ABSTRACT
Coiled bodies (CBs) are nuclear organelles whose morphological structure and molecular composition have been conserved from plants to animals. Furthermore, CBs are often found to co-localize with specific DNA loci in both mammalian somatic nuclei and amphibian oocytes. Much as rDNA sequences are called nucleolus organizers, we term these coiled body-associated sequences `coiled body organizers' (CBORs). The only sequences that have been shown to be CBORs in human cells are the U1, U2 and histone gene loci. We wanted to determine whether other snRNA genes might also act as CBORs. In this paper we show that human U3 genes (the RNU3 locus) preferentially associate with CBs in interphase cells. In addition, we have analyzed the genomic organization of the RNU3 locus by constructing a BAC and P1 clone contig. We found that, unlike the RNU1 and RNU2 loci, U3 genes are not tandemly repeated. Rather, U3 genes are clustered on human chromosome 17p11.2, with evidence for large inverted duplications within the cluster. Thus all of the CBORs identified to date are composed of either tandemly repeated or tightly clustered genes. The evolutionary and cell biological consequences of this type of organization are discussed.
U3 RNA is the most abundant of the small nucleolar RNAs (snoRNAs), with an estimated 2 × 105 copies/cell (1 ). U3 is essential for growth in yeast (2 ) and is known to be required for very early pre-rRNA processing events in both yeast and mammals (2 -4 ). Along with U8 and U13, U3 RNA is one of only three known vertebrate snoRNAs to be transcribed by their own promoters (1 ); the vast majority of snoRNAs are processed from within the introns of upstream transcripts (5 ). U3 RNA is transcribed by RNA polymerase II, however, its promoter sequences lack TATA boxes and polyadenylation signals (reviewed in 6 ). Instead, initiation and termination of snRNA transcription occur by mechanisms different from those of other polymerase II transcripts (7 -9 ).
Human U3 genes are part of a multigene family located on chromosome 17p11-p12 (the RNU3 locus), with an estimated 5-10 copies/haploid genome (10 ). However, further characterization of the RNU3 locus has been hampered by the relatively small insert size of [lambda] clones (10 ) and the instability of large insert YAC clones (11 ). We became interested in the organization of U3 genes as part of a search for genomic loci that preferentially co-localize with coiled bodies (CBs) in human interphase cells.
Coiled bodies are nuclear organelles, conserved from plants to animals that contain high concentrations of small nuclear ribonucleoproteins (snRNPs) involved in a variety of cellular processes, including pre-mRNA splicing, rRNA processing and histone mRNA 3'-end maturation (12 -14 ). Due to their similar molecular composition, CBs are thought to be the somatic equivalent of the sphere organelles found in amphibian oocyte nuclei (12 ). Although CB function(s) remains obscure, recent studies have shown that histone, U1 and U2 genes co-localize with CBs in human cells (15 ,16 ). Of particular significance is the fact that sphere organelles are often found adjacent to the histone gene loci on amphibian lampbrush chromosomes (17 ,18 ). Thus the co-localization of CBs and sphere organelles with specific DNA sequences appears to have been evolutionarily conserved.
We wanted to determine whether other snRNA genes might also function as so-called coiled body organizer regions (CBORs). One common theme among these CBORs is that the various genes are multicopy and are either clustered (histone) or tandemly repeated (U1 and U2). Thus we screened human genomic libraries and isolated P1 and BAC clones containing U3 genes. Using the clones for fluorescence in situ hybridization (FISH) we found that, like the U1 and U2 genes mentioned above, U3 genes frequently associate with CBs in interphase HeLa and HT-1080 cells.
Each of the clones map to the same cytogenetic location (17p11.2) and can be assembled into an ~200 kb contig. There are three different U3 genes within this contig, as well as the upstream control region for a putative fourth gene. Furthermore, we found that the genes and flanking sequences within this cluster are highly conserved. Two of the genes are located within a 90 kb inverted duplication. Each half of this 90 kb element has the same (mirror image) restriction map and blotting and sequence analysis of coding regions and end clones reveals extensive sequence identity throughout the repeat, suggesting either a very recent duplication event or the existence of mechanisms for maintaining sequence homogeneity throughout the cluster.
