The Rsb Phosphoregulatory Network Controls Availability of the Primary Sigma Factor in Chlamydia trachomatis and Influences the Kinetics of Growth and Development

Chlamydia trachomatis is an obligate intracellular human pathogen that exhibits stage-specific gene transcription throughout a biphasic developmental cycle. The mechanisms that control modulation in transcription and associated phenotypic changes are poorly understood. This study provides evidence that a switch-protein kinase regulatory network controls availability of σ66 , the main sigma subunit for transcription in Chlamydia. In vitro analysis revealed that a putative switch-protein kinase regulator, RsbW, is capable of interacting directly with σ66, as well as phosphorylating its own antagonist, RsbV1, rendering it inactive. Conversely, the putative PP2C-like phosphatase domain of chlamydial RsbU was capable of reverting RsbV1 into its active state. Recent advances in genetic manipulation of Chlamydia were employed to inactivate rsbV1, as well as to increase the expression levels of rsbW or rsbV1, in vivo. Representative σ66-dependent gene transcription was repressed in the absence of rsbV1 or upon increased expression of RsbW, and increased upon elevated expression of RsbV1. These effects on housekeeping transcription were also correlated to several measures of growth and development. A model is proposed where the relative levels of active antagonist (RsbV1) and switch-protein anti-sigma factor (RsbW) control the availability of σ66 and subsequently act as a molecular 'throttle' for Chlamydia growth and development.


Introduction
Chlamydia trachomatis is the leading cause of bacterial sexually transmitted infection worldwide [1], as well as the leading cause of infection-associated blindness [2]. Members of the Chlamydiaceae are obligate intracellular parasites that must complete a unique intracellular development cycle in order to propagate. This cycle is characterized by phenotypic variation between an Elementary Body (EB) that is infectious but metabolically quiescent, and a Reticulate Body (RB) that is replicative but not infectious. Cellular infection is initiated by the EB, which attaches and induces its own intake via translocation of effector cargo. Once inside the cell, the EB differentiates into an RB, which replicates via binary fission. During the late stages of infection, RBs asynchronously re-differentiate into the EB form (reviewed in [3]). Alternatively, certain stress conditions mediate the onset of a separate growth mode (termed persistence), which was defined as a 'viable but non-cultivable' state, in which chlamydiae fail to complete the development cycle and instead differentiate into aberrant, enlarged particles; this phenotype is reversible upon abatement of the mediating stress [4]. While modulations in transcript levels are associated with the stages of acute development [5][6][7] and the onset of persistence [8,9], the regulatory mechanisms that govern these phenotypic shifts are not understood.
Sigma factors are responsible for the recruitment of the core RNA polymerase (RNAP) to cognate promoter elements, and, thus, their function dictates the subset of genes transcribed within a cell. Chlamydia encodes three sigma factors, σ 66 , σ 54 , and σ 28 , whose individual expression patterns [10][11][12] fail explain the stage-specific transcription profiles observed in acute chlamydial development [12]. Assuming these sigma factors participate in distinct functions, then post-expression mechanisms of regulation must theoretically be employed.
Switch-protein kinase modules are common effectors of energy and stress responses in bacteria. One of the best studied is the 'Regulator of SigmaB' or Rsb system in Bacillus subtilis. Within this type of module, a component called the switch-protein can either bind to and affect the function of a target protein, or function as a kinase in the phosphorylation of a network antagonist. Phosphorylation of the antagonist prevents further interaction with the switch-protein. The function of a competing PP2C-like phosphatase controls the level of active antagonist. In the absence of active (i.e. non phosphorylated) antagonist, the switch protein is driven towards its regulatory function. Commonly, the target of the switch-protein is a sigma factor, as is the case for the RsbW protein in B. subtilis (reviewed in [13]). Analogues of a switch-protein kinase regulatory module are conserved in the C. trachomatis genome [14] and were named after the B. subtilis module. Putative components of the Rsb switch-protein regulatory system, outlined in Table 1, include one analogue of the switch-protein kinase (RsbW), two analogues of the module antagonist (RsbV1 and RsbV2), and three proteins that contain recognized PP2C-like phosphatase domains (RsbU, CT589, and CT259) [14]. Henceforth, chlamydial analogues with common family-member protein names will be designated with a 'Ct' subscript, e.g. RsbW Ct . Previous studies have revealed that all potential module members are expressed [12,15] and that RsbW Ct is a kinase specific for RsbV1 and RsbV2-although a 'switch' function regulatory target has not been identified [16,17]. The current study expands on these works, and further outlines the functions of Rsb module members in Chlamydia. In vitro and in vivo evidence indicate that the regulatory target of RsbW Ct is the primary σ-factor for Chlamydia, σ 66 , and that opposing functions of RsbW Ct and the antagonist, RsbV1, contribute to the amount of σ 66 that is available for association with the core RNAP complex.

