S. lividans 1326 was obtained from the John Innes Culture Collection. In this S. lividans background, we constructed an adpA mutant in which adpA was replaced with an apramycin-resistance cassette 25 .

Streptomyces was grown on NE plates 26 and in YEME liquid medium 27 in baffled flasks. MS medium was used for sporulation experiments 27 . Apramycin was added to final concentrations of 25 μg mL^-1 to solid media and 20 μg mL^-1 to liquid media as appropriate.

S. lividans microarrays were not available, so S. coelicolor oligonucleotide arrays covering most open reading frames (ORFs) of the genome (for array coverage and design, see 28 29 ) were used. Aliquots of 60 mL of liquid YEME medium were inoculated with about 10⁸ spores and incubated at 30°C with shaking at 200 rpm until early stationary phase (about 30 h of growth). Samples of 12 mL of culture (at OD_450nm = 2.3, corresponding to time point T on Figure 1a) were then collected and RNA extracted as previously described 30 . RNA quality was assessed with an Agilent 2100 Bioanalyser (Agilent Technologies). RNA indirect labelling and array hybridization were performed as described 31 and hybridized microarrays were scanned with a Genepix 4000A scanner (Molecular Devices).

R software 32 was used for normalization and differential analysis. A Loess normalization 33 was performed on a slide-by-slide basis (BioConductor package marray; 34 ). A paired t-test was used for differential analysis. Variance estimates for each gene were computed under the hypothesis of homoscedasticity, together with the Benjamini and Yekutieli P-value adjustment method 35 . Only genes with a significant (P-value < 0.05) fold change (Fc) were taken into consideration. Empty and flagged spots were excluded, and only genes with no missing values were analysed. A few genes which displayed excessive variation were analysed using the Vmixt method from the VarMixt package 36 . We defined our cut-off for microarray data acquisition as Fc <0.625 or Fc > 1.6 with P-value < 0.05. The genome of S. lividans 1326 was sequenced only recently 24 , so we used the StrepDB database 7 , and in some cases a basic local alignment search tool (Blast), to identify S. lividans orthologs (SLI gene number) of S. coelicolor genes. We also used the protein classification scheme for the S. coelicolor genome available on the Welcome Trust Sanger Institute database 37 .

Oligonucleotide pairs specific for cchA (SLI0459), cchB (SLI0458), SLI0755, SLI6586, ramR (SLI7029), hyaS (SLI7885) and hrdB (SLI6088, MG16-17) (Additional file 1: Table S1) were designed using the BEACON Designer software (Premier BioSoft). RNA samples were extracted from cultures in YEME liquid medium at OD_450nm values of about 0.3, 0.8, 1.5, 1.9 and 2.3 (time points A, B, C, D and T, respectively). Aliquots of 20 μg of RNA were treated twice with 2 Units of DNase I with the TURBO DNA-free reagent (Ambion) for 30 min at 37°C. Reverse transcription and quantitative real-time PCR were performed as previously described 25 . PCRs involved a hybridization step of 55°C, except for ramR, SLI0755 and cchB where a temperature of 58°C was used. Each assay was performed in triplicate and repeated with at least two independent RNA samples. The critical threshold cycle (C _T ) was defined for each sample. The relative amounts of cDNA for the tested genes were normalized to that of the hrdB gene transcript which did not vary under our experimental conditions (and thus served as an internal standard). The change (n-fold) in a transcript level was calculated using the following equations: ΔC _T  = C _T(test DNA) - C _{T(reference cDNA)}, ΔΔC _T  = ΔC _{T(target gene)} - ΔC _T(hrdB), and ratio =  2 - Δ Δ C T 38 . Student’s t test was used to evaluate the significance of differences between the expression level of tested genes and that of a reference gene. A P-value < 0.05 was considered significant.

Several AdpA-binding site sequences, identified in S. griseus by DNase I footprinting experiments 10 13 18 23 , were used with the PREDetector software (version 1.2.3.0) 39 to generate a S. griseus matrix 25 . This matrix was used with the S. coelicolor genome sequence (the S. lividans genome sequence was not available during the course of this study and is still not available on PREDetector software) to identify putative AdpA-binding sites upstream from S. lividans AdpA-dependent genes (scores > 3). The StrepDB database 7 and Blast were used to identify S. lividans, S. coelicolor and S. griseus ortholog gene names.

