|Home | About | Journals | Submit | Contact Us | Français|
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
Naturally occurring RNAs contain numerous enzymatically altered nucleosides. Differences in RNA populations (RNomics) and pattern of RNA modifications (Modomics) depends on the organism analyzed and are two of the criteria that distinguish the three kingdoms of life. If the genomic sequences of the RNA molecules can be derived from whole genome sequence information, the modification profile cannot and requires or direct sequencing of the RNAs or predictive methods base on the presence or absence of the modifications genes.
By employing a comparative genomics approach, we predicted almost all of the genes coding for the t+rRNA modification enzymes in the mesophilic moderate halophile Haloferax volcanii. These encode both guide RNAs and enzymes. Some are orthologous to previously identified genes in Archaea, Bacteria or in Saccharomyces cerevisiae, but several are original predictions.
The number of modifications in t+rRNAs in the halophilic archaeon is surprisingly low when compared with other Archaea or Bacteria, particularly the hyperthermophilic organisms. This may result from the specific lifestyle of halophiles that require high intracellular salt concentration for survival. This salt content could allow RNA to maintain its functional structural integrity with fewer modifications. We predict that the few modifications present must be particularly important for decoding, accuracy of translation or are modifications that cannot be functionally replaced by the electrostatic interactions provided by the surrounding salt-ions. This analysis also guides future experimental validation work aiming to complete the understanding of the function of RNA modifications in Archaeal translation.
Post-transcriptional modification of transfer and ribosomal RNAs is essential for their cellular activities as core molecules of the translation apparatus. To date, the chemical structure of more than one hundred RNA modifications have been identified in all domains of life [1-3]. In transfer RNAs, modified nucleotides are found predominantly within the 3D-core of molecules and in the anticodon arm, especially at the wobble position 34 and at position 37, 3' adjacent to the anticodon (conventional numbering of tRNA positions is as defined in , http://www.tRNA.uni-bayreuth.de). These particular modifications allow the molecules to adopt the canonical L-shaped conformation and modulate interactions with various interacting macromolecules such as aminoacyl:tRNA-synthetases, initiation, elongation and termination factors, mRNA and/or elements of the decoding and peptidyl-centers of the ribosome (reviewed in [5-10]). In ribosomal RNAs, modified nucleotides are located mostly in regions corresponding to the functional centers of the ribosome [11-14]. Their location suggests a role in accuracy and efficiency of translation, however the specific function of each modified nucleoside is still largely unknown. This lack of knowledge stems from peculiarities of the rRNA molecule itself: it is a large molecule (molecular mass between 1.5 to 3.9 MDa); some nucleotides are only partially modified and their function(s) are most certainly dependent on a network of synergistic interactions with different elements of the ribosome, including other modified nucleosides that may act cooperatively. Nevertheless, function has been attributed to modified nucleosides in rRNA in a few cases [13,15-22].
Difference in profile and type of RNA modifications (Modomics) is one of the criteria that distinguish the three kingdoms of life. While universal modifications such as Ψ, m5U, t6A or m1G are found in a large numbers of archaeal, bacterial and eukaryal tRNAs, each kingdom has a set of signature modifications. For examples mimG, G+, m22Gm, ac6A or m1Ψ are typical of archaeal tRNAs, while yW, mcm5U and manQ, or k2C, mo5U and m6t6A are typical of tRNAs from Eukarya or Bacteria respectively (for review see Figure 8.1 in ). The same conclusion applies for modified nucleotides in rRNAs (see ; http://people.biochem.umass.edu/fournierlab/snornadb/main.php).
In Archaea, our knowledge of the diversity of RNA modifications is largely founded on the lifework of Jim McCloskey and Pamela Crain, who analyzed bulk tRNA and rRNA preparations from a phylogenetically diverse set of Archaea. The technique used combined separation of nucleosides of bulk RNA RNase hydrolysate by liquid chromatography, followed by comparison of the derived modified nucleosides to synthetic ones by mass spectrometry techniques  (for more recent development of the technique, see  and references therein). However, to date Haloferax volcanii, a Halobacteriaceae that lives optimally at 42°C in the presence of 1.5–2.5 M NaCl , is the only Archaea for which both the chemical identities and positions of almost all modified ribonucleosides have been mapped for nearly the whole set of the 52 sequenced tRNAs with distinct anticodons [28,29]. In addition, 13 tRNA sequences of two closely related mesophilic halophiles are available, Halobacterium cutirubrum (12 sequences) and Halococcus morrhuae (one sequence) . For H. volcanii ribosomal RNAs, the type and position of modifications are available in the case of the 16S RNA [30,31], but not for the 23S nor the 5S RNAs. These can be inferred from studies on another closely phylogenetically related halophilic Archaea Haloarcula marismortui [32,33]. However, while the RNA modifications have been mapped in RNAs of halophiles, including H. volcanii, the identity of the genes that code for the corresponding RNA modification enzymes remains largely ignored.
Using a comparative genomic analysis method, that we have recently applied to the only other organism with an almost complete set of sequenced tRNAs, the pathogenic bacteria Mycoplasma capricolum , we set out to predict all the RNA modification genes in the halophilic archaeaon H. volcanii. Some were easily predicted by homology with experimentally validated RNA modification genes from other organisms, while a few are original predictions based on comparative genomic analysis  (not based on homology). This computation work provides predictions that can now guide the experimental validation work with the goals of elucidating the role of RNA modifications in Archaeal translation, and ultimately obtaining a better understanding of the emergence of this extraordinary complex enzymatic machinery during evolution.
Thanks to the "tour de force" of Gupta [28,29], a list of almost all of the modified nucleosides present in the different tRNAs of Haloferax volcanii, a typical mesophilic halophile, is available. Figure Figure1A1A shows their distribution (identity and location) in the general 2D-cloverleaf structure of tRNA, while Figure Figure1B1B shows their positions in a schematic 3D-architecture model. The modifications that are unique to archaeal tRNAs are shown in gray. For example G+-15 (for Archaeosine), C*-34 (for a lysidine-type of nucleoside) and m1Ψ-54 are unique to all archaeal tRNAs analyzed so far, both by their chemical structure and their positions in the nucleic acid, while others such as m22G-10, Ψ-22, Ψ-52, Cm-56 and m1I-57 (I for inosine) are unique only because of their position, rather than their chemical structure (see in [1,4], reviewed in ). Since tRNA modification enzymes are usually site-specific, we expect the corresponding genes to be different from those in the other kingdoms. In terms of chemical structure (and not their position), the modified nucleosides m22G, m5C and m1I found in tRNAs of H. volcanii are characteristic of eukaryal rather than eubacterial tRNAs, while D (for dihydrouridine), I (for inosine), m5U, i6A (i for isopentenyl), Q (for queuosine) and m7G, which are common in tRNAs of Bacteria and Eukarya, are absent in H. volcanii (but not necessarily in other Archaea – see Fig 8.1 in ). Also, as mentioned in the Methods section, m1G at position 9 and m5C at positions 50–52 are found in tRNAs from another halophile, H. cutirubrum, but absent in these positions in all H. volcanii tRNAs.
As illustrated in Figure Figure1B,1B, the modified nucleosides in tRNA can be classified in two categories: those that are present in the 3D-core (gray background) and presumably implicated mostly in the formation and/or the control of flexibility of the L-shaped molecule (reviewed in [9,36]); and those present in the decoding region (anticodon hairpin), implicated in the efficacy and accuracy of interaction with selected amino-acyl tRNA synthetases (reviewed in ) and the various codons within the mRNA:ribosome complex (reviewed in [5,8]).
The identification of a modified nucleotide does not imply it is present in a one to one ratio with the RNA molecule. Indeed, the presence and final chemical structure of certain modified nucleotides, particularly the hypermodified ones, may vary according to the physiological constraints of the cell (aerobic/anaerobic conditions, temperature, availability of intermediate metabolites or cofactors of the modification enzymes, various metabolic stress conditions; discussed in: [37-41]). The A-15, C-34, U-52 and U-54 residues in some H. volcanii tRNAs were reported to be only partially modified into G+-15, ac4C-34 (ac for acetyl), Ψ-52 or Ψ-54/m1Ψ-54 respectively, giving rise to distinct iso-tRNA species that sometimes can be separated by liquid chromatography or 2D-gel electrophoresis . When a modification requires multiple modification enzymes like m1Ψ-54, G+-15 and few U-34 derivatives (see below), only intermediate products may exist under certain physiological conditions. However, the genes corresponding to all of the expected modified nucleotides (present or not) in the cellular tRNA population should be present in the genome.
