Introns in protein-coding genes and pre-mRNA splicing are ubiquitous in Eukarya and, to a lesser extent, in Bacteria. Until 2001, pre-mRNA splicing had not been reported in Archaea. In 2002, we reported the first examples of archaeal pre-mRNA splicing for homologs of the eukaryotic CBF5 (centromere binding factor 5 in yeast, or dyskerin in humans) protein in Aeropyrum pernix, Sulfolobus solfataricus, and S. tokodaii [1, also in S. acidocaldarius, see 2 in 2006]. We found that the cleavage of the pre-mRNA depends on the recognition of a bulge-helix-bulge (BHB)-like structure in the precursor 12 by the splicing endonuclease EndA 3. In Archaea, pre-tRNA and pre-rRNA splicings also depend on the same system [45; reviewed in 1]. Although most species from the orders Desulfurococcales and Sulfolobales have the (putative) cbf5 intron, H. butylicus, P. occultum, P. fumarii, and I. islandicus in the order Desulfurococcales do not contain the intron 12. This observation suggested putative secondary loss of the intron. However, phylogenetic analysis of the Cbf5 protein sequences did not resolve the relationships between species from different orders of Crenarchaeota, likely due to the short sequence (about 70 amino acid residues) studied in the analysis 2.

In the present study, we determined a formerly undetermined region of cbf5 sequences from the previously characterized 27 species and new sequences from an additional 16 species. We studied 43 species, which were almost all the available species from type culture collections. We determined up to two-thirds of the coding region, corresponding to about 220 amino acid residues, and then examined the timing of the gain and the possible loss of the intron in the archaeal protein-coding gene. We found that the intron existed in the cbf5 gene in the common ancestor of the orders Desulfurococcales and Sulfolobales, and then the intron was lost in some lineages in the order Desulfurococcales.

Our previous analysis of crenarchaeal cbf5 genes showed that only orders Desulfurococcales and Sulfolobales have the (putative) intron in their cbf5 genes, although some species in the order Desulfurococcales do not have the intron. However, phylogenetic analysis with the previous dataset did not strongly support the sister grouping of orders Desulfurococcales and Sulfolobales without species from other orders, and the phylogenetic positions of the species in Desulfurococcales, which do not have the intron, were unclear 2.

To improve the phylogenetic analysis of the cbf5 gene, we extended the analyzed region of the genes from 27 species to include an additional area in the 3' region (from about 70 to 220 amino acid residues), and we added new sequences from an additional 16 crenarchaeal species. We also added the recent information from the newly determined crenarchaeal and korarchaeal genomes. The species and the intron size information are summarized in [Additional file 1]. When the presence of the intron was expected, the new putative exon-intron borders from seven species among the additional 16 species were subjected to a prediction of their secondary structures (Figure 1. For 18 species which have the (putative) intron among the previously characterized 27 species, see reference 2). Except for the cases of 'Caldococcus noboribetus' and Acidianus brierleyi, the predicted structures in the pre-mRNAs have an unconventional BHB structure 28, which should be recognized and cleaved by the hetero-oligomeric splicing endonuclease, as demonstrated previously 2. Recent X-ray crystallography has revealed that hetero-oligomeric splicing endonuclease is a dimer of hetero-dimers 29. The predicted cleavage sites between the second and the third residues in the bulges of the BHB motif were consistent with the expected exon-intron borders, suggesting that the predicted exon-intron borders were convincing. In fact, partial cDNA sequences of spliced cbf5 mRNA from Desulfurococcus amylolyticus, Desulfurococcus mucosus, Staphylothermus hellenicus, Acidianus brierleyi and Ignisphaera aggregans were consistent with the predictions (Watanabe, Y. and Itoh, T. unpublished results), although the definite identification of the borders of the remaining species requires a cDNA sequencing and cleavage study using splicing endonuclease. Results from our present study, together with the previous study 2, indicate that among the order Desulfurococcales, Ignicoccus spp. and all species from family Pyrodictiaceae do not have the cbf5 intron.

