 region genes and their mouse orthologs were studied by northern blot analysis. As predicted from cDNA and genomic analyses (see above), a single 2.4-kb transcript is seen for human NIPA2 (fig. 6a), whereas 1.9 kb and 3.2 kb transcripts occur for mouse Nipa2 (fig. 6b), with constitutive expression in both species including pronounced expression in the placenta (figs. 6a and 6b). In mouse, a 1.9-kb Nipa1 transcript is also constitutive but at relatively low levels, whereas a larger (7.5-kb) mRNA is seen in several tissues, with both mRNA isoforms enriched in brain (fig. 6b). Similarly, the human NIPA1 locus expresses a short (2.2-kb) and a long (7.5-kb) mRNA at low levels, with enrichment of the larger isoform in neuronal tissues (Rainier et al. 2003 [in this issue]). Similar studies with CYFIP1 demonstrate a 4.4-kb constitutively expressed mRNA in human (fig. 6c) and mouse (fig. 6d), with highest levels in placenta (figs. 6c and 6d) and equal levels throughout the brain (fig. 6c). GCP5 has a 3.7-kb transcript in human (fig. 6e) and mouse (fig. 6f), with the former enriched in muscle tissues, whereas a smaller (2.8-kb) GCP5 mRNA is also seen in some tissues. Mouse Gcp5 is also enriched in placenta (fig. 6f), as is true for NIPA2 and CYFIP1 in human and mouse.

Implications for Genomic Imprinting Mechanisms

All four genes between BP1 and BP2 in 15q11.2 were shown here to be nonimprinted in lymphoblastoid cell lines and human-rodent somatic cell hybrids, with the mouse orthologs nonimprinted in brain. The latter observation is important because most, if not all, the imprinted loci in the PWS/AS imprinted domain are predominantly expressed and imprinted in the brain (Nicholls and Knepper 2001). We conclude that the mammalian NIPA1, NIPA2, CYFIP1, and GCP5 genes are nonimprinted. Although the four mouse genes map ∼2 Mb from the imprinted domain, the human orthologs are separated from the PWS imprinted domain solely by BP2 sequences; however, the size of BP2 is poorly defined because of the complex and incompletely known composition of duplicons. Since the HERC2 duplicons within BP2 include the CpG-rich promoter and are derived from a nonimprinted locus (Amos-Landgraf et al. 1999; Ji et al. 1999), we suggest that one model to explain the proximal imprint boundary might simply be the evolutionary integration of “nonimprintable” promoter sequences adjacent to the imprinted domain. In contrast, when genes such as mammalian MKRN3 and rodent Frat3 were duplicated by retroposition, they had promoters that were subject to the imprint process when these genes integrated into or adjacent to the imprinted domain (Chai et al. 2001).

Several studies have shown replication asynchrony in domains with monoallelic gene expression, including those imprinted (Kitsberg et al. 1993; Chess et al. 1994; Knoll et al. 1994; Hollander et al. 1998; Simon et al. 1999; Mostoslavsky et al. 2001) and with a transition from imprinted to nonimprinted sequences at mouse H19, defined by an asynchronous to synchronous replication change (Greally et al. 1998). The use of marked mouse chromosomes 7 gave us an extremely useful tool for analysis of replication patterns at the Nipa1-Nipa2-Cyfip1 domain, where we found asynchrony that was random with respect to parental origin. A similar pattern has been described elsewhere for the P (OCA2) locus in human 15q13 (Knoll et al. 1994), whereas imprinted loci replicate the chromosome derived from one parent (usually the paternal allele) earlier than the other (Kitsberg et al. 1993; Knoll et al. 1994; Simon et al. 1999). The OCA2 locus is ∼1.5–2 Mb from the imprinted domain, whereas the mouse region studied here is ∼2–2.5 Mb from the mouse 7C imprinted domain. It is possible that the transition from asynchrony determined by parent of origin to the pattern of random asynchrony that we have demonstrated here is detectable only with the use of a marked chromosome such as that used in this study. Our finding that the four genes in the BP1–BP2 region are nonimprinted suggests that the asynchronous DNA replication observed proximal to deletion breakpoint BP1 (Ritchie et al. 1998) may be another example of the same phenomenon and does not indicate a more extensive imprinted domain. Rather, replication asynchrony may identify those chromosomal regions with the capacity for monoallelic gene expression: however, we propose that other factors, including DNA methylation and/or elements of the histone code (Jenuwein and Allis 2001; Turner 2002), are mechanistically needed to complete the silencing or activation of individual alleles.