In their early work, Gupta and Woese identified four positions with modified bases in H. volcanii 16S RNA. These were later confirmed  and identified as acp3U-910 (position 966 by E. coli numbering; acp for 3-amino-3-carboxypropyl, thus an amino acid) in hairpin 31 located in the 3' major domain, m6A-1432 (position 1500) in helix 44, the tandem m62A-1450 and m62A-1451 (positions 1518 and 1519) in hairpin 45 and a modified cytidine (C*) of still unknown structure (MW:330.117 as determined by mass spectrometry) at position 1352 (1404) in helix 44 in the 3' minor domain of SSU RNA. Their locations are shown in the schematic 2D-structure (Figure (Figure2A)2A) and 3D-structure (Figure (Figure2B)2B) of 16S rRNA.
The characteristic pair of tandem dimethylated adenosine (m62A m62A) is universally present at analogous positions in rRNA of all organisms examined so far. These are located at the interface of the two ribosomal subunits [11,12,14] and their formation may serve as a checkpoint in quality control of ribosome biogenesis [42-44]. Likewise, acp3U-910 (966) in hairpin 31 appears to be nearly universally modified, although the type of base and corresponding modification vary from one organism to another: m2G 3'-adjacent to a m5C in 16S RNA of both E. coli and Thermotoga maritima , m22G 3'-adjacent to m5C in Thermus thermophilus , m1Ψ-acp3U in Drosophila melanogaster SSU RNA and designated as unknown modified nucleoside in SSU RNA of other organisms, mostly archaeons . This modified nucleotide is above the P-site-bound tRNA and directly contacts the anticodon stem-loop of tRNA at position 34 [47-51], and is also often modified (see Figure Figure2A2A and Additional file 1 ). Several studies indicate this nucleotide is important in decoding genetic information, particularly at the step of initiation [21,22,52].
Helix 44 is the dominant structural component of the 30S subunit interface. It's upper end lies just below where the mRNA transverses the subunit in the P site [53,54]. This portion forms a significant intersubunit bridge while at the same time is directly functionally important for efficient and accurate decoding since two bases, at least in the E. coli ribosome (bases 1492 and 1493) flip out of an internal loop in this region [53,54]. This allows the monitoring by direct contact of the mRNA-tRNA base pairing in the A site, a conformational transition facilitated by the binding of aminoglycoside antibiotic, e.g. paromomycin, to a pocket in the major grove of the top of helix 44 . Modified nucleoside m6A-1432 (1500 E. coli numbering) at the bottom of helix 44 is present in SSU RNA of most (if not all) Archaea, and only a few Eukarya, but never in Bacteria (for references see ). It is also termed a 'decoding site nt'  because it is present in the functionally significant region of helix 44, adjacent to a critical intersubunit bridge (B2a). Contrary to the others above, the unknown N-330 (C*1352; 1404 E. coli numbering) is found in Archaea and in many Bacteria, but not in Eukarya. While it directly contacts paromomycin bound to helix 44, its function remains an enigma and its chemical structure remains to be elucidated.
No data are available for H. volcanii 23S RNA modifications. We therefore used the analysis performed in the closely related organism Haloarcula marismortui [32,33,56] that led to the identification of modified nucleotides at eight positions. Their locations in the generalized schematic 2D and 3D-stucture of 23S rRNA are shown in Figures Figures2C2C and and2D2D.
Three Ψ residues are present: two of which, Ψ-1956 and Ψ-1958, are located at universally conserved positions (1915 and 1917; E. coli numbering) in helix 69 loop of domain IV. The helix 69 stem-loop contacts A- and P-site tRNAs, contributes to bridge regions B2a and B2b of 23S rRNA, is involved in translation termination, contacts ribosome recycling factor, plays an active role in dissociation of subunits at the end of translation, and is important for subunit association [17,33,49,57-62]. Specifically, Ψ-1956 (1915) contacts the D stem of tRNA in the A site (positions 11 and 12) and Ψ-1958 (1917) is immediately adjacent to bridge B2a contacts, as well as direct contacts to A-site tRNA; they are important for the conformational flexibility of helix 69 and their loss affects subunit association, dissociation, and translation termination [17,58-61]. The peptidyl transferase center of the large ribosomal subunit is the rRNA that directly surrounds the active site of peptide bond formation and is made of the rRNA of the central loop and proximal nucleotides of Domain V, as well as the A-loop (loop of hairpin 92) and P-loop (loop of hairpin 80, Figure Figure2)2) . The third Ψ is conserved only in eukaryotes and is located at position 2621 (2586 E. coli numbering) between hairpin 90 and 93 in the central loop of domain V, immediately adjacent to rRNA contacts with the P-site tRNA (terminal A of CCA at position 76) [49,63].
Three 2'-O-ribose methylations were also found: one Gm is located at position 1950 (1909) in the structurally conserved helix 69 of domain IV contacting nucleotide 12 of the tRNA in the P site , while the conserved tandem UmGm, is located at positions 2587/2588 (2552/2553 E. coli numbering) in hairpin 92, also called A-loop or peptidyltransferase loop of domain V, that contacts acceptor end of tRNA in the A-site of the ribosome. In the case of H. marismortui, mutagenesis studies  and X-ray crystallography  demonstrated the existence of base pairing between Gm-2588 (2553) and C-75 of CCA end of tRNA in the A-site of the ribosome, thus implicating a role in the peptidyl transfer reaction.
Lastly, two base methylations (m1A, a modified nucleoside bearing a positive charge and m3U) are present respectively at position 628 (571) in hairpin 25.1 of domain II and at position 2619 (2584) of domain V, between hairpins 90 and 93. The importance of each of these two modified nucleotides (not modified in E. coli) for maintenance of the archaeal ribosome architecture or translation is not known. However, since m3U- 2619 (2584) is in the central loop of Domain V and immediately adjacent to rRNA contacts with P-site tRNA, a fact revealed early on by affinity labeling results showing E. coli 2584 adjacent the CCA end of P-site RNA [65,66], this suggest a role in peptidyltransferase activity. Moreover, the absence of the U-methylation in the homologous position in H. salinarium 23S rRNA confers resistance to sparsomycin, and antibiotic that normally binds to the peptidyl transferase center .
In domain V of 23S RNA of H. salinarium, an additional Ψ-2606 (2580) at the bottom of helix 90, not present in rRNA of H. marismortui, has been reported . However, as no homolog of the bacterial RluC-type, or the eukaryotic Pus5p-type of enzymes responsible for the formation of this Ψ in the LSU rRNA of respectively E. coli and S. cerevisiae can be found in the genome H. volcanii (see below), its presence in the 23S rRNA of H. volcanii is highly unlikely.
As no modifications were detected in the 5S rRNA of the two halophilic archaea H. halobium and H. marismortui [68,69], we predicted that H. volcanii would also lack modifications in this rRNA. A 2'O-methylcytosine (Cm) at a conserved C-position (position 32) has been reported only in the 5S rRNA of the thermophiles Sulfolobus acidocaldarius  and S. solfataricus , while in the hyperthermophilic Pyrodictium occultum the base at the same location (position 35 in P. occultum) is further acetylated into ac4Cm. Both derivatives ac4C and ac4Cm coexist, indicating incomplete modification of C-35 under the conditions the cells were grown before extraction of the RNA [68,69]. The same is true for other modified nucleotides in the 16+23 S rRNAs.
The genome of Haloferax volcanii, strain D2 (4,012,900 nt) comprises one chromosome (2,847,757 bp) in several identical copies (up to 20 ) and four smaller plasmids (pHV1:85,092 bp, pHV2:6,359 bp, pHV3:437,906 bp and pHV4:635,786 bp). All tRNA genes are located on the chromosome. This mesophilic halophile exhibits the typical archaeal tRNA set  which is characterized by 46 distinct anticodons able to read all 61 sense codons (see details below). The extra G nucleotide at position 0 of tDNA-His (GTG) is encoded in the genome but none of the CCA 3' terminal sequence of tDNAs are present. The list of tDNA sequences in linear and cloverleaf forms is given in Additional file 2. Remarkably the sequences of each mature tRNA (as sequenced by Gupta [28,29]) and corresponding tDNAs as identified above perfectly match. The only sequences of mature tRNAs that are missing from Gupta's analysis are those specific for tRNA-Val (anticodon UAC), tRNA-Ser (UGA), tRNA-Thr (UGU), tRNA-Gln (UUG), tRNA-Arg (UCU) and tRNA-Arg (CCU), all but one harboring a T34 wobble base in the corresponding tRNA gene. As stated in the original work , the missing tRNAs probably correspond to minor isoacceptor species that co-migrated with one of the major species and therefore were impossible to isolate and sequence.
