Precise timing of transcription by c-di-GMP coordinates cell cycle and morphogenesis in Caulobacter

Bacteria adapt their growth rate to their metabolic status and environmental conditions by modulating the length of their G1 period. Here we demonstrate that a gradual increase in the concentration of the second messenger c-di-GMP determines precise gene expression during G1/S transition in Caulobacter crescentus. We show that c-di-GMP stimulates the kinase ShkA by binding to its central pseudo-receiver domain, activates the TacA transcription factor, and initiates a G1/S-specific transcription program leading to cell morphogenesis and S-phase entry. Activation of the ShkA-dependent genetic program causes c-di-GMP to reach peak levels, which triggers S-phase entry and promotes proteolysis of ShkA and TacA. Thus, a gradual increase of c-di-GMP results in precise control of ShkA-TacA activity, enabling G1/S-specific gene expression that coordinates cell cycle and morphogenesis.

T he bacterial cell cycle is divided into three periods: after cell division and before initiation of chromosome replication (B or G1); chromosome replication (C or S); and cell division (D or G2) 1 . Since chromosome replication and cell division (C and D periods) remain constant over a wide range of growth rates 2,3 , the step committing cells to initiate chromosome replication largely determines bacterial proliferation rates. Bacteria like Escherichia coli or Bacillus subtilis can increase their growth rate by bypassing the B period and by initiating replication multiple times per division cycle 2 . In contrast, Caulobacter crescentus strictly separates its cell cycle stages. An asymmetric division generates a sessile stalked (ST) cell, which directly reenters S-phase, and a motile swarmer (SW) cell that remains in G1 for a variable time depending on nutrient availability 4,5 . Coincident with G1/S transition, motile SW cells undergo morphogenesis to gain sessility (Fig. 1a). But what determines the length of G1 in this organism has remained unclear.
In C. crescentus, replication initiation is regulated by the cell cycle kinase CckA. CckA is a bifunctional enzyme that acts as a kinase for the replication initiation inhibitor CtrA in G1, but switches to being a phosphatase in S phase, resulting in the inactivation of CtrA and clearance of the replication block 6 (Fig. 1a). The CckA switch is governed by two response regulators, DivK and PleD. While DivK controls CckA activity through protein-protein interactions [7][8][9] , PleD is a diguanylate cyclase, which is responsible for the characteristic oscillation of cdi-GMP during the cell cycle 10 . The concentration of c-di-GMP, below the detection limit in G1, increases during G1/S to reach peak levels at the onset of S-phase 11 where it allosterically stimulates CckA phosphatase 9 (Fig. 1a).
Activation of DivK and PleD is directed by the SpmX scaffolding protein, which accumulates during G1/S and recruits DivJ, the kinase of DivK and PleD, to the incipient stalked cell pole (Fig. 1a) 12 . This makes SpmX accumulation the earliest known event to trigger S-phase entry. However, it is unclear how spmX expression is timed during G1/S to initiate replication. Transcription of spmX is regulated by the response regulator TacA, which in its phosphorylated form also induces the expression of a large set of genes required for SW-to-ST cell morphogenesis 12,13 . TacA is activated via a multistep phosphorylation cascade (Fig. 1b) consisting of the sensor histidine kinase ShkA, and the phosphotransferase protein ShpA 12,13 . ShkA is a multidomain protein kinase with a catalytic domain (CA) that binds ATP and transfers a phosphate via the conserved His residue of the dimerization histidine-phosphotransfer domain (DHp) to a conserved Asp residue of the C-terminal receiver domain (REC2) 12,13 (Fig. 1b). TacA controls a large set of target genes including staR, which encodes a transcription factor important for stalk elongation. Thus, this phosphorelay system coordinates replication initiation with morphogenesis. In accordance, cells lacking TacA are elongated, stalk-less and show altered DNA profiles compared with wild type 13,14 . However, the signal activating ShkA has remained unknown (Fig. 1b).
Here we demonstrate that an upshift in levels of the second messenger c-di-GMP determines precise gene expression during G1/S. A combination of genetic, biochemical and cell biology data revealed that c-di-GMP controls the ShkA-TacA pathway by directly binding to the ShkA sensor histidine kinase and strongly stimulating its kinase activity. We demonstrate that binding of cdi-GMP to a pseudo-receiver domain of ShkA abrogates ShkA auto-inhibition leading to the activation of the ShkA-TacA pathway. C-di-GMP-mediated activation of ShkA and the subsequent c-di-GMP-mediated proteolysis of ShkA together define a window of ShkA activity during the cell cycle. This window is sharpened to a narrow, G1/S-specific period by TacA degradation, which precedes the degradation of ShkA. Thus, the exact timing of G1/S-specific gene expression results from the consecutive c-di-GMP-mediated activation and degradation of ShkA-TacA phosphorylation components.

C-di-GMP stimulates ShkA kinase activity in vivo and in vitro.
