Genome-wide dynamics of a bacterial response to antibiotics that target the cell envelope

Exposure of S. coelicolor to antibiotics that inhibit different stages of peptidoglycan biosynthesis induced changes in transcription in up to 27% of the genome

Transcriptome analysis identified 2094 genes whose expression was significantly different at the 1% probability level, and more than 1.5-fold up- or down-regulated, as a result of exposure of exponentially growing S. coelicolor cells to vancomycin, bacitracin or moenomycin A (Figure 1, Additional files 1, 2, 3 and 4). This corresponds to 27% of the 7,825 genes predicted to make up the genome [1], and includes 1162 genes that were repressed at one or more time points after antibiotic addition, and 1062 genes whose expression was up-regulated. 130 genes were common to both the repressed and induced lists. For each antibiotic treatment, the majority of the genes identified as induced or repressed were present in at least two of the three time points analysed (Figure 1A and 1C). Strikingly, only a minority (0.06%) of the 1147 genes (68/1147) changing in expression in response to bacitracin treatment responded uniquely to this compound; the vast majority were similarly affected by vancomycin or by all three drugs (see Figure 1B). In contrast, 42% (454/1079) of the genes significantly down-regulated following addition of vancomycin responded only to this compound, and 31% (235/744) were also uniquely induced. Moenomycin treatment notably down-regulated the expression of far fewer genes than either vancomycin or bacitracin, and 16% of these (30/184) responded uniquely. Moenomycin however up-regulated a similar number of genes to either vancomycin or bacitracin, approximately 60% (364/616) of which were also induced by one or more of the other antibiotics, while 40% (252/616) were uniquely induced. Interestingly, a core set of 243 genes were up-regulated in response to any one of the compounds and 118 were similarly down-regulated, representing a common response to the antibiotic treatments.

Common changes in expression of known stress-response genes following antibiotic treatment suggest a general environmental stress response system in S. coelicolor

i) The σ^E cell envelope stress response system is induced by all three drugs

σ^E is an extracytoplasmic function sigma factor required for normal cell envelope integrity in S. coelicolor, and its expression is induced by a wide-variety of cell wall damaging agents [19]. Reassuringly, transcription of the four-gene operon containing the σ^E gene was significantly induced following treatment with each of the antibiotics (Table 1, Figure 2). In addition, similarity analysis indicated that the expression of 158 genes was closely correlated with that of σ^E (Pearson correlation >0.9; Additional file 13). This group is likely to include genes directly dependent on σ^E for their transcription, and indeed contains σ^hrdD (SCO3202) previously characterised to be dependent on σ^E for expression [20]. Interestingly, expression of the only aminoacyl-tRNA synthetase gene (SC03397) to be up-regulated in response to any drug correlates with σ^E transcription (similarity coefficient = 0.93); of the remaining 27 tRNA synthetase genes annotated in the genome sequence, transcription of 21 was significantly repressed following antibiotic treatment. Mutation in SCO3397, encoding a lysyl-tRNA synthetase, markedly reduced the MIC towards both vancomycin and bacitracin (Table 3) indicating that it plays an important role in resistance to these compounds.