Six tRNA genes are present in two copies, raising the total number of tRNA genes from 46 to 52. Among these six pairs, five are perfect duplicates (from positions 1 to 73), while the two tDNA-Gly (GCC) differ by the two base pairs 4–69 and 5–68 (CG and TA versus TA and CG, respectively) as previously noted . Three of these tRNA pairs are organized in direct tandems with a short distance between the two genes: 2 × tDNA-Gly (GCC), 12 nt; 2 × tDNA-Asp (GTC), 29 nt; 2 × tDNA-Val (GAC), 45 nt probably revealing a recent gene duplication. The two tDNA-Ala (TGC) are each embedded in the two copies of the ribosomal operon (between 16S and 23S rRNA genes). Other tDNAs are randomly distributed throughout the genome; the next closest distance between two tDNAs being 96 nt.
As only one gene exists for the majority of tRNAs harboring each a distinct anticodon, large differences must exist either in the expression levels of individual tDNAs, or in the half-life of individual mature tRNAs (or both). Indeed the steady state concentrations within the cell of the major tRNAs (reading most used codons) must be higher than those of minor tRNAs (reading rare codons). The regulation of the expression of the different tDNAs is yet to be elucidated in H. volcanii and in all other Archaea (discussed in ). Its is possible that tRNA stability depends on factors similar to those identified in yeast (reviewed in ).
Only three genes carry introns and in contrast with many other archaea (see ), all are found at the canonical position 37/38. The three genes, tDNA-Met (ATG) (intron of 75 nt), tDNA-Gln (CAA) (intron of 31 nt) and tDNA-Trp (intron of 103 nt), display a nearly perfect hBHBh' motif  with the so-called h helix being the anticodon stem and the so-called h' helix being 3-, 8- and 2-bp long, respectively (see  and Additional file 3). Pre-tRNA-Trp is unique as it contains the C/D and C'/D' boxes that allow methylation of 2' hydroxyl of the ribose at positions 34 and 39 in the intron sequence [74-77] – see also below).
As always in Archaea and Bacteria but not in Eukarya , three different tDNAs bearing the (CAT) anticodon are present: the initiator tDNA-Met (CAT), the elongator tDNA-Met (CAT) and the tDNA-Ile (CAT). In this last case, the final identity of the mature functional tRNA-Ile depends on post-transcriptional modification of C-34 into an as yet unknown modified C-derivative (see below).
The sequences of the 46 tRNAs harboring a distinct anticodon (or tDNA when the sequence of mature tRNA is not available) are listed in Figure Figure33 from the wobble base at position 34 to nucleotide-39 (the proximal first base pair of the anticodon stem). This figure allows us to define the codon decoding strategy in a halophilic archaeon. It appears that: i) a systematic 'A-34 sparing' strategy is found, allowing the decoding of all pyrimidine ending codons (NN.U/C) by one tRNA harboring a G34. N'N' anticodon (N stating any of the 4 canonical nucleotides, N' its complementary Watson-Crick counterpart and G-34 is never post-transcriptionally modified). The presence of G-34 in tRNA-Arg (GCG) of the four codons decoding box is remarkable as the corresponding nucleoside is always an A-34 (in fact post-transcriptionally deaminated into inosine-34) in all Eukarya and most Bacteria ; ii) no 'C-34 sparing' strategy is used, that would require a U34-containing tRNA to decode a codon ending with G-3, while in the majority of Bacteria such a situation is frequent (see ). Thus in H. volcanii, the only wobbling-type case of decoding during translation of the mRNA is between a G34-containing tRNA and a codon ending with a U-3. An acetyl group is present on N4 of C-34 in many C34-containing tRNAs, and many of these tRNAs seem to be only partially modified . The presence of ac4C at the wobble position of tRNAs is unique to Archaea, with the exception of the elongator tRNA-Met (ac4C.AU) in E. coli . However, the same modification has been found at position 12 in the D-arm of some tRNA-Leu and tRNA-Ser molecules of S. cerevisiae  and in the 5S rRNA of some thermophilic archaea (see above). This modified nucleotide exhibits an exceptional conformational rigidity when embedded in an RNA molecule [80,81]. Its presence in the wobble position probably allows better binding of the tRNA to the cognate codon, possibly helps the tRNA to discriminate against codons ending with A  and to aid in phase maintenance during translation; iii) the rare isoleucine AUA codon is translated by a minor tRNA-Ile, like in all bacteria. It harbors a unique type of modified cytidine able to discriminate against the Met-AUG codon. In E. coli, (and all Bacteria and eukaryotic mitochondria), this C-34 residue is always modified into lysidine (k2C, , reviewed in ); while in Archaea, the chemical structure of the modified cytosine-34 remains to be identified ( see also below); iv) due to lack of sequence information about many of the mature U34-containing tRNAs the identification of the chemical nature of the modified U (indicated as U* in the original works of Gupta and '?U' in Figure Figure3)3) will require the discovery of potential U-34 modifying enzymes in the genome of H. volcanii (see below) or additional analytical experiments; v) without exception, three isoacceptor tRNAs are always used to decode four synonymous codons in the four codons decoding boxes and two isoacceptor tRNAs for decoding the two purine-ending codons (NN.G/A) in the split codon boxes. Thus altogether 45 elongator tRNA and one additional initiator tRNA-Met are required to decode the 61 sense codons in H. volcanii. From the early work of Bayley and Griffiths , it is known that accuracy of translation of synthetic homopolymers by extracts of the extreme halophilic bacterium H. cutirubrum, and probably all halophiles, requires the presence of very high salt concentration (up to 4 M).
Biochemical analysis using as substrate T7-transcripts of tRNA genes lacking all the modified nucleosides, allows enzymatic activities for producing pseudouridine and several base-methylated derivatives in tRNA, such as m1A-57, m1I-57, Cm-56, m1Ψ-54, m5C-49 and m22G-26 to be demonstrated in cell-free extracts of H. volcanii , but none of the corresponding genes were identified. Only recently were the genes coding for the multiprotein complex that use guide RNA to methylate the 2'-hydroxyl of cytosine-34 and uridine-39 in H. volcanii tRNA-Trp characterized [74,75]. In other Archaeal species (mainly in M. jannaschii and P. furiosus or P. abyssi), genes coding for several tRNA modification enzymes have been not only identified, but also experimentally validated. These were used to easily predict the H. volcanii orthologs with good confidence (Table (Table1).1). These include the enzymes that introduce the m2G/m22G-10, m1G-37, and m1A-57 modifications (references are given in Table Table1).1). A protein homologous to the key enzyme transglycosyltransferase (TGT) responsible for the insertion of the G+ precursor preQ0 in M. jannaschii tRNA  is also found in H. volcanii. The genes involved the synthesis of preQ0 and in the conversion of preQ0 to G+ after its insertion in tRNA, are not known in any Archaeal organisms to date. They are currently being identified in our laboratory at the University of Florida in Gainesville and will be described elsewhere. Another set of tRNA modification enzymes that introduce the Ψ-13, m2G/m22G-26 and t6A-37 modifications respectively can be predicted by homology with yeast and/or E. coli experimentally validated orthologs (Table (Table1).1). For the 12 remaining modifications, the prediction process is less straightforward because the homology scores with the experimentally validated yeast or bacterial homologs are too low, paralog families complicate the analysis or the corresponding gene has not been identified in any species. These are discussed separately below.
In H. volcanii tRNA, 2'-O-methylation of ribose occurs in four positions, 32, 34, 39 and 56 (Figure (Figure1A).1A). As stated above, Cm-34 and Um-39 in the anticodon branch of tRNA-Trp are formed by the guide RNA machinery that includes the Fibrillarin enzyme (aFib) and accessory proteins Nop56/58 and L7Ae [88-90], all encoded in the genome of H. volcanii (Table (Table2).2). The RNA antisense bearing the C/D and C'/D' boxes is part of the pre-tRNA sequence and includes the long intron of 103 nt, a situation that exists also in pre-tRNA-Trp from at least 29 archaea (see Additional file 4) [76,77]. The mechanism by which the 'intronic' antisense sequence acts in vivo in cis to 'self' induce the 2'-O-methylations of C-34 and U-39 in pre-tRNA-Trp, or in trans by acting on an other molecule of pre-tRNA-Trp, is still an open question. However in vitro experiments favor a trans-acting box C/D snRNA guided mechanism . In the case of intron-containing tRNA-Met, Cm-34 is also guided by a sRNA (see Additional file 5) but here the C/D antisense RNA is not intronic but exonic as described for the C/D box sRNA sR49, which was predicted to guide the modification of Cm-34 in the tRNA-Met of Pyrococcus . We identified 18 sRNA candidates to guide the modification of Cm-34. We found a candidate in the genome of H. walsbyi for which no tRNA-Met containing an intron could be identified in the genomic sequence available at NCBI. Thanks to the target region, we also identified the tRNA-Met containing the intron. Our analysis reveals that sRNA guiding formation of 2'-O-methyl ribose at position 34 and 39 in pre-tRNA-Trp is always intronic, while the formation of the same Cm-34 in pre- tRNA-Met is always exonic. In both cases, part of the intron sequence is involved in base pairing with sRNA, thus 2'-O-methylation at ribose in position 34 and 39 in pre-tRNA-Trp and in position 34 in pre-tRNA-Met have to occur before intron splicing .