A C. crescentus strain lacking all diguanylate cyclases (cdG 0 strain), which was generated to remove all traces of the second messenger c-di-GMP, shows strong developmental and morphological abnormalities with mutant cells being irregularly shaped, elongated and lacking all polar appendages 11 . Because stalk biogenesis depends on an active ShkA-TacA phosphorelay (Fig. 1b) 13 , we tested if c-di-GMP controls the ShkA-TacA pathway. Expression of TacA D54E , a phospho-mimetic form of TacA 13 , restored stalk biogenesis, transcription of the TacA targets staR and spmX, as well as localization of a SpmX-mCherry fusion to the stalked pole (Fig. 1c, d; Supplementary Fig. 1). spmX and staR transcription was also restored when c-di-GMP was reintroduced by expression of the heterologous diguanylate cyclase (DGC) dgcZ from E. coli in the cdG 0 background or in a strain that lacks all DGCs and phosphodiesterases (PDE) (rcdG 0 strain) ( Fig. 1d; Supplementary Fig. 1). Phos-tag PAGE analysis revealed that TacA and ShkA were unphosphorylated in the cdG 0 strain. This phosphorylation was restored upon expression of the constitutively active DGC PleD* 15 (Fig. 1e). In contrast, expression of the heterologous phosphodiesterase PA5295 from Pseudomonas aeruginosa, which also lowers c-di-GMP levels, in wildtype cells reduced TacA and ShkA phosphorylation (Fig. 1f). Stalk formation is independent of the ShkA-TacA pathway when cells are grown in a low-phosphate medium 9 . In agreement with this, a strain lacking c-di-GMP formed stalks at low-phosphate concentrations, but failed to localize DivJ ( Supplementary Fig. 1). Many cells displayed bipolar stalks, indicating that at lowphosphate c-di-GMP controls stalk positioning through a pathway distinct from ShkA-TacA. Altogether, these results indicate that c-di-GMP acts upstream of and is required for the activity of the ShkA-TacA phosphorelay.
In vitro phosphorylation assays with purified ShkA or with all components of the phosphorelay demonstrated that c-di-GMP strongly and specifically stimulates ShkA autokinase activity (Fig. 1g, h; Supplementary Fig. 2a). Moreover, purified ShkA bound c-di-GMP with a K D in the sub-micromolar range ( Supplementary Fig. 2b). Altogether, we conclude that c-di-GMP activates the ShkA-TacA pathway by directly binding to and allosterically stimulating ShkA kinase.
C-di-GMP activates ShkA by binding to its pseudo-receiver domain. We next sought to dissect the mechanism of c-di-GMPmediated ShkA activation. We devised a genetic selection (see "Methods") to isolate shkA mutations that restored spmX expression in a rcdG 0 background. Independent mutations were identified in two residues (D369, R371) within a short stretch of three highly conserved amino acids in the linker region between REC1 and REC2 (hereafter referred to as the "DDR" motif) ( Fig. 2a; Supplementary Fig. 3). These results identified the REC1-REC2 linker region as a critical determinant of ShkA activation.
We further characterized the D369N variant both in a wild-type background and in a strain expressing a stabilized version of TacA (TacA DD ), and confirmed that TacA phosphorylation levels and spmX transcription were restored to wild-type levels in the rcdG 0 background harboring the shkA D369N allele (Fig. 2b, c). Also, purified ShkA D369N protein showed strong autophosphorylation activity even in the absence of c-di-GMP (Fig. 2d). The mutant retained its ability to bind c-di-GMP in vitro ( Supplementary  Fig. 2c)  both in vivo (Fig. 2b) and in vitro ( Supplementary Fig. 2d). Thus, mutations in the DDR motif uncouple ShkA activity from c-di-GMP without interfering with c-di-GMP binding. We used the c-di-GMP independence of the D369N ShkA variant to identify residues involved in c-di-GMP binding. We reasoned that mutations in ShkA specifically interfering with cdi-GMP-mediated activation should be rescued when combined with D369N. In contrast, mutants causing more general kinase defects (ATP-binding, phosphotransfer, protein stability etc.) would not be recuperated when combined with D369N. An alignment of ShkA orthologs from C. crescentus and related organisms revealed a total of 25 candidate residues for c-di-GMP binding distributed throughout the entire protein ( Supplementary  Fig. 3). Of all the mutants that severely affected ShkA activity ( Supplementary Fig. 4a, b) and that were rescued in vivo (Fig. 2e) and in vitro (Fig. 2f) when combined with D369N, only one, Y338A, in REC1, interfered with c-di-GMP binding (Fig. 2g, h), while the other mutations likely affect c-di-GMP-mediated conformational changes required for activation.
