ArgR of Streptomyces coelicolor Is a Versatile Regulator

ArgR is the regulator of arginine biosynthesis genes in Streptomyces species. Transcriptomic comparison by microarrays has been made between Streptomyces coelicolor M145 and its mutant S. coelicolor ΔargR under control, unsupplemented conditions, and in the presence of arginine. Expression of 459 genes was different in transcriptomic assays, but only 27 genes were affected by arginine supplementation. Arginine and pyrimidine biosynthesis genes were derepressed by the lack of ArgR, while no strong effect on expression resulted on arginine supplementation. Several nitrogen metabolism genes expression as glnK, glnA and glnII, were downregulated in S. coelicolor ΔargR. In addition, downregulation of genes for the yellow type I polyketide CPK antibiotic and for the antibiotic regulatory genes afsS and scbR was observed. The transcriptomic data were validated by either reverse transcription-PCR, expression of the gene-promoter coupled to the luciferase gene, proteomic or by electrophoresis mobility shift assay (EMSA) using pure Strep-tagged ArgR. Two ARG-boxes in the arginine operon genes suggest that these genes are more tightly controlled. Other genes, including genes encoding regulatory proteins, possess a DNA sequence formed by a single ARG-box which responds to ArgR, as validated by EMSA.


Introduction
Arginine metabolism is feedback repressed by arginine in different Gram-positive and Gram-negative bacteria. This effect is mediated by ArgR, a hexameric protein that represses arginine biosynthesis genes, using L-arginine as co-repressor, in Escherichia coli [1] and Pseudomonas [2]. A similar effect is exerted by the homologous regulatory protein AhrC in Bacillus subtilis [3] or Lactococcus [4] and by ArgR in Corynebacterium [5].
Characterization of argR, encoding the ArgR repressor [6] and the use of AhrC protein permitted to understand the arginine biosynthesis cluster regulation in Streptomyces clavuligerus and to locate ARG-boxes, for ArgR binding, upstream of several arginine biosynthesis genes [7]. This provided the basis to study arginine regulation in other Streptomyces species. This group of soil-dwelling bacteria produces many secondary metabolites that use arginine, or arginine-related molecules, as precursors. Streptomycin, mitomycin, streptothricin or clavulanic acid are metabolites that contain moieties of guanidine, carbamoyl groups, ornithine or arginine, all of which are compounds related to the arginine biosynthesis or catabolism [8][9][10][11]. The C5 moiety of clavulanic acid, produced by S. clavuligerus, derives directly from an arginine molecule [11]; in addition, undecylprodigiosin [12] and other metabolites contain proline, which as arginine derives from glutamate, and might share common regulatory mechanisms with arginine.
DNA microarrays and proteomic studies are useful tools to understand metabolic pathway regulation in microorganisms. In Pseudomonas, which has four arginine catabolic pathways: arginine deiminase (ADI), arginine succinyl transferase (AST), arginine decarboxylase (ADC) and arginine dehydrogenase (ADH), the transcriptomic studies identified 38 genes related to the ADH pathway that are induced by arginine in the absence of ArgR, as well as 27 arginine-induced genes of the AST pathway [2,13]. Using transcriptomic studies Caldara et al. [14] have also found that arginine uptake systems encoded by the art and the hisJQMP genes are repressed by arginine in E. coli.
Transcriptomic studies have been very useful to understand the molecular mechanisms of global control by high hierarchy regulators, for example, A-factor regulation in Streptomyces griseus [15] and phosphate and nitrogen regulation in S. coelicolor [16][17]. These studies allowed us to understand networks interconnecting phosphate and nitrogen metabolism in Streptomyces [18], carbon regulation [19] and the connection between the global regulator AfsR, a protein controlling antibiotic production in S. coelicolor, and the phosphate regulation exerted by PhoP, the response regulator of the two components of the PhoRP system that controls the pho regulon [20].
Therefore, it was convenient to analyze gene expression under control and arginine-supplemented conditions, to obtain better knowledge about arginine transport and catabolism in this model microorganism, and to determine whether ArgR is a regulatory protein involved only in arginine control, or if it has wider regulatory functions.

