|Home | About | Journals | Submit | Contact Us | Français|
The online version of this article has been published under an open access model. Users are entitled to use, reproduce, disseminate, or display the open access version of this article for non-commercial purposes provided that: the original authorship is properly and fully attributed; the Journal and Oxford University Press are attributed as the original place of publication with the correct citation details given; if an article is subsequently reproduced or disseminated not in its entirety but only in part or as a derivative work this must be clearly indicated. For commercial re-use, please contact firstname.lastname@example.org
We present the first comprehensive analysis of RNA polymerase III (Pol III) transcribed genes in ten yeast genomes. This set includes all tRNA genes (tDNA) and genes coding for SNR6 (U6), SNR52, SCR1 and RPR1 RNA in the nine hemiascomycetes Saccharomyces cerevisiae, Saccharomyces castellii, Candida glabrata, Kluyveromyces waltii, Kluyveromyces lactis, Eremothecium gossypii, Debaryomyces hansenii, Candida albicans, Yarrowia lipolytica and the archiascomycete Schizosaccharomyces pombe. We systematically analysed sequence specificities of tRNA genes, polymorphism, variability of introns, gene redundancy and gene clustering. Analysis of decoding strategies showed that yeasts close to S.cerevisiae use bacterial decoding rules to read the Leu CUN and Arg CGN codons, in contrast to all other known Eukaryotes. In D.hansenii and C.albicans, we identified a novel tDNA-Leu (AAG), reading the Leu CUU/CUC/CUA codons with an unusual G at position 32. A systematic ‘p-distance tree’ using the 60 variable positions of the tRNA molecule revealed that most tDNAs cluster into amino acid-specific sub-trees, suggesting that, within hemiascomycetes, orthologous tDNAs are more closely related than paralogs. We finally determined the bipartite A- and B-box sequences recognized by TFIIIC. These minimal sequences are nearly conserved throughout hemiascomycetes and were satisfactorily retrieved at appropriate locations in other Pol III genes.
In eukaryotes, RNA polymerase III (Pol III) transcribes a few hundreds short non-coding RNA genes, the bulk of which are the transfer RNA genes (tDNA) and the 5S RNA genes (1,2). In Saccharomyces cerevisiae, a few other non-coding RNAs are also synthesized by Pol III: (i) SNR6 which is the U6 RNA component of the spliceosome (3); (ii) RPR1, the RNA component of ribonuclease P (4) and (iii) SCR1, the RNA component of the signal recognition particle (SRP) (5). Recently, genome wide investigation on Pol III transcription machinery occupancy in S.cerevisiae showed that all known Pol III genes are occupied and revealed new potential Pol III genes. These are SNR52 (6), a C/D snoRNA responsible for the 2′-O-methylation of small subunit rRNA at A420 (7,8) which was previously considered to be a Pol II product; and ZOD1, whose function remains unkwown (9).
The transcription of all Pol III genes is dependent on two general transcription factors: the assembling factor TFIIIC and the recruiting factor TFIIIB (10). The 5S RNA genes require a specific recognition factor, TFIIIA (11), while all other Pol III genes are first recognized by TFIIIC (which also recognizes the 5S RNA-TFIIIA complex). After binding to DNA, TFIIIC directs the upstream binding of TFIIIB, thus forming a pre-initiation complex able of recruiting Pol III for multiple cycles of transcription (12,13). Among the three eukaryotic RNA polymerases, Pol III is the only one featuring a reinitiation mechanism that makes the production of Pol III transcripts extremely efficient (13,14). Pol III terminates transcription at short tracks of T's (preferentially followed by A or G) in the RNA-like strand (15). Unlike Archaea and Bacteria, the use of sequences upstream of eukaryotic tRNA (and other Pol III) genes as promotor elements is rare, as demonstrated in S.cerevisiae (16–18). For review on tRNA genes and Pol III transcription, see also (2,10,12).
For all Pol III genes, but the 5S RNA gene, the primary recognition by TFIIIC relies on two short promoter sequences, which are generally internal to the genes. In tRNA, the nucleotides implicated are those that make up the universally conserved tertiary base pairs bridging the D- and T-loops. At the genomic DNA level, these nucleotides are the two variably distant linear promoter regions recognized by TFIIIC (traditionally referred to as A- and B-boxes) (19,20). Early definitions of the A- and B-box consensus sequences [TGGCnnAGTGG and GGTTCGAnnCC, respectively (19)] appear now too restrictive as more sequences become available. Updated consensus have been later proposed for the A-box; all terminate with (or extend beyond) the two universally conserved nucleotides G18 and G19 [see, e.g. (5)]. The A- and B-promoter sequences must also be present in other Pol III genes, but no accurate definition and genome wide compilation of these promoter sequences were presented yet.
A recent work, based on the comparative analysis of genomes showed that tRNA genes from Eukaryotes, Archaea and Bacteria display both common and domain-specific features (21). However, only two yeasts (S.cerevisiae and Schizosaccharomyces pombe) among seven eukaryotes were available. The large number of hemiascomycetous genomes now sequenced (22,23) offers the opportunity to perform a detailed comparative genomics of Pol III genes. With compact genomes (less than 20 Mb) these organisms give access to a wide evolutionary range, even larger than that of Chordates if one considers the phylogenetic distance between S.cerevisiae and Yarrowia lipolytica (24). Pol III genes were analysed in nine yeast species across the evolutionary tree of hemiascomycetes: S.cerevisiae (25), Saccharomyces castellii (26) which is now placed in the Naumovia clade (27) close to the Saccharomyces, Candida glabrata (24), Kluyveromyces waltii (28), Kluyveromyces lactis (24), Eremothecium gossypii (29), Debaryomyces hansenii (24), Candida albicans (30) and Yarrowia lipolytica (24). The archiascomycete S.pombe (31) was used as an outgroup.
Over 2300 Pol III genes were extracted from these ten yeast genomes. The majority of them are the tRNA genes (a detailed list of the 2335 tRNA genes is available as Supplementary Data). Whether these tDNAs from the ten yeast genomes obey the rules previously defined for eukaryotic tDNA was tested. Several sequence deviations to the cloverleaf tRNA model that may possibly affect the tertiary structure of some tRNAs were discovered. Peculiarities in the decoding of leucine and arginine codons, previously seen in S.cerevisiae only, are extended to related yeasts. Eight of the genomes harbour head-to-tail tDNA pairs, with a maximum of 17 cases in D.hansenii. We also have performed a global distance analysis over all tDNA sequences. Despite the short length of tDNA (hence presumed low informational content), the results suggest a common phylogenetic origin inside each amino acid-specific family, confirm a case of tRNA capture and suggest a novel one. Finally, from the compilation of all tDNA data, the sequences TRGYnnAnnnG (11 nt) and GWTCRAnnC (9 nt) were derived as the hemiascomycetous signatures of the Pol III transcriptional promoters for tDNA. These identity elements are also found at appropriate locations in the other Pol III RNA genes SNR6, SNR52, RPR1 and SCR1.
