Core of the partner switching signalling mechanism is conserved in the obligate intracellular pathogen Chlamydia trachomatis

Authors


E-mail cwprice@ucdavis.edu; Tel. (+1) 530 752 1596; Fax (+1) 530 752 4759.

Summary

Chlamydia trachomatis is an obligate intracellular bacterial pathogen that can cause sexually transmitted and ocular diseases in humans. Its biphasic developmental cycle and ability to evade host-cell defences suggest that the organism responds to external signals, but its genome encodes few recognized signalling pathways. One such pathway is predicted to function by a partner switching mechanism, in which key protein interactions are controlled by serine phosphorylation. From genome analysis this mechanism is both ancient and widespread among eubacteria, but it has been experimentally characterized in only a few. C. trachomatis has no system of genetic exchange, so here an in vitro approach was used to establish the activities and interactions of the inferred partner switching components: the RsbW switch protein/kinase and its RsbV antagonists. The C. trachomatis genome encodes two RsbV paralogs, RsbV1 and RsbV2. We found that each RsbV protein was specifically phosphorylated by RsbW, and tandem mass spectrometry located the phosphoryl group on a conserved serine residue. Mutant RsbV1 and RsbV2 proteins in which this conserved serine was changed to alanine could activate the yeast two-hybrid system when paired with RsbW, whereas mutant proteins bearing a charged aspartate failed to activate. From this we infer that the phosphorylation state of RsbV1 and RsbV2 controls their interaction with RsbW in vivo. This experimental demonstration that the core of the partner switching mechanism is conserved in C. trachomatis indicates that its basic features are maintained over a large evolutionary span. Although the molecular target of the C. tra-chomatis switch remains to be identified, based on the predicted properties of its input phosphatases we propose that the pathway controls an important aspect of the developmental cycle within the host, in response to signals external to the C. trachomatis cytoplasmic membrane.

Introduction

The obligate intracellular pathogen Chlamydia trachomatis is an agent of significant human illness, including blindness, reactive arthritis and sexually transmitted diseases with potentially serious sequelae (Schracter, 1999). The bacterium has a biphasic developmental cycle, beginning as an inert but infectious elementary body which develops into a metabolically active reticulate body only after entering the host cell, where it replicates in a parisitophorous vacuole called the chlamydial inclusion (Moulder, 1991). During the developmental cycle the organism communicates with its host and modifies the inclusion environment, in part by means of effector proteins secreted via a type III mechanism (Fields et al., 2003; Clifton et al., 2004; Subtil et al., 2005). One consequence of this modification is to circumvent host-cell defences by preventing interaction of the endocytic or lysosomal pathway with the inclusion while establishing contact with the exocytic pathway (reviewed in Fields and Hackstadt, 2002; Rockey et al., 2002; Salcedo and Holden, 2005). Towards the end of the replication period, the reticulate body differentiates into the inert elementary body, which is then released by host-cell lysis (Moulder, 1991).

This specialized biphasic cycle and the ability to survive intracellular defences imply differential control of protein activity and gene expression, presumably in response to external and internal signals, and transcriptional profiling experiments have provided evidence for a changing pattern of expression as the infection unfolds (Shaw et al., 2000; Belland et al., 2003; Nicholson et al., 2003). The genome of C. trachomatis is greatly reduced compared with most free-living bacteria, with only 893 chromosomal genes (Stephens et al., 1998), and it encodes few obvious signalling elements that might sense and convey the requisite cues. These elements include HcrA, which controls a heat shock response via the CIRCE sequence (Wilson and Tan, 2004); the Fur-like regulator DcrA, which potentially controls about 30 genes, including some implicated in type III secretion (Rau et al., 2005); and a solitary two-component system, CtcB-CtcC, which is thought to regulate the NtrA-like σ54 transcription factor (Koo and Stephens, 2003). Notably, these signalling elements also include the essential components of a partner switching signalling pathway, which is the focus here.

The partner switching mechanism has been studied most intensively in the Gram-positive bacterium Bacillus subtilis, where it functions in the signalling pathways that activate the stress-responsive transcription factors σF and σB (reviewed in Price, 2002; Hilbert and Piggot, 2004). These σ factors control the sporulation process and the general stress response respectively. Partner switching pathways contain a minimum of four proteins, and Fig. 1A shows the σB example: an input phosphatase (RsbU or RsbP), an antagonist protein (RsbV), a switch protein/kinase (RsbW) and a target protein (σB). According to this model, the RsbW switch protein has two important functions in unstressed cells. First, it acts as an anti-σ factor which binds and sequesters its σB target in an inactive complex (Benson and Haldenwang, 1993; Alper et al., 1996). Second, it also acts as a serine kinase which phosphorylates the RsbV antagonist protein, rendering it unable to bind RsbW (Dufour and Haldenwang, 1994; Alper et al., 1996; Yang et al., 1996). Following an activating stress, RsbV-P is dephosphorylated by one of the input phosphatases (Voelker et al., 1996a; Yang et al., 1996; Vijay et al., 2000), thereby allowing RsbV to bind RsbW and induce the release of σB. RsbW in effect switches its binding partner, depending on the phosphorylation state of the RsbV antagonist protein. In this example, association of the switch protein negatively controls its target protein, a transcription factor. However, there is a second partner switch further upstream in the σB signalling pathway (not shown), in which association of the RsbT switch protein positively controls its target protein, which happens to be the RsbU environmental phosphatase (Yang et al., 1996; Kang et al., 1998; Delumeau et al., 2004). In this case, the RsbT switch protein serves as a regulatory subunit for the phosphatase. Thus, the output of the switch can be either positive or negative, and the regulatory effect is mediated by direct interaction with the target protein, which can be either an enzyme or a transcription factor (Price, 2002).

Figure 1.

