Capreomycin susceptibility is increased by TlyA-directed 2′-O-methylation on both ribosomal subunits

Authors

  • Tanakarn Monshupanee,

    Corresponding author
    1. Department of Biochemistry and Molecular Biology, University of Southern Denmark, Campusvej 55, DK-5230 Odense M, Denmark.
    Search for more papers by this author
    • Present address: Department of Biochemistry, Faculty of Science, Chulalongkorn University, Bangkok 10330, Thailand.

  • Shanna K. Johansen,

    1. Department of Biochemistry and Molecular Biology, University of Southern Denmark, Campusvej 55, DK-5230 Odense M, Denmark.
    Search for more papers by this author
  • Albert E. Dahlberg,

    1. Department of Molecular Biology, Cell Biology and Biochemistry, Brown University, Providence, RI 02912, USA.
    Search for more papers by this author
  • Stephen Douthwaite

    Corresponding author
    1. Department of Biochemistry and Molecular Biology, University of Southern Denmark, Campusvej 55, DK-5230 Odense M, Denmark.
    Search for more papers by this author

E-mail tanakarn.m@chula.ac.th; Tel. (+66) 2218 5438; Fax (+66) 2218 5418;

E-mail srd@bmb.sdu.dk; Tel. (+45) 6550 2395; Fax (+45) 6550 2476.

Summary

The binding site of the cyclic peptide antibiotics capreomycin and viomycin is located on the ribosomal subunit interface close to nucleotides C1409 in 16S rRNA and C1920 in 23S rRNA. In Mycobacterium tuberculosis, the 2′-hydroxyls of both nucleotides are methylated by the enzyme TlyA. Loss of these methylations through inactivation of TlyA confers resistance to capreomycin and viomycin. We report here that TlyA orthologues occur in diverse bacteria and fall into two distinct groups. One group, now termed TlyAI, has shorter N- and C-termini and methylates only C1920; the second group (now TlyAII) includes the mycobacterial enzyme, and these longer orthologues methylate at both C1409 and C1920. Ribosomal subunits are the preferred substrates for both groups of orthologues. Amino acid substitutions at the N-terminus of TlyAII reduce its ability to methylate these substrates. Growing pairs of recombinant TlyAIIEscherichia coli strains in competition shows that even subtle changes in the level of rRNA methylation lead to significant differences in susceptibility to sub-inhibitory concentrations of capreomycin. The findings reveal that 2′-O-methyls at both C1409 and C1920 play a role in facilitating the inhibitory effects of capreomycin and viomycin on the bacterial ribosome.

Introduction

Post-transcriptional modifications of rRNAs facilitate ribosomal subunit assembly and conformation changes during protein synthesis (Lapeyre, 2005; Chow et al., 2007; Woodson, 2008; Grosjean, 2009). In bacteria, rRNA modifications are mainly uridine isomerizations and methylations on the nucleobase or on the 2′-hydroxl of the ribose. In Escherichia coli, the most intensely studied bacterium, rRNAs modifications are generally added by enzymes that recognize single specific nucleotide targets (Purta et al., 2009). A subset of rRNA methylations confer resistance to many ribosome-targeting antibiotics (see reviews by Poehlsgaard and Douthwaite, 2005; Conn et al., 2009; Long and Vester, 2009; McCusker and Fujimori, 2012), and are often acquired by bacterial pathogens via an exogenous methyltransferase gene. There are, however, a few cases where drug resistance is conferred by loss of rRNA methylation after inactivation of an intrinsic methyltransferase gene (Helser et al., 1971; Okamoto et al., 2007). This latter type of mechanism has been observed in virulent strains of Mycobacterium tuberculosis where inactivation of the tlyA methyltransferase gene confers resistance to capreomycin and viomycin (Maus et al., 2005a; Johansen et al., 2006).

