TPR-mediated interaction of RapC with ComA inhibits response regulator-DNA binding for competence development in Bacillus subtilis



The Bacillus subtilis Rap family of proteins are characterized by protein–protein interaction modules containing the so-called tetratricopeptide repeats (TPRs). The six TPR motifs of RapC mediate its interaction with the pentapeptide inhibitor PhrC (ERGMT) or with its target protein ComA, a phosphorylation-dependent response regulator transcription factor for genetic competence. Our results show that RapC interaction with ComA inhibits the response regulator's ability to bind its target DNA promoter but does not affect its phosphorylation state. RapC binds equally well to ComA or to ComA∼P. The PhrC pentapeptide binds to RapC and inhibits its interaction with ComA. The D195 residue in TPR3 and the P263 residue in TPR5 of RapC are critical for the interaction with PhrC as their mutation to asparagine or leucine, respectively, prevents peptide inhibitory activity. The RapC mechanism of regulating ComA activity is a new example of how TPR motifs and their structural organization have been adapted for different specific functions within the B. subtilis Rap family.


Tetratricopeptide (TPR) repeats are structural domains found in a wide variety of proteins from eukaryotes to prokaryotes (Blatch and Lassle, 1999). Each repeat consists of 34 amino acids that fold into a structure consisting of a pair of antiparallel α-helices of equivalent length generally associated with a packing angle of ∼24° between the helix axes. Tetratricopeptide motifs are usually found in tandem arrays ranging between three and 16 motifs per protein. Multiple TPR motifs organize as a scaffold with a right-handed superhelical structure featuring a helical groove. This groove is thought to be suitable for recognizing interacting proteins. Tetratricopeptide domains are believed to be an ancient module promoting protein–protein interactions (Sikorski et al., 1990; Das et al., 1998; Groves and Barford, 1999; Scheufler et al., 2000).

In Bacillus subtilis, the 11 members of the Rap family of proteins each contain six tandem repeated motifs related to the TPR (Perego and Brannigan, 2001). The RapA, RapB and RapE members of this family act as negative regulators in the pathway to sporulation initiation by promoting the dephosphorylation of the Spo0F response regulator intermediate of the phosphorelay signal transduction system (Perego et al., 1994; Jiang et al., 2000). The dephosphorylation activity of RapA and RapE is inhibited by the PhrA (ARNQT) and PhrE (SRNVT) pentapeptides respectively (Perego, 1997; Jiang et al., 2000). The Phr pentapeptides are encoded by a rap-associated gene as 44 amino acid precursors that follow an export–import processing pathway resulting in the production of the active inhibitors: the PhrA active inhibitor comprises the five carboxy-terminal residues of the precursor whereas the PhrE pentapeptide is located internally, nine amino acids from the carboxy-terminal end of the phrE gene product. The intracellular activity of Phr pentapeptides is dependent upon reimportation by the oligopeptide permease (Opp) transport system (Perego et al., 1991; Rudner et al., 1991; Perego and Hoch, 1996; Lazazzera et al., 1997).

Phr pentapeptides are specifically active against their cognate Rap protein. Their specificity is intrinsic to the amino acid sequence of the pentapeptide and strictly dependent upon the presence of the carboxylic acid at the carboxy-terminal end of the peptide. A 1 : 1 stoichiometry was shown to characterize the interaction of RapA with its inhibitor PhrA pentapeptide or its Spo0F∼P target. RapA and RapB were found to be active in solution as dimer proteins (Core et al., 2001; Ishikawa et al., 2002).

The genetic characterization of another member of the Rap family, RapC, suggested a role for this protein in regulating competence development through modulation of the activity of the ComA response regulator and transcription factor (Solomon et al., 1996; Lazazzera et al., 1997) (M. Perego, unpubl. data). ComA, with its mated histidine kinase ComP, constitute the signal transduction system required for development of competence to DNA uptake in B. subtilis through regulation of the srfA operon (Nakano and Zuber, 1991; Roggiani and Dubnau, 1993). An additional role is played by the ComA–ComP proteins during the transition phase of growth from vegetative to stationary (or sporulation) phase, by controlling degradative enzyme production through DegQ (Msadek et al., 1991) and sporulation through RapA and RapE (Mueller et al., 1992; Jiang et al., 2000a).

The activity of RapC on ComA was found to be inhibited by the PhrC pentapeptide (ERGMT) (Solomon et al., 1996). This peptide is the result of the export–import processing pathway followed by the 40 amino acid product of the phrC gene and the release of the carboxy-terminal pentapeptide. By extrapolation from the biochemical characterization of RapA-dependent dephosphorylation of Spo0F∼P, it was proposed that RapC affected competence development by dephosphorylating ComA∼P (Solomon et al., 1996; Perego, 1997; Lazazzera et al., 1999a). In this report we show that RapC does not dephosphorylate ComA∼P. Instead, RapC binds to ComA, independently of its phosphorylation state, and it inhibits its DNA-binding activity. Thus, Rap proteins use the TPR motifs to establish protein–protein interaction and modulate the activity of their targets through at least two distinct mechanisms.