The ten genomes investigated are listed in Supplementary Table 1; all but the archiascomycete S.pombe belong to hemiascomycetes. Genomes are also referred to with a four-letter acronym made of the two first letters of the gender name followed by the two first letters of the species name (e.g. SACE stands for Saccharomyces cerevisiae). The tRNA genes and the tRNA are designated as in this example: ‘tDNA-Leu (TAG)’ and ‘tRNA-Leu (UAG)’, respectively. The anticodons are always written between brackets with nt 34, indicated or not, in the first position. The conventional IUB/IUPAC degenerate DNA alphabet (32) and special symbols used for base pairings combinations are defined in the legend to Figure 2; ‘n’ is often used, instead of ‘N’, for clarity. The universal conventional numbering system for tRNA positions is that adopted in the tRNA database (33). The sequences of all tDNA identified in the ten genomes are given in the Supplementary Table 4.
The full set of nuclear tRNA genes were searched in each genome using the procedure described earlier (21). This search method is based on the detection in a given genome of the nucleotide sequences corresponding to the eukaryotic-type conserved nuclear tRNA cloverleaf structure [Figure 2, see also Supplementary Table 4 in (21)]. However, in the case of Y.lipolytica (acronym YALI), this procedure failed to reveal a number of tDNA, otherwise correctly detected by tRNAscan-SE (34). These tDNA contained an unexpected number of GT pairs (in tDNA, GU in tRNA stems) and/or Watson–Crick mismatched pairs within the stems of the cloverleaf structure. Our initial search parameters were therefore adapted (for this particular genome only) as follows: number of GT pairs allowed in the anticodon stem: three (instead of two); total number of mismatches in the four stem: three (instead of two); total number of GT and mismatched pairs: six (instead of five) (Figure 2).
Two possible pseudogenes were identified in Y.lipolytica: one encoding tRNA-Ala (anticodon AGC) (cove score 61.36) which differs from the other 29 copies by a T instead of a G at position 63, thus creating a second mismatched pair in the T-stem; the second encoding tRNA-Leu (AAG) (cove score 51.86) which, among 21 copies, has a T instead of G at position 19, thus creating a mismatch in place of the usual G19C56 tertiary base pair. The functionality of these two gene products is therefore questionable. In the genome of K.waltii (KLWA), a mitochondrial origin was suspected for 13 single copy tDNAs for the following reasons: (i) these tDNAs were located in three short contigs (G194contig_278, G194contig_341 and G194contig_362); (ii) these three contigs display continuous low GC content (about 20% compared to 44% for the total of all contigs); (iii) each of these 13 single copy tDNA was markedly different from other bona fide nuclear and multiple copy tDNA bearing the same anticodon; (iv) Blast search of these three contigs revealed high scores with the mitochondrial genome of the close species K.lactis. These three low GC content contigs were therefore considered as actual fragments of the mitochondrial genome of K.waltii and not as ancient permanent inclusions of its mitochondrial genome into the nuclear genome.
In K.waltii (KLWA), the genes encoding tRNA-Leu (CAA, decoding the UUG codon) and tRNA-Arg (CCG, decoding the CGG codon) were not identified (in Figure 1, these missing genes are indicated by a ‘/’ sign). In S.castellii (SACA), tRNA-Pro (AGG, decoding the CCU and CCC codons) is also missing. For these two genomes, (as well as for C.albicans (CAAL), the genomic sequence is not complete.
In order to align perfectly the sequences, introns (if any, located between nt 37 and 38), the base 47 (not always present) and the V-arm extension (from positions 47 to 48, present only in Leu and Ser isoacceptors) were removed. All sequence variations due to the polymorphism of some genes (e.g. a GC to AT base pair change in a stem) and not located in the eliminated regions listed above were selected for the p-distance analysis (some examples are given in Figure 4A). Only one tDNA copy was retained per family of strictly identical sequences and this lead to a total of 603 different sequences out of a total of 2335 tDNA sequences examined in this work. This number represents an intermediate between the total number of different types of tRNA for all ten genomes (426 tRNA/anticodon types) and the total number of tRNA genes (2335 genes). The comparative analysis of the 603 sequences required the computation of 181 503 pairwise p-distance values. The p-distance is defined as the number of nucleotide sites, which are different between any pair of sequences compared, divided by the total number of common nucleotides. The largest p-distance we observed was that between tDNA-Leu (AAG) from Y.lipolytica and tDNA-Glu (CTC) from D.hansenii (54 positions different out of 75). A histogram of the p-distance is shown in Figure 4B. The p-distance tree presented in Figure 4C was built by the Neighbor-Joining method (35) implemented within the MEGA2 software (36).
Four ncRNA Pol III genes were also considered: SNR6 (U6), SNR52, RPR1 and SCR1. For S.cerevisiae, the boundaries of the mature products of these four genes were taken from the gene definition in SGD (URL's given in Supplementary Data). These four Pol III genes were identified in the four recently sequenced genomes (C.glabrata, K.lactis, D.hansenii and Y.lipolytica) as follows: SNR6: this gene was previously identified by a BlastN search (24); SNR52: following a genomic BlastN search run on each genome with the S.cerevisiae gene as entry. In other genomes, SNR6 and SNR52 genes were identified with BlastN when not annotated. RPR1: this gene was recently identified in the ten genomes explored and more (4). SCR1: this gene, previously identified in C.glabrata, K.lactis, D.hansenii and Y.lipolytica on the basis of a structural identity with the S.cerevisiae genes (see Supplementary Table S6 in (24)), was identified thanks to the conservation of the P6 and P8 helices (for nomenclature, see (37)). All RPR1 or SCR1 RNAs from the ten genomes were structurally aligned (Figure 7) and the boundaries of the mature products deduced from those of the S.cerevisiae mature RNAs.
The A- and B-boxes of all four genes from S.cerevisiae had already been identified and verified experimentally: SNR6 (3); SNR52 (6); RPR1, (38); SCR1, (5). A- and B-boxes of these four genes were searched in other genomes in and around the mature product sequences using the consensus TRGYnnAnnnG and GWTCRAnnC as determined from tDNA sequences analysis (this work, Figure 5E). The putative A-box of SCR1 genes was located eight bases downstream that previously proposed (5). Detailed data about A- and B-boxes of these four ncRNA genes are presented in Figure 6.