Model of a partner switching signalling pathway in C. trachomatis.
A. A model of the B. subtilis pathway controlling the general stress transcription factor σB. Two different classes of stress signals are conveyed to the switch via independent pathways, each terminating with a differentially regulated PP2C phosphatase that targets RsbV-P (with shaded P indicating phosphoryl group; see text for full description of model).
B. A model of the C. trachomatis pathway under consideration here, with two membrane-bound PP2C phosphatases each specific for one of the two RsbV antagonist proteins (see text). According to this model, the phosphorylation state of RsbV1 and RsbV2 are determined by competing PP2C and RsbW serine kinase activities. RsbV1-P and RsbV2-P (with shaded phosphoryl) are unable to bind RsbW, which is then free to modulate activity of its target protein via direct protein–protein interaction (+ and – arrows).
C. The domain organization of the C. trachomatis partner switching regulators, as predicted by Conserved Domain Search (Marchler-Bauer and Bryant, 2004). RsbU and CT589 each have two potential transmembrane domains that would place their PP2C serine phosphatase domains within the cytoplasm (Claros and von Heijne, 1994; toppred program as implemented by Deveaud and Schuerer at the Institut Pasteur, run with an upper cut-off of 1.5). RsbU also possesses a HAMP domain between the second transmembrane (TM) region and the PP2C domain (Aravind and Ponting, 1999), and CT589 is most similar to RsbU within this region (data not shown). RsbV1 and RsbV2 each have a STAS domain containing the conserved serine on which other antagonist proteins are phosphorylated, and RsbW contains a serine-threonine kinase domain with conserved N, G1 and G2 boxes.

Genome analysis suggests that the partner switching mechanism is widespread among eubacteria and can be adapted to different signalling tasks (Koonin et al., 2000; Mittenhuber, 2002; Price, 2002). However, the only experimental characterization of the mechanism outside the Gram-positive lineage has come in the beta-proteobacterium Bordetella bronchiseptica. In this pathogen, partner switching orthologues function essentially like their B. subtilis counterparts and are required for delivery of protein effectors into host cells via a type III secretion system (Mattoo et al., 2004; Kozak et al., 2005). The available evidence indicates that this signalling pathway is not required for transcription of type III secretion genes but instead controls the secretion process at the post-translational level.

Here we ask whether the partner switching mechanism is conserved in the Gram-negative C. trachomatis, which is phylogenetically distinct from free-living Bordetella species and which occupies a very different environmental niche. Because C. trachomatis cannot yet be grown in cell-free media and lacks a system for genetic analysis, an in vitro approach was used to demonstrate that the core of the mechanism operates much as it does in the B. subtilis paradigm. We speculate that this partner switching pathway senses and transmits signals from the host-cell environment to control an important aspect of chlamydial physiology.

Results

Model of a partner switching signalling pathway in C. trachomatis

The C. trachomatis genome encodes five potential components of a partner switching signalling network, four of which are currently annotated as Rsb to reflect similarity with their B. subtilis counterparts: two PP2C-like phosphatases, RsbU and CT589; two antagonist proteins, RsbV1 and RsbV2; and one serine kinase, RsbW (Stephens et al., 1998; Mittenhuber, 2002). In contrast to the case with most free-living organisms (Koonin et al., 2000; Price, 2002; Mattoo et al., 2004), the genes encoding these potential regulators are not clustered, with the exception of the RsbU and CT589 phosphatase genes, which lie adjacent to one another. Transcriptional profiling studies have shown that all five genes are expressed, although there is some disagreement regarding the timing of this expression during the developmental cycle (Douglas and Hatch, 2000; Belland et al., 2003; Nicholson et al., 2003). We have incorporated these five components into the working model shown in Fig. 1B.

In this model the signalling pathway begins with RsbU and CT589. Characteristic of partner switching phosphatases (Mittenhuber, 2002; Price, 2002), these predicted products contain a C-terminal PP2C-like domain and an N-terminal domain capable of controlling phosphatase activity. Here, the N-terminal domains each contain two potential transmembrane helices that would position a significant loop external to the C. trachomatis membrane while maintaining the C-terminal phosphatase domain within the cytoplasm (Fig. 1C). For simplicity, the model shown in Fig. 1B suggests that each phosphatase is specific for one of the two antagonist proteins, RsbV1 or RsbV2. However, the absence of key residues within the PP2C domain of CT589 raises some uncertainty regarding its phosphatase activity, leading us to speculate that it may function as a regulatory subunit of RsbU (see Discussion).

As shown in Fig. 1C, RsbV1 and RsbV2 possess the sulphur transporter anti-sigma (STAS) domain that defines the antagonist protein family (Aravind and Koonin, 2000; Mittenhuber, 2002), and they also possess a conserved serine residue which is important for the phosphorylation of other family members by their cognate kinases (Najafi et al., 1995; Yang et al., 1996; Kozak et al., 2005). These cognate serine kinases are related to bacterial histidine protein kinases (Min et al., 1993; Mittenhuber, 2002), not to eukaryotic serine-threonine kinases, and RsbW is the only such kinase encoded on the C. trachomatis genome. As a member of the GHKL superfamily (Dutta and Inouye, 2000), RsbW possesses three conserved motifs of the Bergerat fold, the N, G1 and G2 boxes, which together bind Mg++ and ATP (Fig. 1C).

In the model shown in Fig. 1B, the phosphorylation states of the RsbV1 and RsbV2 antagonists are controlled by the balance between the activities of the input phosphatases and the RsbW kinase. The output of the partner switch could be either positive or negative, as shown. In either case, the model holds that signals external to the C. trachomatis membrane would modulate phosphatase activity to ultimately control the amount of free RsbW that can bind and regulate its target protein.

Whether the output of the switch is positive or negative, the model makes two strong predictions which allow us to test the validity of its core. First, the model predicts that RsbW has a serine kinase activity specific for the RsbV1 and RsbV2 antagonist proteins. Second, it predicts that the phosphorylation state of the RsbV1 and RsbV2 antagonists controls their physical interaction with RsbW. A third strong prediction – the specificity of the input phosphatases – could not be examined here because thus far it has proven problematic to expresses active recombinant RsbU or CT589, either as full-length or truncated proteins (L. Hua and C.W. Price, unpublished).

RsbW specifically phosphorylates RsbV1 and RsbV2

To test the prediction that RsbW has serine kinase activity towards RsbV1 and RsbV2, we first purified the three proteins from Escherichia coli as histidine-tagged recombinant proteins, together with B. subtilis RsbV and RsbW to serve as controls. C. trachomatis RsbW was then mixed with either RsbV1 or RsbV2 in the presence of [γ-32P]-ATP. As shown in Fig. 2A, RsbW phosphorylated RsbV1 (lane 1) and RsbV2 (lane 3), but not B. subtilis RsbV (lane 5). In complementary experiments, B. subtilis RsbW phosphorylated its cognate RsbV antagonist (lane 8), but not RsbV1 (lane 6) or RsbV2 (lane 7). We conclude that C. trachomatis RsbW has a kinase activity specific for RsbV1 and RsbV2.