Thus, Y338 in REC1 is likely a part of the c-di-GMP binding site. Indeed, the purified REC1 domain alone was able to bind cdi-GMP, although with lower affinity than full-length ShkA (Fig. 2i). NMR spectroscopy with REC1 revealed a fold reminiscent of prototypical REC domains, except that helix α3 is not present irrespective of the presence or absence of c-di-GMP ( Supplementary Fig. 4c-f). Upon the addition of c-di-GMP, REC1 shows chemical shift perturbations (CSPs), which cluster on the β5-α5 surface ( Fig. 2j; Supplementary Fig. 4g). Importantly,  these residues include Y338 and are well conserved ( Fig. 2k; Supplementary Fig. 3). Some of the other residues implicated in cdi-GMP binding by NMR were also important for ShkA activity in vivo ( Supplementary Fig. 4h). Altogether, these experiments revealed a REC1 pseudo-receiver domain of ShkA as the primary docking site for c-di-GMP, and identified residues directly involved in c-di-GMP binding and c-di-GMP-mediated activation of ShkA. In silico analysis revealed that pseudo-receiver domains, although often not annotated, are widespread among histidine kinases ( Supplementary Fig. 5, Supplementary Data 1). We propose that pseudo-receiver domains have lost their original phosphotransfer function but during evolution have adopted signal perception functions and may thus represent a large class of so-far unrecognized kinase input domains.
Similar to the two identified DDR mutants, the kinase activity of ShkA is also uncoupled from c-di-GMP when conserved residues of REC2 are mutated ( Supplementary Fig. 2e) or when the C-terminal REC2 domain is removed from the catalytic core (DHp, CA, and REC1 domain) (Fig. 1h). Thus, the REC2 domain and the DDR linker motif inhibit ShkA autophosphorylation, maintaining it in an inactive state. We propose that c-di-GMP binding to REC1 overrides this auto-inhibition and activates the enzyme.
C-di-GMP defines a narrow window of ShkA activity during G1/S. The c-di-GMP-independent variants of ShkA allowed us to more carefully investigate the role of ShkA in the temporal control of events during G1/S. Introduction of the shkA D369N allele into the rcdG 0 strain restored SpmX protein levels, stalk biogenesis, DivJ localization to the incipient stalked pole, and normal cell morphology (Fig. 3a, b; Supplementary Fig. 6a). Of note, shkA D369N failed to restore G2-specific processes in the rcdG 0 strain like motility and type IV pili assembly (Supplementary Fig. 6b, c). Restoration of cell morphology and DivJ localization was entirely dependent on SpmX (Fig. 3b), arguing that c-di-GMP and ShkA contribute to cell cycle progression and morphogenesis via spmX expression control. Consistent with this notion, the shkA D369N allele rescued the synthetic-lethal phenotype of a strain lacking c-di-GMP and suffering from reduced DivK levels 9 in a SpmX-dependent manner ( Supplementary  Fig. 6), indicating that ShkA influences the CckA kinase/phosphatase balance. A strain harboring the shkA D369N allele prematurely produced SpmX already in newborn G1 cells (Fig. 3c). Accordingly, the shkA D369N mutant exited G1 prematurely, as indicated by the strong reduction of G1 cells (Fig. 3d). Likewise, cells failed to properly arrest in G1 after entry into the stationary phase (Fig. 3d, e). Growth rate of the shkA D369N mutant was not increased in either PYE or M2G medium compared with wild type, suggesting that the overall length of the cell cycle (combined G1, S, and G2 phases) is not or only marginally affected (Supplementary Fig. 6). These results support a role of c-di-GMP, and the ShkA-TacA pathway, in the timely execution of the G1/S transition.
TacA was previously shown to be degraded by the ClpXP protease, a process that requires the adapter proteins RcdA and CpdR and depends on the C-terminal Ala-Gly degradation motif of TacA 16 . ShkA harbors the same C-terminal degradation signal and was also degraded upon S-phase entry about 10 min after the removal of TacA (Fig. 3f). When the C-terminal Ala-Gly motif was replaced by Asp residues (ShkA DD ), ShkA and ShkA~P were stabilized and prevailed throughout the cell cycle (Fig. 3f). Sequential degradation of TacA and ShkA may be explained by their differential requirements for protease adapters. While TacA degradation by ClpXP depends on CpdR and RcdA 16 , ShkA degradation also requires PopA, the third member of the adapter hierarchy regulating ClpXP protease activity during the C. crescentus cell cycle (Fig. 3g, h). Because PopA needs to bind cdi-GMP to act as a protease adapter 17 , the ShkA-TacA pathway is confined to G1/S by the sequential c-di-GMP-dependent activation and c-di-GMP-mediated degradation of the ShkA kinase.