Comparison of S. coelicolor strains grown in the presence and absence of arginine
Experiments were performed to define the optimal conditions to achieve similar growth kinetics in MG medium, as well as good reproducibility, for the parental strain S. coelicolor M145 and the argR (SCO1576)-deleted mutant S. coelicolor DargR. A high arginine concentration is required in Streptomyces to produce a clear effect on enzymes of the arginine biosynthesis pathway in MG medium [6,21], therefore, the MG cultures were supplemented with 25 mM arginine. The pattern of arginine utilization was similar for both strains; approximately 15% of the arginine was consumed at 32 h of growth and no arginine was left at 96 h. Antibiotic onset in liquid MG medium using S. coelicolor M145 occurred at about 40 h for actinorhodin and 50 h for undecylprodigiosin. Under non-supplemented conditions, S. coelicolor DargR produced only about 20% undecylprodigiosin in relation to the parental strain and lacked completely actinorhodin formation. Arginine supplementation strongly impaired the production of the pigments in both strains (Fig. 1).
Transcription profiles of S. coelicolor M145 and the DargR mutant in response to arginine The statistical analysis of four biological replicates for each experimental condition indicated in Materials and Methods identified 459 genes with significant differential transcription (Dataset S1) in at least one out of the five contrasts shown in Fig. 2. Only 27 genes (6.1%) were differentially expressed in the argininesupplemented conditions, with respect to the control, indicating a weak transcriptional response to the presence of arginine (Fig. 2, contrasts 2 and 3). Most of the 459 differentially expressed genes corresponded to the comparisons between the wild-type strain and the DargR mutant in control (contrast 1) or arginine-supplemented (contrast 4) cultures. About 50% of the genes differentially expressed encode membrane proteins, secreted proteins or proteins with unknown functions.
The transcriptional profiles analysis permitted to establish a classification of the genes into five types. The patterns of genes differentially transcribed in the two strains in unsupplemented or supplemented conditions fell mainly into types I and II (Fig. 3) with different modulations of the expression (subtypes 1-5). Fortythree genes showed the pattern of expression indicated in types III to V (Fig. 3). Here we will analyze general genes included in the five different types, while genes with specific functions will be analyzed in subsequent sections.
Type I genes. For type I genes, ArgR appeared to act as a repressor protein. Arginine itself did not affect expression of 76 genes (subtype I.1) and produced a small negative effect in the presence of ArgR or a stimulatory effect in the absence of ArgR in the other 69 genes (subtypes I.2 and I.3, respectively), which was stronger in subtypes I.4 and I. 5. The arginine effect, if detected, was negative in the parental strain, but positive in the mutant strain (subtypes I.2-I.5). Thus, arginine appeared to function as a co-repressor for the genes classified into types I.2, I.4 and I.5. Surprisingly, subtypes I.1-I.3 comprise 14 genes for regulatory proteins including regulators of the TetR, DeoR, AraC, GntR or LipR families and sigma, anti-sigma or anti-sigma antagonist factors (Table S1). Other genes of type I encode proteases, peptidases or aminotransferases, adenosine deaminases and glutamyl-tRNAcharging amidotransferases, or are related to amino acid metabolism, such as those for serine hydroxymethyltransferase (SCO4837), NAD-dependent glutamate dehydrogenase (SCO2999), glutamate binding protein (SCO5776), and the B 12 -dependent and the B 12 -independent methionine synthases MetH and MetE (SCO1657 and SCO0985, respectively). Two genes involved in sporulation regulation, and five genes for aerial mycelium formation are also included in this group (Table S1).
Type II genes. For type II genes, ArgR behaved as an activator. Addition of arginine (contrasts 2 and 3) did not modify these genes expression (42 genes in type II.1) or produced a slightly positive or negative differential expression in the presence or absence of ArgR, respectively (56 and 32 genes in subtypes II.2 and II.3). Genes encoding regulators of AsnC, AraC, PadR, AbaA, TetR or DeoR type or members of two component systems are present in these subtypes. Type II includes genes for proteases, peptidases and protease inhibitors (Table S1). Some secondary metabolism genes, such as those encoding the putative actinorhodin transporter (SCO5083) and the butyrolactonebinding protein ScbR (SCO6265), belong to this type.
Subtypes II.4 and II.5 include 20 genes that were upregulated by arginine supplementation in the parental strain (contrast 1), but not in the argR mutant. The most interesting genes in these subtypes are the sigma-like afsS gene (SCO4425), a gene that encodes a tetracenomycin-like transporter (SCO2373) and a member of the two component system for nitrate sensing (SCO1160).
Type III genes. This type comprises 34 genes whose transcription was stimulated by arginine in the strain containing ArgR but downregulated in the absence of ArgR (i.e. the opposite behaviour to that of types I.4 and I.5). Genes of relevance in this type are transcriptional regulators (SCO2184 and SCO3769), a member of a two-component system (SCO6421), and the ATPdependent polyphosphate kinase (SCO1781).
Type IV genes. The expression of the four genes of this type was stimulated by arginine, independently on the presence of ArgR. Two of these genes encode, respectively, an amidinotransferase (SCO1222) and the c-aminobutyric acid aminotransferase (gabT, SCO5676).
Type V genes. The five genes in this type showed patterns similar to those of types I.2 or I.5, with a more marked effect in the M145 strain. Two genes related to amino acid metabolism, a cysteine desulfurase and a threonine synthase (SCO2146 and SCO4293) are included in this type.

Regulation of arginine biosynthesis and transport genes as measured by transcriptomics
Arginine biosynthesis genes in S. coelicolor are organized in the argCJBDR cluster (SCO1580-1576) and in three separate genes: argH (SCO1570), arcB (SCO5976, located in the arc cluster) and argG (SCO7036) (Fig. 4). Genes involved in the formation of carbamoyl-phosphate (carAB) and required to form citrulline from ornithine will be considered in the next section. Our transcriptional data of the S. coelicolor parental strain (containing ArgR) did not show significant arginine-dependent repression for most arginine biosynthetic genes (argCJD, argG and arcB are in the arginine-not-affected type I.1), in contrast to E. coli [14]. On the other hand, arginine supplementation in the DargR mutant cultures caused a slight upregulation of some arg genes as argH (type I.3) and argB (type I.5) with a 2.4 and 2.1-fold increase, respectively. The main transcriptional differences were the higher transcription in the DargR mutant (from 1.6-fold in argD to seven fold in argB or arcB) in relation to the wild-type strain. A putative gene, SCO1581, divergent to argC, is not differentially expressed in any condition.
The SCO5258-5260 genes, homologous to Pseudomonas aeruginosa aotJQMOP encoding proteins for arginine/ornithine uptake, were similarly transcribed in all conditions. However, transcription of the SCO5776-5777 genes, annotated as a putative ATPase and binding proteins for glutamate uptake (an arginine precursor) were negatively modulated by arginine and activated in the DargR mutant.