P1 phage bearing a 266 bp PCR fragment from the region immediately upstream of the U3 coding region were identified within a pooled human genomic library by Genome Systems Inc. (St Louis, MO). The primers were GM1 (5'-GCTGTCATTCAGTATTATGCTAAT-3') and GM2 (5'-AATGATCCCTGAAAGTATAGTCTT-3') (19 ). Ten of the 25 top pools had positive PCR signals. From these 10 positives three P1 phage (clones 2981, 2982 and 2983) were identified in the lower pools. A pooled human BAC library was purchased from Research Genetics (Huntsville, AL) and screened with the RNU3 primer pair. The primers, called RNU3-5' (5'-TGGGGTGCAGTGGGGTAAC-3') and RNU3-3' (5'-TTCTCGCTCCCTCTTTTGTCGC-3'), produce a 1.1 kb band corresponding to positions -570 to +461 with respect to the start site of U3 transcription. Four clones were identified: 28C2, 53L14, 212M14 and 374C9. Clones 28C2 and 374C9 were found to be sister clones, with identical inserts, and are called C2/C9 throughout the text. Similarly, 53L14 and 212M14 were also sister clones and are called L14/M14.
P1 and BAC clone DNAs were restricted and analyzed by pulsed field gel electrophoresis (PFGE) using both CHEF-mapper and FIGE-mapper gel systems (BioRad, Burlingame, CA). FIGE was used primarily for fragments in the 1-150 kb size range. DNAs were fractionated in 1% agarose in 0.5× TBE with pulse times ranging from 0.1 to 6.8 s. Following electrophoresis, gels were stained with ethidium bromide, photographed and transferred to nylon membranes by standard methods. Filters were hybridized using either Church's buffer overnight (20 ) or Rapid-hyb buffer (Amersham, Arlington, IL) for 2.5 h. Initial washing was performed at room temperature in 2× SSC, 0.1% SDS for 15 min, followed by two stringency washes at 0.1 or 0.2× SSC, 0.1% SDS at 60°C for 20 min.
U3 cDNA and promoter region PCR products were radiolabeled using standard techniques. RNA end fragment probes were radiolabeled using T7 or SP6 RNA polymerase (Promega, Madison, WI). Riboprobes were routinely preannealed with 0.2 mg/ml (final) human C0t1 DNA (Gibco-BRL, Gaithersburg, MD) for 5 min at 65°C prior to hybridization in order to quench cross-hybridization of interspersed repeated sequences.
In order to analyze the individual U3 genes from the P1 and BAC clones PCR fragments were generated with the GM1/RNU3-3' primer pair (-242 to +461). All of the PCR-based subcloning was done using a thermostable polymerase mixture that included an editing activity (e.g. Pfu polymerase; Stratagene, La Jolla, CA). Fragments were subcloned using the TA cloning system (pCRII vector; Invitrogen). In addition, a 3.5 kb fragment containing the U3b2 gene was amplified from P1 2983 using a T7 vector primer in combination with the RNU3-3' primer and then TA subcloned. A 3.0 kb BamHI fragment containing the U3b1/b2 gene pair from 2983 as well as two NotI-KpnI end fragments from P1 clones 2981 (3.3 kb) and 2982 (5.0 kb) were isolated and subcloned by conventional techniques [pBluescript SK(-); Stratagene, La Jolla, CA]. Sequencing was accomplished using a Taq-based cycle sequencing system incorporating fluorescent dideoxy terminators and run on an automated ABI 377 sequencer (Applied Biosystems, Foster City, CA).
