Response Regulator Heterodimer Formation Controls a Key Stage in Streptomyces Development

The orphan, atypical response regulators BldM and WhiI each play critical roles in Streptomyces differentiation. BldM is required for the formation of aerial hyphae, and WhiI is required for the differentiation of these reproductive structures into mature spores. To gain insight into BldM function, we defined the genome-wide BldM regulon using ChIP-Seq and transcriptional profiling. BldM target genes clustered into two groups based on their whi gene dependency. Expression of Group I genes depended on bldM but was independent of all the whi genes, and biochemical experiments showed that Group I promoters were controlled by a BldM homodimer. In contrast, Group II genes were expressed later than Group I genes and their expression depended not only on bldM but also on whiI and whiG (encoding the sigma factor that activates whiI). Additional ChIP-Seq analysis showed that BldM Group II genes were also direct targets of WhiI and that in vivo binding of WhiI to these promoters depended on BldM and vice versa. We go on to demonstrate that BldM and WhiI form a functional heterodimer that controls Group II promoters, serving to integrate signals from two distinct developmental pathways. The BldM-WhiI system thus exemplifies the potential of response regulator heterodimer formation as a mechanism to expand the signaling capabilities of bacterial cells.


Introduction
Two-component signal transduction systems are of central importance in regulating gene expression in bacteria. Canonically they consist of a response regulator, which functions as a homodimer, and a cognate sensor histidine kinase (which may also function as a cognate phosphatase). The activity of the kinase/phosphatase is modulated in response to a perceived stimulus. The sensor kinase autophosphorylates on a conserved histidine residue, and the phosphoryl group is then transferred to a conserved aspartate in the response regulator. The addition of the phosphoryl group stabilizes a conformation of the response regulator that drives an output response, most often the activation of gene expression. However, the intrinsic modularity of these systems has allowed bacteria to evolve variations on this basic theme, including more complex multicomponent phosphorelays, and changes in the nature of the response regulator effector domain such that the output can be, for example, an enzymatic activity rather than DNA binding [1][2][3]. Recognizing the diversity of mechanisms associated with these systems is therefore critical to understanding the full potential of the signaling capabilities of bacterial cells. This study concerns the behavior of two response regulators required for morphological development in the filamentous bacteria Streptomyces.
When streptomycete spores germinate, one or two germ tubes emerge and grow by tip extension and branching to form an extensive, multicellular vegetative mycelium [4][5][6]. Streptomycetes differentiate by forming specialized reproductive structures called aerial hyphae, which emerge from the colony surface into the air. The formation of aerial hyphae requires the activity of a class of developmental master regulators encoded by the bld (bald) genes [4][5][6]. Subsequently, in the most dramatic event of the lifecycle, each multigenomic aerial hypha arrests tip growth and undergoes a massive, synchronous septation event, giving rise to ,50-100 unigenomic prespore compartments that ultimately develop into mature, pigmented exospores [4][5][6]. The differentiation of aerial hyphae into mature spores is coordinated by the activity of a second class of developmental master regulators encoded by the whi (white) genes. The focus of this work is the interaction between two of these global regulators, BldM and WhiI.
BldM and WhiI are both atypical response regulators (ARRs). In canonical response regulators, the aspartate residue that is subject to phosphorylation sits in a highly conserved pocket within the N-terminal receiver domain. ARRs usually lack essential residues within this phosphorylation pocket, suggesting that their activity is not controlled by phosphorylation [7][8][9][10][11]. WhiI has a degenerate phosphorylation pocket, lacking a universally conserved lysine and one of a pair of adjacent aspartate residues essential for binding Mg 2+ [12,13]. Although BldM does have a conserved phosphorylation pocket, a bldM allele carrying a D54A substitution at the putative site of phosporylation fully complements a bldM null mutant [14]. Moreover, BldM could not be phosphorylated in vitro [14]. Finally, BldM and WhiI are both 'orphan' response regulators -their genes are not adjacent to a sensor kinase gene, as is most often the case for canonical response regulators. Taken together, these observations strongly suggest that BldM and WhiI are not controlled by phosphorylation as part of conventional two-component systems. BldM and WhiI both belong to the NarL/FixJ subfamily of response regulators. bldM and whiI are found in all sequenced streptomycete genomes and their chromosomal context is conserved throughout. Strikingly, the amino acid sequence of BldM is 100% identical across all sequenced streptomycetes (the only response regulator that is 100% identical across all streptomycetes). WhiI is at least 93% identical, with amino acid variations found mainly in the linker region between the degenerate receiver domain and the DNA-binding domain. All sequenced WhiIs have a degenerate phosphorylation pocket.
During differentiation, the whiI and bldM genes are activated by two cognate, development-specific sigma factors, s WhiG and s BldN , respectively. whiI expression is activated from a single s WhiG target promoter, and thus whiI is not expressed in a whiG mutant [12]. s BldN directs transcription of the p1 promoter of bldM (the other promoter, bldMp2, is s BldN -independent, and so transcription of bldM is developmentally activated from only one of its two promoters in a bldN mutant [15,16]). In addition to bldM, the other key targets of s BldN are the genes encoding the chaplins and rodlins, the major proteins of the hydrophobic sheath that coats the aerial hyphae and spores in Streptomyces [16][17][18][19].
Streptomyces venezuelae has recently emerged as an attractive new model system for the analysis of Streptomyces development because it sporulates in liquid culture [16,20]. Here we take advantage of the S. venezuelae system to apply global microarray transcriptional profiling and ChIP-Seq to characterize the BldM and WhiI regulons. Through this route we go on to show that a key stage in Streptomyces development is controlled by response regulator heterodimer formation between BldM and WhiI, and to greatly expand our understanding of the regulatory network that controls morphological differentiation in these multicellular bacteria. The BldM-WhiI system thus exemplifies the potential of response regulator heterodimer formation as a mechanism to expand the signaling capabilities of bacterial cells.

