Identification of 86 candidates for small non-messenger RNAs from the archaeon Archaeoglobus fulgidus

Afu-14, Afu-64, Afu-113, and Afu-122 display all canonical features of archaeal methylation guide sRNAs (9, 14, and 17). Afu-14, Afu-64, and Afu-122 had been predicted in the A. fulgidus complete genome, termed sR1, sR4, and SR2, respectively (14). Their existence had not been tested experimentally. The fourth previously predicted A. fulgidus C/D sRNA, sR3, had been characterized experimentally as a methylation guide (18). It was not detected by us, probably because it corresponds to a pre-tRNA intron, presumably suggested to be active in cis before tRNA splicing and because tRNAs (or pre-tRNAs) are underrepresented in the cDNA library (see above). However, we identified a fourth, previously unpredicted C/D sRNA candidate, Afu-113: it harbors a 9-nt-long 5′ antisense element to 23S rRNA predicted to direct 2′-O-methylation of U2129.

Two additional members of this group, Afu-67 and Afu-180, significantly differ from previously reported archaeal sRNAs by the presence of nucleotide deviations from the box C and C′ consensus and by unusually long D′–C′ and C′–D intervals. Both RNAs lack an antisense element of at least 8 nt upstream of box D or box D′. Thus, a potential rRNA or tRNA target site is absent (the statistical significance of shorter, unperfect matches with rRNAs seems marginal). The function of these two sRNAs remains elusive.

Based on hallmark sequence and structural features, Afu-4, Afu-46, Afu-52, and Afu-190 are likely to represent archaeal counterparts for the abundant class of H/ACA-box snoRNAs, which guide pseudouridylation of ribosomal or spliceosomal RNAs in eukaryotic cells (19–22). The few archaeal rRNAs analyzed thus far contain only a very small number of pseudouridines, similar to bacteria but in marked contrast to Eukarya (23, 24). However, integral protein components of box H/ACA small nucleolar ribonucleoprotein particles, Gar1p, Nhp2p, and Cbf5p, the common pseudouridine synthase, are present in archaeal genomes, raising the possibility that pseudouridine formation in archaeal rRNAs involves counterparts of eukaryotic box H/ACA small nucleolar ribonucleoprotein particles (25).

Afu-4 features a highly stable secondary structure organized in three long stems, each containing a typical pseudouridylation pocket immediately followed by a downstream, single-stranded H- or ACA-box motif (for structure see Fig. 5, which is published as supporting information on the PNAS web site). Although the AGA-box downstream from Afu-4 stem III deviates from the canonical motif, the presence of AGA has been observed in yeast and Trypanosoma H/ACA snoRNAs (20, 26). Sequences in the large internal loop of each stem can form canonical bipartite guide duplexes of 9–13 bp around a target uridine in 16S rRNA or 23S rRNA (Fig. 1A). Remarkably, the target uridine is always separated from the box motif by the typical distance, 15–16 nt (20). Based on such indirect but statistically highly significant evidence, Afu-4 is predicted to guide pseudouridylation of three uridines, U1167 in 16S and U2601 and U1364 in 23S rRNA (Fig. 1A). One of these uridines, U2601, is universally pseudouridylated in Eukarya. Interestingly, this latter nucleotide is not pseudouridinylated in the crenarchaeote Sulfolobus acidocaldarius (23, 24). A homolog of Afu-4 was discovered computationally in the archaeon Pyrococcus furiosus in the lab of Sean Eddy (R. J. Kline, Z. Misulovin, and S. Eddy, personal communication).