Capreomycin and viomycin are cyclic-peptide antimicrobial compounds, and capreomycin is particularly important in the treatment of tuberculosis caused by strains of M. tuberculosis that are resistant to first-line drugs (Zhang and Yew, 2009; Almeida Da Silva and Palomino, 2011). Capreomycin and viomycin both bind to the bacterial ribosome where interbridge B2a extends across the subunit interface and links the decoding region of 16S rRNA helix 44 with the highly conserved helix 69 of 23S rRNA (Johansen et al., 2006; Stanley et al., 2010). Within these helices, nucleotides C1409 of 16S rRNA and C1920 of 23S rRNA (E. coli rRNAs numbering throughout) are methylated at their 2′-hydroxyls by the M. tuberculosis methyltransferase TlyA (Johansen et al., 2006). Nucleotides C1409 and C1920 are brought into proximity upon association of the ribosomal subunits and formation of interbridge B2a (Fig. 1) (Yusupov et al., 2001; Selmer et al., 2006) and failure to methylate these nucleotides due to loss of TlyA activity results in resistance to capreomycin and viomycin (Maus et al., 2005a; Johansen et al., 2006).

Figure 1.

Capreomycin and viomycin binding site on 70S ribosomes showing nucleotides C1409 and C1920 and methylation by TlyA orthologues. A. Ribosome with capreomycin (green) bound showing interbridge B2a formed between 16S rRNA helix 44 and 23S rRNA helix 69 (boxed). The sites of methylation at nucleotides C1409 and C1920 are highlighted (red) within the box. This region is shown relative to the positions of the A-, P- and E-site tRNAs (purple); backbones of 16S rRNA (yellow), 30S r-proteins (orange), 23S rRNA (grey) and 50S r-proteins (blue) are shown. The ribosome is shown in the classical state (Stanley et al., 2010; Dunkle et al., 2011). B. Enlargement of the boxed region. C. Ribose targets for TlyAI and TlyAII methylation (red) relative to the binding site of capreomycin (drug carbons, green; oxygens, red; nitrogens, blue). In the classical state T. thermophilus (Stanley et al., 2010) and E. coli (Dunkle et al., 2011) ribosome structures, the distance between the 2′-O positions of C1409 and C1920 is approximately 21 Å, and this distance remains virtually unchanged in the intermediate translocation (hybrid)-state of the ribosome (Dunkle et al., 2011). D. Same view with viomycin bound. Drug binding structures redrawn from the X-ray crystal PDB files 3KNH through 3KNO (Stanley et al., 2010).

Orthologues of the mycobacterial TlyA are found in numerous unrelated bacteria including Thermus thermophilus (Monshupanee et al., 2008). The T. thermophilus enzyme was shown to 2′-O-methylate the 23S rRNA nucleotide C1920 (Mengel-Jorgensen et al., 2006) without concomitant methylation at 16S rRNA nucleotide C1409 (Guymon et al., 2006), which indicated that the two methylation activities of TlyA could be differentiated (Monshupanee et al., 2008). Here, we investigate how the mycobacterial TlyA recognizes two distinct methylation targets on separate ribosomal subunits, and we evaluate the relative importance of the two methylations for inhibition by cyclic peptide drugs. The database sequences of TlyA show that the lengths of the N- and C-termini differ such that orthologues can be divided into two groups. Recombinant versions of TlyA orthologues from both groups were constructed and expressed in E. coli to determine their substrate specificities. Structural features of the orthologues that could be linked to substrate specificity were then altered by site-directed mutagenesis. Strains harbouring the mutant variants of TlyA were grown in competition in the presence of drug, and their growth was followed using a quantitative fluorescence PCR assay. The results demonstrate that even subtle differences in rRNA methylation levels produce measurable changes in susceptibility to capreomycin.

Results

Comparative analysis of TlyA sequences

Protein blast searches using M. tuberculosis TlyA as the query against 2011 bacterial genome sequences revealed 852 genomes that contain an open reading frame with better than 40% amino acid identity, and these were classed as TlyA orthologues. Sequences outside the Bacteria were observed with significant similarity to TlyA, although they generally fell below the threshold for inclusion as orthologues. These non-bacterial sequences comprised three insects (31–43% identity), 11 plants (36–41%), five algae (34–40%) and two Archaea (29% and 31%). Domains indicative of an RNA binding region, an S-adenosylmethionine (SAM) binding site (Martin and McMillan, 2002) and 2′-O-methyltransferase function (Aravind and Koonin, 1999; Feder et al., 2003; Maus et al., 2005a) are particularly well conserved in the orthologue sequences (Fig. 2).

Figure 2.

