INTRODUCTION

Top Abstract Introduction Materials and Methods Results Discussion References

The nucleolus contains a large number of small nucleolar RNAs (snoRNAs), which function in the nucleolytic processing and nucleotide modification of precursor rRNAs (pre-rRNAs) (reviewed in references 34, 48, and 49). The vast majority of snoRNAs fall into two families, which can be distinguished on the basis of common sequence boxes. Many snoRNAs share the conserved C (consensus sequence: RUGAUGA) and D (CUGA) boxes (50; reviewed in reference 34), and the rest of the snoRNAs possess the H (AnAnnA) and ACA motifs (4, 16). A few C and D box-containing (box C/D) snoRNAs (U3, U8, U14, and U22) and an H and ACA motif-containing (box H/ACA) snoRNA (snR30) are required for the production of mature-sized rRNAs (21, 22, 30, 35, 40, 46, 53). However, the majority of box C/D and box H/ACA snoRNAs direct site-specific 2'-O-ribose methylation (10, 26, 32, 51) and pseudouridylation (17, 37) of rRNAs, respectively.

Genes coding for snoRNAs have been found in diverse genomic contexts (reviewed in references 34, 49, and 59). While the majority of vertebrate snoRNAs are synthesized as parts of pre-mRNA introns, most yeast and plant snoRNAs are transcribed as monocistronic or polycistronic precursor snoRNAs (pre-snoRNAs) from independent transcription units. In each case, mature snoRNAs are released from the primary transcript by nucleolytic processing. The intronic snoRNAs are processed from the removed and debranched host introns (23, 25, 39, 41) or, less frequently, from endonucleolytically released intronic fragments (7, 15, 42, 55) by exonucleolytic activities (2, 9, 11, 23-25, 41, 52). From yeast polycistronic pre-snoRNA transcripts, the RNase III (Rnt1p) endoribonuclease releases individual pre-snoRNA fragments and, again, exonucleolytic trimming forms the correct 5' and 3' termini of the snoRNA (12, 13, 43). The yeast Rat1p and Xrn1p 5'3' exonucleases play an essential role in the 5'-end formation of both intronic and polycistronic snoRNAs (41, 43, 55). Maturation of some yeast snoRNAs by trimming of short 3'-terminal trailer sequences specifically requires the Rrp6p 3'5' exonuclease that is a component of the yeast exosome (1, 2).

The exonucleolytic processing of intronic and polycistronic snoRNAs is directed by cis-acting elements which have been found within the snoRNA itself. Consistent with this finding, snoRNAs placed into nonnatural sequence contexts are accurately processed both in vitro and in vivo (5, 16, 17, 25-27, 44, 57, 60). The conserved sequence boxes and the neighboring stem structures constitute the signal elements directing the correct snoRNA formation (4, 6, 9, 16, 20, 44, 57, 60). Processing and accumulation of the intron-encoded and polycistronic box C/D snoRNAs depend on the C and D boxes and an adjacent 4- or 5-bp helix that folds together the 5' and 3' ends of the snoRNAs. This terminal structure, called the box C/D core motif (57, 60), functions as a binding site for snoRNP core proteins (8, 28, 33, 58). Most probably, snoRNA proteins bound to the pre-snoRNA protect the snoRNA sequences from the processing exonucleases and therefore delineate the correct snoRNA termini and provide metabolic stability for the mature snoRNA.

Several laboratories have demonstrated that a base-paired 5',3'-terminal stem is absolutely required for the processing and accumulation of box C/D snoRNAs (6, 9, 20, 31, 57, 60). However, mysteriously enough, many intron-encoded box C/D snoRNAs expressed in mammalian cells (26, 54) or polycistronic snoRNAs accumulating in yeast (32) lack the canonical 5',3'-terminal helix. Elements supporting the processing and accumulation of this large group of box C/D snoRNAs remain largely speculative. In this study, we demonstrate that processing of mammalian box C/D snoRNAs lacking a 5',3'-terminal stem is supported by external intronic stem structures that are fully or partially degraded during the exonucleolytic processing of these snoRNAs.