RapC and PhrC control transcription of ComA-dependent promoters

The completion of the B. subtilis genome sequence revealed that, in addition to the first rap genes identified (rapA and rapB), nine other rap genes, rapC to rapK, were present on the chromosome (Perego et al., 1996; Kunst et al., 1997; Perego, 1999). A deletion analysis carried out on all the rap genes (M. Perego, unpubl. data) revealed that deletion of rapC increased the transcription of rapA, a promoter known to be under control of the ComA response regulator for competence development (Mueller et al., 1992). Transcription analysis of the rapC promoter by means of an E. coli lacZ-fusion construct (plasmid pOM54, Fig. 1), indicated that a deletion of the rapC gene also increased the level of transcription of the rapC promoter in cells grown in sporulation growth conditions (Fig. 2A) as well as in defined minimal medium as reported by Lazazzera et al. (1999b). This is due to the presence of a ComA binding site in the rapC promoter region (Fig. 1) (Nakano et al., 1991; Lazazzera et al., 1999b). Because rapC is in an operon with the phrC gene, we constructed a chromosomal deletion of the latter and tested its effect on the transcription of a ComA-dependent promoter in sporulation growth conditions. As shown in Fig. 2A, in the absence of the phrC gene product (strain JH11299), transcription of the rapC promoter was significantly lower than in the wild-type strain (strain JH12923), i.e. the effect of the phrC deletion was opposite to the effect of the rapC deletion and both gene products affected ComA-dependent gene transcription. This is the expected result if the phrC peptide serves as an inhibitor of RapC modulation of ComA activity (Lazazzera et al., 1999b).

Figure 1.

Chromosomal region carrying the rapC and phrC genes. The restriction sites relevant to this study are shown. Parenthesis indicate the restriction sites introduced by PCR amplification. The length of the fragments cloned in the plasmids used is indicated by the lines. The position of the ComA-binding box and of the SigmaA- and SigmaH-dependent promoters transcribing rapC-phrC and phrC alone, respectively, are shown (_). The SigmaH promoter of phrC was originally identified by Carter et al. (1990). The position of the putative transcriptional terminator is indicated by the symbol Ω.

Figure 2.

Transcription analysis of a ComA-dependent promoter in rapC and phrC mutant strains. β-galactosidase assays were carried out on the rapC-lacZ carrying strains as described in Experimental procedures.
A. Strains: JH12923 (wild-type) -•-; JH12963 (rapC) -▪-; JH11299 (phrC) -▴-.
B. Strains: JH12923 (wild-type) -•-; JH19106 (pOM 102 wt) -▴-; JH11407 (pOM 102-P263L) -◆-.

As a further indication of the mechanism of peptide inhibition, an effect on rapC transcription similar to the one observed in the phrC mutant strain was obtained in a strain that expressed the RapC protein with a substitution of D195 to asparagine or P263 to leucine (see Fig 2B and Fig. 8 below). These mutations correspond to the D194N and P261L substitution in RapA that were shown to make this protein insensitive to the inhibitory activity of the PhrA pentapeptide (Perego et al., 1994; Perego, 1997). Thus, PhrC is an inhibitor of RapC activity and its mechanism of action is most likely identical to the mechanism of PhrA inhibition of RapA.

Figure 8.

TPR motifs in the Rap proteins.
A. Distribution of TPR motifs in the Rap protein (382 aa).
B. Amino acid sequence alignment of the six TPR repeats of the RapC protein. Residues D195 (TPR3) and P263 (TPR5) are represented in bold. Residues at positions 8, 20, 24 and 27 are generally conserved even among functionally different TPR domains and thus are considered important for the structural integrity of the TPR domain. The position of the two α-helices predicted for the Rap's TPR domains is shown.
C. Ribbon representation of the Rap TPR homology model. Only TPR1 to TPR5 were modelled as a result of the greater distances between TPR6 and TPR5 compared to the other interdomain regions. For each domain, α-helix A is shown in blue, whereas α-helix B is shown in red. Two orthogonal views are presented; the left is parallel while the right is perpendicular to the central helices. The side chains of residues D195 and P263 are shown to illustrate how they protrude into the groove generated by the packing of the antiparallel α-helices. The figure was drawn using MOLSCRIPT (Kraulis, 1991).

RapC does not dephosphorylate ComA∼P

Based on the known phosphatase activity of RapA, RapB and RapE on the Spo0F∼P response regulator, the genetic results described above were interpreted to mean that RapC was dephosphorylating the phosphorylated form of ComA. Thus, deletion of rapC would result in a higher level of ComA∼P accumulated in the cell and, as a consequence, a higher level of transcription of ComA-dependent genes. We overexpressed and purified ComP (MBP-ComP, see Experimental procedures), ComA and RapC to test this hypothesis.

The in vitro assay shown in Fig. 3 revealed that MBP-ComP, in the presence of ATP, promoted the phosphorylation of its response regulator ComA. However, the addition of RapC to this reaction did not significantly affect the phosphorylation level of ComA indicating that RapC did not promote its dephosphorylation. Additionally, this experiment suggested that the MBP-ComP-dependent phosphorylation of ComA, or phosphotransfer reaction, was not significantly affected by the presence of RapC. Thus RapC did not seem to affect the overall level of phosphorylation of the ComP-ComA two-component system.

Figure 3.