The ShkA-TacA pathway limits gene expression to G1/S. To carefully assess the contribution of ShkA activation and degradation for the temporal control of spmX, the spmX promoter was fused to the fluorescent protein Dendra2. The photoconvertible properties of Dendra2 18 allowed determining both "ON" and "OFF" kinetics of spmX promoter activity during the cell cycle ( Fig. 4b; Supplementary Fig. 7a). These experiments revealed that spmX promoter activity peaks during G1/S roughly 15-30 min after passing through the predivisional stage (Fig. 4b). The spmX promoter was active in cells progressing through G1/S, but not in newborn ST progeny that reenter S-phase immediately at the end of the asymmetric cell cycle (Fig. 4a, c; Supplementary Fig. 7a-c). Stage-specific expression of spmX required a combination of c-di-GMP oscillations during the cell cycle and TacA degradation: expression of combinations of dgcZ, a gene encoding a highly active DGC from E. coli 19 , and of tacA DD (encoding stable TacA 16 ) and shkA D369N (encoding constitutive ShkA) alleles resulted in a gradual loss of G1/S-specific spmX expression  Supplementary Fig. 7c). These experiments provided direct evidence that the activity of the ShkA-TacA pathway is strictly limited to G1/S and SW progeny that need to passage through the G1 phase of the cell cycle.
To investigate the importance of limiting the ShkA-TacA pathway to G1/S for accurate cell cycle progression, we examined the consequences of ShkA-TacA dysregulation using stable (TacA DD , ShkA DD ), or constitutively active (ShkA D369N ) variants or combinations thereof. All alleles increased overall spmX expression and showed additive effects when combined ( Supplementary Fig. 8a). Cell division and cell morphology were normal in all strains that either retained TacA degradation or cdi-GMP-mediated ShkA activity control. However, when mutations that stabilize ShkA or TacA were combined with a   mutation constitutively activating ShkA, strains showed strong cell division and morphology aberrations, effects that were strictly dependent on an intact copy of spmX ( Fig. 4d; Supplementary Fig. 8b). Thus, dysregulation of the SpmX morphogen leads to aberrant cell morphogenesis. Together, these experiments provide evidence that dysregulation of ShkA-TacA activity leads to severe cell cycle and morphological defects.
The diguanylate cyclase PleD activates ShkA during G1/S. The above data support a model in which an upshift of c-di-GMP stimulates ShkA kinase activity thereby initiating the G1/S-specific genetic program (Fig. 5a). If so, the G1/S transition should be kick-started by one of the C. crescentus diguanylate cyclases. Screening a spmX-lacZ reporter strain for transposon insertions with reduced lacZ expression identified mutations in pleD.  Accordingly, a ΔpleD deletion strain showed strongly reduced spmX expression (Fig. 5b). While a second diguanylate cyclase enzyme, DgcB, had a more modest effect, spmX promoter activity was almost completely abolished in a ΔpleD ΔdgcB double mutant, akin to a strain lacking c-di-GMP (Fig. 5b). Thus, PleD is the major diguanylate cyclase driving the G1/S-specific transcriptional program.
PleD phosphorylation is regulated by the antagonistic ST cellspecific kinase DivJ and the SW cell-specific phosphatase PleC (Figs. 1a and 5a) 8 . However, PleC but not DivJ, was required for spmX expression, and spmX expression was restored to normal levels in a strain lacking both PleC and DivJ (Fig. 5b). This indicated that neither DivJ nor PleC is responsible for the initial activation of PleD and for ShkA stimulation, and that the role of PleC is likely indirect. PleC phosphatase was previously shown to reduce DivK phosphorylation leading to the activation of CtrA. Because tacA is a direct target of CtrA, it was proposed that in a ΔpleC mutant, spmX expression is impaired because TacA fails to accumulate 12 . Expression of tacA and spmX indeed required the PleC phosphatase (Fig. 5c, d). However, when the pleC deletion was combined with the constitutive shkA D369N allele, SpmX protein levels and polar localization of DivJ, but not tacA expression, were fully restored (Fig. 5c, e; Supplementary Fig. 9), arguing that TacA levels are not the limiting factor in cells lacking PleC. Rather, ShkA activity and its stimulation by c-di-GMP are switched off in the ΔpleC mutant. Indeed, a constitutively active variant of PleD, PleD* 15 , restored SpmX protein levels in the ΔpleC mutant, an effect that was entirely dependent on ShkA (Fig. 5f). These results suggest that c-di-GMP levels are limiting in a ΔpleC mutant. In line with this, LC-MS measurements showed that the ΔpleC mutant has significantly reduced c-di-GMP levels as compared with the isogenic pleC + strain (9.5 ± 1.3 µM vs. 16 The finding that the ShkA-TacA pathway is OFF in the ΔpleC mutant because c-di-GMP concentrations are limiting, together with the observation that PleD serves as the main c-di-GMP donor for ShkA activation, argues for the existence of an as yet unidentified PleD kinase, the expression of which will likely depend on the PleC-CckA-CtrA cascade. We speculate that activation of this kinase and its downstream target PleD represents a key event in the decision of C. crescentus to exit G1 (Fig. 5a).