Figure 2.

RsbW specifically phosphorylates the RsbV1 and RsbV2 antagonist proteins.
A. Purified RsbV substrates (10 µg) and RsbW kinases (1 µg) were mixed and incubated with [γ-32P]-ATP as described in Experimental procedures. After 30 min at 37°C, the reaction mixtures were analysed by SDS-PAGE and autoradiography. Lanes 1–5, C. trachomatis (Ct) RsbW was incubated with RsbV1 (lane 1), RsbV1 S56A (lane 2), RsbV2 (lane 3), RsbV2 S55A (lane 4) and B. subtilis (Bs) RsbV (lane 5). Lanes 6–8, B. subtilis RsbW was incubated with C. trachomatis RsbV1 (lane 6), RsbV2 (lane 7) and B. subtilis RsbV (lane 8). Positions of molecular mass standards (kDa) are on the left.
B. clustal w alignment (Thompson et al., 1994) of C. trachomatis RsbV1 and RsbV2 with the B. subtilis RsbV, RsbS and SpoIIAA antagonist proteins. Identical residues are indicated by asterisks (*) and conserved substitutions by colons (:), with the conserved serine in larger size. Serine 58 in SpoIIAA is the site of phosphorylation by its cognate kinase (Najafi et al., 1995), and serine 56 in RsbV and serine 59 in RsbS are required for phosphorylation (Yang et al., 1996). For B. subtilis RsbS and SpoIIAA, alteration of this conserved serine to either alanine or aspartate has been shown to have opposite regulatory effects in vivo (Diederich et al., 1994; Kang et al., 1996) and, for RsbV and RsbS, opposite consequences in a yeast two-hybrid interactive assay (Yang et al., 1996). The arrows indicate the serine to alanine (A) and serine to aspartate (D) substitutions introduced into C. trachomatis RsbV1 and RsbV2.

We also performed the kinase reactions using mutant forms of RsbV1 and RsbV2 in which the presumed sites of phosphorylation were changed to alanine (Fig. 2B). In this case, there was no detectable phosphorylation of RsbV1 and RsbV2 by RsbW (lanes 2 and 4 of Fig. 2A). These results indicate that the conserved serine 56 and serine 55 are important for phosphorylation of RsbV1 and RsbV2, respectively, and suggest these residues are the sole sites of phosphorylation on the C. trachomatis proteins.

Tandem mass spectrometry identifies the phosphorylated residues in RsbV1 and RsbV2

To directly identify the phosphorylated peptides in RsbV1 and RsbV2, we first performed the in vitro kinase reactions with unlabelled ATP, then carried out matrix-assisted laser desorption ionization-time of flight (MALDI-TOF) mass spectrometry-based peptide mass mapping of the proteins both before and after alkaline phosphatase treatment. This analysis (not shown) identified one phosphorylated tryptic peptide from RsbV1 (NIILDCGDLDYISSAGIR) and one from RsbV2 (VFYMSSAGLR). Each of these peptides contains three potential sites for serine or tyrosine phosphorylation. In order to identify the exact site of phosphorylation, the protein mixtures that directly resulted from the in vitro kinase reactions were cleaved with trypsin and the peptides analysed by capillary liquid chromatography-tandem mass spectrometry (LC-MS/MS).

The LC-MS/MS results for the tryptic RsbV1 peptides are illustrated in Fig. 3A. From the LC-MS survey scan, the doubly charged, phosphorylated peptide was identified and fragmented. The fragment ion assignment in the CID spectrum revealed the presence of the y5 ion at m/z 583.2, y5-H3PO4 at m/z 485.3 and the b13 ion at 1492.7 for the sequence NIILDCGDLDYISSpAGIR, and the absence of any y5, b13 or b13-H3PO4 ions for the sequence NIILDCGDLDYISpSAGIR. This assignment of the y5 and b13 ions unambiguously identified S56 as the only detectable site of phosphorylation in RsbV1.

Figure 3.

RsbW phosphorylates RsbV1 and RsbV2 on a conserved serine residue. MS/MS spectra of the phosphopeptide ions [NIILDCGDLDYISSpAGIR+2H]2+ from RsbV1 and [VFYMSSpAGLR+2H]2+ from RsbV2, obtained from LC-MS/MS.
A. y5 and b13 are characteristic fragment ions indicating that RsbV1 is phosphorylated on S56. This spectrum also shows the neutral loss of phosphoric acid typical of phosphopeptides (y6-H3PO4 and y6-H3PO4-H2O).
B. y5 and b5 are characteristic fragment ions indicating that RsbV2 is phosphorylated on S55.

The results for the tryptic RsbV2 peptides are shown in Fig. 3B. After the doubly charged, phosphorylated peptide was identified and fragmented, analysis of the CID spectrum found the y5, y5-H3PO4 and b5 ions for the sequence VFYMSSpAGLR, but no y5, b5 or b5-H3PO4 ions for VFYMSpSAGLR. Thus, S55 can be unequivocally assigned as the only detectable site of phosphorylation in RsbV2.

The sequence coverage in these LC-MS/MS experiments was 100% for RsbV1 and 95% for RsbV2, missing only two short tryptic peptides. To address the uncovered sequence of RsbV2, we performed another independent digestion with subtilisin and confirmed the S55 site (data not shown). These MS/MS data, together with the results of the in vitro kinase assays with wild and mutant proteins, indicate that RsbW phosphorylates RsbV1 primarily on conserved serine 56 and RsbV2 primarily on conserved serine 55.

Alteration of the conserved serine in RsbV1 or RsbV2 affects their interaction with RsbW

The yeast two-hybrid system was used to test the prediction that the phosphorylation state of RsbV1 and RsbV2 controls their interaction with RsbW. In this system, the strength of protein–protein interactions can be estimated by the degree to which transcription of a lacZ reporter gene is activated by complementary fusions of the tested proteins with the yeast GAL4 DNA-binding and -activation domains (Fields and Song, 1989). This two-hybrid system has proven to be an accurate predictor of the interactions between wild and mutant partner switching regulators from the B. subtilisσB network (Voelker et al., 1996b; Yang et al., 1996; Kang et al., 1998; Akbar et al., 2001; Delumeau et al., 2004; Kim et al., 2004).