RapC does not dephosphorylate ComA∼P. MBP-ComP (1 µM) and ComA (2.5 µM) were incubated without RapC (lanes 1–6) or with RapC (5 µM) (lanes 7–12) and aliquots were taken at the time indicated and run on SDS-PAGE. The gel was visualized with a PhosphorImager (Amersham-Molecular Dynamics).

RapC does not inhibit ComP autophosphorylation or phosphotransfer to ComA

To further confirm that RapC did not affect the phosphorylation level of the ComP-ComA system, we examined the role of ComP autophosphorylation and the rate of phosphoryl transfer between MBP-ComP∼P and ComA in the presence and absence of RapC. As shown in Fig. 4A, RapC did not inhibit the MBP-ComP autophosphorylation reaction. We then tested the rate of phosphotransfer by first purifying MBP-ComP∼P from the autophosphorylation reaction buffer (as described in Experimental procedures) and then incubating it with ComA in the presence and absence of RapC. The results of this assay (Fig. 4B, lanes 5–8) showed that the same level of phosphoryl transfer occurred in both reaction conditions and at the time-points analysed (15 s and 60 min). Additionally, RapC did not affect the stability of MBP-ComP∼P as seen in Fig. 4B, lanes 1–4. Thin-layer chromatography analysis of the reactions shown in Fig. 4B to detect any cleaved phosphate further supported this conclusion (data not shown).

Figure 4.

RapC does not inhibit MBP-ComP autophosphorylation or phosphoryl transfer to ComA.
A. Rate of MBP-ComP (1 µM) autophosphorylation in the presence (–▴-) or absence (–•-) of RapC (5 µM). Time-point aliquots were analysed by SDS-PAGE and the radioactivity was quantified by PhosphorImager exposure and IMAGEQUANT software analysis.
B. MBP-ComP∼P (1 µM) (lanes 1–4) and MBP-ComP∼P (1 µM) with ComA (2.5 µM) (lanes 5–8) were incubated in the presence (+) or absence (–) of RapC (5 µM) for the time indicated in the figure. The reactions were stopped by the addition of SDS-loading dye and run on SDS-PAGE. ComA and RapC were premixed for 20 min before the addition of MBP-ComP∼P.

These experiments allowed us to conclude that RapC did not dephosphorylate ComA∼P and did not interfere with the autophosphorylation of ComP, its stability or its ability to phosphorylate ComA in vitro. The question remained how RapC affected ComA activity in vivo.

RapC interacts specifically with ComA

We previously showed that the RapA, RapB and RapE proteins target the phosphorylated form of the Spo0F response regulator and promote its dephosphorylation (Perego et al., 1994; Jiang et al., 2000a). With the general assumption that all B. subtilis Rap proteins were dephosphorylating a target response regulator, the results shown above raised the possibility that ComA may have not been the direct target of RapC. Because RapA and RapB were previously shown to form a specific complex with Spo0F∼P in a native gel binding assay (Ishikawa et al., 2002), we employed the same assay to screen response regulator proteins for their ability to interact with RapC. We first tested proteins readily available in the laboratory, i.e. Spo0A, Spo0F, Spo0F∼P and ComA. The results shown in Fig. 5A clearly indicated that RapC specifically interacted with ComA and formed a stable complex in the native gel assay. No interaction was observed with Spo0F, Spo0F∼P or Spo0A. This finding strongly restored the belief that RapC regulation of competence development was exerted through the ComA response regulator. The native gel assay also showed that RapC was a dimer protein as was the case for RapA and RapB (Ishikawa et al., 2002) (data not shown).

Figure 5.

RapC interaction with ComA and PhrC.
A. RapC interacts specifically with ComA. RapC (10 µM) (lanes 1–5) was incubated with Spo0F (lane 2), Spo0F∼P (lane 3), ComA (lane 4) or Spo0A (lane 5) (all at 10 µM). The reactions were run on 10% Tris-Tricine-EDTA native gel and stained with Coomassie blue. A complex is visible only in lane 4 containing ComA and RapC.
B. Interaction of RapC with ComA and PhrC. RapC, ComA and PhrC (all at 10 µM) were analysed on a 10% Tris-Tricine-EDTA native gel stained with Coomassie blue. Lane 1: RapC; lane 2: RapC and PhrC; lane 3: RapC and ComA; lane 4: RapC, ComA and PhrC (RapC and ComA were mixed before the addition of PhrC); lane 5: ComA.

The PhrC pentapeptide prevents the formation of the ComA:RapC complex

We previously showed that the complex formed by RapA with Spo0F∼P was prevented by the PhrA pentapeptide (Ishikawa et al., 2002). Because PhrC was known to counteract the activity of RapC in vivo, we tested whether the mechanism was through inhibition of ComA:RapC complex formation. The native gel binding assay was first employed to determine whether PhrC could bind to RapC. As shown in Fig. 5B, lane 2, PhrC formed a complex with RapC and this complex had a slower mobility than the RapC dimer alone. When PhrC was added to the reaction containing ComA and RapC (Fig. 5B, lane 4), more than 50% of response regulator protein remained free while more than 90% was complexed with RapC in the reaction lacking PhrC. Thus PhrC counteracted the activity of RapC itself by binding to RapC and preventing its interaction with ComA.