BldM directly activates the expression of key developmental genes
Having established conditions in which S. venezuelae sporulates abundantly in liquid culture [16,20], immunoblotting of samples taken at 14, 15 and 16 h of growth in MYM liquid sporulation medium showed that BldM was abundant at each of these time points ( Figure S1). To gain greater insight into BldM function, we defined the genome-wide BldM regulon using ChIP-Seq. As described in Materials and Methods, wild-type S. venezuelae was subjected to formaldehyde cross-linking, lysis and sonication after 16 h of growth. After immunoprecipitation using a BldM-specific polyclonal antibody, the resulting DNA was subjected to deep sequencing. As a negative control, a ChIP-Seq experiment was performed on the congenic bldM null mutant. Several wellcharacterized developmental loci, including ssgR, rshA, smeA-sffA, whiB and whiE, were among the direct BldM targets identified.
Next, in order to determine how BldM influences the expression of its target genes, wild-type S. venezuelae and the congenic DbldM mutant were subjected to time-resolved, genome-wide transcriptional profiling during vegetative growth and sporulation. Strains were grown under the same conditions used for the ChIP-Seq experiments. RNA samples were prepared at 2-hour intervals from 8 to 20 hours, by which time sporulation was nearing completion, and following cDNA synthesis and labeling, samples were hybridized to Affymetrix DNA microarrays. Three independent biological replicates were performed for each strain, and analysis of the resulting data showed that the expression of 131 direct BldM targets was significantly down-regulated in the DbldM mutant (p,0.01) in comparison to the wild type. In contrast, only six genes were up-regulated in the DbldM mutant (p,0.01) in comparison to the wild type. These results suggest that BldM functions mainly as a transcriptional activator.
Many BldM target genes cluster into two discrete groups based on whi gene dependency We next determined the time-resolved transcriptional profiles of the BldM target genes in seven constructed white mutants: DwhiA, DwhiB, DwhiD, DwhiG, DwhiH and DwhiI. Strikingly, many of the BldM target genes clustered into two well-defined groups according to their dependencies on the whi genes. Group I genes consisted of developmentally induced genes that depend on bldM, but were activated normally in all the whi mutants ( Figure 1A and Table S1). Group II BldM target genes were also developmentally induced, but depended not only on bldM, but also on whiG and whiI ( Figure 1B and Table S2).

Group I genes-identification of a consensus Group I BldM binding site
To gain further insight into Group I binding sites, we fed the sequences of Group I promoter regions into the MEME algorithm [21] to search for over-represented sequences, using as input the entire intergenic region in each case. This analysis revealed a wellconserved copy of a 16 bp palindromic sequence, 59-TCACcCgnncGgGTGA-39, for which the sequence logo is shown in Figure 2A. The palindromic nature of this sequence would be consistent with BldM binding as a homodimer to Group I promoters.
To test the validity of the MEME output, and to confirm and extend the ChIP-Seq analysis, we overexpressed and purified BldM from E. coli as a soluble, N-terminally His 6 -tagged protein.
The resulting BldM protein was used in DNase I footprinting analysis on the intergenic regions upstream of two Group I BldM targets, sven1998 and sven4150. In both cases, BldM protected a region containing a well-conserved copy of the palindromic, MEME-predicted binding site ( Figure 2B), consistent with this sequence serving as a high-affinity binding site for a BldM homodimer.