	MATERIALS AND METHODS

Top Abstract Introduction Materials and Methods Results Discussion References

General procedures and oligos. Unless stated otherwise, all techniques used for manipulating DNA, RNA, and oligodeoxynucleotides (oligos) were performed according to standard laboratory protocols (45). The following oligos were used in this study: 1, ATAATCGATGTAGGGTGATGAAAAAGAATCCTTAGGCG; 2, ATAACGCGTAGTCAGTGTAACTATGACAAG; 3, ATAATCGATGGAGGCAAGTAGGGTGATGAAAAAGAATCC; 4, ATAACGCGTGGAGGCAACAGTCAGTGTAAC; 5, ATAATCGATGGCGCCTCGCGTC ATAGTCAG; 6, ATACTCGAGACCAAGGAGAGCAAGGCAAGTGAG GG; 7, ATAATCGATCTCTATTCGTAGGGTGATGAAAAAGAATCC; 8, ATAACGCGTCTCTATTCCAGTCAGTGTAACTATG; 9, CGATGCAACAGCATGATGATCGACCGC; 10, GGTAAGATCTCTGATCAATGATGACATATGGTCAAG; 11, TCCATTTACCTGACTGTTGCA; 12, TTGATCAGAGATCTTACCGCGGTCGATCATCATGCTGTTGCAT; 13, CGCGTGCAACAGTCAGGTAAATGGACTTGACCATATGTCATCA; 14, ATAATCGATTCGAGACATGATGATCGACCGCGG; 15, ATAACGCGTGCAACAGTCAGATAAATGG; 16, ATAATCGATCAAGAGACATGATGATCG; 17, ATAATCGATTGCAAGAGACATGATGATGCG; 18, ATAATCGATCGTGCAAGAGACATGATGATCG; 19, ATAATCGATCGCGTGCAAGAGACATGATGATCGACCGCGG; 20, ATAATCGATTCGACACATGATGATCGACCGCGG; 21, CGCGTGCAACAGTCAGGTAAATGACCGCGGTCA; 22, TATGACCGCGGTCATTTACCTGACTGTTGCA; and 23, GGCCACAACCACGCCTAAGG.

Construction of plasmids for transfection of COS7 cells. Construction of the pGL_CXM mammalian expression vector has been described elsewhere (25). In brief, the human -globin gene carrying three novel restriction sites (ClaI, XhoI, and MluI) in its second intron was placed under the control of the cytomegalovirus (CMV) promoter. To generate pGL-U46(0/0), the coding region of human U46 snoRNA was PCR amplified using oligos 1 and 2 as 5'- and 3'-end-specific primers, respectively. The amplified DNA fragment was digested with ClaI and MluI and inserted into the same sites of the pGL_CXM expression construct. A similar approach was used to generate pGL-U46(8/9) and pGL-U46(93/54). Fragments of the human S8 ribosomal protein gene carrying the U46 gene and its 8- or 93-nucleotide upstream and 9- or 54-nucleotide downstream flanking sequences were amplified using oligos 3 and 5 and oligos 4 and 6, respectively. The amplified (8-U46-9) and (93-U46-54) fragments were inserted into the ClaI/MluI and ClaI/XhoI sites of pGL_CXM, respectively. To obtain pGL-U46(8/9) and pGL-U46(8/9) (underlining indicates the destroyed part of the construct [see Results]), the U46 gene was PCR amplified using oligos 7 and 4 and oligos 3 and 8 as primers, respectively. The amplified fragments were cloned into the ClaI/MluI sites of pGL_CXM. The same strategy was used to generate pGL-U46(8/9), except that oligos 7 and 8 were used as 5'- and 3'-end-specific PCR primers, respectively.