RsbW associates directly with σ 66
The regulated target of the putative switch-protein kinase analogue in Chlamydia (RsbW Ct ) had not been identified at the onset of this study. Therefore, a bacterial adenylate cyclase two-hybrid (BACTH) system [18] was employed to screen for interactions between RsbW Ct and the three chlamydial sigma factors: σ 28 , σ 54 , and σ 66 (Fig 1A). In order to gauge the relevance of this method for the screening of RsbW-type interactions, homologous proteins of known function from B. subtilis [19,20] were used as positive (σ B + RsbW Bs ) and negative (σ B + RsbT Bs ) control combinations. As expected, interaction of σ B with its cognate anti-sigma factor, RsbW-Bs , complemented adenylate cyclase (AC) activity, and expression of the cAMP-dependent lacZ reporter cassette was detected via the Miller Assay. In contrast, co-expression of σ B with RsbT Bs (a paralogue of RsbW Bs that does not bind to σ B ) did not exhibit complementation. When RsbW Ct was expressed with the chlamydial alternative sigma factors, σ 28 and σ 54 , similar activities to the empty vector controls were observed (t = 0.124 and t = 0.023, respectively; One-way ANOVA, Bonferroni's multiple comparison post test). In contrast, an interaction between RsbW Ct and σ 66 was clear (t = 3.633), as LacZ levels approached those of the positive control (GCN4 leucine zipper domains; 'zip'). Interaction between σ 66 and RsbW Ct was observed in both cloning permutations of the BACTH assay, and expression of σ 66 did not artificially activate the cAMP dependent reporter LacZ cassette in the absence of an RsbW Ct fusion protein (S1 Fig).
To validate the BACTH screen, molecular interactions were monitored in real time by surface plasmon resonance (SPR). Recombinant, purified σ-factors (σ 66 , σ 28 , and σ 54 ) were immobilized to consecutive flowcells of an individual CM5 sensorchip, before the chip was charged with RsbW Ct analyte at various concentrations. The change in response units (RU) over time, relative to a reference flowcell (immobilized Glutathione-S-transferase; GST), served as a function of binding between the ligand (sigma factor) and analyte (RsbW Ct ). Curves of best fit were applied to RU sensorgrams and the rate of equilibrium (R eq ) for each curve was determined (for example, a σ 66 -binding plot overlay is shown in Fig 1B). In order to compare the relative binding of RsbW Ct to the three σ-factors, R eq values were normalized by the theoretical maximum RU (R max ) for each ligand, which was based on the original immobilization levels for each sigma factor. At every concentration tested, RsbW Ct bound a greater percentage of immobilized σ 66 than either of the alternative sigma factors (Fig 1C). This trend was consistent in multiple experiments (n = 3), although the overall percentage of theoretical R max bound appeared to decrease with the age of the immobilized ligand on each chip (from 1 to 3 days), indicating that the activity or conformation of the immobilized ligands decreased over time.
When a GST-antibody capture system was used, in which GST-tagged sigma factors were captured by immobilized anti-GST immunoglobulin immediately prior to addition of RsbW Ct , up to 40% of the σ 66 R max was bound at a concentration of 5 μM (S2 Fig). Thus, these data indicate that RsbW Ct associates with the sigma factor, σ 66 . Interaction of RsbW with RsbV1 or RsbV2 depends on ATP and the phosphorylation state of the antagonist RsbW Ct is a kinase specific for RsbV1 and RsbV2, and the phosphorylated residues of the antagonists have been mapped to Serine-56 and Serine-55, respectively [17]. In order to gain insight into the kinetics of the kinase activity for RsbW Ct against RsbV1 and RsbV2, aliquots from phosphorylation reactions were removed iteratively and resolved on a polyacrylamide gel supplemented with Phos-tag, an agent which causes an electromobility shift of phosphorylated proteins. When incubated in the presence of ATP, RsbW Ct induced a mobility shift of both RsbV1 and RsbV2, although there was a clear distinction in the rates at which the two antagonists were phosphorylated; after only 10 minutes, 100% of RsbV1 migration was shifted, whereas only a fraction of RsbV2 was shifted after 2 hours of incubation with RsbW Ct (Fig 2A  and 2B).
To determine the direct effects of phosphorylation on the interaction between RsbW Ct and antagonist, SPR was again utilized; RsbV1 and RsbV2 antagonists were immobilized onto individual flow-cells of a sensorchip and RsbW Ct was then supplied in the absence or presence of ATP. Without ATP, RsbW Ct did not bind to RsbV1 (Fig 2C) or to RsbV2 ( Fig 2D). When ATP was supplied along with RsbW Ct , both ligands produced abnormal binding curves, in which an initial increase in response gradually returned to the baseline level. Subsequent applications of RsbW Ct plus ATP yielded no response with either ligand. We reasoned that upon addition of ATP, RsbW Ct was able to associate initially with, then phosphorylate RsbV1 and RsbV2, but that phosphorylation prevented their further interaction.
To explore this hypothesis, similar SPR experiments were carried out on derivatives of RsbV1 and RsbV2, in which the phospho-accepting residues were mutated to mimic either an immutable non-phosphorylated state (serine to alanine), or a permanent phosphorylation state (serine to aspartic acid). The addition of RsbW Ct without ATP exhibited no binding response to any of the four antagonist derivatives (Fig 2E and 2F). Supplementation of ATP with RsbW Ct yielded no binding response to the RsbV1 S56D or RsbV2 S55D derivatives, suggesting that a negative charge (from the phosphate group or, in this case, the aspartic acid residue) at position 56/55, respectively, of the antagonist abrogated association with RsbW Ct . Supplementation of ATP with RsbW Ct produced a typical saturation-binding curve to the RsbV1 S56A and, to a lesser extent, RsbV2 S55A antagonist derivatives. The S56A or S55A derivative antagonists gave a similar binding response on subsequent RsbW Ct plus ATP injections, presumably because of their inability to accept phosphorylation and, rendering them inactive. Together these data indicate that the binding of RsbW Ct to its RsbV1 and RsbV2 antagonists depends on both the phosphorylation state of RsbV1 or RsbV2 and ATP.