Using a combination of cloverleaf structure detection algorithm (21) and tRNAscan-SE (34), 2335 genes encoding nuclear tRNA molecules (tDNA) were identified from genomic sequences of ten yeast species (listed in Supplementary Data). Among them, 47 different anticodons were identified. The numbers of genes encoding each isoacceptor are given in Figure 1. Note the significant variations between species.
With a few exceptions, all tRNAs obey the canonical eukaryotic cloverleaf model (Figure 2) initially established by comparing tDNA sequences from S.cerevisiae and S.pombe and five other eukaryotes (21). Features specific to unique tDNAs are listed in Supplementary Table 2 and illustrated in Figure 2. For example, in S.pombe, the three copies of the tDNA-Ser (anticodon GCT) harbour an extra ‘20c’ base in the D-loop. In D.hansenii and C.albicans, a special tDNA-Ser (CAG) that reads the CUG codon as Ser instead of Leu contains an unusual G at position 33 (39–41). In the same organism, the tDNA-Leu (AAG) that reads the CUU, CUC and CUA codons, contains an unusual G at position 32 (see also below and in Figure 3). A few tDNAs contains unusual accumulations of non-Watson–Crick base in the cloverleaf stems. For example, in Y.lipolytica, tDNA-Ala (TGC) has two mismatched base pairs and four GT base pairs, three of which are present in a row in the anticodon stem. Earlier analysis (21) showed that a maximum of two GT pairs in the anticodon stem and five GT or mismatched base pairs in all stems are present in the canonical eukaryotic cloverleaf. A near perfect conservation of the features characteristic of tDNA-iMet (21) is observed: AT pair in 1–72 (TA in S.pombe), positions 17, 17a, 20a and 20b unoccupied, GGGCT in 29–33 (AGGCT in C.albicans), and A in 54 and 60. Unusual sequence features occurring in the D- and T-loops, that may affect the transcription of tRNA genes, are discussed below.
All members of a multicopy tDNA gene family in a given yeast species (tRNAs harbouring the same anticodon) display neighbour (sometimes strictly identical) sequences (from nt 1 to 73), with the exception of two cases. This is why the total number of distinct tRNA molecules in each yeast species exceeds the number of isoacceptor tRNA (‘tRNA species’ and ‘variant tRNAs’, respectively, indicated at bottom of Figure 1). Slight sequence variations are frequent (e.g. a GC pair changed into an AT in some copies). Except in two cases, no markedly different tDNAs coding for the same amino acid were identified. The tDNA-Arg (CCG) departs from other tDNA-Arg in five of the ten genomes investigated. Also, the two copies of tDNA-Thr (CGT) of Y.lipolytica differ at 20 of the 75 positions of the tRNA molecule and, moreover, one has an intron (13 nt) whereas the other does not. These two cases are examined in details below.
In eukaryotic pre-tRNA molecules encoded by nuclear genes, the introns, when present, are always located between 37 and 38 nt (1 nt downstream of the anticodon). One exception is the nucleomorph (remnant eukaryotic nucleus) of the cryptophyte Guillardia theta (42) where introns (as short as 3 nt) have recently been found at non-canonical positions. The conservation of the intron position is dictated by the eukaryotic pre-tRNA-specific splicing machinery that has probably evolved from the unique ancestral machinery of Archaea (43,44) [reviewed in (45–47)]. In Bacteria, the only introns found in the tRNA genes are autocatalytic group I introns located within the anticodon loop [reviewed in (48,49)].
Among the 47 tRNA species identified in the ten yeast genomes studied, 40 display an intron in at least one species (Figure 1). tRNA-iMet (CAU) is the only isoacceptor that never bears an intron. Introns are universal to all pre-tRNA-Ile (UAU), pre-tRNA-Ser (CGA), pre-tRNA-Ser GCU) and pre-tRNA-Tyr (GUA). With 27 of the 44 pre-tRNAs containing intron, Y.lipolytica is the richest intron-containing yeast and eukaryotic species known to date. Next comes C.albicans with 24 intron-containing pre-tRNAs (over a total of 42). The remaining eight genomes display 10 to 16 pre-tRNAs harbouring introns.
Intron sizes range from 7 nt (pre-tRNA-Met (CAU) in S.pombe) to 71 nt (pre-tRNA-Ile (UAU) from C.glabrata). Sequence variation is observed among the different copies of a same tRNA species. For example, among the 27 copies of Y.lipolytica pre-tRNA-Glu (CUC), two copies have a 20 nt intron and the 25 other ones harbour 9 or 10 nt introns. Morover, in some other types of pre-tRNA of Y.lipolytica, the intron is simply missing in some tDNA copies. For example, among the 12 copies of pre-tRNA-His (GUG) of Y.lipolytica, only four copies have introns of 14, 15, 16 and 22 nt, respectively and the remaining seven copies lack the intron. Such polymorphism is not reported in other Eukarya, except a single case in Caenorhabditis elegans [see in (21)]. Preliminary analysis of recently available Cryptococcus neoformans genome shows that the presence or absence of introns in the multiple copies of the same isoacceptor is more encountered than initially thought (C. Marck, unpublished data).
No obvious correlations emerge from the presence or absence of introns in the various isoacceptors. However, the universal presence of intron in pre-tRNA-Ile (UAU) and pre-tRNA-Tyr (GUA) in yeasts and in all other sequenced eukaryotes is probably related to the need of specific posttranscriptional isomerization of uridine into pseudouridine (Ψ) (50,51). This modification is catalysed by tRNA pseudouridine synthases during pre-tRNA maturation, respectively PUS1 [catalyzing the intron-dependent formation of Ψ34 and Ψ36 in pre-tRNA-Ile (52,53)] and PUS7 [catalyzing the intron-dependent formation of Ψ35 in pre-tRNA-Tyr (54)]. Likewise, in all yeast genomes analysed (except D.hansenii), tRNA-Leu (CAA) harbours an intron that, in S.cerevisiae, is known to be essential for the intron-dependent formation of m5C34 catalysed by the TRM4 methyltransferase (55,56). The identity of C34 modification in intron-less tRNA-Leu (C34AA) of D.hansenii is not known but we may anticipate it is 2′-O-methyl-C34 (Cm34) as in the intron-less tRNA-Leu (Cm34AA) of Candida cylindracea and Drosophila melanogaster [see in (33)]. In S.cerevisiae, the methylation of residue 34 in tRNA-Leu (Cm34AA), as well as tRNA-Phe (Gm34AA) and tRNA-Trp (Cm34CA) is catalysed by the TRM7 methyltransferase only in intron-less pre-tRNA, thus subsequently after the removal of intron (57,58). No such correlation can be made between the apparently universal presence of introns in pre-tRNA-Ser (CGA) and (GCU) and any modified nucleotides that are present in the corresponding mature tRNA, or possible alternate base pairing configurations [see (59)].
