Cyanobacteria are phototrophic bacteria which perform oxygenic photosynthesis and populate widely diverse environments such as freshwater, the oceans, rock surfaces, desert soil or the Polar Regions. Therefore, various types of regulatory RNA can be expected that interplay with the different signal transduction pathways and stress responses. Antisense RNAs found within the gas vesicle operon in Calothrix PCC 7601 4, or covering the ferric uptake regulator gene furA in Anabaena PCC 7120 over its full length 6, or regulating the light-absorbing protein IsiA under conditions of iron limitation and redox stress in the unicellular Synechocystis PCC 6803 5 are three known examples for such RNAs in cyanobacteria. Based on comparative computational analysis, we predicted a whole set of putative trans-acting ncRNAs with unknown function, which we called Yfr1-Yfr7 7.

We have previously shown the existence of Yfr1 in three out of four tested marine cyanobacteria belonging to the genera Prochlorococcus and Synechococcus 7. Unicellular marine cyanobacteria of these genera provide an excellent dataset for computational predictions that require comparative genome information since currently 19 different genome sequences from very closely related isolates are available. Thus, there is an extensive dataset to take compensatory base mutations into account for the prediction of novel ncRNA candidates. In these genomes, orthologs of Yfr1 can be found even by simple BLAST searches (Fig. 1A).

Despite some recent progress 10, marine unicellular cyanobacteria are not trivial in functional studies requiring genetic manipulation. Therefore, finding possible orthologs of Yfr1 in one of the cyanobacterial model species amenable for genetic manipulation would be desirable. The direct identification of such an ncRNA is not intuitively possible based on sequence similarity alone. In Table 1, BlastN hits are given for three such model strains, Synechocystis PCC 6803, Synechococcus PCC 7942/6301, Anabaena PCC 7120 and the marine Synechococcus WH 7803. Using the yfr1 full length sequence of Prochlorococcus MED4 (5'-atgggggaaaccccatactcctcacacaccaaatcgcccgatttatcgggcttt-3') as a query, only for Synechocystis PCC 6803 a hit to the correct sequence was found, ranking only at the third position. Similar results were obtained when the yfr1 full length sequence of Synechococcus WH 8102 was taken, except that the ortholog in the very closely related Synechococcus WH 7803 was found with clear statistical support (score = 77.8 bits, e-value = 4e-16; not shown).

Thus, sequence homology alone is not sufficient and we reasoned to complement it by comparative structure information. Taking up this idea, the initial set of the three previously identified Yfr1 sequences from Prochlorococcus MED4, MIT9313 and Synechococcus WH8102 plus the novel putative orthologs from the marine Synechococcus/Prochlorococcus were subjected to a comparative sequence/structure analysis. As a result we derived an initial sequence/structure-model (Fig. 1B) consisting of two stem loops and a central unpaired section of 18–19 nt. Interestingly, both loop sequences appeared unconserved and therefore these were set to aNy nucleotide in our subsequent searches. In contrast, we noted the presence of an eleven nt highly conserved sequence stretch within the unpaired central section. We reasoned if this sequence element would be essential and tried different RNA motif descriptors in which the undecanucleotide had to be fully conserved or allowed one or two mismatches. Furthermore, we relaxed the required length of the central unpaired section to a range of 12–25 nt and the lengths of both terminal helices to 5–10 base pairs and permitted a single bulge or mismatch in the 5' helix as observed in the data in Fig. 1. This model was used to search for orthologs within genomes of other cyanobacteria and also other eubacteria. Candidates found by this method were incorporated into the model and, finally, this procedure led to the identification of putative Yfr1 orthologs in 31 (28 new) out of 33 available cyanobacterial genomes. All here identified Yfr1 candidates are summarized in a multiple sequence/structure alignment shown in Fig. 2. An outstanding property appearing in this alignment is the perfectly conserved 11 nt long sequence motif (5'-ACUCCUCACAC-3') within the unpaired region. The reason for this perfect conservation might be that this short sequence is targeting a conserved sequence element in one or several other RNAs, mRNAs or ncRNAs, through base-pairing interactions as it is known for many ncRNAs. Alternatively, it could also be a protein binding motif. The rather high concentration of Yfr1 in the different organisms is rather untypical for an ncRNA acting through base-pairing.