Comparison of TlyA sequences. The sequence of TlyA from M. tuberculosis H37Rv (NCBI accession number: CAB10951) is compared with orthologues from M. smegmatis MC2 155 (NCBI number YP_888048.1), S. coelicolor A3 (NC_003888), G. stearothermophilus (AB126617), T. thermophilus HB8 (AP008226) and S. hyodysenteriae (X61684). The total numbers of residues in the enzymes together with the percentages that are identical and functionally similar to the M. tuberculosis sequence are listed on the right; these conserved regions are shown as black bars within the sequences. Based on their lengths, the sequences fall into the two distinct groups TlyAI and TlyAII. A consensus from these orthologues is shown highlighting the conserved S4-type RNA binding motif (S4), the S-adenosylmethionine (SAM) binding domain and three motifs indicative of 2′-O-methyltransferases (K, D and K) (Aravind and Koonin, 1999; Feder et al., 2003; Maus et al., 2005b).

Alignment of bacterial TlyA orthologues showed that they could be differentiated into two distinct groups on the basis of the lengths of their N- and C-termini (Fig. 2). The first group, now designated TlyAI, has the shorter termini. The second group, TlyAII, consists of enzymes that are at least four residues longer at the N-terminus with a highly conserved arginine or lysine at position 4; the C-termini of TlyAII proteins are up to fifty residues longer, but lack any obvious conservation.

TlyA orthologues methylate single or dual targets

The rRNA methyltransferase activities of five bacterial TlyAI and TlyAII orthologues (Fig. 2) were tested in vivo in recombinant strains of E. coli, a species that has no natural tlyA orthologue of its own. Methylation of rRNAs was analysed by primer extension and matrix assisted laser desorption/ionization mass spectrometry (MALDI-MS). Recombinant TlyAI enzymes from T. thermophilus and Serpulina hyodysenteriae methylated 23S rRNA nucleotide C1920, but not nucleotide C1409 in 16S rRNA; TlyAII enzymes derived from Mycobacterium smegmatis, Streptomyces coelicolor and Geobacillus stearothermophilus methylated at both C1409 and C1920 (Fig. 3). The minimum inhibitory concentration (MIC) measurements showed that the susceptibility of the E. coli recombinants to capreomycin and viomycin increased between twofold and eightfold upon expression of the different tlyA orthologues (Table 1). We noted considerable variation in the sequences, codon usage and expression levels of these orthologues, and concluded that a more uniform system was needed to ascertain whether tlyAI and tlyAII had significantly different effects on drug susceptibility.

Figure 3.

Gel autoradiographs of primer extensions showing 2′-O-methylations on E. coli rRNAs by recombinant TlyA orthologues. rRNAs were isolated from E. coli strains that contained the empty plasmid, pLJ102 (no tlyA), or a plasmid expressing recombinant tlyA from M. smegmatis (M. smeg), T. thermophilus (T. therm), S. coelicolor (S. coel), S. hyodysenteriae (S. hyod) and G. stearothermophilus (G. stear). Decreasing dGTP concentration (wedges) intensifies reverse transcription termination at 2′-O-methylated C1409 in 16S rRNA (upper panel) and C1920 in 23S rRNA (lower panel). Run-through transcription terminated at G1405 and A1919 upon incorporation of ddCTP and ddTTP respectively. Lanes C, U, A and G are dideoxy-sequencing reactions on unmodified E. coli rRNAs. Two additional, independent repetitions of the experiment yielded identical results.

Table 1.  Antibiotic susceptibilities of E. coli expressing recombinant tlyA orthologues.
E. coli with plasmid-encoded tlyA orthologuesMIC (µg ml−1)
CapreomycinViomycin
  1. Each MIC (minimum inhibitory concentration) represents a minimum of three measurements, and the values were consistently the same under the conditions used.

Empty plasmid, no tlyA128128
T. thermophilus tlyAI 3232
S. hyodysenteriae tlyAI 3232
M. smegmatis tlyAII 3216
S. coelicolor tlyAII 6464
G. stearothermophilus tlyAII 6432

Ribosomal substrates of TlyA methyltransferases

The methylation substrates for TlyAI and TlyAII were identified in vitro using purified recombinant enzymes with a range of ribosomal substrates (16S and 23S rRNAs, 30S and 50S subunits and 70S ribosomes) isolated from E. coli and from the ΔtlyA strains of M. smegmatis and T. thermophilus. In the in vitro assays, TlyAI from T. thermophilus efficiently methylated C1920 in 50S subunits and, consistent with the results obtained in vivo, it failed to methylate C1409 in any of the substrates containing 16S rRNA (Fig. S1A, B and C). The 50S subunit was the preferred substrate for TlyAI, although a modest amount of methylation was seen with free 23S rRNA and 70S ribosomes.