Afu-46, Afu-52, and Afu-190 also represent likely rRNA pseudouridylation guides. They feature single, long, stable stems positioned immediately upstream from an ACA motif (for structures see Supporting Results and Discussion and Fig. 3). They resemble Trypanosoma H/ACA-box snoRNAs, which also consist of a single hairpin (26), and depart from the two-domain structures widespread in yeast, animal, and plant snoRNAs (ref. 20; see ref. 27 for review). In each case, the long stem contains an internal loop with all the features of a pseudouridylation pocket. The respective internal loops are able to form a 10–13-bp bipartite RNA duplex around a target uridine in 16S or 23S rRNA (Fig. 1A). Also, the targeted uridine is separated from the ACA- (or H-) box by 14–16 nt. Afu-46, Afu-52, and Afu-190 therefore are predicted to guide pseudouridylation of U2639 and U2878 in 23S rRNA and U1004 in 16S rRNA, respectively (Fig. 1A). Presence of pseudouridines at the expected rRNA sites was confirmed by primer extension performed on N-cyclohexyl-N′-d-(4-methyl-morpholinium)ethylcarbdiimide p-tosylate-treated total cellular RNA. As predicted, the six rRNA pseudouridines were detected unambiguously (Fig. 1B and data not shown). All four snmRNAs from this group are relatively rich in GC as compared with the A. fulgidus genome (48.6%), especially Afu-4 (60%).

A total of 22 cDNA clones map to three different A. fulgidus genomic loci within clusters of a peculiar class of short tandem repeats, termed SRSRs for short regularly spaced repeats (Fig. 2 A–C; for Northern blot see Fig. 2D). A similar family of repeats, apparently devoid of any protein-coding potential, is present in most completely sequenced archaeal genomes (28). The tandemly repeated short sequence elements (24–37 nt depending on the organism) are always organized in a few large clusters in which 10–100 identical copies are interspersed by divergent spacers of constant size (31–51 nt depending on the organism). Sequences of the repeat units but not the spacers are conserved among Archaea including the crenarchaeon Sulfolobus solfataricus (Fig. 2E). In A. fulgidus, this repeat family is distributed among three genomic loci. Locus 1 and locus 2, designated as SR-1A and SR-1B, respectively (13), contain 48 and 60 copies, respectively, of a 30-nt repeat (Fig. 2 A, B, and E) interspersed with 38-nt-long unique spacers. The third locus, SR-2 (Fig. 2C), contains 42 copies of a 36-nt repeat (related in sequence to the 30-nt SR-1 repeat, Fig. 2E) interspersed with 41-nt-long unique spacers. The detection of snmRNA candidates mapping to all three loci (Table 1) provides evidence that short, regularly spaced repeats are transcribed in all cases from the same DNA strand. Transcription of tandemly repeated genetic elements in prokaryotes has not been reported before (for a review see ref. 28).

Our findings are corroborated by Northern analysis with probes against the respective repeat motifs. We observed ladder-like band patterns (Fig. 2D), with multiples of ≈68 nt (locus 1 and 2 repeats) or 75 nt (locus 3 repeat). These increments almost precisely correspond to the interval between successive repeats in each cluster (68 nt for locus 1 and 2 and 77 nt for locus 3). We observed a similar ladder-like pattern in a Northern blot from S. solfataricus by using a corresponding probe (data not shown). These patterns suggest that in all Archaea examined the clustered repeats are transcribed in the form of long precursor(s), subsequently processed into monomers or multimers of the repeat motif. Accordingly, clone Afu-79 spans two consecutive repeat units. Moreover, except for one clone, Afu-98, the 3′ end of the 22 cDNA sequences derived from the three A. fulgidus repeat clusters locates in the middle of the repeat motif.

A. fulgidus repeats in loci 1 and 2 are preceded by a highly similar region (285 nt, 90% identity), which also is related to the sequence upstream from locus 3. This finding could reflect the importance of these regions for promoting cluster transcription (Fig. 2A–C, and data not shown). Each of the three A. fulgidus repeat clusters spans one or two hypothetical ORFs (Fig. 2A–C), However, none of these ORFs exhibit sequence similarity to ORFs in any archaeal genome. Furthermore, with the exception of a hypothetical protein assigned as AF1880 in locus 3 (Fig. 2C), these hypothetical mRNAs would be transcribed from the opposite DNA strand.

Clusters of this family of repeats are present also in the genomes of the halophile Haloferax volcanii and its megaplasmid (29). Remarkably, the introduction of extra copies of the repeats within a plasmid alters its distribution among the daughter cells, suggesting that the tandem repeats are involved in replicon partitioning (29). Our results raise the possibility that their role in this control might be mediated by transcription of the repeat clusters.