The number of copies encoding the same tRNA varies widely between different isoacceptors within a single yeast species and for the same isoacceptor between different yeast species. For example, several tDNAs exist as single copies while 34 copies of the (DNA-Lys (CTT) are present in Y.lipolytica. The same tDNA-Glu (CTC) is encoded by 1 up to 27 copies according to the species (Figure 1). Highly redundant tRNA genes usually correspond to abundant cellular tRNA molecules and poorly redundant tDNA to minor cellular tRNA (60–62) [reviewed in (63), see also (64)]. The total number of tRNA genes varies from 131 (in C.albicans) to 510 (in Y.lipolytica—see bottom of Figure 1), while 274 tDNA were reported for S.cerevisiae (62,65). After correction for genome length, it appears that the tRNA gene density is three times higher in Y.lipolytica than in C.albicans.
For all hemiascomycetous species, tRNA genes appear scattered throughout the genome. In S.cerevisiae, 39 pairs of tRNA genes result from the ancestral duplication of the whole genome (66). Consistent with their variation in total number, the average distances between two successive tRNA genes on the chromosome maps range from 40 kb in Y.lipolytica to 110 kb in C.albicans. No gene cluster was found in the hemiascomycetous genomes examined except in D.hansenii where eight identical tandem co-oriented copies of a tDNA-Lys (CTT) are present on chromosome B. The distances separating these genes (188 to 1855 bp) indicate independent transcription. This is consistent with the frequent formation of tandem genes in this particular species (22).
Clusters of tDNAs have been detected in a variety of eukaryotic genomes including D.melanogaster (67) and the archiascomycete S.pombe. In the latter case 27 tDNAs are found in the 50 kb region surrounding chromosome B centromere (68), and 20 other tDNA in the 75 kb region around the chromosome C centromere (31). It is remarkable that no such cluster exists in any of the hemiascomycetes studied. Instead, most tDNA are scattered throughout the genome in a random orientation relative to flanking genes.
In studying the localization of tDNA in hemiascomycetes, we were surprised to observe numerous cases of head-to-tail pairs of tDNA. In such pairs, the distance between the two genes ranges from 5 to 26 nt (Supplementary Table 3). This distance is shorter than the minimal sequence required for the independent transcription of the second gene [about 100 nt, (2)]. A few of the pairs had already been noted, as in S.cerevisiae and S.pombe (69–71) and two more recently discovered in the S.cerevisiae (62,65), but their almost universal presence in hemiascomycetes was not suspected. In yeast and Xenopus oocyte nuclear extract, the co-transcription of the paired tDNAs into a single precursor followed by processing to mature-size tRNA molecules was experimentally demonstrated (70,72). It is likely that the same mechanism operates for all pairs now identified. In agreement with this hypothesis, all genes are always co-oriented in a given pair and the short intergenic sequences show no obvious Pol III terminator (tracks of T's), although the strength of terminators is difficult to predict (15).
Interestingly, the tDNA pairs differ from one yeast species to the next, with only limited conservation (e.g. tDNA-Arg/tDNA-Asp pairs found in S.cerevisiae, S.castellii and K.lactis) and the pairs are often found in multiple copies within a genome, e.g. six occurrences of the tDNA-Ile/tDNA-Ala pair in D.hansenii. In some cases, the pair is composed of two identical tDNA but in most cases, two distinct tDNA are involved. The expansion of identical pairs in a genome suggests successive duplications of the pairs within each phylogenetic branch through a yet unknown mechanism. Single copies of the tRNA genes identical to those involved in pairs are also present in the same genome. No correlation can be made yet between the decoding capacity of each tRNA of the pairs and the level of their expression. It is possible that some enzymatic modification of nucleotides (like in the case of intron-containing tRNA, see above) or correct folding is however dependent on the expression of such paired pre-tRNAs.
Three major sparing strategies allow an organism to read the genetic code information (insets in Figure 1) with a limited repertoire of anticodons in the tRNAs (21). The first sparing strategy is the universal ‘A34 or G34-sparing’ in which either an A34- or G34-containing tRNA decodes the two pyrimidine ending codons. In these cases, the A34 is always posttrancriptionally modified into inosine (I34) (73) [reviewed in (74)] while the G34, is often modified into Gm34 or Q34 derivatives [reviewed in (75,76)]. The mutually exclusive existence of an A or G at the first position of the anticodon is true in the 16 four-codon boxes of the genetic code, despite the fact that a given tRNA species (given anticodon) is usually encoded by multiple genes. Such cases of tRNA sparing are indicated by arrows and symbols ‘ΔA34’ or ‘ΔG34 in Figure 1. As a consequence of this first sparing rule, the maximum number of tRNA species in any organism cannot exceed 46 (64 codons, minus 3 stop codons, minus 16 cases of ‘A34 or G34-sparing’ and plus 1 for the iMet codon leads to 46). This rule holds for all three biological domains: Archaea, Bacteria or Eukarya. This number of 46 tRNA/anticodon species is reached in S.pombe but hemiascomycetous yeasts lower the number of tRNAs (Figure 1 bottom). The use of anticodon starting with either A34 or G34 is the same in all three kingdoms for Phe, Tyr, His, Gln, Asn, Lys, Asp, Glu, Cys, Trp, Arg, Ser (AGY codons only) and Gly (Figure 1, Phe in upper panel, others in lower panel). In contrast, this choice differs between Eukarya and Archaea/Bacteria for the other amino acids (Figure 1, upper panel). Eukarya use the anticodons starting with A34 (G34-sparing, noted ‘ΔG34’ in Figure 1) while Archaea and Bacteria use anticodons starting with G34 (A34-sparing, noted ‘ΔA34’) (21).
