INTRODUCTION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERNCES

Coiled bodies (CBs) are nuclear organelles whose precise functions are unknown. They are highly enriched in small nuclear ribonucleoproteins (snRNPs) and several basal transcription factors and have thus received a great deal of recent interest (for review, see Lamond and Earnshaw, 1998; Matera, 1998). Evidence that CBs and their twin structures, called gems, are involved in aspects of snRNP biogenesis and maintenance has intensified this interest, especially given that defects in this pathway result in the genetic disease spinal muscular atrophy (Fischer et al., 1997; Liu et al., 1997; Pellizzoni et al., 1998; for review, see Matera, 1999).

Although CBs are now clearly implicated in snRNP biogenesis, several lines of evidence suggest that they are multifunctional organelles. First, CBs contain many apparently disparate cellular components in addition to the aforementioned snRNAs and associated Sm protein epitopes, including certain nucleolar proteins and small nucleolar RNPs, histone pre-mRNA processing factors, and members of the transcriptional apparatus (Matera, 1998). Second, CBs preferentially associate with specific DNA loci within mammalian nuclei and amphibian oocytes, including genes that encode U1, U2, and U3 snRNAs as well as the cell cycle-regulated histone loci (Gall et al., 1981; Callan et al., 1991; Frey and Matera, 1995; Smith et al., 1995; Gao et al., 1997; Schul et al., 1998). If CBs and gems are simply involved in snRNP assembly (and possibly regeneration; Pellizzoni et al., 1998), why do they contain basal transcription factors and associate with snRNA and histone genes?

We suggest that CBs may also participate in regulating snRNA levels by an autogenous feedback loop (Frey and Matera, 1995; Frey et al., 1999) in which a fraction of mature (or partially mature) U snRNPs return, salmon-like, to CBs adjacent to the sites of snRNA synthesis and regulate transcription. Furthermore, we have shown that association of human U2 genes (the RNU2 locus) with CBs is mediated by nascent U2 RNA (Frey et al., 1999). Based on these data, we hypothesized that when snRNP concentrations are high, the resulting association of the CB with nascent U snRNA would then result in transcriptional attenuation or stalling. Conversely, when snRNP concentrations within the cell are low, CBs would no longer associate with the genes, and transcription might even be up-regulated. Although many mechanisms can be envisioned, it is plausible that excess snRNP-specific proteins (in the absence of the mature snRNAs) might bind directly to the nascent snRNA transcripts and block their interaction with CBs (Frey et al., 1999).

Previously characterized CB-associated loci, including the U1, U2, and U3 genes (RNU1, RNU2, and RNU3 loci, respectively) as well as the two major histone gene loci (HIST1 and HIST2), are either tightly clustered or tandemly repeated in the human genome (Frey and Matera, 1995; Gao et al., 1997). In addition to being clustered, these genes share a number of common features. For example, neither snRNA nor histone genes contain introns, and their transcripts are not polyadenylated. The snRNA genes are transcribed from a special class of non-TATA promoters. Furthermore, both histone mRNAs and snRNAs have conserved, 3'-terminal stem-loop structures. But for the presence of an ORF, histone mRNAs are structurally quite similar to snRNAs. In light of these results, we wondered whether genomic organization plays any role in the association frequency of genes with CBs. We therefore wanted to ask whether single-copy snRNA genes could also preferentially associate with CBs.

To investigate this issue, we isolated and mapped the chromosomal locations of bacterial artificial chromosome (BAC) clones containing U4, U6, U7, U11, and U12 snRNA loci by radiation hybrid (RH) and fluorescence in situ hybridization (FISH) analysis and found that they are dispersed among multiple human chromosomes. Each of these loci is single copy, except for U4, which has two copies. We then examined whether these genes associated with CBs and found association values lower than those of multicopy U1 and U2 genes, but higher than other single-copy control loci such as collagen type I (COL1A1), c-myc, makorin, and nerve growth factor receptor (NGFR). To put these observations on a more quantitative basis, we developed a theoretical model that generated random colocalization values in which the frequency of association was solely dependent on the respective areas of the gene signals and those occupied by CBs. This model was used to generate expected values for ² tests against the observed values for each locus. Three of the snRNA loci (RNU4, RNU11, and RNU12) were determined to preferentially associate with CBs, whereas no significant association was found for the RNU6, RNU7, or the control genes.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERNCES

