A phosphorelay system controls stalk biogenesis during cell cycle progression in Caulobacter crescentus


  • Emanuele G. Biondi,

    1. Bauer Center for Genomics Research, Harvard University, 7 Divinity Avenue, Cambridge, MA 02138, USA.
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Jeffrey M. Skerker,

    1. Bauer Center for Genomics Research, Harvard University, 7 Divinity Avenue, Cambridge, MA 02138, USA.
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Muhammad Arif,

    1. Bauer Center for Genomics Research, Harvard University, 7 Divinity Avenue, Cambridge, MA 02138, USA.
    Search for more papers by this author
  • Melanie S. Prasol,

    1. Bauer Center for Genomics Research, Harvard University, 7 Divinity Avenue, Cambridge, MA 02138, USA.
    Search for more papers by this author
  • Barrett S. Perchuk,

    1. Bauer Center for Genomics Research, Harvard University, 7 Divinity Avenue, Cambridge, MA 02138, USA.
    Search for more papers by this author
  • Michael T. Laub

    Corresponding author
    1. Bauer Center for Genomics Research, Harvard University, 7 Divinity Avenue, Cambridge, MA 02138, USA.
      E-mail laub@cgr.harvard.edu; Tel. (+1) 617 384 9647; Fax (+1) 617 495 2196.
    Search for more papers by this author

E-mail laub@cgr.harvard.edu; Tel. (+1) 617 384 9647; Fax (+1) 617 495 2196.


A fundamental question in developmental biology is how morphogenesis is coordinated with cell cycle progression. In Caulobacter crescentus, each cell cycle produces morphologically distinct daughter cells, a stalked cell and a flagellated swarmer cell. Construction of both the flagellum and stalk requires the alternative sigma factor RpoN (σ54). Here we report that a σ54-dependent activator, TacA, is required for cell cycle regulated stalk biogenesis by collaborating with RpoN to activate gene expression. We have also identified the first histidine phosphotransferase in C. crescentus, ShpA, and show that it too is required for stalk biogenesis. Using a systematic biochemical technique called phosphotransfer profiling we have identified a multicomponent phosphorelay which leads from the hybrid histidine kinase ShkA to ShpA and finally to TacA. This pathway functions in vivo to phosphorylate and hence, activate TacA. Finally, whole genome microarrays were used to identify candidate members of the TacA regulon, and we show that at least one target gene, staR, regulates stalk length. This is the first example of a general method for identifying the connectivity of a phosphorelay and can be applied to any organism with two-component signal transduction systems.


In many organisms, programs of morphogenesis and development are temporally coupled to cell cycle progression, yet the molecular basis of this coupling remains poorly understood, particularly in bacteria. Caulobacter crescentus has emerged as a tractable model system for studying the regulation of these fundamental processes as cell cycle progression in this organism is accompanied by a series of morphological transitions culminating in the production of two distinct daughter cells, a swarmer cell and a stalked cell (Ohta et al., 2000; Ryan and Shapiro, 2003; Skerker and Laub, 2004). Motile swarmer cells possess a single polar flagellum and polar pili, and they cannot initiate DNA replication. Swarmer cells differentiate into stalked cells by shedding their polar flagellum, retracting the pili, and synthesizing a stalk at the former site of the flagellum. This motile-to-sessile transition accompanies a G1-S transition and the initiation of a single round of DNA replication. During the ensuing S-phase and a brief G2 phase, the predivisional cell continues to extend the polar stalk, while also building a new polar flagellum and pili opposite the stalked pole. An asymmetric cell division then ensues, producing a new swarmer and new stalked cell (see also Fig. 7).

Figure 7.

Model of the genetic circuitry regulating polar morphogenesis in C. crescentus.
A. Transcriptional regulation of stalk and flagellar genes. Phosphorylated CtrA activates transcription of rpoN54) and tacA. σ54 and FlbD activate transcription of flagellar genes whereas σ54 and phosphorylated TacA activate transcription of stalk biogenesis genes including staR. A phosphorelay, composed of ShkA (CC0138) and ShpA (CC1114), phosphorylates TacA.
B. Cell cycle-regulated transcription of genes involved in regulation of stalk and flagellum biogenesis (see text for details).

The polar stalk of Caulobacter has been suggested to serve a number of important functions. The adhesive polysaccharide substance called holdfast at the distal tip of the stalk allows cells to bind to abiotic surfaces or to other cells (Poindexter, 1964; Kurtz and Smit, 1994; Smith et al., 2003). As Caulobacter is an aquatic, aerobic bacterium, the stalk has also been suggested to affect buoyant density and help cells stay associated with the air–water interface (Poindexter, 1978). Finally, the stalk may play a significant role in nutrient acquisition. Cells starved for phosphate extend their stalks up to 10 or 20 times the length of the cell body, possibly as a mechanism to increase surface area and nutrient uptake ability (Poindexter, 1964; 1984; Gonin et al., 2000).

Little is known about the regulation of stalk biogenesis. However, many other regulatory processes central to cell cycle progression and morphogenesis in Caulobacter are carried out by two-component signal transduction systems (Ohta et al., 2000; Skerker and Laub, 2004). These signalling systems, comprised of histidine kinases and response regulators, are widespread in the prokaryotic kingdom, as well as being found in plants, yeast and slime moulds (Hoch and Silhavy, 1995). The Caulobacter genome encodes 62 histidine kinases and 44 response regulators (Nierman et al., 2001; Skerker et al., 2005). The canonical view of these signalling systems involves a sensor histidine kinase which autophosphorylates on a conserved histidine residue and then transfers the phosphoryl group to an aspartate residue on the receiver domain of a response regulator (Fig. 1A). Phosphorylation of the receiver domain activates the output domain, which often has the ability to bind DNA, thereby coupling signalling pathways to changes in gene expression and the execution of various cellular events.

Figure 1.