RapC binds equally to ComA or ComA∼P

Our previous studies showed that the RapA protein formed a stable complex with the phosphorylated form of the Spo0F response regulator (Ishikawa et al., 2002). Interaction with the unphosphorylated form of Spo0F, although detectable through a cross-linking reaction, was not sufficiently stable to be detected by the native gel binding assay. Thus the question arose whether RapC had any specificity for one form or the other of ComA: the assays shown in Fig. 5 were carried out with unphosphorylated ComA and they indicated that approximately 90% of protein was bound to RapC in a stable complex. Because it is established that phosphorylated ComA directly activates the transcription of competence genes such as srfA (Roggiani and Dubnau, 1993), specificity for the unphosphorylated form of ComA rather than the phosphorylated form would be inconsistent with the phenotypes observed in vivo in the rapC and phrC mutant strains.

If RapC does not affect the phosphorylation state of the ComP-ComA proteins, how does the deletion of rapC result in increased transcription of ComA∼P-dependent genes? Similarly, how does a deletion of phrC, whose product inhibits RapC, result in a decrease of ComA∼P-dependent gene transcription? To address this problem of target specificity, we assayed the binding of RapC to ComA and ComA∼P. The ComA response regulator and the MBP-ComP protein kinase were incubated in the presence and absence of ATP and the autophosphorylation and phosphoryl transfer reactions were allowed to occur. At equilibrium, the RapC protein was added and allowed to interact before the reactions were loaded on the native gel shown in Fig. 6. After Coomassie staining, a quantification of protein concentration indicated that there was no significant difference in the amount of ComA bound to RapC between the reaction containing or the reaction lacking ATP. In order to determine the percentage of phosphorylated ComA present in the reaction, an identical MBP-ComP-dependent phosphorylation reaction of ComA was carried out in the presence of γ32P-ATP. The subsequent analysis by SDS-PAGE and quantification of radioactivity indicated that approximately 30% of ComA was phosphorylated in the assay conditions used (data not shown). The phosphorylation state of ComA was also shown not to affect the inhibitory role played by PhrC in ComA:RapC complex formation (Fig. 6, lane 6 and 7). Thus RapC did not seem to differentiate between ComA and ComA∼P.

Figure 6.

Interaction of RapC with ComA and ComA∼P. MBP-ComP (0.12 µM) and ComA (12 µM) were incubated with ATP (lanes 2, 4 and 7) or without ATP (lanes 1, 3 and 6) for 60 min before the addition of RapC (12 µM) (lanes 3 and 4) or the addition of RapC and PhrC (12 µM) (lanes 6 and 7). Lane 5 contains RapC (12 µM) alone. The reactions were analysed on a 10% Tris-Tricine-EDTA native gel stained with Coomassie blue.

RapC inhibits ComA binding to its target DNA

In order to rationalize the genetic and biochemical data available thus far on the mechanism of RapC regulation of ComA activity, we hypothesized an effect of RapC:ComA interaction on the DNA-binding property of the response regulator transcription factor. Thus, a DNA retardation gel electrophoresis assay was carried out using the rapC promoter as a substrate. This promoter contains a consensus-binding sequence to which ComA is known to bind from work on the srfA promoter (Roggiani and Dubnau, 1993). Additionally, transcription from a rapC-lacZ fusion construct was abolished in a comA deletion mutant (M. Perego, unpubl. data) (Lazazzera et al., 1999b). As shown in Fig. 7A (lanes 1 and 2), ComA bound to the [α32P]-dATP-labelled DNA fragment containing the rapC promoter and promoted its retardation whereas the RapC protein did not affect the migration of the fragment (Fig. 7A, lane 3) However, the DNA-binding capability of ComA was affected by the presence of RapC with the maximal inhibition observed at a 1:1 stoichiometry (Fig. 7A, lanes 4–7). The DNA-binding properties of ComA was restored in the presence of RapC when the PhrC pentapeptide was added to the reactions (Fig. 7B, lanes 5–10). A 1:1 stoichiometry of RapC and PhrC resulted in the maximal restoration of ComA binding to the rapC promoter fragment.

Figure 7.

Gel retardation assay of ComA binding to the rapC promoter. A 313-bp fragment containing the rapC promoter was labelled with [α32P]-ATP and 2 µM were used in each lane.
A. Titration of ComA by RapC. ComA was used at 20 µM final concentration. Lane 1: no protein; lane 2: ComA; lane 3: RapC (25 µM); lane 4: ComA and RapC (5 µM); lane 5: ComA and RapC (10 µM); lane 6: ComA and RapC (20 µM); lane 7: ComA and RapC (25 µM).
B. Titration of RapC by PhrC. ComA was used at 15 µM and RapC at 20 µM final concentration. Lane 1: no protein; lane 2: ComA; lane 3: RapC; lane 4: ComA and RapC; lanes 5, 6 and 7: RapC and PhrC (10, 20, 50 µM respectively) were preincubated for 20 min, ComA was then added and incubation continued for 20 min before the addition of the rapC promoter fragment; lane 8, 9 and 10: RapC and ComA were preincubated for 20 min, PhrC (10, 20, 50 µM respectively) was then added and the incubation continued for 20 min before the addition of the rapC promoter fragment.