From this set of orthologs, a comprehensive sequence/structure-model for Yfr1 was derived (Fig. 3). The 5'-end forms a 6–10 bp stem-loop, which may be interrupted by one bulge or internal loop. At the 3'-end the sequences show a 6–9 bp GC-rich hairpin, followed by a U-tail. These features make this element very likely to be a Rho-independent terminator. Furthermore, together with the 5'-terminal helix the 3'-helix may ensure the unpaired state of the conserved sequence motif and/or protect the RNA against degradation. One would expect approximately 0.5 instances of a specific 11 nt motif in a 2 MB genome and equal base distribution of 0.25. The here identified ultraconserved 5'-ACUCCUCACAC-3' occurs more often then by chance in the investigated genomes (for instance, three times in Prochlorococcus sp. MED4, genome size 1.66 MB; four times in Synechocystis PCC6803, genome size 3.57 MB, but also twice in Prochlorococcus sp. SS120 where it is not related to Yfr1 and is not located within a comparable context of possible secondary structure). Therefore, this motif alone is not sufficient to find homologs of Yfr1. However, it is interesting to note that the two additional occurrences in Prochlorococcus MED4 both are located in intergenic spacer regions, close to two other ncRNAs: 160–150 nt upstream the mapped 5' end of Yfr2 and 128–138 nt upstream of Yfr5 7. The only two cyanobacterial genomes for which no Yfr1 ortholog was found were Prochlorococcus sp. SS120 and MIT9211. These latter two examples were chosen to test when the RNA motif prediction would find likely false positives. When the score was reduced to 0.0 still no candidate Yfr1 homologs were found in these two genomes. Only when in addition one or two mismatches were allowed in the central consensus element, candidates were found (see additional file 1). However, these candidate sequences could not be aligned to each other or to the other Yfr1 sequences. In case of Prochlorococcus sp. SS120 also a previous experimental screen had remained negative 7. Therefore, the candidates found under these very relaxed conditions (additional file 1) do not appear realistic. Thus, the here desribed search method allows for an excellent discrimination between true positives and false positives. This finding is in agreement with studies in which RNA motif prediction was integrated into computational pipelines for the high-throughput prediction of cis-regulatory RNA sequences 11.

Conservation of genomic location and flanking genes can also be a powerful tool for finding related ncRNAs. Indeed, the yfr1 gene is at a conserved position in the majority of the here investigated genomes. In 25 out of 31 genomes, it is found upstream a guaB gene, coding for inosine-5'-monophosphate dehydrogenase (Table 2). Even more obvious is the frequent association with a downstream located trxA gene (27 of 31 genomes, Table 2).

We extended the search also to non-cyanobacterial photosynthetic bacteria, such as the alpha-proteobacteria Erythrobacter litoralis, Rhodobacter sphaeroides, Silicibacter sp. TM1040, the chlorobiaceae Chlorobium tepidum, Chlorobium chlorochromatii and Cytophaga hutchinsonii, as well as one spirochaete (Leptospira borgpeterseni) and, finally, Escherichia coli, but did not identify a reasonable ortholog of Yfr1 within these. So far, the occurrence of Yfr1 seems to be restricted to cyanobacteria. Since cyanobacteria represent a separate eubacterial phylum for at least 2.5 billion years of evolution it is very likely that Yfr1 originated very early in the progenitors of this group and must have co-evolved with its (hypothetical) target(s) for a very long time.

According to morphologic criteria, cyanobacteria can be organized into five different sections 12. These sections are only partially congruent with molecular phylogenetic data, therefore we chose six cyanobacterial species which clearly are very different from each other, judged by their morphology, 16S rRNA sequence and life style, to test our result by Northern hybridization. We also included RNA from Microcystis PCC7806 for which no genome sequence was available and, hence, no candidate predicted. The results are given in Fig. 4 and clearly show that Yfr1 is present in all tested species. Together with the three previous identified ones 7 our experiment raises the number of validated Yfr1 homologs to ten. In our previous study we did not identify an Yfr1 homolog in Prochlorococcus SS120. This is substantiated by the result of this study as we also do not predict one here. Summarizing, we found reasonable orthologs of Yfr1 in 31 out of 33 cyanobacterial strains and, furthermore, their existence could be validated for ten out of ten in this and a previous study.

Superimposing our results with a phylogenetic analysis based on 16S rRNA yields the tree shown in Fig. 5. The widespread existence of Yfr1 over large evolutionary distances points to an important function of Yfr1 throughout the cyanobacterial domain. Interestingly, Prochlorococcus SS120 and MIT9211, neither of which is predicted to hold an Yfr1 homolog, appear in one separate cluster. This lets us conclude, that these two or a common ancestor of them lost Yfr1 secondarily. This may be related to their special habitat. Different Prochlorococcus isolates can broadly become subdivided in an ecotype adapted to low light and another one adapted to high light. These ecotypes are genetically and physiologically distinct 13 and show a distinct distribution under natural conditions 1415. Prochlorococcus SS120 and MIT9211 have both been isolated from greater depths in the water column and are adapted to very low light conditions.

Marine unicellular cyanobacteria were grown as previously published 7. Cultures of Microcystis aeruginosa, Synechococcus elongatus PCC 7942, Thermosynechococcus elongatus BP-1, Synechocystis sp. PCC 6803, Nostoc sp. PCC 7120 and Gloeobacter violaceus were obtained from the Pasteur Culture Collection and one, Nostoc punctiforme ATCC 29133, from American Tissue Culture Collection. All cultures were grown as recommended by the Pasteur Culture Collection.

