Allosteric Regulation of Histidine Kinases by Their Cognate Response Regulator Determines Cell Fate

SUMMARY

The two-component phosphorylation network is of critical importance for bacterial growth and physiology. Here, we address plasticity and interconnection of distinct signal transduction pathways within this network. In Caulobacter crescentus antagonistic activities of the PleC phosphatase and DivJ kinase localized at opposite cell poles control the phosphorylation state and subcellular localization of the cell fate determinator protein DivK. We show that DivK functions as an allosteric regulator that switches PleC from a phosphatase into an autokinase state and thereby mediates a cyclic di-GMP-dependent morphogenetic program. Through allosteric activation of the DivJ autokinase, DivK also stimulates its own phosphorylation and polar localization. These data suggest that DivK is the central effector of an integrated circuit that operates via spatially organized feedback loops to control asymmetry and cell fate determination in C. crescentus. Thus, single domain response regulators can facilitate crosstalk, feedback control, and long-range communication among members of the two-component network.

INTRODUCTION

Asymmetric cell division underlies the fundamental basis for the developmental evolution of organisms. It refers to the capability of stem cells to simultaneously produce a continuous output of differentiated cells and to maintain their own population of undifferentiated cells. The regulation of asymmetric division is achieved by the controlled segregation of basally localized cell fate determinants, which leads to the polarization of the stem cell along its axis (Betschinger and Knoblich, 2004). In the bacterium Caulobacter crescentus, asymmetry is established by members of the two-component signaling systems, which control various aspects of bacterial physiology including cell differentiation and virulence. Two sensor histidine kinases, DivJ and PleC, are positioned to opposing poles of the Caulobacter predivisional cell (McAdams and Shapiro, 2003). A cylindrical extension of the cell body, the stalk, and an adhesive holdfast occupy the DivJ-marked pole, while the PleC-occupied pole bears a rotating flagellum and adhesive pili (Figure 7A). Upon division the stalked cell re-enters S-phase immediately, whereas the motile-swarmer cell takes advantage of the replication inert G1 phase to spread out before it undergoes reprogramming into a surface adherent stalked cell.

Figure 7. Model for the Regulation of the PleC-DivJ-DivK Cell Fate Control System.

(A) Schematic of the C. crescentus cell cycle with polar appendices and localization patterns for PleC, DivJ, DivK, DivK~P, PleD, and PleD~P. A summary of the posttranslational feedback circuitry with the DivJ kinase, the PleC phosphatase, and the DivK response regulator is indicated.

(B) Diagram of the phosphorylation circuit controlling cell fate and cellular asymmetry in Caulobacter crescentus. The model predicts that the sessile-stalked cell program is directed through polar DivK~P interacting with and activating both PleC and DivJ kinases. Likewise, the motile-swarmer cell program results from the absence of the DivJ kinase, dissolution of DivK from the cell pole, and establishment of the PleC phosphatase activity. The two known molecular targets of the PleC-DivJ-DivK cell fate machinery, PleD and CtrA, are indicated. Stippled lines indicate PleC and DivJ polar localization control through CtrA~P-regulated expression of podJ and spmX, respectively (Crymes et al., 1999; Radhakrishnan et al., 2008).

C. crescentus cell fate is implemented by the essential single domain response regulator DivK (Hecht et al., 1995; Matroule et al., 2004). DivK localizes to both poles of the predivisional cell in a phosphorylation-dependent manner, but is released from the flagellated pole after completion of cytokinesis (Figure 7A) (Jacobs et al., 2001). While the DivJ autokinase is the main phosphodonor for DivK and responsible for its sequestration to the cell poles (Lam et al., 2003), the PleC phosphatase activity displaces DivK from the flagellated pole by maintaining DivK~P levels low in the swarmer cell (Lam et al., 2003; Matroule et al., 2004). Compartmentalization of the DivJ kinase and the PleC phosphatase during cell division results in the sudden reduction of DivK~P levels in the swarmer cell and the initiation of the swarmer-specific developmental program (Matroule et al., 2004). Conversely, a rapid DivJ-mediated increase of DivK phosphorylation is critical for G1-to-S transition and cell differentiation (Hung and Shapiro, 2002; Jacobs et al., 2001; Wu et al., 1998). Activated DivK has recently been proposed to control cell cycle progression and development via the CckA-ChpT pathway, which regulates the activity of the master cell cycle regulator CtrA (Biondi et al., 2006).

Although it is clear that DivK phosphorylation by DivJ and dephosphorylation by PleC are vital for C. crescentus cell cycle control and development, the significance of the spatial behavior of DivK and its molecular role remain unclear. Here we propose that DivK together with PleC and DivJ form the core of an integrated regulatory circuitry that operates via spatially organized cellular feedback loops. We show that DivK directly interacts with both polar proteins to strongly boost their kinase activities. By switching PleC from the phosphatase into the autokinase mode and by forming a strong positive feedback loop with the DivJ autokinase, DivK effectively and robustly mediates G1-to-S transition.