Isolation of BAC Clones and RH Mapping

Primers were constructed and used to screen a human BAC library (Research Genetics, Huntsville, AL) and RH panels (Genebridge 4, Research Genetics; Hudson et al., 1995). Sequence-tagged site mapping data for each of these loci were created and deposited in GenBank using the following primer pairs: RNU4 (5'-AATACGGCTGGTGGAGTGGGAACA-3' and 5'-TCGCGCCTCGGATAAACCTCATT-3'), accession number AF114984; RNU6 (5'-CGAATTTGCGTGTCATC-3' and 5'-AGGTCGGGCAGGAAGAG-3'), AF114985; RNU7 (5'-CGCGAACTCTAGAAATGAATGAC-3' and 5'-TGCTGCGTATGTCTTTGGAG-3'), AF114986; RNU11 (5'-TGGCTAGGGGTGGCACAAGATACA-3' and 5'-GTCGATTCCGCACGCAGAGCA-3'), AF114982; and RNU12 (5'-CCGCTAGGGAGCGACGAACTAC-3' and 5'-TTCGTGGGTCACAACGTCAA-TAC-3'), AF114983. The following clone addresses were identified: RNU4, 189A14; RNU6, 330B12; RNU11, 190M11; and RNU12, 133E16. Each of the loci was then mapped using the Genebridge 4 RH panel, except the RNU7 primer pair, which was mapped on the higher resolution Stanford G3 RH panel. The complete sequence of a P1 artificial chromosome (PAC) clone containing the U7 gene had already been deposited in the database as part of a large-scale genome-sequencing project (Anisari-Lari et al., 1997).

PCR reactions consisted of 25 ng of each hybrid, 20 mM Tris, pH 8.3, 2.5 mM MgCl₂, 50 mM KCl, 0.2 mM dNTPs, 0.8 µM primers, and 0.5 U of Taq polymerase in a 10 µl-volume. After a 1-min denaturation at 94°C, 30 cycles of 94°C (20 s), a 30-s annealing (initial annealing at 62°C, final annealing at 47°C), and 72°C extension (45 s, final extension for 5 min), most reactions were performed using a "touchdown" protocol on a PTC-100 thermocycler (MJ Research, Watertown, MA). A 2% agarose gel was used to detect the presence or absence of the respective PCR products. Data vectors for each of the primer pairs are available through their accession numbers.

Histone Processing Assay

Oocytes were injected in the cytoplasm with 35 nl of H₂O or a control oligonucleotide directed against Xenopus heterogeneous nuclear RNP (hnRNP) A1 mRNA (5-TCGCCCATCCACTTT-3') or the anti-frog U7 snRNA (5'-AAGAGCTGTAACACTT-3') oligonucleotide (2 ng/nl) and incubated for 4 h at 18°C. Oocytes were subsequently injected in the nucleus with a plasmid bearing the mouse H2a614 reporter gene (225 pg of DNA for each plasmid per nucleus). All oocytes were coinjected either with the plasmid pTT005 (containing the putative human U7 snRNA gene) or a control pGEM7zF plasmid and incubated at 18°C. Eighteen hours later, RNA was extracted as previously described (Williams et al., 1994) and assayed by S1 nuclease analysis for processing of the mouse H2a614 mRNA (Wang et al., 1999). The percentage of processed histone mRNA was quantified on a PhosphorImager (Storm 840; Molecular Dynamics, Sunnyvale, CA).

