A novel brain-specific box C/D small nucleolar RNA processed from tandemly repeated introns of a noncoding RNA gene in rats.

Antisense box C/D small nucleolar RNAs (snoRNAs) guide the 2'-O-ribose methylations of eukaryotic rRNAs and small nuclear RNAs (snRNAs) through formation of a specific base pairing at each RNA methylation site. By analysis of a box C/D snoRNA cDNA library constructed from rat brain RNAs, we have identified a novel box C/D snoRNA, RBII-36, which is devoid of complementarity to rRNA or an snRNA and exhibits a brain-specific expression pattern. It is uniformly expressed in all major areas of adult rat brain (except for choroid plexus) and throughout rat brain ontogeny but exclusively detected in neurons in which it exhibits a nucleolar localization. In vertebrates, known methylation guide snoRNAs are intron-encoded and processed from transcripts of housekeeping genes. In contrast, RBII-36 snoRNA is intron-encoded in a gene preferentially expressed in the rat central nervous system and not in proliferating cells. Remarkably, this host gene, which encodes a previously reported noncoding RNA, Bsr, spans tandemly repeated 0.9-kilobase units including the snoRNA-containing intron. The novel brain-specific snoRNA appears to result not only from processing of the debranched lariat but also from endonucleolytic cleavages of unspliced Bsr RNA (i.e. an alternative splicing-independent pathway unreported so far for mammalian intronic snoRNAs). Sequences homologous to RBII-36 snoRNA were exclusively detected in the Rattus genus of rodents, suggesting a very recent origin of this brain-specific snoRNA.

uration involves a large number of small nucleolar ribonucleoprotein particles (snoRNPs), 1 which can be defined as a metabolically stable association between small nucleolar RNA (snoRNA) and a specific set of proteins (1). All snoRNAs to date (except the RNA component for RNase MRP) fall into two major classes, box C/D and box H/ACA snoRNAs, based on the presence of short consensus sequence motifs (2). Whereas a few of them are involved in pre-rRNA nucleolytic cleavages, most box C/D and box H/ACA snoRNAs play a key role in specifying the two major types of rRNA nucleotide modification (2Ј-O-ribose methylations and pseudouridylations, respectively) through formation of a specific RNA duplex at the modification site (3)(4)(5)(6). While function(s) of these nucleotide modifications are still elusive they are proposed to fine tune rRNA folding and interactions with ribosomal proteins. Thereby, they could play a role in the biogenesis and activity of mature cytoplasmic ribosomes, particularly with regard to their peptidyl transferase activities (7). Antisense box C/D snoRNAs contain conserved sequence motifs box C (5Ј-RUGAUGA-3Ј) and box D (5Ј-CUGA-3Ј), generally brought together by a short 5Ј-3Ј-terminal stem, and one (sometimes two) antisense element(s) (8). Each antisense element, which exhibits an extended (10 -21nt) complementarity to a site of rRNA ribose methylation, is located immediately upstream from box D (or another copy of this motif, box DЈ, in the 5Ј half of the snoRNA). In the corresponding snoRNA/rRNA duplexes, the 2Ј-O-ribose methylated nucleotide is systematically paired to the fifth nucleotide upstream from box D (or DЈ) (9,10). Expression of an artificial box C/D snoRNA carrying an appropriate antisense element is sufficient to target a novel ribose methylation on the predicted pre-rRNA nucleotide and also, to a lesser extent, to RNApolymerase II transcripts (4,11), showing that the antisense element associated with box D (or DЈ) is the sole determinant of the site of methylation. Antisense box C/D snoRNAs able to direct the formation of three distinct ribose methylations in U6 and U5 snRNA have been reported (12)(13)(14) as well as several other box C/D snoRNAs likely to represent bona fide guides for additional snRNA ribose methylations in U6, U2, and U4 snRNAs (15). Moreover, several novel mammalian box C/D snoRNAs devoid of complementarity to rRNA or snRNA remain without a presumptive RNA target (15,16). Finally, the intriguing detection of one H/ACA-and three C/D-box snoRNAs exclusively expressed in the brain and also devoid of rRNA or snRNA complementarity opens new perspectives as to the potential range of functions of this unexpectedly large snoRNA family (17). Remarkably, one of the novel brain-specific snoRNA displays an 18-nt antisense element that could play a key role in the processing of a brain-specific mRNA.
Methylation guide snoRNAs are also intriguing by their unusual genomic organization and modes of biosynthesis. In vertebrates, they are encoded within introns and are not independently transcribed but processed from the pre-mRNA introns, in most cases by exonucleolytic digestion of the debranched lariat (18 -21). In contrast, in Saccharomyces cerevisiae, only a few of them are intronic, while most of them are synthesized from independent mono-, di-, or polycistronic RNA transcripts processed by endo-and exonucleases (22)(23)(24)(25)(26). Polycistronic snoRNAs have also been reported in plants (27)(28)(29) and in Trypanosoma brucei (30 -32). Finally, the detection of archaeal homologs of box C/D snoRNAs also points to novel aspects in their biosynthesis, with the coding sequences of several of them partially overlapping upstream and/or downstream open reading frames (33,34). Earlier observations that snoRNA host genes coded for proteins involved in ribosome biosynthesis and/or nucleolar function (e.g. ribosomal proteins, translation factors, nucleolin) suggested that this gene organization might provide a regulatory link between partners in the same biological process (1,35). However, it is now clear that this hypothesis cannot be generalized. First, some intronic snoRNAs are hosted by different genes in different eukaryotic species, thus appearing as mobile genetic elements. Second, the function of some snoRNA host genes lack any obvious direct relationship with translation. Finally, an increasing number of snoRNAs have been found in introns of non-protein-coding genes of elusive function (36 -40). However, all snoRNA host genes known to date are housekeeping genes exhibiting hallmarks of 5Ј TOP genes (i.e. a C residue at the ϩ1-position followed by a 5-15-nt-long polypyrimidine tract in a short 5Ј-untranslated region), which could provide the basis for a coordination of snoRNA biosynthesis at the transcriptional level (37,40).