Strains	Relevant genotype/comments	MIC (μg/ml)			Source/reference
		Vancomycin	Bacitracin	Moenomycin
S. coelicolor A3(2) (M600)	SCP1- SCP2-	~80	40-45	>300	Kieser et al. 2000 [81]
ΔSCO0600 (H1001)	ΔsigB::apr SCP1- SCP2-	60-70	25-30	>300	This study
ΔSCO3356 (J2130)	ΔsigE SCP1- SCP2-	60-70	35-40	>300	Paget et al. 1999 [20]
ΔSCO1513 (M570)	ΔrelA SCP1- SCP2-	60-70	20-25	>300	Chakraburtty and Bibb. 1997 [23]
ΔSCO3089-3090 (H1003)	ΔSCO3089-3090::apr SCP1- SCP2-	70-80	15-20	>300	This study
ΔSCO3110-3111 (H1002)	ΔSCO3110-3111::apr SCP1- SCP2-	70-80	30-35	>300	This study
ΔSCO3089-3090 + ΔSCO3110-3111 (H1004)	ΔSCO3089-3090 + 3110-3111::apr SCP1- SCP2-	70-80	<10	>300	This study
SCO4005::Tn (M900)	SCO4005::Tn SCP1- SCP2-	70-80	25-30	>300	Hesketh et al. 2007 [21]
SCO5147::Tn (M903)	SCO5147::Tn SCP1- SCP2-	80-90	40-45	>300	Hesketh et al. 2007 [21]
ΔSCO1875 (H1005)	ΔSCO1875::apr SCP1- SCP2-	<50	10-15	>300	This study
ΔSCO2608 (H1006)	ΔSCO2608::apr SCP1- SCP2-	60-70	40-45	>300	This study
SCO2897::Tn (H1007)	ΔSCO2897::Tn SCP1- SCP2-	~60	30-35	>300	This study
SCO3580::Tn (H1008)	ΔSCO3580::Tn SCP1- SCP2-	<50	40-45	>300	This study
SCO3771::Tn (H1009)	ΔSCO3771::Tn SCP1- SCP2-	<50	40-45	>300	This study
ΔSCO3847 (H1010)	ΔSCO3847::apr SCP1- SCP2-	~60	40-45	>300	This study
SCO3901::Tn (H1011)	ΔSCO3901::Tn SCP1- SCP2-	~60	40-45	>300	This study
ΔSCO4013 (H1012)	ΔSCO4013::apr SCP1- SCP2-	70-80	40-45	>300	This study
SCO5039::Tn (H1013)	ΔSCO5039::Tn SCP1- SCP2-	60-70	40-45	>300	This study
ΔSCO3397 (J3355)	ΔSCO3397::apr SCP1- SCP2-	<50	10-15	>300	This study

Changes in expression of genes with functions related to production and maintenance of the cell wall show antibiotic-specificity

i) Genes required for cell wall peptidoglycan biosynthesis

Bacterial cell wall peptidoglycan biosynthesis can be separated into reactions occurring intracellularly to generate the lipid II precursor subunits, and those taking place extracellularly to join the subunits together into a three dimensional structure [35–42] (Figure 3A). Transcription of many of the genes for the cytoplasmic steps required for precursor biosynthesis, particularly the mur operon (SCO2083-2089) and the femX gene (SCO3904), were significantly repressed following treatment with either bacitracin or vancomycin, but were unchanged following moenomycin addition (Table 2, Figure 3B). In contrast, genes encoding penicillin-binding proteins (PBPs) required for the extracellular reactions in peptidoglycan production were over-represented in the genes induced by all three antibiotics (Table 1, Figure 3B). The secreted PBPs SCO1875, SCO2897 and SCO5039 were significantly induced following treatment with each compound, while SCO2608 was induced by vancomycin and moenomycin but not by bacitracin. In MIC tests, deletion of SCO1875 markedly increased susceptibility to both bacitracin and vancomycin (Table 3) indicating that this PBP is essential for optimum resistance to these compounds, and suggesting that up-regulation of its transcription is an important response to the antibiotic treatments. Deletion of SCO2897 also increased susceptibility to both antibiotics (lowered the MIC), while mutation in SCO2608 and SCO5039 increased sensitivity to vancomycin only (Table 3). Vancomycin treatment alone down-regulated expression of SCO2090, SCO3901 and SCO4013, and was the only antibiotic to up-regulate transcription of SCO3156 and SCO3847. Consistent with this, deletion of SCO3847 slightly lowered the MIC for vancomycin but showed no effect with bacitracin (Table 3). The PBP encoded by SCO3157 was up-regulated by both vancomycin and bacitracin but not by moenomycin.