Identification of a putative histidine phosphotransferase in C. crescentus.
A. Schematic of a canonical two-component signal transduction pathway versus a multicomponent phosphorelay.
B. CC1114 amino acid sequence was aligned against the phosphotransferase domain of ArcB (gi48994928, fragment 655–778 aa), the P1 domain of CheA (gi16129840, fragment 1–140 aa), LuxU (gi39721593) and Ypd1 (gi6319966), using 3D-Coffee. Alpha helices are indicated (grey bars) according to the secondary structure of ArcB.
C. Homology model of CC1114 using ArcB as template.

Complicating this simple paradigm for two-component signalling are hybrid histidine kinases in which a histidine kinase contains a response regulator receiver domain at its C-terminal end. These hybrid histidine kinases often participate in more complex pathways called phosphorelays (Fig. 1A). In such pathways, hybrid kinases first autophosphorylate on a conserved histidine residue in the kinase domain and then transfer the phosphoryl group intramolecularly to a conserved aspartate in its receiver domain. The phosphoryl group is then transferred to another conserved histidine on a protein called a histidine phosphotransferase, which can in turn transfer its phosphoryl group to an aspartate on a diffusible response regulator. Unlike histidine kinases and response regulators, histidine phosphotransferases show little sequence identity to one another, with conservation of only a few critical residues and a four-helix bundle structure (Kato et al., 1997; Song et al., 1999; Xu and West, 1999).

Previous genetic screens as well as a systematic deletion of each of the 106 two-component genes in Caulobacter have identified 39 genes which are required for normal cell cycle progression or morphogenesis (Skerker et al., 2005). Of these, 19 were impaired in motility and nine had defects in stalk biogenesis. Many of those involved in motility had been previously identified as regulators of flagellar biogenesis, the best understood aspect of morphogenesis in Caulobacter. The assembly of the polar flagellum in predivisional cells is controlled by a four-tiered transcriptional cascade (Ramakrishnan et al., 1994; Gober and England, 2000). This cascade is set in motion by the DNA-binding response regulator CtrA (Quon et al., 1996). The phosphorylation and activity of CtrA are under extensive cell cycle control, coupling flagellar biogenesis to cell cycle progression (Quon et al., 1996; Domian et al., 1997). CtrA directly regulates the expression of at least 95 genes, including rpoN, which encodes the σ54 subunit of RNA polymerase (Laub et al., 2002). rpoN mutants are non-motile and stalkless, suggesting that σ54 regulates genes involved in flagellum and stalk biogenesis (Brun and Shapiro, 1992).

σ54-dependent gene expression typically requires a phosphorylated transcriptional activator of the NtrC-family of response regulators (Kustu et al., 1989). In Caulobacter, flbD encodes a σ54-dependent activator which is induced at approximately the same time as rpoN, and together with RpoN, activates expression of genes required late in flagellar assembly (Ramakrishnan et al., 1991). FlbD also represses genes required early in flagellar assembly, and thus coordinates the timing of multiple steps in flagellum assembly (Wingrove et al., 1993; Benson et al., 1994; Wingrove and Gober, 1994). However, unlike rpoN, flbD mutants are not impaired in stalk synthesis, suggesting that another NtrC-like response regulator functions with rpoN to regulate stalk biogenesis.

In a systematic deletion of two-component signalling genes the NtrC-like regulator tacA was found to lack stalks (Skerker et al., 2005), suggesting it regulates stalk biogenesis. In that study and previous genetic screens, deleting the histidine kinase pleC was also found to produce stalkless cells (Sommer and Newton, 1989; Skerker et al., 2005). PleC, however, exclusively phosphorylates the response regulators DivK and PleD, and not TacA (Skerker et al., 2005). In the systematic deletion study no other deletions of canonical histidine kinases produced a stalkless phenotype. However, the hybrid kinase CC0138 was found to be stalkless, suggesting that a phosphorelay controls phosphorylation of the σ54-activator TacA. Here, using genetics and a systematic biochemical method called phosphotransfer profiling, we describe the identification of ShpA, the first histidine phosphotransferase in Caulobacter, and demonstrate that it is central to a phosphorelay which regulates stalk biogenesis by controlling the activity of TacA. Using whole-genome DNA microarrays, we identify the likely target genes of TacA. Our results provide evidence for a transcriptional cascade that regulates stalk biogenesis and suggest a mechanism for the cell cycle timing of stalk synthesis.


Identification of C. crescentus histidine phosphotransferases

A systematic study of all C. crescentus two-component signalling genes identified two genes whose deletion produced stalkless cells, the gene CC0138 which encodes a hybrid histidine kinase and CC3315 which encodes a σ54-dependent response regulator named TacA (Skerker et al., 2005). The discovery of these mutants suggested that a phosphorelay may control stalk formation in C. crescentus. However, phosphorelays require histidine phosphotransferases to shuttle phosphoryl groups from hybrid kinases to response regulators (Fig. 1A), and none were identified in the initial annotation of the C. crescentus genome based on sequence homology methods (Nierman et al., 2001). To identify putative phosphotransferases in the C. crescentus genome, we searched for genes predicted to encode proteins bearing other common features. Specifically, we sought proteins with (i) fewer than 250 amino acids, (ii) greater than 70% predicted alpha helical secondary structure, and (iii) the conserved motif HXXKG within a predicted alpha helix. Two genes satisfied these criteria, CC1114 and CC1220. This report focuses on the gene CC1114 which is predicted to encode a protein of 112 amino acids with high alpha helical content (Fig. 1B). The primary sequence of CC1114 was successfully threaded onto the known four-helix bundle structures of three monomeric histidine phosphotransferases: Ypd1 from Saccharomyces cerevisiae (Xu and West, 1999), LuxU from Vibrio harveyi (Ulrich et al., 2005), and the phosphotransferase domain of ArcB from Escherichia coli (Kato et al., 1997) (Fig. 1C). These threaded structural alignments further predicted that His-56 of CC1114 lies within the middle of an alpha helix and is solvent exposed. Homologues of CC1114 are found in the genomes of closely related alpha-proteobacteria (Fig. S1). In sum, these bioinformatic analyses suggested that CC1114 may encode a histidine phosphotransferase.