To achieve additional reduction (down to 44 or 42 tRNA types), hemiascomycetous yeasts use a second sparing strategy, known as ‘C34-sparing strategy’, which is also used in all three domains of life (indicated as ‘ΔC34’ in Figure 1). When the tRNA with anticodon starting with C34 is absent, the cognate G3-ending codon is read by the U34-containing isoacceptor tRNA. In this case, U34 is always modified [reviewed in (75–77)]. In the genome of Y.lipolytica, a set of 44 tRNA is enough to read all 62 codons, the tRNA-Arg (C34CG) and tRNA-Gly (C34CC) being both absent. Other genomes, like S.cerevisiae, C.glabrata, K.waltii, C.albicans and S.castellii, lack a few more tRNAs with anticodon starting with C34 (Figure 1). Note that C.glabrata and K.lactis display exactly the same set of 42 tRNA (this work) as S.cerevisiae (62,65). As a matter of fact, this moderate variation in the tRNA ‘repertoire’ (between 42 and 46 tRNA) hides drastic changes in the way each individual yeast decodes Leu and Arg with respect to other eukaryotes (discussed below).
The largest variability of tRNA repertoire between yeasts occurs in the decoding of the Leu CTN and Arg CGN codons (boxed in Figure 1). This situation is illustrated more in details in Figure 3A–D together with additional eukaryotes. Genes coding for each of the tRNA reading one of the two purine-ending Leu codons UUA and UUG are universally present. In contrast, two distinct strategies are used to read the four Leu CUN codons. Four of the nine hemiascomycetes and S.pombe use the ‘Eukaryotic-type G34-sparing’ strategy, as expected (tRNA-Leu (A34AG). The five hemiascomycetes S.cerevisiae, C.glabrata, K.waltii, K.lactis and E.gossypii, which belong to the same evolutionary branch, use the ‘Bacterial/Archaeal A34-sparing’ (tRNA-Leu (G34AG). This case is unique among eukaryotes.
The other two Leu CUR codons (CUA and CUG in Figure 3A) are read by either a unique tRNA-Leu harbouring a UAG anticodon (as in S.cerevisiae, C.glabrata, K.lactis, E.gossypii, S.castellii and S.pombe) in which U34 is not posttranscriptionally modified (78), or by two tRNAs (in K.waltii and Y.lipolytica), one with U34AG and the other with C34AG anticodon. Since no RNA sequence of the fully mature tRNA-Leu (UAG) other than that of S.cerevisiae is available, we do not know whether U34 of this tRNA-Leu (UAG) is modified in other yeasts. The peculiar decoding of the Leu CUN codons in D.hansenii and C.albicans (because of a change in the amino acid assignment for one of these codons—shown in Figure 3A and B) is commented below.
Another example of an imitation of bacterial sparing strategy is found in the decoding of the Arg CGN codons (Figures 3C and D). Whereas Y.lipolytica (as the archiascomycete S.pombe), uses the typical ‘Eukaryotic-type G34-sparing’ strategy, all other hemiascomycetes use a third type of sparing, known as ‘U34-sparing’ strategy, which is specific to arginine CGN codons and only known in Bacteria (21). To read Arg CGN codons, Y.lipolytica and S.pombe use a tRNA-Arg (A34CG) reading CGU and CGC codons and a tRNA-Arg (U34CG) reading CGA codons [and also CGG codons if the tRNA-Arg (U34CG) is absent, as in Y.lipolytica]. In this case, U34 is possibly modified into a yet unknown derivative of the type mcm5U like in tRNA-Arg [mcm5U34CU) (33)]. In all other eight hemiascomycetes, the tRNA-Arg (U34CG) is and a single tRNA-Arg (A34CG) reads the three Arg codons CGU, CGC and also CGA. In this case, A34 is probably modified into I34 and the decoding of CGA codon involves a wobble I34A3 pairing mode which was initially anticipated (79,80) but only recently demonstrated to occur during mRNA decoding on the ribosome (81). The CGG codons are read by a second tRNA-Arg (C34CG) and, in these eight organisms obeying the U34-sparing strategy, this tRNA becomes essential, while it may be absent when the tRNA-Arg (U34CG) is present as in Y.lipolytica, D.melanogaster and Encephalitozoon cuniculi (usual C34-sparing strategy) (Figure 3C).
In D.hansenii and C.albicans, the ‘Leu’ codon, CUG, is read as Ser (40) (Figure 3A and B). The tRNA-Leu (U34AG), which reads the CUA codon in all other eight yeasts investigated, is missing in these two genomes. Consequently, the Leu codon CUA must be read by the tRNA-Leu (A34AG), which cannot be the major tRNA-Leu according to the codon usage and the existence of only two gene copies in each genome. This hypothesis implies that an I34A3 wobble pairing exists in the codon-anticodon pairing during translation on the ribosome. This decoding strategy is analogous to the ‘Bacterial-type U34-sparing’ mode of reading the three Arg codons CGU, CGC and CGA as discussed above [see also left part of Figure 3C and D and Figure 6 in (21)].
In summary, the absence of a tRNA-Leu (U34AG) in both D.hansenii and C.albicans (U34-sparing strategy) appears consistent with the need to avoid any misreading of the Ser codon CUG of the CUN decoding box. The four codons of this box are read by only two types of tRNA isoacceptors. The first type charges Ser for the CUG codon and possesses an uncommon G at position 33 of the anticodon loop (39–41,82). The second one charges Leu for the three codons CUU, CUC and CUA and possesses also an unusual G, but located at position 32, instead of the universal pyrimidine found in 4000 tDNAs analysed (21). The G at position 32 cannot result from sequencing errors because it is found in the two gene copies in each genome (D.hansenii and C.albicans). For this tRNA, we do not know what is its decoding capability compared to a more ‘normal’ tRNA and whether A34 is posttrancriptionnally modified into I34 (83,84).
The presence of G32 or G33 instead of the universal pyrimidines [C or U, see in (47)] probably alters the anticodon stem–loop structure and allows accurate readings of CUU, CUC and also CUA as leucine in the case of tRNA-Leu (A34AG) [possibly (I34AG)] and CUG as serine in the case of tRNA-Ser (C34AG). The coexistence of these two types of unusual tRNA among the tRNA population within the same organism (D.hansenii and C.albicans) is therefore not a coincidence but rather an important novel feature of the decoding strategy in these microorganisms [see also (85)].
Evolutionary relationship between the tRNA gene species of the different yeasts were investigated using a pairwise p-distance matrix analysis carried over the 603 variant tDNA sequences (see bottom of Figure 1) identified in the ten yeasts. Some examples of tDNA sequences prepared for p-distance computation are shown in Figure 4A. All pairwise distances were computed after removal of the intronic sequences (if any), and of base 47 and V-arm sequences in tDNAs-Leu and tDNAs-Ser. The repartition of the pairwise distances obained (Figure 4B) shows a majority of p-distance values in the range 0.5–0.6 (50–60% difference). The p-distance tree derived from this matrix (according to Materials and Methods) is shown in Figure 4C. Remarkably, for tDNA specific for Gln, Ala, Pro, His, Leu, Ser, iMet, Val and Lys, orthologous tRNA genes (coding for the same amino acid) belonging to the nine hemiascomycetes or even to all ten yeast species cluster together. In such instances, the evolutionary divergence of sequences between orthologous tDNA of different yeast species is less than the divergence between paralogous tDNA species (charging different amino acids) within a single yeast species.