The M. smegmatis TlyAII enzyme methylated efficiently at C1920 in 50S subunits in vitro, although weaker activity was also detected on 70S ribosome and 23S rRNA substrates (Fig. S1F). The M. smegmatis TlyAII modified C1409 in E. coli 30S subunits, and activity at this site on 16S rRNA and 70S ribosomes was negligible (Fig. S1D). Surprisingly, the M. smegmatis enzyme displayed only weak activity on its endogenous 30S subunits derived from the M. smegmatis ΔtlyA strain (Fig. S1E).

Residues essential for target recognition

The contribution of the N- and C-termini towards recognition of the two rRNA targets was investigated by mutagenizing the M. smegmatis TlyAII sequence. Deletion of four N-terminal residues (ΔA2-R3-R4-A5) abolished methylation at 16S rRNA nucleotide C1409, while methylation at 23S rRNA C1920 was retained, albeit with reduced efficiency (Fig. 4A). The same effect was obtained by substituting both Arg-3 and Arg-4 for alanines (R3A-R4A), however, methylation was unaltered when both of these basic residues were replaced by similarly charged lysines (R3K-R4K). The more important of the two arginines for 30S subunit recognition appears to be R4, as the single R4A substitution causes complete loss of C1409 methylation. Defective methylation by this latter mutant was partially rescued by substitution of R3 for lysine (R3K-R4A) (Fig. 4B and C), while substitution of R3 on its own (R3A) did not affect methylation. Deletion of the 27 C-terminal residues in TlyAII abolished methylation activity at both C1409 and C1920 (Fig. 4A), irrespective of whether the recombinant protein was constructed with or without a C-terminal His-tag.

Figure 4.

Methylation of rRNA by TlyA mutant enzymes in vivo. A. Mutations in the N- and C-terminals of M. smegmatis TlyAII that affect the enzyme's ability to methylate E. coli 16S and 23S rRNAs in vivo at nucleotides C1409 and C1920 (++, full; +, partial; −, no methylation). Methylation levels were measured by both MALDI-MS and primer extension (Fig. S2). Altered susceptibility of E. coli to capreomycin (Cap) is shown as MICs (µg ml−1); measurements in triplicate with identical results. Δ, deleted amino acid residue. B. MALDI-MS analyses of 16S rRNA nucleotides 1377–1435 after RNase T1 digestion; rRNA from cells expressing wild-type TlyAII from pSJ101; boxes show the enlarged spectral region with nucleotide C1409 expressed from cells with different TlyA variants. C. Corresponding MALDI-TOF MS analyses of 23S rRNA nucleotides 1889–1948 showing the enlarged peaks containing C1920. The measured m/z-values (above the peaks) match the theoretical values (boxed) to within 0.1 Da.

In another set of recombinants, we extended the N- and C-termini of the T. thermophilus TlyAI enzyme in an attempt to expand its substrate range to include C1409. Recombinant versions of T. thermophilus TlyAI were constructed with the N- and/or C-termini from the M.  smegmatis, S. coelicolor and G. stearothermophilus TlyAII sequences. The T. thermophilus TlyAI recombinants retained their ability to methylate at C1920 although none of them became capable of methylating C1409 (data not shown).

Assessing the effects of rRNA methylation on capreomycin susceptibility

The varying degrees of modification at nucleotides C1409 and C1920 added by the different TlyAII mutants caused a fourfold variation in drug sensitivity with MICs of 32 to 128 µg ml−1 (Fig. 4A). These MIC values were highly reproducible, but they were nevertheless insufficiently differentiated for the evaluation of subtle changes in drug sensitivity. We therefore opted for a growth competition assay over a large number of cell divisions, during which any difference in drug susceptibility would become apparent. For growth competition, pairs of E. coli strains expressing variants of the M. smegmatis tlyAII gene were grown together in the same medium for over 50 generations. Strains were identical except for substitutions of one or two residue at the N-terminal part of TlyAII. The substitutions affected rRNA methylation, which in turn was seen to influence growth rates in capreomycin even at concentrations well below the MICs. The proportion of each strain in the population was followed during the growth cycles using a fluorescence PCR assay that used the sequence differences in the tlyAII genes to distinguish the strains.