In the present study, we have focused our attention on the identification of novel brain-specific box C/D snoRNAs, in order to better assess the complexity of their repertoire and provide insights into their biosynthetic pathway and function(s) in brain cells. Following the construction of a rat brain cDNA library specific for box C/D snoRNAs, we have identified and further characterized a novel brain-specific specimen, RBII-36, without apparent equivalent in mouse or human brain cells. Also devoid of complementarity to rRNA or snRNAs, RBII-36 shares with two brain-specific box C/D snoRNAs recently identified in mouse and human (17) an outstanding feature; the snoRNA is transcribed from a complex transcription unit spanning an array of tandemly repeated snoRNA-containing units. Moreover, RBII-36 snoRNA appears to be processed from intronic RNA through a superimposition of splicing-dependent and splicing-independent processing events.

EXPERIMENTAL PROCEDURES
Unless otherwise noted, all techniques for cloning and manipulating nucleic acids were performed according to standard protocols.
Construction of a Rat Brain snoRNA cDNA Library-Half a dissected rat brain was resuspended in 6 ml of NET-150 buffer (50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 0.05% Nonidet P-40) and homogenized at 1.53 kilobar using the One Shot Model (Constant Systems Ltd.). Brain extract was then clarified by centrifugation (10 min, 10,000 rpm in a Sorvall SS34 rotor at 4°C), and the supernatant was submitted to immunoprecipitation as described below. RNA extracted from 0.5 ml of PAS-Ig pellet was ligated to phosphorylated primer 1 (10 pmol) in the presence of T4 RNA ligase as described (10). Ligation product was used as template for cDNA synthesis with primer 2 and Superscript II reverse transcriptase (Life Technologies, Inc.). The cDNA product was then used as template for PCR by Taq polymerase (Appligene) with primers 2 and 3, and the PCR product was digested by BamHI and cloned in pKS vector. Individual clones were manually sequenced (T7 sequenase version 2.0, U.S. Biochemical Corp.).
Immunoprecipitations-Ten microliters of rabbit antitrimethyguanosine (anti-TMG R1131 (7 mg/ml), kindly provided by R. Lührmann) or monoclonal anti-fibrillarin 72B9 (kindly provided by M. Pollard) were incubated with gentle agitation for 120 min at room temperature with 2.5 mg of swollen protein A-Sepharose (Sigma) in 0.5 ml of NET-150 buffer (50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 0.05% Nonidet P-40). PAS-Ig pellets were washed eight times with 1 ml of NET-150 buffer. Ten micrograms of total RNA purified from rat brain or 0.5 ml of total brain extract (prepared as described above for construction of the rat cDNA library) was then added to PAS-R1131 (in 0.5 ml of NET-150) or PAS-72B9, respectively, and incubated for 60 min at 4°C with gentle agitation. PAS pellet was then collected and washed eight times with 1 ml of NET-150 buffer. RNAs from pellet was recovered by SDS/phenol extraction and analyzed by Northern blot as described below.
Primer Extension and RNA Sequencing-Primer extension was carried out as described previously (41) using avian myeloblastosis virus reverse transcriptase (Promega) at 42°C for 60 min. Direct RNA sequencing by primer extension was performed using the same conditions with dideoxynucleotides elongation mixes.
Isolation and Fractionation of RNA from Tissues and Rat Brain Areas-Rat tissue freshly prepared from a dead animal according to French institutional guidelines and rat brain areas manually dissected were quickly frozen in liquid nitrogen and stored at Ϫ80°C. Total cell RNA was isolated by the method of Chomczynski and Sacchi (42) adapted to our conditions (43). Poly(A) ϩ RNA purification was performed as recommended by manufacturer (Promega). Fractionation of RNAs was carried out by electrophoresis in 1.2% agarose, 2.2 M formaldehyde or 6% acrylamide, 7 M urea gels. Fractionated RNAs were transferred electrophoretically (120 min in 0.5ϫ TBE at 1 mA) to nylon membrane (Hybond N ϩ ; Amersham Pharmacia Biotech) and crosslinked by UV irradiation (120,000 J/cm 2 , Stratalinker, Stratagene, La Jolla, CA) before Northern blot hybridization.
Cloning of Spliced Bsr cDNAs-A rat brain cDNA library (Lamba ZAP II library from Stratagene) has been screened by hybridization as recommended by the manufacturer with a 32 P-labeled oligonucleotide specific for spliced Bsr RNA.
Processing of RBII-36 snoRNA from Intronic Sequences-The 0.9-kblong Bsr repeat unit was PCR-amplified from genomic DNA with 5Ј Bsr exon and 3Ј Bsr exon primers and inserted into eukaryotic vector pRCEN downstream from the cytomegalovirus promoter. Transfection of the construct pCMV/RBII-36 into mouse cell line L929 and processing assays were as described (11).
In Situ Hybridization-Detection of RBII-36 snoRNA and Bsr RNAs by in situ hybridization was performed on adult rat brain 10 -15 m cryostat sagittal sections as described (44) using digoxigenin-labeled antisense snoRNA or a spliced Bsr RNA riboprobe spanning four spliced exons.