These results allowed us to conclude that, contrary to the prediction based on the mechanism of action of other members of the Rap family of proteins, RapC does not promote the dephosphorylation of ComA∼P but it prevents ComA, independently of its phosphorylation state, from interacting with its target DNA. Thus this mechanism explains the phenotypes observed in vivo, i.e. increased transcription of ComA-dependent genes in the rapC mutant and decreased transcription in the phrC strain.


This work shows how the TPR protein structural organization has been adopted by the Rap proteins of B. subtilis to promote protein–protein interactions that result in at least two distinct regulatory mechanisms controlling the output of two-component signal transduction systems. On one hand the RapA, RapB and RapE members of the Rap family regulate sporulation initiation through protein–protein interaction with the Spo0F∼P intermediate response regulator of the phosphorelay pathway. This interaction is specific for the phosphorylated form of its response regulator and it promotes the dephosphorylation reaction thus inhibiting the initiation of the sporulation process (Perego et al., 1994; Jiang et al., 2000a; Ishikawa et al., 2002).

On the other hand, RapC, another member of the Rap family of proteins and as such highly similar to RapA in structure (six TPR motifs) and in amino acid sequence (44% identical residues) (Perego et al., 1996; Perego and Brannigan, 2001), uses a different mechanism to regulate the activity of a response regulator through protein–protein interaction: RapC-specific interaction with ComA prevents this response regulator from binding to its target DNA promoters and thus it prevents the initiation of the competence development pathway. RapC can both inhibit the binding of ComA to DNA as well as dissociate a preformed complex ComA:DNA (Fig. 7 and data not shown). Furthermore, RapC can interact with the unphosphorylated form of ComA as well as the phosphorylated form, ComA∼P (Fig. 6).

Spo0F and ComA are both members of the response regulator family of two-component systems and as such they share the common structural organization of the regulator domain of response regulators with five β-strands surrounded by five α-helices (Volz, 1995; Madhusudan et al., 1996; West and Stock, 2001). However, Spo0F is a single domain response regulator whose main function is to transfer phosphoryl groups within the phosphorelay pathway from the sporulation kinases (KinA-E) to the Spo0A response regulator and transcription factor through the Spo0B phosphotransferase (Burbulys et al., 1991; LeDeaux and Grossman, 1995; Kobayashi et al., 1995; Jiang et al., 2000b). The Spo0F protein interacts with RapA, RapB or RapE mainly through residues localized on the top of the molecule near the active site. Residues in the β1-α1, β3-α3 and β4-α4 loops connecting the β-strands to the α-helices of Spo0F were shown by alanine-scanning mutagenesis to mediate the interaction with RapB (Tzeng and Hoch, 1997; Tzeng et al., 1998). This interaction results in the dephosphorylation of Spo0F∼P. Regulating the phosphorylation level of Spo0F seems a very effective and energy-saving way of regulating the entire phosphorelay pathway for sporulation initiation.

ComA is a two-domain response regulator and, as such, two distinct functions can be reasonably envisioned as a target for a regulatory control: the phosphorylation state of the amino-terminal receiver domain or the DNA-binding and transcription regulation function of the carboxy-terminal domain (Roggiani and Dubnau, 1993). Here we have shown that the latter characterizes the mechanism of RapC regulation of competence gene expression in B. subtilis. The question, however, remains of whether RapC inhibits ComA ability to bind its target DNA by directly interacting with the amino terminal or with the carboxy-terminal domain of the response regulator or both. It is believed that the carboxy-terminal DNA-binding domain of response regulators may be kept in an ‘off’ position by interaction with its unphosphorylated amino-terminal domain (Stock et al., 1995). Accordingly, ComA∼P was shown to have a higher binding affinity for the srfA promoter than its unphosphorylated counterpart (Roggiani and Dubnau, 1993). Thus RapC may interact with the amino-terminal portion of ComA, independently of its phosphorylation state, and keep the C-end in the ‘off’ position. Alternatively, RapC could interact directly with the carboxy-terminal domain in such a way to prevent DNA-binding. One helix of the helix–turn–helix structure of this domain could provide the interacting surface for RapC binding. We favour this second explanation for the reason that binding of RapC to ComA does not seem to affect the phosphoryl transfer from ComP to the response regulator (Fig. 3). This would be expected if the ComA:RapC interaction involved the corresponding protein surfaces that form the Spo0F:RapA interface (Tzeng et al., 1998). Support to this hypothesis also comes from a yeast two-hybrid study that shows RapC specific interaction with the carboxy-terminal domain of ComA and lack of interaction with the N-terminal domain (S. Ishikawa, S. Stephenson, N. Ogasawara and M. Perego, manuscript in preparation).

Still undefined is the surface of interaction of Spo0F∼P and ComA on the RapA and RapC proteins respectively. Because mutations in RapA or RapC that would abolish their interaction with the response regulators are not known at this time, we cannot directly correlate the structural organization in TPR domains with their functional role. However, many studies on eukaryotic proteins containing TPR motifs have proven the role of these domains as protein–protein interaction modules (Gatto et al., 2000; Scheufler et al., 2000; Brinker et al., 2002; Ward et al., 2002). X-ray crystallographic structures have shown that a TPR motif contains two antiparallel α-helices (A and B) such that tandem arrays of TPR motifs generate a right-handed helical structure with an amphipatic channel that can accommodate the target protein (Das et al., 1998). The co-crystals of TPR-ligand complexes of the Hsp70-Hsp90 multichaperone machinery has shown the ligand peptides binding in an extended conformation and spanning the channel formed by the TPR domains (Scheufler et al., 2000). Ligand binding is mediated by both electrostatic and hydrophobic interactions, the latter being most likely the critical ones for establishing binding specificity. The surface of the channel is made mainly by the side chains of the amino acids in the helix A of each TPR motif, whereas the opposite side of the channel is formed by amino acids from both helix A and B. Furthermore, five to six TPR motifs could accommodate an α-helix of a target protein (Blatch and Lassle, 1999).