One of the readouts of the DivK-driven network is the synthesis of the second messenger cyclic di-GMP via the activation of the response regulator PleD (Aldridge et al., 2003; Paul et al., 2004). PleD phosphorylation results in dimerization-based activation of the C-terminal diguanylate cyclase domain and sequestration of the regulator to the differentiating pole where it directs flagellar ejection, holdfast biogenesis, and stalk formation (Levi and Jenal, 2006; Paul et al., 2007, 2004). Genetic experiments indicated that DivJ, PleC, and DivK are upstream components required for the activation of the PleD diguanylate cyclase (Aldridge et al., 2003; Sommer and Newton, 1991). Consistent with this, in vivo and in vitro experiments had shown a direct role for PleC and DivJ in modulating phosphorylation, diguanylate cyclase activity, and polar localization of PleD (Aldridge et al., 2003; Paul et al., 2004). This indicated that PleC, in addition to functioning as phosphatase in swarmer cells, also plays a role as autokinase and contributes to the stalked cell-specific program via PleD activation. We present in vitro and in vivo evidence that DivK together with the DivJ and PleC kinases serves to activate and sequester the PleD diguanylate cyclase during the G1-to-S transition. DivK contributes to the specific phosphorylation of PleD by acting as a specific and effective enhancer of the PleC and DivJ autokinases. Our findings propose a regulatory role for single domain response regulators in two component signal transduction pathways as diffusible modulators of their cognate sensor histidine kinases.

RESULTS

DivK Stimulates the PleD Diguanylate Cyclase Activity in a PleC-Dependent Manner

To explore the regulatory link between PleD, PleC, and DivK, activation of PleD diguanylate cyclase activity was assayed in vitro in the presence of PleC and DivK. In line with earlier results (Paul et al., 2004) PleC alone only marginally stimulated PleD activity (Figure 1A). Surprisingly, in the presence of DivK or DivK_D53N, a mutant that cannot be phosphorylated because it lacks the phosphoryl acceptor site, PleD diguanylate cyclase activity was dramatically stimulated (Figures 1A and 1B). Stimulation by DivK wild-type was less effective, presumably because DivK itself can use PleC~P as phosphodonor (Hecht et al., 1995) and thus can sequester some of the available phosphoryl groups. DivK-dependent stimulation of PleD activity required ATP, PleC autokinase activity, and the PleD phosphoryl acceptor site Asp53 (Figure 1B). In particular, purified PleC_F778L, a mutant that lacks autokinase activity but shows normal phosphatase activity (Matroule et al., 2004) failed to support DivK-dependent stimulation of PleD. Because PleD is activated by dimerization (Paul et al., 2007), residual diguanylate cyclase activity can be detected at this protein concentration (5 μM) even in the absence of phosphorylation (the K_d for dimerization of nonactivated PleD is 100 μM [Wassmann et al., 2007]) (Figure 1B). It is interesting to note that when all three proteins were present PleD diguanylate cyclase activity decreased below the basal level of nonactivated PleD under conditions that did not allow phosphorylation (no ATP, no PleC autokinase activity) (Figure 1B). Altogether, these experiments demonstrate that the single domain response regulator DivK is able to efficiently stimulate the in vitro activity of PleD and that this activation requires an active histidine protein kinase PleC.

Figure 1. In Vitro Activation of the Diguanylate Cyclase PleD by the Histidine Kinase PleC and the Response Regulator DivK.

(A) PleC and PleD were incubated with DivK or DivK_D53N in the presence of ATP and [α-³³P]GTP. The formation of c-di-GMP was determined as initial velocities of the enzymatic reactions (Paul et al., 2007). Reactions contained 1.25 μM PleD, 10 μM PleC, and increasing amounts of DivK_D53N or DivK (0.1 μM, 0.3 μM, 1 μM, 10 μM or 30 μM). Error bars represent the mean ± standard deviation (SD).

(B) Reaction conditions were as indicated in (A). Reactions contained 10 μM PleC wild-type or PleC_F778L, 5 μM PleD or PleD_D53N, and 10 μM DivK_D53N or DivK_D53ND90G.

(C) Reaction conditions were as indicated in (A). Reactions contained DivJ (10 μM), PleD (10 μM), and DivK, DivK_D53N, or DivK_D90G (10 μM, 1 μM, 0.3 μM, or 0.1 μM).

DivK Stimulates PleC Autokinase but Not Phosphatase Activity

The above experiments suggested that DivK activates PleD by interfering with PleC autokinase activity. To test this, PleC autophosphorylation activity was monitored in the presence and absence of DivK. Because PleC readily phosphorylates DivK in vitro (Hecht et al., 1995; Wu et al., 1998), we used DivK_D53N to avoid a reduction of PleC~P by phosphotransfer to DivK. As shown in Figure 2A and Figure S1A (available online), DivK_D53N stimulated levels of PleC~P in a concentration dependent manner. Because the stability of PleC~P was not affected by DivK (Figure S1B), DivK_D53N seems to specifically stimulate PleC autophosphorylation activity. As a consequence of this stimulation, phosphotransfer from PleC to PleD (Figure 2B) as well as PleD dimerization (Figure S2) was increased when both response regulators were present in the reaction. This is consistent with the observed increase in c-di-GMP synthesis in reactions containing PleD, PleC and DivK (Figure 1). Likewise, phosphotransfer from PleC~P to DivK was stimulated in the presence of DivK_D53N (Figure 3A). Stimulation of PleC autokinase activity did not result from a DivK-mediated in vitro artifact (e.g., through DivK assisted folding of PleC), as solubility, quarternary structure, and activity of PleC preincubated with or without DivK_D53N was indistinguishable (Figure S3). Finally, we tested if DivK_D53N was able to stimulate PleC phosphatase activity. Purified DivK~P was mixed with PleC in the presence or absence of DivK_D53N to assay dephosphorylation rates. In contrast to PleC mediated phosphorylation of DivK, dephosphorylation of DivK was not increased in the presence of DivK_D53N (Figures 3A–3C). Rather, the rate of DivK~P dephosphorylation was reduced in the presence of DivK_D53N. In conclusion, these experiments provide strong evidence that the response regulator DivK is able to selectively stimulate the kinase but not the phosphatase activity of its cognate histidine kinase PleC.