Arginine catabolism
S. coelicolor M145 grew well on plates containing ornithine, arginine, urea or citrulline as sole carbon or nitrogen source. No significant differences were found in the growth of the S. coelicolor DargR mutant, with only a slightly lower growth on citrulline.
Of the multiple catabolic pathways for arginine catabolism found in other bacteria, only two catabolic genes, arcA and arcA2 (SCO0613 and SCO5975), both for arginine deiminase, have been identified in S. coelicolor genome. They have a high similarity to Pseudomonas or E. coli arcA genes but they were not differentially expressed in our experimental conditions. All the genes of the ADC pathway of Pseudomonas, forming succinate from arginine, have homologous genes scattered in S. coelicolor genome (SCO7311, SCO5527, SCO6414, SCO5655, SCO4913, SCO5676 and SCO7035) with e-value scores of 10 280 to 10 2101 . These genes were not affected by arginine or by the ArgR regulator, with the exception of gabT (SCO5676), which encodes the enzyme that catabolizes 4-aminobutyrate to form succinate semialdehyde. Expression of this type IV gene was twice as high in cultures supplemented with arginine than in unsupplemented cultures of both the M145 and the DargR strain.
The AST pathway is encoded by the aruCFGDBE genes in P. aeruginosa [4]. In S. coelicolor, an aruC homologue, gabT (SCO5676, discussed above) was stimulated by arginine and SCO1865 (subtype I.1), encoding an aminotransferase, was overexpressed in the DargR mutant. Homologues of aruF or aruG are not present in S. coelicolor and those genes encoding proteins similar to AruD did not show differential expression under any conditions. Genes encoding standard arginases (EC.3.5.3.1) have not been found in the S. coelicolor genome. Nevertheless, arginase activity is possibly associated with the carbamoyl-phosphate transferase activity of ArcB, as occurs in S. clavuligerus [22]. Expression of arcB (and the putative arginase activity) was strongly repressed in the presence of ArgR, as indicated above. No effect due to arginine or to the lack of ArgR was observed in the urea-utilizing genes (ureDGFCBA, SCO1231-1236, and ureABC, SCO5525-5526).

Nucleotides and nucleic acid-biosynthesis-related genes
Pyrimidine biosynthesis is related to arginine biosynthesis through the utilization of carbamoyl-phosphate (CP). CP is formed by the CP synthase, encoded by pyrA and pyrAa (Figs. 4 and 5) and condensed to ornithine by the ornithine carbamoyltransferase (arcB) to yield citrulline, and to aspartate to form carbamoyl-aspartate, by the aspartate carbamoyl transferase (pyrB). Both arcB (I.1 subtype) and pyrB genes were upregulated in the DargR mutant.
In the pyrimidine pathway, carbamoyl-aspartate is sequentially converted in UMP through the action of enzymes encoded by the pyrB, pyrC, pyrD, pyrE and pyrF genes [23], clustered in S. coelicolor, with exception of pyrE (Fig. 5A). Several genes in this cluster showed a significantly higher expression in the DargR mutant than in the parental strain (1.6-2.2-fold; contrast 1). Addition of arginine slightly reduced their transcription in the M145 strain but it increased in the DargR mutant (2.7-4-fold; contrast 4). Two sets of these genes were especially affected: pyrR-pyrB and SCO1485-pyrA-pyrAa showing profiles of type I.5, except for SCO1485 which was of type I.4. No significant differential expression of the pyrE gene, located outside of the cluster, was detected.
Two sets of independent ribonucleotide reductases (RNRs) in Streptomyces are encoded by the nrdABS (SCO5226-5224) and nrdRJ clusters (SCO5804-5805). The first cluster encodes the oxygendependent RNR and is regulated by a B 12 -dependent riboswitch located upstream of nrdA [24]. The microarray probes for the riboswitch region (associated to the fake gene nrdX because of the original genome annotation), the nrdS (encoding an AraC-like regulator) and the nrdJ (encoding the class II RNR) genes, indicated a 30-40% expression decrease in arginine supplemented cultures of the wild-type strain, and a higher expression in the DargR-mutant (up to 1.5-fold), especially in the presence of arginine (Fig. 5B). Two genes, cobQ and cobO (SCO1848 and SCO1851) for B 12 biosynthesis had the type I.1 profile.

Effect of arginine on the nitrogen metabolic network
The nitrogen utilization network in Streptomyces uses the glutamine synthetase, encoded by glnA, as a key enzyme and is strictly regulated by the OmpR-like regulators GlnR and GlnRII [17].
Our data showed that the transcription of several nitrogen metabolism genes was controlled by ArgR. The glnK gene (SCO5584), encoding a PII-like regulator, was downregulated in the DargR mutant (II.2 type). The same regulation was found for the glnK flanking genes, amtB and glnD, encoding respectively an ammonium tranporter and the PII-modifying adenylyltransferase. Similarly, glnII (SCO2210) that encodes the eukaryotic-like glutamine synthetase GlnII, and glnA (SCO2198) that encodes the prokaryotic type I glutamine synthetase, showed profiles of type II.1 and II.4. Therefore, all these genes appeared to require ArgR for optimal expression. No differences in transcription were found for the regulators glnRII and glnR (SCO2213 and SCO4159, respectively). Thus, it seems that the positive regulation of ArgR on nitrogen metabolism is not mediated by GlnR or GlnRII.