Remarkably, Halobacteria show more degenerated C, C', D and D' boxes and a longer region between D' and C' boxes (19 to 21 pb) than other orders (4 to 10 pb). In contrast, the insertion of Cm-32 found in the anticodon loop of four tRNAs (two specific for Lys, one for Tyr and one for Trp) and of Cm-56 found in all H. volcanii tRNAs (with no exceptions; see Additional file 1), is almost certainly catalyzed by non RNA guided enzymes. Indeed, a solid homolog of the TrmH (YhfQ) protein that has been found to catalyze the formation of Xm-32 in E. coli , is present in the genome of H. volcanii (Table (Table1).1). It is the only member of the SpoU family [92,93] found in this organism. For Cm-56 in the Ψ-loop, a strong homolog of the P. abyssi protein found to catalyze this reaction in vitro ( and reviewed in ), can be identified in the genome of H. volcanii (Table (Table11).
Apart from Ψ-13 which is most certainly modified by the TruD ortholog (HVO_0658, belonging to COG0585, see table table1),1), seven other Ψ residues? are present in H. volcanii tRNA at positions 22, 28, 38, 39, 52, 54 and 55 (Figure (Figure1).1). Ψ-55 is a universal modification inserted in yeast by Pus4p  and in E. coli by TruB , both belonging to the same COG0103. The only homolog of these two proteins that can be identified in the H. volcanii genome is Cbf5p, which is the catalytic subunit of the guide machinery (see below). However, recent work from different laboratories have shown that in vitro, Cbf5p can modify U-54 in tRNA, as well as in rRNA, in a guide-independent fashion, the enzymatic reaction being stimulated by the presence of Nop10p [98-100]. Psu10p from P. furiosus, that is not part of the TruB/Cbf5p family of proteins (COG0103) but is instead a member of the COG1258 family (Table (Table1),1), can also introduce the Ψ-55 modification in archaeal tRNAs in vitro. This observation has been validated by complementation experiment using an E. coli truB mutant . It is however still not clear which of the two enzymes (Cbf5p and/or Pus10p) is responsible for the formation of Ψ-55 (as well as of Psi-54) in Archaeal tRNAs in vivo. As discussed in , the possibility exists that each of the two Ψ-55 forming enzymatic systems act on distinct sets of tRNAs. It is worthy of mention that no Psu10p homolog is found in N. equitans, whereas a genes coding for Cbf5p and Nop10 homologs are detected (see "Archaeal rRNA modification" subsystem in the SEED database for sequences). Unfortunately, no evidence for the presence or absence of Ψ-55 in any of the tRNAs, or of Ψ in rRNA is available for this organism.
Other quasi universal Ψ. modifications are Ψ-38/39 of the anticodon branch inserted in yeast by Pus3p  and in E. coli by TruA , both members of the COG0101 family. Only one protein of this family could be identified in H. volcanii (Table (Table1).1). Its homology with both the E. coli and the yeast GOG0101 members is quite low but multiple sequence alignments using clustalw  confirmed that the critical TruA specific active site consensus sequence (XXXRTD)  is conserved in the Haloferax protein. No homologs of yeast Pusp 1- 9  or E. coli TruABCD (reviewed in ) could be identified in the H. volcanii genome, leaving Ψ 's at positions 22, 28 and 52 with no corresponding candidate pseudouridine synthase. The corresponding uridines are located in helical regions (at least in the fully mature tRNA), thus perhaps relatively inaccessible to the enzyme, and these are isomerized to Ψ only in one tRNA species, whereas the other Ψ modifications, at least those in positions 38, 54 and 55 are located in loops and found in several tRNAs (see Additional file 1). This led us to suppose they may be introduced at a very early stage of the tRNA biosynthesis, possibly still during transcription and when the cloverleaf structure of the nascent pre-tRNA is not yet formed, potentially by the RNA-guided machinery including Cbf5p, Nop10p, Gar1p and L7Ae homologs (Table (Table2).2). However no H/ACA snRNA could be identified in H. volcanii. This could be due to a high divergence of H/ACA RNA structures in H. volcanii or, as discussed above in the case of Ψ-55 formation, to a guide RNA independent pseudouridylation by the Cbf5p/Nop10p dependent machinery acting during transcription on an as yet unfinished tRNA molecule composed of stems and loops. One cannot however rule out that an as yet unidentified pseudouridine synthase family is present in this organism.
Four positions 39, 40, 48, 49 are modified to m5C in H. volcanii. In yeast, Trm4p is not site-specific and introduces this modification at several positions in tRNA molecules . Members of this huge family of proteins (COG0144) are however difficult to annotate by sequence alone as some also modify rRNA [107-109]. Recently one of the five COG0144 members from P. abyssi (PAB1947) was found to catalyze in vitro the formation of m5C at several positions in tRNA, including positions 48 and 49 . H. volcanii has just one member of this family (HVO_1594) that is highly similar to PAB1947, and ribosomal RNA of this organism does not contain any m5C (see Figures 2A–D). Hence, it is highly probable that HVO_1594 is the only RNA:m5C methyltransferase that modifies the four cytosines found in the sequenced tRNAs of H. volcanii (Figure (Figure1).1). The presence of additional m5C residues at positions 50–52 in some tRNAs of H. cutirubrum probably also result from the action in this organism of a unique multi-site specific tRNA:m5C methyltransferase.
In H. volcanii, the only inosine (deaminated adenosine, in the form of m1I) residue is found at position 57 of the majority of tRNAs (Figure (Figure1,1, see also Additional file 1). Enzymatic formation of the doubly modified m1I occurs in two strictly sequential steps. The first step is the methylation of A-57 catalyzed by the tRNA:m1A methyltransferase of the P. abyssi TrmI family (COG2519, ) (Table (Table1).1). Then deamination of m1A-57 occurs by a tRNA:m1A-specific deaminase [86,112], that is different from other tRNA deaminases such as Tad1p and Tad2p/Tad3p catalyzing the formation of inosine from adenosine in position 37 and 34 respectively in S. cerevisiae tRNAs [113-115] or TadA catalyzing the site-specific formation of inosine-34 exclusively in tRNA-Arg (anticodon AGC), as we could not identify any homologs of these families in the H. volcanii genome. We searched for protein families specifically conserved in all Archaea but absent in Eukarya, with RNA binding domains. One candidate is the COG1491 family. It is annotated as an RNA-binding protein as the structure of the A. fulgidus family member (PDB: 2I5H) showed that the N-terminal domain is similar to many nucleic acid binding protein with the presence of a characteristic S1 domain . Analysis of a clustalw sequence alignment reveals a highly conserved histidine residue in a motif [R,K] [L,M]H [A,S,T,Q,M]L [E,Q,N] (Figure (Figure4A)4A) that is similar to the adenosine deaminase "motif I" found in all adenosine deaminases .
The methylation at position 54 is a hallmark of Archaea, except in Thermococcales where m5U54 is found . COG1901 proteins that are part of SPOUT superfamily have been predicted as candidates for this missing methylase  and genes of this family do indeed cluster with Psu10p in several genomes (Figure (Figure4A).4A). However, it is present in organisms that are expected to have m5U and not m1Ψ at this position such as the Pyrococci . Experimental validation is required to ascertain the function of this putative Ψ-dependent methyltransferase (work in progress).
As a rule U34-harboring tRNAs belonging to the split codon boxes corresponding to Leu (UAA), Gln (UAG), Lys (UUU), Glu (UUC) and Arg (UCU) need to discriminate for the NN-Purine codons and not miscode for the NN-pyrimidine codons. Only certain types of modified U-34 can perform this task (reviewed in: ). In contrast, tRNAs bearing unmodified U-34 are able to decode codons ending with purines and pyrimidines, such as those found in the four codon decoding boxes as in Mycoplasma for example . As expected a modified U! (of which the chemical identity remains to be determined) was identified in naturally occurring tRNAs of H. volcanii specific for Lys (U!UU) and Glu (U!UC) [28,29]. However for tRNA-Leu (?UAA) and tRNA-Arg (?UCU), the identity of U-34 is yet to be determined as these tRNAs remain to be sequenced (Figure (Figure33 and Additional file 1). Curiously, U-34 of H. volcanii tRNA-Leu (UAA) was reported not to be modified, while the U-34 residue in tRNA-Leu (?UAG), tRNA-Arg (?UCG) and tRNA-Gly (?UCC) belonging to the four codons boxes appears to be (Figure (Figure3).3). Unexpectedly, U-34 in tRNA-Pro (UGG) and tRNA-Ala (UGC) is not modified [28,29], while the corresponding four codons decoding boxes contains two other tRNAs (one with G-34 and the other one with C-34) able to decode the other codons, except the one ending with A (see Figure Figure3).3). Thus the pattern of modified/unmodified U-34 in H. volcanii tRNAs is non-canonical and the exact chemical nature of the U! in the different tRNAs of H. volcanii remains an enigma [as well as in the few other archaeal tRNAs sequenced so far .