Search for Novel Rat Brain-specific Box C/D snoRNAs in
Rat-Taking advantage of the fact that nucleolar protein fibrillarin specifically binds all members of this large snoRNA family, we first constructed a rat brain cDNA library specific for box C/D snoRNAs (Fig. 1). Briefly, box C/D snoRNPs were immunoprecipitated from a total rat brain extract by an antifibrillarin antibody (72B9). Extracted RNAs were oligonucleotide-tagged at both ends by T4 RNA ligase and amplified by RT-PCR as described (10). RT-PCR primers were designed in such a way as to only amplify RNA molecules containing box C (RUGAUGA) and D (CUGA) motifs at the appropriate distance from their extremities (Fig. 1). Amplified DNAs were then cloned, and individual clones were systematically sequenced. A large proportion of the clones corresponded to rat homologues of ubiquitous box C/D snoRNAs already identified in humans, mice, or other mammals (3,15). Another substantial fraction corresponded to novel, ubiquitously expressed rat snoRNAs either able to target rRNA or snRNA ribose methylations or devoid of any obvious presumptive RNA target. Finally, a few clones corresponding to brain-specific box C/D snoRNAs were also isolated, three of them corresponding to rat homologues of brain-specific snoRNAs previously detected in mouse, MBII-52, MBII-85, and MBII-48 (15,17). In this study, we have focused our attention on three closely related clones without equivalent among previously detected mouse snoRNAs.
Three Related Clones Correspond to a New Box C/D snoRNA-Data base searches revealed that closely related sequences of clones IPLIB-36, IPLIB-76, and IPLIB-256 ( Fig. 2A) shared 93-100% identity with four EST sequences (Gen-Bank TM accession numbers AA900746, BE099838, AI599685, and AW919728) and also, interestingly, with the 5Ј-end of a brain-specific non-protein-coding RNA, Bsr, recently identified in rat (45). As depicted in Fig. 2B (top), this region of Bsr contains a series of 0.9-kb-long tandemly repeated units. Within the snoRNA-like sequences of Bsr, the hallmark box motifs (C, DЈ, CЈ, and D, in that order) are generally conserved among the different repeated units, with polymorphism essentially restricted to length variation of a 5-19-nt-long adenosine tract located a few nt downstream from box C (Fig. 2B). Moreover, segments immediately upstream from box D or DЈ do not exhibit a complementarity to rRNA or an snRNA of at least 9 bp, thus making it unlikely that the snoRNA could direct a 2Ј-O-methylation of one of these potential targets (41). Examination of all specimens of the snoRNA-like sequence of Bsr available in GenBank TM showed that a perfect 5-bp-long 5Ј-3Ј-terminal stem could form in all cases at the expected location relative to the C and D motifs (Fig. 2B). This reinforces the notion that part of the Bsr transcript actually encoded a box C/D snoRNA. To test this possibility, a Northern blot analysis of total rat brain RNA was performed with 32 P-labeled probe IPLIB36-2. Three distinct radioactive signals were detected at about 76, 80, and 88 nt (Fig. 2C, left), in good agreement with the size expected for an RNA exhibiting the above-mentioned length polymorphism and ending at the expected 5Ј-3Ј-terminal stem. Intriguingly, we also observed a radioactive band of much faster mobility, at about 40 nt. Reverse transcription performed on total rat brain RNAs using IPLIB36-2 primer confirmed the expected location of the snoRNA 5Ј end and revealed a snoRNA size heterogeneity consistent with its being expressed from multiple units of the Bsr repetitious array (Fig.  2C, right). In agreement with the presence of the 40-nt RNA band detected by Northern blot, an intense cDNA band doublet was also detected at 39 -40 nt, indicating that the 5Ј-end of the shortened form was located 3-4 nt upstream from box CЈ. As expected, the RNA bands of 76, 80, and 88 nt detected in Northern blot with the IPLIB36-2 probe were all specifically immunoprecipitated with the 72B9 anti-fibrillarin antibody (Fig. 2D, left) and also with anti-hNop58 and to a lesser extent with anti-Nop140 antibodies (data not shown). This was also the case for the 5Ј truncated 40-nt-long species. Taken together, these data indicate that the three closely related sequences isolated from our cDNA library correspond to a bona fide box C/D snoRNA encoded in the Bsr locus. This snoRNA is devoid of a 5Ј-trimethyl cap, as indicated by its lack of immunoprecipitation with an anti-trimethyl cap (R1131) antibody (Fig. 2D, right). In the course of their study, Komine et al. (45) detected with the 0.9-kb Bsr cDNA probe, in addition to the long Bsr transcript, an intriguing cross-hybridizing 80-nt-long RNA of elusive origin. In hindsight, it is now clear that this 80-nt-long RNA must correspond to the novel box C/D snoRNA identified in the present study, termed RBII-36 (for rat brain fraction II 36).
FIG. 1. Construction of a cDNA library for box C/D snoRNAs in rat brain. A, experimental strategy. A rat brain extract was immunoprecipitated by an anti-fibrillarin antibody (72B9), and extracted RNAs were tagged at both ends by oligonucleotide primer 1 using T4 RNA ligase. Tagged RNAs were amplified by RT-PCR using primers 2 and 3, designed so as to amplify RNAs containing box C (RUGAUGA) and box D (CUGA) motifs at the appropriate distance from their 5Ј-and 3Ј-ends; the two primers included 4 degenerated positions between the primer 1-matching sequence and their 3Ј-terminal sequence (matching either box C or box D). Amplified DNAs were cloned into pKS vector, individual clones were systematically sequenced, and new snoRNAs were assayed by Northern blot as described in the legend to Fig. 2B.