Discussion
C. crescentus SW cells are born with low levels of c-di-GMP 11,20 . This is imposed by two cell type-specific regulators, the phosphodiesterases PdeA 21 and the phosphatase PleC, which maintains the diguanylate cyclase PleD in its inactive, unphosphorylated form 8 (Fig. 5a). During the G1/S transition, PdeA is proteolytically removed 21 and PleD is activated by phosphorylation. This results in a gradual increase of c-di-GMP 11,20 , which leads to a series of accurately timed events prompting exit from G1, cell morphogenesis and entry into S-phase. First, ShkA is allosterically activated resulting in TacA phosphorylation and the expression of a large group of G1/S-specific genes that orchestrate the morphological restructuring of the motile SW cells into sessile ST cells 13,14 . At this stage, c-di-GMP levels are high enough to activate ShkA-TacA but have not reached the peak levels needed to trigger the CckA cell cycle switch and S-phase entry 8,9 . This leads to the execution of the morphogenetic program before cells commit to chromosome replication and division. The next step is then catalyzed by the expression of one of the G1/S-specific proteins, the morphogen SpmX, which is responsible for the polar sequestration and activation of DivJ 12 . This, in turn, leads to the production of more c-di-GMP via reinforced activation of PleD and, together with DivJ-mediated phosphorylation of DivK, switches CckA into a phosphatase and ultimately licenses replication initiation 7-9 (Fig. 5a). ShkA binds c-di-GMP with fiveto tenfold higher affinity than CckA 9,17 explaining how the ShkA-TacA pathway and the CckA switch can be sequentially activated. Thus, at least four kinases form a hierarchical cascade (HK→ShkA→ DivJ→CckA) that is responsible for the accurate temporal control of events during G1/S. The activity and timing of this cascade is coordinated by the second messenger c-di-GMP, a stepwise increase of which enforces consecutive cell cycle steps by modulating the activity of ShkA and CckA, respectively (Fig. 5a).
By contributing to the CckA phosphatase switch via SpmX and DivJ, the ShkA-TacA pathway initiates its own termination. The CckA phosphatase activates a protease adapter cascade that includes CpdR and PopA 16 and that leads to the consecutive degradation of TacA and ShkA by the ClpXP protease (Fig. 5a). This negative feedback constitutes an intrinsic, self-sustained timer that shuts down ShkA-TacA activity as soon as c-di-GMP has reached peak levels required to activate the CckA phosphatase and the PopA protease adapter, thereby irreversibly committing cells to S-phase. Because TacA not only controls genes involved in cell cycle progression but also regulates morphological restructuring of the motile SW cells into sessile ST cells 13,14 , accurate temporal control of this pathway secures the tight coordination between replicative and behavioral processes.
Our results show that the diguanylate cyclase PleD is largely responsible for the c-di-GMP upshift during G1-S transition. We postulate that the initial event leading to PleD activation during G1/S must be executed by a kinase other than DivJ or PleC, and that DivJ is part of a positive feedback loop that reinforces PleD Fig. 4 Oscillation of c-di-GMP and protein degradation limit TacA activity to G1/S. a Time-lapse fluorescence microscopy of C. crescentus wild-type cells expressing dendra2 from the spmX promoter. The time of photoconversion of Dendra2 (green to red) in the predivisional cell (PD) is indicated in the schematic of the C. crescentus cell cycle. A representative example of a dividing cell is shown below with separate green and red channels and with the ST cell pole of the PD cell marked (orange arrow). spmX expression (green) is restricted to the SW cell during G1/S transition (white arrows). The bar is 1 μm. b ON and OFF kinetics of spmX promoter activity during G1/S in wild-type cells harboring the spmX-dendra2 reporter plasmid. ON kinetics were determined as outlined in a with cells being photoconverted before division (0 min) and the fraction of ST (gray circles) and SW cells (green circles) with induced green fluorescence plotted over time. OFF kinetics were determined by the fraction of SW cells with induced green fluorescence after photoconversion at the indicated time points during the cell cycle (blue boxes). The overlap of the two curves (green area) defines the window of spmX promoter activity during the cell cycle. c Activity of spmX promoter in lineages of individual C. crescentus cells of different strains through three consecutive generations as indicated by the schematic on the left. The projected G1/S-specific expression of spmX is indicated in green. For each strain 10-16 late PD harboring the spmX-dendra2 reporter were photoconverted roughly 15 min before cell division and followed by time-lapse microscopy through three cell division events. Right: The fraction of SW and ST offspring (% of total cells analyzed) with active spmX promoter (green) was plotted over three generations with the x-axis representing generations 1-3. For experimental details and data analysis see "Methods". C-di-GMP levels were manipulated by Plac-driven dgcZ with 0.1 mM IPTG. EV, empty vector control. d Representative phase-contrast micrographs of strains carrying different shkA and tacA alleles encoding stabilized (shkA DD and tacA DD ) or c-di-GMP-independent (shkA D369N ) versions of the respective proteins. The scale bar represents 4 µm.
activity upon S-phase entry (Fig. 5a). Although the nature of this kinase is currently unknown, we speculate that its expression or activity is CtrA-dependent and that it plays a key role in orchestrating exit from G1 as it may not only serve to activate PleD and provide the initial boost of c-di-GMP but may also contribute to the activation of DivK, a factor required for the CckA kinase/phosphatase switch. The essential nature of DivK but not of DivJ, its only known activating kinase, argues for regulatory redundancy in the upstream components required to boost DivK phosphorylation during G1/S (Fig. 5a).