In the case of tRNAs specific for Gln, Lys and Glu in H. volcanii, a thiolated U-34 derivative should exist, as for their bacterial and eukaryal counterparts (for examples see [121,122]). Indeed, in the genome of H. volcanii, a gene homolog to eukaryal Tuc1 belonging to COG0037 is found. In S. cerevisiae, this Tuc1 protein has recently been shown to be involved in the formation of s2U-34 in yeast cytoplasmic tRNAs . Also, clustering of Tuc1 with IscS and IscU, the two proteins required for donating the thio compound (Figure (Figure4B)4B) strengthens the prediction that this family of proteins do participate in the formation of thiolated compounds. However, analysis of the H. volcanii genome suggests that ?U34 in tRNAs, as in many other archaeal tRNAs, is more complex than just a s2U. Indeed, homologs of the yeast Trm9p methylase and of the radical SAM enzyme Elp3 are also found in H. volcanii genome (Table (Table1).1). Trm9p is the yeast mcm5U/mcm5s2U tRNA carboxyl methyltransferases  and Elp3 is in yeast part of the elongation complex comprised of 6 proteins Elp1-6 that have all been shown to have a role in the formation of mcm5(s2)U . In vivo the pleiotropic effect of mutations in the yeast Elp genes appear to be due to the absence of the modified base in tRNA . However, out of the six eukaryal Elp proteins only homologs of Elp3 can be found in Archaea. This protein is part of the radical SAM family . In S. solfataricus, the Elp3 and Trm9p encoding genes are also clustered (Figure (Figure4B).4B). Tuc1 is present in all sequenced genomes of Archaea, Elp3 is lacking only in N. equitans, and Trm9p homologs are found only in a limited subset of Archaea (see "Archaeal tRNA modification" subsystem referenced in the methods section). Taken together, the data suggest that the type of U-34 modification in H. volcanii tRNAs belonging to two codon split decoding boxes is similar but not identical to the mcm5s2U derivative found in eukaryal tRNAs.
Gupta proposed that mo5U-34 may exist in H. volcanii tRNAs. In bacteria it has been shown that chorismic acid is a precursor to mo5U-34 formation through the ho5U intermediate, with the product of cmoB catalyzing the conversion of ho5U-34 to mo5U-34 . CmoB protein is part of the methyltransferases type 11 family http://www.ebi.ac.uk/interpro/IEntry?ac=IPR013216 that is difficult to annotate because it is so widespread. A distant homolog of CmoB was indeed identified (Table (Table2)2) and could be the potential mo5U synthase but this prediction is not very robust and absolutely requires experimental validation. The genes that are responsible for the formation of ho5U have not been identified in any organism.
Is found only at position 34 in many H. volcanii tRNAs. In certain Archaea and Eukaryotes it has also been detected in ribosomal RNA (see above). The only known enzyme involved in ac4C formation is yeast Tan1 (YGL232W ) that does not have any homolog in H. volcanii. However it was predicted that Tan1 binds tRNA and carries the recognition determinants but must function in complex with yet unidentified acetylation enzymes , reviewed in . One enzyme family COG1444, that contains an ATPase domain fused to an acetyltransferase domain was identified as a potential candidate. Genes of this family cluster with translation genes (Figure (Figure4A).4A). In yeast, the homolog (YNL132W) is essential , whereas the E. coli ypfI homolog is not , and has recently been shown to be responsible for ac4C formation in tRNA initiator in E. coli .
Finally, like Bacteria, all Archaea but N. equitans have a minor tRNA-Ile (CUA) [71,84]. This requires the modification of C-34 to a C*. Otherwise this tRNA-Ile would be charged erroneously with methionine , reviewed in . In all bacteria except in Mycoplasma mobile , a lysinyl group is inserted by the ATP-dependent TilS family of enzymes , but in Archaea the structure of C-34 modification is still not known, and no tilS gene has been found in H. volcanii. Potential candidates for a gene coding for an enzyme catalyzing the selective modification of C-34 in tRNA-Ile (CAU) should be found in all Archaea but N. equitans and should also be absent in the genomes of E. coli and S. cerevisiae. Using the OrthoMCL phylogenetic distribution search tool , we identified 10 protein families that conform to the above criteria. We favor two candidates in this list for missing C*-synthase genes: i) the nucleic acid binding protein-OB fold family (COG1571) that contains a potential RNA binding domain, and ii) the COG2047 family, annotated as an ATP-grasp superfamily. Both these families cluster with tRNA modification genes in several Archaea (Figure (Figure4C)4C) and follow the expected phylogenetic distribution (see "Archaeal tRNA modification" subsystem in SEED).
Very few Archaeal enzymes involved in the modification of ribosomal RNA have been experimentally characterized to date with the exception of the guide rRNA methylation and peudouridylation machinery enzymes (see for examples: [100,135,136]). However thus far little investigation has been performed in the case of halophilic organisms, including H. volcanii. The analysis below (Table (Table2,2, and references therein) is hence mainly based on comparative genomics predictions that await experimental validations.
Among modification enzymes acting on rRNA that can be easily identified in H. volcanii by sequence homology is the archaeal member of the RsmA/Dim1p family. This enzyme introduces four methyl groups in the conserved tandem adenosine in hairpin 45 of the 16S RNA to form m62A m62A (positions 1450 and 1451; 1518 and 1519, E. coli numbering). The function of the M. jannaschii RsmA ortholog was experimentally confirmed . A strong homolog of this enzyme is found in H. volcanii (Table (Table22).
Likewise, a homolog of RrmJ (or RlmE belonging to COG 1189) that catalyzes the formation of the quasi universally conserved Um at position 2552 of the hairpin 92 of bacterial 23S RNA  in E. coli (Um-2587 in H. marismortui) is also found in H. volcanii (Figure (Figure2C2C and Table Table2).2). In S. cerevisiae a site-specific Mrm2 enzyme introduces the same modification in the mitochondrial 21S (Um-2791)  but a guide RNA (SnR52) machinery is responsible for the equivalent cytoplasmic yeast 28S rRNA methylation (Um-2921) . In this later case, Spb1p (of COG 1189 as Mrm2 and RlmE) can also catalyze the formation of Um-2921 (2552 E. coli numbering) even if its normal function is to catalyze the 2'-O-methylation of adjacent G-2922 . Since the snRNP-dependent formation of Um-2921 occurs within the nucleus at an early step of the rRNA maturation process, and the action of Spb1 enzyme proceed later within the cytoplasm, only if U-2921 has not previously been fully modified in the nucleus, can Spb1p then complete the reaction . In the case of U-2587 (2552) and/or G-2588 (2553) methylation in 23S RNA of H. volcanii, searches for potential guide RNA have been unsuccessful by using both pattern matching approaches and the dedicated SnoScan software, while in P. abyssi, a sR25 C/D box sRNA was predicted to guide the methylation of U-2669 (U-2552 in E. coli) . Failure to detect the guide in the halophiles might be due to a divergent structure of the snRNAs in these organisms or could reflect the real absence of such guide RNAs for these particular methylation targets. The possibility exists that the halophilic RlmE homolog, identified above, is a multi-site specific enzyme and catalyzes both the formation of Um-2587 and Gm-2588. A precedent for such a situation is found in the enzymatic formation of m1A at both positions 57 and 58 of tRNA by the Pyrococcale TrmI enzyme , while the bacterial and eukaryal homologs (TrmI and Trm6p respectively) are strictly site-specific and methylate only A-58 in tRNAs [142,143]. Another possibility is that Halophiles have multiple paralog copies (from 3–6 copies, see 'Archaeal RNA subsystem") of COG3269 family that contain the RNA binding TRAM (TRM2 And MiaB, domain)  and one of these enzymes could be responsible for the formation of Um-2921.