RBII-36 Is Mainly Expressed in Rat
Brain-To gain more information about the pattern of expression of RBII-36, a panel of total RNA samples extracted from various adult rat tissues was analyzed by Northern blot using the IPLIB36-2-specific probe (Fig. 3A). We detected a signal only in the brain RNA sample (lane 1), although very low levels of RBII-36 snoRNA could also be detected in ovaries and stomach after much longer exposures (data not shown). This tissue-specific expression pattern is in striking contrast to what has been observed for previously identified snoRNAs guiding rRNA or snRNA methylation. Thus, as shown in Fig. 3A, U40 and U21 snoRNA were detected at roughly similar levels in all of the rat tissues checked. The same membrane was also probed with sequences derived from two brain-specific box C/D snoRNAs recently identified in mice, MBII-52 and MBII-85 (17). As expected, rat homologs of these snoRNAs were detected, which were also specifically expressed in the brain, although a very weak signal could also be observed in other tissues for RBII-85 snoRNA after much longer exposures (not shown). We conclude that RBII-36 snoRNA, like MBII-52 and MBII-85 snoRNAs, belongs to a novel subset of box C/D snoRNAs specifically expressed in the brain. However, the abundance of RBII-36 snoRNA in rat brain is much lower than that of rat homologs of MBII-52 and MBII-85 snoRNAs, both present at levels similar to U3 snoRNA, and is in the same range of magnitude as ubiquitously expressed box C/D snoRNAs U21 and U40 (Fig. 3B). The expression of RBII-36 snoRNA was also studied during rat brain development (Fig. 3C). Rat brains were dissected from embryos at days 15, 17, and 19 (E15, E17, and E19) or newborns at days 2, 5, 13, and 21 (P2, P5, P13, P21) and analyzed by Northern blot. Using 5.8 S rRNA as an internal reference, quantification of the RBII-36 Northern blot signal indicated that the RBII-36 snoRNA expression profile is remarkably constant throughout rat brain development.
RBII-36 snoRNA Expression in Different Brain Areas-We next investigated the distribution of RBII-36 snoRNA within various rat brain areas. In a first step, several brain areas were isolated from adult rat brain, total RNA was extracted and analyzed by Northern blot using IPLIB36-2 as a probe. Fuji Bas-1000 imager analysis (Molecular Dynamics, Inc., Sunnyvale, CA) of Northern blot signals revealed that RBII-36 snoRNA was nearly uniformly expressed in all brain areas checked (Fig. 4A, lanes 2-9) with the notable exception of choroid plexus (Fig. 4A, lane 10), which corresponds to highly vascular epithelial structures arising from the wall of ventricles. Such a brain-regional distribution is highly reminiscent of that observed for mouse brain-specific MBII-52 and MBII-85 snoRNAs (17). To gain more information at the cellular and subcellular level, we next performed in situ hybridization to rat brain sections using a digoxigenin-la-FIG. 2. IPLIB clones representing a novel rat box C/D snoRNA. A, sequences of the three related cDNA clones isolated in this study. B, alignment of EST and Bsr clones exhibiting homology with the three related cDNA clones described in the legend to Fig. 2A. Top, schematic representation of a previously characterized Bsr genomic clone (GenBank TM accession number AB014883) containing four 0.9-kb-long repeated units (45). The position of the IPLIB snoRNA-like sequence (80 -90 nt) within 1 unit is indicated by a black box. EST and Bsr clones are designated by their GenBank TM accession numbers, with the letter in parentheses referring to a particular repeat within the clone according the previously used terminology (45). Box C and D motifs are boxed, and nucleotides presumably involved in the formation of a 5Ј-3Ј-terminal stem are delineated by horizontal arrows in opposite orientation. Below the alignment, nucleotide positions conserved in at least 16 of the 18 Bsr repeats are denoted by asterisks, and the location of the different oligonucleotide probes or primers is shown by bars. The arrowheads denote the 5Ј-end of the truncated 40-nt-long snoRNA form. C, detection of the cognate snoRNA in total rat brain. Left, Northern blot hybridization with 32 P-labeled probe IPLIB36-2 using 10 g of total brain RNA (lane M, size marker). Right, 5Ј-end identification by primer extension with 32 P-labeled oligonucleotide IPLIB36-2 using 10 g of total brain RNA. Top, cDNA bands corresponding to full-size snoRNA. Bottom, 5Ј-truncated, 40-nt-long cDNA. In both cases, the corresponding RNA sequencing pattern is also shown. Correspondences between Northern bands and primer extension products are denoted by arrows. D, immunoprecipitation by anti-fibrillarin 72B9 (left) and anti-trimethyl cap R1131 (right) antibodies. RBII-36 snoRNA was detected by Northern blot hybridization with IPLIB36-2 probe, and membranes were also probed with U3-or U2-specific probes, respectively, as positive controls, and with a 5.8 S rRNA probe as a negative control. Lanes 1 and 3, input RNA; lanes 2 and 5, RNA recovered from the pellet; lane 4, RNA recovered from the supernatant. beled snoRNA riboprobe. Results clearly confirm that RBII-36 snoRNA is expressed in most brain areas (Fig. 4B  and data not shown). Moreover, RBII-36 snoRNA was mainly detected within neurons and more pronounced within granular and pyramidal layers of the hippocampus and granular layers and Purkinje cells in the cerebellum. About 60 and 80% of neurons were labeled in thalamus and in cortex, respectively, while no snoRNA signal was detected within the most external cortical layers. RBII-36 snoRNA localization within the brain thus closely parallels that reported previ- FIG. 3. Brain-specific expression of RBII-36 snoRNA. A, RBII-36 snoRNA abundance in various adult rat tissues. Equal amounts of total RNA prepared from adult rat tissues (as specified at the top of each lane) were fractionated in a 6% polyacrylamide gel, and RBII-36 snoRNA was detected by Northern blot as in Fig. 2A. The membrane was also probed for MBII-52 and MBII-85 and for U3, U21, and U40 snoRNAs (M, DNA size marker in nt). B, abundance of RBII-36 snoRNA in rat brain. Decreasing amounts of total rat brain RNA (15,5 , and 1 g, in lanes 1, 2, and  3, respectively) were separated on a 6% polyacrylamide gel, and RBII-36 snoRNA was detected by Northern blot as in Fig. 2A. The same membrane was also hybridized with oligonucleotide probes specific to snoRNAs MBII-52, MBII-85, U3, U21, and U40, 32 P-end-labeled to the same specific activity (all blots were exposed for identical times). C, RBII-36 snoRNA expression pattern in the developing rat brain. Ten micrograms of total RNA from developing rat brains were fractionated in a 6% polyacrylamide gel, and RBII-36 snoRNA was detected by Northern blot as in Fig. 2A.  4. RBII-36 snoRNA expression pattern in rat brain. A, Northern blot hybridization of RNA from major rat brain areas. Five micrograms of total RNA extracted from manually dissected brain areas from adult rat were analyzed by Northern blot as indicated in Fig. 2A, and the same membrane was also hybridized with a 5.8 S rRNA probe for gel loading controls. Signal intensities were quantified with a Fuji Bas-1000 imager. B, in situ hybridization on brain sections with a digoxigenin-labeled snoRNA probe. Panel of different brain areas in sagittal sections of adult rat brain, as indicated in each image. DG, dentate gyrus; CGC, cerebellar granular cells; PC, Purkinje cells. ously for unspliced Bsr RNA (45) and also for spliced Bsr RNA (see below). 2 We next investigated the subcellular localization of RBII-36 snoRNA by in situ hybridization (Fig. 4C). The position of the nucleolus was detected by using antibodies to nucleolar protein nucleolin, and RBII-36 snoRNA was visualized with a digoxigenin-labeled antisense riboprobe. As shown in Fig. 4C (top), the RBII-36 snoRNA signal in Purkinje cells colocalized with the nucleolin signal, showing that the brain-specific snoRNA is mainly located within the nucleolus. Intriguingly, a punctate secondary staining was also systematically detected within the nucleoplasm (indicated by an arrowhead), not only with the snoRNA antisense probe but also with the riboprobe specific for spliced Bsr RNA (Fig.  4C, bottom, and see below). This extranucleolar signal was typically detected only once per nucleus and was generally surrounded by tractlike structures.