Metaphase spreads were prepared from human lymphocyte cultures by standard methanol:acetic acid fixation. P1 DNA was nick translated with biotin-dUTP and slides were hybridized to metaphase spreads and washed essentially as described (21 ). HeLa and HT-1080 (a human diploid fibrosarcoma) cells were grown in monolayers on chambered glass slides, pre-extracted with Triton X-100 and then fixed with paraformaldehyde as described (15 ). Slides were incubated with anti-p80 coilin antibodies, fixed, denatured and then hybridized with the biotinylated U3 probe as described (15 ). Slides were scored for co-localization of U3 DNA and p80 coilin fluorescence using a dual bandpass filter set (Chroma Tech., Brattleboro, VT). For each slide the numbers of red and green signals were counted as well as the number whose signals were at least partially overlapping. Digital imaging microscopy was performed as described (15 ).
In addition to the major spliceosomal snRNPs, CBs are thought to contain U3 snoRNP (22 ,23 ). We wanted to see if, like U1 and U2 genes, CBs associated with U3 loci. Previous studies had shown that there are roughly 5-10 functional U3 genes in the haploid human genome and that the sequences flanking the genes are highly conserved (10 ,24 ). Using a PCR fragment from the 5'-flank (essentially the upstream control region) as a probe for in situ hybridization, Mazan et al. (19 ) detected labeling only on human chromosome 17p11-p12. As part of our ongoing effort to identify genetic loci that interact with CBs, we isolated three P1 and four BAC clones containing U3 genes. Sequencing of PCR products or subclones thereof showed that each of them contained at least one U3 gene. The U3 coding regions on P1 clones 2981 and 2982 were identical, while some of the subclones derived from P1 2983 (as well as the BAC clones) differed at several positions, including some in the coding region. A schematic of the observed coding region changes is shown in Figure 1 . The regions flanking the U3 genes (both the 5'- and 3'-ends) in each of the clones were also extremely similar and were a close match to the sequence in the database (data not shown; accession nos AF020531, AF020534 and AF020535).
Consistent with the idea that U3 genes are clustered, we found that the clones all mapped to the same cytological location when used as FISH probes (Fig. 2 A and B). FLpter measurements (25 ) on multiple metaphases averaged between 19.8 and 24.5% of the length of the chromosome and are completely consistent with previous mapping studies that placed the RNU3 cluster at 17p11.2 (26 ). Once we had established that the clones contained U3 coding sequences and that they hybridized to the proper cytogenetic location, we assayed the frequency with which the hybridization signals overlapped with CBs in both HeLa and HT-1080 cells (15 ). Interestingly, RNU3 signals co-localized with at least one CB in ~54% of HeLa cells and 20% of HT-1080 cells (Fig. 2 C). Previous work showed that histone, U1 and U2 gene signals co-localized with CBs in a similar fraction of cells, while those of control loci did not (15 ). The slightly lower rate of association of CBs with U3 genes in HT-1080 cells is still significantly higher than that of the controls, especially given that the association rate of U2 genes with CBs in HT-1080 cells is also reduced (~75% in HeLa versus 35% in HT-1080; M.R.Frey and A.G.Matera, unpublished observations). This lower rate likely results from the fact that HT-1080 cells are diploid and have an average of only 1-2 CBs/cell, whereas HeLa cells are notoriously aneuploid and have 4-6 CBs/cell (see 15 for a discussion).