The previously characterised van gene cluster responsible for remodelling peptidoglcan precursor biosynthesis to produce subunits terminating in D-alanyl-D-lactate (rather than D-alanyl-D-alanine) was, as expected, induced strongly only by vancomycin, and not by either of the other two antibiotics [43, 44](Figure 3B). vanH, encoding D-lactate dehydrogenase, exhibited the largest change in expression observed for any gene: a 300-fold increase 90 min after vancomycin addition.

The biosynthesis of antibiotics is induced in response to cell wall specific drugs

In the gene ontology analysis (Additional files 5, 6,7, 8, 9, 10 and 11), the GO term for acyl carrier activity was prominent in the lists from the term-for-term and MGSA analysis of genes induced by vancomycin or bacitracin in the 90 min sample, with as many as 8 genes associated with this term present in the lists of up-regulated genes. These correspond to four genes each from the biosynthetic clusters responsible for producing the antibiotics CDA and undecylprodiginine (Red), and indicate that activation of antibiotic production is part of the cellular response to exposure to both vancomycin and bacitracin. Indeed, 39 of the 47 CDA cluster probe sets (corresponding to 32 of the 40 genes) were up-regulated after addition of vancomycin, along with all 23 Red cluster probe sets (22 of the 22 genes; Table 1, Additional files 14 Figures 1d and 1i). The corresponding figures for bacitracin treatment were 35 and 9 probe sets, respectively, while moenomycin induced 11 probe sets from the CDA cluster. Activation of CDA synthesis is therefore a common response to treatment with all of the antibiotics. The expression of the CDA and Red biosynthesis genes, and that of the only other antibiotic biosynthetic gene cluster identified in the genome, the actinorhodin (Act) cluster, is ultimately dependent on activation by regulatory proteins encoded by each cluster [51–53]. The changes in expression of these activator genes brought about by treatment with the cell wall antibiotics is summarised in Table 4, and indicates that vancomycin and bacitracin in particular had a strong inducing effect on transcription of all four activators. Although an increase in expression of the actinorhodin biosynthesis genes was not observed in the 90 min in liquid culture following addition of any of the antibiotics, vancomycin treatment induced production of the blue pigmented antibiotic in a paper disc assay using growth on a solid agar medium (Figure 4). This response was markedly reduced, but not abolished, in an M570 mutant defective in ppGpp synthesis, indicating that the signalling molecule is not essential for this response to vancomycin.

Antibiotic cluster	Activator gene vancomycin	Fold-change increase in expression in the 90 min sample1
		Vancomycin	Bacitracin	Moenomycin
CDA	cdaR (SCO3217)	5.4	2.4	1.5
Act	actII-ORF4 (SCO5085)	2.8	1.6	1.5
Red	redD (SCO5877)	11.4	4.4	2.1
Red	redZ (SCO5881)	5.7	3.6	NA

¹NA = the gene was not present in the list of significantly differently expressed genes in Additional files 1, 2 and 3.

Sigma factor gene expression is a major response to exposure to cell wall antibiotics

S. coelicolor RNA polymerase can interact with any one of 64 different sigma factors, each specifying recognition of a different cognate consensus promoter sequence [54]. Changes in the relative abundance of sigma factors within a cell can therefore significantly influence the pattern of gene transcription. Sigma factor genes were significantly over-represented in the lists of genes induced by each antibiotic (Table 1) indicating that alteration in sigma factor gene expression is an important response to treatment with antibiotics targeting cell wall biosynthesis. This is also reinforced by the prominence of genes known to be regulated by σ^E, σ^B and σ^R in the differentially expressed gene lists (see Tables 1 and 2), and by the observed increase in susceptibility of σ^E and σ^B mutant strains towards bacitracin and vancomycin (see Table 3). Eight sigma factor genes, including σ^E (SCO3356) and σ^B (SCO0600), were up-regulated following treatment with all three compounds. Of these, the genes encoding the putative extra-cytoplasmic function (ECF) sigma factors encoded by SCO4005, SCO4908 and SCO5147 were particularly strongly induced (5- to 7-fold by vancomycin; Additional file 14 Figure 1l). A mutant strain in which SCO4005 was disrupted by insertion of a transposon showed increased susceptibility to bacitracin, but was not significantly altered in its resistance to vancomycin (Table 3). However, a similar mutant in which SCO5147 had been disrupted showed no increase in susceptibility to either compound and was slightly more resistant to vancomycin than the parental strain (Table 3). Transcription of the alternative sigma factor σ^hrdD (SCO3202) known to be dependent on σ^E[20] was also up-regulated by all three antibiotics, while the expression of σ^bldN (SCO3323), a sigma factor important for S. coelicolor differentiation [55], was repressed by all three. Vancomycin alone repressed transcription of four other sigma factor genes (SCO3068, SCO3626, SCO5216 (σ^R), and SCO5243); two of these, including σ^R, were up-regulated after treatment with at least one of the other two antibiotics. Consistent with this, genes from the σ^R-regulon were significantly over-represented in both the vancomycin-repressed and moenomycin-induced gene lists (see Tables 1 and 2).