Phosphorylated RsbV1 is a substrate of the C-terminal PP2C-like domain of RsbU Ct
Three proteins contain recognized PP2C-like phosphatase (IPR001932) domains in C. trachomatis: CT259, RsbU Ct and CT589 ( Table 1). RsbU Ct and CT589 are paralogues, both predicted to transverse the membrane with a hypothetical domain localized in the periplasm and a PP2C-like domain localized in the cytoplasm [14,17]. In contrast, CT259 appears to consist of a single cytosolic domain ( Fig 3A). When aligned with the PP2C-like domains of RsbU Bs and SpoIIE from B. subtilis, the residues involved in two Mn 2+ coordination sites, which are essential for phosphatase activity, are conserved in RsbU Ct , but not in CT589 [17]. CT259 appears to maintain both divalent metal coordination sites, although the first site exhibits conservative aspartic acid to glutamic acid mutations (S3 Fig). To test if the carboxy-terminal PP2C-like domain of RsbU Ct was an active phosphatase, the sequence corresponding to the final 258 amino acids was cloned into an expression vector for recombinant purification. This C-terminal domain of RsbU Ct (referred to as C-RsbU Ct ) was incubated with RsbV1 or RsbV2 that had been previously phosphorylated by RsbW Ct . The phosphorylation states of the antagonists were distinguished by resolution on a Phos-tag supplemented acrylamide gel. Introduction of C-RsbU Ct caused a rapid shift from the phospho-to non-phosphorylated form of RsbV1, however phospho-RsbV2 remained phosphorylated throughout the entire time-course ( Fig 3B). CT259 was also examined for phosphatase activity were incubated in the presence of RsbW Ct + ATP and monitored longitudinally (minutes) for phosphorylation, detected via electromobility shift during Phostag supplemented gel electrophoresis. Each panel represents two experiments in which the same results were obtained. For SPR analysis, antagonists RsbV1 (C) and RsbV2 (D) were immobilized and charged with 2500nM RsbW Ct-in the absence (black) or presence (first attempt, green; repeated attempts, red) of ATP. Response over time was normalized by the response in an empty flow-cell. Sensorgrams shown represent two experiments (separate sensorchips) in which the same trends were observed. Derivative mutants, in which the phospho-accepting residues were changed to aspartic acid (S56D, S55D) or alanine (S56A, S55A) for RsbV1 (E) and RsbV2 (F), respectively, were tested for their ability to associate with RsbW Ct . Serine to aspartic acid mutations, mimicking a permanently phosphorylated state, prevented the association of RsbW Ct in both the absence (blue) and presence (green) of ATP for both antagonists. Alanine derivatives of both antagonists, which cannot be phosphorylated and mimic a permanently active state, were bound in the presence of ATP (black), but not without ATP (red). against the two phospho-antagonists, though no activity was observed ( Fig 3C). Thus, while the C-terminal domain of RsbU Ct appears to maintain phosphatase activity for phospho-RsbV1, no phosphatase capable of recognizing RsbV2 was identified.

Manipulation of Rsb pathway components in vivo
A chlamydial shuttle vector for the controlled expression of a target cassette in Chlamydia was engineered in order to further investigate the role of Rsb components in vivo. Briefly, the shuttle vector, pGFP::SW2 [21], was modified such that the constitutive promoter driving the gfpcat cassette was replaced with the tetracycline inducible promoter system from pRPF185 [22], producing pCT308-GFP (S4A Fig). The gfp cassette was then replaced with sequences corresponding to rsbW or rsbV1 from C. trachomatis serovar D/UW/CX genomic DNA to make pCT1310-RsbW and pCT1310-RsbV1, respectively. These three shuttle vectors transformed C. trachomatis strain L2/25667R (originally lacking any plasmid; henceforth referred to as L2R) and were stable through several passages in penicillin-supplemented medium, prior to plaque purification, expansion, and density gradient harvest. Strains harboring these shuttle vectors were as resistant to penicillin as the original parent shuttle vector, pGFP::SW2 [21] with concentrations of 0.01-0.05 U/ml failing to have any effect on generation of infectious progeny, despite the same concentrations inhibiting the C. trachomatis L2R strain lacking any plasmid (S4B Fig). Endogenous fluorescence was not detected in Chlamydia harboring pCT308-GFP in the absence of ATc, but was detected upon supplementation of the growth medium (S4C Fig).
Additionally, a strain in which rsbV1 was inactivated using insertional mutagenesis [23] was engineered. Briefly, a GII intron carrying the aadA-marker (to confer resistance to spectinomycin) was targeted for insertion into the 5' region of rsbV1, creating DFCT15 (rsbV1::GII[aadA], S5 and S6 Figs). The mutant strain was resistant to spectinomycin and sequencing of the disrupted rsbV1 locus confirmed the GII intron insertion, resulting in alteration of the wild type RsbV1 sequence after 10 amino acids. Serial passage in the absence of spectinomycin followed by PCR analysis of the insertion-site using rsbV1-specific primers confirmed marker stability in the absence of selection, as previously observed for the GII intron carrying the bla-marker [23]. Consequently, experiments were performed without spectinomycin when matched with non-spectinomycin resistant strains.