Figure 2. DivK Stimulates Autophosphorylation of the PleC Histidine Kinase.

(A) In vitro autophosphorylation with PleC (10 μM) and DivK_D53N (0.3, 10 μM).

(B) Phosphotransfer reactions with PleC (10 μM), DivK_D53N (10 μM), and PleD (10 μM).

(C) In vitro autophosphorylation of PleC (10 μM) in the presence of DivK_D53N (10 μM) or DivK_D53ND90G (10 μM).

(D) In vitro phosphotransfer between DivJ (2.5 μM) and the response regulators DivK (2.5 μM) and PleD (2.5 μM or 50 μM). The bands corresponding to the phosphorylated proteins are marked.

Figure 3. DivK Does not Stimulate PleC Phosphatase Activity.

(A) PleC-mediated phosphorylation of DivK. PleC (10 μM) and DivK (10 μM) were incubated with [γ-³²P]ATP in the presence or absence of DivK_D53N (15 μM) for the times indicated.

(B) PleC-mediated dephosphorylation of DivK. Purified DivK~P was incubated alone (top row), with DivK_D53N (15 μM; second row), with PleC (2.5 μM; third row), or with DivK_D53N and PleC (bottom row) for the times indicated.

(C) Quantification of PleC-mediated DivK phosphorylation (squares) and dephosphorylation (circles) in the presence (open symbols) or absence of DivK_D53N (closed symbols). The phosphatase assays were normalized for DivK autophosphatase activity.

DivK and PleD Compete for Phosphorylation by the DivJ Kinase

Because DivK and PleD also interact with DivJ, we tested if the response regulators also showed some synergistic behavior with respect to DivJ or if they would compete for DivJ kinase. As shown in Figure 1C, the PleD diguanylate cyclase was activated by DivJ, but the addition of DivK efficiently blocked PleD activation. PleD activation was restored only when DivK was diluted below a molar ratio of 1:10. DivK_D53N also reduced DivJ-mediated PleD activation but was less efficient than DivK wild-type (Figure 1C). These results suggested that PleD and DivK compete for the phosphodonor DivJ~P. This was confirmed by monitoring phosphotransfer from DivJ~P to the two response regulators. Although DivJ~P readily served as phosphodonor for PleD in the absence of DivK, phosphoryl groups were transferred exclusively to DivK when both response regulators were present in the reaction mixture (Figure 2D). Altogether, these experiments show that DivK and PleD compete for phosphorylation by the stalked pole specific kinase DivJ, and that in vitro DivJ~P prefers to transfers phosphate to DivK.

DivK Stimulates Autophosphorylation of DivJ

The observation that DivK is able to stimulate PleC autokinase activity in vitro raised the possibility that DivK might also modulate the activity of its other cognate kinase, DivJ. Similar to its observed effect on PleC, DivK_D53N efficiently stimulated DivJ autophosphorylation (Figures 4A and S1C), and did not affect the stability of DivJ~P (Figure S1D). As for PleC, pre-incubation of DivJ with DivK_D53N did not affect kinase solubility or quarternary structure, and stimulation by DivK was indistinguishable in untreated and DivK pre-treated samples (Figure S4). Activation of DivJ by DivK_D53N led to increased phosphotransfer to the response regulators DivK and PleD (Figure 4B). In conclusion, DivK acts both as phosphoryl acceptor and as a potent activator of the stalked pole-specific kinase DivJ. A nonphosphorylatable form of PleD, PleD_D53N, had no stimulatory effect on DivJ or PleC (Figure S1C), arguing that the activation of the DivJ and PleC autokinase activities by DivK is specific for this response regulator.

Figure 4. A Positive Feedback Mechanism Stimulates DivK Phosphorylation by DivJ.

(A) DivK stimulates DivJ autophosphorylation. DivJ (10 μM) was incubated with [γ-³²P]ATP and DivK_D53N (30 μM) or DivK_D53ND90G (30 μM) for the indicated times.

(B) Phosphotransfer reactions with DivJ (10 μM), DivK (10 μM), PleD (10 μM), DivK_D53N (10 μM), and DivK_D53ND90G (10 μM). The bands corresponding to the phosphorylated proteins are indicated. The band labeled with an asterisk is not visible on the Coomassie-stained gel and most likely represents a gel artifact.

The Developmental Mutant DivK_D90G Fails to Stimulate PleD Activation by PleC

Consistent with the postulated cell-cycle role of DivK during the G1-to-S transition, the cold-sensitive divK_D90G mutant arrests in G1 at the restrictive temperature (Wu et al., 1998). At the permissive temperature, this strain shows a developmental phenotype strikingly similar to mutants lacking PleD (see below). To test if altered interactions of the mutant protein with DivJ or PleC account for this phenotype we analyzed the ability of DivK_D90G to activate the DivJ and PleC autokinases. DivK_D90G was slightly less efficient in competing with PleD for the activation by DivJ (Figure 1C). To analyze the interaction of DivK_D90G with the DivJ and PleC kinases, we constructed a double mutant protein that lacked the phosphoryl acceptor Asp53. DivK_D53ND90G was still able to stimulate DivJ autophosphorylation and phosphotransfer to DivK or PleD (Figure 4B). In contrast, DivK_D53ND90G was unable to stimulate PleD and PleC autophosphorylation (Figures 1B and 2C), and failed to stimulate PleD activation by PleC (Figure 1B). Phosphotransfer from PleC~P to DivK_D90G was similar to DivK wild-type (Figure S5), and DivK_D53ND90G did not affect PleC phosphatase activity (Figure S6). Together, these experiments show that the D90G mutation specifically affects the ability of DivK to stimulate PleC autokinase.