Radioactively labelled DNA fragments (180 bp to 496 bp) corresponding to promoter regions of putative S. lividans AdpA-regulated genes were obtained by PCR. Primers (named GSgene in Additional file 1: Table S1) were used to amplify the promoter regions of cchA (opposite orientation to cchB), SLI0755, SLI6586 (opposite orientation to SLI6587), ramR and hyaS as described elsewhere 25 . Purified radiolabelled fragments (10,000 cpm) were then used with purified AdpA histidine-tagged protein (AdpA-His₆) in EMSA as previously described 25 40 .

We had previously inactivated adpA in S. lividans and found that this adpA mutant failed to produce aerial mycelium on rich media and that its growth was comparable to that of the parental strain 1326 in liquid YEME medium at 30°C 25 . Expression studies with this S. lividans adpA mutant cultivated in liquid medium identified two differentiation-regulating factors (STI1 and the ClpP1ClpP2 peptidases) whose ORFs were under the direct control of AdpA 25 . We used transcriptome analysis of this adpA mutant to identify other AdpA-dependent pathways in S. lividans; however, this analysis was performed using S. coelicolor microarrays 29 because the S. lividans genome sequence was not yet available 24 and the two species are very closely related 41 . Total RNA was isolated from S. lividans 1326 and adpA cells during early stationary phase (time point T in Figure 1a) because at this growth phase, S. coelicolor adpA is expressed 4 ; also the expression of genes involved in secondary metabolism in a S. coelicolor bldA mutant 42 , a strain defective for AdpA translation, starts to diverge from that in the wild-type.

Global gene expression in the mutant was compared to that in the parental strain. The expression of more than 300 genes was affected in the adpA mutant at early stationary phase (Table 1 and Additional file 2: Table S2): 193 genes were significantly down-regulated (1.6-to 30-fold i.e. 0.033 < Fc < 0.625), and 138 were up-regulated (1.6-to 3.6-fold) with a P-value < 0.05 (see Additional file 2: Table S2 for the complete data set). Theses genes encode proteins of several different classes according to the Welcome Trust Sanger Institute S. coelicolor genome database 37 : 72 of the genes are involved in metabolism of small molecules, including seven playing a role in electron transport (e.g. SLI0755-SLI0754, cydAB operons) (Table 1); 18 encode proteins involved in secondary metabolism, for example the cchA-cchF gene cluster (SLI0459-0454) involved in coelichelin biosynthesis 43 and the SLI0339-0359 cluster encoding the putative deoxysugar synthase/glycosyltransferase. Deletion of adpA in S. lividans also affected the expression of 32 genes involved in regulation including ramR (SLI7029), wblA (SLI3822), bldN (SLI3667), hrdD (SLI3556) and cutRS (SLI6134-35) 1 6 . Sixty-two genes involved in the cell envelope 37 were differentially expressed in the adpA mutant; they include hyaS (SLI7885) 44 , chpE, chpH 1 , SLI6586 and SLI6587 which were strongly down-regulated in the adpA mutant (Table 1). Thirty-nine genes encoding proteins involved in various cellular processes (osmotic adaptation, transport/binding proteins, chaperones, and detoxification) 37 were also deregulated in the absence of AdpA (Additional file 2: Table S2). The expression of 111 genes coding for proteins with unidentified or unclassified function was altered in the adpA mutant. Thus, deletion of adpA influenced the expression of a large number of genes involved in a broad range of metabolic pathways, and indeed other functions, in S. lividans.

^aGene expression in the S. lividans adpA mutant was compared to that in the wild-type, using S. coelicolor microarrays. Table 1 shows a selected subset of the genes (see Additional file 2: Table S2 for the complete list). The genes presented here were further studied or are discussed in the text because of their role in Streptomyces primary or secondary metabolism 1 6 17 .

^bGene names for S. coelicolor (SCO) and S. lividans (SLI) and annotated function are from the StrepDB database 7 .