The E. coli strain harbouring wild-type TlyAII was rapidly out-competed by the R3K/R4A-TlyAII strain (Fig. 5), despite having been indistinguishable on the basis of MIC measurements (both 32 µg ml−1). At a sub-inhibitory level of capreomycin (6 µg ml−1) the proportion of cells with the wild-type TlyAII gradually declined and disappeared from the total population after two growth cycles corresponding to about 22 generations. Increasing the capreomycin concentration to 12 µg ml−1 accelerated the selection process and cells with wild-type TlyAII were out-competed already by the 11th generation of growth. When competing against strains such as R4A-TlyAII that are less effective at methylating their rRNAs, the strain with wild-type TlyAII was even more rapidly outgrown at sub-inhibitory drug levels (Fig. 5). In growth experiments with capreomycin, where both competing strains showed less than stoichiometric rRNA methylation patterns (shown here for the R4A versus the R3K/R4A mutant), the strain with the lower level of methylation invariably came to dominate the culture.

Figure 5.

Cell growth competition assay. Pairs of E. coli strains with TlyA variants, differing in the degree of 2′-O-methylation at C1409 and C1920, were grown together through five cycles (with approximately 11 generations per cycle) in the presence of sub-inhibitory concentrations of capreomycin. The proportion of each strain was determined by PCR with fluorescently labelled primers. Left column: wild-type strain expressing TlyAII from plasmid pSJ101 (A) grown in competition with the R3K/R4A-TlyAII strain (B, pTM113); the upstream PCR primers complementary to these tlyA sequences were respectively labelled with Cy3 (fluorescing green after excitation at 512 nm) and Cy5 (fluorescing red after excitation at 625 nm). Centre column: wild-type TlyAII strain grown in competition with the R4A-TlyAII strain (C, pTM106) and analysed with Cy3 and Cy5 primers respectively. Right column: R4A- and R3K/R4A-TlyAII strains grown in competition (note Cy3 is used here for the R4A strain). Upper row, no drugs; middle row, competition in 6 µg capreomycin ml−1; lower row, competition in 12 µg capreomycin ml−1. The superimpositions of the 170 bp PCR products appear yellow when the amounts of strains are approximately equal, and the red or green colour becomes predominant when one strain out-competes the other. Each data set is representative of a minimum of three growth assays.

In the absence of capreomycin, all E. coli tlyAII recombinant strains grew identically over 55 generations (Fig. 5), and extended culturing to over 100 generations revealed no growth advantage for any strain in the absence of drug (not shown).

Discussion

Capreomycin and viomycin inhibit protein synthesis on the bacterial ribosome by interacting across the interface of the ribosomal subunits (Yamada et al., 1978; Johansen et al., 2006; Stanley et al., 2010) and blocking the translocation of tRNA (Modolell and Vazquez, 1977). The crystal structure places the drugs between the A- and P-sites with the ribosome in the classical state for tRNA binding (Stanley et al., 2010), while functional studies indicate that the drugs are at the same site but with the ribosome in the pre-translocational hybrid state (Johansen et al., 2006; Ermolenko et al., 2007). Methylation of the rRNAs in both subunits by TlyA facilitates inhibition by capreomycin and viomycin, and this could be linked with the relative movement of the subunits between the different states during translocation. However, no adverse effect was seen on growth of the parent strains after inactivation of TlyA methylation (Maus et al., 2005a; Monshupanee et al., 2008). Furthermore, in the absence of drug, the competition growth assay revealed neither growth advantage nor biological cost in connection with TlyA-directed methylations (Fig. 5, upper row). Viewed together with the occurrence of TlyA orthologues in over 40% of bacterial species, it seems likely that the methylations confer some advantage that has not yet become apparent under laboratory growth conditions.