We postulate that in its default state, the ShkA kinase is inhibited by the C-terminal REC2 domain and that c-di-GMP binding liberates the kinase by interfering with this off-state conformation. A ShkA variant lacking the REC2 domain is active without c-di-GMP 13 . Similarly, mutations of residues important for REC2 function lead to constitutive, c-di-GMP-independent autokinase activity. Thus, the conserved DDR motif in the REC1-REC2 linker likely serves to lock ShkA in the inactive state when no c-di-GMP is present. REC2 may be closely tethered to REC1 in the inactive state through an interaction of the DDR linker motif with REC1, a conformation that may prevent the productive interaction of the catalytic CA with the DHp domain. We hypothesize that binding of c-di-GMP to REC1 interferes with this tethering, thereby liberating REC2 and facilitating the productive interaction between CA and DHp for autophosphorylation and eventually for phosphotransfer between DHp and REC2. This model is strongly supported by an accompanying structural analysis of ShkA 22 .
These findings demonstrate that degenerate REC domains, also called pseudo-receiver domains, can function as docking sites for small regulatory molecules. Hybrid histidine kinases with pseudoreceiver domains located between the CA and the phosphorylated receiver domain are widespread in bacteria and include the wellstudied virulence factors of the GacS/BarA family 23,24 or the global stress regulator RcsC 25,26 (Supplementary Fig. 5). For instance, the pseudo-receiver of RcsC can only be recognized by f Immunoblots of indicated strains probed with anti-SpmX and anti-MreB antibodies. pPleD* indicates a pMR20-based plasmid (pPA114-47) expressing the constitutively active, phosphorylation-independent PleD* allele. EV denotes the empty plasmid control (pMR20). Source data are provided as a Source Data file.  Strains and plasmids. Plasmids are listed in Supplementary Data 2. Oligonucleotides used for plasmid construction were purchased from Sigma and are listed in Supplementary Data 3. Plasmids were constructed as follows: pAK503: nptII without RBS and start codon was PCR-amplified from pAK405 using primers 8959/9495, the product was digested with BsrGI/XbaI and cloned into pAK501 cut with Acc65I/XbaI. pAK502-tacA: the spmX promoter and part of the codon sequence were PCRamplified from C. crescentus gDNA using primers 10015/10016, the product was digested with XbaI/KpnI and cloned into pAK502 digested with the same enzymes.
pAK502-spmX: the spmX promoter and part of the codon sequence were PCRamplified from C. crescentus gDNA using primers 8966/9064, the product was digested with XbaI/KpnI and cloned into pAK502 digested with the same enzymes.
pAK503-spmX: the spmX promoter and part of the codon sequence were PCRamplified from C. crescentus gDNA using primers 8966/9064, the product was digested with XbaI/KpnI and cloned into pAK503 digested with the same enzymes.
pNPTStet: tetA and tetR were PCR-amplified from pQF using primers 9552/ 9553, the product was digested with SpeI/NdeI and cloned into pNPTS138 digested with AseI/XbaI. pNPTStet-ampG: part of ampG located downstream of shkA was PCR-amplified from C. crescentus gDNA with primers 9554/9555, the product was digested with SpeI/KpnI and cloned into pNPTStet digested with the same enzymes. pET28a-shkA: shkA was PCR-amplified from C. crescentus gDNA with primers 9745/9746, the product was digested with NdeI/EcoRI and cloned into pET28a digested with the same enzymes.
The product was digested with NdeI/EcoRI and cloned into pQF digested with AseI/EcoRI. This shkA allele contains a second mutation, A451V.
pET28a-shkA-REC1: a fragment of shkA encoding REC1 was PCR-amplified from C. crescentus gDNA with primers 10862/10866, the product was digested with NdeI/KpnI and cloned into pUC18 digested with the same enzymes. A NdeI/EcoRI from the resulting plasmid was subcloned into pET28a digested with the same enzymes.
pQF-pleC: a DNA fragment including pleC ORF and the mapped pleC promoter as PCR-amplified from plasmid pJS14-pleC with primers 10781/10783, the product was digested with SpeI/KpnI and cloned into pQF digested with the same enzymes.
pdivJ-mCherry: the 3′ part of divJ was PCR-amplified from C. crescentus gDNA with primers 10205/10206, the product was digested with KpnI/AgeI and cloned into pCHYC-4 digested with the same enzymes.