PleD-Dependent Pole Morphogenesis Requires the PleC Autokinase

Although the phosphatase activity of PleC is sufficient to initiate pole development and induce motility in the swarmer cell (Matroule et al., 2004), it has not been tested if PleC kinase activity plays a role during the next, PleD-dependent step of pole morphogenesis. Because PleC and DivK are able to efficiently activate PleD in vitro, we analyzed the role of PleC kinase in promoting PleD-dependent developmental processes like motility, stalk formation, holdfast synthesis, and surface attachment (Aldridge and Jenal, 1999; Aldridge et al., 2003). For this purpose, we generated a ΔpleC single and a ΔpleC ΔpleD double mutant in the surface-binding wild-type strain CB15 (ATCC19089) and used these strains for complementation experiments with pleC alleles encoding PleC variants with kinase and phosphatase activity (K⁺P⁺; pleC_WT), with only phosphatase activity (K⁻P⁺; pleC_F778L), or lacking both activities (K⁻P⁻; pleC_T614R) (Matroule et al., 2004). To corroborate that the F778L mutation specifically affects PleC autokinase but not phosphatase activity in vivo, we showed that DivK~P levels were similarly reduced in swarmer cells of strains harboring pleC_WT or pleC_F778L, respectively (Figure S7).

C. crescentus surface binding is developmentally controlled and requires an active flagellum, the formation of polar pili, and synthesis of an adhesive holdfast (Levi and Jenal, 2006). Because of its block in pole development, the ΔpleC mutant completely failed to attach to plastic surfaces, whereas a ΔpleD mutant inefficiently adhered to surfaces because of a delay in holdfast formation (Levi and Jenal, 2006) (Figure 5A). Complementation with pleC wild-type fully restored surface attachment of the ΔpleC, but not of the ΔpleC ΔpleD double mutant. In contrast, the pleC_F778L (K⁻P⁺) allele partially restored surface binding of the ΔpleC mutant to the same level observed for a ΔpleD mutant (Figure 5A), arguing that in strains lacking PleC kinase activity development is initiated but cannot proceed past the stage of PleD activation. In comparison, expression of PleC_T614R (K⁻P⁻) in the ΔpleC strain showed no effect (Figure 5A). Reduced attachment was the result of inefficient formation of the primary cell adhesin, the holdfast, during development. Synchronized cultures of ΔpleC mutant cells complemented with pleC wild-type developed a visible holdfast 15–30 min after swarmer cells were released into fresh medium (Figure 6A). A similar pattern was observed for C. crescentus wild-type cells (Levi and Jenal, 2006). In contrast, cells expressing pleC_F778L (K⁻P⁺) showed a severe delay in holdfast synthesis and reduced surface attachment during the swarmer-to-stalked cell differentiation (Figure 6A). Because activation of the PleD response regulator results in the production of c-di-GMP (Paul et al., 2004), we next tested if PleD-dependent pole morphogenesis correlates with the production of the second messenger during development. A pronounced peak of c-di-GMP was observed early in the C. crescentus cell cycle coincident with the onset of holdfast biogenesis both in wild-type (data not shown) and in the ΔpleC mutant complemented with wild-type pleC (Figure 6B). Because c-di-GMP levels failed to increase in pleD mutant cells or in cells lacking PleC kinase activity (K⁻P⁺), we conclude that both activities are necessary to increase second messenger concentration during Caulobacter cell differentiation.

Figure 5. DivK Stimulates PleC and DivJ Autokinase Activities in Vivo.

(A) PleC kinase activity is required for PleD-dependent surface attachment. Attachment of cells to polystyrene was measured for the following strains as described in (Levi and Jenal, 2006): CB15 wild-type, ΔpleC (ΔC), and ΔpleC ΔpleD (ΔCD). As indicated, the strains contained an empty vector (pMR) or a plasmid copy of the following pleC alleles: pleC wild-type (pC_WT), pleC_F778L (pC_F778L), or pleC_T614R (pC_T614R). Error bars represent the mean ± SD.

(B) and (C) Surface attachment of the following strains was measured: CB15 wild-type, ΔpleD (ΔD), ΔpleC (ΔC), divK_D90G ΔpleC (K_D90GΔC), and divK_D90G ΔpleC ΔpleD (K_D90GΔCD). The strains contained an empty vector (pMR) or a plasmid copy of the pleC wild-type (pC_WT) or pleC_T614R (pC_T614R) allele as indicated.

(D) Levels of PleC~P in the ΔpleC deletion strain containing an empty vector (ΔC+pMR), the ΔpleC deletion strain containing a plasmid copy of the pleC wild-type (ΔC+pC_WT), and the divK_D90G ΔpleC double mutant containing a plasmid copy of the pleC wild-type (K_D90GΔC+pC_WT). The band corresponding to PleC~P is marked.

(E) Levels of DivJ~P per cell in a C. crescentus divK⁺ strain and a divK:Ω null mutant (left panel). Immunoblots with anti-DivJ and anti-DivK antibodies, respectively, are shown on the right. Error bars represent the mean ± SD.

Figure 6. PleD-Dependent c-di-GMP Synthesis and Holdfast Formation during Development Requires DivK and PleC Kinase Activity.

(A) Holdfast formation (open symbols) and attachment (closed symbols) during the G1-to-S transition. Swarmer cells of the following strains were isolated and suspended in M2G medium: ΔpleC deletion strain containing a plasmid copy of the pleC wild-type (squares), ΔpleC deletion strain containing a plasmid copy of the pleC_F778L allele (diamonds), and a divK_D90GΔpleC double mutant containing a plasmid copy of the pleC wild-type (circles). Error bars represent the mean ± SD.