Effect on morphology, differentiation and secondary metabolism genes
Arginine addition or deletion of argR resulted in a drastic reduction of actinorhodin and undecylprodigiosin, or even in complete lack of production of the latter pigment ( Fig. 1). Indeed, two genes of the act pathway, actII-orf2 (SCO5083) encoding a putative actinorhodin transporter [25] (profile II.3) and the ketoacylreductase ActIII (SCO5086; II.4 profile) were affected. Probably due to the early time of culture used to extract RNA, no differences were found in expression of the undecylprodigiosinencoding genes. The effect on production of this antibiotic ( Fig. 1) was not related to the supply of the proline precursor, since the three genes of the proline pathway (proABC, SCO2585, SCO2587 and SCO3337) were equally expressed in both strains or culture conditions. Two regulatory genes, essential for antibiotic formation, afsS (SCO4425) encoding a sigma-like protein and scbR (SCO6265) for a butyrolactone-receptor protein [26][27] showed II.5 and II.1 profiles, respectively, with a strong decrease in expression in the DargR mutant. The repression in afsS may also contribute to the drastic reduction of both pigmented antibiotics.
A very strong effect due to the lack of ArgR was observed on the biosynthesis of rodlins and chaplins, required to form aerial mycelium and spores [28], and on the genes encoding the yellow polyketide pigment CPK [29] (Fig. 6). Genes for chaplins are located in five different sites in the genome being chpA and chpD clustered with rdlA and rdlB for rodlins (SCO2716-2719). Although expression of chpF and chpA was not affected, the other genes showed profiles of the I.4 type (chpG and chpE, SCO2699 and SCO1800, respectively) and I.5 type (chpH, rdlA and chpC, SCO1675, SCO2718 and SCO1674, respectively).
Most genes of the cpk cluster were downregulated in the DargR mutant. In the parental strain, higher transcription of some genes, especially cpkE, was detected in arginine-supplemented cultures. The profile of these genes was of II.1 type (kasO, cpkJ, cpkN, SCO6280, SCO6283 and SCO6288, respectively) or II.4 type (cpkDE, cpkHI, cpkK, SCO7276-7277, SCO6281-6282 and SCO6284, respectively) (Fig. 6B). The observed effect was probably due to a cascade effect mediated by the Streptomyces type of regulators encoded by kasO and cpkN, which were downregulated in the DargR mutant, but it may also reflect a general effect of ArgR on the glutamate supply, because the CPK antibiotic production is stimulated by glutamate [29]. Two genes located adjacent to the cpk cluster, scbR (already mentioned) and scF for a putative secreted FAD-binding protein, were also affected by the lack of ArgR (Fig. 6B).

Proteomic studies
Comparative analysis of the proteomes of S. coelicolor wild-type and DargR mutant grown with and without arginine supplementation revealed no significant differences, confirming the low effect of arginine detected in the transcriptomic studies. The largest proteome differences were found in the comparison of the wildtype and the DargR mutant. Twenty-six differentially represented proteins were identified, several of which were consistent with the transcriptomic results. They are listed in Table 1 (see Fig. 7 and Dataset S2 for details). Nitrogen metabolism proteins. Another transcriptomic and proteomic matching result was for glutamine synthetase I, GlnA (SCO2198, spot IN 6143), which was under-represented in the mutant proteome, confirming that glnA expression required ArgR as transcriptional activator. As glutamine is the amino donor for carbamoyl-phosphate synthesis, ArgR may influence the urea cycle and the pyrimidine biosynthesis by controlling the glutamine levels.
Pentose phosphate and glycolysis enzymes. Several proteins related to these pathways are under-represented in the S. coelicolor DargR proteome as compared with the wild-type. These the proteins are the transketolase TktA1 (encoded by SCO1935, spot IN 6445), that interconverts D-fructose-6-phosphate and Dxylulose-5-phosphate in the pentose phosphate pathway, which finally results in formation of D-glyceraldehyde- 3 IN 6317), were under-represented in the DargR mutant when compared to the wild-type strain. These proteomic results revealed that glycolytic and pentose phosphate pathways were more active in the wild-type strain.
Proteins related to succinate metabolism. Two of the enzymes involved in arginine biosynthesis, ArgH and ArgG, have succinate either as substrate or reaction product. Surprisingly several proteins using succinate were under-represented in the mutant proteome. This is the case of MetB (SCO4958, spot IN 6257) a cystathionine c-synthase that uses O-succinyl/acetyl homoserine as substrate and SucD (SCO4809, spot IN 6400) the a subunit of succinyl-CoA synthetase. Although no significant changes in the expression of these genes were observed, other genes related to succinate metabolism showed a differential expression in transcriptomic experiments. That is the case for gabT (forming succinate), the adenylosuccinate-synthetaseencoding gene purA (SCO3629, profile II.2) and for dhsA (SCO4856), dhsB (SCO4855) and SCO0922, three out of the 10 genes encoding succinate dehydrogenases which showed I.3 profile.
Purine metabolism. In the proteome of the wild-type strain, three protein spots involved in purine metabolism were detected, which were under-represented in the mutant strain. Proteins with inosine 59-monophosphate (IMP) dehydrogenase (SCO4770) and 5-inositol-5-monophosphate dehydrogenase (SCO4771) were identified in spots IN 6264 and IN 6269, respectively. These activities carry out the conversion of IMP to xanthosine 59-phosphate. The same profile was shown by the PurH protein, (spot IN 6125), encoded by SCO4814 involved in the synthesis of 1-(59-phosphoribosyl)-5-formamide-4imidazole carboxamide which connects purines and histidine pathways.
Other proteins. Three polyketide and fatty acid biosynthesis enzymes were also under-represented in the mutant proteome. The 3-oxoacyl-ACP reductase CpkI (encoded by SCO6282, spot IN 6648), required for the production of the yellow-pigmented CPK associated metabolite [29] which was consistent with the transcriptomic results. The enoyl-ACP-reductase FabI/InhA (SCO1814, spot IN 6550) was also under-represented.   (1) SCO n6 (2) gene name annotated function (3)   Post-transcritional regulation, protein turnover, mRNA stability or translation efficiency might account for the low correlation in the transcriptomic and proteomic results in some cases [30].

Validation experiments: RT-PCR and heterologous genes expression
In addition to proteomic studies, validation of the transcriptomic results was performed for nearly 40 representative genes by either (i) RT-PCR, (ii) coupling promoters to the Vibrio harveyi promoter-less luxAB genes to measure luciferase activity and (iii) EMSA using a recombinant Strep-ArgR fusion protein (Table 2).
Promoters of differentially expressed genes were fused to the promoter-less luxAB genes; these constructions were introduced into S. coelicolor M145 and S. coelicolor DargR and cultured in MG medium with or without arginine supplementation. The promoters tested were those belonging to argH and arcB for arginine biosynthesis, pyrR, pyrB and nrdA (SCO5226) related to nucleotide biosynthesis, leuA for amino acids biosynthesis, whiB for morphological differentiation, and SCO1086, a strongly ArgR affected gene, of unknown function. The promoter of the pyk1 gene, which is constitutively expressed in our culture conditions, was used as a negative control. In general, the luminescence values in S. coelicolor DargR exconjugants correlated well with the transcriptomic experiment, although S. coelicolor M145 exconjugants in the presence of arginine gave a slightly higher activity than that expected from the microarray expression data (Fig. 9).