To perform the yeast two-hybrid analysis, two kinds of substitutions were made at the conserved serine residues of C. trachomatis RsbV1 and RsbV2: a serine to alanine alteration, which cannot be phosphorylated, and a serine to aspartate alteration, which is thought to mimic the serine residue in its phosphorylated state (Diederich et al., 1994; Kang et al., 1996; Yang et al., 1996). As shown in Table 1, the alanine-substituted RsbV1 and RsbV2 interacted strongly with RsbW in the yeast two-hybrid system, whereas the aspartate-substituted versions manifested no significant interactions with RsbW. Interestingly, wild-type RsbV1 and RsbV2 also failed to interact with RsbW. A likely explanation for this latter result is that wild-type RsbV1 and RsbV2 were so highly phosphorylated by RsbW that the interactions were abolished. This explanation is reinforced by the level of interaction noted between B. subtilis RsbV and RsbW in yeast cells, where the wild-type RsbV–RsbW interaction gave only 24% of the activity of the RsbVS56A–RsbW interaction (Table 1).

Table 1.  Yeast two-hybrid analysis of RsbW interactions with wild and mutant RsbV antagonist proteins.a
Protein fused to GAL4 DNA-binding domainProtein fused to GAL4 trans-activating domainb
Chlamydia trachomatis RsbW Bacillus subtilis RsbW
  • a

    . Values shown are the average β-galactosidase activities (in units per mg of protein) of the indicated pairwise comparisons, each determined from three independent double transformants. Intrinsic activity of each single transformant was 1–2 units, which defines the background activity for the double transformants.

  • b

    . Both RsbW proteins had high intrinsic activity when fused with the GAL4 DNA-binding domain, so only the values for RsbW proteins fused with the GAL4 trans-activating domain are shown here.

Chlamydia trachomatis
 RsbV1 S56A570 5
 RsbV1 S56D 2 1
 RsbV1 (wild type) 1 1
Chlamydia trachomatis
 RsbV2 S55A448 1
 RsbV2 S55D 2 1
 RsbV2 (wild type) 1 1
Bacillus subtilis
 RsbV S56A 4795
 RsbV (wild type) 6190

The interactions observed between C. trachomatis RsbV1, RsbV2 and RsbW were specific because RsbW did not interact with either wild or mutant B. subtilis RsbV, and B. subtilis RsbW did not interact with either wild or mutant RsbV1 and RsbV2 (Table 1). We conclude that the conserved serine residue is important for the interaction of either RsbV1 or RsbV2 with RsbW in the yeast two-hybrid system. Moreover, we infer from the opposite phenotypes of the alanine and aspartate substitutions that the phosphorylation state of RsbV1 and RsbV2 controls their interaction with RsbW.

Is a σ factor the target of the C. trachomatis switch?

In Gram-positive bacteria the target of a partner switch is often a σ factor, with the switch protein/kinase serving as an anti-σ factor that negatively controls transcriptional activity by its direct association with σ (Benson and Haldenwang, 1993; Duncan and Losick, 1993; Min et al., 1993; Beaucher et al., 2002; Lee et al., 2004). Considering that the pattern of transcription changes in an ordered manner as infection proceeds (Shaw et al., 2000; Belland et al., 2003; Nicholson et al., 2003), the three σ factors encoded by the C. trachomatis genome become potential switch targets. This possibility was explored using two methods effective in the B. subtilis system: (i) the yeast two-hybrid assay and (ii) a direct biochemical test for anti-σ activity.

We first conducted a yeast two-hybrid analysis of the potential interaction between RsbW and the three C. trachomatisσ factors: σ66, σ54 and σ28. As shown in Table 2, none of these σ factors appeared to interact with RsbW in the yeast two-hybrid system. In contrast, the B. subtilisσB control did interact strongly with RsbW from B. subtilis, as was previously found (Voelker et al., 1996b; Yang et al., 1996), and not with RsbW from C. trachomatis. The yeast two-hybrid analysis therefore provides no support for the notion that C. trachomatis RsbW complexes any of the σ factors encoded by the C. trachomatis genome.

Table 2.  Yeast two-hybrid analysis of RsbW interaction with σ factors.a
Protein fused to GAL4 DNA-binding domainProtein fused to GAL4 trans-activating domainb
Chlamydia trachomatis RsbW Bacillus subtilis RsbW
  • a

    . Values shown are the average β-galactosidase activities (in units per mg of protein) of the indicated pairwise comparisons, each determined from three independent double transformants. Intrinsic activity of each single transformant was 2–3 units, with the exception of B. subtilisσB, which was 7 units.

  • b

    . Both RsbW proteins had high intrinsic activity when fused with the GAL4 DNA-binding domain, so only the values for RsbW proteins fused with the GAL4 trans-activating domain are shown here.

Chlamydia trachomatisσ28 4  5
Chlamydia trachomatisσ54 4  6
Chlamydia trachomatisσ66 5  9
Bacillus subtilisσB102510

It is possible that the negative results from the yeast two-hybrid analysis reflect features of the assay that are not directly related to protein–protein interactions. For example, the σ factor of interest might be weakly expressed or unstable in yeast cells, or poorly targeted to the yeast nucleus. We therefore directly tested the ability of RsbW to inhibit its most likely target, σ28, using the σ28-dependent hctB promoter (Yu and Tan, 2003) as template for in vitro transcription assays. Our reasoning was that σ28 belongs to the flagellar-stress-sporulation branch of the σ70 family (Gruber and Gross, 2003; Shen et al., 2004), and in Gram-positive bacteria members of this branch are under the direct control of anti-σ factors that are RsbW orthologues (Beaucher et al., 2002; Price, 2002; Lee et al., 2004). Conversely, σ66 is the major σ factor of C. trachomatis (Koehler et al., 1990), and the only known anti-σ for such factors belong to the AsiA or Rsd families (Pineda et al., 2004), none of which are encoded by the C. trachomatis genome. Finally, σ54 is a member of the NtrA family of σ factors, none of which have been reported to be under anti-σ control.