We found at least one C/D box sRNA candidate to guide the 2'O-methylation of the ribose at position G-1950 (1909 in E. coli). Homologous sequences were found in 25 archaeal genomes (Additional file 7). Moreover, our results suggest that in Pyrococcus, the sR41 orphan C/D box sRNA http://lowelab.ucsc.edu/snoRNAdb/Archaea/Pyro-annote.html could modify the equivalent of G-1950 position.
No homologs of the multiple known E. coli pseudouridine synthases that modify rRNA (for review see ) could be identified in H. volcanii. As demonstrated for rRNA of Eukarya and some Archaea, such as S. solfataricus, A. fulgidus or P. furiosus and P. abyssi, Ψ residues could also be introduced by the guide RNA machinery and indeed, all the enzymes needed are presents in the H. volcanii genome (Table (Table2).2). In Eukaryotes, the equivalent of Ψ-1956 and Ψ-1958 are modified by the same H/ACA sRNA, respectively U19 in Human and snR191 in Yeast. In Archaea the equivalent of Ψ 1956 was proposed to be modified in Pyrococcales and A. Fulgidus respectively by Pf7 in P. furiosus and Afu4 in A. Fulgidus . Recently a combination of in silico and experimental work identified seven H/ACA involved in pseudouridylation of rRNA in P. abyssi while a total of 17 Ψ residues were detected . Some of these sRNA modify several positions in rRNA but clearly not all the 17 Ψ residues are accounted for, and for certain positions (such as Ψ-2603 of P. abyssi rRNA) the modification can be introduced in vitro by the Cbf5p/Nop10p dependent complex in the absence of any guide RNA . Indeed Ψ-2016 (Ψ-1956) was introduce by the Pf7 homolog (Pab40) and the Afu-4 H/ACA sRNA in vitro but the modification equivalent to Ψ-1958 was not . Pf7 contains three hairpin motifs, namely Pf7-stem-I, Pf7-stem-II and Pf7-stem-III, each one able to guide a modification. The in silico approach used in the present analysis allowed to identify two H/ACA sRNA hairpins, respectively HP1 and HP2, candidate to the modifications of Ψ-1956, Ψ-1958 and Ψ-2621. We did not find the homolog of Pf7-stemI hairpin, which is consistent with the absence of a fourth modification in H. marismortuii. HP2 is clearly the homolog of Pf7-stemIII and HP1 appear to be the homolog of Pf7-stemII. Remarkably in H. volcanii and other halobacteria, both hairpins are conserved but are separated by a long spacer whereas they are adjacent in thermococcales. HP2 would be able to guide Ψ 1956 and Ψ 1958 by forming alternative structures around the position to modify (see Additional file 8B) while the HP1 could target Ψ-2621 (see additional file 8A). Remarkably in P. abyssi, this last modification was not found experimentally  whereas Pab40 could adopt an alternative structure able to target this position (see Additional file 8A). Finally we did not find the homolog of Pf7-I and Pf7-II in the Crenarchaeote Ignococcus hospitalis. Certainly, the modification targeted is not present.
This methylated adenosine is located in hairpin 25.1 (position 628) of Domain II (Figure (Figure2C).2C). A weak homologs of RlmA that introduces a m1G in the E. coli large subunit in position 745  can be identified (Table (Table2)2) and is a possible candidate for the formation of m1A, even if it is a different purine base. Indeed, during evolution, an enzyme able to methylated N1 in guanosine might have adapted to methylation of N1 in adenosine, exactly as an ancient C5-methylated enzyme has derived to become a C5-methyltransferase of uridine by simply changing few aminoacids in the active site in order to accommodate U instead of C [147,148].
This N3 methylated uridine (position 2619 in 23S RNA, 2584 in E. coli) is located between hairpins 92 and 93 (Figure (Figure2C).2C). A good candidate for the missing m3U inserting enzyme in H. volcanii is the protein belonging to COG2106 (Table (Table2).2). Indeed, analyzing the SPOUT family enzymes, Bujnicki and coll.  found that COG1385, exemplified by E. coli RsmE that introduces the m3U modification in 16S RNA , has a complementary phylogenetic distribution to the COG 2106 family found in Archaea and eukaryotes. Moreover, genes encoding COG2106 proteins are inserted in operons encoding for ribosomal proteins in phylogenetically diverse Archaea such as the Pyrococci, Archeoglobus fulgidus and H. salinarium (data not shown).
This methylated exocyclic NH2 of adenosine is located at position 1432 (1500 E. coli numbering) of 16S RNA in helix 44 in the 3' minor domain (Figure (Figure2A).2A). Compilation of modification data in the SSU RNA modification database  shows that the m6A modification found in the H. volcanii 16S RNA can also be found at the same position in S. solfataricus and in three eukaryotes Homo sapiens, Xenopus laevis and Rattus norvegicus. By searching for genes that are present in these four organisms (and that are generally annotated as methylansferase) we identified the COG2263 family. Annotated as RNA methyltransferase or N6-DNA-methylase, members of this family are present in most archaea and many eukaryotes. The structure of the COG2263 member PH1948 was determined in complex with S-AdoMet  and revealed that this protein was a structural homolog of ErmC' (pdb :1QAN) that confers resistance to macrolides by introducing an N6-methylation at adenine 2058 (as E. coli numbering) of 23S rRNA . We propose that the H. volcanii COG2263 member (Table (Table2)2) is also involved in m6A formation but in the 16S RNA, not the 23S RNA.
This uridine bearing a 3-amino-3-carboxypropyl group on N3 of uridine is located at position 910 (966 E. coli numbering) of hairpin 31 in 3'major domain of 16S RNA (Figure (Figure2A).2A). It is modified to m1acp3Ψ. in all eukaryotes analyzed so far but is never present in small RNA subunits of bacteria that always have a non modified G (or m2G) in this position . Using the phylogenetic pattern tool of the OrthoMCL database  we searched for genes that are conserved in mammals, S. cerevisiae and all Archaea but absent in all bacteria. A large collection of genes follow this pattern (89 altogether), most of them are ribosomal proteins and other translation related genes. One candidate stood out as a potential acp3U inserting enzyme, the COG2016 family. Proteins of this family are found in all sequenced Archaea and eukaryotes and contain a C-terminal PUA domain (Pseudo Uridine synthase and Archaeosine transglycosylase) that is often involved in RNA binding . The yeast member of the COG2016 family, YER007C-A, has been shown to associate with ribosomes and a null mutant has clear translation defects .
Beside the putative genes identified above, a few other genes corresponding to as yet unidentified modified nucleotides need to be discovered, such as for the currently unidentified C*(N330) derivative located at position 1352 in hairpin 44 in the 3' minor domain SSU RNA (Figure (Figure2A).2A). N-330 is also found at the same position in the bacteria Thermotoga maritima . Lastly, while the possibility is meager, one or two additional modified nucleotides might still exist in rRNAs of H. volcanii. Indeed, the full lengths of the 16S and 23S (1472 nt and 2922 nt respectively) of H. volcanii or of H. marismortui have not been explored, only the most critical regions where the probability was high to discover conserved or semi-conserved nucleotides have been investigated.
The archaeaon Haloferax volcanii has the particularity of being a 'salt-loving' prokaryote that lives in the mildly hot and hypersaline environment of the Dead Sea (40- 50°C, 1.5-3M NaCl) where it was first isolated [27,154]. Life at such high salt concentrations is energetically costly. Indeed, to insure the osmotic balance between the cytosol and the high salt environment in which they thrive, halophiles have to accumulate and maintain high concentrations of solutes. These are mainly inorganic ions, such as KCl that can reach molar concentrations or Mg2+, but various organic osmotic solutes such as glycerol, trehalose and/or glycine betaine are also used [154,155]. As there are no visible compartments in the Haloferax cell , this lifestyle requires the adaptation of the entire intracellular enzymatic machinery, including RNA maturation and mRNA translation processes. Indeed, at high salt concentrations, the high molecular weight rRNA and the majority of proteins from non halophilic prokaryotes simply precipitate (reviewed in ).
Here we combine the identification of the whole set of functional tRNAs, including the presence of modified nucleosides (tRNomics), with the identification of most of the corresponding RNA modification enzymes (Modomics) in H. volcanii. This analysis allows to address: i) the peculiarities of the decoding strategies used by H. volcanii to read the 62 (61+1 initiator) sense codons of mRNAs; and ii) to emphasize the relative low number of modifications in halophilic t+rRNAs. This work is a logical continuation of a similar tRNomics analysis of fully sequenced genomes from the three kingdoms of life [71,73,158], later extended to the Modomics analysis of Mollicutes, a family of parasites that underwent a drastic reduction of their genomes during evolution . On an evolution point of view, Halobacteriales like the euryarchaeon H. volcanii, and other distantly related organisms able to grow at salt concentrations above 100 g/L (1.7 M NaCl), such as certain Methanosarcinales (Archaea), Flavobacteria, Cyanobacteria and Proteobacteria (Bacteria) or a few Flagellated, Ciliates and Fungi (Eukarya), are all located on a relatively 'recent' branches of the small subunit rRNA based phylogenetic tree (see Figure Figure11 in ). Thus, emergence of halophilic organisms likely results from an adaptive-type of cellular evolution from a non-halophilic ancestor arising independently several times during the evolution of the three domains of life.