Author Summary
Two-component signal transduction systems are a primary means of regulating gene expression in bacteria. Recognizing the diversity of mechanisms associated with these systems is therefore critical to understanding the full signaling potential of bacterial cells. We have analyzed the behavior of two orphan, atypical response regulators that play key roles in controlling morphological differentiation in the filamentous bacteria Streptomyces-BldM and WhiI. We demonstrate that BldM activates its Group I target promoters as a homodimer, but that it subsequently activates its Group II target promoters by forming a functional heterodimer with WhiI. BldM-WhiI heterodimer formation thus represents an unusual mechanism for the coactivation of target genes and the integration of regulatory signals at promoters, enhancing the known repertoire of signaling capabilities associated with twocomponent systems.
Group II BldM target genes were expressed later than Group I genes (see insets in Figure 1). Further, and in contrast to Group I, the expression of Group II genes depended not only on bldM but also on whiG and whiI ( Figure 1 and Table S2). It is straightforward to account for the dependence of Group II gene expression on whiG. In S. coelicolor, whiI expression is activated from a single s WhiG target promoter, and thus whiI is not expressed in a whiG mutant [12]. This s WhiG target promoter appears well conserved at the sequence level in S. venezuelae ( Figure S2A) and whiI is not expressed in an S. venezuelae whiG mutant ( Figure S2B). Thus all genes that depend on whiI must necessarily also depend on whiG. Expression of whiI was not significantly affected in the DbldM mutant and vice versa (p,0.01) ( Figure S2B), implying independent s WhiG -WhiI and s BldN -BldM regulatory pathways. The challenge then was to determine why Group II BldM target genes depend on whiI.

Group II genes are direct in vivo targets of WhiI
To determine if the dependence of Group II BldM target genes on WhiI was direct or indirect, we characterized the in vivo WhiI binding sites across the S. venezuelae genome using ChIP-Seq. Wild-type S. venezuelae was harvested at 16 h of growth and treated as described for the BldM ChIP-Seq, except that a WhiIspecific polyclonal antibody was used. As a control, a ChIP-Seq experiment was performed using the congenic whiI null mutant. The data showed that all of the BldM Group II targets were also direct targets of WhiI ( Figure 3). As an independent confirmation, we repeated the ChIP-Seq experiment using a FLAG-tagged WhiI protein. An N-terminally 3xFLAG-tagged allele of whiI (TF-WhiI) was constructed such that it was expressed from its native promoter and cloned into the single-copy vector pMS82, which integrates site-specifically into the chromosome at the phage WBT1 attB site [22]. This construct fully complemented the phenotype of the whiI null mutant ( Figure S3), and the complemented strain was used for the ChIP-Seq experiment, now using wild-type S. venezuelae as the negative control. The ChIP-Seq results seen using FLAG immunoprecipitation were almost identical to those obtained using WhiI polyclonal antibodies, confirming that Group II genes are directly regulated by both BldM and WhiI (Table S2).

At Group II promoters, in vivo WhiI binding depends on BldM and vice versa
Our data showed that expression of Group II genes depends on bldM and whiI and that both BldM and WhiI bind directly to the RNA samples were prepared at 2-hour intervals from 8 to 20 hours, by which time sporulation was nearing completion. For each panel, the x-axis indicates the age of the culture in hours, and the y-axis indicates the per gene normalized transcript abundance (log 2 ). Group-I genes (blue) depend on bldM but are independent of all the whi genes. Group-II genes (red) depend on bldM but also depend whiG and whiI. The average expression profile is indicated by the black line. Group I genes are activated at least two hours earlier than Group-II genes (see insets). Red bars indicate vegetative growth (V), fragmentation of mycelium (F), and spore formation (S). doi:10.1371/journal.pgen.1004554.g001 promoters of these genes. One possible model consistent with these observations would be that BldM and WhiI co-activate Group II promoters by binding as two separate homodimers. An alternative model would be that these two proteins activate Group II promoters by binding as a functional BldM-WhiI heterodimer. To begin to differentiate between these models, we performed BldM ChIP-Seq in a DwhiI mutant and WhiI ChIP-Seq in a DbldM mutant, using the same conditions described above. In a DwhiI mutant BldM binding was still observed at all Group I promoters (which depend solely on bldM), but no BldM binding was seen at Group II promoters ( Figure 3). Equally, no WhiI binding to Group II promoters was observed in a DbldM mutant ( Figure 3). Thus, in vivo, BldM and WhiI show mutual dependence for binding to Group II promoters.