(B) Measurements of c-di-GMP levels during the G1-to-S transition. The following strains were analyzed: ΔpleC deletion strain containing a plasmid copy of the pleC wild-type (squares), ΔpleC deletion strain containing a plasmid copy of the pleC_F778L allele (circles), ΔpleC ΔpleD deletion strain containing a plasmid copy of the pleC wild-type (diamonds), and a divK_D90GΔpleC double mutant containing a plasmid copy of the pleC wild-type (gray circles).

(C) PleC-mYFP and DivJ-tDimer2 localization was examined by epi-fluorescence microscopy in synchronized cell populations of UJ4507 at the indicated time points. Arrows indicate polar foci.

Likewise, motility control and stalk biogenesis during C. crescentus cell differentiation required both PleD and an active PleC autokinase (Figure S8). Together these findings suggest that while PleC phosphatase activity is sufficient to activate the motility program in newborn swarmer cells, consecutive steps in pole development during the G1-to-S transition require PleC autokinase activity to activate the PleD diguanylate cyclase.

DivK Stimulates the PleC-PleD Signal Transduction Pathway Involved in Cell Fate Determination

To investigate the in vivo role of DivK in pole development we first tested if increased cellular concentrations of DivK_D53N could stimulate PleC- and PleD-dependent surface attachment. Cells containing an additional plasmid-borne copy of divK_D53N showed increased surface attachment as compared to cells harboring a control plasmid (Figures 5B and 5E). Importantly, attachment was not increased in strains lacking PleD or PleC autokinase activity, arguing that higher levels of DivK_D53N increased attachment in a PleC- and PleD-dependent manner.

Next, we compared surface attachment of C. crescentus wild-type with that of a divK_D90G mutant strain. The observation that DivK_D90G failed to stimulate PleD activity through the PleC kinase in vitro suggested that the PleD-dependent pathway controlling pole development might be inactive in the divK_D90G mutant at the permissive temperature. As shown in Figure 5C, the complete deficiency of the ΔpleC mutant to attach to surfaces was partially suppressed in the presence of the divK_D90G allele. This is consistent with the observation that divK_D90G is able to bypass the cell division checkpoint of cells lacking the PleC phosphatase (Matroule et al., 2004; Sommer and Newton, 1991). However, attachment of a ΔpleC divK_D90G double mutant remained at an intermediary level typically observed for cells lacking PleD, and an extrachromosomal copy of the pleC wild-type gene could not restore wild-type levels of attachment in this mutant background (Figure 5C). Finally, a ΔpleC ΔpleD divK_D90G triple mutant when complemented with pleC wild-type showed the same intermediary attachment level, arguing that PleD was not activated in the presence of the divK_D90G allele. Like mutants lacking PleD or the PleC autokinase activity, divK_D90G mutant cells showed a severe delay in holdfast synthesis during the swarmer-to-stalked cell differentiation (Figure 6A) and were unable to produce the characteristic PleD-dependent peak of c-di-GMP during the G1-S transition (Figure 6B).

Altogether, these experiments corroborate the results obtained in vitro and suggest that DivK is required for the activation of the PleC autokinase in vivo. Based on this we propose that PleC, DivK and PleD define a pathway required for c-di-GMP mediated pole morphogenesis during C. crescentus cell differentiation.

DivK Stimulates PleC and DivJ Kinase Activity In Vivo

An attractive model to explain how DivK activates PleC autokinase during the G1-to-S transition combines the in vitro and in vivo data presented above with the spatial dynamics of DivK during the cell cycle (Jacobs et al., 2001; Matroule et al., 2004). In such a model the appearance of DivJ during swarmer cell differentiation activates DivK and mediates its localization to the PleC-occupied pole. An increase of DivK levels at this site would then activate the PleC autokinase and trigger c-di-GMP production through PleD phosphorylation. If so, one would predict that in differentiating cells the autokinase activities of both PleC and DivJ are dependent on DivK. To test this we measured the in vivo levels of PleC~P and DivJ~P in dependence of DivK. PleC~P was readily detectable in wild-type cells but was greatly reduced in divK_D90G mutant cells at the permissive conditions (Figure 5D). Because DivK_D90G was still able to stimulate DivJ in vitro (Figure 4), we generated a divK null mutant strain to test the influence of DivK on DivJ activation in vivo. The divK gene was inactivated in the ctrA401 mutant background, which shows reduced activity of the cell cycle master regulator CtrA (Quon et al., 1998). Consistent with the observation that ctrA mutant alleles can restore viability of divK null mutants (Wu et al., 1998), a divK::Ω ctrA401 double mutant was fully viable. As shown in Figure 5E, DivJ~P levels were significantly reduced in ΔdivK ctrA401 double mutant as compared to the isogenic ctrA401 single mutant. These results support the conclusion that DivK is required to stimulate PleC and DivJ autokinase activities in vivo.