As shown in Fig. 4A. C. trachomatisσ28 was purified from E. coli as a histidine-tagged recombinant protein with a predicted mass of 31.5 kDa; like other acidic sigma factors it migrates with a greater apparent mass in SDS-PAGE (Shen et al., 2004). As shown in Fig. 4B, transcriptional activity of RNA polymerase core on the hctB template was increased about 350-fold with the addition of this purified σ28. Notably, pre-incubation of σ28 with up to 160-fold molar excess of C. trachomatis RsbW had no effect on the amount of hctB message produced. The fact that this same preparation of RsbW retained its serine kinase activity (Fig. 2A) suggests that expression of the protein in E. coli did not cause significant alterations to its structure or activity.

Figure 4.

In vitro transcription directed by C. trachomatisσ28 is unaffected by RsbW.
A. Coomassie-stained SDS-PAGE of purified recombinant σ28. Lane 1, molecular mass standards (in kDa); lanes 2–4, 0.2, 1 and 5 µg of σ28.
B. Levels of σ28-dependent transcript from an hctB–lacZ template measured by quantitative PCR and expressed as fold increase relative to the reaction with RNA polymerase core lacking σ28. The left-hand bar (Eσ28) shows the control with no added RsbW, and the remaining bars show that pre-incubation of σ28 with up to 160-fold molar excess of RsbW had no significant effect on σ28-dependent transcription.
C. Levels of σ28-dependent transcript measured as described in (B). Left-hand bar (Eσ28) shows the untreated control and the remaining bars show the effects of pre-incubation with either preserum or anti-Eσ28.
D. Levels of σ70-dependent transcript measured from dnaK–lacZ template. Left-hand bar (Eσ70) shows the untreated control and the remaining bars show the effects of pre-incubation with either preserum or anti-Eσ28.

Because C. trachomatis and E. coliσ28 have similar recognition specificities (Yu and Tan, 2003; Shen et al., 2004), it was important to demonstrate that the transcriptional activity shown in Fig. 4B was dependent on chlamydial σ28 and not the result of a contaminating E. coli protein. Yu and Tan (2003) reported that polyclonal antibody raised against chlamydial σ28 does not cross-react with E. coliσ28. We confirmed this observation with our own polyclonal antibody (data not shown), then used the antibody to test the specificity of the transcription reaction. Addition of pre-immune sera reduced transcription by 30% in both the σ28-dependent (Fig. 4C) and E. coliσ70-dependent control reactions (Fig. 4D), suggesting a modest, non-specific inhibition by serum alone. Notably, the anti-chlamydial σ28 antibody caused a 25-fold further reduction of transcription in the σ28 reaction, but no significant further reduction in the σ70 control. These experiments show that the transcriptional activity measured at the hctB promoter is dependent on C. trachomatisσ28, and that the assay can detect transcriptional inhibition by proteins which specifically bind C. trachomatisσ28. We therefore conclude that RsbW alone does not function as an anti-σ factor in this in vitro assay. Thus, two independent methods – the yeast two-hybrid assay and in vitro transcription assays – provide no support for the idea that σ28 is the target of RsbW.

Discussion

The genome of the intracellular pathogen C. trachomatis encodes all the components of a partner switching signalling pathway, allowing us to develop the working model shown in Fig. 1B. Here we have tested two key predictions of this model, and in so doing have demonstrated that its central components have the same functional properties as well-characterized partner switching regulators from B. subtilis. First, from the biochemical analysis reported here (Figs 2 and 3), we conclude that RsbW has a serine kinase activity specific for RsbV1 and RsbV2. Second, from the yeast two-hybrid analysis (Table 1), we draw the strong inference that phosphorylation of RsbV1 on serine 56 and RsbV2 on serine 55 controls their interaction with RsbW. Thus, the core of the partner switching mechanism is highly conserved between B. subtilis and C. trachomatis. Because the core is conserved over such a large evolutionary span, the versatility to accomplish different signalling tasks depends on alterations to the input and the output of the switch, as we shall now discuss.

The input

Here the likely input phosphatases RsbU and CT589 each have a predicted extracytoplasmic region which could sense signals exterior to the C. trachomatis membrane and modulate activity of its companion phosphatase domain within the cytoplasm. The switch controlled by these input phosphatases could conceivably function immediately upon binding of the elementary body to the host cell or during events subsequent to entry, including establishing and maintaining the highly specialized inclusion environment. Because the extracytoplasmic regions of RsbU and CT589 have different primary sequences, each would be capable of detecting a different signal and acting at a different time in the developmental cycle.

Although this sensing model is attractive, analysis of the RsbU and CT589 phosphatase domains suggests that it is unlikely to be correct in its simplest form. Figure 5 shows an alignment of the phosphatase domains from B. subtilis SpoIIE and RsbU – which control the σF and σB partner switches, respectively – together with the corresponding domains from C. trachomatis RsbU and CT589. Conspicuously absent from the CT589 sequence are three of four invariant aspartate residues which from structural studies are known to co-ordinate Mn++ (or Mg++) atoms within the active sites of PP2C phosphatases (Das et al., 1996; Pullen et al., 2004). Notably, alteration of one or another of these aspartate residues has a profound negative effect on the activity of B. subtilis SpoIIE (Schroeter et al., 1999), yeast TPD1 (Adler et al., 1997) or human PP2Cα (Jackson et al., 2003). Although it remains possible that CT589 is sufficiently divergent that other residues could assume this catalytic function, it is more probable that CT589 lacks phosphatase activity entirely.

Figure 5.

Alignment of PP2C phosphatase domains shows four invariant aspartate residues. clustal w alignment showing the invariant aspartate residues that co-ordinate two Mn++ atoms in the active sites of solved PP2C phosphatase structures (Das et al., 1996; Pullen et al., 2004), numbered here according to the residues of human PP2C (not shown). C. trachomatis CT589 lacks D38, D60 and D239 which are required for function in B. subtilis SpoIIE (Schroeter et al., 1999) and other PP2C phosphatases (see text). The asterisks (*) indicate identical residues in all four proteins and the colons (:) indicate conserved substitutions.