The zinc-responsive zur regulon is induced exclusively by vancomycin

Zinc plays an indispensable role in cellular biochemistry as a catalytic or structural cofactor for a wide variety of metalloproteins. It can however be toxic if accumulated to excess, and its intracellular concentration is controlled within safe limits by homeostatic regulatory mechanisms. In bacteria this is largely achieved by tight control of import and export mechanisms via the zinc-responsive regulatory protein Zur [56, 57], and the Zur regulon in S. coelicolor has recently been characterised [58–60]. Interestingly in this work the majority of the genes in the regulon, including the znuABC operon (SCO2505-2507) encoding the high-affinity Zn²⁺ import system, were transiently up-regulated following addition of vancomycin but not by either of the other two antibiotics (Table 1, Additional file 14 Figure 1p). Expression of znuA (SCO2505) was induced 12-fold 30 min after vancomycin addition, decreasing to about 2-fold up-regulated by 90 min (see Additional file 3).

Genes required for branched chain amino acid and biotin biosynthesis are induced by moenomycin

GO terms associated with amino acid biosynthesis were over-represented in the genes induced in response to moenomycin treatment (see Additional files 5, 6, 7, 8, 9 and 10). Comparison of the significantly differently expressed genes with the groups of functionally related genes in Tables 1 and 2 supported this observation, indicating that moenomycin induced genes associated with branched chain amino acid (leucine, isoleucine and valine), lysine and methionine biosynthesis. Indeed, the ilvD (SCO3345), leuB (SCO5522) and leuC (SCO5553) genes for the biosynthesis of the branched chain amino acids leucine, valine and isoleucine were up to 7-fold up-regulated by moenomycin treatment (Additional file 14 Figure 1c). In addition, genes associated with the transport of amino acids were over-represented in both the moenomycin up-regulated genes, and those down-regulated by the antibiotic (Tables 1 and 2). Interestingly, the down-regulated genes included four from the putative branched-chain amino acid import operon SCO2008-2012 (Additional file 14 Figure 1a).

GO analysis of the 252 genes specifically up-regulated by moenomycin in Figure 1B indicated a significant over-representation not only of amino acid biosynthesis genes, but also of biotin biosynthesis. The bioB-D operon (SCO1244-1246) was approximately 2-fold up-regulated in the 90 min moenomycin-treated sample. Biotin is a cofactor implicated in catabolism of fatty acids and branched chain amino acids, and up-regulation of its production may indicate an increased use of branched chain amino acids as an energy source in the moenomycin treated cells. Indeed, expression of accD1 (SCO2776) encoding the biotin-dependent acetyl-CoA carboxylase enzyme, a key enzyme in branched chain amino acid degradation, was 5-fold up-regulated 90 min after moenomycin addition (see Additional file 3).