Four different snmRNA species, designated 16S-D, 16S-U, 23S-D, and 23S-U RNA, representing processing products of pre-16S or pre-23S rRNAs, can be assigned to this group (Table 1). All four RNAs are generated by cleavage and subsequent ligation of pre-rRNA spacers at the BHB motifs flanking pre-16S and pre-23S processing intermediates in the rRNA primary transcript, thus revealing an additional link between rRNA processing and RNA splicing in Archaea (30).

Interestingly, 16S-D RNA, by far the most abundant RNA present in our cDNA library, contains structural motifs typical of methylation guide sRNAs. The 16S-D RNA binds to the L7Ae protein, a core component of archaeal C/D-box ribonucleoprotein particles, supporting the notion that it might have an important, but still-unknown role in pre-rRNA biogenesis or even might target RNA molecules other than rRNA (30).

In general, candidate clones located in intergenic regions should represent the most likely previously uncharacterized functional snmRNAs. Indeed, more than 30 candidates for snmRNAs could be identified in E. coli intergenic regions by using different computational and/or experimental approaches (6–8). We have detected nine candidates for snmRNAs located in the intergenic regions of the A. fulgidus genome. Three snmRNAs, Afu-226, Afu-274, and Afu-277, are remotely reminiscent of eukaryotic H/ACA snoRNAs because of the presence of sizeable stems and an ACA motif in a single-stranded 3′ tail. Among them, Afu-274 seems to be the best candidate with its two major stems separated by a hinge exhibiting a potential H-box. Although its 3′ stem could well accommodate a pseudouridylation pocket, the best potential bipartite guide sequence (for U1349 in 16S rRNA) in this pocket is only 8 nt long, however, and the internal loop is unusually large for a typical H/ACA snoRNA. Because we could not unambiguously assign them as bona fide H/ACA RNAs, Afu-226, Afu-274, and Afu-277 were annotated as class II/group 1 RNAs.

From the group of snmRNAs that are transcribed in the antisense orientation of an ORF encoding a known or hypothetical protein we identified 33 snmRNA candidates (Table 2). In E. coli, small antisense RNAs complementary to the translational initiation sites of specific mRNAs have been implicated in the regulation of translation. Such activity may be mediated via masking of the Shine–Dalgarno sequence and the AUG start codon (for reviews see refs. 1 and 31). We therefore assigned the 33 previously uncharacterized antisense RNAs to three different subgroups on the basis of the location of complementarity to the presumed respective mRNA targets. Four snmRNAs (Afu-32, Afu-97, Afu-142, and Afu-239) are complementary to the putative translation initiation region and might be involved in regulation of translation in analogy to E. coli snmRNAs. The potential target of Afu-142, gsp-E4, resembles a type II secretion protein, whereas the three other candidate antisense RNAs in this subset correspond to hypothetical proteins. The five snmRNAs complementary to the translation termination site and the remaining snmRNAs complementary to regions located within an ORF still might regulate translation by as-yet-unknown mechanisms.

From this group, five independent candidates for snmRNAs could be identified that overlapped ORFs. Expression of all these candidates was confirmed by Northern blot analysis, a prerequisite to be assigned to this group. At the same time, the ORF of the respective overlapping gene was required to be considerably larger than the size of the snmRNA, thereby excluding identification of false positives derived from degradation of mRNAs (Table 2). Coexpression of the potential snmRNA genes and their respective overlapping protein genes would require either processing of a fraction of a presumptive common transcript into the snmRNA or separate transcription of the snmRNA and mRNA from two independent promoters.

For this last group, it remains to be demonstrated that its RNA members actually represent defined previously uncharacterized snmRNAs rather than degradation products of known mRNAs. Based on the detection of a Northern hybridization signal with a size considerably smaller than that of the predicted ORF, three potential snmRNAs candidates, Afu-158, Afu-225, and Afu-301, of a total of 182 RNA species derived from the coding region of mRNAs can reasonably be assigned to this group. Any of them, however, might still represent a (stable) degradation product of an mRNA. Conversely, it is noteworthy that the three snmRNAs all map within genes for hypothetical proteins, opening the possibility that the proposed ORFs have been annotated wrongly and encode a snmRNA instead.