A model of the first five TPR motifs of Rap proteins (TPR1-5, Fig. 8A) was generated based on the tri-dimensional structure of the TPR elements of the PP5 protein phosphatase (Fig. 8C) (Das et al., 1998). Positioning of residues D195 and P263 of helix A of TPR3 and TPR5, respectively, clearly shows the extensions of their side chains toward the groove of the hypothetical structure thus supporting our hypothesis that these residues are critical for Phr peptide interaction. It is unlikely however, that these Asp and Pro residues are relevant for establishing specificity as they are conserved in all the Rap proteins known to be regulated by a Phr peptide (RapA, B, C, E and F) (Perego and Brannigan, 2001). Because the D to N and P to L mutations result in Rap proteins insensitive to the inhibitory activity of the corresponding peptide, these two residues are probably critical for establishing the interaction and allowing complex formation. At least in the case of RapA and PhrA, the D194N and P261L mutant proteins are not affected in their ability to promote Spo0F∼P dephosphorylation (S. Ishikawa and M. Perego, unpubl. data) suggesting that these residues may not be important for the interaction with the target response regulator. However, kinetic analysis of the RapA-induced dephosphorylation of Spo0F∼P and the inhibition by PhrA indicate that the substrate and the inhibitor compete for binding to a common regulatory site on RapA (L. Core and M. Perego, unpubl. data). Thus the TPR motifs are the key structural feature that provides Rap proteins with their regulatory functions by promoting the binding to the Phr peptides or to the substrate response regulators.

One major unsolved issue regarding the RapA-dependent dephosphorylation of Spo0F∼P was whether the Rap proteins were enzymatically involved in the reaction or whether they only acted allosterically, thus promoting autodephosphorylation which is an intrinsic function of many response regulators (Stock et al., 1995). The finding that RapC does not affect the phosphorylation state of ComA but rather acts as an allosteric effector of its DNA-binding function, contributes to push forward the theory that RapA, RapB and RapE are also allosteric effectors whose specific interaction with Spo0F∼P stimulates its autodephosphorylation activity. This theory is also supported by the fact that we never identified a phosphorylated intermediate in the Rap-dependent dephosphorylation of Spo0F∼P as in the case of Ser/Thr or Tyr phosphatases of eukaryotic signal transduction systems (Zhang et al., 1994; Barford, 1996).

Thus Rap proteins are regulatory molecules that employ the TPR structural organization to interact with their targets and affect their activity. Whether the targets of the remaining members of the B. subtilis Rap proteins will also be response regulators remains to be determined. As Raps are in fact protein-binding proteins, we cannot exclude the possibility that some of them may have evolved an interaction surface for specifically binding proteins other than response regulators. This must be the case of a family of proteins comprising the NprA of B. stearothermophilus, PreL of Lactobacillus, NprR and PlcR of B. thuringiensis that shares a significant level of conservation with the B. subtilis Rap proteins (Uehara et al., 1979; Lereclus et al., 1996; Perego and Hoch, 2002). However, the homology is mainly the result of the presence of TPR domains in the NprA-like proteins. These proteins in fact distinguish themselves from the Rap proteins by the presence of an amino-terminal DNA-binding domain that contributes to their function as transcriptional regulators. The genes encoding the NprA, PreL, NprR and PlcR are also followed by a small open reading frame encoding a peptide structurally similar to the Phr peptide associated with the Rap proteins (M. Perego, unpubl. obs.). Recently, the carboxy-terminal pentapeptide of the papR gene product associated with the B. thuringiensis PlcR was shown to be required to activate PlcR DNA-binding activity (Slamti and Lereclus, 2002).

Thus Rap proteins and NprA-like proteins, although functionally different, must have evolved from a unique regulatory protein–protein interaction mechanism that exploited the TPR structural motif and its evolutionary ability to specifically interact with many different target proteins (Blatch and Lassle, 1999).

Experimental procedures

Bacterial strains and growth conditions

The B. subtilis strains used in this study were derivatives of the parental strain JH642 (trpC2, phe-1) carrying the rapC-lacZ fusion construct of plasmid pOM54 (KmR) integrated in the amyE region to give strain JH12923. Competent cells (Anagnostopoulos and Spizizen, 1961) of JH12923 were transformed with uncut plasmid pOM47 to give strain JH12963 (rapC::cat) or with linearized plasmid pOM88 to give strain JH11299 (phrC::cat) (Fig. 1). Cultures for β-galactosidase assay were grown in Schaeffer's sporulation medium (Schaeffer et al., 1965). Samples were taken at hourly intervals and processed as previously described (Ferrari et al., 1986). β-galactosidase activity was expressed in Miller units (Miller, 1972).