CC1114 has histidine phosphotransferase activity in vitro

To determine whether the CC1114 protein exhibits phosphotransferase activity, we tagged CC1114 with a His6-tag and purified the protein by affinity chromatography to greater than 90% homogeneity. To determine whether CC1114 can receive a phosphoryl group from a hybrid histidine kinase, we incubated CC1114 with the purified kinase and receiver domains from each of four hybrid kinases which possess high in vitro activity: CC0138, CckA, CC3102 and CC3219. We used separate kinase and receiver domains for each hybrid kinase because full-length constructs had little to no activity (data not shown). Reactions were initiated by the addition of [γ-32P]-ATP, incubated for 30 min, and then resolved by SDS-PAGE (Fig. 2A). In each reaction, radiolabel was incorporated into the kinase domain by autophosphorylation and subsequently transferred to the corresponding receiver domain. For the reactions with CC0138 and CC3102, significant accumulation of label was also found at a position corresponding to CC1114. Incubation of CC1114 with the CC0138 kinase domain in the absence of the CC0138 receiver domain was indistinguishable from incubating the CC0138 kinase domain alone (Fig. 2B, lane 1 vs. 4) indicating that (i) the receiver domain is required for phosphotransfer from the kinase domain to CC1114, and (ii) CC1114 does not have autophosphorylation activity. Together these observations support the conclusion that CC1114 is a histidine phosphotransferase.

Figure 2.

Biochemical identification of the TacA phosphorelay.
A. CC1114 can be phosphorylated by Caulobacter hybrid kinases. The kinase (HY-HK) and receiver domains (HY-RD) for four different hybrid kinases were each purified and tested in vitro for their ability to phosphorylate CC1114. The gene name or CC number is indicated for each kinase.
B. Phosphorylation of CC1114 is dependent on the presence of a receiver domain. Pluses and minuses indicate presence or absence of reaction components: purified kinase domain of CC0138 (CC0138-HK), its receiver domain (CC0138-RD), and CC1114.
C and D. Identification of a CC1114-dependent phosphorelay using phosphotransfer profiling. Phosphorylated CC1114 was profiled against all 44 purified Caulobacter response regulators (C) and against all 27 purified receiver domains of Caulobacter hybrid kinases (D). Each phosphotransfer reaction was incubated for 2 min at room temperature.
E. CC1114 exhibits a kinetic preference for CC0138 and TacA-RD. Phosphotransfer from CC1114 to the receiver domains of CC0138, CC0921, CckA, CC3102, CC3219 and TacA were tested at a 10 s time point. Open arrowheads in panels C–E indicate lanes with evidence of phosphotransfer, manifested as loss of radiolabel from the CC1114 band and/or incorporation of label on a receiver domain.
F. Schematic of the phosphorelay between the hybrid kinase CC0138, the histidine phosphotransferase CC1114 and the response regulator TacA.
G. Reconstitution of the phosphorelay in vitro using purified components. Pluses and minuses indicate the presence or absence of purified CC0138-HK, CC0138-RD, CC1114 and TacA-RD.

CC1114 participates in a phosphorelay culminating in the phosphorylation of TacA

We previously developed an assay called phosphotransfer profiling which allows the rapid, systematic identification of response regulator targets of a given histidine kinase (Skerker et al., 2005). In this assay, one histidine kinase is systematically tested in vitro for its ability to transfer a phosphoryl group to each purified response regulator encoded in the genome. As histidine kinases exhibit a large kinetic preference in vitro for their in vivo cognate substrates, kinase-regulator pairs are easily identified by testing phosphotransfer at multiple time points and identifying the kinetically preferred substrate(s) (Skerker et al., 2005). Here, we adapted this assay (also see Fig. S2) to examine the substrate specificity, and hence probable in vivo partners, of CC1114. CC1114 was first phosphorylated (producing CC1114∼P) and then split and incubated individually for 2 min with each of the 44 purified, full-length response regulators encoded in the C. crescentus genome (Skerker et al., 2005). Phosphorimaging demonstrated a single preferred target of CC1114∼P, the response regulator TacA (Fig. 2C).

Next, we wished to identify the hybrid kinase which phosphorylates CC1114. In vivo hybrid histidine kinases can pass a phosphoryl group to a phosphotransferase or vice versa, with the directionality typically driven by mass-action (Uhl and Miller, 1996; Georgellis et al., 1998). We took advantage of this reversibility to identify the hybrid kinases which interact with CC1114, by profiling phosphotransfer from CC1114∼P to the purified receiver domains of each of the 27 hybrid kinases encoded in C. crescentus (Fig. S3). With phosphotransfer reaction times of 2 min, two receiver domains, from the hybrid kinases CC0138 and CC0921, were efficiently phosphorylated, resulting in depletion of radiolabel from CC1114∼P (Fig. 2D). However, with a shorter reaction time (Fig. 2E), CC1114 exhibited a kinetic preference for phosphotransfer to the CC0138 receiver domain. Furthermore, the rate of transfer to the CC0138 and TacA receiver domains was approximately equal, suggesting that these are the kinetically preferred and hence most likely in vivo partners for CC1114 (Fig. 2E). Finally, as a control, we examined the phosphotransfer profile of the CC0138 kinase domain with respect to each of the 44 soluble, full-length C. crescentus response regulators. This experiment verified that CC0138 does not directly phosphorylate TacA or any response regulator other than its own receiver domain (data not shown).

Taken together, the phosphotransfer profiles suggest that CC1114 mediates a phosphorelay between the hybrid kinase CC0138 and the response regulator TacA (Fig. 2F). To verify the complete biochemical pathway, we mixed all components with [γ-32P]-ATP and incubated reactions at 30°C (Fig. 2G). Phosphorimaging showed that label accumulated mostly in TacA (Fig. 2G, lane 4). Excluding the CC0138 receiver domain (Fig. 2G, lane 2) or CC1114 (Fig. 2G, lane 3) from the reaction eliminated phosphorylation of TacA. These data support the conclusion that autophosphorylation of CC0138 leads to a phosphorelay through its own receiver domain and CC1114, culminating in the phosphorylation of TacA.