Genes encoding two-component signal transduction systems are specifically induced by bacitracin

GO analysis of the 39 genes whose transcription was induced only in response to bacitracin (see Figure 1B) indicated genes associated with two-component system signal transduction as the most over-represented process (see Additional file 11), with ten of the genes encoding response regulator or sensor kinase proteins. These included two clusters of three genes encoding a response regulator-kinase-kinase (SCO4596-4598), and a regulator-regulator-kinase (SCO3638, SCO3640 and SCO3641). This implies that they represent multi-component signal transduction systems that are important for coordinating a response to bacitracin (but not moenomycin or vancomycin). Other two-component system genes uniquely up-regulated in response to bacitracin were the SCO5683-5684 regulator-kinase pair, and the kinases encoded by SCO2359 and SCO7649. Mutant strains deleted for SCO3638-3641, SCO4596-4598 or SCO5683-5684 did not however show any significant change in their resistance to bacitracin (data not shown).

Identification of novel ABC transport systems important for bacitracin resistance

SCO3089-3090 and SCO3010-3011 encode putative ABC transport proteins that show a high degree of homology: SCO3089 is 75% identical to SCO3111 at the amino acid level, while SCO3090 shows 38% identity over its full-length to SCO3110 (Figure 5A). Transcription of all four genes was strikingly up-regulated 30-fold or higher in response to each of the three compounds (Figure 5B, Additional files 1, 2 and 3), suggesting an important role for these transporters in the response to the antibiotic treatments. To test this further, mutant strains carrying deletions in one or both of the transport systems were constructed and analysed for their susceptibility to vancomycin and bacitracin (Table 3 and Figure 5C). While the single and double mutant strains showed at most only a marginal increase in susceptibility to vancomycin, the double mutant strain H1004 exhibited markedly increased susceptibility to bacitracin, reducing the MIC to below 10 μg/ml compared to 40-45 μg/ml for the parental strain. Both single mutant strains also showed an increased susceptibility to bacitracin, but MIC values (15-20 μg/ml for H1003 (ΔSCO3089-3090) and 30-35 μg/ml for H1002 (ΔSCO3110-3111)) were not as low as for the double mutant.

Stress response genes contribute towards antibiotic resistance in S. coelicolor

Vancomycin, bacitracin and moenomycin target distinctly different stages of cell wall biosynthesis (see Figure 3A), yet exposure of S. coelicolor to each antibiotic resulted in changes in expression of many of the same genes (see Figure 1). Previous studies on the role of the ECF (e xtrac ytoplasmic f unction) sigma factor σ^E in maintaining cell wall homeostasis in S. coelicolor identified a key role in responding to antibiotics that damage the cell wall [19]. This study reveals that the transcriptional response to vancomycin, moenomycin and bacitracin extends far beyond the involvement of σ^E, incorporating elements of the heat shock response, the stringent response, and the response to osmotic and oxidative stress. Whether this reflects different, condition-specific regulatory cascades influencing the same genes, or triggering of the same regulatory cascade by different stimuli is unclear. The failure of vancomycin to increase synthesis of the stringent factor ppGpp while modulating expression of many stringently controlled genes suggests the former, but simultaneous induction of transcription of σ^B and the σ^B regulon, and of σ^E and the σ^E-dependent gene SCO3202, by all three antibiotics would be indicative of the latter. A combination of the two appears likely. These stress responses were not necessary for antibiotic resistance, but were required for maximal resistance. A σ^B-deletion mutant strain and a mutant defective in ppGpp synthesis both exhibited increased susceptibility towards bacitracin and vancomycin. S. coelicolor carries an inducible system that specifically confers high-level (MIC ~80 μg/ml) resistance to vancomycin, and functions by reprogramming cell wall precursor biosynthesis [43, 44]. Although the van genes involved are among the most highly induced in response to vancomycin treatment at all time points tested (see Figure 3B, Additional file 1), it is unlikely that cells become instantly fully resistant to the effects of the antibiotic and expression of the stress response genes may facilitate survival during this transition period. A transient slowing in growth was observed following treatment with both vancomycin and bacitracin, consistent with the existence of such a "stress" phase (data not shown).