BldM and WhiI interact in E. coli
To explore the possibility of BldM-WhiI heterodimer formation, we tested BldM and WhiI for direct interaction in E. coli using a bacterial two-hybrid (BACTH) system [23]. bldM was fused to the gene encoding the T18 fragment of adenylate cyclase in the vector pUT18 such that BldM was at the N-terminus of the fusion protein, and was also fused to the gene encoding the T25 fragment of adenylate cyclase in the vector pKT25 such that BldM was at the C-terminus of the fusion protein. Parallel pUT18 and pKT25 constructs were made carrying fusions to WhiI.
Interacting pairs of proteins were screened initially by transforming E. coli BTH101 with the appropriate plasmids and monitoring restoration of adenylate cyclase activity on X-gal indicator plates; clones of each pair were then assayed for bgalactosidase activity (Figure 4). Interaction of BldM with itself was readily observed (,750 Miller units), but a stronger interaction (,3000 Miller units), was observed between BldM and WhiI, regardless of which protein was fused to the T18 fragment of adenylate cyclase and which was fused to the T25 fragment ( Figure 4). WhiI showed no detectable interaction with itself ( Figure 4).

BldM and WhiI interact in Streptomyces
To confirm and extend the BACTH analysis, we tested the interaction of BldM and WhiI in Streptomyces by coimmunoprecipitation. A C-terminally 3xFLAG-tagged allele of bldM (BldM-TF) was constructed such that it was expressed from its native promoter and cloned into the integrative vector pMS82 [22]. This construct complemented the phenotype of the DbldM null mutant to wild-type levels of sporulation ( Figure S3). This strain was used in conjunction with the DwhiI mutant complemented with the Nterminally 3xFLAG-tagged allele of whiI (TF-WhiI) described above. The BldM-TF and TF-WhiI proteins were immunoprecipitated directly from 16 h MYM liquid cultures using the M2 anti-FLAG monoclonal antibody. WhiI coimmunoprecipitated with BldM-TF and BldM coimmunoprecipitated with TF-WhiI, but neither was detected in the negative controls when the wildtype strain was used ( Figure S4). Thus BldM and WhiI interact in Streptomyces in vivo.

Co-expression of BldM rescues WhiI from inclusion bodies
An early frustration in the in vitro analysis of WhiI function was that it overexpressed in an insoluble form in E. coli under all conditions tested. The realization that WhiI might act as part of a functional BldM-WhiI heterodimer led us to try an alternative approach. Where proteins form a complex, it is often observed that individual components are insoluble when expressed in isolation but become soluble when expressed with their cognate partner protein. Two examples are the a and b subunits of lambda integrase [24], and Streptomyces s BldN and its cognate anti-sigma factor, RsbN [16]. Accordingly, BldM and WhiI were coexpressed in E. coli using the pETDuet-1 system (Novagen). Initially, BldM was N-terminally His 6 -tagged and WhiI was left untagged. Co-expression of BldM was found to solubilize WhiI completely. Further, when His 6 -BldM was purified on a HisTrap Ni column, WhiI copurified with His 6 -BldM in approximately equal amounts, despite the fact that WhiI was untagged ( Figure 5A) (the identity of the two proteins was confirmed by tryptic mass fingerprinting). Thus BldM rescues WhiI from inclusion bodies and the two proteins copurify in approximately stoichiometric amounts via a His 6 -tag present on BldM only. This approach was extended by coexpressing BldM and WhiI carrying two compatible affinity tags. N-terminally His 6 -tagged BldM and N-terminally StrepII-tagged WhiI were co-expressed from the pETDuet-1 vector. As before, both BldM and WhiI were found in the soluble fraction. The BldM-WhiI complex was then purified over consecutive HisTrap Ni and StrepTrap HP affinity columns and the BldM and WhiI proteins were found to be present in stoichiometric amounts in the resulting preparation ( Figure 5B).

In silico analysis of Group II promoters
To further understand Group II binding sites, we searched for over-represented sequences in Group II promoter regions using MEME [21], again using as input the entire intergenic region in each case. This analysis revealed a well-conserved 16 bp non-palindromic sequence, 59-TGnnCCGnnCGGGTGA-39, for which the sequence logo is shown in Figure 6A. Strikingly, the 39 half of the Group II logo was equivalent to a half-site of the Group I palindrome, but the other half was different in sequence, potentially consistent with a BldM-WhiI heterodimer binding to Group II targets.
BldM-WhiI binds to the MEME-predicted motifs in Group II promoters but BldM or GST-WhiI do not To directly test the model that Group II promoters are controlled by a functional BldM-WhiI heterodimer, and to validate the MEME-predicted binding motif for these promoters, the doubly-tagged BldM-WhiI that had been purified over consecutive HisTrap Ni and StrepTrap HP affinity columns was used in DNase I footprinting analysis on the intergenic regions upstream of two Group II targets, sven1263 and murA2 (sven5810). BldM-WhiI footprinted on both promoters and in each case the protection region contained a copy of the nonpalindromic, MEME-predicted binding site ( Figure 6B), consistent with this sequence serving as a high-affinity binding site for a BldM-WhiI heterodimer. In contrast, neither BldM alone nor WhiI (produced as a soluble GST fusion), footprinted on either promoter ( Figure 6B).