The model outlined above predicts that PleC and DivJ colocalize at the differentiating pole at the time when PleD is activated and the holdfast is synthesized. To test this we analyzed the localization patterns of PleC and DivJ in synchronized cells expressing a PleC-YFP and a DivJ-tDimer2 fusion (Matroule et al., 2004; Wheeler and Shapiro, 1999). PleC-YFP is localized in swarmer cells but disperses from the emerging stalked pole between 30 and 45 min after the initiation of development when cellular levels of PleC drop (Figures 6C and S9). At later stages of the cell cycle, PleC-YFP concentrates at the pole opposite the stalk (Figure 6C). In contrast, in newborn swarmer cells DivJ levels are low and DivJ-tDimer2 is absent at the flagellated pole. In parallel with an increase of DivJ, DivJ-tDimer2 concentrates at the emerging stalked pole already after 15 min (Figures 6C and S9). As a result, DivJ-tDimer2 and PleC-YFP coexist at the differentiating pole during a time window that overlaps with PleD-mediated c-di-GMP production and holdfast synthesis (Figures 6A–6C and S9). Localization studies with PleC-YFP and DivK-CFP as well as with DivJ-YFP and DivK-CFP indicated that DivK-CFP is also present at the differentiating cell pole during the same time window (data not shown).

DISCUSSION

Our data establish posttranslational feedback loops as important elements of the regulatory machinery that determines cell fate in C. crescentus. At the heart of this regulatory mechanism is the diffusible single domain response regulator DivK, which is required for cell cycle progression and the establishment of asymmetry (Hecht et al., 1995; Hung and Shapiro, 2002; Jacobs et al., 2001; Lam et al., 2003; Matroule et al., 2004). Two antagonistic players, PleC and DivJ, localized at opposite poles of the predivisional cell, determine the phosphorylation status and polar localization of DivK during the cell cycle (Jacobs et al., 2001; Lam et al., 2003; Matroule et al., 2004; Wheeler and Shapiro, 1999) (Figure 7). DivK implements cell cycle control through CtrA (Biondi et al., 2006; Hung and Shapiro, 2002; Wu et al., 1998), a DNA-binding response regulator that is controlled by phosphorylation (Biondi et al., 2006; Domian et al., 1997; Quon et al., 1996), but the molecular basis for this interaction is unknown. Based on our data we propose that DivK primarily acts by modulating the activities of DivJ and PleC, and possibly other sensor histidine kinases involved in C. crescentus polarity and cell cycle control (Ohta and Newton, 2003; Wheeler and Shapiro, 1999). DivK~P negatively controls the CckA-ChpT pathway that regulates CtrA phosphorylation and stability during the cell cycle (Biondi et al., 2006). However, as there is no evidence that this interaction is direct, important regulatory components of the DivK network might still be missing. A potential candidate for such a component is the unorthodox sensor kinase DivL (Wu et al., 1999). Not only were divL alleles isolated in a genetic screen together with divK, divJ, pleC, and pleD (Ohta et al., 1992; Sommer and Newton, 1991), but the DivL kinase also localizes to the cell poles (Sciochetti et al., 2005). Because in vitro experiments failed to identify additional kinases stimulated by DivK_D53N (E.G.B. and M.T.L., unpublished data), we propose that DivK is a specific regulator of DivJ, PleC, and possibly DivL.

In vitro and in vivo studies suggest that DivK is a potent activator of the PleC kinase that changes the cell’s developmental program by switching PleC from its default phosphatase into an autokinase mode during the G1-to-S transition. These data are in agreement with the observed synergistic effect between PleC and DivK (Hecht et al., 1995). The only direct targets of bifunctional PleC identified so far are DivK and PleD, with DivK~P being a substrate of the PleC phosphatase in swarmer cells (Matroule et al., 2004; Wheeler and Shapiro, 1999) and PleC~P acting as a phosphodonor for PleD in stalked cells. PleC, once activated by DivK, might also contribute to DivK phosphorylation in stalked and predivisional cells. But if DivK~P is a substrate for the PleC phosphatase in swarmer cells and then switches PleC into the autokinase mode during differentiation, what triggers this transition? DivJ, the main kinase of DivK is present at very low concentrations in swarmer cells (Jacobs et al., 2001; Wheeler and Shapiro, 1999). As DivJ levels rise during G1-to-S, DivJ localizes to the differentiating pole. At this stage DivJ-mediated phosphorylation localizes DivK~P to the cell pole, where it forces PleC into the autokinase mode (Figure 7). Because DivK also activates DivJ, the two proteins might form a positive feedback loop that leads to a strong stimulation of DivJ in stalked cells. Upregulation of the DivJ autokinase by DivK might be particularly important to quickly and robustly switch PleC from the phosphatase to the kinase mode during the G1-to-S transition and in the predivisional cell (Figure 7). Conversely, the feedback loops can also explain how the system is quickly reset as cells enter G1. The dominant role of DivJ in DivK phosphorylation and polar localization suggests that its loss results in the dissolution of DivK from the poles (Jacobs et al., 2001; Matroule et al., 2004). Therefore, exclusion of DivJ from the swarmer compartment during cytokinesis will reduce DivK~P levels and decrease DivK concentration at the flagellated pole below a threshold level required for PleC autokinase activation. As a result, PleC switches back into its phosphatase mode, thereby installing the swarmer cell program. Thus, the feedback loops described here might give rise to sharp and robust developmental transitions. The recurring spatial mixing and separation of the default phosphatase PleC and its dominant inhibitor, the DivJ-DivK kinase loop, would contribute to the oscillation of the system between stalked and swarmer programs (Figure 7A). The role of DivK is to facilitate long-range communication between the asymmetric DivJ and PleC antagonists and coordinate their activities.