RNA analyses. RNase A/T₁ mapping was performed as described earlier (19, 25). Complementary RNA probes were synthesized in vitro with T7 RNA polymerase. All probes were purified on 5% denaturing polyacrylamide gels. To generate templates for the preparation of antisense RNA probes, the HindIII-EcoRI fragments of the pGL-U46, pGL-U46(8/9), pGL-U46(93/54), pGL-U46(8/9), pGL-U46(8/9), pGL-U46(8/9), pGL-X1, pGL-X1(0), pGL-X1(3), pGL-X1(5), pGL-X1(7), pGL-X1(9), pGL-X1(5t), pGL-X1(0)-(9), and pGL-X1(5t)-(9) expression constructs were cloned into the HindIII/EcoRI sites of pBluescribe KS() (Stratagene). The resulting recombinant plasmids were linearized with HindIII and used as templates for T7 RNA polymerase transcription. The 5' terminus of the human U46 snoRNA was determined by primer extension analysis with 5 µg of template RNA extracted from a nuclear fraction of human HeLa cells (50) and 5 pmol of terminally labeled oligo 23 as a primer. Computer folding of human and mouse intronic sequences was performed by using the RNAdraw program (http://rnadraw.base8.se/).

	RESULTS

Top Abstract Introduction Materials and Methods Results Discussion References

Intronic flanking sequences are required for the accumulation of U46 intron-encoded snoRNA. Contrary to the fact that a short 5',3'-terminal helix structure has been demonstrated to be fundamental to the accumulation of both intron-encoded and polycistronic box C/D snoRNAs, a large number of naturally occurring box C/D snoRNAs lack a terminal stem (see above). To solve this apparent contradiction, we have set about identifying the cis-acting elements that direct the processing of the human U46 box C/D snoRNA, which has been reported to lack a canonical 5',3'-terminal stem (26). Primer extension (Fig. 1A) and oligo ligation-PCR amplification (Fig. 1B) experiments demonstrated that the C and D boxes of the mature U46 snoRNA are preceded by five 5'-terminal and followed by two 3'-terminal nucleotides which are apparently unable to form a canonical terminal stem. The human U46 snoRNA is encoded within the second intron of the S8 ribosomal protein gene (26). To dissect elements essential for the accumulation of U46 snoRNA, fragments of the human S8 gene encompassing the coding region of U46 snoRNA with or without intronic flanking sequences were inserted into the second intron of the human -globin gene (Fig. 2A). The resulting constructs were placed under the control of the CMV promoter and transfected into simian COS7 cells (25). The accumulation of U46 snoRNA and the spliced globin mRNA was monitored by RNase A/T₁ mapping using antisense RNA probes specific for each expression construct (Fig. 2B). Control mapping with RNAs obtained from human HeLa cells (Fig. 2B, lanes 2, 6, and 10) and nontransfected COS7 cells (lanes 3, 7, and 11) revealed that the human and simian U46 snoRNAs can be readily distinguished based on their different migration properties in a denaturing polyacrylamide gel. RNA fragments protected by the simian U46 snoRNA migrated slightly faster than those protected by the human U46 snoRNA.

View larger version (65K): [in this window] [in a new window]   FIG. 1.   Characterization of the 5' and 3' termini of human U46 snoRNA. (A) Primer extension analysis. A terminally labeled oligo complementary to the human U46 snoRNA from positions 21 to 40 was annealed to HeLa cell snoRNAs and extended by use of avian myeloblastosis virus reverse transcriptase (lane R). Lanes G, A, T, and C represent dideoxy sequencing reactions performed on a recombinant plasmid carrying the human U46 gene. The extended DNA products were fractionated on a 6% sequencing gel. (B) Determination of the 3' end of the human U46 snoRNA by the T4 RNA ligase-PCR procedure. Sequence analysis of three cDNA clones obtained from independent PCRs resulted in the same 3'-terminal snoRNA sequences. Sequences of the oligoribonucleotide ligated to the snoRNA are also indicated.