Here, we show that TlyA orthologues segregate into two groups. The first group TlyAI is exemplified by the T. thermophilus orthologue and 2′-O-methylates nucleotide C1920 in 23S rRNA (Monshupanee et al., 2008), while the second group TlyAII includes the mycobacterial orthologues, and these enzymes methylate C1409 in 16S rRNA in addition to nucleotide C1920 (Johansen et al., 2006). The differences in substrate recognition and modification by TlyAI and TlyAII correlate with distinctive variations in the lengths of their N- and C-termini. The N-terminal domain corresponds to an S4-type RNA-binding motif commonly found in bacterial and eukaryotic RNA methyltransferases, pseudouridine synthases and ribosomal proteins (Aravind and Koonin, 1999; Arenas et al., 2011). In TlyAII this domain is preceded by several residues that have been modelled in silico as a finger-like projection with the conserved positively charged Lys/Arg at positions 3 and 4 on the surface (Arenas et al., 2011), and we show here that residue R4 plays an important role in broadening the range of RNA interaction to include 16S rRNA nucleotide C1409. However, this sequence on its own is not enough to direct the enzyme to nucleotide C1409 and, after extending the N-termini of TlyAI enzymes with the corresponding TlyAII sequences, recombinants still modified C1920 but without gaining the ability to methylate at C1409. Similar results were found for the C-termini of the enzymes. The longer C-terminus of TlyAII is clearly important for function (pTM103, Fig. 4), but its addition to TlyAI did not change enzyme specificity. The N- and C-terminal extensions in TlyAII appear to have evolved to be structurally and functionally integrated with the enzyme core, and thus merely grafting these sequences onto TlyAI does not extend its function.

Nucleotides C1409 and C1920 are located at the interface sides of their respective 30S (Schlünzen et al., 2000; Wimberly et al., 2000; Demirci et al., 2010) and 50S subunits (Schlünzen et al., 2003; Tu et al., 2005). Subunit association to form 70S ribosomes (Yusupov et al., 2001; Selmer et al., 2006; Bulkley et al., 2010; Dunkle et al., 2011) would prevent molecules as large as the TlyA methyltransferases from gaining access to nucleotides C1409 and C1920 (Fig. S3). Genetic evidence from a M. tuberculosis mutant with a 23S rRNA mutation (ΔA1916), which prevented methylation at C1920 but not at C1409, suggested that TlyA recognizes the two sites individually and presumably prior to subunit association (Johansen et al., 2006). This idea was verified here by in vitro methylation assays showing that both TlyAI and TlyAII prefer unassociated subunits as their substrates. The weak modification seen at C1920 with 70S ribosomes is consistent with partial dissociation to separate subunits during the assay. For C1409 methylation by TlyAII, the reaction kinetics is slower than at C1920, and no significant methylation was seen at C1409 when 70S ribosomes were used as the substrate. A moderate amount of modification was noted with free 23S rRNA, and C1920 is clearly accessibility in this substrate but presumably in a less than optimal conformation. It was noted that modification in vitro by the mycobacterial TlyAII on its own 30S subunits was relatively poor compared to E. coli 30S subunits, and contrasts with the stoichiometric modification observed in vivo (Johansen et al., 2006). This might suggest that modification at C1409 in mycobacteria occurs prior to completion of 30S subunit assembly.

Inhibition by capreomycin and viomycin is affected in a similar manner by the recombinant TlyA orthologues with MICs changing by up to eightfold (Table 1). Such modest changes in MIC are sufficient to cause clinical resistance in virulent strains of M. tuberculosis that have lost tlyA function (Maus et al., 2005a,b). Despite this, quantifying the relative effects on drug sensitivity of single and dual methylations by TlyAI and TlyAII proved difficult by MIC determination alone. The orthologues differ greatly in their gene and protein sequences (Fig. 2) and thus lacked uniformity in expression and enzymatic activities in the E. coli host. This obstacle was overcome by creating a series of strains that possessed different degrees of methylation at C1409 and C1920, and were genetically identical except for substitutions at N-terminal residues in recombinant M. smegmatis TlyAII.

We conclude from the data on these TlyAII strains that increases in the level of methylation at both the target nucleotides C1409 and C1920 result in heightened sensitivity to the cyclic peptide drugs. The importance of the C1409 is revealed by the strain with partial C1409/full C1920 methylation which, in capreomycin, out-competes the wild-type strain with full C1409/full C1920 methylation (Fig. 5). The role of C1920 methylation in increasing capreomycin sensitivity is established by the strains expressing TlyAI, where there is full C1920 methylation but no methylation at C1409 (Table 1).

The growth competition assay provides a means of detecting relatively small amounts of compounds with the same inhibitory mechanism as the cyclic peptides capreomycin and viomycin. We note that the assay should in principle be capable of detecting other compounds such as thermorubin, which also binds across the subunit interface in the region of interbridge B2a but does so through direct interaction with the 2′-hydroxyl of C1409 (Bulkley et al., 2012). Methylation by TlyAII would thus be expected to confer resistance to thermorubin and, as a consequence of this, thermorubin-type and capreomycin-type compounds would have inverse effects upon competing pairs of TlyA strains.