Using a new dataset, we reconstructed phylogenetic trees of the cbf5 protein sequence by using maximum likelihood (not shown) and Bayesian methods [Additional file 4]. These trees suggested the monophyly of the cbf5 protein sequences from orders Desulfurococcales and Sulfolobales. We verified this monophyly with several statistical tests (analysis 1, [Additional file 5]). To finish the computation within a reasonable time (approximately 1 week) using the available computational environment with a reduced number of trees to be considered, we first reduced the number of sequences in the dataset and reconstructed the phylogenetic tree (Figure 2). There was no significant difference in the tree topology before and after the reduction of the sequence (compare [Additional file 4] and Figure 2). Then, we fixed the relationships within each of the eight groups (Figure 2) and examined the relationships between the groups (analysis 1, Additional file 5). The results of the tests supported the monophyly of the sequences from orders Desulfurococcales and Sulfolobales (AU; P = 0.938, NP; P = 0.799, KH; P = 0.907) and also suggested the inclusion of the sequence of 'Korarchaeum' into the crenarchaeal sequences. The result is consistent with the phylogenetic association of rRNA and protein sequences from 'Korarchaeum' and Crenarchaea 30.

The sequences from the species of family Pyrodictiaceae and Ignicoccus spp. are grouped independently, and these monophylies were strongly supported with high statistical values in the trees (Figure 2, see also [Additional files 5 and 6]). Although among orders Desulfurococcales and Sulfolobales, these groups are not likely to be the earliest branching (Figure 2, see also [Additional file 4]), the branching order among order Desulfurococcales, particularly of Ignisphaera aggregans, was uncertain. Thus, we examined whether the sequences of family Pyrodictiaceae and/or Ignicoccus spp. branched earliest among the order Desulfurococcales, except for Ignisphaera aggregans, by using AU, NP, and KH tests of an alternative grouping set (analysis 2, Figure 2, [Additional file 7]). The monophyly of the Desulfurococcaceae (i.e., the earliest branching of the Pyrodictiaceae sequence) was rejected by the AU test (P = 0.029) and NP test (P = 0.001) (95% significance level) but not by the KH test (P = 0.075). If Ignisphaera aggregans was not considered, the monophyly of the Desulfurococcaceae (excluding Ignisphaera aggregans and Pyrodictiacean species) would be supported by only small probabilities by the AU test and KH test (P = 0.062, and 0.071, respectively) and rejected by the NP test (P < 0.001). The monophyletic grouping of the Desulfurococcaceae (group d in Figure 2) with the intron and the Pyrodictiaceae was supported by the AU, NP, and KH tests (P = 0.831, 0.697, and 0.829, respectively). These results suggest that the sequences of the Pyrodictiaceae (as seen in the Bayesian tree of Figure 2) are unlikely to be the earliest branching. The monophyletic grouping of Desulfurococcaceae (c) with the intron and Ignicoccus spp. (as seen in the tree of Figure 2) was also supported by the AU, NP, and KH tests (P = 0.82, 0.605, and 0.78, respectively). These results also suggest that the sequence of Ignicoccus spp. is not likely to be the earliest branching as seen in the Bayesian tree of Figure 2. The monophyly of Desulfurococcaceae (b) + Desulfurococcaceae (c) (appeared in the Bayesian tree of Figure 2) could not be rejected by the AU and KH tests because of their medium probabilities (P = 0.155 and 0.187, respectively), but this monophyly was rejected by the NP test (P = 0.02). The monophyly of Ignisphaera aggregans + Pyrodictiaceae also cannot be rejected by the AU, NP, and KH tests because of their medium probabilities (P = 0.313, 0.212, and 0.219, respectively). The monophyly of Desulfurococcaceae (c and d) + Pyrodictiaceae was not rejected by the tests (AU; P = 0.287, NP; P = 0.078, KH; P = 0.163). Finally, the monophyly of species with the intron was not rejected by the tests (AU; P = 0.329, NP; P = 0.058, KH; P = 0.194). Therefore, the sequence of both Ignicoccus spp. and the Pyrodictiaceae was unlikely to be the earliest simultaneous branching, as seen in the tree presented in Figure 2. These results suggest that the sequences of these groups are not likely to be the earliest branching, although the possibility was not completely excluded. As a reference, we constructed a phylogenetic tree of 16S rRNA of the corresponding species by using the Bayesian method [Additional file 8]. The 16S rRNA tree also supported the monophyletic groupings of orders Desulfurococcales and Sulfolobales, Ignicoccus spp. and Desulfurococcaceae (c), and Pyrodictiaceae and Desulfurococcaceae (d), suggesting that there was no obvious gene transfer of cbf5 from outside of orders Desulfurococcales and Sulfolobales. About 6% of protein-coding genes in Ignicoccus hospitalis are thought to be transferred from its symbiont 'Nanoarchaea' 31. However, in our analysis, the monophyletic grouping of cbf5 genes in Ignicoccus spp. with the nanoarchaeal sequence was not supported. Thus, the cbf5 gene in Ignicoccus spp. is not likely due to gene transfer of the intron-less nanoarchaeal cbf5 gene.