Nonetheless, the preservation of similar RsbU–CT589 gene pairs in other members of the Chlamydiaceae(Table 3) argues that CT589 has an important function. In addition, the fact that such CT589-like proteins are found only within this family implicates them in sensing a signal unique to the chlamydial developmental cycle. We can imagine two ways by which information sensed by the extracytoplasmic domain of CT589 could be conveyed to the partner switch. First, the cytoplasmic phosphatase domain of CT589 could interact with that of RsbU and directly control its activity. A precedent for such an interaction between phosphatase domains comes from the unusual two-domain PP2C phosphatase of Plasmodium falciparum (Mamoun et al., 1998). Second, CT589 could recognize and bind RsbV1 or RsbV2 in a long-lived, catalytically inactive complex, thereby removing it from interaction with RsbW. These hypotheses might be tested in vitro when active recombinant RsbU and CT589 become available.

Table 3.  Conservation of adjacent RsbU and CT589 loci among the Chlamydiaceae.
OrganismRsbU orthologueCT589 orthologue
Accession No.a E-valuebAccession No.a E-valueb
  • a

    . Each orthologue is indicated by its accession number in the NCBI Protein Database.

  • b

    . Random expectation (E) value for the alignment resulting from ablastp search (Altschul et al., 1997) of proteins available in the NCBI Microbial Genomes Database, using the default parameters of the on-site program.

Chlamydia trachomatis (serovar D)NP_220103.10.0NP_220104.10.0
Chlamydia muridarum NP_297250.10.0NP_297251.10.0
Chlamydia abortus YP_220318.10.0YP_220319.13e-146
Chlamydia caviae NP_829825.10.0NP_829826.11e-145
Chlamydia pneumoniae NP_2249880.0NP_2249871e-126
Bordetella brochiseptica[BtrU]NP_888187.15e-21  
Bacillus subtilis[RsbU] P40399 3e-11  
Chlamydia trachomatis[CT589]NP_220104.16e-08  

The output

Although control of σ factor activity as the developmental cycle unfolds is a logical function for the C. trachomatis partner switch, we could find no experimental support for the notion that the switch target is a transcription factor. RsbW does not activate the yeast two-hybrid system when paired with any of the three σ factors encoded by the C. trachomatis genome (Table 2), nor does it manifest anti-σ activity towards its most likely target, σ28, during an in vitro transcription assay (Fig. 4). These are two criteria met by well-characterized anti-σ factor orthologues such as B. subtilis RsbW (Benson and Haldenwang, 1993; Alper et al., 1996; Voelker et al., 1996b; Yang et al., 1996). It is of course possible that C. trachomatis RsbW uniquely requires another protein to bind its target, and this hypothetical protein would be missing from the yeast two-hybrid and transcription assays. However, based on the prior example of the B. subtilisσB signalling network, where one switch target is the RsbU environmental-signalling phosphatase (Yang et al., 1996; Kang et al., 1998; Delumeau et al., 2004), we consider it more likely that the target of the C. trachomatis switch is not a σ factor but is instead another protein that post-translationally controls an important response function.

An intriguing clue to this response function comes from the work of Miller and colleagues, who found that a similar partner switching pathway exerts post-translational control on type III secretion in the Gram-negative pathogen B. bronchiseptica (Mattoo et al., 2004; Kozak et al., 2005). The B. bronchiseptica genome encodes clear orthologues to C. trachomatis RsbU, RsbV and RsbW (Mattoo et al., 2004), but the second RsbV antagonist and the non-conforming CT589 protein are missing. Notably, the B. bronchiseptica RsbU orthologue, called BtrU, has a predicted N-terminal, extracytoplasmic domain much like that of C. trachomatis RsbU, leading to the proposal that BtrU functions as a membrane-bound sensor that transmits signals to the cytoplasmic BtrV and BtrW proteins (Mattoo et al., 2004; Kozak et al., 2005). Because the B. bronchiseptica BtrW switch protein is a positive regulator of type III secretion that does not appear to affect transcription of type III genes, Miller and colleagues have further proposed that the direct target of their switch is a hypothetical protein which essentially gates the secretory apparatus. Thus, the current model for the B. bronchiseptica pathway (Kozak et al., 2005) echoes that of the environmental stress signalling pathway in B. subtilis, where the output of the switch is positive and the direct target of the RsbT switch protein is not a transcription factor but an enzyme – the RsbU environmental-signalling phosphatase.

Secretion of protein effectors via the type III secretion system allows the extracellular pathogen B. bron-chiseptica to resist host defences and persistently colonize the trachea of immunocompetent mice (Yuk et al., 2000). A similar secretion of protein effectors is proposed to mediate pivotal events in the developmental cycle of the intracellular pathogen C. trachomatis, including initial uptake (Clifton et al., 2004), early inclusion modifications (Fields et al., 2003) and later evasion of host adaptive immunity (Zhong et al., 2001). Based on the B. bronchiseptica example, it becomes an intriguing possibility that C. trachomatis RsbU (and CT589) sense host signals in order to modulate secretion of key effector proteins by means of RsbV1, RsbV2 and RsbW. In this view, the target of the C. trachomatis switch would be a protein that regulates activity of the type III secretion apparatus at the post-translational level. It is possible that the identity of the hypothetical target protein will be suggested using in vitro approaches. But a definitive test of this proposal will likely require development of a workable genetic system and the ability to grow the organism in cell-free media.

What advantage might be conferred by two RsbV antagonist proteins?

Among experimentally characterized partner switching networks, the presence of multiple antagonist proteins interacting with a single switch protein is uncommon but not altogether unknown. For example, the Mycobacterium tuberculosis genome encodes seven RsbV-like antagonists (Koonin et al., 2000). The interactions of two of these antagonist proteins with the UsfX switch protein has been studied in vitro (Beaucher et al., 2002), and they respond to signals via different mechanisms – one by cysteine residues sensitive to reducing conditions and the other by a serine residue presumably modified by an unidentified kinase. In contrast, the data presented here support the hypothesis that both RsbV1 or RsbV2 of C. trachomatis respond via the same mechanism – serine modification.