RBII-36 snoRNA Is Intron-encoded in Bsr RNA-Bsr RNA is a noncoding RNA expressed in rat central nervous system only and consists of a direct tandem repetition of 0.9-kb-long units that are 87-100% identical to each other (45). Since comparison of genomic and cDNA clones did not suggest the presence of intervening sequences in the Bsr transcript, our detection of RBII-36 snoRNA within the Bsr RNA locus was quite unexpected. Analysis of various Bsr cDNA sequences by the Net-Gene neural network program predicted several potential donor and acceptor splice sites (not shown). In order to assess whether Bsr RNA was actually spliced, we performed RT-PCR using primers positioned over the predicted exon with the highest score (Fig. 5A). Analysis of DNA products on ethidium bromide-stained agarose gel revealed a ladder-like pattern of at least four bands differing by about 100 bp increments (data not shown), a result in agreement with the presence of a tandem repetition of spliced exons of the shortest predicted size (i.e. 86 nt). This notion was confirmed by Southern blot hybridization of the RT-PCR product using a 32 P-labeled oligonucleotide probe specific for religated exons (as depicted in Fig. 5A), which revealed the same ladder-like pattern (Fig. 5B, left). Finally, we sequenced the three shortest DNA bands of the ladder-like pattern generated by RT-PCR, which confirmed the predicted location of exonic junction, as shown for the shortest, 86-nt 2 E. Basyuk, et al., unpublished data.

FIG. 5. Identification of an intron in Bsr RNA.
A, sequence of a single arbitrarily chosen Bsr RNA repeat unit (GenBank TM accession number AB01879). Exon sequences are in uppercase type, intron sequence in lowercase, and the snoRNA sequence is underlined. Major spliced sites are underlined, whereas minor ones, identified by sequence analysis of a rat brain cDNA library, are boxed. The location of RT-PCR primers used in B) (filled bars) and the two moieties of the probe specific for spliced exons (hollow bars) are also shown. B, detection of the Bsr spliced exons by RT-PCR analysis. Total rat brain RNA (10 g) was annealed with 3Ј Bsr exon primer, and cDNA was synthesized with avian myeloblastosis virus reverse transcriptase before amplification by PCR with 5Ј and 3Ј Bsr exon primers. PCR products were fractionated on a 2% agarose gel and analyzed by Southern blotting with 32 P-labeled Bsr junction oligonucleotide probe specific for spliced exons (lane 1, control experiment performed without reverse transcriptase). Right, sequence analysis at the spliced junction for band I. A nucleotide polymorphism relative to the repeat unit sequence shown in Fig. 5A is denoted by a star. C, schematic representation of 12 partially sequenced Bsr cDNAs clones isolated upon screening of a rat brain cDNA library. Portions of cDNA clones that have not been sequenced are schematized by dots. The size of each Bsr cDNA insert (estimated by enzymatic digestion analysis) is indicated in parenthesis (a star denotes the presence of RBII-36 snoRNA sequence within the unsequenced portion of the clone, as assessed by dot blot hybridization). Repeated Bsr exons E1 are represented by open boxes, and an additional unrelated exon, E2, is represented by a black box (its sequence is in capital letters (top)). An unspliced intron I1 (thick line, with the snoRNA coding region depicted by an open arrow) is occasionally detected. Other portions of the clones exhibiting homologies to rat ESTs are also indicated. Gray box, Lx7#LINE/L1 repetitive element (GenBank TM accession numbers AI072426 and BE105131); hatched box, GenBank TM accession numbers AI180055, BF551987, and AI574801), The wavy line denotes the 5Ј part of an unspliced intron, I2, located between E1 and E2 exon (GenBank TM accession number AB014882).