Functional Considerations of a New Gene Family

Two of the four genes identified in the BP1–BP2 region are the related NIPA1 and NIPA2 genes. Phylogenetic studies indicate that both paralogous gene members are ancient and likely arose 450–600 million years ago, around the time of the origin of vertebrates. Despite the adjacent chromosome linkage of NIPA1 and NIPA2 and retention of this syntenic relationship in all examined vertebrates, there is no evidence that one served as the precursor to the other. Indeed, four additional NIPA1/2-related paralogous family members are dispersed in vertebrate genomes, and each appears to have arisen early in vertebrate evolution (J-H.C. and R.D.N., unpublished data). Alternatively, NIPA1 or other family members may be evolutionarily younger, and, after their origin, they may have gone through a period of rapid mutation, followed by fixation and a dramatic slowdown in mutation rate. Further study of this gene family in extant vertebrate species will shed light on the evolutionary history of this novel gene family.

NIPA1 and NIPA2 differ in gene structure and expression patterns, reflecting their ancient origins; however, in both cases, there are five coding exons, and these show homology in intron placement. In particular, the only intron in the insect NIPA1/2 gene is conserved in position in all vertebrate NIPA2, NIPA1, invertebrate, and plant family members, despite variation in exon number. During evolution of NIPA2, two 5′ noncoding exons arose along with the coding potential for a small uORF. The presence of one or more uORFs in NIPA2 suggests that this gene undergoes translational regulation (Morris and Geballe 2000) and that translation of NIPA2 may involve an internal ribosome entry site mechanism (Fernandez et al. 2002), although further work will be necessary to examine these possibilities. In contrast to the constitutive expression of NIPA2 and the smaller NIPA1 transcript in all tissues of human and mouse, NIPA1 shows high levels of brain-enriched expression of a 7.5-kb transcript (as shown here and in the article by Rainier et al. 2003 [in this issue]). The mechanism for this may reflect stabilization of the mRNA in brain for the long form of NIPA1, transcriptional activation plus coordinate polyA site selection in a neuron-specific manner, or silencing of the latter transcription-polyadenylation pattern in nonneural tissues, as occurs with the neuron-restrictive silencer factor (Zhao et al. 1999; Lunyak et al. 2002; Proudfoot et al. 2002; Worthington et al. 2002).

Our prediction of a nine-TMH helix domain structure for the NIPA1 and NIPA2 polypeptides suggests that they function as either receptors or transporters. NIPA1/2 show low homology, spanning three TM domains (∼29% over ∼93 amino acids), with some G-protein coupled receptors, such as extracellular calcium receptors and olfactory receptors (data not shown); however, this may reflect the hydrophobic nature of TM domains, and any functional or evolutionary relationship remains conjecture. Indeed, polypeptides with nine TM spanning passes of the lipid bilayer are rare. Exceptions are glucose-6-phosphatase (Pan et al. 1998), a cardiac Na^+/−-Ca²⁺ exchanger (Qiu et al. 2001), and the TM9 domain superfamily (TM9SF) (Chluba-de Tapia et al. 1997); TM9SF members are localized to endosomes (Schimmöller et al. 1998), bind the synthetic ligands cyanopindolol and the β-adrenergic agonist SM-11044, and have roles in colon relaxation and eosinophil chemotaxis (Sugasawa et al. 2001), or are involved in adhesion and phagocytosis in Dictyostelium (Cornillon et al. 2000). However, the TM9SF family is unrelated in primary sequence to the NIPA1/2-family. Furthermore, whereas TM9SF members have a signal peptide (type I topology), members of the NIPA1/2-family do not, and the latter must have internal membrane targeting sequences. Consequently, the function of the NIPA1/2 family remains unknown. Nevertheless, the presence of distinct family members in vertebrates, invertebrates, plants, and bacteria (the present article; J-H.C. and R.D.N., unpublished data) indicates that these conserved polypeptides are likely to be critical for signaling within or between cells. Studies with antibodies and with animal models deficient in or overexpressing these genes, as well as identification of ligands for these polypeptides, are now necessary to determine their functions.