We also aligned the (putative) cbf5 introns with the flanking sequences using the program R-coffee with the RNA secondary structure prediction option (Figure 3). The alignment showed some conservation in the intron region beyond base-pairing with the exon regions to maintain the motif required for cleavage by the splicing endonuclease, suggesting a common origin for these introns. Note that the internal region of the introns was highly variable likely due to the independence of recognition by the splicing endonuclease during the cleavage at the exon-intron borders.

The origin of the archaeal cbf5 intron is still unclear. We previously proposed that relaxed substrate specificity 2323334 of the hetero-oligomeric splicing endonuclease 335 led to the birth of the pre-mRNA intron, which frequently contains the relaxed cleavage motif (2 and this study). In particular, the recognition of the relaxed cleavage motif within a non-tRNA context has been shown to be characteristic of crenarchaeal hetero-tetrameric splicing endonuclease 2293233. Although the intron sizes in cbf5 and rRNA are different from one another, as discussed below, archaeal rRNA introns are observed mainly in crenarchaeal species, which are expected to have the crenarchaeal hetero-tetrameric splicing endonuclease 36. In some cases, archaeal rRNA introns also have the relaxed cleavage motifs 37. The size of archaeal tRNA introns (11 to 175 nucleotides) are more similar to those in crenarchaeal cbf5, and accumulation of tRNA introns in crenarchaeal species is observed 36. The unconventional cleavage motif at the exon-intron borders and the intron location at the position rather than the usual position "37/38" of tRNA intron are also observed more frequently in crenarchaeal species 2836. The contribution of the hetero-tetrameric splicing endonuclease is suggested for the cleavage of the unconventional motif, and has been demonstrated by the crenarchaeal hetero-tetrameric splicing endonuclease (reviewed in 29).

Numerous archaeal rRNA introns contain the open reading frame for DNA endonuclease, which functions as a homing endonuclease to make the intron as a mobile element (reviewed in 38). Apparently, the archaeal cbf5 intron is too short (from 16 to 44 bp, see [Additional file 1]) to encode such a nuclease. Nomura et al. found that A. pernix isolates have variations in the number, sequence, and positions of rRNA introns 6 (see also Table 2). In the present study, we determined partial cbf5 sequences of these A. pernix isolates. Together with the results of the previous studies for type strain K1 1, we found that at the corresponding positions, all of the analyzed cbf5 genes have a putative intron, classified as type 1 or type 2 (Figure 4, the distribution is mentioned in Table 2), which contains only two base substitutions. There was no correlation between the variation of cbf5 and rRNA introns (Table 2). Although sequence variation of rRNA introns between A. pernix isolates (one to two substitutions in I beta or one substitution in I epsilon) were observed, this was not correlated with the variation of the cbf5 intron. However, a correlation between the cbf5 intron and radA phylogeny shown by Nomura et al. 6 was observed (not shown). Our results show that, as for the large-scale in-del event, the cbf5 intron was more conserved than the rRNA introns with the homing DNA endonuclease gene. However, Nomura et al. 6 also found that some of the rRNA introns are deletion derivatives of the introns with an open reading frame. For example, A. perinix introns I delta and I zeta are deletion derivatives of I alpha and I gamma, respectively 6. The contemporary cbf5 introns may be examples of such deletion derivatives. Proof of this possibility requires further taxonomic sampling of cbf5 genes to find the intron that includes the protein-coding sequence.