The simple model shown in Fig. 1B suggests that RsbV1 and RsbV2 are each specifically dephosphorylated by one of the input phosphatases, allowing each antagonist to respond to a discrete signal detected by the extracellular domain of its cognate phosphatase. If the CT589 protein lacks phosphatase activity as we suspect, then the model shown in Fig. 1B is unlikely to be correct in detail. Instead, the phosphorylation states of RsbV1 and RsbV2 would reflect the balance of the RsbU phosphatase and RsbW kinase activities. Depending on the binding constants that the RsbV1 and RsbV2 antagonists manifest towards RsbW, their cellular concentrations during the developmental cycle, and their relative fitness as substrates for the RsbU phosphatase and RsbW kinase, one can imagine a two-stage release of RsbW by RsbV1 and RsbV2. Presumably this two-stage release would be orchestrated by the RsbU phosphatase in response to a signal received by its extracellular domain, and further modulated by the regulatory activity of the CT589 protein.

In summary, our results and those of others link the partner switching mechanism with the transmission of stress signals. Among free-living Gram-positive bacteria, this mechanism is commonly used in signalling pathways that control an environmental stress response (Beaucher et al., 2002; Price, 2002; Lee et al., 2004). Based on the predicted membrane topologies of C. trachomatis RsbU and CT589, we suggest that the partner switching mechanism is similarly employed here within the more specialized environment of the host cell.

Experimental procedures

Bacterial strains and genetic methods

Escherichia coli DH5α (Gibco BRL) was the host for all plasmid constructions. Standard recombinant DNA methods were used for polymerase chain reaction (PCR), restriction digests, ligations and transformations of E. coli, as described by Sambrook et al. (1989). Oligonucleotide primers were designed according to the published genome sequence (Stephens et al., 1998) and are available upon request. Each open reading frame was amplified by PCR using C. trachomatis serovar D genomic DNA. These products were ligated into the pCR-Blunt II-TOPO vector (Invitrogen, Carlsbad, CA) to create plasmids which then served as the basis of subsequent constructions. The DNA sequence was confirmed in these and in all following constructs.

To introduce the coding sequences for wild-type RsbV1, RsbV2, RsbW and the three C. trachomatisσ factors into the yeast two-hybrid vectors, we used the Matchmaker Two-Hybrid System (Clontech, Palo Alto, CA). The relevant plasmids were digested with restriction enzymes and cloned into either pGBT9 (for fusions to the yeast GAL4 DNA-binding domain) or pGAD424 (for fusions to the yeast GAL4-activation domain). In each case, the expected fusion junction was verified by DNA sequence analysis. For controls we used the previously constructed vectors in which B. subtilis RsbV and RsbW had been fused to the GAL4-binding or -activation domains (Yang et al., 1996).

The four-primer method of site-directed mutagenesis (Ho et al., 1989) was used to make alanine and aspartate substitutions of RsbV1 and RsbV2 in the yeast two-hybrid vectors. Each PCR-mutagenized product was cloned into the pCR-Blunt II-TOPO vector. We then treated the resultant plasmid with EcoRI and BamHI and ligated the fragment of interest into the yeast two-hybrid vectors pGBT9 and pGAD424.

To facilitate purification of wild-type and mutant RsbV1, RsbV2 and RsbW proteins, the relevant plasmids were digested with restriction enzymes and the fragments of interest cloned into the pET15b expression vector (Novagen, Madison, WI). pET15b fuses a six-histidine tag and a thrombin cleavage site to the protein encoded by the fragment and also places the fusion construction under control of a LacI-repressible, IPTG-inducible T7 promoter. These fusion junctions were also confirmed by DNA sequencing. In a slight variation, the σ28 coding sequence was directly amplified and cloned into a related vector, pET21a, then sequenced to confirm both the insert and the junctions.

Overexpression and purification of recombinant proteins

Plasmids encoding three wild-type and two mutant Rsb proteins in the pET15b vector were transformed into E. coli BL21 (DE3). After inducing their expression by adding 1 mM IPTG (final) to the growth medium, proteins were purified under native conditions on nickel affinity columns (Qiagen, Valencia, CA) according to the manufacturer's protocol. To purify the B. subtilis Rsb proteins, we used the previously constructed pET15b clones that carry the wild-type rsbV and rsbW coding regions (Yang et al., 1996; Kang et al., 1998). Concentration of each purified Rsb protein was determined with the Protein Assay Reagent (Bio-Rad Laboratories, Richmond, CA).

The limited solubility of C. trachomatisσ28 required a modification of this basic method. The protein was expressed in freshly transformed E. coli BL21 (DE3) grown to an A600 of 0.6 at which point the system was induced with 1 mM IPTG for 1 h. Subsequent purification used the Novagen His-bind kit, essentially according to the manufacturer's protocol (Novagen, Madison, WI). To increase σ28 solubility, 0.25% (w/v) sarkosyl was added before sonication, after which σ28 was purified from the filtered supernatant using a Ni+ column, as described by Novagen. Eluted fractions of purified σ28 were then dialysed at 4°C against three changes of TEG buffer [10 mM Tris-HCl pH 7.6, 0.1 mM EDTA and 5% (v/v) glycerol] containing decreasing amounts of sarkosyl – 0.025%, 0.0025% and 0 – with 12 h for each buffer exchange.

In vitro kinase assay

The phosphorylation assay was performed in a 30 µl reaction volume containing kinase buffer (50 mM Tris-HCl pH 7.6, 50 mM KCl, 10 mM MgCl2, 1 mM DTT and 0.1 mM EDTA). Ten-microgram substrate and 1 µg kinase were added, and the reaction was started by the addition of a mixture of 15 µCi [γ-32P]-ATP and 1 mM unlabelled ATP. The reaction was incubated at 37°C for 30 min, then terminated by the addition of 7 µl of 5 × sample loading buffer [2% (w/v) SDS, 5% (v/v) glycerol, 60 mM Tris-HCl pH 6.8, 100 mM DTT and 0.1% (w/v) bromphenol blue]. Samples were separated on 15% SDS-polyacrylamide gels. Protein bands were stained with Coomassie blue and the phosphorylated proteins detected by autoradiography.

