Findings

The bldA mutants of Streptomyces coelicolor A3(2) were first isolated almost 40 years ago (Merrick 1976) and 11 years later were shown to carry mutations within gene for leucyl \({\text{tRNA}}^{\text{Leu}}{}_{\text{UAA}}\). (Lawlor et al. 1987) This mutation abolishes aerial mycelium formation (Bald phenotype) and antibiotic production by streptomycetes; currently bldA is extensively used as a tool to activate cryptic secondary metabolome (Hackl and Bechthold 2015). Codon TTA, whose decoding is controlled by bldA, is very rare in GC-rich Streptomyces genomes, and present only in genes with unknown and auxiliary functions, such as colony morphological development and antibiotic production. As accumulation of translation-competent, charged \({\text{tRNA}}^{\text{Leu}}{}_{\text{UAA}}\) is confined to late stages of growth, so does the expression of TTA-containing genes (Chater 2006). UUA codon and its cognate tRNA were long time ago suggested to form a genetic switch that operates at the level of translation (Hopwood 1987). Use of TTA for regulatory purposes is somewhat controversial. On one hand, rarity of this codon ensures that only certain genes are influenced by bldA-based switch. On the other hand, rare codons are thought to be associated with higher missense error rates, which would not favor their proper operation as a switch. Although it was suggested implicitly that UUA is decoded accurately, there is a number of notable exceptions. Particularly, bldA mutants showed no Bald phenotype on certain solid media (Hopwood 1987); several TTA-containing genes were expressed in bldA-deficient strains (Trepanier et al. 2002; Makitrynskyy et al. 2013), particularly when their transcription is artificially elevated (Gramajo et al. 1993). All these observations imply that efficient mistranslation of UUA codon is possible at least under some conditions. It is not known what tRNAs could potentially recognize UUA in the absence of \({\text{tRNA}}^{\text{Leu}}{}_{\text{UAA}}\) and what structural and functional peculiarities of \({\text{tRNA}}^{\text{Leu}}{}_{\text{UAA}}\) contribute to its regulatory function (Pettersson and Kirsebom 2011). Here we took bioinformatic approach to obtain new insight into this issue and to chart new directions for experimental verification.

Although there are several databases of tRNA genes, such as GtRNAdb and tRNADB-CE, they provide contradictory information on tRNA content for model strain Streptomyces coelicolor A3(2) and lack data on several species relevant to this work. Furthermore, available online resources do not show what kind of tRNA may decode certain codon via wobble interaction. We therefore compiled all available information on tRNA genes and their decoding capacity for six Streptomyces species with known cases of bldA-based regulation using several databases and search tools detailed in Additional file 1. It could be concluded that overall tRNA gene content is highly conserved in six analyzed Streptomyces genomes, although copy number of individual tRNA genes varies (Table 1). For several codons there were no acceptor tRNAs (for example, alanine codon GCT); those apparently are recognized by isoacceptor tRNAs (e.g. GCT is read by GCC isoacceptor; see Table 1), which are encoded within the analyzed genomes. It is common for all known organisms that the entire set of sense codons (61 + 1 initiator) is read by far fewer than 62 isoacceptors; extreme cases of anticodon-sparing are documented in some archaea and mycoplasmas, where only 26–33 anticodons are required to read the genetic code (Marck and Grosjean 2002). All genomes contain single tRNA gene for UUA decoding. Therefore, differences in bldA mutant phenotypes across different species could not be ascribed to variations in tRNA gene content. Codon UUA could be poorly recognized by phenylalanine \({\text{tRNA}}^{\text{Phe}}{}_{\text{AAA}}\) via wobble interactions (Lim and Curran 2001). However, no respective tRNA gene is present in all studied Streptomyces genomes. Cytidine posttranscriptionally modified with lysidine (k2C34) is known to recognize adenosine in third codon position. This kind of modification to date was described only for anticodon CAU, which normally decodes methionine codon AUG. The k2C34-containing tRNACAU is charged with isoleucine and recognizes isoleucine codon AUA. It is not possible for \({\text{tRNA}}^{\text{k2CAU}}{}_{\text{Ile}}\) to recognize codon UUA because of mismatch in second codon position. Hence, there are no tRNA genes in Streptomyces genomes that would allow UUA codon reading (via correct or wobble interactions) in the absence of cognate \({\text{tRNA}}^{\text{UAA}}{}_{\text{Leu}}\). We therefore looked into possibility of UUA misreading. According to Lim and Curran (2001), three anticodons could misread UUA: UAC, GAA, CAA. Of these, first two would lead to aminoacid misincorporation (Val and Phe, respectively).