Discussion
The focus of this study is to elucidate the mechanism underlying the direct co-activation by BldM and WhiI of the Group II genes, required for the late stages of development. Our data show that BldM activates transcription of these Group II genes as a BldM-WhiI heterodimer, while activating transcription of the Group I genes required for the early stages of development as a BldM homodimer. This work also significantly expands our knowledge of the regulatory network that controls morphological differentiation in Streptomyces (Figure 7), an advance made possible by exploiting S. venezuelae as a new model species for the genus.

Group I genes
BldM homodimer activates several genes known to play key roles in the differentiation of aerial hyphae into spores, including whiB and ssgR. In addition to their positive regulation by BldM homodimer, these genes are also subject to repression during vegetative growth by the master regulator BldD, as is bldM itself (Figure 7) [25]. SsgA and SsgB are homologous proteins directly involved in the positive control of cell division in Streptomyces [26]. Sporogenic aerial hyphae undergo a synchronous round of cell division, initiated by the polymerization of a ladder of 50 or more FtsZ rings. SsgA and SsgB function in the recruitment and accurate positioning of these FtsZ rings, and DssgA and DssgB mutants of S. coelicolor lack sporulation septa [26][27][28]. ssgR encodes an IclR-family transcriptional regulator that directly activates the expression of ssgA in S. coelicolor [29]. Here we show that ssgR is directly activated by, and is completely dependent on BldM in S. venezuelae. ssgB is developmentally induced in S. venezuelae, but despite being a direct BldM target, its expression is only weakly affected in the DbldM mutant, suggesting complex regulation of this gene.
BldM homodimer also activates expression of whiB, which plays a vital role in developmentally controlled cell division. whiB null mutants fail to arrest aerial tip growth, the normal prelude to sporulation, and are completely blocked in the initiation of sporulation sepatation, producing abnormally long, undivided aerial hyphae [30]. WhiB is the founding member of a family of proteins confined to the actinomycetes, and several of these WhiBlike (Wbl) proteins have been shown to play key roles in the biology of streptomycetes and mycobacteria. Wbl proteins carry a [4Fe-4S] iron-sulfur cluster coordinated by four invariant cysteines in a C(X 29 )C(X 2 )C(X 5 )C motif [31][32][33][34][35], and although the biochemical role of these unusual proteins has been controversial [36], it seems increasingly certain that they function as transcription factors [33,[37][38].

Group II genes
Among the Group II targets controlled by the BldM-WhiI heterodimer are two loci with well characterized roles in sporulation: smeA-sffA and whiE. The smeA-sffA operon encodes a DNA translocase (SffA) involved in chromosome segregation into spores that is specifically targeted to sporulation septa by the small membrane protein SmeA [39]. Deletion of smeA-sffA in S. coelicolor results in a defect in spore chromosome segregation and has pleiotropic effects on spore maturation [39]. Like the Group I targets ssgR and whiB, expression of the smeA-sffA operon is also repressed by the master regulator BldD during vegetative growth [25]. whiE is a complex locus that specifies the spore pigment. The structure of the spore pigment has not been determined in any Streptomyces species but its polyketide nature was first predicted from the sequence of the whiE locus in S. coelicolor, because it encodes proteins that closely resemble the components of type II polyketide synthases involved in the synthesis of aromatic antibiotics [40][41][42]. Based on their coordinate regulation and proposed functions, we predict the whiE locus of S. venezuelae consists of an operon of seven genes (sven6798-6792) and the divergently transcribed gene sven6799. Two distinct ChIP-Seq peaks were seen in the intergenic region separating sven6799 and the sven6798-6792 operon, and all eight genes fail to be expressed in the bldM and whiI mutants, implying the BldM-WhiI heterodimer controls expression of the entire locus.

All WhiI-regulated genes belong to Group II
The work presented here suggests that there is no set of genes regulated by a WhiI homodimer and that WhiI functions as an auxiliary protein to modulate BldM binding specificity through heterodimerization. With no exceptions, all the genes down regulated in a DwhiI mutant were also down regulated in a DbldM mutant. Although some promoters were exclusively enriched as peaks in the WhiI ChIP-Seq experiment, without exception the transcriptional profile of such targets was unaffected in a DwhiI mutant, showing that WhiI has no regulatory influence on these genes. Further, in a bldM null mutant, some WhiI peaks are seen in ChIP-Seq, but these sites, often internal to ORFs, show no correlation with the wild-type WhiI regulon and the targets lacked a consensus binding motif. These results suggest that, in the absence of BldM, any DNA binding by WhiI is aberrant and unrelated to its behavior in the wild type. In contrast, in a DwhiI mutant, BldM fails to bind to Group II promoters but binds normally to its Group I promoters.
In a recent study, evidence was presented that the DNA-binding domain of S. coelicolor WhiI (in the absence of the receiver domain) can bind in vitro to the promoter of the sco3900-sco3899 operon encoding a transcriptional regulator (InoR) and an inositol 1-phosphate synthase (InoA), respectively [43]. Our ChIP-Seq data show that WhiI does not regulate these genes in S. venezuelae. Further, in wild-type S. venezuelae both genes are actively expressed during vegetative growth but are downregulated during development, and this expression pattern is unaffected in a whiI mutant.