Total RNA was isolated as previously described 7 but with modified lysis conditions for PCC strains, Anabaena and Nostoc as follows: Disruption by adding an equal volume of glass beads, 33.3 μl 20 % SDS solution and 583 μl acidic phenol to 500 μl concentrated cell solution, following several cycles of vigorous agitation, freezing in liquid nitrogen and thawing in a water bath. Centrifugation of the mixture for 15 min at maximal speed at 4°C yielded an upper aqueous phase which could be cleaned up by standard phenol-chloroform-extraction. Finally, the phases were separated by centrifugation at 9.000 rpm for 15 minutes at 4°C. The RNA was precipitated from the aqueous phase with 100 μl isopropanol for several hours at -20°C, pelleted by centrifugation, washed with 75% ethanol, dried and resuspended in 100 μl of DEPC-treated water. Total RNA was separated in 10 % polyacrylamide-urea gels. Polyacrylamide gels were stained with ethidium bromide (0.3 μg/l) in 1× TBE buffer, rinsed with water and analyzed with a Lumi-Imager F1 system (Roche). Transcript sizes were determined by correlation to MspI-digested DNA of plasmid pUC19.

The Nostoc punctiforme ATCC 29133 (=PCC 73102) genome sequence was downloaded from the JGI website 16. Synechococcus RCC307 and WH7803 were obtained from the Genoscope 17.

All other sequences were obtained from the finished or unfinished genomes website at Genbank 18 with the following accession numbers: Anabaena variabilis ATCC 29413, CP000117; Crocosphaera watsonii WH8501, AADV00000000; Gloeobacter violaceus PCC 7421, BA000045; Lyngbya aestuarii CCY9616; Nodularia spumigena CCY9414, Anabaena PCC 7120, NC_003272; Prochlorococcus strains: MIT9312, NC_007577; NATL2A, NC_007335; MIT9313, NC_005071; SS120, NC_005042; MIT9211, AALP00000000; MIT9303, NC_008820; AS9601, NC_008816; NATL1A, NC_008819; MIT9515, NC_008817; MED4 (=CCMP1986), NC_005072; Synechococcus strains: PCC7942, NC_007604; PCC 6301, NC_006576; OS-B' (JA-2-3B'a(2-13)), NC_007776; OS-A (JA-3-3Ab), NC_007775; CC9605, NC_007516; CC9902, NC_007513; WH8102, NC_005070; WH7805, AAOK00000000; WH5701, AANO00000000; RS9917 (=RCC556), AANP00000000; RS9916 (=RCC555), NZ_AAUA00000000; Synechococcus BL107, NZ_AATZ00000000; Synechococcus CC9311, NC_008319; Synechocystis sp. PCC 6803, NC_000911; Thermosynechococcus elongatus BP-1, NC_004113; Trichodesmium erythraeum IMS101, NC_008312.

Multiple sequence alignments were generated using ClustalW 19 with default parameters for DNA. Comparative structure prediction was done with RNAlishapes 20, a tool which predicts a consensus structure for a set of aligned sequences by taking covariance and free energy into account. The resulting consensus structure was analysed together with the multiple sequence alignment using RALEE 21. The latter served also for manual optimisation of the alignment and the consensus structure, respectively, and for the production of color annotated alignments. Homology searches were done using RNAMotif 22 based on a combined sequence structure motif description for Yfr1. The final RNAMotif descriptor was:

descr

   h5(minlen = 5, maxlen = 10, mispair = 1, pair+={"g:u","u:g"}) ss(minlen = 3, maxlen = 8) h3

   ss(minlen = 12, maxlen = 20, seq="ACTCCTCACAC", mismatch = 0)

   h5(minlen = 6, maxlen = 9, mispair = 0, pair+={"g:u","u:g"}) ss(minlen = 3, maxlen = 15) h3

   ss(seq=" ^ [AUCG]UU$")

score

   { gcnt = 0;

      len = length(h5 [5]);

      loop = length(ss [6]);

      for(i = 1; i < = len; i++){

         j = len - i + 1;

         b1 = h5 [5, i, 1];

      b2 = h3 [7, j, 1];

         if((b1 == "g" && b2 == "c") || (b1 == "c" && b2 == "g"))

         gcnt++;

      }

   # require 65% GC in the stem!

      SCORE = 1.0 * gcnt/len;

      if(SCORE < .65)

      REJECT;

   }

Yfr1_multi_2 5'-GTGTGGTGTGAGGAGTGAACGGAA-3'

Yfr1_gloeo_2 5'-CATGGTGTGAGGAGTGAACGGAAAC-3'

SsA_Thermorev 5'-GCGACGCCGTTTTACCT-3'

SsaA_6803rev 5'-CACCACGCCGTTTTACCT-3'

SsaA_7120rev 5'-CGCAACGCCGTTTTACCT-3'

Hybridization was performed at 54°C for Yfr1 and 50°C for 6S RNA in hybridization buffer (50% deionized formamide, 7% SDS, 250 mM NaCl, 120 mM Na(PO₄), pH 7.2).