CC0138, CC1114 and TacA control stalk biogenesis and cell division

The phosphotransfer experiments suggested that CC1114 mediates an exclusive phosphorelay which leads to phosphorylation of TacA. As deletion strains for both tacA and CC0138 were found previously to lack stalks (Skerker et al., 2005), we predicted that a CC1114 deletion should also be stalkless. We therefore constructed a strain in which almost the entire coding region of CC1114 was replaced with a tetracycline resistance cassette. Examination of this strain by light microscopy confirmed that ΔCC1114, like ΔCC0138 and ΔtacA, is stalkless in rich (Fig. 3A, D, G and J) and minimal media (data not shown). However, deletion strains for each phosphorelay component were able to synthesize stalks when grown in low phosphate medium, a condition known to trigger extensive stalk growth (Poindexter, 1964; 1984; Gonin et al., 2000), suggesting that these strains are impaired in the regulation of stalk biogenesis, not stalk synthesis per se (Fig. 3C, F, I and L). Each deletion strain also appeared mildly filamentous with at least 20% of cells showing obvious cellular elongation relative to wild type, suggesting a defect in cell division or control of cell growth. Motile cells of ΔCC0138, ΔCC1114 and ΔtacA were easily observed by microscopy, but due to elongation cells did not swim as rapidly as wild type. Consistent with this observation, each phosphorelay mutant showed partial defects in a swarm plate motility assay (Fig. 3M). ΔCC0138, ΔCC1114 and ΔtacA each exhibited a swarm size intermediate between that of the wild-type CB15N and the non-motile strains ΔflbD and rpoN::Tn5, presumably due to the inability of the filamentous, mutant cells to swim efficiently through the agar.

Figure 3.

Phenotypic characterization of ΔCC0138, ΔCC1114 and ΔtacA.
A, D, G, J. Differential interference contrast (DIC) images of wild-type CB15N (A), ΔCC0138 (D), ΔCC1114 (G) and ΔtacA (J) grown to mid-log phase in PYE medium.
B, E, H, K. Fluorescent images of CB15N (B), ΔCC0138 (E), ΔCC1114 (H) and ΔtacA (K) stained with Syto-9 to visualize stalks.
C, F, I, L. DIC images of CB15N (C), ΔCC0138 (F), ΔCC1114 (I) and ΔtacA (L), grown in low phosphate medium, showing long stalks in this growth condition.
M. Swarm plate analysis of wild-type and mutant strains. CB15N forms a large swarm at the point of inoculation. ΔflbD and rpoN::Tn5 are both non-motile control strains. ΔtacA, ΔCC1114 and ΔCC0138 are motile, but produce smaller swarms due to defects in cell division.

The mutant phenotypes of ΔCC0138, ΔCC1114 and ΔtacA were each complemented by providing the deleted gene on a plasmid (Fig. 4A, C and H), demonstrating that the cell division and stalk phenotypes were due to a single gene deficiency in each case. ΔCC1114 cannot however, be complemented by an allele of CC1114 bearing the H56A mutation (Fig. 4B), consistent with the prediction that His-56 is the sole phosphorylation site on CC1114. Moreover, purified CC1114(H56A) also did not exhibit phosphotransferase activity in vitro (data not shown).

Figure 4.

Genetic analysis of the TacA phosphorelay. CC0138 = shkA, CC1114 = shpA.
A. Deletion of CC1114 results in a stalkless phenotype (Fig. 3G), which is complemented by providing a copy of the CC1114 gene in trans on the plasmid pHXM-CC1114 (ML691). The presence of stalks is indicated by white arrowheads.
B. CC1114 bearing a mutation of the active site histidine to an alanine (H56A) is unable to complement stalk formation in vivo.
A, C, H. Phosphorelay mutants can be complemented by providing the gene in trans on a xylose-inducible plasmid. (A) ML691 (ΔshpA+ pHXM-shpA), (C) ML690 (ΔshkA + pHXM-shkA), (H) ML692 (ΔtacA+ pHXM-tacA). All strains were grown in PYE + glucose, which provides a low level of expression of the gene being tested. Stalks are indicated by white arrowheads. DIC images of strains grown to mid-log phase.
D–I. Expression of tacA(D54E) but not tacA, is able to rescue the stalk defect of both ΔshkA (D vs. E) and ΔshpA (F vs. G). (H–I) Either tacA or tacA(D54E) can rescue a ΔtacA mutant.

In sum, deleting any component of the CC0138-CC1114-TacA phosphorelay led to a minor defect in cell division or control of cell growth resulting in elongated cells, and a major defect in stalk biogenesis. As growth, division, and stalk biogenesis each involve the coordination of peptidoglycan and cell membrane metabolism (Wagner et al., 2005), a single deficiency may underlie all of the morphological phenotypes of ΔCC0138, ΔCC1114 and ΔtacA. Moreover, the similarity of mutant phenotypes for CC0138, CC1114 and tacA supports the notion that they act in the same signalling pathway in vivo, as suggested by the in vitro results demonstrating a phosphorelay comprised of these proteins. As the major defect of the deletion strains is in stalk biogenesis we have named CC0138 shkA for stalk biogenesis histidine kinase A and named CC1114 shpA for stalk biogenesis histidine phosphotransferase A.