Metaphase FISH and Image Acquisition

Human metaphase spreads were prepared from normal peripheral blood according to established procedures. BAC and PAC clones containing elements of human U4, U6, U7, U11, and U12 genes were labeled with biotin-16-dUTP (Boehringer Mannheim, Indianapolis, IN) by nick translation. Approximately 100 ng of each labeled clone were then ethanol precipitated with 2 µg of human Cot1 DNA (Life Technologies, Gaithersburg, MD) and 8 µg of salmon sperm DNA per metaphase slide. The DNA was then dissolved in 50% formamide and 2× SSC; after an overnight incubation at 37°C, slides were washed (three times for 5 min each) in 50% formamide and 2× SSC at 42°C and then in 1× SSC at 60°C. Fluorescein-conjugated avidin (Vector Laboratories, Burlingame, CA) was used to detect the hybridization signals. A DAPI (Boehringer Mannheim) counterstain generated the G/Q banding pattern.

Images were obtained using a Zeiss (Thornwood, NY) Axioplan epifluorescence microscope equipped with a cooled charge-coupled device (CCD) camera (Photometrics, Tucson, AZ). The 16-bit source images were stored as normalized 8-bit gray scale data files by the software program CCD Image Capture (Yale University, New Haven, CT). Highly plane-parallel bandpass filters maintained proper image registration (Ballard and Ward, 1993). Gene Join (Yale University) and an Apple Macintosh computer were used to merge and pseudocolor the images. Finished color prints were produced using Adobe Photoshop 2.5.1 (Adobe Systems, Mountain View, CA) and a dye sublimation printer (Codonics, Middleburg Heights, OH).

Interphase FISH Analysis and Cell Scoring

HeLa cells were grown in a monolayer on chambered slides (Nunc, Rochester, NY), prepermeabilized in Triton X-100, and fixed with paraformaldehyde as described (Frey and Matera, 1995). The cells were then incubated with anti-p80 coilin antibody (to mark CBs), fixed again, and hybridized with a biotinylated (Boehringer Mannheim) gene probe as described (Frey and Matera, 1995).

Scoring of slides for colocalization of genes and CBs was performed using a dual-bandpass filter set (Chroma Technologies, Brattleboro, VT). Signals that were either immediately adjacent or coincident to CBs were scored as positive for association. The number of cells in each set that had none, one, two, or three or more CB-signal associations was recorded. The number of CBs and probe signals per cell was also noted.

Nuclear and nucleolar areas were determined by measuring x and y orthogonal diameters of the DAPI images in the case of nuclei or the DAPI-negative areas in the case of nucleoli and calculating areas according to the formula for an ellipse: A_ellipse = (D_x/2) × (D_y/2). The resultant values were averaged for >100 nuclei or nucleoli. CB areas and signal diameters were determined similarly using anti-coilin or FISH locus signals.

Statistical Analysis

The number of CBs and probe signals per cell was averaged for each set of experiments. A theoretical model of random colocalization between the CBs and probe signals using these average values was used. Treating the HeLa nucleus as a two-dimensional area within which one or more CB-sized areas and gene signal-sized areas can randomly colocalize yields a colocalization probability proportional to their areas. Our model uses empirical measurements of the areas of HeLa cell nuclei, nucleoli, and CBs to calculate the probability of colocalization within a sample of nuclei.

The probability of a single CB colocalizing with a single FISH signal area is P_{(colocalization given c = 1, s = 1)} = A_cb/A_N, where c is the number of CBs, and s is the number of locus signals. A_N is equivalent to the nuclear area minus the nucleolar area, and A_cb is the effective area available for colocalization between a CB and a signal according to the formula A_cb = (r_cb + d_s)², where r_cb is the radius of a CB and d_s is the diameter of a locus signal. This allows the effective area of a CB to include the zone around the CB within which a signal may occur and still colocalize with that CB, permitting the probe signals to be treated as points for the purpose of calculating colocalization probabilities. For more than one CB, the areas can be summed, yielding P_{(colocalization given s = 1)} = cA_cb/A_N, as long as c  (A_N/A_cb).

For more than one signal, the random colocalization probablity is more complex. The absence of multiple signals associated with the same CB in our present data set suggests the constraint that only one signal may associate with one CB. The probability of the first gene signal-CB colocalization in a nucleus is as derived above: P_{(first colocalization)} = cA_cb/A_N. The probability of the second colocalization will be lower, because one fewer CB area is available for colocalization: P_{(second colocalization)} = (c  1)A_cb/A_N. For the nth colocalization, P_{(nth colocalization)} = (c  n)A_cb/A_N, for 0 > c > n; i.e. the number of gene signals is less that the number of CBs. This is consistent with the observed data.