View larger version (32K): [in this window] [in a new window]   FIG. 2.   Processing of the U46 snoRNA from the second intron of the human -globin pre-mRNA expressed in COS7 cells. (A) Schematic representation of the expression constructs used for transfection of simian COS7 cells. The CMV promoter, the exon regions (E1 to E3), and the polyadenylation (PA) site of the human -globin gene are indicated. The relevant restriction sites are shown (C, ClaI; E, EcoRI; H, HindIII; M, MluI; X, XhoI). The human U46 intronic snoRNA (open arrow)with or without its authentic flanking sequenceswas inserted into the ClaI/MluI or ClaI/XhoI sites of the globin (GL) expression construct. (B) RNase A/T1 mapping of U46 snoRNA accumulation. RNAs isolated from human HeLa (H) cells and from transfected (T) or nontransfected (N) COS7 cells were mapped with sequence-specific RNA probes as indicated above the lanes; C, control mapping with Escherichia coli tRNA. RNA fragments protected by the first (E1) and second (E2) exons of the spliced globin mRNA and the human (hU46) and simian (sU46) U46 snoRNAs are indicated. Lane M, size markers in nucleotides.

An intronic stem structure is required for the accumulation of U46 snoRNA. We noticed that the upstream and downstream intronic sequences flanking the human U46 snoRNA in the S8 pre-mRNA possess the potential to form an 8-bp perfect double helix (Fig. 3A). In the GL-U46(8/9) transcript, this interaction is accidentally elongated with an additional U-A base pair originating from the terminal nucleotides of the ClaI and MluI restriction sites used for the insertion of U46-containing fragments of the S8 gene. A possible function of the 5',3'-terminal stem might be the folding of C and D boxes into close proximity with each other (57, 60). For processing of the human U46 snoRNA, instead of the canonical 5',3'-terminal stem, an intronic helix may juxtapose the C and D boxes. To test this hypothesis, the putative intronic stem of the GL-U46(8/9) construct was destroyed by substitution of either the 5' or the 3' side of the helix (Fig. 3A). Upon transfection of the resulting GL-U46(8/9) and GL-U46(8/9) constructs into COS7 cells, the expression of the U46 snoRNA and the globin mRNA was tested by RNase mapping (Fig. 3B). The GL-U46(8/9) and GL-U46(8/9) pre-mRNAs were correctly and efficiently spliced, but no detectable amount of U46 snoRNA was processed from the excised introns of these transcripts (Fig. 3B, lanes 8 and 12). When the two stem mutations were combined and, therefore, the intronic stem structure of U46 was reestablished, the accumulation of faithfully processed U46 snoRNA was restored (Fig. 3B, lane 16). These results demonstrate that the formation of an external helix structure is essential for the production of U46 snoRNA and that the nucleotide composition of the intronic stem structure does not influence the efficiency or fidelity of the processing reaction.

View larger version (47K): [in this window] [in a new window]   FIG. 3.   Processing of U46 snoRNA with altered external stems. (A) Proposed secondary structure of the 5',3'-terminal region of human U46 snoRNA in the GL-U46(8/9) pre-mRNA. Nucleotides derived from the human S8 pre-mRNA are in uppercase letters. Lowercase letters indicate nucleotides originating from the ClaI/MluI sites of the GL expression construct. Sequences retained in the mature U46 snoRNA are boxed. The C and D box motifs are highlighted. Sequences used to destroy the external stem of U46 in the GL-U46(8/9) and GL-U46(8/9) constructs are in italics. (B) RNase A/T1 mapping of U46 accumulation. The GL-U46(8/9), GL-U46(8/9), GL-U46(8/9), and GL-U46(8/9) constructs were transfected into COS7 cells. RNAs isolated from transfected (T) or nontransfected (N) COS cells were mapped with sequence-specific RNA probes as indicated above the lanes; H and C, control mappings with HeLa cell or E. coli RNAs, respectively. Lane M, size markers in nucleotides. See the legend to Fig. 2 for designations.