This model makes the critical assumption that localization of DivK~P to the cell poles increases DivK concentration at this subcellular site above a threshold level required for the activation of the PleC and DivJ kinases. The model predicts that polar localization of DivJ and PleC is critical for the DivK-mediated feedback loops to operate. Indeed, mutants lacking the SpmX muramidase fail to recruit DivJ to the emerging stalked pole and are unable to activate DivJ kinase and spark DivK phosphorylation in stalked cells (Radhakrishnan et al., 2008). Conversely, mutants lacking PodJ fail to localize PleC to the flagellated cell pole and show distinct pole development defects (Viollier et al., 2002; Wang et al., 1993). Intriguingly, podJ and spmX expression is controlled reciprocally by activated CtrA~P (Figure 7B) (Crymes et al., 1999; Radhakrishnan et al., 2008) opening up the possibility that DivJ and PleC, through DivK, control their own spatiotemporal behavior.

What could be the molecular mechanism through which DivK stimulates the autokinase activities of PleC and DivJ? Efficient stimulation of PleC and DivJ autophosphorylation was obtained at a 1:1 ratio of kinase and DivK_D53N and a soluble kinase fragment containing only the DHp (dimerization and histidine phosphotransfer) and CA (catalytic and ATP-binding) domains (Parkinson and Kofoid, 1992; Stock et al., 2000). Thus, DivK modulates autokinase activity by binding to the catalytic core fragment. This is in accordance with results from a yeast two-hybrid screen that indicated DivK binding to a shared 66 amino acid sequence forming the core of the DHp domain of several kinases including PleC and DivJ (Ohta and Newton, 2003). Because DivK did not affect the oligomerization state of PleC and DivJ, interference with kinase dimerization seems unlikely. The observed drop of basal level activity of the PleD diguanylate cyclase under conditions where autophosphorylation is absent or inefficient, indicates the formation of a nonproductive ternary complex that engages PleD in an inactive monomeric form (Paul et al., 2007). Possibly, DivK and PleD monomers interact with the same kinase dimer and DivK bound to one kinase protomer can activate autophosphorylation and phosphotransfer to PleD via the other protomer. The failure of the DivK_D90G mutant to stimulate PleC kinase activity in vitro and in vivo could result from an altered interaction with the DHp domain of the kinase. This view is consistent with the position of Asp90 at the N terminus of the DivK receiver domain helix α4. Because this region undergoes structural changes upon phosphorylation (Robinson et al., 2000) and makes specific contacts with the DHp domain (Zapf et al., 2000) it is a good candidate for the interaction surface that mediates the phosphatase-kinase switch of PleC.

In conclusion, our results indicate a role for the abundant class of single domain response regulators in spatially interconnecting different components of the two-component signal transduction circuitry. While recent system level approaches indicated that a one-to-one relationship between histidine kinases and their cognate response regulators prevents unwanted cross-talk (Skerker et al., 2005; Yamamoto et al., 2005) our results add a new level of complexity to this sensory network demonstrating the possibility of widespread inter-connections between apparently insolated signaling systems through retrograde information transfer from response regulators to cognate kinases.

EXPERIMENTAL PROCEDURES

Strains, Plasmids, and Media

The bacterial strains and plasmids used in this study are shown in Table S1. Caulobacter crescentus strains were grown in peptone yeast extract (PYE), or in minimal glucose media (M2G, [Ely, 1991]). Newborn SW cells were isolated by Ludox or Percoll gradient centrifugation (Jenal and Shapiro, 1996), and released into fresh M2G medium. For conjugal transfer into C. crescentus, Escherichia coli strain S17-1 was used as donor. The divK_D90G mutation (Sommer and Newton, 1991) was introduced into CB15ΔpleC (UJ731) by allelic exchange. To generate the divK null mutant, the divK::Ω allele was transduced from strain CJ403 (Lam et al., 2003) into the C. crescentus ctrA401 mutant strain (Quon et al., 1996). The exact procedures of strain and plasmid construction are available on request. Unless stated otherwise, pooled data consist of at least three independent experiments and are represented as mean ± standard deviation (SD).

Attachment Assays and Holdfast Staining

For attachment assays, logarithmically growing C. crescentus cells were diluted 1:30 in PYE, and cultivated for 24 hr in 96-well microtiter plates at 200 rpm on a rocking platform. Attachment of cells to the polystyrene surface was quantified as described (Levi and Jenal, 2006). For attachment assays of synchronized populations, cell aliquots were transferred to microtiter plates at the time points indicated, and allowed to bind to the plastic surface for 15 min at room temperature before staining. C. crescentus holdfast was stained with a mixture of Oregon Green-conjugated wheat-germ agglutinin (0.2 mg/ml) and Calcofluor White (0.1 mg/ml) and was visualized by fluorescence imaging on an Olympus AX70 microscope with a Hamamatsu C4742-95 digital camera (Levi and Jenal, 2006). Images were recorded and processed with Improvision Openlab and the Photoshop CS v8.0 (Adobe, CA) software packages.

Fluorescence Microscopy

For fluorescence imaging bacteria were placed on a microscope slide coated with 2% agarose dissolved in water. An Olympus IX71 microscope equipped with an UPlanSApo 100×/1.40 Oil objective (Olympus, Germany) and a cool-SNAP HQ CCD camera (Photometrics, AZ, United States) were used to take differential interference contrast (DIC) and fluorescence photomicrographs. YFP (Ex 500/20 nm, EM 535/30 nm) Rhodamine filter sets (Ex 555/28 nm, EM 617/73 nm) were used. DIC pictures were taken with 0.15 s and fluorescence pictures with 1.0 s exposure time. Images were processed with soft-WoRx v3.3.6 (Applied Precision, WA) and Photoshop CS v8.0 softwares.