Instead of calculating the probability of each individual colocalization in this way, the probability of a nucleus having one or more colocalizations between CBs and locus signals was derived as follows. The probability of zero colocalizations given one signal is P_{(no colocalizations, s = 1)} = 1  P_{(first colocalization)}, and for s signals, P_{(no colocalizations)} = (1  P_{(first colocalization)})^s. For a given nucleus with s signals and c CBs, the probability of that nucleus having one or more colocalizations is therefore P_{(one or more)} = 1  P_{(no colocalizations)}, or P_{(one or more)} = 1  (1  P_{(first colocalization)})^s. This not only reduces the colocalization probability calculation to a simple binomial distribution but also provides a more conservative test for significance and reduces the effect of the sparse multiple colocalizations in the data set.

A test was constructed to obtain significance levels for the experimental outcomes. The random colocalization probabilities generated according to the formulae above for P_{(one or more)} and P_{(no colocalizations)} were used to generate expected numbers of nuclei with one or more colocalization and expected numbers of nuclei with no colocalizations. For the observed data sets, data were grouped into the number of nuclei per set with one or more colocalizations and the number with no colocalizations. These values were compared with the expected values using a ² test with 1 df. A significance level of 0.01 was chosen to provide a stringent test of significance, indicating that there was a <1% chance that those experimental outcomes found to be significant were actually due to random colocalization. As a further test of the model, we ran a computer simulation of random gene signals and CBs localizing within a nucleus and found that the statistical tests agreed well with the theoretical calculations (see RESULTS).

	RESULTS

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERNCES

Isolation and Chromosomal Localization of Dispersed Single-Copy snRNA Loci

To ask whether single-copy snRNA genes also associate with CBs, we first needed to obtain suitable, large-insert FISH probes. To date, only small plasmid clones containing bona fide U4 (Bark et al., 1986), U6 (Kunkel and Pederson, 1988), U11 (Suter-Crazzolara and Keller, 1991), and U12 (Tarn et al., 1995) genes had been reported in the literature. We therefore used sequence information from these clones to generate PCR primer pairs and screened a human BAC library. These same primer pairs were used to perform RH mapping with the Genebridge 4 panel (Hudson et al., 1995). The results of these RH and cytogenetic mapping experiments (performed in parallel) are shown in Figure 1 and are completely consistent with each other. We found that the two U4 genes mapped to 12q23-24.1, the U6 gene to 15q21-22, U11 to 1p34.2-34.3, and U12 to 22q13. The cytogenetic locations of previously identified CB associated loci are shown for comparison.

View larger version (32K): [in this window] [in a new window]   Figure 1.   Chromosomal locations of snRNA and histone genes used in this study. Map locations of the single-copy snRNA genes (determined by RH and cytogenetic mapping) as well as those of previously characterized snRNA and histone loci are shown. Bars to the right of the ideograms show the range of FISH signals. The nearest markers from the RH mapping are shown to the right of the chromsomes; MIT framework markers are shown in bold.

Although clones corresponding to functional human U4, U6, U11, and U12 genes have been described in the literature (see above), no such genomic sequence has been reported for U7 genes. Nagafuchi et al. (1994) noted that a nearly identical copy of a U7 snRNA sequence was located ~1.3 kb downstream of the dentatorubral and pallidolusian atrophy (DRPLA) gene on chromosome 12p13. This sequence differed by only a single base pair from the reported U7 RNA sequence (Mowry and Steitz, 1987), and its genomic location is syntenic to the region of mouse chromosome 6 that is known to contain the bona fide mouse U7 gene (Turner et al., 1996). Subsequently, we noted that a large-scale genome-sequencing project had sequenced this region of chromosome 12 (Anisari-Lari et al., 1997). Alignment of the regions surrounding the human (accession number U47924) and mouse (X54165) U7 sequences (Figure 2) revealed that the human sequence was unlikely to be a pseudogene, because it contains significant upstream similarity, including recognizable snRNA transcriptional control elements (proximal and distal sequence elements). Thus using a subcloned 3.4-kb BamHI fragment from this region (a kind gift from M. Yamada, National Children's Medical Research Center, Tokyo, Japan), we decided to test whether the human U7 gene was functional in a histone pre-mRNA processing assay.