S. coelicolor is susceptible to bacitracin (MIC 40-45 μg/ml) but highly resistant to moenomycin (MIC >300 μg/ml). All of the mutants tested were also resistant to 300 μg/ml moenomycin. It is perhaps therefore not surprising that moenomycin repressed the transcription of many fewer genes than either bacitracin or vancomycin (see Figure 1A). Interestingly however, moenomycin activated the transcription of a comparable number of genes to both vancomycin and bacitracin, including genes from the heat shock regulon, the stringent response, and the osmotic and oxidative stress regulons. Treatment with 10 μg/ml moenomycin, although only a fraction of the MIC, clearly induced a physiological response, perhaps reflecting induction of mechanisms for cell wall homeostasis well below inhibitory concentrations of the antibiotic or even a signalling role for the compound (see below).

Identification of new genes with roles in antibiotic resistance

Genes whose transcription was activated in response to two or more of the compounds may be components of a general response to treatment with antibiotics inhibiting cell wall biosynthesis. Inactivation of two putative ABC transporters, SCO3089-3090 and SCO3110-3111, both markedly up-regulated after treatment with all three antibiotics, significantly increased susceptibility to bacitracin. The integral membrane proteins encoded by these two systems are similar to lysophospholipase L1 and transport system components associated with lipoprotein release (http://streptomyces.org.uk), possibly suggesting a role in remodelling the cell envelope. Interestingly, orthologous and syntenous systems also occur in the genomes of Streptomyces clavuligerus, Streptomyces avermitilis and Streptomyces griseus, suggesting conserved evolution and a common biological function. The ECF sigma factor encoded by SCO4005 also contributes to bacitracin resistance; identifying the SCO4005 regulon could provide new insights into how bacterial cells cope with the antibiotic.

Of the other genes inactivated and analysed for their effect on antibiotic resistance, only disruption of the PBP encoded by SCO1875 and of the putative lysyl-tRNA synthetase encoded by SCO3397 markedly increased susceptibility to both bacitracin and vancomycin. Transcription of both SCO1875 and SCO3397 is markedly up-regulated following treatment with each of the antibiotics and is therefore an important part of the antibiotic resistance response. Navratna et al. [61] reported that deletion of the gene encoding PBP4 in Staphylococcus aureus increased susceptibility to vancomycin and was associated with a reduction in cross-linked muropeptide. PBP5 in Enterococcus faecalis is required for resistance to cephalosporins [62]. None of these three PBPs are essential for growth in their respective organisms, yet modulation of their activities appears to be an important component of the response to compounds that inhibit peptidoglycan biosynthesis. The response to vancomycin treatment included the up-regulation of transcription of eight genes encoding putative PBPs, including two only induced by this antibiotic. These observations, together with three PBPs genes uniquely repressed following vancomycin addition are discussed later. Of the 28 genes predicted to encode tRNA synthetases in the S. coelicolor genome, only SCO3397 was up-regulated following drug treatment, the majority (21) typically being down-regulated. In addition to their central role in protein biosynthesis, aminoacyl-tRNAs are involved in porphyrin biosynthesis, N-terminal protein modification, peptidoglycan biosynthesis and aminoacyl-phosphotidylglycerol biosynthesis [63]. SCO3397 is required for maximal resistance to bacitracin and vancomycin in S. coelicolor, and homologous MprF proteins in Staphylococcus aureus and Listeria monocytogenes have also been implicated in resistance to antimicrobial peptides [64, 65]. Although it is not known whether SCO3397 possesses phospholipid lysylination activity similar to MprF, this protein class may represent a useful target for counteracting resistance to antibiotics active against the cell envelope.

Genes responding uniquely to individual drug treatments offer new insights into activity and resistance

While there were many similarities in the changes in gene expression following treatment with vancomycin, moenomycin or bacitracin, genes responding uniquely to only one of the antibiotics were also identified (see Figure 1B). Given the different activities of the three compounds, these results potentially provide new insights into the mode of action of, and mechanisms of resistance towards, the individual antibiotics.

Induction of antibiotic biosynthesis following antibiotic treatment: chemical warfare in the soil?