Implications of BP1–BP2 Region Genes for Hereditary Disease

Since blocks of duplicated sequences that are implicated in homologous recombination flank the BP1–BP2 region genes, chromosome rearrangements involving the BP1–BP2 region may lead to dosage imbalance or rearrangements of the NIPA1, NIPA2, CYFIP1, and GCP5 genes. This includes loss of one allele in the larger class I chromosome deletions in PWS and AS, or gain of one or two alleles, in 15q11-q13 duplications and triplications as well as for inv dup (15) marker chromosomes (Knoll et al. 1990; Cheng et al. 1994; Leana-Cox et al. 1994; Huang et al. 1997; Amos-Landgraf et al. 1999; Christian et al. 1999; Roberts et al. 2002). On the basis of the finding of small inv dup (15) marker chromosomes with the BP1–BP2 region but not the PWS/AS region in a series of patients with generally normal phenotypes, Huang et al. (1997) concluded that no genes in these small markers have clinically relevant dosage effects. In contrast, recent studies by Butler et al. (2003) found that PWS subjects with class I deletions (BP1–BP3) have a more severe phenotype than those with class II deletions (BP2–BP3), including greater self injurious behavior, deficits in adaptive behavior (including motor skills), obsessive-compulsive behavior, and difficulties with reading, mathematics skills, and visual-motor integration. These data provide evidence that deletions of the four genes in the BP1–BP2 region may be associated with dosage-sensitive behavioral and psychological phenotypes. We propose that some individuals may have a condition with just these types of clinical phenotypes and a deletion or duplication limited to the BP1–BP2 region due to recombination between duplicated sequences within BP1 and BP2. Further study of BP1–BP2 region genes and chromosome rearrangements will determine whether the presence of four highly conserved genes flanked by unstable DNA sequences can have a significant phenotypic impact.

A chromosome 15 rearrangement limited to the unique 250-kb BP1–BP2 region was recently found in a subject with PWS due to a 15q11-q13 deletion that arose on the paternally inherited chromosome containing a familial duplication limited to BP1–BP2 (Butler et al. 2002). Consequently, duplications, deletions, or inversions limited to the BP1–BP2 region may predispose to subsequent larger rearrangements of proximal 15q. Indeed, a polymorphic 1.5-Mb inversion within a similar class of duplicated sequences in chromosome 7q11.23 is associated with susceptibility to deletion of the 1.5-Mb region in Williams-Beuren syndrome (Osborne et al. 2001). Similarly, inversions of BP2–BP3 were identified in four of six mothers of patients with AS deletion (Gimelli et al. 2003), which supports the previous finding of a familial 15q11-q13 inversion in the mother and father of deletion patients with AS and PWS, respectively (Clayton-Smith et al. 1993), consistent with the large inversion being a predisposition allele. Indeed, BP1–BP2 deletions or other rearrangements may be more common than the PWS/AS deletions, which occur at an overall frequency of ∼1/10,000 newborns, since the distance between the putative recombining segments is only a fraction of that of the 4–4.5-Mb PWS/AS deletions.

All four BP1–BP2 region genes are candidates for dominantly inherited spastic paraplegia, locus 6 (SPG6 [MIM 600363]). SPG6 is characterized by insidiously progressive lower-extremity spasticity in which axonal degeneration affects primarily the longest axons in the CNS and is caused by a mutation within a 7.3-cM segment spanning the PWS/AS region (Fink at al. 1995). The finding of obligate recombinants for markers within the cluster of GABA_A receptor genes in distal 15q11-q13 ruled these out (Fink et al. 1996). As a dominant disorder, SPG6 is not expected to be associated with imprinted genes, ruling out most other genes in the PWS/AS deletion region. Since no functional genes are known proximal to BP1, the BP1–BP2 region genes, particularly CYFIP1 and GCP5, became prime candidates. CYFIP1 associates with Rac1 and F-actin (Kobayashi et al. 1998), with Rac1 implicated in regulation of the actin cytoskeleton, including axon growth, guidance, and branching (Ng et al. 2002). CYFIP1 also associates with FMRP (Schenck et al. 2001), the fragile X mental retardation protein, which is implicated in neurite extension, guidance, and branching (Morales et al. 2002). In addition, GCP5 encodes a protein that is part of the human γ-tubulin complex required for microtubule nucleation at the centrosome (Murphy et al. 2001). Microtubules are critical in axonal transport, and defects in this process are known in spastic paraplegias (Crosby and Proukakis 2002). However, although NIPA1 and NIPA2 are of unknown function, the CNS-specific expression of NIPA1 made it an attractive and correct candidate (Rainier et al. 2003 [in this issue]).