View larger version (33K): [in this window] [in a new window]   Figure 2.   Sequence comparison of human and mouse RNU7 loci. The U7 genes at human 12p13 (accession number U474924) and mouse chromosome 6F (X54165) along with flanking regions are compared. Note the relatively high degree of upstream similarity, including the functionally conserved control regions called the proximal and distal sequence elements (PSE and DSE, overlined). The human U7 snRNA coding region is also overlined. A single-nucleotide difference between the published RNA sequence (accession number M17910) and the human 12p13 sequence is illustrated. Downstream regions, with the exception of a putative 3' box terminator sequence (overlined), are not conserved.

The Human U7 Gene on 12p13 Encodes a Functional snRNA

Xenopus oocytes were used as a model system for in vivo processing of histone pre-mRNA (Williams et al., 1994; Wang et al., 1999). Unprocessed RNAs are stable in oocytes (Bentley and Groudine, 1988; Middleton and Morgan, 1990), and quantitative results from a processing assay can be generated. Previous results demonstrate that oocytes are competent to process histone pre-mRNA transcribed from the mouse histone H2a-614 gene when injected into the nucleus (Williams et al., 1994). In addition, antisense oligonucleotides directed against Xenopus U7 snRNA have been shown to effectively reduce histone mRNA processing in the oocyte (Scharl and Steitz, 1996). Antisense oligonucleotides will bind to the target snRNAs, and the DNA-RNA hybrid is destroyed by endogenous RNase H (Pan and Prives, 1988).

Oocytes were injected with water, a U7 snRNA antisense oligonucleotide, or a control oligonucleotide targeted to hnRNP A1 mRNA. Subsequently, oocytes were injected with the mouse H2a-614 gene and either a control plasmid (pGEM7zf) or plasmid pTT005, containing the putative human U7 snRNA gene. Eighteen hours later, oocytes were analyzed for the unprocessed and processed mouse H2a-614 RNA using an S1 nuclease assay. The efficiency of processing of histone pre-mRNA in this batch of oocytes was 60-65% (Figure 3A, lanes 2 and 4). Oocytes treated with anti-U7 oligonucleotides processed histone pre-mRNA much less efficiently (Figure 3A, lane 6). In contrast, oocytes injected with the putative human U7 gene processed histone pre-mRNA with >90% efficiency (Figure 3A, lanes 3, 5, and 7). Even the oocytes treated with U7 antisense oligomers were able to process histone pre-mRNA efficiently when supplemented by injection of the human U7 snRNA gene (Figure 3A, compare lanes 6 and 7). Similar results were obtained with oocytes from several frogs (our unpublished results). We conclude that a U7 snRNA is transcribed from pTT005 and that this snRNA assembles with endogenous proteins in the oocyte to form a functional U7 snRNP.

View larger version (51K): [in this window] [in a new window]   Figure 3.   The human U7 sequence at 12p13 encodes a functional snRNA. (A) Xenopus oocytes were injected with H20 (lanes 2 and 3), anti-hnRNP A1 mRNA control oligonucleotide (lanes 4 and 5), or anti-U7 snRNA oligonucleotide (lanes 6 and 7). Four hours later, oocytes were injected with either a control plasmid and the mouse H2a-614 gene (lanes 2, 4, and 6) or plasmid pTT005, containing the putative human U7 gene (hU7) and the mouse H2a-614 gene (lanes 3, 5, and 7). Eighteen hours later, total RNA was prepared from oocytes and analyzed by S1 nuclease mapping using an assay that detects both processed (Proc) and unprocessed (Unproc) histone mRNA. NS, nonspecific transcript derived from the hU7 plasmid. Lane 1, pUC18 DNA digested with HpaII. (B) Schematic of the S1 nuclease assay used in A. The probe contains 34 nucleotides beyond the end of the histone mRNA followed by plasmid sequences that are not present in the transcript from the mouse histone H2a-614 gene.