The accumulation of an intronic artificial box C/D snoRNA is supported by an external stem structure. To exclude the formal possibility that internal sequence and/or structural elements located in the U46 snoRNA might contribute to the processing of this snoRNA, we constructed an artificial box C/D snoRNA, called X1 RNA. Apart from the essential box C and D motifs and the 3'-terminal nucleotides that were designed to form an 8-bp helix with sequences preceding the C box, the sequence of the X1 RNA was randomly generated. The artificial X1 locus was inserted into the ClaI and MluI sites of the pGL_CXM expression vector (Fig. 2A) and transfected into COS7 cells. A predicted two-dimensional structure of the 5',3'-terminal stem-box C/D domain of the putative pre-X1 RNA is shown in Fig. 4A. In the GL-X1(0) construct, base pairing of the terminal stem of X1 RNA was disrupted (Fig. 4A). In the GL-X1(3), GL-X1(5), GL-X1(7), and GL-X1(9) constructs, novel helices of increasing length were constructed outside of the predicted mature X1 RNA sequences.

View larger version (36K): [in this window] [in a new window]   FIG. 4.   Processing of an artificial box C/D snoRNA from the second intron of the human -globin pre-mRNA. (A) Schematic representation of artificial box C/D snoRNAs. The X1, X1(0), X1(3), X1(5), X1(7), and X1(9) artificial snoRNAs were expressed in COS7 cells within the second intron of the human globin pre-mRNA. Asterisks indicate the 5'-terminal nucleotides of the processed X1, X1(7), and X1(9) RNAs. (B) RNase mapping of the accumulation of X1 artificial snoRNAs. RNAs extracted from COS7 cells transfected with the GL, GL-X1, GL-X1(0), GL-X1(3), GL-X1(5), GL-X1(7), or GL-X1(9) expression construct were mapped with sequence-specific RNA probes as indicated above the lanes. Lane N, control mapping performed with RNAs obtained from nontransfected cells; lane M, size markers in nucleotides. See the legend to Fig. 2 for designations.

Compilation of mammalian intronic pre-snoRNA sequences lacking a canonical 5',3'-terminal stem. A closer inspection of the available human and mouse precursor sequences of box C/D intronic snoRNAs revealed that each snoRNA that lacks a 5',3'-terminal stem possesses upstream and downstream intronic sequences which are able to form short stem structures (Fig. 5). The length of these putative double helices varies between 5 bp (mouse U27) and 16 bp (human U44). Unfortunately, the 5'- and 3'-terminal nucleotides have been experimentally determined for only a few snoRNAs (Fig. 5). Nevertheless, the available data show that in some instances, the putative external stem may include nucleotides that are preserved in the 5'- and/or 3'-terminal sequences of mature snoRNAs (e.g., hU31, hU36b, hU42, hU48, hU55, and mU36b). In other instances, the proposed external stem is composed of intronic sequences that are fully degraded during snoRNA formation (e.g., hU30, hU37, hU43, hU44, and hU46). While snoRNAs with a canonical box C/D-stem core motif possess five or six 3'-terminal nucleotides after the D box motif, snoRNAs lacking a terminal stem usually feature only two unpaired 3'-terminal nucleotides. In both groups of snoRNAs, the C box is preceded by four or, more frequently, five nucleotides (Fig. 5). In summary, the presence of potential external helices in the pre-RNAs of mammalian intron-encoded box C/D snoRNAs lacking a canonical 5',3'-terminal stem further underlines the functional importance of intronic stem structures in snoRNA processing and accumulation.

View larger version (46K): [in this window] [in a new window]   FIG. 5.   Proposed structures of the intron-snoRNA junction regions of human and mouse pre-mRNAs encoding box C/D snoRNAs without a canonical 5',3'-terminal stem. Sequences of the human (h) U27, U30, and U31 (54), U36b (18), U37 (38), U42, U43, U44, U46, U48, U55, and U56 (26), and U76 and U78 (47) pre-snoRNAs and the mouse (m) U25, U27, U30, and U31 (54), U36b (18), and U44 and U78 (47) pre-snoRNAs were taken from the literature. Solid lines represent snoRNA sequences located between the C and D box motifs. Asterisks indicate the experimentally determined 5',3'-terminal nucleotides of mature snoRNAs.