Experimental procedures

Comparative analysis of TlyA sequences

Protein blast searches were carried out in the NCBI Genomic Blast database (http://www.ncbi.nlm.nih.gov/sutils/genom_table.cgi) using the M. tuberculosis TlyA as the query against the data available in April 2012. The cut-off was set at 40% identity, which has been reported to give clear discrimination between different protein families and domains (Gandhimathi et al., 2012). Data were processed using Cluster W (Larkin et al., 2007).

Bacterial strains, growth conditions and MIC determination

Escherichia coli strain DH1 was grown at 37°C in Lauria–Bertani (LB) medium containing ampicillin at 100 µg ml−1 to maintain plasmids, and IPTG at 1 mM for full induction of the lac promoter (Sambrook and Russell, 2001). MICs of capreomycin and viomycin were determined as previously described (Johansen et al., 2006). Briefly, overnight cultures of E. coli were diluted to 10−5-fold and plated on to LB agar containing capreomycin or viomycin with the concentrations increasing in twofold steps; measurements were made a minimum of three times. The MICs were defined as the lowest concentration at which no growth was observed after incubation for 48 h. The tlyA null-mutants M. smegmatis strain P2U (EZ::TN-kan tlyA nt.163) (Maus et al., 2005a) and T. thermophilus HB8 strain TM469 (ΔtlyA::htk1) (Monshupanee et al., 2008) were grown in Middle-brook 7 H9 medium (Remel and Difco) containing 0.2% (v/v) glycerol at 37°C and in ATCC medium 1598 (Alfredsson et al., 1985) at 72°C respectively.

Cloning and mutagenesis of tlyA

Orthologues of tlyA were amplified from the genomic DNA of M. smegmatis ATCC 700084, T. thermophilus HB8, S. coelicolor A3, Serpulina (Brachyspira) hyodysenteriae and G. stearothermophilus using complementary PCR primers (Table S1). Specific mutations were introduced into M. smegmatis tlyAII by PCR using mutagenic primers, and tlyA genes were inserted into the NdeI and BglII sites in plasmids pSJ101 or pLJ102 bringing them under control of the lac promoter (Johansen et al., 2006). Attention was paid to codon usage at the N-terminus that could affect expression (Gonzalez de Valdivia and Isaksson, 2004), such that the third codon in the wild-type M. smegmatis tlyAII sequence (CGG for arginine) was suitably matched in growth competition assays. E. coli strain DH1 was transformed with the recombinant plasmids (Table S2).

Protein expression and purification

Escherichia coli cells were grown to an OD600 of 0.6 and TlyA expression was induced by adding IPTG to 1 mM. The cultures were grown for a further 4 h before harvesting the cells by centrifugation, cell pellets were resuspended and washing in 20 mM Tris-HCl, pH 7.5, 10 mM MgOAc2, 100 mM NH4Cl and then lysed by sonication (Hansen et al., 2001). The TlyA proteins possessed a C-terminal histidine-tag were purified using Ni-NTA affinity agarose (Qiagen). The proteins were dialysed in 20 mM Tris-HCl, pH 7.5, 10 mM MgOAc2, 250 mM NH4Cl, 6 mM 2-mercaptoethanol, 10% glycerol, using Amicon Ultra-4 Centrifugal Filter Devices (Milipore), and were analysed by SDS gel electrophoresis.

Preparation of ribosomes and rRNAs

Escherichia coli, M. smegmatis and T. thermophilus cells were harvested at an OD600 of 0.6 for preparation of ribosomal particles (Johansen et al., 2006; Monshupanee et al., 2008). Tight couple ribosomes, 50S and 30S subunits were isolated by sucrose gradient centrifugation, and fractions were pelleted by ultracentrifugation (Monshupanee et al., 2008). 16S and 23S rRNAs were isolated from their respective subunits by phenol extraction (Hansen et al., 2001).