The sequences of three other isoacceptor families (specific for Gly, Asp and Glu) cluster less perfectly: indeed, tDNA-Gly split into two clusters while the tDNA specific for Asp and Glu are fused in the same cluster. The tDNA for amino acids Cys, Trp, Ile, Thr, Tyr, Asn and Phe cluster together for all hemiascomycetous yeasts but the clusters do not contain orthologs from S.pombe. The fact that initiator tDNA-Met and elongator tDNA-Met do not cluster together was expected due to clear singularities in the sequences for initiator tRNA [discussed in detail in (21)].
A novel case of ‘tDNA mimicry’ was identified: the only tDNA missing in the elongator tDNA-Met cluster (light box) is that of Y.lipolytica that clusters inside the Thr cluster (indicated by + YALI Met (CAT) on the right side of the Figure 4C). This tDNA-Met (present in nine identical copies) is very close in sequence to one of the two copies of tDNA-Thr (TGT) of Y.lipolytica (57 positions identical) while these two copies diverge at 20 positions. These data are indicative of a possible tDNA capture (tDNA-Met derived of tDNA-Thr in Y.lipolytica) similar to the case of tDNA-Arg (CCG) commented below.
It is worth mentioning that the different tDNA-Leu and tDNA-Ser form a unique cluster, despite the fact that these amino acids correspond to two distinct decoding boxes (4 + 2 codons for each). This observation is consistent with the fact that these tDNAs are phylogenetically related (86). Interestingly, the tDNA-Ser harbouring a CAG anticodon, hence reading CUG as Ser instead of Leu in C.albicans and D.hansenii (detailed in Figure 3A and B and discussed above), clusters with the Ser-tDNAs and not with the Leu-tDNAs, thus attesting to its clear affiliation to the tDNA-Ser family. In contrast, the sequences of the five tDNA-Arg isoacceptors, which also belong to two different decoding boxes (CGN and AGR), are split into two separated clusters. The first cluster (noted ‘3,4/5 Arg’, at bottom of Figure 4B) contains all tDNA-Arg except six of the eight tDNA-Arg (CCG) that form a separate cluster (noted ‘0,1 0,1 Arg (CCG)’) close to the tDNA-Asp/tDNA-Glu cluster. Fender and coworkers proposed that the arginine specific tRNA (CCG) gene from S.cerevisiae (as well as those of Saccharomyces uvarum, Zygosaccharomyces rouxii, C.glabrata and K.lactis) is a remnant of a former aspartate acceptor (87). The conversion of only two bases (G38 and U73 into C38 and G73, respectively) in an in vitro transcript of tDNA-Arg (CCG) is sufficient to allow mutant tRNA-Arg (CCG) to become an aspartate acceptor (87). We now show that this tDNA-Arg (CCG), which does not exist in Y.lipolytica, is also presumably derived from the tDNA-Asp (GTC) in S.castellii and E.gossypii but not in D.hansenii and C.albicans (tDNA-Arg (CCG) allowing us to date the recruiting event on the hemiascomycete tree (see Discussion).
The large collection of tDNA sequences extracted from ten yeasts allows for a better definition of the A- and B-consensus sequences which are recognized by the transcription factor TFIIIC. Only one G remains in the final genomic consensus of the A-box if the variable occupancy of the optional bases 17 and 17a of the D-loop is considered. The consensus sequence (in the form of a cloverleaf) of the 274 tDNAs from S.cerevisiae is shown in Figure 5A, while Figure 5B illustrates the variable distance between the A- and B-boxes and Figure 5C lists the conserved and semi-conserved nucleotides found in the tDNAs of each of the yeasts examined in this work. The position numbered 17a (indicated by arrow and asterisk in Figure 5A and B) is never occupied in any of eukaryotic tDNAs sequenced so far. This is a major difference with the situation in tDNAs of archaeal and bacterial genomes where position 17a (and 17) is occupied in 64 and 7% of the tDNAs, respectively [see Supplementary Table 1 in (21)]. Also few nucleotides are strictly conserved (T8, G10, A14, G19 and A21 in the A-box; G53, T55, C56, A58 and C61 in the B-box). Most of the sequence variability occurs in the evolutionarily distant yeast S.pombe (sequence exceptions are indicated in the boxes surrounding the cloverleaf in Figure 5A).
In the A-box, exceptions to the conserved G10 are mostly found in the tDNA-Leu and tDNA-Ser of D.hansenii, C.albicans and Y.lipolytica and S.pombe (these tDNAs are shown as grey background in Figure 1). At position 18, an A (instead of G) is found twice. The tDNA-Pro (TGG) (3 copies) of Y.lipolytica harbours an unusual A18 (instead of G18), which probably allows a A18U55 tertiary base pair instead of the bifurcated GU pair in the tRNA trancript (88,89). In S.pombe, the single copy tRNA-Arg (CCG) harbours an unusual A18G55 tertiary base pair, and no other tDNA bearing G at position 55 has been identified.
Taking into account the various combinations of bases present or absent at the four positions 17, 17a, 20a and 20b of the cloverleaf, six different DNA patterns are possible (Figure 5D). Similar results and conclusions are obtained with the tDNAs of the nine other yeasts (data not shown). Remarkably, if the six patterns are combined, a G (either G18 or G19, as shown in Figure 5E) is always found, in the genomic sequences, four bases downstream of the universally conserved A14. We therefore hypothesized that the minimal identity elements of the A-box (A-box signature) recognized by the S.cerevisiae transcription factor TFIIIC is the 11 nt sequence TRGYnnAnnnG, ending with only one G (n being any base, R a purine and Y a pyrimidine).
Similarly, we assume that the minimal identity elements (sequence signature) of the B-box are the 9 nt sequence GWTCRAnnC (W meaning A or T, Figure 5E). A consensus sequence comprising 11 bases (i.e. including 52–62 bp) was previously reported for the B-box (19), but it clearly appears that 52–62 bp is not conserved for tRNA, even within S.cerevisiae [see also data in Supplementary Table 2 in (21)].A remarkable exception concerns the nine copies of tDNA-Ala (AGC) from S.pombe that harbour an A at position 53 and a T at position 61. This GC to AT base pair change at both edges of the B-box is probably counterbalanced by the greater role of upstream TATA sequence that helps the binding of the second factor TFIIIB in S.pombe (90). Given these minimal consensus for the A- and B-boxes, extracted from the tRNA genes (Figure 5E), we checked whether they could be retrieved in other Pol III genes from the ten genomes.