The constitutively active allele tacA(D54E) rescues ΔshkA, ΔshpA and ΔtacA

The data presented thus far suggest that ShkA and ShpA function in vivo to phosphorylate TacA. We therefore predicted that expression of a constitutively active form of tacA, tacA(D54E), should rescue deletion of shkA or shpA. In the NtrC family of response regulators, which includes TacA, changing the active site aspartate to a glutamate often mimics phosphorylation of the response regulator and hence renders its activation independent of upstream kinases or phosphorelays (Klose et al., 1993). We therefore placed a tacA(D54E) allele on the vector pJS71 under control of the xylose-inducible promoter PxylX (Meisenzahl et al., 1997) and transformed this plasmid into the strains ΔshkA, ΔshpA and ΔtacA. Each strain was grown in rich medium supplemented with glucose. As the pJS71 vector is a relatively high copy vector, growth in the presence of the repressor glucose leads to a low level of expression from PxylX but prevents overexpression. For ΔshkA, ΔshpA and ΔtacA, expression of tacA(D54E) rescued both the cellular elongation and stalk defects of the mutant strains (Fig. 4E, G and I). In contrast, expression of wild-type tacA from the same plasmid rescued ΔtacA (Fig. 4H), but did not rescue the mutant phenotypes of ΔshkA or ΔshpA (Fig. 4D and F). Even overexpression of wild-type tacA, by growth in 0.1% xylose, did not rescue the ΔshkA and ΔshpA mutants (data not shown). These data support the conclusion that the primary in vivo function of ShkA and ShpA is to phosphorylate, and hence activate, TacA.

Next, we examined the effect of overproducing TacA(D54E). Wild-type cells harbouring the PxylX-tacA(D54E) allele on pJS71 were grown in the presence of either glucose or xylose. In glucose, these cells appeared wild type in terms of morphology (Fig. 5A) and growth rate (data not shown). However, in xylose the cells grew almost twice as slowly as in glucose and rapidly became filamentous. Cells had abnormal stalks that were shorter or longer than wild type and in some cases appeared only as a small bud or protrusion from the cell pole (Fig. 5B). These growth and stalk defects could be the result of high levels of constitutively active TacA and/or the synthesis of active TacA at the wrong time during cell cycle progression. Together, the overexpression and deletion analyses indicate that TacA activity must be tightly regulated during the Caulobacter cell cycle to ensure proper stalk biogenesis.

Figure 5.

staR (CC2251) is involved in stalk formation.
A–D. Overexpression of tacA(D54E) in wild-type Caulobacter (CB15N) leads to cell filamentation, and defective stalk formation (A vs. B). Similarly, overexpression of CC2251 leads to filamentation and long stalks (C vs. D). White arrowheads indicate long stalks and open arrowheads indicate defective stalks.
E–G. StaR regulates stalk length. DIC images of wild-type CB15N (E) ΔCC2251 (F) and ML760 (CB15N + pLXM-CC2251) in PYE xylose (G) for comparison of stalk length. The deletion mutant ΔCC2251 has stalks approximately 1/2 the length of wild type, whereas overexpression of CC2251 results in stalks which are about twice the length of wild type (see text for details).

Identifying the TacA regulon with DNA microarrays

TacA is a member of the NtrC family of response regulators, also known as enhancer-binding proteins (EBPs), which typically function to activate σ54-dependent gene expression. σ54 is encoded in Caulobacter by the gene rpoN and is required for both flagellar and stalk biogenesis (Brun and Shapiro, 1992). FlbD was identified as the EBP response regulator which works with σ54 in predivisional cells to regulate the expression of flagellar genes (Ramakrishnan et al., 1991; Wingrove et al., 1993; Benson et al., 1994; Wingrove and Gober, 1994). Based on our results, we hypothesize that TacA collaborates with σ54 to control genes involved in regulating stalk biogenesis. To identify possible target genes, we used whole genome DNA microarrays to interrogate the expression patterns of TacA phosphorelay mutants (ΔshkA, ΔshpA, ΔtacA) and a rpoN disruption mutant rpoN::Tn5 (Fig. 6). To partition the σ54 regulon into FlbD- and TacA-dependent genes, we also examined a ΔflbD strain (Fig. 6).

Figure 6.

Expression profiles of ΔshkA, ΔshpA, ΔtacA, rpoN::Tn5 and ΔflbD. Colours, as indicated in the legend, represent upregulation (yellow) or downregulation (blue) in comparison with expression in wild-type cells.
A. Correlation between expression profiles of ΔshkA,ΔshpA,ΔtacA (see text for details). Columns, corresponding to ΔshkA, ΔshpA, ΔtacA, rpoN mutant and ΔflbD, show only genes upregulated or downregulated more than twofold in at least one of the mutants. For complete data, see Table S1.
B. Genes consistently downregulated in phosphorelay mutants and rpoN form a putative TacA regulon, and genes consistently downregulated in rpoN and ΔflbD the FlbD regulon.

Strains were grown to mid-log phase (OD600 ∼0.4) in rich medium, then RNA was harvested and used to probe whole-genome DNA microarrays. Each mutant strain was compared with the wild-type CB15N and each comparison done in duplicate or triplicate with results averaged. Genes whose expression level changed at least twofold in rpoN::Tn5 and in any one of the phosphorelay or flbD mutants were selected for further analysis (Fig. 6). Visual inspection of expression profiles for each mutant relative to wild type, clearly demonstrated the similarity of ΔshkA, ΔshpA, ΔtacA and rpoN::Tn5 (Fig. 6). Correlation coefficients for each pairwise comparison among these mutants were greater than 0.73. Similarly high correlation coefficients have been seen for mutants of other two-component signalling genes which function in the same pathway (Jacobs et al., 2003). These results thus add further support to the biochemical and genetic data demonstrating that ShkA-ShpA-TacA constitutes a phosphorelay which controls gene expression through phosphorylation of TacA.

We predicted that the TacA target genes involved in stalk biogenesis would be similarly affected in ΔshkA, ΔshpA, ΔtacA and rpoN::Tn5, each of which lacks stalks, but unchanged in the ΔflbD mutant which is non-motile, but retains stalks. In total, 30 genes satisfied these criteria, two of which are consistently downregulated in each stalkless mutant and 28 upregulated in each. The two downregulated genes contained σ54 binding sites (TGGCCC-N5-TTGC) (Wu et al., 1995) in their upstream regulatory regions, suggesting that each is a direct target of TacA and σ54. CC3218 has no predicted function based on sequence homology but is transcriptionally cell cycle-regulated (Laub et al., 2000) with expression highest in stalked cells and again in late predivisional cells. The other putative TacA target, CC2251, is predicted to encode a transcription factor of the Cro/CI family. During wild-type cell cycle progression CC2251 mRNA levels peak in predivisional cells, shortly after tacA expression peaks, suggesting that these regulators may form a transcriptional cascade which controls stalk biogenesis (Fig. 7).