In vitro methylation

In vitro methylation assays were carried out on 10 pmol substrate in 100 µl of 20 mM Tris-HCl, pH 7.5, 10 mM MgCl2, 100 mM NH4Cl, 6 mM 2-mercaptoethanol and 10% glycerol. RNA substrates were pre-incubated at 50°C for 10 min and cooled to room temperature. Reactions were started by addition of 0.1 to 10 pmol recombinant TlyA enzyme and SAM to a final concentration of 1 mM. Reactions on E. coli and M. smegmatis ribosomal substrates were at 37°C for 45 min, whereas T. thermophilus substrates were incubated at 65°C. Reactions were terminated by extraction with phenol/chloroform; rRNAs were recovered by ethanol precipitation prior to analysis by primer extension and mass spectrometry.

Primer extension

DNA primers were 5′-labelled by polynucleotide kinase using 32P-γ-ATP (New England Biolabs) and were hybridized to complementary rRNA regions (Stern et al., 1988). The 2′-O-ribose methylation on 16S rRNA nucleotide 1409 and 23S rRNA nucleotide 1920 were analysed by varying dGTP concentrations (1 µM, 0.5 µM and 0.1 µM) in the presence of 40 µM (dATP, dCTP and ddTTP) or 40 µM (dATP, ddCTP and dTTP) and extending with AMV reverse transcriptase (Finnzymes) (Maden et al., 1995). Extension products were run on polyacrylamide-urea gels and visualized by phosphorimaging (Typhoon, Amersham Biosciences). Measuring 2′-O-methylation in this manner is non-quantitative, and therefore methylation levels were simply tabulated as ‘++’ (full), ‘+’ (partial) or ‘−’ (no methylation) after standardizing against rRNAs that had previously been defined as being fully methylated or without methylation (Johansen et al., 2006).

MALDI-TOF mass spectrometry

The fragments consisting of 16S rRNA nucleotides 1377 to 1435 and 23S rRNA nucleotides 1889 to 1948 were isolated by hybridization to complementary oligodeoxynucleotides (Table S1) (Johansen et al., 2006). The isolated RNA sequences were digested with RNase T1 and cyclic phosphates were hydrolysed with HCl. The RNA digested products were analysed on a PerSeptive Voyager-DE STR MALDI mass spectrometer (Applied Biosystems) in positive ion mode (Kirpekar et al., 2000; Andersen et al., 2004). Spectra were calibrated using m/z software (Proteometrics).

Competition growth assay and florescence PCR

Approximately equal numbers of two E. coli strains harbouring different tlyA variants were mixed in LB medium with ampicillin, IPTG and capreomycin at 0 to 32 µg ml−1. After 24 h incubation (approximately 11 generations), the stationary phase cells were diluted 2000-fold and the cycle was repeated. Aliquots of cells were harvested by centrifugation at the end of each cycle; the cell pellets were resuspended in 10 vols of deionized water, boiled for 2 min, and centrifuged. The mixture of plasmid-encoded tlyA genes in the supernatant was assessed by Taq DNA Polymerase (Promega) PCR using one non-labelled reverse primer and two forward primers that were 5′-fluorescence-labelled with either cyanine 3 (Cy3, green) or cyanine 5 (Cy5, red) and complementary to a specific tlyA variant (Table S1). Reactions were carried out such that the amount of PCR product was proportional to amount of the original template: 50 µl PCR reactions with 2 µl of the supernatant template, 1.25 U Taq DNA polymerase (goTaq, Promega), 2.5 nmol of each dNTP, 50 pmol of each primer. PCR was over 20–22 cycles with denaturation at 98°C for 2 min in the initial step and subsequently for 15 s; annealing at 65°C for 30 s; and extension at 72°C for 20 s. The PCR products of 170 base pairs were run on 3% agarose gels and were detected using an Amersham Biosciences Typhoon calibrated for Cy3 at 512 nm and Cy5 at 625 nm. The two fluorescence scans were superimposed using the Image Quant (Amersham Biosciences) program to reveal the relative proportions of tlyA variants in each sample.

Acknowledgements

We thank Märit Pringle for providing S. hyodysenteriae DNA, Gilles van Wezel for providing the S. coelicolor strain and Claus Asger Lykkebo for the graphics in Figs 1 and S3. We gratefully acknowledge support from the Danish Research Agency (FNU-rammebevillinger 09-064292/10-084554) (S. D.), the Nucleic Acid Center of the Danish Grundforskningsfond (S. D.), the US National Institutes of Health (GM19756) (A. D.) and the A1B1 project of the Faculty of Science, Chulalongkorn University (T. M.). This manuscript is dedicated to the memory of a dear friend and colleague, the late Shanna K. Johansen.

Ancillary