The potential of response regulator heterodimerization as a regulatory mechanism
Transcription factor heterodimerization can coordinate responses to different cues by integrating signals from distinct regulatory pathways. Although heterodimerization is prevalent as a regulatory mechanism in eukaryotes [44], it is rare in bacteria. Prior to the work described here, the only response regulator reported to heterodimerize with an auxiliary regulator was RcsB.
In Escherichia coli, the typical response regulator RcsB plays a central role in the regulation of capsule synthesis. Once phosphorylated by the histidine kinase RcsD, RcsB directly activates target genes including rprA, osmC, osmB and ftsZ, functioning as a homodimer [45,46]. It also activates exopolysaccharide synthesis genes, required for capsule formation, as a heterodimer with RcsA, which is distantly related to response regulators (like RcsB, RcsA has a typical LuxRtype C-terminal DNA-binding domain, but its N-terminal domain is not related to typical response-regulator receiver domains). Like WhiI, RcsA appears to lack the capacity to activate genes by itself, and therefore functions solely as a modulator of RcsB binding specificity. RcsA is actively degraded by the Lon protease and in lon mutants capsule genes are highly upregulated causing a mucoid phenotype, due to enhanced activation by the stabilized RcsB-RcsA heterodimer [45,46]. The capacity of RcsB to complex with other transcription factors is not limited to its interaction with RcsA, since RcsB also forms functional heterodimers with BglJ [47] and with GadE [48]. Using in vitro FRET analysis, weak hetero-pair interactions were detected between several members of the OmpR sub-family of response regulators from E. coli [49]. While these in vitro data may suggest potential crosstalk between distinct signaling pathways, their physiological significance has yet to be demonstrated.
The activities of numerous bacterial promoters respond to multiple cues, and there are many examples of promoters that depend on two activators for their activity. Several different regulatory mechanisms underpinning such codependence have been identified [50,51]. The most widely documented is found at promoters where both activators bind independently, and both activators make independent contacts with RNA polymerase. However, there are rare examples of coactivators that exhibit cooperative binding, such as MelR and CRP at the E. coli melAB promoter [52]. There are also examples in which DNA binding by a secondary activator leads to the repositioning of the primary activator from a site where it cannot activate transcription to a site where it can, such as the repositioning of MalT by CRP at the malK promoter [53]. Response regulator heterodimer formation provides a new model for coactivation of target genes and the integration of regulatory signals at promoters. BldM-WhiI heterodimer formation serves to integrate signals from two independent pathways (s WhiG -WhiI and s BldN -BldM) and it may also function as a timing device, since Group II genes are activated later than Group I genes. Thus the BldM-WhiI system exemplifies the potential of response regulator heterodimer formation as a mechanism to expand the signaling capabilities of bacterial cells.

Bacterial strains, plasmids, oligonucleotides and growth conditions
Bacterial strains and plasmids are listed in Table S3 and the oligonucleotide primers with corresponding restriction sites used in   Table S4. For microarray and ChIP-Seq experiments, S. venezuelae strains were grown at 30uC in MYM liquid sporulation medium [16] made with 50% tap water and supplemented with 200 ml trace element solution [54] per 100 ml. The phenotypes of mutants and complemented strains were scored after 3-4 days growth on MYM-agar at 30uC. bldM and whiI deletion mutants were constructed by 'Redirect' PCR targeting [55] and their chromosomal structures were confirmed by PCR analysis and by Southern hybridization using the parental cosmids as probes.