pNPTS138-3xFLAG-tacA: the upstream region of the tacA ORF was PCRamplified from C. crescentus gDNA with primers 6438/6439 (fragment 1). A 5′ part of tacA was PCR-amplified from C. crescentus gDNA with primers 6440/7604 (fragment 2), and then used as a template with primers 6441/7604 to insert the coding sequence of the 3xFlag-Tag (fragment 3). SOE-PCR was performed with fragment 1 and 3 as template and primers 6438/7604. The resulting product was then digested with PstI/EcoRI and ligated into pNPTS138 digested with the same enzymes.
pNPTS138-3xFLAG-shkA: the upstream region of the shkA ORF was PCRamplified from C. crescentus gDNA with primers 6442/6443 (fragment 1). A 5′ part of shkA was PCR-amplified from C. crescentus gDNA with primers 6444/7605 (fragment 2), and then used as a template with primers 6445/7605 to insert the coding sequence of the 3xFlag-Tag (fragment 3). SOE-PCR was performed with fragment 1 and 3 as template and primers 6442/7605. The resulting product was then digested with PstI/EcoRI and ligated into pNPTS138 digested with the same enzymes.
pNPTS138-shkA(DD): the 5′ region of shkA was PCR-amplified from C. crescentus gDNA with primers 8169/8170 (fragment 1). The downstream region of the shkA ORF was PCR-amplified from C. crescentus gDNA with primers 8171/ 8172 (fragment 2). SOE-PCR was performed with the two fragments as template and primers 8169/8172. The resulting product was then digested with BamHI/ EcoRI and ligated into pNPTS138 digested with the same enzymes.
pNPTS138-tacA(DD): the 5′ region of tacA was PCR-amplified from C. crescentus gDNA with primers 7925/7716 (fragment 1). The downstream region of the tacA ORF was PCR-amplified from C. crescentus gDNA with primers 7717/ 7718 (fragment 2). SOE-PCR was performed with the two fragments as template and primers 7925/7718. The resulting product was then digested with HindIII/ EcoRI and ligated into pNPTS138 digested with the same enzymes.
pNPTS138-tacA(D54E): the mutant allele was generated by SOE-PCR using C. crescentus gDNA as template and flanking primers 7958/7961 and mutagenic primers 7959/7960. The product was digested with PstI/EcoRI and cloned into pNPTS138 digested with the same enzymes.
pNPTS138-ΔCC0168: the region upstream of CC0168 was amplified from C. crescentus NA1000 gDNA with primer pairs 3490 and 3491. The PCR product was cloned into pGEM-T Easy, sequenced, and cut out with HindIII and KpnI. The region downstream of CC0168 was amplified from C. crescentus NA1000 gDNA with primer pairs 3492 and 3493. The PCR product was cloned into pGEM-T Easy, sequenced and cut out with BamHI and KpnI. Both cut PCR products were ligated in a triple ligation with pNPTS138 digested with HindIII and BamHI.
pET28a-His-shkA: shkA was PCR-amplified from C. crescentus gDNA with primers 5988/5486, the product was digested with NdeI/HindIII and cloned into pET28a digested with the same enzymes.
pET28a-His-shpA: shpA was PCR-amplified from C. crescentus gDNA with primers 6015/6016, the product was digested with HindIII/EcoRI and cloned into pET28a digested with the same enzymes.
pET28a-His-tacA-RD: the sequence of tacA encoding the receiver domain was PCR-amplified from C. crescentus gDNA with primers 7925/6013, the product was digested with HindIII/EcoRI and cloned into pET28a digested with the same enzymes.
pET32b-Trx-His-shkA-HK: the sequence of shkA encoding the kinase catalytic core including the REC1 domain was PCR-amplified from C. crescentus gDNA with primers 6110/6011, the product was digested with HindIII/EcoRI and cloned into pET32b digested with the same enzymes.
pET32b-Trx-His-shkA-RD2: the sequence of shkA encoding the REC2 domain was PCR-amplified from C. crescentus gDNA with primers 6109/6009, the product was digested with HindIII/EcoRI and cloned into pET32b digested with the same enzymes.
pMT687-tacA: tacA was PCR-amplified from C. crescentus gDNA with primers 5487/5490, the product was digested with NdeI/KpnI and cloned into pMT687 digested with the same enzymes.
pMT687-tacA(D54E): the mutant allele was generated by SOE-PCR using C. crescentus gDNA as template and flanking primers 5487/5490 and mutagenic primers 5488/5489. The product was digested with NdeI/KpnI and cloned into pNPTS138 digested with the same enzymes.
pRKlac290-spmX: the promoter region of spmX was PCR-amplified from C. crescentus gDNA with primers 6864/6865. The resulting product was digested with EcoRI/PstI and ligated into pRKlac290 digested with the same enzymes.