S. coelicolor M600 produces three antibiotics, actinorhodin (Act), undecylprodiginine (Red) and the calcium-dependent antibiotic CDA. Strikingly, expression of the regulatory genes required for activating transcription of all three biosynthetic gene clusters was up-regulated within 30-60 min of exposure to both vancomycin and bacitracin, and was accompanied by a marked increase in transcription of the Red and CDA biosynthesis genes. Moenomycin also induced the CDA biosynthetic gene cluster. Production of the pigmented antibiotics Act and Red in response to vancomycin treatment was readily detectable in an agar disc assay (see Figure 4). S. coelicolor shares its native soil environment with a host of other microbial strains capable of secreting their own antibiotic metabolites into their immediate surroundings, and it appears to have evolved to respond by activating production of its own bioactive metabolites. Whether it does so as a form of retaliatory chemical warfare or as a means of inter-species communication is open to speculation.

Antibiotics as signalling molecules?

The production of metabolites with antibiotic activity is a widespread trait of soil microorganisms, and must confer some evolutionary advantage. While some of these compounds may indeed be used to inhibit competing species, they may also act as signalling molecules, particularly at sub-inhibitory concentrations [74–77], where they can modulate the transcriptional profiles of other bacteria [78–80]. The significant transcriptional response of S. coelicolor to treatment with moenomycin at levels corresponding to <5% of the MIC may represent another such example.

Bacterial strains and culture conditions

S. coelicolor M600, a prototrophic plasmid-free derivative of S. coelicolor A3(2), was used for the microarray study. Spores of S. coelicolor M600 were germinated and grown to mid-log phase (OD600 nm approx. 0.5) in NMMP as described previously [19, 43, 44]. Biological triplicate cultures were treated at this point by adding a sub-lethal concentration (10 μg/ml) of vancomycin, bacitracin or moenomycin A and samples taken 0, 30, 60, 90 minutes after addition of the antibiotic. A negative control which received no antibiotic treatment was also performed. All of the antibiotics used for this study except moenomycin A were purchased from Sigma-Aldrich. Purified moenomycin A was a kind gift from Professor Peter Walzel in the University of Leipzig. All strains used are listed in Table 3.

Paper disc assays were used to study the induction of pigmented antibiotics by vancomycin. Spore lawns of M600 and M570 were spread on SMMS agar plates [81], and paper discs impregnated with vancomycin (or blank) were immediately applied to the surface. Cultures were incubated for 3 days at 30°C.

RNA isolation and DNA microarray analysis

RNA was isolated from liquid cultures according to Hesketh et al. (2007) [21]. Purified total RNA (10 μg) was processed into labelled and fragmented cDNA for hybridisation to Streptomyces diS_div712a Affymetrix GeneChip arrays as previously described [21]. Following scanning of the arrays, the data quality was verified using a variety of tools including the 'affyPLM', 'affy' and 'simpleaffy' packages for the statistical computing environment R [82], quality control methods available within GeneSpring 11 (Agilent), and data from report files generated in the Affymetrix Genechip Operating Software. One replicate from the bacitracin experiment failed quality control and was excluded from further analysis, leaving duplicate data for this antibiotic treatment. All microarray data are available from ArrayExpress (http://www.ebi.ac.uk/arrayexpress/) under accession number E-MEXP-3032.

To identify differentially expressed genes, array data were first imported into GeneSpring 11 with normalisation using the Robust Multichip Average (RMA) algorithm of Irizarry et al. [83]. The data were then filtered to remove genes deemed to be expressed at a level below reliable quantification by determining those with a raw signal value below a defined background cut-off value of 100 in all samples. The results were further filtered to remove genes deemed to be unresponsive to the conditions under test in the experiment by identifying those with normalised expression values between 0.667-1.5 (1.5-fold change limit) in all conditions. The filtered data (2789 genes) were then subjected to two-way ANOVA to identify genes significantly altered under the experimental conditions, contrasting the expression values in each antibiotic-treated sample with the corresponding time point from the untreated control. This was performed using the parametric test option with a false discovery rate of P < 0.01. P-values were corrected using the Benjamini and Hochberg false discovery rate multiple testing correction procedure, and were computed asymptotically. Details of the statistical calculations used in the software can be accessed through the manufacturer's manual.