ChIP (Chromatin immunoprecipitation)-Seq
For each strain, two flasks containing 35 ml of MYM were inoculated with spores (or mycelium in case of the DbldM mutant) to give an OD 600 ,0.35 after 8 h of growth. The crosslinking reagent formaldehyde was added to a final concentration of 1% (v/v) to the cultures at 16 h of growth and incubated at 30uC with shaking for 30 min before glycine was added to a final concentration of 125 mM to quench the crosslinking reaction. The samples were incubated at room temperature for 5 min and washed twice in PBS buffer pH 7.4 (Sigma). Mycelial pellets were resuspended in 0.5 ml lysis buffer (10 mM Tris-HCl pH 8, 50 mM NaCl, 15 mg/ml lysozyme, 1x protease inhibitor) and incubated at 37uC for 20 min. The lysate was resuspended in 0.5 ml IP buffer (100 mM Tris-HCl pH 8, 250 mM NaCl, 0.1% Triton-X-100, 1x protease inhibitor) and the lysate was kept on ice for 2 min before sonication. The samples were subjected to seven cycles of sonication, 15 s each, at 10 microns, to shear the chromosome into fragments ranging in size from 300-1000 bp. The sample was then centrifuged twice at top speed, 4uC for 15 minutes to clear cell extracts. To pre-clear non-specific binding, 90 ml protein A sepharose (Sigma) was added to cell lysate (about 900 ml) and incubated for 1 h at 4uC with mixing. The beads were cleared by centrifugation at top speed for 15 min. 100 ml BldM or WhiI antibodies were added to the corresponding cell lysates overnight at 4uC with mixing. 100 ml Protein A Sepharose 1:1 suspension was added to immunoprecipitate antibody-BldM or WhiI chromatin complexes and incubated for 4 h at 4uC with mixing. The samples were centrifuged at 3500 rpm for 30 s and the beads were washed four times with IP buffer. The pellets were eluted in 150 ml IP elution buffer (50 mM Tris-HCl pH 7.6, 10 mM EDTA, 1% SDS) overnight at 65uC to reverse crosslink. The samples were centrifuged at top speed for 5 min to remove the beads and the pellets were re-extracted with 50 ml TE buffer (10 mM Tris-HCl pH 7.4, 1 mM EDTA). The supernatants were combined and incubated with 3 ml 10 mg/ml proteinase K (Roche) for 2 h at 55uC. The samples were extracted twice with phenol-chloroform to remove protein followed by chloroform extraction to remove traces of phenol and purified with Qiaquick columns (Qiagen). The IP DNA was eluted in 50 ml EB buffer (Qiagen). Sequencing libraries were generated and the IP DNA was sequenced as described previously [20]. The BayesPeak package was used to identify significantly enriched regions and the default parameters were applied [56].

Microarray transcriptional profiling
Microarray transcriptional profiling experiments were carried out as described previously [16,20]. Multi-experiment viewer software (MeV 4.8) was used for viewing and statistical analysis [57]. The non-parametric tool 'Rank Products' [58] was used in MeV to assign 'down regulated', 'up regulated' and 'not significant' genes based on expression at 16, 18 and 20 h of growth. Group I genes were defined as direct BldM ChIP-Seq targets that were significantly down regulated in DbldM and not significantly changed in all Dwhi mutants (Table S1). Group II genes were defined as direct ChIP-Seq targets of both BldM and WhiI that were significantly down regulated in DbldM, DwhiG and DwhiI (Table S2).

Protein expression and purification
The open reading frame of interest was PCR-amplified using Expand High-Fidelity DNA polymerase (Roche). Plasmids containing the correct inserts were confirmed by sequencing and introduced into electrocompetent E. coli BL21(DE3)/pLysS. The transformed cells were spread on LB-carbenicillin/chloramphenicol and one colony was used for inoculation. Proteins were expressed in two 2.5 litre volumetric flasks each containing 400 ml LB culture and expression was induced with 0.25 mM IPTG. The optimised temperature for expression varied with the protein: His 6 -BldM was expressed at 25uC for 5 h; GST-WhiI was expressed at 15uC overnight; His 6 -BldM/WhiI or S II -WhiI were co-expressed at 30uC for 5 h. The pellets were lysed in a buffer containing 50 mM Tris-HCl pH 8, 250 mM NaCl, 10% glycerol, 0.1% Triton X100, protease inhibitor (complete mini, EDTA-free, Figure 7. Schematic representation summarizing the regulatory network involved in controlling BldM and WhiI expression and the activation of Group-I and Group-II genes. During vegetative growth, BldD represses bldC, bldN, whiG and bldM. At the onset of differentiation, this repression is relieved, s BldN activates bldM and s WhiG activates whiI. BldM homodimer activates early sporulation genes such as whiB and ssgR. Subsequently, BldM and WhiI form a heterodimer that activates late sporulation genes such as whiE and smeA-sffA. Thus heterodimer formation serves to integrate signals from two independent pathways (s WhiG -WhiI and s BldN -BldM). bldM expression is tightly controlled by at least four global regulators -BldD, BldC, s BldN and PhoP (a response regulator activated during phosphate depletion [59] Roche) and incubated at room temperature for 20 min. HisTrap HP Ni and StrepTrap HP affinity columns (GE Healthcare) were used to purify the His 6 -and S II -tagged proteins in a tandem manner. The Gst-WhiI was purified with 1 ml GSTrap FF column (GE Healthcare).