Identification of ArgR binding sites
With the aim of finding putative ArgR binding sites that correlated with the differential expression results, we made a matrix (model 1) from the alignment of the 16 experimentally tested S. clavuligerus ARG boxes upstream of argC, argG, argH and arcB [6] and their complementary sequences. The S. coelicolor 59 regions were scanned by means of the Patser algorithm and the RSA tools server. A neutral separator of two positions was inserted between a duplicated matrix to take into account the tandem arrangement of the ARG boxes. Using model 1 the only sequences with Ri values (individual information content) higher than 12.2 were those of the above ARG boxes, and a new one in the promoter region of SCO1086. This gene encodes a protein with a transglutaminase-like motif (PF01841) that might be a protease according to Makarova et al. [31] and is repressed by ArgR (1.5 profile) as predicted by Castro-Melchor et al. [32].
Although no more ArgR binding sites were detected, more ArgR binding sites in S. coelicolor genome could have been expected. They might include low-conserved ARG boxes, boxes with different length of separation, or single (half-length) ARG boxes, as occurs in Thermotoga or B. subtilis [33][34]. A single putative ARG box of Ri 10.4 was found upstream of the leuA (SCO2528) gene. This region was found to be clearly retarded by ArgR. The 22 ARG boxes of argC, argG, argH, arcB, SCO1086 and leuA were integrated in a new weight matrix. In the new ARG box Streptomyces model (model 2, Fig. 10C), the binding site was formed by an imperfect 20 nt palindromic sequence. This matrix allowed us to locate 1583 ARG box candidates with a Ri.5. Manual selection of them was performed, paying attention to those located in the neighbouring of differentially transcribed genes.
Nearly thirty DNA fragments containing putative ARG boxes were used in binding assays in the presence of Strep-ArgR in EMSA experiments. They correspond to upstream regions of: (i) genes related to amino acids biosynthesis (other than arginine) such as SCO4293 for a putative threonine synthase, leuA or the bidirectional promoter region hisD-SCO2055; (ii) genes for nucleotide biosynthesis or regulation as nrdA, SCO2015 or rstP; (iii) genes related to nitrogen or carbon metabolism: glnII, amtB, glnR (internal), SCO1086, ureA (SCO1236) or the bidirectional promoter region pyk1-SCO2015; (iv) genes for transcriptional regulators: whiB, malE-malR, SCO2686, SCO7302-7303, SCO1220-1221, SCO3978-3979, SCO0800-0801, or for sigma or anti-sigma factors: SCO3067-sig15, SCO7314; (v) genes encoding hypothetical conserved proteins (SCO5864); and (vi) genes related to arginine (argH, arcB; argG, argC and SCO1086) or pyrimidine (pyrR, pyrB and pyrA) biosynthesis as positive controls. The tested ARG boxes have Ri values that range from 5 (intergenic SCO3067-sig15 region) to 17 (upstream of argH). In most of the analyzed regions the ARG box (20 nucleotides) is unique, but in some cases two ARG boxes, separated from 1 to 208 nucleotides, can be identified. The two ARG boxes upstream of argH, arcB, argG and argC are adyacent.
Twenty-four fragments were bound by ArgR. In general a good correlation with the gene expression in transcriptomic studies was found (Figs. 8, 10 and Table 2). Only occasionally, probe shift was unclear or not present (i.e. for glnII, glnR, SCO7302-7303, SCO0800-0801 and SCO1220-1221). The occasional lack of relation between ARG box and differential expression in the microarray experiments will be discussed below.