An internal helix structure supports the accumulation of X1 artificial snoRNA. Although the data presented thus far are consistent with the idea that external helices play an important role in the biogenesis of box C/D snoRNAs, it remains questionable whether external helices alone could support the accumulation of some naturally occurring snoRNAs. The minimal length of the external stem capable of supporting production of the X1 artificial snoRNA was established as 7 bp (Fig. 4). Therefore, it seems unlikely that, for example, processing of the human U31 or the mouse U25 and U27 snoRNAs, which possess only 5- or 6-bp-long external helices, could rely exclusively on such short interactions (Fig. 5). Computer-aided modeling of the human U31 and the mouse U25 and U27 snoRNAs revealed that they contain 7- to 10-bp-long internal helices that could be formed by snoRNA sequences located between the C and D boxes (Fig. 6A). Moreover, extension of the structural investigation to other snoRNAs lacking a canonical 5',3'-terminal stem showed that these snoRNAs frequently possess potential internal stem structures (data not shown).

View larger version (32K): [in this window] [in a new window]   FIG. 6.   Processing of artificial box C/D snoRNAs carrying internal stem structures. (A) Proposed internal stems of the human (h) U31 and the mouse (m) U25 and U27 snoRNAs. For other details, see the legend to Fig. 5. (B) Schematic representation of artificial X1 snoRNAs. (C) Accumulation of X1 RNAs processed from the human -globin pre-mRNA. RNA samples obtained from COS7 cells transfected with the GL-X1, GL-X1(0), GL-X1(0)-(9), GL-X1(5t), or GL-X1(5t)-(9) expression construct were analyzed by RNase A/T1 mapping. For other details, see the legends to Fig. 2 and 4.

	DISCUSSION

Top Abstract Introduction Materials and Methods Results Discussion References

Most box C/D snoRNAs are processed from intronic or polycistronic pre-snoRNAs by exonucleolytic activities (see above). The box C/D terminal core motif, including the phylogenetically conserved nucleotides of boxes C and D and the base-paired 5',3'-terminal stem, plays a pivotal role in the biogenesis and function of these snoRNAs. The box C/D-stem motif directs the correct processing of the snoRNA (6, 9, 57, 60), provides metabolic stability for the mature snoRNA (6, 20, 31), directs the nucleolar localization of the snoRNA (29, 36, 44), and functions in the rRNA 2'-O-methylation reaction (10, 26, 27). The multiple functions of the box C/D-stem motif are accomplished by snoRNP proteins which directly or indirectly bind to this structural motif (8, 14, 28, 58). In vitro reconstitution experiments suggested that, in addition to the conserved C and D boxes, the adjacent 5',3'-terminal stem also plays a fundamental role in snoRNP assembly (8, 58). Consistent with this conclusion, the terminal stem is essential for the processing (57, 60) and accumulation (9, 20, 31) of box C/D snoRNAs.

Contrary to the observed essentiality of the 5',3'-terminal stem, many box C/D snoRNAs lack a terminal helix (26, 32, 54). Elements underlying the accumulation of these snoRNAs remain conjectural. In this study, systematic modeling of pre-mRNA introns hosting a box C/D snoRNA suggested that the presence of an external intronic stem is a common feature of mammalian snoRNAs lacking a terminal stem (Fig. 5). Indeed, in vivo processing studies demonstrated that an intronic stem is required for the accumulation of the human U46 snoRNA, which features no terminal stem (Fig. 2 and 3). Apparently, the nucleotide composition of the external stem of U46 snoRNA has no functional importance, since replacement of the cognate intronic stem of U46 snoRNA with an artificial helix influenced neither the efficiency nor the fidelity of snoRNA excision (Fig. 3). The above conclusion was further corroborated by the fact that, in the presence of an external stem, the X1 artificial snoRNA was efficiently processed from the -globin pre-mRNA (Fig. 4). Together, these findings suggest that the processing of mammalian box C/D snoRNAs lacking a canonical 5',3'-terminal stem is supported by external stem structures that are fully or partially formed by intronic sequences flanking the snoRNA.