We then investigated whether the minimal A and B sequences obtained from the tDNA analysis (shown in Figure 5E) are also retrieved in the four other Pol III genes SNR6, SNR52, RPR1 and SCR1 common to the nine hemiascomycetes and S.pombe. The multicopy 5S gene, which is also transcribed by Pol III was not investigated here because it is recognized by its specific transcription factor, TFIIIA. In the case of S.cerevisiae, the A- and B-promoter sequences of all four genes have been experimentally investigated [SNR6 (3); SNR52 (6); RPR1 (38); SCR1 (5)]. In these genes, the promoters are always internal to the primary transcript but, in contrast with tRNA genes, they are, in some cases, external to the mature product (see schemes in Figure 6).
Among the four genes considered, SNR6 (U6) appears the most conserved in sequence. In S.cerevisiae the B-box is exceptionally located about 120 bases beyond the gene and an upstream TATA promoter element is also present (3). The extragenic B block of SNR6 was located in the orthologous genes at comparable distance (109–177 nt, Figure 6). A specialized chromatin structure appears to dictate this peculiar organization of the two promoter sequences (91). Remarkably, the U6 gene of S.pombe is the only Pol III gene known to be interrupted by a spliceosomal intron (92), and the B-box (which perfectly fits the consensus) is located inside this intron rather than downstream of the gene. This peculiar type of organization of the U6 gene is specific to the Schizosaccharomyces genus (93).
The next two genes, SNR52 (6) and RPR1 (94) share a common organization (at least in S.cerevisiae and related genomes): the A- and B-boxes are internal to the transcript but external to the matured product since a leader sequence (dashed lines in Figure 6) is cleaved posttranscriptionally. No structural constraint applies, at the RNA level in this leader sequence and larger variations in the locations of A- and B-boxes occur in SNR52 (−208 to −65 for the A-box, −96 to −9 for the B-box). In the S.cerevisiae gene, a TTTTTT sequence is present 3′ of the A-box; this sequence was shown to be a weak Pol III transcriptional terminator (30% efficient) in the SNR52 context (15). In the gene of S.pombe, we did not identify any A- or B-box nor a Pol III terminator poly-T, arguing for a Pol II transcription for this gene as is the case of most snRNAs in yeasts.
The RPR1 genes previously identified (4) were structurally aligned. Two types of promoter organization (internal or external to the mature product) can be distinguished (Figures 6 and and7A)7A) in accordance with the phylogenetic distances. From S.cerevisiae to K.lactis, both the A- and B-boxes are located upstream of the matured product (as in SNR52 genes). In E.gossypii and D.hansenii, the B-box terminates in the mature product (inside the 5′ strand of the P1 helix) while, in C.albicans and Y.lipolytica, the B-box is fully internal to the mature product (boxed in Figure 6). In C.albicans, the B-box is located in the 5′ strand of the P7 helix and, for Y.lipolytica, in the P3 helix. Similarly to the SNR52 gene, the RPR1 gene of S.pombe is probably a Pol II gene (no A- or B-box and no poly-T terminator could be identified). Exceptions to the B-box consensus were found for S.cerevisiae, D.hansenii and C.albicans. In S.cerevisiae, an A nucleotide at the third position (instead of a T, also seen in SCR1 of S.cerevisiae and S.castellii) does not prevent TFIIIC recognition (38). We noticed a common variation (C at the fifth position) in D.hansenii and C.albicans genes and C at the second position (instead of W) is also observed in the C.albicans SCR1 gene.
The SCR1 genes from the ten genomes were structurally aligned, based on the conservation of P6 and P8 helices (Figure 7B) and the location of the A- and B-boxes carefully examined with respect to the RNA secondary structure. The A-box was previously located at position 10 of the S.cerevisiae gene by Dieci and coworkers (5) (starting nucleotide is U with light grey background in the UGU motif, Figure 7B). Alternatively, the A-box might be located 8 nt downstream (at position 18 of SCR1, green background), where the A-box consensus (TRGYnnAnnnG) is nearly satisfied for nine out of the ten genomes. Mutation of GG at position 19–20 of SCR1 (positions 18 and 19 in tRNA) affects TFIIIC binding, thus suggesting that these 2 nt do belong to the A-box (5). This experimental result fits with the two possible locations for the A-box (starting at 10 or 18 in S.cerevisae SCR1). Clearly, in the case of K.lactis, none of the two A-box positions reasonably fits the consensus while an A-box, with a single variation (at 3rd position), can be found slightly upstream at position −7. In SCR1, the B-box is located 24 to 50 nt downstream the A-box in a region of weak sequence conservation, except in D.hansenii and C.albicans where the B-box overlaps the 5′ strands of P5e and P5f helices.
We present the first comprehensive genome wide analysis of Pol III-dependent genes in ten eukaryotes (nine hemiascomycetes and the archiascomycete S.pombe). This exhaustive analysis unearthed several original observations. Unexpected features for decoding were first revealed. Yeasts close to S.cerevisiae follow the bacterial sparing rules to decode Leu CUN and Arg CGN codons. Such changes, which are unique among eukaryotes, can be precisely dated on the phylogeny of hemiascomycetes. As shown in Figure 8, the most ancient switch appears to be the change of decoding Arg CGN codons from the regular eukaryotic to a bacterial-type (node #1). The change in the genetic code that reassigned the CUG codon to Ser occurred later, in the branch leading to the Candida genus (D.hansenii and C.albicans, node #2). Independently, in another branch leading to other hemiascomycetes, including S.cerevisiae, the decoding of Leu CUN codons switches from the eukaryotic to bacterial mode (G34- to A34-sparing, node #3). Remarkably, S.castellii has reverted to the usual eukaryotic G34-sparing (node #5). The capture of tDNA-Asp leading to a novel tDNA-Arg (CCG) appears to be also concomitant with the events occurring at node #3. Finally, the loss of tDNA-Leu (CAG) seems to have occurred several times independently (in these cases, the CUG codon is read by tRNA-Leu (UAG)).