RsbW Ct and RsbV1 affect transcription from σ 66 -dependent promoters in Chlamydia
Based on the results of in vitro binding assays and the respective functions of homologous proteins in other bacteria, we hypothesized that RsbW Ct was a negative regulator of σ 66 and that non-phosphorylated RsbV1 would act as a positive regulator of σ 66 by antagonizing RsbW Ct . To test this hypothesis in vivo, we monitored the expression of bona fide σ 66 -dependent genes in our L2R transformant and rsbV1 knockout strains. Normalization of transcript expression in Chlamydia is not trivial (e.g. [8,24]), especially considering our hypothesis of differential 'housekeeping' transcription. We tested a number of different normalization techniques that did not assume that any chlamydial gene would correlate unequivocally to the number of chlamydiae present within the sample. While use of C. trachomatis-specific gDNA from parallel samples as an exogenous control has been employed frequently (e.g. [8,25]), this type of normalization fails to account for variation in RNA loading / efficiency of reverse transcription as a source of error. We found that use of the geometric mean of Chlamydia-specific gDNA from parallel samples and endogenous host cell gapdh as a normalization factor for each sample produced a data set with the lowest cumulative intrasample error (S7 Fig). Primary Cq values, along with normalization calculations, are available in S1 Dataset. As expected, rsbW transcript expression was elevated in the L2R pCT1310-RsbW strain (Fig 4A), whereas rsbV1 transcript expression was elevated in the L2R pCT1310-RsbV1 strain ( Fig 4B). Transcript of rsbV1 was not detected by RT-qPCR in DFCT15 (rsbV1::GII), confirming genetic analysis (S6 Fig).
Expression of the gfp-cassette in L2R pCT308-GFP strain reached similar levels of the other expression cassettes (data available in S1 Dataset).
To test whether 'housekeeping' transcription was affected in the transformant / knockout mutant strains, we assessed the transcription of two genes with bona fide σ 66 -dependent promoter systems, 16S rRNA [26,27] and ompA [28,29], limiting our analysis to time-points prior to the typical RB to EB re-differentiation in order to avoid re-differentiation as a confounding variable. As predicted, σ 66 -dependent transcription was elevated in the L2R pCT1310-RsbV1 strain, with the pooled relative expression of ompA and 16S rRNA reaching a geometric mean of 2.32-fold over the control (95% CI = 1.72-to 3.11-fold). Conversely, the relative expression of σ 66 -dependent transcription was reduced to 0.362 of control (95% CI = 0.224 to 0.583) in the rsbV1::GII strain. Notably, σ 66 -dependent transcription in the L2R pCT1310-RsbW strain was reduced from the control expression (95% CI = 0.585 to 0.988), however did not reach similar repression levels as the rsbV1::GII strain. This negative effect has been observed in each of two previous analysis of expression experiments (summary figures shown S8 and S9 Figs). The observation that the rsbV1::GII strain exhibits lower amounts of σ 66 -mediated transcription compared to the RsbW Ct expression strain may indicate that the levels of rsbW achieved in these experiments were insufficient to completely overcome antagonism from endogenous RsbV1 levels, resulting in a subtle but consistent phenotype.
As a control, we also assessed whether σ 28 -dependent transcription of hctB [30], was affected in the transformant / knock-out mutant strains, with the caveat that all known σ 28dependent genes are transcribed during the late stages of the developmental cycle due to repression by an early-stage transcriptional repressor (EUO) [31,32]. Indeed, we were not able to detect hctB transcription in all samples at 12 hpi, and levels at 18 hpi neared the limit of detection. Thus, we analyzed expression of hctB during the late stages of the developmental cycle. At these time-points, no differential expression was observed. Taken together with the in vitro binding data, our results suggest RsbW Ct is an anti-sigma factor of σ 66 , and that RsbV1 is a bona fide antagonist of the RsbW Ct to σ 66 association.

Ectopic expression of RsbW Ct and RsbV1 affects chlamydial replication and development
We hypothesized that modulating the activity of the 'housekeeping' sigma factor would have detectable effects on Chlamydia, yet no obvious effect on growth or development was observed during the passage and selection of pCT1310-series transformants. Immunofluorescent analysis (IFA) revealed that all transformant strains exhibited a morphology consistent with acute development, and that no aberrant, enlarged chlamydial particles consistent with the persistent phenotype were observed. However, we did observe that inclusions from the L2R pCT1310-RsbV1 strain appeared much larger than the GFP or RsbW Ct expression strains (p<0.0001; One-way ANOVA, Tukey's multiple comparison post-test versus both other samples; S10 Fig). We reasoned that increased inclusion size observed in the L2R pCT1310-RsbV1 were monitored in L2R strains harboring shuttle vector plasmids pCT308-GFP, pCT1310-RsbW, or pCT1310-RsbV1, as well as in strain DFCT15, in which rsbV1 had been inactivated via transposon insertion (rsbV1::GII). The expression levels of target cassettes were elevated in their respective expression strains, whereas no rsbV1 transcription was detected in the rsbV1::GII strain. Expression of target cassettes are shown as the mean normalized expression as a function of time (hours post infection). Error bars represent the 95% CI. Normalized expression of 16S rRNA (C), ompA (D), and hctB (E) were calibrated to the geometric mean of the GFP-expression control. Lines indicate the geometric mean of relative expression and error bars represent the 95% CI. Pvalues are derived from One-way ANOVA with Tukey's multiple comparisons post-test performed in R. Transcription of ompA and 16S rRNA are σ 66dependent [26][27][28][29], whereas transcription of hctB is σ 28 -dependent [30]. Only σ 66 -dependent genes experienced differential expression. strain might be attributed to increased metabolic capacity upon elevated 'housekeeping' transcription, and if so, this may correlate to modulations in growth and development. To test this hypothesis, we analyzed the transformant and rsbV1 mutant strains for genomic replication (1-step growth), recoverable infectious progeny (2-step growth), and plaque expansion.
For both 1-step and 2-step growth analysis, data was normalized by the empirical IFU input for each experiment, such that the data presented accounts for the actual number of infection events for each sample. Differential genomic replication was not overtly evident between the strains, though analysis of the area underneath each growth curve did reveal a trend that mirrored levels of σ 66 -dependent transcription, with L2R pCT1310-RsbV1 exhibiting the highest chlamydial load, followed by L2R pCT308-GFP, and then L2R pCT1310-RsbW and the rsbV1 mutant ( Fig 5A). As a second indicator of development, recoverable infectious progeny (2-step growth) was also monitored (Fig 5B). At 24 hpi, the L2R pCT1310-RsbV1 strain exhibited a 2.3-fold increase in infectious progeny compared to control (p = 0.0026; One-way ANOVA with Tukey's multiple comparisons post-test), whereas L2R pCT1310-RsbW exhibited a 3.1-fold reduction (p = 0.181). The rsbV1 mutant also displayed a reduction in infectious progeny that exceeded that of L2 pCT1310-RsbW (p = 0.108 versus control). Thus recoverable infectious progeny and, to a lesser extent, genomic replication exhibited a pattern that correlated with observed σ 66 -dependent transcription.
However, because neither one-step or two-step growth analysis alone is a perfect measure of Chlamydia fitness, we chose to analyze transformant and mutant strains in a plaque expansion assay. Because plaque expansion depends on all aspects of chlamydial development (e.g. replication, re-differentiation, cell exit, and secondary host cell entry), the relative size of plaques can be used as an indicator of overall chlamydial fitness.
Specific locations within HeLa monolayers were inoculated with the modified Chlamydia strains and plaques were measured after 8-9 days, using a semi-automated procedure in FIJI [33]. An example of this process is shown in Fig 5C, in which plaques from day 9 post-inoculation are shown. In concordance with the 1-step and 2-step growth profiles, the L2R pCT1310-RsbV1 strain exhibited the highest rate of plaque expansion (Fig 5D). Moreover, the L2R pCT1310-RsbW and DFCT15 (rsbV1::GII) strains yielded smaller plaques than the GFP expression control. Thus, in concert with in vitro and transcript expression data, these results support a model in which the availability of σ 66 is affected by relative levels of RsbW and RsbV1 and that their experimental manipulation was capable of influencing the rate at which C. trachomatis infection progresses in a cell culture model.