Discussion
ArgR is known as the regulator of the arginine biosynthesis pathway and, in most bacteria, the argR gene is clustered with arginine biosynthesis or transport genes [7,35].
The transcriptomic and proteomic studies shown in this work indicated that in addition to arginine biosynthesis ArgR controls directly or indirectly 452 genes (contrasts 1 and 4, Fig. 2) involved in different aspects of nitrogen metabolism, purine and pyrimidine biosynthesis, cell morphology, and antibiotic production. More interesting is the effect of ArgR on the expression of genes encoding general regulators (GntR-, AbaA-and TetR-like regulators), sigma factors (BldN, SigR, AfsS), or two component regulatory systems, which could be mediated by a cascade regulatory mechanism. Binding assays validated the ARG boxes detected in Figure 10 and in most of the promoters shown in Table 2.
In E. coli arginine acts as co-repressor binding ArgR and controlling arginine biosynthesis. However, in Streptomyces the effect of arginine as co-repressor is weak and high levels of arginine (25 mM) were required to exert effect on the expression of only 27 genes (contrasts 2 and 3, Fig. 2).
Addition of arginine to S. coelicolor cultures reduced the production of actinorhodin and undecylprodigiosin by 92 and 99%, respectively, even though none of the pigmented antibiotic derives directly from arginine. Even more interesting was the observation that the DargR mutant showed a strong reduction (80%) in undecylprodigiosin production in relation to the wild-type strain and lacked detectable actinorhodin production. Therefore, ArgR is a regulator required for production of these secondary metabolites.
This drastic effect of ArgR on the production of both antibiotics might be mediated by the transcriptional response of the global positive regulator afsS. In addition, the effect on actinorhodin production could be correlated with the lower expression of actIII and actII-orf2 in the DargR mutant. The transporter protein encoded by actII-orf2 is essential for actinorhodin secretion [25] because accumulation of intracellular antibiotic feedback represses (chpC) (9) qDargR -SCO1800 (chpE) (9) qDargR -SCO2198 (glnA) (9) QDargR -SCO2718 (rdlA) (9) qDargR -SCO5584 (glnK) (10) QDargR -SCO6265 (scbR) (11) QDargR -SCO6282 (cpkI) (9) QDargR the biosynthetic genes. Identification of additional control mechanisms related to antibiotic production will require further experiments using RNA samples taken during the production phase, because the samples for this omics analyses were taken at an earlier time (32 h) than that of intense antibiotic production. The ArgR protein acts as a transcriptional regulator after binding to the promoters of the arg regulon genes. The binding sites are composed of imperfect palindromes, known as ARG boxes [36][37][38]. Both the argR genes and the ARG boxes are well conserved among different bacteria [33,39]. In E. coli the binding sites are composed of two ARG boxes of 18 bp that, with only one known exception, overlap the promoter. The separation between boxes is 3 bp except the 2 bp of separation of the argR operator [40,41]. ArgR functions in E. coli as a repressor on arginine biosynthesis and transport genes [14,42] and on the glutamate synthase operon [43], but as activator of the ast operon [44]. The B. subtilis AhrC protein -the ArgR orthologue -, which was found to bind S. coelicolor argC promoter [45], is also a repressor of the arginine biosynthesis genes [46,47] and an activator of the arginine catabolism genes in Bacillus [48]; similar regulatory behavior is found for ArgR in P. aeruginosa and Salmonella [35,49]. In B. subtilis argC gene two operators have been found. The operator with the highest affinity for AhrC is formed by two ARG boxes, separated by 11 bp. The second operator, within argC coding region, has a single ARG box. A model for the holo-AhrC-argC complex proposes that the AhrC hexamer interacts with the three ARG boxes [34]. Other operators formed by a single ARG box are found in the Bacillus rocA and rocD catabolic genes and show a lower affinity for AhrC [48].
We identified by DNase I footprinting the ArgR binding site that controls the argCJBDR operon of S. clavuligerus. This operator is formed by two 18 nucleotides ARG boxes separated by 2 bp (6). This arrangement is conserved in ArgR putative binding sites of other Streptomyces promoter and in the upstream sequences of S. coelicolor genes (argC, argG, arcB and argH) involved in arginine biosynthesis.
To fit with the experimental transcriptomic and gel-shift data presented in this work a new ArgR binding site model has been designed (Fig. 10C). This sequence is formed by a 20 nucleotidesingle ARG box and permitted the location of new ARG boxes with different Ri values. The presence of two tandem ARG boxes in the arginine operon genes suggests that these genes are more tightly controlled than others containing a single ARG box. In some of the DNA fragments tested by EMSA two separated ARG boxes were predicted (Table 2). In the intergenic malE-malR region the two putative boxes with Ri values of 9.3 and 8.6 are separated 83 nucleotides. The upstream region of nrdA and the intergenic SCO7302-SCO7303 or SCO2055-hisD regions also presented two putative sites with lower Ri. Additional studies will be necessary to demonstrate whether or not ArgR binds both or only one of these ARG boxes.
In most of the cases shown in Fig. 10, the gel shift correlated with a statistically significant differential expression of the gene whose promoter was tested, indicating that the model 2 is functional, although additional EMSA should be done with increasing amounts of protein to determine the different ArgR affinity to the operators. However, this was not always the case, as occurred with the amtB, SCO3067-sig15 or SCO2686 probes. This apparent lack of correlation in a few cases might be explained by the specific experimental conditions used in this study. The differential transcription observed in the microarray reflects the sampling of a 32-h culture grown in MG medium; however, in other media or culture times, these genes giving only in vitro binding might also show in vivo significant differential expression.
Gel-shift assays confirmed that the ArgR-dependent luciferase activity of these promoters results from the direct control of ArgR on these regions. As shown in Fig. 10 and Table 2, ArgR is a transcriptional regulator of pyrimidines (pyrR, pyrA, pyrB), nucleotides (i.e. SCO2015, nrdA), several amino acids (SCO4293), nitrogen metabolism (amtB, ureA) and several regulators (whiB, rstP, SCO2686 and SCO3979). ArgR also bound the intergenic region SCO3067-SCO3068 (for anti-anti-sigma and sigma factors) and the upstream region of SCO7314 encoding a sigma factor suggesting that many of the observed transcriptional effects are likely to be due to a cascade mechanism promoted by ArgR.
Regarding nitrogen control, putative ARG boxes were observed upstream of amtB (Ri 8.3), upstream of glnII (into the SCO2209 coding region, Ri 5.3) and in the 39 end of glnR (Ri 7.4). However, only the one located upstream of amtB was controlled by ArgR, as indicated by the luciferase assay and EMSA. Therefore, ArgR did not regulate glnR and the down-expression of glnII observed in the deleted mutant was not directly due to ArgR. Interestingly the (cpkJ) (9) QDargR (1) ND, not detected. (2) Values are ratios of specific luminescences at 32 h of culture. Values are shown when t-test p-value ,0.05 (4 biological replicates for each condition). NS, no significant differences. 1, 2, 3 and 4 as in Figure 2.
(3) Relative up or down expression of the gene in mutant strain. (4) Luminescence measures too low (,1 arbitrary unit). (5) Microarray results (when uncorrected p-value ,0.05 for the contrast, if not, no comparison was done) showed the same sign of regulation. (6) The probe does not correspond to the promoter region but to the upstream gene SCO2209 which contains the predicted ARG box. (7) The predicted ARG box is located in the 25 nucleotides of the 39end of SCO4159 (glnR). (8) Predicted polycistronic transcript from SCO1488 (pyrR). (9) Putative ARG boxes not found (Ri$3, 800-nt region centered at the start codon). (10) Predicted polycistronic transcript from SCO5583 (amtB). (11) Predicted ARG box overlapping the start codon, Ri 7.7 bits. doi:10.1371/journal.pone.0032697.t002 amtB operator also binds GlnR but this regulatory protein and ArgR have different binding sites [50].