The detailed mechanism by which mRNAs are accurately decoded without slippage by the ribosome in an extremely halophiles is largely ignored. The only published study using cell-free system from Halobacterium cutirubrum shows that incorporation of radiolabeled amino-acids into polypeptides under the direction of synthetic polyribonucleotides, follows the same decoding rules found in non halophilic organisms, but that the accuracy of amino acids incorporation was dependent on the presence of various salts at high concentrations (KCl, NaCl, NH4Cl – ). This lead to the conclusion that the codon recognition processes are only secondarily dependent on ionic interactions and that the effect of salts was probably to enable all the macromolecular components to assume their correct secondary and tertiary configuration, a conclusion that is evident nowadays.
What is clear from the present work, is that the 52 (45 elongators + 1 initiator + 6 duplicants) tRNAs found in H. volcanii that read the 62 universally used sense codons (61+1 initiator) are typical of the Archaea that have been analyzed to date with a few minor differences discussed above ( and unpublished data). What is more interesting is that H. volcanii uses only 16 different types of modified nucleotides at 18 positions in the 46 mature tRNA isoacceptors, while both E. coli and S. cerevisiae use at least 28 different types of modified nucleotides at 20 and 35 different positions respectively [4,71]. As far as the type and position of modified nucleoside in tRNAs, the archaeon H. volcanii resembles Eukarya in some ways and Eubacteria in others (Table (Table3).3). The cases where an identical modification is found at the same position in the three kingdoms are rare (indicated in bold Table Table3).3). The modifications that are archaeal specific by their chemical structure and/or their positions in tRNAs are also not numerous (underlined in Table Table3).3). Examples include G+-15, ac4C-34, m1Ψ-54, Cm-56 and m1I-57 (Fig (Fig1).1). Phylogenetic and structural analysis of the transglycosylase TGT catalyzing the incorporation of precursor of Archaeosine (G+) into tRNA, points to a common evolutionary origin with the present-day enzyme catalyzing the formation of queuosine at position 34 in many bacteria and higher eukaryotes , a typical case of divergent evolution. In contrast, the enzymatic formation of m1I at position 57 in archaeal tRNA involves a totally different set of enzymes than those needed to catalyze the formation of the same modification at position 37 in eukaryal tRNA-Ala , this time a case of convergent evolution. Our analysis has raised several questions that await experimental follow-up. Several predictions such as those for the genes involved in m1I or ac4C formation need to be validated. The nature of s2U-34 derivative that was predicted from the comparative genomic analysis but was not found in the initial tRNA sequencing work [28,29] has to be identified. We failed to find any gene coding for putative (multi) site-specific RNA pseudouridine synthase(s), nor for 'classical' box H/ACA guide RNAs with that catalyze the formation of pseudouridines at positions 22, 28 and 52. As point out above, it might well be that these Ψ 's are formed at very early step of the tRNA maturation (possibly during transcription) by the non RNA guide Cbf5p/Nop10P/Gar1 complex.
The type and location of modified nucleotides found in 16S rRNA of H. volcanii and in the 23S rRNA of the closely phylogenetically related H. marismortui were compared to those found in E. coli and S. cerevisiae (Table (Table4).4). There again the surprising feature in halophiles is the paucity of rRNA modifications with only 4 different modified nucleotides in 5 positions in the 16S rRNA (out of 1472 nt) and 6 in 8 positions in the 23S rRNA (out of 2922 nt). In E. coli there are 16 different types of modified nucleotides within 35 positions of the 16+23S rRNAs and in S. cerevisiae 18+23+5S rRNAs contain at least 8 different modified nucleotides located at more than 100 positions [12,13]. Only a few of these modifications are found in all the three biological domains in rRNA analyzed to date from (in bold in Table Table4).4). Without exception, they are located in critical functional domains of the RNA molecules, e.g. in the decoding center of the SSU rRNA (Figure (Figure2B)2B) and near the peptidyl transferase center of LSU rRNA (Fig (Fig2D)2D) manifesting their functional importance in various aspects of the dynamic process or mRNA translation (as discussed above in the data section). Their importance is further supported by the fact that the genes coding for the corresponding enzymes, as well as the sRNA guided modification machinery allowing the formation of these conserved t+rRNA modifications, are also remarkably conserved among the different domains of life, except for Gm-2588, m1A-628 and acp3U-910 that are present in eukaryotic rRNA and absent in bacterial rRNA (see Table Table44).
A clear positive correlation has been observed between the total number of ribose methylation sites, the eventual corresponding number of methylation guide sRNAs and the optimal temperature at which an organism is growing, suggesting an important role of this type of modification in RNA stabilization (reviewed in ). Clearly as the number of 2'- O-methyl ribose is exceptionally low in rRNA of halophiles, the rules guiding the faithful maturation of rRNA molecule, as well as the stabilization of their quaternary structure within the ribosome, might differ from other Archaea (psychrophiles, mesophiles and hyperthermophiles). Of note also is the absence of polyamines in extreme halophiles, as the slight amount of polyamines that can be detected actually originate from the culture medium (Oshima Tairo, personal communication). Polyamines, like Mg2+ and other ions stabilize nucleic acids (reviewed in ) and also facilitate protein synthesis [162,163]. The 3D structure of the large 50S subunit of H. marismortui has been solved to 2.4 Angstrom resolution . Analysis of the structure reveals a great number of monovalent and divalent ions as well as water molecules that are critical for the formation and stabilization of that rRNA structure. Hence, we propose that the presence of high concentration of salts (mono- and divalent) in the cytosol of H. volcanii has allowed the elimination of numerous rRNA and tRNA modifications as well as of polyamines biosynthesis, whose 'global' functions are to allow faithful maturation of pre-t+rRNAs and/or to stabilize the mature t+rRNAs and their association with other proteins (e.g. quaternary structure in the case of ribosome). If the functional replacement of many RNA modification by salts had indeed occurred, then modified nucleotides remaining in t+rRNA of halophilic organism must serve purposes other than stabilization of RNA architecture, such as decoding, accuracy of translation or other functions that cannot be functionally replaced by the electrostatic interactions provided by the surrounding salts. This hypothesis is corroborated with the fact that most, if not all of the modified nucleoside found in H. volcanii/H. marismortui rRNA are among the most evolutionary conserved modified nucleosides along organisms of the three biological domains (Table (Table44 and discussed above in data section). They are also among those we have pointed out as being the most refractory to reductive evolution in Mycoplasma . This would suggest that the modifications remaining in H. volcanii tRNA are also critical for functions that cannot be replaced by salt and we are currently mutating all the corresponding genes to address the functional of these modifications in vivo.
This tRNomics and Modomics analysis of H. volcanii reinforces the necessity to integrate the knowledge of both t+rRNA sequences and modifications in order to understand the decoding properties of a given organisms. For most organisms this information can be derived only from comparative genomic analysis as sequence information of mature RNAs are lacking. However, to predict the presence or absence of modified nucleotides just from the analysis of the encoded genes is still quite dangerous and requires the type of systematic analysis performed here as a foundation in order to analyze other Archaeal genomes and understand of the function of RNA modification in Archaeal translation and its evolutionary importance.
The complete genome of H. volcanii DS2 (April 2007 (haloVolc1) assembly) was obtained using the UCSC Archaeal Genome Browser http://archaea.ucsc.edu/cgi-bin/hgTracks?hgsid=84889&chromInfoPage=. The full set of tRNA genes (tDNAs) was first identified by searching the nucleotide sequence corresponding to all the archaeal-type conserved tRNA cloverleaf structures (for details see ). Verification with tRNAscan-SE disclosed two more genes displaying anomalously low Cove score values. Close examination of the sequences revealed the presence of an anomalous G at position 58 (instead of the universal A58) in elongator tDNA-Met (CAT) (Cove score: 54.0); this remarkable sequence exception is confirmed by the tRNA sequencing . The other exception is a G at position 8 (instead of the universal pyrimidine T8/C8) in tDNA-Thr (TGT) (Cove score: 44.6). This tRNA however was not sequenced, but one can observed in this tRNA that base 14, which is usually paired with base 8 (Watson-Crick A-T pair), is also exceptionally G instead of A suggesting a Hoogsteen G8-G14 base pair. The complete list of the 52 tDNAs of H. volcanii tabulated in a linear, as well as in a cloverleaf representations is given in Additional file 2. These 52 genes correspond to 46 different tRNAs (different anticodons) since 6 genes are present in two copies (the two copies of tDNA-Gly (GCC) slightly differ in the amino acid stem only). Only three genes bear introns: tRNA-Trp (CCA), tRNA-Gln (TTG) and tRNA-Met (CAT) – for details see text below.