^c S. coelicolor microarrays were used for transcriptome analysis of the S. lividans adpA mutant (the complete microarray data set is presented in Additional file 2: Table S2). The S. lividans genome sequence was recently made available 24 and SLI ortholog gene numbers were identified as SCO gene orthologs with StrepDB database 7 . The expression of genes shown in bold was analysed by qRT-PCR. Intergenic DNA regions between genes labelled with asterisks were analyzed by EMSA (Figure 2). A SCO7658-orthologous sequence (98% nucleotide identity according to BLAST) was detected in S. lividans, downstream from hyaS, but it was not annotated as a S. lividans coding DNA sequence (CDS). However our microarray data suggest that this sequence is indeed a CDS or alternatively that the S. lividans hyaS CDS is longer than annotated.

^dSCO genes and their S. griseus orthologs studied and described under another name found on StrepDB database 7 or see “References”.

^eFold change (Fc) in gene expression in the S. lividans adpA mutant with respect to the parental strain with P-value < 0.05, as calculated by Student’s t-test applying the Benjamini and Hochberg multiple testing correction. ± indicates average Fc of some gene operons (see Additional file 2: Table S2 for details).

^fFrom a protein classification scheme for the S. coelicolor genome available from the Welcome Trust Sanger Institute database 37 : macromolecule metabolism (m. m.), small molecule metabolism (s. m.).

To confirm that S. lividans AdpA controls the expression of genes identified as differentially expressed in microarray experiments, six genes were studied in more detail by qRT-PCR. The six genes were selected as having biological functions related to Streptomyces development or the cell envelope (ramR 1 , hyaS 44 and SLI6586 37 ) or primary or secondary metabolism (SLI0755, cchA, and cchB 43 ), and for having very large fold-change values (Table 1). The genes in S. coelicolor and griseus orthologous to SLI6586 and SLI6587 encode secreted proteins 12 42 . The expression levels of these genes in S. lividans wild-type and adpA strains were measured after various times of growth in liquid YEME media (Figure 1b), as shown in Figure 1a.

The S. lividans hyaS gene was strongly down-regulated in the adpA mutant compared to the wild-type (Fc < 0.03) (Figure 1b) as previously observed for the SCO0762 homolog also known as sti1 25 . This suggests that hyaS expression is strongly dependent on S. lividans AdpA or an AdpA-dependent regulator. SLI0755, SLI6586 and ramR, were also expressed at a lower level in the adpA mutant than wild-type, particularly after mid-exponential phase (Figure 1b, times C, D and T); cchB seemed to be mostly affected by AdpA during stationary phase (Figure 1b, time T). The expression of cchA was strongly down-regulated by the absence of AdpA at times D and T (Figure 1b): note that despite repeated efforts, cchA expression could not be detected in samples corresponding to times A to C for unknown reasons. The findings for gene expression as determined by microarrays and by qRT-PCR were consistent, with the exception of those for ramR. The expression of ramR observed by qRT-PCR at time T differed from that determined in microarray experiments (Table 1), suggesting that some of our microarray data are flattened. Nevertheless, these qRT-PCR experiments confirmed that the expression of the six selected genes is indeed AdpA-dependent in S. lividans at every growth time studied.

To determine whether S. lividans AdpA directly controls these genes, we searched for potential AdpA-binding sites in their promoter regions in silico. A consensus AdpA-binding sequence (^5′TGGCSNGWWY^3′) has been established in S. griseus, and AdpA can bind up to five sites between positions -260 bp and +60 bp with respect to the transcriptional start point of the target gene 10 . BLAST analysis revealed that the S. griseus AdpA DNA-binding domain is conserved in S. coelicolor and S. lividans AdpAs (data not shown) suggesting that all three species share the same AdpA-binding consensus sequence.