Single-Copy snRNA Genes Associate with CBs Less Frequently than Their Multicopy Counterparts

Using an interphase FISH and cell-scoring protocol developed previously (Frey and Matera, 1995), we analyzed the distributions of U4, U6, U7, U11, and U12 snRNA genes with respect to CBs in unsynchronized HeLa-ATCC cells (Table 1). If the CB and gene signals were found to overlap, they were scored as a colocalization event. Examples of typical CB colocalizations with single-copy snRNA gene loci are shown in Figure 4. We also scored a number of control loci, including the NGFR, HPRT, c-myc, COL1A1, makorin, and 5S rRNA (RN5S) genes (Table 1). Previous studies have shown that only transcriptionally active U2 gene constructs associate with CBs (Frey et al., 1999). Therefore, the only stipulation we made regarding choice of the control genes is that they are single copy (with the exception of RN5S) and that their transcripts are expressed in HeLa cells (with the exception of NGFR).

View larger version (42K): [in this window] [in a new window]   Figure 4.   Single-copy snRNA loci also colocalize with CBs in interphase HeLa cells. Nuclei were stained with DAPI (blue); CBs are displayed in red; and snRNA gene signals are in green. In A (RNU11) and B (RNU12), BAC clones corresponding to the appropriate locus were used to generate the FISH signals; CB immunofluorescence was performed, and cells were scored as described previously (Frey and Matera, 1995). In each panel, a typical colocalization event is shown; the full data set is presented in Table 1.

To develop a locus association test, empirical measurements of the areas of HeLa cell nuclei, nucleoli, and CBs were used to calculate the probability of colocalization within a sample of nuclei. The probability of a single CB colocalizing with a single FISH signal is proportional to the respective areas of the two structures and the nuclear area, as diagrammed in Figure 5. In our data set, all the FISH signals were excluded from nucleoli. Thus the effective nuclear area is equivalent to the area of the nucleus minus that of the nucleoli. Conversely, the effective area of the CB is actually larger by one gene signal diameter (Figure 5), because the two objects (CBs and gene signals) must touch to be scored positively for association (see MATERIALS AND METHODS for additional details). Therefore, the probability that a random gene would colocalize with at least one of n CBs is proportional to the effective nuclear areas of all the CBs in the cell, divided by the effective nuclear area (i.e., the nucleoplasm).

View larger version (14K): [in this window] [in a new window]   Figure 5.   Diagram showing a nucleus with CBs, locus signals, a nucleolus, and a CB-locus association. The dashed line around the CB represents the effective area of a CB that is available for colocalization if locus signals are considered points. The nucleoplasmic area is essentially equal to the area of the nucleus minus the nucleolar area. Average values were nuclear area, 133.7 µm2; nucleolar area, 10.9 µm2; nucleoplasmic area (AN), 123 µm2; and effective CB area (Acb), 0.66 µm2.

Using these parameters, we generated expected values of nuclei with one or more random associations and used them for a ² test against experimental data. As expected, the association of CBs with the previously characterized histone and snRNA loci was highly significant (with p values in the 1.0E-30-1.0E-200 range; see Table 1). Interestingly, we found that among our test probes, the RNU4, RNU11, and RNU12 loci also displayed significant association frequencies (Table 1). However, neither the RNU6 and RNU7 loci nor the control gene loci met the 0.01 stringency threshold. Taken together with the RN5S experiments, the RNU6 data imply that pol III genes do not preferentially associate with CBs (albeit with several caveats; see DISCUSSION).