In Vivo Phosphorylation

In vivo phosphorylation experiments were performed as described previously (Domian et al., 1997; Jacobs et al., 2001) with the following modifications. The strains were grown in M5G minimal medium supplemented with 1 mM glutamate until an optical density at 660 nm of 0.15, collected by centrifugation and resuspended in filtered culture medium to an optical density of 0.3. Cells were labeled at 30°C for 5 min with 100 μCi [³²P]H₃PO₄ or 30 μCi [³²P]ATP, and after lysis immunoprecipitated with 10 μl anti-PleC, anti-DivK, or anti-DivJ serum. Radiolabeled proteins were separated on a 10% polyacryamide gel and visualized with a phosphoimager (Molecular Dynamics, GE Healthcare, NJ).

Nucleotide Analysis

For the analysis of c-di-GMP levels in synchronized C. crescentus cultures (OD₆₆₀ 0.4), nucleotides were extracted with 1 M formic acid. Lyophilized samples were analyzed with a 125/4 Nucleosil 4000-1 PEI column (Macherey-Na-gel, Germany) on SMART- (GE Healthcare, NJ) or Pro Star HPLC-Systems (Varian, CA). The nucleotides were applied to the column dissolved in buffer A (7 mM KH₂PO₄ [pH 4]), and eluted with a gradient of buffer B (0.5 M KH₂PO₄, 1 M Na₂SO₄ [pH 5.5]) at a flow rate of 50 μl/min. Concentrations were determined by comparison with a standard of chemically synthesized c-di-GMP.

Expression and Purification of Proteins

E. coli cells carrying the respective expression plasmid were grown in LB medium with ampicillin (100 μg/ml), and expression was induced by adding either arabinose (final concentration of 0.2%) or IPTG (final concentration of 0.4 mM). After harvesting by centrifugation, the cells were resuspended in TN-buffer (50 mM Tris-HCl [pH 8.0], 500 mM NaCl, 5 mM β-mercaptoethanol), lysed by passage through a French pressure cell, and clarified by centrifugation. The supernatant was loaded onto Ni-NTA affinity resin (QIAGEN, Germany), washed with TN-buffer, and eluted with an imidazol-gradient. PleD, DivK, a catalytic fragment of DivJ’ (C-terminal 296 amino acids), and a catalytic PleC’ fragment containing the complete cytoplasmic domains (540 amino acids) remained soluble and were purified under native conditions. Two additional catalytic PleC’ fragments (containing the C-terminal 313 or 489 amino acids, respectively) were solubilized from inclusion bodies and renatured after purification (Hecht et al., 1995; Paul et al., 2004). Proteins were examined for purity by SDS-PAGE and fractions containing pure protein were pooled and dialyzed. DivK was concentrated by precipitation with 60% saturated ammonium sulfate (pH 7.0) prior to dialysis; the kinase fragments and PleD were concentrated using Amicon Ultra- or Microcon- Centrifugal Filter Units (Millipore, MA). Analytical size exclusion chromatography was performed with a Superdex 200 column on an ÄKTApurifier (GE Healthcare) system at a flow rate of 0.5 ml/min. Concentrations and molecular weights of proteins were determined by an online refractometer (Optilab rEX, Wyatt Technology) and a miniDAWN light scattering instrumentation (Wyatt Technology, CA). Untagged DivK for in vitro phosphatase assays was obtained by cleaving His-DivK with Thrombin (Novagen, WI, United States) for 1 hr at 25°C, followed by preparative size exclusion chromatography with a Superdex 75 column on an ÄKTApurifier (GE Healthcare, NJ) at a flow rate of 1.5 ml/min.

Enzymatic Assays

Diguanylate cyclase assays were adapted from procedures described previously (Paul et al., 2007, 2004). The standard diguanylate cyclase reaction mixtures contained 50 mM Tris-HCl (pH 7.8), 250 mM NaCl, 10 mM MgCl₂ in 50 μl volume and was started by the addition of a mixture of 100 μM GTP/[α-³³P]GTP (PerkinElmer; 0.01 μCi/μl). All kinase reactions were supplemented with 25 mM KCl. To calculate the initial velocity of product formation, aliquots were withdrawn at regular time intervals and the reaction was stopped with an equal volume of 50 mM EDTA (pH 6.0). Reaction products (2 μl) were separated on polyethyleneimine-cellulose plates (Macherey-Nagel) in 1.5 M KH₂PO₄/5.5 M (NH₄)₂SO₄ (pH 3.5) in a 2:1 ratio. Plates were exposed to a phosphorimager screen, and the intensity of the various radioactive species was calculated by quantifying the intensities of the relevant spots using the Image-Quant software (Molecular Dynamics). Measurements were always restricted to the linear range of product formation. In vitro kinase and phosphatase assays were performed as described earlier (Hecht et al., 1995; Matroule et al., 2004; Paul et al., 2004; Skerker et al., 2005). For kinase assays the proteins were incubated at 25°C for 15 min in phosphorylation buffer (50 mM Tris-HCl at pH 7.8, 25 mM NaCl, 25 mM KCl, 5 mM MgCl₂) containing 5 μCi [γ-³²P]ATP (Amersham Biosciences, GE Healthcare, NJ), unless stated otherwise. Reactions were stopped with 4× SDS-PAGE sample buffer (250 mM Tris-HCl at pH 6.8, 40% glycerol, 8% SDS, 2.4 M β-mercaptoethanol, 0.06% bromophenol blue, 40 mM EDTA), and ³²P-labeled proteins were separated by electrophoresis on 10% SDS-PAGE gels followed by autoradiography on a phosphor-imager screen. DivK-P for phosphatase assays was purified from contaminating His-tagged PleC and DivK_D53N by two rounds of batch purification with Ni-NTA resin (QIAGEN, Germany), followed by gel filtration for the removal of ATP.