Mutation studies of the mouse orthologs of the four BP1–BP2 region genes will shed light on the potential phenotypic role of recessive loss-of-function mutations in these genes. Indeed, a locus required for embryo implantation, l71Rl, was identified just proximal to the pink-eyed dilution (p) gene on the basis of recessively inherited deletions of the region (Wu et al. 2000). Genetic complementation and molecular mapping define the critical interval for l71Rl as a small region between D7Mit70 and 5′ Herc2 (Wu et al. 2000), to which we have mapped the mouse Nipa1, Nipa2, Cyfip1, and Gcp5 genes, making one or more of these four genes likely candidates for l71Rl. Although the homologous human chromosome BP1–BP2 region is flanked by unstable DNA sequences, as discussed above, the mouse l71Rl data indicate that it is unlikely that individuals will exist who are homozygously deleted for the BP1–BP2 region, since such homozygous deletions would be predicted to be associated with a failure of the embryo to implant. Should BP1–BP2 deletions occur in association with a normal or mild neurobehavioral phenotype (see above), their inheritance from each parent might be associated with early pregnancy loss and thus may represent an important genetic counseling issue.

Evolutionary Transposition of BP1–BP2 Region Genes Mirrors Recombination Events in Genomic Disease

We have shown that a block of at least six genes (OCA2, HERC2, NIPA1, NIPA2, CYFIP1, and GCP5) are syntenic in human, mouse, and Fugu; however, in human, the latter four genes have transposed to a site ∼4 Mb away from the OCA2 and HERC2 genes. These and previously published data allow us to propose a model for the evolution of human 15q11-q13 (fig. 7). Mouse and Fugu are representative of the ancestral vertebrate arrangement of these genes (fig. 7). The evolutionarily young imprinted domain was subsequently added in mammals, adjacent to this ancestral domain (fig. 7) (Chai et al. 2001; Nicholls and Knepper 2001). During evolution of an ancestral primate, 25–60 million years ago (Amos-Landgraf et al. 1999; Christian et al. 1999; Ji et al. 1999; Locke et al. 2001), the HERC2 gene began its evolutionary odyssey, and we propose that duplications of this gene first formed duplicons flanking a NIPA1-NIPA2-CYFIP1-GCP5 cassette at a position equivalent to BP3 (fig. 7). Subsequently, the ∼250-kb four-gene unique cassette (with all required cis regulatory elements) and flanking HERC2 duplicons were transposed by a duplicon-mediated process in an ancestral primate to a site ∼4 Mb away, on the other side of the imprinted gene domain. Here, the flanking HERC2 duplicons would now have formed the BP1 and BP2 domains, as found in human (fig. 7). This transposition may have involved HERC2 duplicons directly, as supported by the location of a HERC2 duplicon in BAC 26F2 immediately adjacent to NIPA1 at BP1 (data not shown). Alternatively, other duplicons have been added and interspersed with HERC2 duplicons at unknown times during primate evolution (Buiting et al. 1999; Christian et al. 1999; Pujana et al. 2001) and may have contributed to the genomic transposition. Genomic studies of primates will help determine the timing of these events and mechanisms, although secondary rearrangements within and between duplicated sequences in different extant primate species might complicate interpretation of the exact evolutionary events. Nevertheless, it is clear that an evolutionary genetic instability has been conferred on this domain by the origin and expansion of large blocks of duplicated DNA sequences, now mirrored by the events that mediate chromosome rearrangements in genomic diseases involving 15q11-q13.