Culture conditions
Growth and manipulation of Streptomyces strains were carried out according to standard procedures [51]. Spore suspensions were obtained in TBO medium: 20 g/L tomato paste, 20 g/L oat flakes, and 20 g/L agar, pH 6.5. S. coelicolor cultures were grown at 30uC, 300 rpm (2.5 cm orbit), in starch and glutamate defined MG medium [16] that contained 2.5 mM potassium phosphate and 25 mM arginine when indicated. Baffled flasks (500 mL), that contained 100 mL of MG medium, were inoculated with 10 6 spores/mL. Dry weight was determined in 2-mL culture samples that were washed twice with MilliQ water and dried at 65uC for 4 days. Arginine consumption was followed using the method of Hess et al. [52]. Cultures of plasmid-bearing cells were supplemented with ampicillin (50 mg/mL), chloramphenicol (25 mg/mL), kanamycin (25 mg/mL) or apramycin (50 mg/mL), as appropriate. E. coli DH5a was used as the general cloning host.
Cultures of 32 h were selected for the expression analysis due to the intense growth and good expression of amino acid biosynthesis pathways at this culture time. Construction of a mutant of S. coelicolor with argR gene deletion An argR-deleted mutant was constructed by PCR targeting [53] using oligonucleotides Coe-argR1 and Coe-argR2 on plasmid pTC123-aphII, a pTC182-derived plasmid [54] that contained the neo gene and a 5.7-kb SphI DNA insert carrying S. coelicolor argBDR genes as well as upstream and downstream sequences. After conjugation, four apramycin-resistant, kanamycin-sensitive S. coelicolor recombinants were characterized by Southern hybridization. We used as probes a SalI-PauI DNA fragment containing argR and the flanking sequences, and the apramycin resistance acc(3)IV gene. The hybridization pattern obtained confirmed the argR replacement deletion in the four identical clones that were named S. coelicolor DargR.

Nucleic acid isolation and purification
Samples (2 mL) from 32-h cultures of S. coelicolor M145 and S. coelicolor DargR grown in MG medium, with or without arginine, were stabilized with RNA Protect Bacteria Reagent (Qiagen). For RNA isolation, mycelia were treated with lysozyme (30 mg/mL); the lysates were extracted with phenol and then transferred to RNeasy Midi Spin Columns (Qiagen), according to the manufacturer's instructions. RNA preparations were incubated with Turbo DNase (Ambion) to eliminate chromosomal DNA contamination. Sample quantification was done with a NanoDrop ND-1000 UV-Vis Spectrophotometer. Total genomic DNA (gDNA) was isolated from a stationary-phase culture following the Kirby mix procedure [51].

PCR and semi-quantitative RT-PCR analysis
Oligonucleotide primers used in this study are shown in Table  S2. All PCRs were performed in a TGradient (Biometra) thermocycler using Platinum Pfx DNA Polymerase (Invitrogen). The dNTP mix contained a higher proportion of G-C (35% each) than A-T nucleotides (15% each), to improve the amplification of high G+C DNA content. PCR products subcloned into pBluescript II SK(+) were sequenced to check the amplification fidelity. Gene expression analysis by RT-PCR was done with the SuperScript One-Step System (Invitrogen) using 150 ng total RNA as a template. For semi-quantitative analysis, samples were taken at three-cycle intervals between cycles 24 and 33, to compare non-saturated PCR product formation. Negative controls were carried out with each set of primers and Platinum Taq DNA polymerase (Invitrogen) to verify the absence of contaminating DNA in the RNA preparations.

Purification of Strep-tagged ArgR
To purify the ArgR protein, plasmid pET-Strep-argR was constructed. The argR gene was amplified by PCR using oligonucleotides ArgR17/18 to place the Strep tag upstream of argR. The amplified product was subcloned in NdeI/HindIII digested plasmid pET-24a(+) (Novagen, Merck). The E. coli BL21(DE3)pLysS (Invitrogen) transformants carrying pET-Strep-argR were grown at 37uC to OD 0.6, induced with 1 mM IPTG, and the growth was continued at 20uC. After 18-20 h, the cells were harvested by centrifugation at 2640 rcf and kept at 280uC. Cells were broken with a Misonix XL-2000 sonifier, centrifuged at 16 100 rcf, and the supernatant was applied to a 1-mL StrepTrap HP column (GE Healthcare) and purified in an Akta Prime FPLC Protein Purification System (GE Healthcare) following the manufacturer's instructions.

Electrophoresis mobility shift assay (EMSA)
The promoters cloned in pBluescript SK+ were sequenced and amplified by PCR using specific or universal 6-FAM labelled oligonucleotides (Table S2). The amplification products were used for EMSA as follows: the reaction contained 5 mL buffer (10 mM Tris-HCl, pH 7.4, 5 mM MgCl 2 , 2.5 mM CaCl 2 , 250 mM KCl, 0.5 mM DTT, 10 mM L-arginine, pH 7.4), poly-(dIdC) 1,3 mg/ mL, 6-FAM-labelled probe 2 nM, glycerol 10% and Strep-ArgR protein 0.8 mM in a total volume of 15 mL. The reaction was maintained for 30 min at 30uC, and then the DNA was separated in a 5% acrylamide gel using 0.56 TBE as developing buffer at 50 V. The bands were visualized in an Ettan DIGE imager (GE Healthcare). In all cases, competition and specificity experiments were done with increasing amounts of unlabelled specific probe and with BSA. The argH promoter-probe was used to test the effect of L-arginine. ArgR affinity was higher in the presence of Larginine, so it was mantained in the binding reaction mixture.