Technical Considerations

Note that our theoretical colocalization model treats the nucleus and the structures within as two-dimensional areas. This two-dimensional model closely approximates the actual nonconfocal image analysis of the experimental data and is appropriate for relatively flattened adherent cells such as HeLa. Taking into account the depth of field of our objective (~1 µm), the volume sampled is approximately one-third to one-half the volume of a HeLa cell. Although area is not generally a good approximation of volume, by constraining the statistical analysis to two dimensions, we actually make the association test more stringent. Random (expected) colocalizations would actually be less frequent if the genes were given more room to roam and would lower the threshold of significance. Thus the two-dimensional approach gives a more conservative picture of CB associations, raising the threshold and detecting only the most significant interactions. We, therefore, cannot completely exclude the possibility that other, less frequent, associations are not of biological interest. It is also noteworthy that if volume measurements are used instead of area, the model can be adapted both to other cell shapes and to alternate image analysis methods such as three-dimensional confocal microscopy.

Significant deviations from the random model could theoretically include both cases of more and fewer nuclei with colocalizations than expected. We therefore used a two-tailed ² distribution but found no cases of significantly fewer colocalizations than expected at the 0.01 stringency level. Makorin and RN5S genes, however, met the 0.05 stringency level with "negative" p values of 0.022 and 0.012, respectively. These genes may be partitioned into regions of chromatin that do not usually come into contact with CBs. Similar negative results have been obtained with various centromeric (alphoid) DNA probes (Smith et al., 1995; Schul, 1998).

The NGFR locus did not meet our stringent criteria for association but was significant at the 0.05 level with a p value of 0.014. As mentioned previously (Frey and Matera, 1995), the elevated CB association frequency of NGFR may simply be attributable to its close proximity (~6-7 Mb; Deloukas et al., 1998) to the RNU2 locus. However, the COL1A1 locus is located ~200 kb proximal to NGFR (i.e., slightly closer to RNU2), and this gene did not significantly associate with CBs (Table 1).

	DISCUSSION

TOP ABSTRACT INTRODUCTION MATERIALS AND METHODS RESULTS DISCUSSION REFERNCES

We wanted to determine whether, like their multicopy U snRNA brethren (Frey and Matera, 1995; Smith et al., 1995; Gao et al., 1997; Schul et al., 1998), single-copy snRNA genes might also preferentially associate with CBs in interphase HeLa cells. To answer this question we first had to isolate BAC clones and map the chromosomal locations of bona fide human U4, U6, U11, and U12 snRNA genes. We found these genes to be dispersed among four different chromosomes (RNU4, 12q23-24.1; RNU6, 15q21-22; RNU11, 1p34.2-34.3; and RNU12, 22q13). Furthermore, database analysis identified a putatively functional U7 gene on chromosome 12p13 (Anisari-Lari et al., 1997), whose map location is syntenic to the known functional RNU7 locus in mouse (Turner et al., 1996). We therefore tested the ability of transcripts from this gene to process histone messages in a Xenopus oocyte expression system and found that this gene is completely functional. An analysis of the interphase distributions of these gene loci with respect to CBs revealed that only the CB colocalization frequencies of the U4, U11, and U12 genes were significant when compared with a random model (Table 1).

We conclude that although clustered snRNA genes colocalize more frequently with CBs than do their single-copy counterparts, the interaction does not depend on the genomic organization of the gene. However, the interaction does appear to depend on the type of RNA polymerase. Both U6 snRNA and 5S rRNA genes (transcribed by pol III) do not associate with CBs, whereas those for U1, U2, U3, U4, U11, and U12 (pol II) are each preferentially localized adjacent to CBs. Taken together with our previous findings that the association of snRNA genes with CBs is mediated by the nascent snRNA transcripts (Frey et al., 1999), these data highlight the possibility that CBs may associate with snRNA genes through interactions with the RNA polymerase II complex bearing the nascent transcript and not simply with the nascent snRNA itself. Clearly, additional experiments will be required to answer this question definitively.

Finally, we are left with a dilemma. Does a statistical preference for colocalization of a given gene with CBs reflect a functional interaction or is it merely fortuitous? The diversity of genes that exhibit such associations (Figure 1), the functional similarities between the products of these genes (they encode small RNAs involved in RNA processing or intronless mRNA precursors that are processed by small RNAs), and the phylogenetic conservation of these interactions from amphibians to mammals (Frey et al., 1999; our unpublished observations) all argue that CBs talk to these genes for a biologically important reason.