Although EBP response regulators such as TacA typically activate gene expression, some can directly repress genes, independent of σ54 (Wingrove et al., 1993; Benson et al., 1994; Wingrove and Gober, 1994). Hence to identify genes which may be directly repressed by TacA, we searched for genes whose expression was significantly upregulated in ΔtacA, but unaffected in the rpoN mutant (Table S1). The only gene fitting these criteria was CC3314, which is annotated as a ‘conserved hypothetical’ and reads divergently from tacA itself. All other genes are similarly upregulated in rpoN::Tn5 and each of the phosphorelay mutants, suggesting that they are regulated indirectly by TacA, perhaps through the transcription factor CC2251. The vast majority of these genes have no annotated function, but their roles, if any, in stalk biogenesis can now be explored.

staR (CC2251) expression is regulated by the TacA phosphorelay and regulates stalk biogenesis

Microarray analysis identified at least two genes, CC2251 and CC3218, as probable direct targets of TacA. To determine their roles, if any, in stalk biogenesis, we constructed deletion and overexpression strains for each gene. Deletion and overexpression of CC3218 had no obvious phenotype (data not shown). Deletion of CC2251 did not eliminate stalks, but the average stalk length for ΔCC2251 cells (0.68 µm ± 0.24, Fig. 5F) was significantly shorter than wild type (1.26 µm ± 0.32, Fig. 5E). Conversely, over-expression of CC2251 from a xylose-inducible promoter on the low-copy vector pMR20 resulted in cells with stalks that were highly variable and longer on average (1.95 µm ± 0.89, Fig. 5G) than wild type. Overexpression of CC2251 from a high-copy (pJS71) vector also produced longer stalks in addition to a growth defect and filamentous cells (Fig. 5D). This growth defect and filamentation was similar to the phenotype of TacA(D54E) overexpression (Fig. 5B), consistent with CC2251 being a direct target of TacA. Based on these observations we have named CC2251 staR for stalk biogenesis regulator. We note though, that because ΔCC2251 does not precisely phenocopy ΔtacA, there must be other TacA target genes which also help to control stalk biogenesis. Candidates from the DNA microarray data will need to be examined to fully account for the stalkless phenotype of tacA, shkA and shpA mutants.


Stalk biogenesis in Caulobacter

For Caulobacter, a developmental program is tightly coupled to cell cycle progression. This development includes three major morphogenetic events, biogenesis of a flagellum, polar pili and a stalk. While the regulation of flagellum assembly and, to a lesser extent, pili biogenesis have been worked out in mechanistic detail (Gober and England, 2000; Skerker and Shapiro, 2000; Viollier et al., 2002a,b), the regulation of stalk biogenesis has remained poorly understood. Here, we have identified a key signal transduction pathway responsible for stalk formation during the cell cycle and have begun outlining the transcriptional pathways involved. This genetic circuitry is summarized in Fig. 7A.

At the heart of this model is σ54 which is necessary for both flagellum and stalk biogenesis. As in other bacteria, σ54-dependent gene expression in Caulobacter requires an activator of the NtrC-family of response regulators, also known as enhancer-binding-proteins, or EBPs. Typically, phosphorylation of an EBP enables it to stimulate the closed-to-open complex transition of a promoter bound by RNA polymerase containing a σ54 subunit, thereby coupling activation of the signalling pathway to changes in gene expression (Kustu et al., 1989). The EBP response regulator FlbD is induced at the same time as RpoN, and together they regulate genes involved in flagellar biogenesis (Fig. 7B) (Wingrove et al., 1993; Benson et al., 1994; Wingrove and Gober, 1994). We have now identified TacA as the EBP which collaborates with RpoN to regulate genes involved in stalk biogenesis. Furthermore, we identified a phosphorelay which regulates stalk biogenesis by modulating the phosphorylation state of TacA: ShkA, a hybrid histidine kinase, first autophosphorylates and then initiates a phosphorelay in which a phosphoryl group is passed to its own receiver domain, then to the histidine phosphotransferase ShpA, and finally to the transcriptional regulator TacA (Fig. 7).

DNA microarray analysis identified two genes controlled by TacA which may help regulate stalk biogenesis: CC2251 (staR) and CC3218. The mRNAs for both staR and CC3218 are cell cycle-regulated with induction in late predivisional cells, immediately after the induction of tacA (M. T. Laub, unpubl. data) (Fig. 7B). StaR appears to play a role in controlling stalk length, as the deletion strain has stalks shorter than wild type, and overexpression leads to longer stalks (Fig. 5). However, as staR mutants do not completely lack stalks or exhibit cellular elongation like tacA mutants, other TacA targets remain to be identified and characterized. The microarray data presented here can now be explored to find these additional targets.

The coupling of stalk biogenesis to cell cycle progression is accomplished in large part by CtrA, the master cell cycle regulator. CtrA is regulated at multiple levels during cell cycle progression such that activity drops during the G1-S transition and then rapidly returns to peak levels as S phase proceeds (Domian et al., 1997). CtrA triggers both flagellum and stalk biogenesis by directly activating rpoN (σ54) and tacA (Laub et al., 2002). Taken together, we propose that stalk biogenesis, like flagellum biogenesis, is regulated by a transcriptional cascade: CtrA is induced after DNA replication initiation and directly activates synthesis of σ54 and TacA, which together activate the expression of stalk regulatory genes, including the transcription factor staR, which in turn probably controls the expression of yet other genes required for stalk biogenesis.

Precisely how PleC, another histidine kinase required for stalk biogenesis, feeds into the stalk regulatory circuit remains unknown. However, suppressor analyses place CtrA downstream of PleC, suggesting that the effect may be through regulation of CtrA activity (Ohta et al., 1992; Wu et al., 1998).