Discussion
The evidence presented in this report indicates that the primary target of the Rsb phosphoregulatory network in Chlamydia trachomatis is σ 66 , the main 'housekeeping' sigma factor of the pathogen. Interaction of RsbW Ct with σ 66 was shown indirectly by bacterial two-hybrid assay and directly by SPR analysis. Moreover, previous results that RsbW Ct is a kinase for both RsbV1 and RsbV2 in vitro [17] were confirmed, although there is a difference in efficiency between these phosphorylation events, with RsbV1 being the preferred substrate. RsbW Ct does not associate with RsbV1 or RsbV2 in their phosphorylated forms, or, interestingly, in the absence of ATP. We further observed in vitro phosphatase activity from the RsbU Ct PP2C-like domain that is specific for phospho-RsbV1, but not for phospho-RsbV2. Thus, a complete signaling module (consisting of a system phosphatase, antagonist, switch-protein, and target) has been characterized in this report. To verify the model generated from in vitro assays, mutant and transformant strains of Chlamydia were engineered for in vivo analysis. Elevated expression of RsbV1 correlated with the enhanced expression of bona fide σ 66 -dependent transcripts and a more rapid growth profile in multiple assays. In contrast, elevated expression of RsbW Ct and the inactivation of its antagonist both resulted in reduced transcription of representative σ 66 -dependent genes and a depressed growth profile.
Taken together, these results provide a mechanism by which the Rsb network could control σ 66 availability, and perhaps growth rate, in response to various stimuli. We postulate the following working model (Fig 6). Under steady-state conditions, the equilibrium of the Rsb network provides a molecular 'speed-limit' on σ 66 activity and subsequently on metabolic activity The effect of rsbW and rsbV1 expression on genomic replication and development of infectious progeny. 1-step and 2-step growth from timecourse experiments were monitored for strain L2R harboring shuttle vector plasmids pCT308-GFP (green), pCT1310-RsbW (red), pCT1310-RsbV1 (blue), or from strain DFCT15 (rsbV1::GII; maroon). The mean of C. trachomatis gDNA, normalized by the empirical IFU input (x10 5 ) is plotted by hours post-infection (A). The inset shows the area under each curve. The mean of recoverable infectious progeny from the same experiments was also normalized by empirical input (B). Error bars for both panels represent the 95% confidence interval. P-values derived from One-way ANOVA with Tukey's multiple comparisons test performed in R. Plaque expansion from infection foci were measured using FIJI freeware (C). Bars represent mean plaque size (mm 2 ) from all foci at days 8 and 9 post-infection (D). Error bars represent 95% confidence intervals. P-values derived from One-way ANOVA with Dunn's test for multiple comparisons using rank sums in R.
doi:10.1371/journal.ppat.1005125.g005 ( Fig 6A). Upon experiencing increased levels of active (i.e. non phosphorylated) RsbV1, the equilibrium of RsbW Ct function would be driven away from sequestration of σ 66 , resulting in amplified levels of 'housekeeping' transcription. Possible inputs for this pathway would be the increased activity/expression of RsbU Ct, or increased expression of RsbV1 ( Fig 6B). Alternatively, upon decreased expression or activity of RsbV1, the equilibrium of RsbW Ct would be driven towards sequestration of σ 66 , limiting 'housekeeping' transcription and possibly restricting metabolism. Potential inputs for this shift would include decreased expression or activity of RsbU Ct , decreased expression of RsbV1 (Fig 6C), or, intriguingly, decreased levels of ATP. A mechanism that links energy stress to decreased general levels of transcription seems plausible, however a procedure for the manipulation or even measurement of Chlamydia ATP levels (i.e. to distinguish host versus pathogen ATP pools) has not been elucidated, making exploration of this hypothesis difficult.
We were initially surprised to observe that the association between RsbW Ct and RsbV1 is dependent on ATP, which represents a mechanism distinct from Rsb components in the in B. subtilis module [34]. However, this distinction may reflect the differences in regulated targets with σ B as an alternative sigma factor responsible for activating transcriptional response to stress, and σ 66 as the primary sigma factor for Chlamydia. Under low ATP conditions in B. subtilis, RsbW Bs is not able to inactivate RsbV Bs such that they associate stably, liberating σ B for the initiation of the stress response [34]. However, under low ATP conditions in Chlamydia, the theoretical association between anti-sigma factor and antagonist would not occur, driving the function of RsbW Ct towards sequestration of σ 66 and the reduction in 'housekeeping' transcription. We postulate that the ATP-dependence of anti-sigma to antagonist interaction Under steady-state conditions, the expression levels and activies of RsbU Ct (purple), RsbV1 (black), and RsbW Ct (red) provide an equilibrium in which σ 66 availability is ample for normal growth and development (A). Upon an accumulation of active (i.e. non phosphorylated) RsbV1 via increased expression or increased activity/expression of RsbU Ct , the equilibrium of RsbW Ct function shifts away from the anti-sigma factor role, allowing increased transcription of σ 66 -promoted genes (B). Upon the decreased availability of active RsbV1 (either through reduced expression or reduced RsbU Ct expression/activity), the equilibrium of RsbW Ct function is driven towards the anti-sigma factor function, limiting the amount of σ 66 that is available for association with the core RNAP complex (C). theoretically must switch the Rsb system from one of stochastic activation in B. subtilis [35] to one of stochastic inactivation in Chlamydia.