Capillary liquid chromatography with electrospray-ionization tandem mass spectroscopy

Kinase reactions were conducted as described above, but omitting the labelled ATP. In-solution trypsin digestion of the reaction mixes (containing both the RsbV substrate of interest and the RsbW kinase) was used to acquire LC-MS/MS data via an high-performance liquid chromatography (HPLC) system (Paradigm MG4, Michrom Bioresources, Auburn, CA) directly coupled to an ion trap mass spectrometer (LCQ Deca XP plus, Thermo Finnigan, San Jose, CA). Tryptic peptides were separated using a fritless reverse phase capillary column (0.1 × 140 mm) packed with C18 (Aqua, 5 µm, 300 Å, Phenomenex, Torrance, CA) and directly sprayed into the mass spectrometer (Gatlin et al., 1998). This chromatographic separation was performed using buffers A (0.1% formic acid with 5% acetonitrile, v/v) and B (0.1% formic acid with 80% acetonitrile), with a 2.5 h long gradient (0–10% B, 10 min; 10–45% B, 110 min; 45–100% B, 20 min; 100% B, 10 min).

All MS/MS spectra were searched against a small database containing only the RsbV1, RsbV2 and RsbW sequences, using the Sequest algorithm (Eng et al., 1994) with possible phosphorylation on tyrosine, threonine, and serine. The MS/MS spectra for the phosphopeptides were then manually re-confirmed. Sequence coverage was calculated using the Sequest criteria suggested by Peng et al. (2003), which gave 99% confidence for the yeast proteome.

Yeast two-hybrid assay

To test interactions between pairs of C. trachomatis proteins, we followed the manufacturer's instructions for the Matchmaker Two-Hybrid System (Clontech). A pGAD424 construction encoding one protein fused to the GAL4 activating domain was co-transformed into the Saccharomyces cerevisiae SFY526 host strain together with a pGBT9 construction encoding another protein fused to the GAL4 DNA-binding domain. Double transformants were selected on minimal medium, and three independent transformants for each combination of plasmids were purified for further use. Each independent transformant was grown in minimal medium, harvested during exponential growth, then assayed for β-galactosidase activity essentially according to Miller (1972), except that yeast extracts were prepared by treating the cell suspensions to freeze–thaw cycles, alternating between dry ice-ethanol and 37°C. Protein levels were determined using the Protein Assay Reagent (Bio-Rad Laboratories, Richmond, CA). Specific activity was defined as ΔA420 × 1000 per minute per mg of protein.

In vitro transcription assays

In vitro transcription reactions were performed by adding 1.5 pmoles of core RNA polymerase (Epicentre, Madison, WI) and 40 units RNasin (Promega, Madison, WI) to transcription buffer (40 mM Tris-HCl pH 8.0, 300 mM KOAc, 10 mM MgCl2 and 5 mM DTT). For the basic assay, 15 pmoles of σ28 in TEG buffer (or TEG buffer alone) was added and the reactions were allowed to incubate for 10 min at 22°C. To analyse the inhibitory effect that RsbW might have on σ28-directed transcription, increasing molar concentrations (60–2400 pmoles) of RsbW was incubated with σ28 for 10 min before the addition of the RsbW–σ28 mix to the core holoenzyme reactions, which were then allowed to incubate for a further 10 min at 22°C. Fifty fmoles of template DNA was then added to the reaction mixes; template DNA consisted of pACYC184 containing the E. coli lacZ gene under the control of the σ28-dependent C. trachomatis (strain L2) hctB promoter (Yu and Tan, 2003). The transcription assays were initiated by the addition of 300 µM NTPs, incubated at 30°C for 20 min, then terminated by heating at 70°C for 10 min. Reactions were placed on ice, treated with 5 units of RQ1 RNase-free DNase (Promega), and the transcribed RNA was purified using an RNeasy kit (Qiagen, Valencia, CA) according to the manufacturer's protocol.

To determine the specificity of C. trachomatisσ28-directed transcription, we raised mouse polyclonal antibody against C. trachomatisσ28 and analysed it for cross-reactivity to E. coliσ28. E. coli lysates containing vector alone (pBAD-TOPO TA), plasmid expressing full-length E. coli fliA or plasmid expressing full-length C. trachomatis fliA were separated by SDS-PAGE and immunoblotted with a 1:500 dilution of anti-C. trachomatisσ28; comparable expression of each σ28 protein was confirmed by a parallel immunoblot with anti-histidine tag antibody (data not shown). To directly test the specificity of C. trachomatisσ28-directed transcription, either preserum or anti-σ28 was diluted 1:100 and incubated with σ28 for 10 min before addition to the transcription reactions. To measure any non-specific effect of these sera, we included control reactions containing an equimolar amount of E. coli70 (Epicentre). Eσ70-directed transcription was measured on template DNA consisting of pACYC184 containing E. coli lacZ under control of the σ70-dependent C. trachomatis (strain L2) dnaK promoter (Schaumburg and Tan, 2003).

The amount of transcribed RNA in each reaction was determined by quantitative PCR. cDNA was generated from equal volumes of RNA eluted from each reaction, using a 3′lacZ primer (GGGCGCATCGTAACCGTGCATCTGC) and Superscript III Reverse Transcriptase (Invitrogen, Madison, WI) according to the manufacturer's protocol. Quantitative PCR was performed on 1:500 dilutions of cDNA using an Applied Biosystems 7500 real-time PCR detection system (Applied Biosystems, Foster City, CA). Reactions contained 300 nM primers (lacZ real-time 5′-GCTGGCGTAATAGCGAA GAGG and 3′ TTGAGGGGACGACGACAGTATC) and 2× SYBR Green PCR Master Mix (Applied Biosystems). Threshold fluorescence was established within the geometric phase of exponential amplification and the cycle of threshold (Ct) determined for each reaction. Fold change in transcript (Eσ28 versus core or Eσ70 versus core) was determined using the 2–ΔΔCT method (Livak and Schmittgen, 2001).

Acknowledgements

This research was supported by Public Health Service Grant GM42077 from the National Institute of General Medical Sciences (C.W.P.); by a UC Davis Faculty Research grant (C.W.P.); and by National Research Service Award AI58490 (P.S.H.) and Public Health Service Grant AI42156 (R.S.S.) from the National Institute of Allergy and Infectious Diseases.

We wish to thank Margaret Brody for her assistance in constructing the alanine-substituted mutant forms of RsbV1 and RsbV2.

Ancillary