RT-PCR product, band I (Fig. 5B, right). Taken together, these data indicate that Bsr RNA harbors a previously uncharacterized 807-nt-long intron containing RBII-36 snoRNA. By screening a rat brain cDNA library with the probe specific for spliced Bsr, we were able to isolate 12 independent Bsr cDNA clones, which have been partially sequenced from both ends (Fig. 5C). They all differ from each other in the number of spliced exons, utilization of cryptic donor and acceptor splice sites (see legend to Fig. 5, A and C), and occasional presence of unspliced intron(s), indicating that Bsr is an alternatively spliced RNA. Interestingly, seven of the positive clones have identical 3Јends. Within the common 3Ј-terminal region and immediately downstream from the 3Ј-most copy of repeated exons E1 (Fig.  5C), we identified by Blast search an 82-nt-long tract matching precisely a sequence located about 700 bp downstream from the repeated Bsr units in the rat genomic clone described previously (45). As expected, genomic sequences flanking the matching segment exhibit hallmarks of splice junction (Fig. 5C, top  right), indicating that the 82-nt-long segment, unrelated to repetitive exon E1 described above, is a novel exon, E2. Moreover, the sequence of cDNA clones revealed that the 3Ј-end of exon E2 is spliced to a 700-bp sequence, absent from the genomic clone mentioned above, which exhibits homology with Lx7/LINE1 repetitive elements. Since Bsr RNA had been proposed to be a nontranslated RNA (45), identification of the previously undetected repeated introns prompted us to reexamine the validity of this conclusion. No sizeable open reading frame was detected in the spliced Bsr sequence, confirming the notion that Bsr is a non-protein-coding RNA.
A Complex Processing Pattern for the Bsr Transcript-To gain more information about RBII-36 snoRNA processing from Bsr transcripts, RNAs extracted from rat brain were assayed by Northern blot hybridization with a riboprobe specific for spliced Bsr RNA (Fig. 6A, right). A heterodisperse signal (from about 0.5 to more than 10 kb) was detected for total RNA (lane 3) as well as for poly(A) Ϫ and poly(A) ϩ RNAs (lanes 4 and 5, respectively), reflecting the presence of a complex spectrum of Bsr RNA processing products. Northern blot analysis with a snoRNA-specific probe also revealed an heterodisperse signal in the same size range in addition to mature RBII-36 snoRNA (Fig. 6A, left). However, superimposed on the heterodisperse signal, discrete bands at about 0.9, 1.8, and 2.7 kb were also apparent, which could result from endonucleolytic cleavages of unspliced Bsr transcripts taking place at a specific site within FIG. 6. RBII-36 snoRNA biosynthetic pathway. A, detection of unspliced and spliced Bsr RNA by Northern blot hybridization. Tenmicrogram samples of total RNA from rat brain (lanes 1 and 3), rat liver (lanes 2 and 6), rat brain poly(A) Ϫ RNA (lane 4), and rat brain poly(A) ϩ RNA (lane 5) were fractionated on 1.2% formaldehyde-agarose gels. Membranes were probed with either a snoRNA RNA riboprobe (left) or with a probe specific for spliced Bsr (right). Probes specific for 5.8 S rRNA and actin mRNA were also used as controls for gel loading and recovery of poly(A) ϩ RNA, respectively. B, schematic representation of previously reported Bsr cDNA clones (45) the 0.9-kb repeat unit. It is noteworthy that among the 33 Bsr cDNA clones described previously (45) seven of them have their 5Ј-end mapping within the same 16-bp segment of the 0.9-kb repeat unit, immediately upstream from the snoRNA coding region. Remarkably, 5Ј boundaries of two of these clones, AB014878 and AB01881, map exactly at the 5Ј-end of mature and truncated RBII-36 snoRNAs, respectively. Moreover, the 3Ј-terminal region of one of the Bsr cDNAs identified in the present work, clone 11.5, corresponds to the entire portion of the intron upstream from the mature snoRNA coding region (Figs. 5C and 6C). Likewise, the 5Ј boundary of two other previously reported clones (AB014877 and AB014879) maps only 10 nt upstream from the 5Ј end of mature RBII-36 snoRNA (Fig. 6, B and C). Moreover, the 3Ј-end of clones AB014877, AB014879, and AB014880 also maps only a few nt upstream from the expected location of the 5Ј-end of a RBII-36 snoRNA copy in the downstream repeat unit (Fig. 6, B and C). Taken together, these observations strongly suggest that the 0.9-, 1.8-, and 2.7-kb-long RNA species might represent hemiprocessed intermediates of RBII-36 snoRNA biogenesis (as depicted in Fig. 6A, left). A single Bsr 0.9-kb repeat unit spanning the snoRNA-containing intron and two flanking exons was transiently expressed under the control of a strong viral promoter in mouse L929 cells. As shown in Fig. 6D, faithfully processed RBII-36 snoRNA was detected in transfected mouse cells. This strongly suggests that brain specificity of the snoRNA expression is exclusively determined at the transcriptional level and that snoRNA processing from brain-specific Bsr transcripts is not dependent upon brain-specific factors. It is also noteworthy that the 5Ј-truncated, 40-nt-long form of RBII-36 was not detected in transfected mouse cells, consistent with the possibility that this form originates from a subset of truncated repeat units.