Peng et al. showed that during the generation of infection, putative 12-bp introns were inserted into protein-coding genes in an archaeal virus genome, although splicing was not demonstrated and the mechanism of insertion of the 12-bp sequence is unknown 39. Interestingly, the sizes of the cbf5 introns from Staphylothermus hellenicus and S. marinus are 36 bp (3 times 12); thus, mechanisms of insertion of archaeal cbf5 introns and the putative introns in the archaeal virus genome may be related. Furthermore, the cbf5 introns of Stetteria hydrogenophila (33 bp) and Ignisphaera aggregans (39 bp), as well as S. hellenicus and S. marinus, do not change the reading frame. The putative introns in the virus genome may not be spliced out and the coding region with such insertions may produce functional proteins. However, in the case of cbf5 introns, the insertion disrupts the codon of the catalytic residue of the protein 140, and thus these must be spliced out if the organism needs the functional protein.

One possible explanation of the putative secondary loss of the cbf5 intron in certain lineages is that the intron-containing gene is replaced with a sequence without the intron, possibly produced by reverse transcription of the spliced mRNA 41, or the spliced mRNA itself. Although reverse transcriptase activity has not been observed in crenarchaeal cells, the presence of a putative reverse transcriptase gene in some archaeal genomes has been suggested 42. In fact, in the sequenced genomes of Ignicoccus hospitalis and Hyperthermus butylicus (family Pyrodictiaceae) with the putative secondary loss of the cbf5 intron, candidate reverse transcriptase genes were identified [Additional files 9 and 10]. An alternative possibility could be the requirement of higher activity of pseudouridine synthase in a certain environment. Previously, we proposed that the cbf5 intron functions as a negative regulator of the expression of pseudouridine synthase 1. Archaeal Cbf5 catalyzes pseudouridine formation in rRNA and tRNA together with other associated proteins using a guide RNA 4043 or without a guide RNA 44. Incorporation of pseudouridine in RNA increases the thermodynamic stability of RNA 45. Furthermore, pseudouridylation of tRNA at position 55 by TruB in mesophilic bacteria Escherichia coli supports the resistance to higher temperature 46. Archaeal Cbf5, a member of truB family 47, also forms a pseudouridine in tRNA at position 55 44. Thus, at extremely high temperatures, the organisms might not prefer the down-regulation system of the pseudouridine synthase and lose it.

The results of the present study suggest that cbf5 gained the intron in the common ancestor of orders Desulfurococcales and Sulfolobales, and that cbf5 lost the intron independently in the ancestors of the family Pyrodictiaceae and Ignicoccus spp. Since we found the first examples of cbf5 introns, sequences of three crenarchaeal genomes with the cbf5 intron have been determined. However, the cbf5 intron in these genomes was misidentified (S. acidocaldarius; 48, see 2) or ignored (Staphylothermus marinus 7, Metallosphaera sedula, [Genbank:CP000682]). Even for the first three examples in A. pernix, S. solfataricus, and S. tokodaii, the gene prediction of these examples was still confused with cases of translational frame-shifting by other researchers 49. Although there was no confirmation of archaeal pre-mRNA splicing for genes other than cbf5, the presence of the putative intron in other protein-coding genes was predicted 3950. To completely understand protein-coding genes in archaeal genomes, tools for effective prediction of introns in archaeal protein-coding genes must be developed with comparative or computational methods 5051. Experimental confirmation of the predictions, including the putative cbf5 introns predicted in our studies, is indispensable.

YW conceived the study and participated in its design, carried out the molecular genetic studies, participated in the sequence alignment, and drafted the manuscript. S. Yokobori participated in the design of the study, the sequence alignment, performed the statistical analysis, and helped draft the manuscript. TI, S. Yoshinari, and NN carried out the molecular genetic studies and helped draft the manuscript. YS, AY, TO, and KK participated in the design and coordination of the study and helped draft the manuscript. All authors read and approved the final manuscript.