This study extends the work of Hua et al [17], which characterized several interactions within this module, including kinase activity of RsbW Ct and detailed in silico analysis of the module members. Yet, they failed to observe any interaction between RsbW Ct and any of the chlamydial σ-factors in a yeast-two hybrid system. Furthermore, in vitro transcription assays showed no interaction between RsbW Ct and σ 28 , and a separate study failed to observe any effect of RsbW Ct in a heterologous σ 28 -mediated transcription assay in Salmonella enterica [16,17]. While we concur that σ 28 does not interact with RsbW Ct , we provide evidence that RsbW Ct indeed interacts with σ 66 . One possible explanation for this discrepancy could be that the σ 66 -RsbW complex was not targeted properly to the yeast nucleus due to its size, stoichiometric ratio, etc. By using a bacterial two hybrid reliant on generation of a cytosolic metabolite, neither complex size nor nuclear import was an issue.
Interestingly, other proteins have been described as modulators of σ 66 function in Chlamydia. For instance, CT663 has been likened to the Rsd protein in E. coli [36], which inhibits σ 70 -mediated transcription during stationary phase growth [37]. Another protein, GrgA, reported as a non-specific DNA binding protein, also associates with the non-conserved region of σ 66 . By interacting with both simultaneously, GrgA enhances the transcription of σ 66 -dependent promoter systems in vitro [38]. Thus, there is a clear precedent for the alteration of σ 66 activity as an evolved strategy in Chlamydia. We add RsbW Ct to this list of σ 66 modulators characterized in vitro, and additionally provide in vivo evidence to support its function in this role.
We would like to note that we also attempted to transform C. trachomatis with shuttle vectors capable of expressing RsbV1_S56A, RsbV2, RsbV2_S55A, RsbU, CT589, and CT259, of which none transformed Chlamydia in at least 2 attempts (in which a positive control for transformation was employed and successful). We also attempted insertional inactivation of rsbV2, rsbU, and ct589, but were unsuccessful. As transformation methodologies are relatively new for Chlamydia, we are unsure whether these observations can be attributed to effects on chlamydial fitness (e.g. such manipulations are lethal or at least highly detrimental) or are simply technical failures. The use of a positive control vector for each transformation, as well as the use of multiple plasmid backbones (pCT1310-and p2TK-SW2 [39]) for attempted transformations suggest the former may be more likely. Interestingly, of the genes targeted for manipulation, those that were successfully altered exhibited mild, but reproducible, phenotypic changes. Thus, it is plausible that any robust alteration in the function of Rsb module (e.g. the overexpression of constitutively active RsbV1 S56A) could be highly detrimental to chlamydial growth under selective cell culture conditions. Further insights into the process of chlamydial transformation should shed light onto the nature of these negative results.
What, then, is the evolutionary function of the Rsb system in Chlamydia? To our knowledge, there is no phenotype (occurring during acute or persistent modes of growth) that demonstrates a global modulation in σ 66 -mediated transcription. Perhaps this is not surprising considering that beyond basic expression and availability of sigma factors, other global transcriptional repressor proteins are likely to be influential at all stages of Chlamydia development. For instance, the protein EUO controls transcription of both σ 66 -and σ 28 -dependent late genes via interaction with operating elements found in both promoter types during the early stages of infection [31]. Thus, even if the Rsb network dictated an increased availability of σ 66 , transcription of σ 66 -dependent late genes would be occluded by the presence of the EUO DNA-binding protein. While it is possible the Rsb module works in concert with other global regulators in order to actualize RB to EB re-differentiation, we favor a model where its function provides a molecular 'speed-limit' for housekeeping transcription (and subsequently metabolism) as a way to avoid overuse of potentially limited nutrients. Manipulation of expression levels of RsbW Ct and RsbV1 via ectopic expression or insertional inactivation resulted in differential growth and development in a nutrient-replete cell culture model. The elucidation of the activities of the system components in response to inimical growth conditions could provide more resolution into the question of the evolutionary role of this system in Chlamydia trachomatis.
In conclusion, this report provides multiple lines of evidence that indicate the Rsb module is a bona fide molecular circuit capable of influencing the availability of the main sigma factor in Chlamydia. The potential of this network to accelerate and restrict growth rate and development substantiates it as an important pathway for further study, and may even constitute a novel target for generation of attenuated, or even accelerated, growth mutants.