RBII-36 snoRNA Is Only Detected in Rats-Most box C/D snoRNAs have homologues in distant vertebrate species that can be detected by Northern cross-hybridizations using oligonucleotide probes directed to their conserved antisense element (9,43,46). Given that the 3Ј half of RBII-36 snoRNA sequence is preferentially conserved among the different tandemly repeated units (Fig. 2B), we first selected the IPLIB36-2 probe to look for homologues in other vertebrate species. At moderate hybridization stringency, RBII-36 snoRNA was detected in rat only, whereas homologs of brain-specific snoRNAs MBII-52 and MBII-85 (17) could be detected in distantly related placental mammals but not in chickens or even in marsupials, as shown in Fig. 3D. At very low stringency (27°C), weak signals were detected for both human and mouse brain RNA (Fig. 3D), but their significance remains uncertain because they were not reproducible (data not shown). Moreover, in these very low stringency conditions we could not detect any signal for humans or mice with four other RBII-36 probes, IPLIB36-1, IPLIB36-3, IPLIB36-4, and IPLIB36-5 (not shown). We also tried to detect the potential presence of RBII-36-like, tandemly repeated sequences in the genome of various rodent species (a generous gift from F. Catzeflis) by PCR amplification performed on genomic DNA with a pair of divergently oriented primers spanning the most conserved part of snoRNA sequence among the rat repeated units. A PCR product of roughly similar size was obtained with each rodent belonging to the genus Rattus sensu lato (47), such as Rattus rattus, Niviventer rapit, Maxomys hylomyoides, and Leopoldamys sabanus. Conversely, no signal was detected for other rodent species outside this genus, such as Mus musculus, Mus pahari, and Mus saxicola (not shown). We therefore conclude that either RBII-36 is a rat-specific snoRNA or that its mouse/human counterparts are very poorly expressed and/or evolutionarily divergent. DISCUSSION RBII-36 snoRNA, a Novel Rat Brain-specific C/D Box snoRNA Lacking an Obvious RNA Target-By sequencing a cDNA library specific for rat brain C/D box snoRNAs (Fig. 1), we have isolated a novel C/D snoRNA, RBII-36, which, in contrast to the vast majority of C/D snoRNAs, is only expressed in the central nervous system (Fig. 3A). RBII-36 accumulates within nucleoli of neuronal cell types, with a particularly strong expression in pyramidal layers of the hippocampus and Purkinje cells of the cerebellum (Fig. 4, B and C). The lack of any potential antisense element able to direct a 2Ј-O-ribose methylation to rRNA or an snRNA raises puzzling questions about its function and the identity of its potential RNA target in neuronal cells, like for the few brain-specific C/D box snoRNAs recently identified in mouse (17). A growing list of C/D box snoRNAs devoid of likely rRNA or snRNA targets has been recently identified (15,16), suggesting the possibility that the repertoire of cellular RNAs targeted by methylation guide snoRNAs could be larger than anticipated. In this regard, several stable ubiquitous noncoding RNA species transit through the nucleolus, which seems to be involved in the metabolism of a variety of cellular RNAs in addition to rRNA (48,49). Moreover, archaeal homologs of C/D box snoRNAs direct methylation of not only rRNAs but also tRNAs. 3,4 The detection of brain-specific C/D box snoRNAs without rRNA or snRNA complementarity might reflect their targeting of brain-specific RNAs, possibly messenger RNAs, although no internal ribosemethylated nucleotide modification has been reported so far in mRNA. In line with this notion, one of the brain-specific C/D box snoRNAs recently reported in mouse, MBII-52, contains a long (18-nt) phylogenetically conserved antisense element directed against a brain-specific mRNA encoding a serotonin receptor (17). Remarkably, the mRNA nucleotide predicted to be methylated by MBII-52 snoRNA is an adenosine undergoing conversion to inosine (50), suggesting that the snoRNA is involved in a control of the editing process, since methylation of an adenosine to be edited strongly inhibits its deamination in vitro (51). In this hypothesis, the presence of inosine at tissuespecific levels, particularly elevated in brain mRNAs (52), hints at the possibility that additional brain-specific box C/D snoRNAs with long complementarities to brain mRNAs still await identification. Searches of genomic data bases have not provided so far any indication as to the identity of a potential mRNA target for RBII-36 snoRNA among a large number of candidate sequences in mammals. The punctate extranucleolar signal detected within nucleoplasm by in situ hybridization with the snoRNA probe (Fig. 4C, top) could correspond to the nuclear site where RBII-36 snoRNA interacts with its still unknown cellular RNA target. Alternatively, it could represent the site of either transcription or early processing of Bsr RNA, in line with our detection of a similar signal with a probe specific for spliced Bsr (Fig. 4C, bottom). Ongoing analyses with additional probes within the Bsr locus should provide further insight into this question.
Bsr, an Intriguing snoRNA-containing Host Gene-Eukaryotic cells contain an unexpectedly large number of noncoding RNAs, which usually have a narrow expression pattern and seem to play pivotal role(s) in many aspects of gene expression (53). Strikingly, several noncoding RNAs of elusive function belonging to the 5Ј TOP family of ubiquitous vertebrate genes host one or several intronic C/D snoRNAs: U19HG (36), U50HG (38), U17HG (40), gas5 (37), and UHG (39). Noncoding RNA Bsr was first described (45) as an intronless transcript isolated upon characterization of genes involved in cerebellar long term depression. It is mainly expressed in the rat central nervous system and consists of tandem repeats of 0.9-kb-long sequences (45). In this study, we have shown that the tandemly repeated units within Bsr RNA actually contain an intron from which RBII-36 snoRNA is processed (Fig. 5) and that the intriguing 75-nt-long RNA species detected during the initial characterization of Bsr RNA corresponds to the novel RBII-36 snoRNA. Curiously, the tandemly repeated organization of snoRNA genes detected for RBII-36 is highly reminiscent of that of two other brain-specific snoRNAs also encoded within noncoding RNA genes in mice and humans (17,54). The relationship, if any, of this particular type of gene organization to the expression of brain-specific snoRNAs remains elusive. The Bsr RNA gene could be identified only within the Rattus genus (Ref. 45 and data not shown), and its function is unknown. The absence of the long open reading frame and the exclusively nuclear location of spliced Bsr RNA (Fig. 4C) strongly argue against its protein coding potential. In this regard, it is noteworthy that genes of noncoding RNAs hosting a ubiquitous snoRNA within one or several of their introns, such as UHG, gas5, and U19HG, generally exhibit highly divergent exons among mice and humans, suggesting that the intron-encoded snoRNAs could well be the only functional entities of these complex transcripts (36,37,39). However, we cannot rule out the possibility that spliced Bsr RNAs have a function per se, like Xlsirt noncoding RNAs, a family of interspersed repeat RNAs of about 80 nt required for the proper intracellular location of Vg1 RNA in Xenopus oocyte (55). Finally, another tantalizing question concerns the potential link between noncoding RNAs and imprinted chomosomal regions, since many noncoding RNAs are imprinted (see Ref. 56 for a review), and imprinted loci are often associated with repeated sequences (57).