The large size of the collection of tDNA sequences originating from a single eukaryotic phylum allows extensive comparisons between both orthologous genes (i.e. between yeast species) and paralogous genes within each species. For a given tDNA species (given anticodon), the large variation in the number of gene copies is particularly remarkable [e.g. 1–27 copies for tDNA-Glu (CTC)]. This variation in number is at least partly correlated to variation in codon usage between yeast species. It is also remarkable that within a yeast species, the various gene copies are always (or nearly) identical. Remarkably, specific deviations with respect to the eukaryotic cloverleaf model apply to all gene copies within a genome. For example, the tertiary base pair T15A48 present in all five tDNA-Phe (TGG) in C.albicans replaces the usual R15Y48 pair; G21 is found instead of the universal A21 in all three tDNA-Met (CAT) in S.pombe; the A53T61 pair, which makes the outer bases of the B-box, is substituted to G53C61 in all nine copies of tDNA-Ala (AGC) in S.pombe). This suggests a specific role for such deviations and also the existence of a survey mechanism permanently unifying the different tDNA copies of the same tDNA (same anticodon) within each species.
The sequence homogeneity between orthologous tDNA (tDNA coding for the same amino acid in different genomes) contrasts with the sequence divergence between paralogous tDNAs (tDNAs bearing different amino acid within a same genome) as shown by our p-distance analysis. Note that a similar histogram of distance (Figure 4B) was already reported several years ago with a much more limited tDNA set, insufficient for phylogenetic analysis (86,95). With our new dataset that includes ~600 different tDNA sequences, single clustering of orthologous tDNA was observed for most amino acids, with the sole exception of tDNA from S.pombe, offering the opportunity to examine the significance of the exceptions to this rule. A first exception is the close relation between the tDNA-Arg (CCG) and the tDNA-Asp (GTC) in yeasts close to S.cerevisiae (87). Actually, the origin of the tDNA-Arg (CCG) in the two related genomes D.hansenii and C.albicans still appears unclear. While the tDNA-Arg (CCG) of D.hansenii sides together with other tDNA-Arg within the main tDNA-Arg cluster, that of C.albicans sides into the extra cluster defined by five other tDNA-Arg (CCG) (Figure 4C). For the time being, it seems reasonable to conclude that tDNA-Arg (CCG) from D.hansenii is a regular tDNA-Arg, not derived from a tDNA-Asp (GUC) anscestor, and that this is also the case for C.albicans. It remains that the emergence of the tDNA-Arg (CCG) (Figure 8, node #3) is complex and that detailed analyses of more genomes are necessary to clarify its origin in the different organisms, including hemiascomycetes. The second exception is the intriguing clustering of the tDNA-Met (CAT) from Y.lipolytica into the Thr cluster that suggests a possible case of capture (Figure 8, node #6). Here again, more genomes (close to Y.lipolytica) will be needed to conclude unambiguously.
Prior to this work, the definition of the promoter elements in the A-box recognized by TFIIIC was uncertain. We used the most representative class of Pol III genes, the tRNA genes, which always amount to more than 41 different types of genes and more than 100 gene copies per genome (up to 500), to extract the A- and B-boxes genomic signatures. These short sequence elements were searched and retrieved in four other Pol III ncRNAs from the ten genomes (except two cases of probable Pol II transcribed genes in S.pombe). Examination of the 39 A- and B-box sequences (Figure 6) shows that the consensus signatures are indeed found always at appropriate locations, with a few sequence exceptions.
Directed mutagenesis experiments have established that the B-box is the most critical region for TFIIIC binding and that the interaction between A-box and TFIIIC is less important to the stability of the DNA–TFIIIC complex (96). Among the 2nd, 4th and 5th positions of the B-box (equivalent to positions 54, 56 and 57 of the tRNA), the 4th position, always occupied by a C, is the most critical and its replacement by G lowers the in vitro binding affinity of TFIIIC by 370-fold (96). Only in one case, divergence at the 4th position of the B-box over 39 exists (a T is present instead of C in the SNR6 gene of C.albicans). In accordance with the less prominent role of the A-box, more numerous cases of sequence deviations were observed. Nevertheless, A-boxes were always localized no more than 21 nt away from the 5′ end of the mature products, which fits with a distance of about 25 nt between the A-box and the start of transcription. The shortest A-B distance observed (24 nt) is greater than the minimal distance experimentally determined for the correct binding of TFIIIC (21 nt) (97). In 35 cases over 39, the terminator (poly-T) is followed by A or G, which is indicative of an efficient Pol III termination (15).
In contrast to the high conservation of the A- and B-promoter elements throughout the ten genomes, their locations are highly variable, depending on the gene and on the genome. For example, the RPR1 B-box, which is external to the mature product in the yeasts from S.cerevisiae to K.waltii, becomes internal in C.albicans and Y.lipolytica. This illustrates the adaptability of the Pol III transcription machinery to overcome the additional constraints exerted on an internal B-box at the RNA level. In these ten genomes, cases of dicistronic Pol III genes (98) were searched, but none except the tDNA pairs were found. Preliminary investigations for tDNA pairs in higher eukaryotes also remained unsuccessful, suggesting that this type of organization and the mechanism that maintain species-specific pairs are restricted to yeasts.
Supplementary Data are available at NAR Online.
The authors thank Jean-Luc Souciet (Strasbourg) and all the members of the Génolevures Consortium for stimulating discussions. The authors thank Yves Boulard (Saclay) for help in running tRNAscan-SE. The authors acknowledge Valérie de Crécy-Lagard (University of Florida) for her help in improving the manuscript. The sequencing projects of C.glabrata, K.lactis, D.hansenii and Y.lipolytica were supported by the Consortium National de Recherche en Génomique (to Génoscope and to Institut Pasteur Génopole), the CNRS (GDR 2354, Génolevures sequencing consortium), the Ministère de la Jeunesse, de l'Éducation et de la Recherche (ACI IMPBio n°IMPB114 ‘Génolevures en ligne’) and the ‘Conseil Régional d'Aquitaine’ (‘Génotypage et Génomique Comparée’). The Magnaporthe grisea sequencing project is performed by Ralph Dean, Fungal Genomics Laboratory at North Carolina State University (www.fungalgenomics.ncsu.edu), and Center for Genome Research (www.broad.mit.edu). The Coprinus cinereus and Fusarium graminearum sequencing projects are performed at the Broad Institute and are supported by the National Research Initiative, which is within the U.S. Department of Agriculture's (USDA's) Cooperative State Research Education and Extension Service, and reviewed through the USDA/NSF Microbial Genome Sequencing Project. E.W. and B.D. are members of Institut Universitaire de France. Funding to pay the Open Access publication charges for this article was provided by Commissariat à l'Énergie Atomique.
Conflict of interest statement. None declared.