The notion that external stem structures can facilitate the processing of intronic box C/D snoRNAs reinforces the previous idea that bringing boxes C and D into close proximity is an important requirement for snoRNA accumulation (52, 57). Most likely, appropriate spatial arrangement of the C and D boxes promotes binding of the snoRNP core proteins that protect the snoRNA sequences from exonucleolytic degradation. Since the intronic sequences involved in external helix formation are degraded during snoRNA excision, we propose that folding of the C and D boxes together is essential only for snoRNA processing and that snoRNP proteins bound to the C and D boxes provide metabolic stability for the mature snoRNA. This idea also implies that, contrary to the previous belief, the canonical 5',3'-terminal stem functions mainly, if not exclusively, in the processing of box C/D snoRNAs. Production of mature intronic snoRNAs can be considered a compromise of two competitive processes, namely, exonucleolytic intron degradation and snoRNP assembly on the intronic precursor RNA. In vitro snoRNP reconstitution experiments suggested that the canonical box C/D core motif, in which the terminal helix is immediately followed by the D box and is separated from the C box by one unpaired nucleotide (Fig. 5), represents the most favorable structural configuration for snoRNP assembly (8, 58). We assume that the accumulation of intronic snoRNAs with a canonical 5',3'-terminal stem depends mainly on the level of synthesis of the host pre-mRNA, since these snoRNAs are processed with high efficacy. In contrast, the expression of snoRNAs lacking a canonical terminal stem can depend greatly on the structural conformation of the pre-snoRNA (Fig. 4 and 6). In other words, intronic pre-snoRNAs with weak snoRNP binding capacity are more likely to be degraded by processing exonucleases; therefore, these snoRNAs may accumulate less efficiently (Fig. 4).

Previously, it has been hypothesized that the processing of box C/D snoRNAs lacking a 5',3'-terminal stem is facilitated by internal stem structures that are formed by snoRNA sequences located between the C and D boxes (57). Although an internal helix introduced into the X1 artificial snoRNA restored a detectable level of snoRNA accumulation, it failed to support the efficient production of the X1 RNA (Fig. 6). This result indicates that folding of the C and D boxes together by an internal stem could not entirely fulfill the structural requirements of efficient snoRNA processing. However, in concert with a short noncanonical terminal stem, the same internal helix could support the efficient processing of X1 RNA, even though the noncanonical stem was unable to sustain snoRNA processing alone (Fig. 6). This result suggests that internal stem structures, in conjunction with other structural elements, such as noncanonical terminal or external stem structures, may significantly influence the accumulation of box C/D snoRNAs.

Very recently, processing of the yeast U18 and snR38 intron-encoded snoRNAs has been reported to be dependent on external intronic stem structures (56), suggesting that parallels can be drawn between the processing mechanisms of yeast and mammalian intronic snoRNAs lacking a terminal stem. Indeed, a closer examination of yeast polycistronic and intronic box C/D snoRNAs devoid of a canonical terminal stem revealed that these snoRNAs are flanked by inverted repeat sequences that, in principle, are able to form helix structures (56; data not shown). In contrast to the mammalian introns, large parts of the yeast snoRNA host introns can be folded into long rod-shaped structures that are built of 27 (U18) to 59 (snR39 and snR59) bp forming small internal loops and bulges (data not shown). Remarkably, the snoRNAs are located invariantly on the top of the long intronic stem structure. The functional importance of this unique organization of the yeast snoRNA host introns, if any, remains to be established.

In conclusion, in vivo processing experiments performed with the human U46 snoRNA and an artificial box C/D snoRNA revealed that the biogenesis of mammalian intron-encoded box C/D snoRNAs is a more complex process than has been anticipated. Besides the previously characterized box C/D-stem core motif, additional cis-acting elements located within the snoRNA and/or in the flanking intronic sequences may contribute to the processing and accumulation of this class of snoRNAs. Most likely, the external and internal stem structures, like the canonical 5',3'-terminal stem, promote the folding of C and D boxes into close proximity with each other and thereby facilitate the assembly of the snoRNP core complex on the pre-snoRNA.