Strains and parent plasmids
Chlamydia trachomatis serovar L2/25667R, pGFP::SW2 shuttle vector, and the BACTH vectors and DHM1 E. coli were provided by Dr. Scot Ouellette (University of South Dakota, Vermillion). The pRFP185 plasmid was received from Dr. Robert Fagan (University of Sheffield, Sheffield). Bacillus subtilis subspecies subtilis strain 168 was a received from Professor Neil Fairweather (Imperial College, London). Chlamydia trachomatis serovar D/UW/Cx was provided by Dr. Rey Carabeo (University of Aberdeen, Scotland).

Cloning
All PCR reactions intended for cloning purposes were performed with Phusion High-Fidelity Polymerase (NEB). Either C. trachomatis genomic DNA (Serovar D/UW/CX) or B. subtilis sbsp. subtilis 168 genomic DNA was used for template in reactions. Primer sequences are listed in S1 Table. Insert and vector ligations were either performed using traditional restriction endonuclease digest (NEB) and ligation (T4 Ligase; NEB), or in a one-step Gibson Assembly cloning reaction (NEB). Plasmids intended for recombinant protein expression are listed in S2 Table, and plasmid shuttle vectors for the transformation of C. trachomatis are listed in S3 Table. Site directed mutagenesis of pGEX expression vectors for the antagonist proteins was also carried out with PCR based Gibson Assembly. All plasmid insert sequences were verified by Sanger sequencing (GATC; Germany or Macrogen, USA).

Bacterial two hybrid assay
The bacterial adenylate cyclase two-hybrid method for assessment of protein-protein interaction was completed as per the instructions of the manufacturer (EuroMedex). Detailed methods are described in S1 Text. The Miller Assay determined expression of the reporter LacZ [40]. Sample groups in all graphs represent an equal number of biological replicates, indicated in figure legends.

Surface plasmon resonance
The Biacore 3000 instrument (GE Healthcare) revealed biomolecular interactions via surface plasmon resonance (SPR). Specifics for each run are described in supplemental information. CM5 sensorchips were used for all experiments. The optimal pH for pre-concentration of ligands was determined using the pH scouting wizard (Biacore 3000 software), and ligands were using the amine coupling kit. All experiments were carried out in a buffer of 10mM HEPES, 150mM NaCl, and 1mM MgCl 2 at 30°C. All consumables were purchased from GE Healthcare.

Protein purification
Proteins were expressed from pGEX-derived vectors in PC2 E. coli [41]. Soluble proteins were purified by standard techniques. Insoluble proteins were liberated from inclusion bodies using sarkosyl as described in the S1 Text. All proteins were assessed for purity via SDS-PAGE and Coomassie-Blue stain. Detailed protocols are available in S1 Text.

Bioinformatics
All DNA and protein sequence diagrams were generated using Geneious version 7.0.2, created by Biomatters. Secondary structures were predicted by the Phyre2 algorithm [42]. Transmembrane regions were predicted using the Hidden Markov Model in Geneious. InterPro domains were identified using the European Bioinformatics Database (EBI). Accession numbers listed are from the UniProtKB database.

Kinase and phosphatase assays
In vitro kinase/phosphatase assays were performed using recombinant, purified preparations of the system components and are described in detail in the S1 Text. Phos-tag reagent (Alpha Laboratories) was utilized to shift the migration of phosphorylated RsbV1 or RsbV2 during SDS-PAGE.

Chlamydia trachomatis transformation and insertional mutategenesis
Transformation of C. trachomatis L2/25667R (plasmid-deficient) with ectopic expression vectors was carried out, as described [21,43] with slight modifications highlighted in the S1 Text. Every transformation attempt utilized pGFP::SW2 as a positive control (100% success rate). DFCT15 was generated from the transformation of C. trachomatis strain L2/434/Bu with plasmid pDFTT6aadA as described in [23] with the exception that spectinomycin was used for selection at 500 ug/ml instead of ampicillin for mutant selection and plaque isolation of clones. The intron was targeted to insert between base pairs 28 and 29 of rsbV1 (TCCCTTGTAAAT GAAGGATGCCTGTTTGGC-intron-CTTGTTCTTCTTTCT) in an antisense orientation. The predicted insertion efficiency values were an E-value of 0.75 and a score of 8.51. Intron retargeting was performed as directed by the TargeTron manual (Sigma-Aldrich). Both transformant and knock-out strains were considered plasmid-competent.

Chlamydia infections
HeLa cells (ATCC) at 80-95% confluence in a 6-or 12-well cluster plate were inoculated with C. trachomatis diluted in Hank's Balanced Salt Solution prior to centrifugation (500xg for 15 minutes at 20°C) and incubation at 37°C for 30 minutes. The inoculum was then aspirated before addition DMEM supplemented with 10% fetal bovine serum (FBS) and other supplements as noted. For cassette induction, anhydrous tetracycline (Sigma) was supplemented to a final concentration of 5 ng/ml at 6 hours post infection.

1-step growth analysis
C. trachomatis strains were grown on the same cluster plate for 6-30 hpi. At designated time points, infected monolayers were washed, trypsin-treated into suspension, pelleted and stored in PBS at -80°C. Total genomic DNA was extracted and diluted to a final concentration of 1 ng/ml. Chlamydia-specific gDNA was assayed twice using qPCR (see Quantitative PCR below) with primers amplifying a region of the hypothetical gene, ct652.1, or the 16S ribosomal subunit gene. Starting quantities were normalized against the empirical IFU input for each strain in order to account for any differences the actual versus intended MOI within each sample.