Gene ontology (GO) analysis was performed on lists of differentially expressed genes to identify over-represented biological functions or processes using the S. coelicolor GOA annotation from EMBL-EBI (ftp://ftp.ebi.ac.uk/pub/databases/GO/goa/proteomes/84.S_coelicolor.goa) as listed on 17/12/2009. GO analyses were realised using the Ontologizer tool (http://compbio.charite.de/index.php/ontologizer2.html), performing term-for-term hypergeometric testing with the Benjamini and Hochberg false discovery rate correction, or a Bayesian modelling approach using the model-based gene set analysis (MGSA) option [84]. Statistical comparison of lists of differentially expressed genes to in-house curated lists of functionally related genes was performed using the 'find similar entity lists' tool of Genespring 11. The results (summarised in Tables 1 and 2) determined whether the number of genes from a particular functional class were over-represented in the data, and therefore biologically meaningful, or could have occurred by chance. Absence of a functional class in Tables 1 and 2 therefore does not mean that all genes in that class are absent in the data, but rather that the number of genes present could have arisen simply by chance at the 5% probability level used. Quality threshold (QT) clustering was performed in Genespring 7.3.

Construction of mutant strains for determining their antibiotic susceptibility

Targeted gene deletions were constructed using the PCR-directed mutagenesis procedure of Gust et al. [85] or by using transposon mutated cosmids from the library of Bishop et al. [86]. Briefly, for PCR-directed replacement of genes with an antibiotic resistance cassette (e.g. aac(3)IV for apramycin resistance, hyg for hygromycin resistance), cosmid DNA in E. coli was targeted with a disruption cassette created by PCR using gene-specific primers and pIJ773 (for aac(3)IV) or pIJ10007 (for hyg) as template. Candidate antibiotic resistant clones were verified by PCR and restriction digestion, and conjugated into S. coelicolor M600. Double crossover integrants were selected as apramycin/hygromycin resistant, kanamycin sensitive clones, and verified by PCR. For transposon mutants, mutated cosmids from the library were conjugated into S. coelicolor M600 and double crossover integrants were selected as apramycin resistant, kanamycin sensitive clones.

Assay for determining the minimum inhibitory concentration (MIC) of the antibiotics

Minimum inhibitory concentration (MIC) values were evaluated using a method similar to that of Andrews (2001) [87]. McFarland turbidity standards of 0 (1.0 ml distilled water), 0.5 (5.0 μl 1.175% BaCl₂, 995 μl 1% H₂SO₄), 1 (10.0 μl 1.175% BaCl₂, 990 μl 1% H₂SO₄), and 2 (20 μl 1.175% BaCl₂, 980 μl 1% H₂SO₄) were prepared. S. coelicolor spores from 20% glycerol stocks were added to 1.0 ml distilled water until the turbidity of the spore suspension was between McFarland standards 1 and 2. Spore suspensions were stored at 4°C and used within 1 week, during which time they showed no loss of viability. 5.0 μl aliquots of the spore suspensions were spotted onto a series of MMCGT agar plates [19] containing the antibiotic under test at a range of concentrations (vancomycin 30-150 μg/ml in 10 μg/ml increments; bacitraicin 10-45 μg/ml in 5 μg/ml increments; moenomycin 0-300 μg/ml in 20 and 50 μg/ml increments). The spots were allowed to dry, and the plates were incubated at 30°C. MIC values were determined by visual inspection of growth over the range of antibiotic concentrations after 48 hrs incubation.

Analysis of intracellular nucleotide pools

Quantification of intracellular nucleotides was performed by HPLC as described in Hesketh et al. [21]. Duplicate cultures of S. coelicolor grown to mid-exponential phase (OD600 nm approx. 0.5) and treated with 10 μg/ml vancomycin were extracted and analysed at 0, 15 30 and 60 min following addition of the antibiotic.