The linear sequences of the 41 naturally occurring mature tRNAs of H. volcanii, as sequenced by Gupta [28,29] are listed in Additional file 1 (including the two variants of tRNA-Gly (GCC)). From comparison with the other fully sequenced tRNAs, the presence of many modified nucleotides in these tDNAs can however easily be inferred. Beside C* in the minor tRNA-Ile (C*AU) and U* in several U34-containing tRNAs, the chemical structures of all modified nucleotides are known. The probability of unknown modified nucleotides remaining in one of the six unsequenced tRNAs of H. volcanii is small. Analysis of the 12 additional sequences from other mesophilic halophiles, reveals the presence of m5C at positions 50 of tRNA-Thr (GGU), position 51 of tRNA-Val (CAC+GAC) and position 52 of tRNA-Arg (GCG), as well as of m1G at position 9 in tRNA-Val (CAC) and probably also in tRNA-His (GUG) and tRNA-Gln (CUG) in H. cutirubrum. These last modifications are not found in any of the sequenced tRNAs of H. volcanii.
Most of the comparative genomic analysis to identify putative RNA modification genes was performed in the integrative SEED database at http://anno-3.nmpdr.org/anno/FIG/subsys.cgi. Results are made available in the "Archaeal tRNA modification" and the "Archaeal rRNA modification subsystem" on the publicly available server http://theseed.uchicago.edu/FIG/index.cgi. Microbes online  was also used for clustering analysis and mining co-expression data. The phylogenetic pattern searches were performed using the signature search tool on the NMPDR server , the COG phylogenetic pattern search at NCBI (, http://www.ncbi.nlm.nih.gov/COG/old/phylox.html), the ortholog table tool at the MBGD database , the phylogenetic search query forms of OrthoMCL  or of the Integrated Microbial Genome (IMG) database . Genome specific BLAST searches  were also performed at NCBI http://www.ncbi.nlm.nih.gov/sutils/genom_table.cgi. Phylogenetic distribution of any given gene family was obtained through the IMG database . Information on the presence of a given modification in RNA was essentially extracted from the RNA modification database , the tRNA database , the small rRNA modification database  and the 3D ribosomal modification map database  (for corresponding http, see above in Introduction section). Databases for rRNA and snoRNA that are involved in RNA-guided modifications are located at. http://people.biochem.umass.edu/fournierlab/snornadb/main.php and http://lowelab.ucsc.edu/snoRNAdb/. Additional information was extracted from specific articles cited throughout the text.
In archaea, C/D box sRNA contains four short conserved sequence motifs called the C box (RUGAUGA), D' box (CUGA), C' box (UGAUGA) and D box (CUGA), and one or two antisense elements. Each antisense element is 8-12nt long, is located immediately upstream of box D or D', and shows conserved complementarities spanning the site of modification. Each antisense element is the determinant of the site-specificity of the methylation site which is always the nucleotide of the target sequence paired to the fifth sRNA nucleotide upstream from the D(D') box (See Additional file 6, and ). Archaeal H/ACA sRNA are composed of one, two, or three stem-loop structure [145,175,176]. Each of these stem-loop structures can be described by two stems separated by an internal loop, a K-turn motif, and a conserved ANA (generally ACA) motif at the 3' end. The internal loop is composed of two single stranded regions which are complementary to a target region around the modified nucleotide. The target region itself encompasses two regions able to form the duplex by forming RNA-RNA interactions with the internal loop. These two regions are separated by UN, U being the uridine which will be converted into a pseudouridine (see Additional file 6
The C/D box and H/ACA box sRNAs responsible for a given set of modifications were searched by using PatScan and Darn! http://carlit.toulouse.inra.fr/Darn/. In principle, the knowledge of presence of 2'-O-methyl derivatives as well as of Ψ in RNA is of great help to identify potential sRNAs. However, as Halobacteria may use non canonical type of sRNAs, the task is not simple. Despite this, for C/D box sRNA, we used a signature describing half of a C/D box sRNA containing a C (C') box motif, a short spacer, the antisense region and a D (D') box motif. The antisense region was modeled as a motif complementary to the sequence spanning four nucleotides before and after the target position. Each candidate was then extended either at its 5' or 3' end to obtain a complete sRNA sequence. In some cases, it was necessary to degenerate the signature (including one or two errors in C, C', D, D' and antisense regions) to obtain a good sRNA candidate. The same strategy was used for H/ACA sRNA. For H/ACA sRNA, the initial signature contained the characteristics of a stem-loop structure with the stem down to the pocket, the two 3–5 nt antisense elements surrounding the residue to modify, a K-turn (K-loop) motif and an ANA motif situated 13–16 nucleotides from the Ψ residue. For each candidate found, we searched for homologous sequences by combining pattern matching approaches and similariy searches (using NCBI-Blast against complete genomes of archaea at http://www.ncbi.nlm.nih.gov/genomes/lproks.cgi. Only candidates found in inter-coding sequences and showing strong homology evidence were kept as good candidates.
Comparison with known sRNA was performed by using data from the literature and available databases http://people.biochem.umass.edu/fournierlab/snornadb/main.php, http://www-snorna.biotoul.fr/,. http://bioinf.scri.sari.ac.uk/cgi-bin/plant_snorna/home and http://lowelab.ucsc.edu/snoRNAdb/).
All tRNA genes and mature tRNA (with their anticodon in brackets) are designated as this example: tDNA-Ile (GAT) and tRNA-Ile (GAU) respectively. The conventional numbering system for tRNA positions and the symbols used for the modified nucleosides are those adopted in the tRNA database . The number after a ribonucleotide (symbolized by A, U, G, C) or its modified counterpart corresponds to its position in the tRNA molecules. In the case of rRNA, unless otherwise specified, numbers correspond to the equivalent position in the E. coli rRNA. Only nucleoside C* is unconventional. As discussed below, C* found at the wobble position 34 of H. volcanii tRNA-Ile (anticodon C*AU) corresponds to a yet incompletely characterized, probably 'lysidine-type' cytosine, while at position 1342 (1404 E. coli numbering) of H. volcanii 16S rRNA, C* corresponds to another uncharacterized C-derivative of a molecular mass of 330.117 Da (N-330). Detailed chemical structures, scientific and common names corresponding to each indicated modified nucleoside and as well as of the corresponding RNA modification enzymes can be found at http://library.med.utah.edu/RNAmods/ and at http://modomics.genesilico.pl
COG: Cluster of Orthologous Group; ORF: open reading frame; SAM or S- AdoMet: S-Adenosyl-L-Methionine; SSU: small subunit; LSU: large subunit; PTC: peptidyl transferase center; NCBI: National Center for Biotechnology Information; RNP: ribonucleprotein; Nt: nucleotide.
HG and VdC-L designed the study and coordinated the analysis. HG carried out the analysis of the tRNA sequences, CM searched and analyzed the tDNA sequences, CG searched and analyzed the sRNA sequences, WD did the analysis of the rRNA modification in the context of the Ribosome structure, VdC-L predicted all the modifications genes and drafted the manuscript. All authors read and approved the final manuscript.
Sequences of the 41 mature tRNAs + 6 tDNA covering the whole decoding set of Haloferax volcanii.
Sequences of the 52 tDNAs of Haloferax volcanii and Genetic Code coverage.
The hBHBh' structure of the introns in pre-tRNA-Met (CAU), pre-tRNA-Gln (UUG) and pre-tRNA-Trp (CCA) in Haloferax volcanii.
sRNAs of Haloferax volcanii predicted to modify Cm34 and Um39 in tRNA-Trp.
sRNA of Haloferax volcanii predicted to modify Cm34 in tRNA-Met.
sRNA of Haloferax volcanii predicted to modify Gm1934 (Gm1950) in 23S rRNA.
H/ACA sRNA sequences of Haloferax volcanii and some homologous sequences.
Representation of C/D box and H/ACA sRNA in archaea.
Legends for Additional files
We wish to thank Jonathan Eisen for making the sequence of Haloferax volcanii publicly available before publication, Ian K. Blaby for improving the manuscript and Jean-Pierre Rousset for stimulating discussions. This work was supported by US Public Health Service grant number GM19351 to Maurille J. Fournier (W.D) and the National Science Foundation grant number MCB-05169448 to V dC.L.