Genomic Caveats

Although uncovering the precise nature of these CB/gene associations will be the subject of future investigations, there are also a number of unknown factors to be considered when interpreting these statistical data. For example, we do not know whether the U6 gene we have identified on chromosome 15q21-22 is the major site of U6 synthesis, even though constructs of the gene are active in vitro (Kunkel and Pederson, 1988), and it remains the only bona fide U6 gene in the database. Thus the reason that this gene does not associate with CBs may not be due to the type of polymerase involved but because it is not highly transcribed. Similarly, the U7 gene on 12p13 also may not be the major site of U7 synthesis. However, several lines of evidence suggest that, in fact, this sequence is the major U7 gene. First, its genomic location is syntenic with the mouse U7 gene (Turner et al., 1996; Nadeau and Sankoff, 1998). Second, despite the fact that the coding region contains a single-nucleotide difference (Figure 2) from the published U7 RNA sequence (Mowry and Steitz, 1987), a subsequent study found that this nucleotide position (49) was a U in five isolated U7 cDNAs (Yu et al., 1996). The original sequencing of the 3' end of U7 was done by RNase A and T₁ mapping and could easily be erroneous. Nevertheless, although there may be multiple active U7 genes in the human genome, the locus at 12p13 does not significantly associate with CBs.

There are ~100 copies of the U4 sequence (including pseudogenes) in the human genome (Bark et al., 1986). Although the two U4 genes contained within our BAC clone are the only known active members of the U4 gene family in the database, the locus on 12q23-24.1 does not correspond to the major U4 locus in human genomic DNA (Bark et al., 1986). The major locus remains uncloned, although based on the conservation of flanking sequences detected by genomic blotting, it seems probable that U4 genes are also tightly clustered (Bark et al., 1986). Based on the CB association frequency observed for the RNU4 locus on 12q (Table 1), it seems likely that these genes are active. It would also be worthwhile to isolate the major U4 cluster and assay both its level of expression and frequency of colocalization with CBs.

CB Biogenesis and Functions

CBs and interchromatin granule clusters both contain mature U snRNPs, yet there are many interchromatin granule clusters but only a few CBs. What limits the number of CBs, and where do they form? One attractive hypothesis is that the CB-associated genes nucleate CBs, just as rDNA (the "nucleolus organizer region") nucleates nucleoli. If CB-associated genes nucleate CBs, why are some CBs attached to specific loci, whereas others appear to be free in the nucleoplasm? The facts that CBs do not accumulate newly synthesized RNA (Fakan and Bernhard, 1971; Callan and Gall, 1991; Jordan et al., 1997; Schul et al., 1998) and that they can form in the complete absence of cognate genomic DNA (Bell et al., 1992; Bauer et al., 1994) argue against this model. Our data demonstrating association of CBs with multiple different snRNA loci are more compatible with spontaneous assembly of CBs, perhaps in nucleoli (Lyon et al., 1997; Issac et al., 1998; Sleeman et al., 1998) followed by RNA-mediated association with the genes.

The dynamic nature of CBs is well described (Andrade et al., 1993; Carmo-Fonseca et al., 1993; Ferreira et al., 1994). One interesting idea that has yet to be explored is that association with the genes may actually stabilize CBs. We suggest that active snRNA genes may be tethered to CBs through multiple weak interactions with their nascent transcripts. Cognate "receptor" molecules would necessarily be localized within or adjacent to CBs. Presumably, the higher transcriptional capacities of multicopy snRNA and histone gene clusters would thereby increase the chances of association with CBs relative to single-copy genes. Furthermore, CBs are capable of simultaneously associating with multiple chromosomal loci (Frey et al., 1999). Regardless of whether CBs are stabilized by interaction with nearby genes, the fact that CBs accumulate high concentrations of mature snRNPs supports the existence of an autogenous feedback loop (Frey and Matera, 1995) through which CBs could participate in coordinate regulation of the expression of several different gene families.