Luciferase assay
For luciferase reporter analysis, promoter regions were amplified with primers containing NdeI and BamHI restriction sites (Table S2) to clone the promoters into the ATG codon of the luxA gene in pLUXAR-neo [55]. Cultures of S. coelicolor exconjugants harbouring the promoter-probe constructs were carried out in MG medium. Samples at 32, 47, 55 and 79 h were taken, spun down, kept frozen and processed simultaneously. Riboflavine was added to the cell suspension to improve the sensitivity of the luciferase assays, which were measured in a Luminoskan luminometer (Labsystems) [18,55]. At least two different cultures from the same strain were analyzed and measured by duplicate.
Labelling and microarray hybridizations S. coelicolor microarrays (SCo3 design) were obtained from the Functional Genomics Laboratory, Surrey University (UK). They contained duplicated probes (50-mer) for 7728 chromosomal genes (out of 7825). The experimental design used S. coelicolor M145 gDNA as a common reference. RNA was extracted from two nutritional states, MG and MG with 25 mM arginine, and from two strains, S. coelicolor M145 and S. coelicolor DargR. Four biological replicates were made for each condition. The Pronto! Universal Microarray Hybridization kit (Corning) was used for prehybridization of the slides. Labelling reactions were performed according to the recommendations described in http://www. surrey.ac.uk/SBMS/Fgenomics. Total RNA was labelled with Cy3-dCTP (Amersham) using random primers and Superscript II reverse transcriptase (Invitrogen). gDNA was labelled with Cy5-dCTP (Amersham) from random primers extended with the Klenow fragment of DNA polymerase (Roche). The final products were purified with MinElute columns (Qiagen) and labelling efficiencies were quantified spectrophotometrically. Cy3-cDNA (100 pmol) and Cy5-labeled gDNA (20 pmol) were mixed, vacuum dried and resuspended in 40 mL Pronto! Long Oligo/ cDNA Hybridization Solution (Corning), to be applied on the microarray surface. Hybridizations were carried out at 42uC and extended to 72 h to improve the quality of the results [56]. Washing, scanning with an Agilent DNA Microarray Scanner G2565BA, and image quantification were carried out as indicated previously [16].

Identification of differentially transcribed genes and transcription profile classification
Microarray data were normalized and analysed with the Bioconductor package limma [57,58]. Spot quality weights were estimated as indicated in Text S1. Local and global normalizations  Fig. 3. (B) EMSA of PCR-amplified DNA fragments containing putative ARG boxes located upstream of the indicated genes. In all cases, P indicates the free probe and B binding reaction using ArgR (0.8 mM protein). The size of the probes is indicated in the panels. (C) Sequence of the putative ARG box present upstream of the indicated gene and Ri value determined using the ARG box model represented on the right. The numbers in parenthesis indicate the nucleotides to the translation start codon. The argH, argC, argG and arcB ARG boxes are included for comparison. doi:10.1371/journal.pone.0032697.g010 were both used [59]. First, weighted medians of log 2 Cy3/Cy5 intensities were calculated for print-tip correction, and afterwards, global Loess was applied [57]. The normalized log 2 of Cy3/Cy5 intensities is referred to in this work as the M g value, which is proportional to the abundance of transcripts for a particular gene [60]. The information from within-array spot duplicates [61] and empirical array weights [62] were taken into account in the linear models [58]. The transcription results of the four experimental conditions were compared using five contrasts. For each contrast, p-values and M c values (log measure of the differential transcription) were calculated. False-discovery rate (FDR) correction for multiple testing was applied. For each contrast or comparison between two experimental conditions, a result was considered as statistically significant if the FDR-corrected p-value was ,0.05. A total of 459 genes showed statistically significant results in at least one contrast.
To classify the transcription profiles observed for this set of genes, we used the H r values, which summarized the results of the hypothesis tests. For each contrast and each gene, the H ri value of each gene was calculated as follows: (i) H ri = 0 indicated that the contrast result was not significant using uncorrected p-values (a = 0.05); and (ii) if the uncorrected p-value was ,0.05, then H ri = 1 (indicating upregulation) if the respective M c value was positive, or H ri = 21, for a negative M c value (downregulation). The set of 459 genes yielded a total of 52 observed combinations of their H r values. Visual inspection of the transcription profiles allowed us to group a subset of 365 genes, which showed the profiles with more likely biological meaning, into 5 main types (I-V) and 13 subtypes as shown in Fig. 3 (see also Dataset S1 and Text S1).

Bioinformatic analysis of ArgR binding sites
ArgR binding sites composed of two palindrome sequences, known as ARG boxes, have been identified previously in S. clavuligerus [6,7]. ARG boxes were easily identified upstream of the arginine biosynthesis genes argH, argG, argC and arcB of S. coelicolor. The sequences of ARG boxes and their complementaries -due to the symmetry of the palindromic site-were used to create information theory models by the Delila programs makebk, encode, rseq, dalvec, ri and makelogo [63,64]. To find new operator sites, the promoter regions (2300, +100 nt) of the S. coelicolor chromosome were scanned by means of the Patser algorithm and the RSA tools server [65].

Proteomic
The mycelia from 20-mL culture samples of S. coelicolor M145 and S. coelicolor DargR grown for 32 h in MG medium were harvested by centrifugation (10 min at 2640 rcf), immediately frozen in liquid nitrogen, and stored at 280uC until use. Proteomic analysis and Mass Spectrometry identification were done as described by Santamarta et al. [66]. Positive identifications were based on the MOWSE score algorithm. A MOWSE score of 52 or higher was significant at the 5% level or better, and proteins typically gave scores well above 70. The analysis was performed in the Proteomic Service of the National Center for Biotechnology (Madrid, Spain).  Text S1 Estimation of spot weights for microarray data analysis.

(DOC)
Dataset S1 Hypothesis-testing results for profile classification.