E. coli DH5α was used for plasmid constructions and propagation. Antibiotics were used at the following concentrations: chloramphenicol 5 µg ml−1, kanamycin 2 µg ml−1 in B. subtilis, 20 µg ml−1 in E. coli.

Plasmid construction

Plasmid pOM48 is a derivative of the integrative vector pJM103 (Perego, 1993) carrying a PCR amplified PstI-EcoRI fragment of 775 bp. This plasmid confers chloramphenicol resistance upon integration into the chromosome via single cross-over recombination. The same fragment was also engineered in the lacZ-transcriptional fusion vector pJM115 (Perego, 1993) thus generating plasmid pOM54. Plasmid pOM54 confers kanamycin resistance in B. subtilis upon integration via double cross-over recombination in the amyE locus.

Plasmid pOM47 is also a derivative of pJM103 carrying the 255 bp EcoRI fragment internal to the rapC gene. Its integration into the chromosome by single cross-over integration and selection for chloramphenicol resistance resulted in inactivation of the rapC gene.

Plasmid pOM88 is a derivative of the cat cassette vector pJM105 (Perego, 1993). The fragments at the left and right of the cassette were obtained by PCR amplification and cloned as SacI-BamHI and BstUI-EcoRI (Klenow treated) into the multiple cloning site at the SacI-BamHI site on the left and HincII site on the right of the cat gene. Integration of linearized pOM88 into the chromosome by double cross-over recombination and selection for chloramphenicol resistance resulted in the deletion of the phrC gene.

Overexpression of ComA was obtained from a derivative of pET16b (Novagen) carrying the PCR amplified comA gene engineered to contain a BspHI site at the 5′ end overlapping the first (ATG) codon and a BamHI site at the 3′ end downstream of the stop codon. This fragment was cloned in pET16b digested with NcoI and BamHI. The expression plasmid for ComP was a derivative of pMAL-c2 (New England BioLabs) carrying a PCR amplified EcoRI-BamHI fragment extending from codon 390 (Phe) to the stop codon. This created a fusion between the maltose binding protein and the C-terminal 379 amino acid of the ComP histidine kinase. The RapC protein was expressed from a derivative of pET16b (Novagen) containing the rapC coding region PCR amplified as an NdeI-BamHI fragment. The cloning resulted in the addition of 10 His codons to the 5′ end of the rapC gene.

All constructs generated from PCR amplified fragments were subject to sequence analysis to verify the fidelity of amplification.

Site-directed mutagenesis

Plasmid pOM 102 is a derivative of the integrative vector pJM 103 (Perego, 1993) carrying a 745 bp fragment of wild-type rapC. This fragment was transferred to pBluescript (Stratagene) in order to generate single stranded DNA and carry out site-directed mutagenesis using the Bio-Rad Muta-gene Phagemid system based on the method described by Kunkel (1985). After sequence analysis, a clone containing the mutations resulting in the D195N or the P263L substitutions were transferred back into pJM103, giving rise to plasmids pOM 102-D195N and pOM 102-P263L. Transformation of these plasmids into B. subtilis results in chromosomal integration by single crossover and generation of a full length, expressed, rapC gene and a truncated, not expressed rapC gene followed by the phrC gene under its own promoter. The presence of the mutations in the full length expressed rapC gene was checked by sequencing of a PCR amplified fragment. Transformation of pOM 102 wild type into strain JH12923, which carries the rapC-lacZ fusion construct in amyE, resulted in strain JH19106. The JH12923 derivative transformed with pOM 102-P263L and expressing the mutant protein was named JH11407.

Protein expression and purification

A colony of E. coli BL21 (DE3) transformed with plasmid pET16b-RapC was used to inoculate 10 ml of LB containing ampicillin (100 µg ml−1) and grown overnight at 30°C. Three ml of overnight culture were used to inoculate 1 L of LB medium (Amp 100 µg ml−1). Cells were grown at 30°C to approximately OD600 = 2.0 without IPTG induction. Cells were harvested at 4°C and the pellet resuspended in lysis buffer (20 mM Tris-HCl pH 7.9, 500 mM NaCl, 20 mM β-mercaptoethanol, 0.1 mM PMSF-phenylmethysulfonilfluoride, 10 mM imidazole). The cells were lysed with three passages through a French press. After removal of cell debris by ultracentrifugation at 45 000 r.p.m. for 1 h at 4°C, the supernatant was loaded onto a nickel nitrilotriacetic acid (NTA) agarose column (Qiagen). The his-tagged protein was eluted with a 10–200 mM imidazole linear gradient in lysis buffer. Fractions were collected and analysed by SDS-PAGE and Coomassie blue staining. The protein was dialysed against the storage buffer (50 mM Hepes pH 7.0, 10 mM DTT), concentrated with an Amicon Centriprep-10 and glycerol was added to 20% final concentration. RapC was stored in aliquots at − 80°C.