An Unusual Processing Pathway for RBII-36 snoRNA-Most intronic snoRNAs are produced by a splicing-dependent processing pathway involving exonucleolytic trimming of the debranched lariat (for a review, see Ref. 6). Another minor, splicing-independent pathway involving endonucleolytic cleavage within the intron followed by exonucleolytic trimming has been identified for U16 and U18 snoRNAs in Xenopus oocytes and for U18 in S. cerevisiae (58 -61). In mouse, brain-specific MBII-52 snoRNA is also encoded in tandemly repeated introns, but boundaries and fate of cognate transcript(s) have not been assessed so far (17). We have identified spliced Bsr RNA products containing at least eight ligated exons (Fig. 5C) together with hemiprocessed snoRNA-containing Bsr RNA precursors (Fig. 6, A and B; Ref. 45), strongly arguing for the superimposition of the two mutually exclusive snoRNA biosynthetic pathways mentioned above (Fig. 7). The heterodisperse Bsr RNA signal (ranging from about 500 bp to more than 10 kb) could reflect transcription from multiple promoters as well as complex alternative splicing of a common transcript synthesized from a unique promoter. The heterodisperse signal, also reported for brain-specific noncoding RNAs IPW and PWCR1, which carry a snoRNA-containing tandem repetition (54,62), might also reflect nuclear degradation of untranslatable transcripts via NMD (63). Boundaries of available Bsr cDNA clones suggest that endonucleolytic cleavages of the Bsr transcript preferentially occur within a few nucleotide positions upstream from the RBII-36 snoRNA sequence in the intron, within an irregular 27-bp-long secondary structure involving both snoRNA flanking sequences in the intron (Fig. 6C). Interestingly, a long base-paired structure in C/D snoRNA-containing introns promotes intronic endonucleolytic cleavages (64,65). The truncated form of RBII-36 snoRNA lacking the box Ccontaining 5Ј-end (Fig. 2C) could result from an impaired ar-rest of 5Ј-3Ј exonucleolytic trimming, possibly due to a noncanonical 5Ј-3Ј-terminal stem-box C/D structure. Alternatively, it could reflect the presence of some truncated Bsr gene copies, possibly separate from the regularly repeated array, as reported for snoRNA-containing PWCR1 genes in humans (54). In line with this notion, we have identified several rat ESTs (GenBank TM accession numbers BF389774, BF386393, and BF390839) exhibiting the truncated snoRNA sequence at their 3Ј-end, while their immediately upstream sequences do not match any known Bsr sequence. Accumulation of the relatively abundant truncated form of the snoRNA (Fig. 2C) and its efficient immunoprecipitation by antibodies against C/D box snoRNA-binding proteins (Fig. 2D, left, and data not shown) suggests that CЈ box might substitute for C box to form the terminal stem-box structure, recognized by snoRNP core proteins, which acts as nucleolar localization signal (66). The detection of RBII-36 snoRNA, the first mammalian intron-encoded snoRNA likely to result from a splicing-independent processing pathway, adds another level of diversity to the strategies used to produce snoRNAs in eukaryotic cells.
Implications of the Tandemly Repeated snoRNA Gene Organization-While exhibiting structural hallmarks of a bona fide methylation guide snoRNA, RBII-36 strikingly departs from known mammalian C/D box snoRNAs by being exclusively detected in the Rattus genus, suggesting it might have originated rather recently in the evolution of rodents (Ref. 45; Fig.  3D and data not shown). In contrast, known ubiquitous intronic C/D box snoRNAs, whether they are encoded in protein genes or in noncoding RNAs, are conserved over a much larger span of vertebrate phylogeny (9,37,39,43,46). The three previously reported brain-specific C/D box snoRNAs also seem substantially more ancient than RBII-36, since their origin must predate the primates/rodents split during mammalian evolution (17). Due to both its recent origin and outstanding gene organization, the novel brain-specific snoRNA should represent a valuable system to catch a glimpse of the genetic processes underlying the evolution of the large family of box C/D snoRNAs in vertebrates. Although a large fraction of rRNA 2Ј-O-methylations are conserved among yeast and vertebrates, the vertebrate pattern is definitely more complex. Moreover, FIG. 7. A model for RBII-36 snoRNA biosynthesis. Bsr genes are transcribed as long RNA precursors from one or several promoters. In the splicing-dependent model (left), Bsr RNAs are alternatively spliced, and RBII-36 snoRNA is processed from the debranched intron 1 lariat. In the splicing-independent model (right), Bsr transcripts are cleaved endonucleolytically within each repeated intron (immediately upstream from the snoRNA coding region), as reflected by the ladder-like pattern of RNA fragments of about 0.9, 1.8, and 2.7 kb detected by Northern blot (Fig. 6A, left). The snoRNA-containing debranched lariat and the product of endonucleolytic cleavages are both further processed by exonucleolytic trimming. this pattern exhibits some substantial differences among distantly related vertebrates, with the presence of a few additional methylations in mammals as compared with amphibian Xenopus laevis (7), suggesting that cognate ubiquitous snoRNAs have originated at relatively late stages during vertebrate evolution. RBII-36 genes could represent a paradigm of an intermediate step in the expansion and functional diversification of the C/D box snoRNA family in vertebrate genomes. Their tandemly repeated organization could provide the structural basis for the generation of a large reservoir of novel C/D box snoRNA sequences among which some might eventually acquire the ability to recognize entirely novel RNA targets. Analysis of the RBII-36-containing repeat unit and derivation of the snoRNA sequence in other representatives of the Rattus genus could provide hints as to the actual function of the recently created snoRNA.