Our cytogenetic results (Fig. 2 A) and those of Mazan et al. (19 ) suggested that U3 genes are linked in the genome. However, Yuan et al. (10 ) isolated multiple independent [lambda] phage clones and found that no two genes were present on the same clone. Hence, they inferred that U3 genes must be at least 15-20 kb apart (10 ). Restriction mapping and Southern analysis of our P1 and BAC clones revealed that these large insert clones share extensive regions of homology (Fig. 3 ). PFGE showed that the three P1 clones were independent, while two of the BACs were sister clones to the other two (Fig. 3 ). The maps were constructed using a combination of U3 coding region and promoter probes as well as various end fragment probes generated via T7 or SP6 transcription (see Materials and Methods). Examples of these blots are shown in Figure 4 . In the first panel the three P1 clones were probed with a U3 cDNA (Fig. 4 A). When digested with MluI P1 clones 2981 and 2982 have only a single U3-positive band, whereas 2983 had three. All other enzymes tested showed only two bands for 2983. Analysis of the BAC clones showed that the L14/M14 sister clones had only two positive MluI fragments, despite otherwise similar restriction patterns (Fig. 4 B). Visual inspection reveals, however, that one band is roughly twice as dark as the other one (e.g. Fig. 4 B, lane 5). This suggested the existence of a nearly perfect inverted repeat. Fortuitously, one side of the inverted repeat was truncated within clone 2983, generating the third band (Fig. 3 ).
We next analyzed the regions around the U3 genes at higher resolution using a combination of conventional gel electrophoresis, restriction mapping, subcloning of end fragments and DNA sequence analysis. The results are summarized in Figure 5 . For clarity of discussion the solitary U3 gene in the contig has been designated U3a and the two inverted U3 genes were named U3b1 and U3b2 (arrowheads, Fig. 5 ). Note that the restriction patterns of the flanking regions of the U3a and U3b1/b2 genes is conserved for over 2.5 kb upstream; downstream similarity extends for at least 700 bp (Fig. 5 ). Subcloning of a 3.5 kb end fragment from 2983, containing only U3b2, was accomplished using a T7 vector primer and a 3'-RNU3 primer. Analysis of this clone proved that the U3b2 gene is oriented as illustrated in Figure 3 and that its sequence is different from that of U3a, which we obtained by sequencing multiple PCR-derived clones from 2981 and 2982 (Fig. 1 ).
The high degree of sequence similarity within the inverted U3b repeats is remarkable. Not only is the restriction pattern identical over a stretch of nearly 45 kb in each direction, but sequence analysis of selected regions within the repeat showed >98.5% similarity at sites at least 6 kb upstream of the U3 genes. Furthermore, pulsed field and conventional Southern analyses revealed that the repeat structure within the RNU3 locus is complex. In addition to the solitary U3a gene and the inverted U3b1/b2 gene pair there is very likely an additional U3b-like gene pair at the 5'-end of our contig (see Figs 2 , 6 and 7 ). The locations of the 49 and 77 kb bands are unknown, but likely to be nearby, since Mazan et al. (19 ) did not detect U3 labeling at additional chromosomal sites.
Our high resolution clone maps (Fig. 5 ) agree well with those of Reddy and co-workers (10 ,24 ). These authors identified four different phage clones that look like U3a (HU3-1-HU3-4) and one clone whose map corresponds to half of a U3b inverted repeat (HU3-178). The authors comment on the fact that the gene in HU3-178 was truncated by the cloning vector and may be the result of an artefact (10 ). We suggest that copies of the inverted repeat were extremely under-represented in the older style [lambda] library and thus only truncated versions of the U3b gene pair were obtained (10 ). Furthermore, our higher resolution maps agree well with previous genomic Southern analyses of the RNU3 locus (10 ,26 ).
It was particularly surprising to find differences among the coding sequences of the U3 genes. The sequences of the inverted U3b genes were identical, but differed at three positions from those of the published U3 gene and at six sites with respect to the U3a gene (Fig. 1 ). Similarly, the U3a gene differed at three other sites from the published U3 RNA sequence. None of the changes interfered significantly with base pairing within stems of the U3 secondary structure (Fig. 1 ), however, one of the mutations (U23G in U3b) was in the conserved box A sequence. Although these differences may be attributed to sequencing artefacts, multiple independent subclones amplified from different genomic clones (e.g. 2981 and 2982) had the same sequence. Interestingly, Yuan and Reddy (10 ) noted that the rat Novikoff hepatoma cell line had three different length isoforms of U3 RNA, whereas HeLa cell U3 migrated as a single band. The point mutations shown in Figure 1 should not alter the length of the U3 transcripts and thus are consistent with previous results. Regardless of whether these point mutations are real or not, these U3 genes all have extensive 5' and 3' homologies and thus are each presumably active genes.