Table 1 tRNA genes in six Streptomyces genomes
Full size table

To gain initial insight into relative mistranslation rate associated with bldA, we applied a computational model of translation accuracy (Shah and Gilchrist 2010) that deduces ratio of abundances of cognate to near-cognate tRNAs (differ from cognate one by one mismatch; see Table 2). The rationale is that error rate would depend not only on the abundance of cognate tRNA, but also on the abundance of all near-cognates, that compete with the former for codon recognition. There was statistically significant positive correlation between the abundance of all leucyl tRNAs and their near-cognates for six Streptomyces species (Fig. 1), suggesting that error rates should not differ for different Leu codon-cognate tRNA pairs (if so, abundances of cognates and near-cognates would be uncorrelated). Similar correlation pattern was observed for most Streptomyces tRNAs (Additional file 1: Fig. S1, S2). We further calculated elongation and error rates for all six leucine codons and revealed that UUA had, in fact, the lowest missense error rate (Table 3). Our findings contrast general belief that low-abundant tRNAs are associated with higher mistranslation rates. Yet, they extend the nuanced view of codon accuracy, based originally on non-actinobacterial, low-GC (less than 70 %) genomes (Shah and Gilchrist 2010), onto GC-rich streptomycetes. Our data also agree with the expectation that proper operation of codon-based genetic switch should be based on accurate translation of UUA.

Table 2 Gene copy number (GCN) for tRNA genes in six Streptomyces genomes
Full size table
Fig. 1
figure 1

Correlation between a focal leucyl tRNA’s abundance tF and the abundance of its neighbors tN, across six Streptomyces genomes (see Additional file 1 for details). Each point represents a leucine tRNA species. The solid lines represent the regression lines between tF and tN for each genome. The data are dependent and nonrandom (Wilcox test, 0.042), and positively correlated (Spearman coefficient, min. 0.354). The mean of the distribution of correlation coefficient values for leucine codons differ from 0 (see Additional file 1: Fig. S2)

Full size image
Table 3 Translation and mistranslation rates for six leucine codons in Streptomyces
Full size table

Rather narrow options for UUA mistranslation, revealed by our analysis, did not take into account that decoding properties of tRNAs can be tuned via posttranscriptional modifications. We identified in genomes of two model streptomycetes a large set of genes for such modifications (including k2C; see above), seven of which are involved in maturation of nascent \({\text{tRNA}}^{\text{UAA}}{}_{\text{Leu}}\) in various non-actinomycete bacteria (Table 4 and Additional file 1: Fig. S3). Of particular interest are the genes for modification of anticodon loop and adjacent bases of \({\text{tRNA}}^{\text{UAA}}{}_{\text{Leu}}\) (see Additional file 1). For example, it is possible that a posttranscriptional modification of nascent bldA transcript important for UUA decoding and/or \({\text{tRNA}}^{\text{UAA}}{}_{\text{Leu}}\) maturation is delayed in Streptomyces. It would temporally limit the occurrence of translationally-competent \({\text{tRNA}}^{\text{UAA}}{}_{\text{Leu}}\), thus explaining late expression of TTA-containing genes. If so, then streptomycetes deficient in certain tRNA modification genes would resemble bld mutants. We are currently studying this idea using S. albus and S. ghanaensis as experimental models and invite verification of this conjecture for other strains. As a conclusion, our work shows that there are no theoretical grounds to consider UUA more error prone than the other leucine codons. We examined, in silico, options for UUA mistranslation and draw the attention of researchers to poorly understood aspects of function of bldA genetic switch.

Table 4 Genes for tRNA posttranscriptional modification in S. coelicolor and S. albus genomes
Full size table