The DNA sequences upstream from the S. coelicolor ramR and hyaS genes and the intergenic region between the divergently transcribed genes cchA/cchB, SCO0774/SCO0775 and SCO6197/SCO6198 were analyzed using PREDetector software 39 and a matrix was generated with identified S. griseus AdpA-binding sequences 10 23 25 . Between three and nine putative AdpA-binding sites were detected within the promoter region of the S. coelicolor genes and by analogy in orthologous S. lividans AdpA-dependent genes (Table 2, location with respect to translation start point). During the course of this study, the S. lividans 1326 genome sequence became available 24 (but not in a form suitable for analysis with PREDetector (version 1.2.3.0) 39 ) and its analysis suggested that the position and composition of AdpA-binding sites were different from those predicted. The putative AdpA-binding sites of S. lividans cchA/cchB at -101 nt and -86 nt are GGGCCGGTTC and TGGCTGGAAC, respectively. The AdpA-binding sites located upstream of SLI0755, SLI6586, and hyaS differ from their S. coelicolor orthologs (see Table 2, changes in the location from translation start site are indicated in bracket).

^a In silico analysis of the S. coelicolor genome using PREDetector software (version 1.2.3.0, the S. lividans database was not available at the time this analysis was performed) 39 to analyse orthologs of S. lividans AdpA-dependent genes. The S. coelicolor AdpA-binding sites identified were checked for their conservation and location using the S. lividans genome StrepDB database 7 (see legend c).

^bGenes are named according to the StrepDB database 7 . *binding sites located between S. coelicolor genes transcribed in the opposite orientation.

^cPutative S. coelicolor AdpA-binding sites were found in silico with PREDetector 39 ; ^#putative site located in the upstream from the CDS of cchB. The site location given corresponds to the position of first nucleotide most distant from the translation start point of the first gene named. The positions of some sites are not the same for the S. lividans orthologs as indicated in brackets (S. lividans StrepDB database 7 ). ^~ putative sites are in the CDS of SLI6587. Predicted CDS diverge between SLI6586 and SLI6587 locus and their orthologs SCO6197 and SCO6198, resulting in a smaller intergenic region in S. lividans.

^dCS, coding strand; NCS, non coding strand with reference to the first gene named in the S. coelicolor gene column.

^eScores given by PREDetector software for S. coelicolor genes 39 .

^fSites present (+) or absent (-) in the S. lividans DNA probes used in EMSA experiments.

We used EMSA to test whether S. lividans AdpA binds to predicted S. lividans AdpA-binding sequence. Recombinant purified AdpA-His₆ bound to the promoter region of S. lividans sti1 (SCO0762 homolog), an AdpA-dependent gene, whereas an excess of AdpA-His₆ (up to 34 pmoles) did not bind to the promoter of SLI4380 (SCO4141 homolog), a gene that is not controlled by S. lividans AdpA. This suggests that the binding of AdpA with the promoter of genes tested in our previous study was specific 25 . AdpA-His₆ was able to bind to the promoter regions of all S. lividans AdpA-dependent genes tested (Table 2, Figure 2), although with different affinities. For SLI6586/SLI6587, ramR and hyaS, displacement of the DNA fragment to the slower migrating protein-DNA complex was nearly complete with amounts of AdpA of less than 11 pmoles (Figure 2, lane 2). For cchA/cchB and SLI0755/SLI0756, larger amounts of AdpA were necessary for near complete displacement of the DNA probe to a protein-DNA complex. In a competition EMSA performed on SLI6586/6587 with an excess of the corresponding unlabelled probe, AdpA-binding to the labelled probe decreased (data not shown). We also tested a hyaS promoter in which one (highest score) of the three putative AdpA-binding sites was mutated (at position -134 to -129, see Additional file 3: Figure S1a): the affinity of AdpA for this promoter region was reduced and one protein-DNA complex disappeared (Additional file 3: Figure S1b). These results suggest that one dimer of AdpA binds the adjacent sites -129 and -123 of S. lividans hyaS promoter and another dimer binds the -100 site resulting in the formation of the two DNA-AdpA complexes depicted in Figure 2.

These EMSA experiments demonstrated that S. lividans AdpA directly binds to five intergenic regions and confirmed the in silico prediction presented in Table 2. S. lividans AdpA directly regulates at least the six AdpA-dependent genes listed above and identified by microarrays and qRT-PCR analysis. These newly identified targets highlight the pleiotropic role of S. lividans AdpA: it is involved in primary (SLI0755) and secondary (cchA, cchB and hyaS) metabolisms, in regulation (ramR), and in cell development (hyaS, ramR and SLI6586).