DNase I footprinting
Singly 32 P end-labelled probes (Table S4) were generated by PCR and purified using Qiaquick columns (Qiagen). Transcription factors were incubated with probe DNA (,150,000 cpm) for 30 min at room temperature in 40 ml reaction buffer [50 mM Tris-HCl pH 7.5, 100 mM NaCl, 10% glycerol, 10 mM MgCl 2 , 2 mM dithiothreitol and 1 mg/reaction poly(dI-dC)], prior to treatment with 1 U DNase I (Promega) for 30-50 s in the case of group-II promoters and 3 U DNase I for 15-20 s in the case of group-I promoters. Reactions were terminated with 140 ml of stop buffer (192 mM sodium acetate, 32 mM EDTA, 0.14% SDS, 70 mg/ml yeast tRNA) and samples were extracted with phenol-chloroform prior to ethanol precipitation. Footprinting samples were loaded on 6% polyacrylamide sequencing gels, next to a G+A ladders prepared according to the Sure Track footprinting kit (Amersham Pharmacia Biotech).
Immunoprecipitation of BldM-FLAG and FLAG-WhiI proteins from S. venezuelae C-terminally 36FLAG-tagged bldM expressed from the native promoter was used to complement the DbldM mutant and Nterminal 36FLAG-tagged whiI carrying native promoter was used to complement the DwhiI mutant. 50 ml dense spore suspension was used to inoculate 300 ml MYM in 2 litre flasks with spring baffles. After 17 h growth, cultures were harvested by centrifugation at 6000 rpm for 15 min at 4uC. Pellets were lysed in a buffer containing 50 mM Tris-HCl pH 8, 250 mM NaCl, 10% glycerol, 0.1% Triton X100, protease inhibitor (complete mini, EDTA-free, Roche) and sonicated for 5 cycles at 15-micron amplitude for 20 s. FLAG-tagged proteins were immunoprecipitiated using M2 beads [anti-FLAG antibodies covalently attached to agarose beads (Sigma)], the beads were washed using TBS buffer containing protease inhibitor (Roche) and 0.05% Triton X100, and the protein was eluted using FLAG peptides as recommended by the manufacturer.

Preparation of crude cell extracts and immunoblot analysis
S. venezuelae strains were grown in MYM medium. 10 ml samples were taken at 14, 15, and 16 hours of growth. Strains were harvested by centrifugation at 3000 rpm for 1 min, washed in 5 ml ice-cold washing buffer (20 mM Tris pH 8.0, 5 mM EDTA) and resuspended in 0.4 ml of ice-cold sonication buffer [20 mM Tris pH 8.0, 5 mM EDTA, 1x protease inhibitor (Roche)]. Samples were sonicated immediately for 5 cycles, 20 s at 10 microns with 1 min intervals of ice incubation, then centrifuged at 13000 rpm at 4uC for 15 min to remove cell debris. Protein concentrations of the supernatant crude cell extracts were measured by Bradford assay and samples (10 mg protein) were separated on a 12.5% SDS-PA gel and blotted onto nitrocellulose membrane. The membrane was incubated in blocking solution [10% dried milk powder in TBS (0.05 M Tris, 0.9% NaCl, pH 7.6, 0.1% Tween)] overnight and then incubated for 1 h at room temperature with the 1/2500 dilution of anti-BldM antiserum in blocking solution. The membrane was rinsed (twice for 10 min) in TBS and then incubated for 1 h with 1/5,000 dilutions of horseradish peroxidase-linked goat anti-rabbit immunoglobulin G antibody (GE Healthcare). Blots were developed using the ECL enhanced chemiluminescence system from GE Healthcare and were typically exposed to X-ray film for between 30 s and 5 min.  Figure S4 Co-immunoprecipitation of BldM and WhiI. A. The DwhiI mutant complemented with the N-terminally 3xFLAG-tagged allele of whiI was grown for 16 h in MYM liquid sporulation medium, and FLAG-WhiI was immunoprecipitated using M2 antibody. Polyclonal BldM antibody was used to detect the presence of BldM, and immunoprecipitates from WT and BldM-FLAG strains were used as negative and positive controls, respectively. B. The DbldM mutant complemented with the C-terminally 3xFLAGtagged allele of bldM was grown for 16 h in MYM liquid sporulation medium and BldM-FLAG was immunoprecipitated as described above. Polyclonal WhiI antibody was used to detect the presence of WhiI, and immunoprecipitates from WT and FLAG-WhiI strains were used as negative and positive controls, respectively. (TIF)