Stalk biogenesis is a cell cycle-regulated event and initially occurs in only one of the two daughter cells produced by cell division, raising the possibility that TacA activity is controlled spatially as well as temporally. FlbD, for example, is phosphorylated and active only in the nascent swarmer cell, where it activates expression of late flagellar genes and represses early flagellar genes (Gober and England, 2000). In the nascent stalked cell, FlbD is present but inactive; this pool of FlbD will be activated as the stalked cell develops, initiating early flagellar gene expression in the predivisional cell. Whether TacA activity is also asymmetrically regulated in a manner that restricts stalk biogenesis to a single daughter cell, the stalked cell, remains to be explored.

What signals lead to activation of TacA?

Why does Caulobacter use a phosphorelay to control stalk biogenesis? It has been suggested that phosphorelays allow the integration of multiple signals by providing numerous points of control. In Bacillus subtilis, a number of signals impinge on the phosphorelay controlling sporulation initiation by affecting the phosphorylation of different pathway components (Perego et al., 1994). Stalk biogenesis is a major developmental event in the life cycle of Caulobacter, and hence may also integrate a wide range of signals. However, the nature of the signal(s) controlling the ShkA-ShpA-TacA phosphorelay is not yet known. ShkA is predicted to be a soluble, cytoplasmic kinase, suggesting that the phosphorelay may respond to an internal cell cycle cue. Caulobacter thus appears to have at least two independent pathways for regulating stalk biogenesis, one dependent on an external cue, phosphate starvation, and one linked to cell cycle progression.

Identification of a histidine phosphotransferase

ShkA-ShpA-TacA is the first phosphorelay identified in C. crescentus as ShpA is the first histidine phosphotransferase identified in this organism. The C. crescentus genome likely encodes other histidine phosphotransferases. CC1220, identified here in the same computational screen as ShpA, is one candidate, but others almost certainly exist, particularly given that the Caulobacter genome encodes 27 hybrid histidine kinases. Hybrid kinases often function in phosphorelays like the one described here; however, in some cases the C-terminal receiver domain of a hybrid kinase may function as an auto-inhibitory domain. Intramolecular phosphorylation may relieve this inhibition, allowing the kinase to phosphorylate another, diffusible response regulator, as with the Agrobacterium tumefaciens hybrid kinase VirA (Chang et al., 1996).

Rapid, systematic mapping of complex phosphorylation cascades

To identify the probable in vivo cognate substrates for ShpA, we adapted a technique called phosphotransfer profiling (Skerker et al., 2005). By examining phosphotransfer from a kinase to each response regulator in parallel, and at multiple time points, kinetic preference is revealed and identifies the probable in vivo targets of the kinase (Skerker et al., 2005). Here, we hypothesized that histidine phosphotransferases would similarly exhibit a kinetic preference with respect to both soluble response regulators and the receiver domains of hybrid histidine kinases. A previous smaller scale study of two phosphorelays also suggested that histidine phosphotransferases exhibit kinetic preference in vitro (Perraud et al., 1998). Our phosphotransfer profiling experiments indeed demonstrated a kinetic preference for ShpA to transfer phosphoryl groups to or from the receiver domains of ShkA and TacA. Our genetic analyses (Figs 3 and 4) and DNA microarray results (Fig. 6) confirmed the in vivo relevance of these interactions. We suggest that histidine phosphotransferases in general may exhibit a global kinetic preference for interaction with their cognate substrates, similar to what has been observed for histidine kinases.

The global kinetic preference of histidine kinases for their cognate response regulator substrates means that the specificity of two-component signal transduction pathways is determined largely at a biochemical level; other factors probably function primarily to enhance this inherent specificity. Our results with ShpA suggest that these observations extend to histidine phosphotransferases and are general properties of two-component signalling pathways. We thus anticipate that our phosphotransfer profiling technique will be generally applicable to the mapping of phosphorelays, both in Caulobacter as well as in other organisms. Two-component signalling systems are prevalent throughout the bacterial kingdom as well as being present in fungi and plants (Stock et al., 2000), and the mapping of pathway connectivity among large numbers of signalling components remains a major challenge.

Concluding remarks

Many organisms couple developmental programs and morphogenesis to cell cycle progression. In the budding yeast S. cerevisiae, mating requires a morphological change called shmooing which prepares haploid cells for fusion. This morphological event is tied to cell cycle progression and can only occur during G1 phase. For B. subtilis the initiation of sporulation, a complex developmental program, is also tightly coupled to cell cycle events (Burkholder et al., 2001). The morphological events of flagellar, pili and stalk biogenesis in Caulobacter are similarly dependent on cell cycle progression. The identification of a complete signalling pathway controlling stalk biogenesis now opens the way to a more complete understanding of how this key morphological event is regulated in space and time during the Caulobacter cell cycle.

Experimental procedures

Bacterial strains, plasmids and growth conditions

Caulobacter crescentus strains were grown in peptone-yeast extract (PYE, rich medium) or M2G (minimal medium) at 30°C (Ely, 1991), supplemented with 3% sucrose, tetracycline (1 µg ml−1), kanamycin (25 µg ml−1), spectinomycin (25 µg ml−1), 0.1% glucose, or 0.1% xylose, as required. Low-phosphate growth experiments were performed as previously described (Quon et al., 1996). E. coli strains were grown at 30°C or 37°C in Luria–Bertani broth supplemented with carbenicillin (100 µg ml−1), chloramphenicol (30 µg ml−1), tetracycline (10 µg ml−1), kanamycin (50 µg ml−1) or spectinomycin (50 µg ml−1) as necessary. PYE swarm plates contained 0.3% bacto agar. Plasmids were transformed into C. crescentus by electroporation. Tetracycline marked deletion strains for CC1114, CC2251 and CC3218 were generated by a two-step recombination protocol (Skerker et al., 2005). pENTR clones of the 27 hybrid kinase receiver domains were constructed as previously described (Skerker et al., 2005). For in vivo expression of Caulobacter genes, pENTR plasmids for CC0138, CC1114, CC1114 (H56A), TacA, TacA(D54E), CC2251 and CC3218 were recombined with either a high-copy destination vector pHXM-DEST or a low-copy destination vector pLXM-DEST (Table S2), as necessary, to generate xylose-inducible constructs (Table S2). For overexpression experiments in Caulobacter, using pHXM-based vectors, saturated overnight cultures grown in PYE plus glucose were washed twice in PYE medium and then diluted to OD600 = 0.03 in either PYE plus glucose or xylose. For experiments using the low copy vector pLXM-CC2251 samples were grown overnight in PYE plus xylose prior to analysis.