The gene duplication(s) within the RNU3 locus is of recent evolutionary origin. Not only are the inverted repeat sequences nearly perfect (>98.5% similar), but the U3a and U3b genes share significant upstream (>2.5 kb) and downstream (>0.7 kb) similarities. Additionally, the mouse U3 gene cluster is located on chromosome 11, in a region that is known to be syntenic with human chromosome 17p (19 ). Three of the four U3 genes from the mouse rnu3 locus are located on the same cosmid clone and all are oriented in the same direction (27 ). Thus the U3 genes appear to evolve in concert (i.e. different U3 genes are more similar within a species than they are between species).
However, gene conversion events within the RNU3 locus are apparently not as frequent as those within the RNU2 locus on chromosome 17q21 (28 -30 ), since U2 genes within that array are essentially identical. Perhaps this greater variability among U3 sequences is due to their clustered (as opposed to tandem) repeat organization. Indeed, concerted evolution may well be a consequence of tandem repetition (29 ,31 ).
Early studies on amphibian oocyte nuclei revealed many morphologically distinct structures, including nucleoli, lampbrush chromosomes and `spheres'. The sphere organelles were first identified structurally, as knobs attached to the lampbrush chromosomes of the newt (32 , reviewed in 33 ). Of particular interest is the fact that some of the spheres within the amphibian oocyte are attached to the so-called `sphere organizer' (SOR) regions of the lampbrush chromosomes. These SORs are located at or near the three different histone gene clusters (17 ,18 ) on the chromosomes of both frogs and newts. Spheres and CBs each contain high concentrations of ribonucleoproteins and other RNA processing machineries (reviewed in 13 ,34 ). One of the most important links between CBs and spheres is that SPH-1, an integral protein component of spheres, shows significant structural homology to p80 coilin (35 ). Furthermore, Gall and co-workers (36 ) demonstrated that the human p80 coilin protein is targeted to spheres when human mRNA was injected into amphibian oocytes. Thus spheres and CBs are homologous organelles that are likely to have similar functions, but to date no functions have been demonstrated (12 ).
We have shown that the RNU3 locus preferentially associates with CBs in interphase human cells. All of the human CBOR loci identified thus far (U1, U2, U3 and histone) are members of repetitive gene families. What factors nucleate CB formation? Is it an RNA- or DNA-based event? It is possible that, like nucleoli and rRNA genes, CBs may form as a consequence of transcription of these clustered genes. It is important to note that Sm and trimethylguanosine epitopes are present in CBs, hence, they are likely to contain mature, rather than nascent, snRNAs (13 ). Thus the analogy of NORs and CBORs is incomplete, since nucleoli form in direct response to transcription of rDNA (37 ), whereas at least some of the snRNAs within CBs are thought to travel to the cytoplasm for snRNP assembly prior to re-import into the nucleus (38 ). In the case of the histone gene-CB association it seems that CBs have the additional function of recruiting U7 snRNA, since the Sm site of U7 is required to specifically target this RNA to CBs following microinjection into Xenopus oocytes (39 ).
Moreover, p80 coilin may be involved in cellular processes affecting many other genes that are not recognized as CBORs. Perhaps the only reason we detect such high concentrations of p80 coilin near the CBORs is because these genes are clustered. Current efforts are underway in the laboratory to characterize additional single and multicopy snRNA gene loci and to ask whether they associate with CBs.
We thank W.Wu for PCR screening of the BAC clones and E.Chan for anti-p80 coilin antibodies. This work was supported by NIH grant GM-53034 (to A.G.M.). A.G.M. was also supported by a Junior Faculty Research Award from the American Cancer Society (JFRA-570).
REFERENCES