Overexpression of ComA was carried out in the E. coli BL21 (DE3) pLysS strain freshly transformed with plasmid pET16b-ComA. Cells were grown in LB medium to OD600 = 0.7, induced with 3 mM IPTG and allowed to express for 3 h at 37°C. The cells were harvested by centrifugation and resuspended in ComA buffer (50 mM Tris-HCl pH 7.8, 2 mM EDTA and 1 mM PMSF). After sonication, the lysate was centrifuged for 30 min at 15 000 r.p.m. Streptomycin sulphate was added to the clear lysate to 1% final concentration. After a second centrifugation for 30 min at 15 000 r.p.m., the supernatant was precipitated with 25% and 50% ammonium sulphate. Precipitants were collected by centrifugation for 30 min at 15 000 r.p.m., resuspended in ComA buffer and dialysed against the same buffer. Proteins were loaded on a Q-Sepharose column and eluted with a gradient of NaCl (0–1 M). The fractions containing ComA were pooled, dialysed in ComA buffer and loaded on a Sephacryl S-100 26–60 gel filtration column (Pharmacia Biotech). The fractions containing ComA were collected, glycerol was added to 20% final concentration and the protein was stored at − 20°C.

The MBP-ComP fusion protein was overexpressed in strain DH5α carrying the pMALc2-ComP plasmid. A culture grown in LB with ampicillin and glucose 0.2% was induced at OD600 = 0.5 with 0.3 mM IPTG and allowed to grow for 2 h. Cells were harvested by centrifugation and the pellet resuspended in lysis buffer (20 mM Tris-HCl pH 7.4, 200 mM NaCl, 1 mM EDTA, 10 mM β-mercaptoethanol, 1 mM PMSF). Cells were lysed by sonication and the debris spun down by a 1 h centrifugation at 45 000 r.p.m. at 4°C. The supernatant was diluted five times in lysis buffer and then loaded on to amylose resin (New England BioLabs). Purification was as described by the manufacturer using 10 mM maltose for protein elution. Fractions were collected and analysed by SDS-PAGE. The protein was dialysed against the lysis buffer lacking PMSF, concentrated twofold and the buffer made 20% glycerol before storing it at − 80°C. The synthetic PhrC peptide (NH2-ERGMT-COOH) was resuspended in H2O and its concentration determined by AAA amino acid analysis.

Native polyacrylamide gel electrophoresis

Native-PAGE was carried out with Tris-Tricine-EDTA buffer as described (Ishikawa et al., 2002). Acrylamide concentrations are specified in the Figure legends.

Purification of ComP∼P

A 500 µl phosphorylation reaction was carried out in 20 mM Hepes pH 7.0, 15 mM MgCl2, 50 mM KCl, 10 mM DTT, 10% glycerol and 12.5 µl of [γ-32P]-ATP (NEN, 6000 Ci mmol−1) with MBP-ComP protein at 50 µM final concentration. After 1 h incubation at room temperature, 0.2 mM ATP was added and the reaction was incubated for one additional hour. EDTA final concentration 30 mM was added to stop the reaction which was transferred to a Slide-A-Lyzer cassette MWCO 10 000 (Pierce) and extensively dialysed against 20 mM Hepes pH 7.5, 10 mM DTT, 0.1 mM EDTA, 50 mM KCl. The dialysis was carried out until the buffer read zero radioactivity counts against the background. MBP-ComPapproxP was then concentrated in a Microcon concentrator YM50 (Amicon) and quick-frozen to − 80°C.

Phosphorylation of ComA by ComP∼P

The phosphorylation reaction of ComA by purified MBP-ComP∼P was carried out in 20 mM Hepes pH 7.0, 50 mM KCl, 15 mM MgCl2, 0.1 mM EDTA, 10 mM DTT and 10% glycerol. MBP-ComP∼P (1 µM) was incubated alone or with ComA (2.5 µM) in the absence or presence of RapC (5 µM). The reaction was carried out at room temperature and samples were taken at 15 s and 60 min. Reactions were stopped by the addition of gel loading buffer and then analysed by SDS-PAGE.

Gel retardation assay

A 313 bp NsiI-EcoRI fragment from pOM48 was labelled with [α-32P]-dATP using the fill-in activity of the Klenow polymerase enzyme (New England BioLabs) according to the manufacturer recommendation. The labelled fragment was purified from a 5% Acrylamide (29:1) gel by electroelution, phenol and chloroform extracted, ethanol-precipitated and resuspended in Tris-EDTA buffer.

Binding of ComA to the DNA fragment was performed at room temperature in 25 mM Hepes pH 7.0, 50 mM KCl, 10 mM MgCl2, 10 mM DTT, 0.1 mM EDTA, 12% glycerol and double-stranded poly [dA-dT] and poly [dG-dC] at 200 mg ml−1 final concentration. The proteins were preincubated 20 min before the addition of the fragment DNA. After an additional 20 min, loading dye was added and the samples were immediately applied to a 5% polyacrylamide gel running at 300 V in Tris-Acetate-EDTA buffer. Electrophoresis was continued at 120 V for approximately 30 min. The gels were dried and exposed to a PhosphorImager screen (Amersham-Molecular Dynamics).


This research was supported, in part, by Public Health Service grant GM55594 from the National Institute of General Medical Sciences, National Institutes of Health. This is publication 15550-MEM from The Scripps Research Institute. The Stein Beneficial Trust supported, in part, oligonucleotide synthesis and DNA sequencing. We thank Samantha Edwards for carrying out the β-galactosidase assays and Debashis Mukhopadhyay for help in generating Fig. 8C.