Strains and plasmids generated are listed in Table S2. Primers and plasmids used to generate deletion strains, pENTR clones and point mutations are listed in Table S3. Site-directed mutagenesis was performed using the QuikChange protocol (Stratagene), with pENTR clones as templates. Morphology of mid-log phase cultures was imaged as described previously using DIC microscopy (Skerker et al., 2005). Stalks were visualized by staining with 5 µM Syto-9 (Molecular Probes).

Sequence analysis and homology modelling

Alpha-helical content was estimated using the Garnier algorithm within EMBOSS (http://emboss.sourceforge.net). A structure-based multiple sequence alignment of CC1114 against other known phosphotransferases was constructed using 3D-Coffee (Poirot et al., 2004). This alignment was used as input to Swiss-Model (Schwede et al., 2003) to predict the structure of CC1114 using the following templates: ArcB (PDB: 1FR0), Ypd1 (PDB: 1QSP) and LuxU (PDB: 1Y6D). ArcB provided the best alignment and the predicted CC1114 structure was analysed and displayed using UCSF Chimera (Pettersen et al., 2004).

Gateway cloning and protein purification

Escherichia coli thioredoxin (TRX)-His6 expression plasmids for all 44 response regulators and 27 hybrid kinase receiver domains were generated using recombinational cloning and purified as described previously (Skerker et al., 2005). CC1114 and CC1114(H56A) were expressed and purified as N-terminal His6-tagged proteins using the destination vector, pHIS-DEST. To construct pHIS-DEST, pET15b (Novagen) was digested with NdeI, blunted with T4 DNA polymerase, and the RfC Gateway vector conversion cassette (Invitrogen) was inserted. A His6-version of the TacA receiver domain (TacA-RD) was also constructed and purified, using pHIS-DEST. TacA-RD yielded a cleaner purification, and acted identical to full-length TacA in our assays, so we substituted this protein as necessary. Entry clones of the kinase domains of CckA, CC0138, CC3102 were recombined into pTRX-HIS-DEST or pHIS-MBP-DEST (for CC3219), and purified as either TRX-His6 or His6-MBP fusions. All purified proteins were stored in HKEDG buffer (10 mM HEPES-KOH, pH 8.0, 50 mM KCl, 10% glycerol, 0.1 mM EDTA, 1 mM DTT) and concentrations were normalized by densitometry.

Phosphotransfer profiling and phosphorelay biochemistry

We used a modified phosphotransfer profiling method (Skerker et al., 2005) to identify the cognate substrates for CC1114 (Fig. 2C and D, Fig. S2). All reactions and dilutions were performed in HKEDG buffer plus 5 mM MgCl2. First, we generated CC1114∼P by incubating 10 µM His6-CC1114 with 1 µM TRX-His6-CC0138-HK and 1 µM TRX-His6-CC0138-RD plus 1 µCi µl−1[γ-32P]ATP (∼6000 Ci mmol−1, Amersham Biosciences) for 30 min at 30°C. This reaction was then diluted 10-fold, and 5 µl of the mixture was added to 5 µl of buffer alone (as a control) or 5 µl of each purified response regulator or hybrid receiver domain (diluted to 5 µM) with incubations for 2 min at room temperature. The final concentrations were 0.5 µM for CC1114, and 2.5 µM for the phosphotransfer substrates. To demonstrate phosphorylation of CC1114 by various hybrid kinases (Fig. 2A), the following four autophosphorylation reactions (30 µl total volume) were performed: (i) TRX-His6-CckA (6 µM), (ii) TRX-His6-CC0138-HK (0.5 µM), (iii) TRX-His6-CC3102-HK (4.5 µM), or (iv) His6-MBP-CC3219-HK (20 µM) were incubated for 30 min at 30°C after addition of 30 µCi [γ-32P]ATP. Then, 1 µl of reaction A, 1 µl of reaction B, 4 µl of reaction C, and 9 µl of reaction D were mixed with 1 µM of the corresponding receiver domains of each histidine kinase, and, as necessary, with 10 µM of His6-CC1114 for 2 or 5 min, in a total volume of 10 µl. For analysis of CC1114 activity (Fig. 2B), reactions were performed the same as for phosphotransfer profiling except that various components were omitted before incubating 30 min at 30°C. For analysis of the complete TacA phosphorelay and kinetics (Fig. 2E and G), CC1114∼P was prepared as above, diluted, and then mixed with 5 µl of 5 µM His6-TacA-RD or other hybrid kinase receiver domains. All phosphotransfer reactions were stopped with 3.5 µl of 4 × SDS-PAGE sample buffer and analysed using 12% or 15% Tris-HCl gels as described previously (Skerker et al., 2005).

Microarray analysis of the TacA regulon

Strains were grown in PYE and harvested at OD600 = 0.4. RNA preparation was performed using the RNA-easy Kit (Qiagen). DNA arrays contained 50mer probes for each gene in the Caulobacter genome (see Table S1 for sequences). Three or more independent cultures were used for each experiment and each sample was labelled either with Cy5-dCTP or Cy3-dCTP (Amersham Biosciences). Hybridizations, scanning and data processing were performed as described previously (Laub et al., 2002) except that hybridizations contained 30% formamide and were incubated at 44°C for 6 h.


We gratefully acknowledge support from the Office of Science (BER), US Department of Energy, Grants No. DE-FG03-01ER63219 and DE-FG02-04ER63922. Support also provided in part by a National Institutes of Health grant to M.T.L. and the Bauer Center for Genomics Research at Harvard University.