2-step growth analysis
Samples intended for quantification of recoverable infectious progeny were dislodged into 1ml SPG buffer (220mM Sucrose, 10mM Na 2 HPO 4, 4mM KH 2 PO 4 , 5mM Glutamic Acid) using sterile glass beads, and stored at -80°C. Sample infectious titers were quantified as described previously [25]. A similar process was used to determine the empirical infectivity of primary sample infections. Titers were normalized by this empirical infection rate for each strain within the individual experiment to exclude any effects of variation in stock aliquots.
Plaque expansion assay. Uninfected HeLa cells (~85-90% confluence) in 6-well cluster plates were covered with 1X DMEM containing 0.8% agarose, which was kept at 42°C until use). Agarose medium was then allowed to congeal at room temperature for~15 minutes. Strains were diluted to a concentration of 10 7 IFU/ml. A bevelled 20μl pipette tip was briefly incubated in the inoculum and then used to puncture the gel overlay at specific sites. After inoculation, gels were overlaid with 3ml of DMEM supplemented with 15% FBS, 1μg/ml cycloheximide, 5ng/ml ATc, and 1 μg/ml gentamicin. All inoculation sites were checked for infection on day 1. Liquid overlay media were changed every 3 days. For plaque area measurement, overlay medium was carefully aspirated and the monolayers fixed by adding 10% formaldehyde in PBS to each well. After~2 hours, gels were carefully removed and the fixed monolayers washed with PBS twice. Each well then received 0.5 ml of a 1% crystal violet in 20% ethanol solution. Monolayers were stained for~30 minutes, prior to gentle washing of the plate in a large reservoir of ddH 2 O. Images of plaques were captured from beneath the well using a Gel-Doc-It Imaging system (Bioimaging Systems) using the exact same zoom, aparature, and focus settings for each plate. Plaques were identified using FIJI [33], by adjusting the "Threshold" and then "Analyze Particles". For infection foci where no plaque was observed, a measurement of 1 pixel was recorded. Image in Fig

Immunofluorescent analysis
Fluorescent imaging for empirical infectivity and for recoverable infectious progeny titer was carried out on a Nikon Eclipse TE2000 epifluoresence microscope. For morphology and inclusion size analysis, images were captured using a Leica SP5 Resonant inverted confocal microscope, using identical settings for each sample. Montage images and scale bars were generated in FIJI. Inclusion sizes were determined in FIJI, using elipses to circumscribe each inclusion within a field of view from which the measurement tool provided the inclusion area (pixels).

Nucleic acid extractions
Genomic DNA extractions were prepared using the DNeasy Blood and Tissue Kit (QIAGEN). For RNA extraction, RNALater fixative was removed from infected monolayers and RNA was extracted by Trizol reagent (Life Technologies). RNA samples were treated with DNase (Turbo DNA-free kit, Applied Biosystems) for 1 hour, before DNase inactivation via the instructions of the manufacturer. cDNA was then generated from 2.5 μg of RNA sample using the Superscript III Reverse Transcriptase kit (Invitrogen), and diluted 1:8 in nuclease-free water before storage at -20°C. No Reverse Transcriptase controls were generated for all samples.

Quantitative PCR
Quantitative PCR was conducted in triplicate on the BioRad CFX96 Real-time system. Each reaction consisted of 1X Power SYBR green Mastermix, 0.45μM of each primer, and 2μl of sample in a volume of 25μl. Each run for a Chlamydia gene contained a standard curve of L2/ 25667R gDNA to assess amplification efficiency. NRT controls were performed for all cDNA samples. Equal amounts of total nucleic acid were loaded for each assay type (2 ng for gDNA, and the cDNA generated from 37.5 ng of RNA).

Transcript expression analysis
To control for both the number of C. trachomatis particles and for differences in reverse transcription efficiency, target gene expression was normalized by the geometric mean of an exogenous (C. trachomatis specific gDNA) and endogenous (host gapdh transcript) controls. Normalized expression was calculated as in [44], using the formula: E Target

Data analysis, statistics, and graphs
Graphs were generated using GraphPad Prism v5.0f. Statistical analysis of BACTH and SPR data were also completed with GraphPad Prism software. The data sets generated for gene expression, plaque expansion size, 1-step growth, and 2-step growth were analyzed and statistical analysis performed using R. R analysis scripts are deposited available at doi: 10 A) Plasmid pCT308-GFP was generated by replacing the promoter driving the gfp-cat cassette from pGFP::SW2 with the tetracycline inducible promoter system from pRPF185. pCT1310-RsbW and pCT1310-RsbV1 were made by exchanging the gfp-cat cassette with the genetic sequence of corresponding genes out of C. trachomatis gDNA Serovar D/UW/ Cx. B) Penicillin inhibitory concentration curves were generated for stable, plaque purified transformant strains. Infections were incubated in media containing the indicated final concentration of Penicillin for 42 hours, prior to cell disruption and quantification of recoverable infectious progeny. All three transformant strains were as resistant to penicillin as the positive control pGFP::SW2. C) The pCT308-GFP strain was examined for fluorescence in non-supplemented and supplemented media (0, 1, 10, or 100 ng/ml ATc). Samples were (mock-) induced at 6 hours post infection and fixed at 28 hours post infection. Chlamydia was immunolabeled using convalescent human sera and a DyLight 594-conjugated secondary antibody. Images were captured by confocal microscope as described in the methods section. We analyzed the 'intrasample' standard deviation from biological replicates (i. e. same strain, time-point harvest, and target gene-ompA, 16S rRNA, and hctB only) upon various normalization strategies. We reasoned that the best normalization strategy would result in the lowest mean of intrasample standard deviation. Standard deviations were converted to % of mean for ease of comparison. Boxplots of the intrasample standard deviation are shown for the primary data (N0) and after various strategies (N05, gDNA of parallel within the subset; N06, mean gDNA of all parallel subsets; N07, gapdh; N12, geometric mean N07 and N05; N13, geometric mean of N07 and N06). The lowest cumulative intra sample standard deviation occurred when expression data was normalized by the host housekeeping reference gene, gapdh (N07). However this strategy failed to account for the number of chlamydiae present within each sample. Therefore we chose to normalize by strategy N12, which is the geometric combination of host gapdh expression and the C. trachomatis gDNA measured from a parallel sample. (TIFF)