pRKlac290-staR: the promoter region of staR was PCR-amplified from C. crescentus gDNA with primers 6621/6622. The resulting product was digested with EcoRI/PstI and ligated into pRKlac290 digested with the same enzymes.
pAH87: pMT375 was digested with XbaI and its insert was replaced with the NsiI-PacI cassette of pAH10 amplified with 7284 and 7287 digested with NheI/SpeI. pAH99: dendra2 was PCR-amplified from pDHL851 using primers 7503 and 7507, which also introduced a stop codon and the RBS of pQE70 upstream of dendra2. The product was digested with HindIII and PacI and ligated into pAH87 digested with the same restriction enzymes. pAH111: the promoter region of spmX was PCR-amplified from C. crescentus gDNA with primers 6864 and 7512, digested with EcoRI/HindIII and ligated into pAH99 digested with the same enzymes.
pAH139: dendra2-ssrA was generated by PCR from pDHL851 with primers 7503 and 6478. The resulting product was digested with HindIII/XhoI and ligated into pAH111 digested with the same enzymes.
Plasmid was delivered into C. crescentus by electroporation or triparental mating using LS980 as helper strain. φCR30 was used for generalized transduction 27 . Strains are listed in Supplementary Data 4. In general, C. crescentus mutant strains were constructed by a two-step homologous recombination selection/counterselection procedure using pNPTS138. Alternatively, pNPTS138based plasmids used for construction of a specific mutant strain were first integrated in this strain and then the mutation was moved into a clean background using φCR30-mediated generalized transduction, followed by resolution of the merodiploid by counterselection on PYE plates containing 0.3% [w/v] sucrose.
Synchronization. Strains were grown overnight in 5 ml PYE supplemented with appropriate antibiotics and inducers in a roller drum at 30°C. The next day, cultures were diluted 20-fold in 25 ml M2G and allowed to grow on a rocking shaker at 30°C until they reached an OD 660 of 0.4-0.6. These cultures were then subcultured in 300 ml M2G such that the cultures reached an OD 660 of 0.6 after overnight incubation on a rocking shaker (160 rpm) at 30°C. Cells were harvested (5000 × g, 10 min, 4°C), the supernatant was aspirated, cells were resuspended in 40 ml cold phosphate buffer (12.3 mM Na 2 HPO 4 , 7.8 mM KH 2 PO 4 ) or in cold M2G for experiments shown in Fig. 3c and kept at 4°C for the rest of the procedure. 20 ml cold Ludox were added and the suspension mixed thoroughly before being transferred to precooled glass Corex tubes and spun to separate the swarmer from the stalked cells (9780 × g, 35 min, 4°C). The top layer of cells was aspirated, the swarmer band moved to a clean precooled Corex tube and washed twice with 30 ml of either cold phosphate buffer or M2G (see above) by centrifugation (8000 × g, 10 min, 4°C). Cells were released to a flask with pre-warmed M2G at an OD 660 of~0.3 and incubated in a water bath (30°C, 150 rpm). Samples were taken at indicated time points in precooled tubes and harvested at max. speed using a cooled table-top centrifuge. The pellets were immediately snap-frozen in liquid nitrogen and stored at −20°C until further processing.
Microscopy and image analysis. Cultures were grown in PYE and imaged in exponential phase (OD 660 of 0.3-0.4) on 1% agarose PYE pads. Microscopy images were acquired using softWoRx 6.0 (GE Healthcare) on a DeltaVision system (GE Healthcare), equipped with a pco.edge sCMOS camera, and an UPlan FL N 100×/ 1.30 oil objective (Olympus). DivJ-mCherry localization was analyzed using Fiji software package 30 with the MicrobeJ plugin 31 . Oufti 32 was used to quantify cell length and data were analyzed in Prism 7 (GraphPad) with statistical testing using ordinary one-way ANOVA and Tukey's multiple comparison test. SpmX-mCherry localization and stalk formation ( Supplementary Fig. 1) was analysed manually.
Time-lapse microscopy and image analysis. For Fig. 4 and Supplementary Fig. 7 spmX promoter activity during the cell cycle of single cells of C. crescentus wild type and selected mutants was analyzed using P spmX -Dendra2 and time-lapse microscopy. Strains were grown overnight in 5 ml PYE supplemented with appropriate antibiotics in a roller drum at 30°C. On the day of the experiment, these cultures were diluted 50-fold in 5 ml PYE supplemented with 2.5 µg ml −1 oxytetracycline and if appropriate with 0.1 mM IPTG and allowed to grow in a roller drum at 30°C until they reached an OD 660 of 0.2. Cells were mounted on 1% PYE medium agar pads containing appropriate supplements sealed with a double layer of gene frames (Thermo Fisher Scientific; 1.7 × 2.8 cm) and subjected to time-lapse microscopy using softWoRx 6.0 (GE Healthcare) on a DeltaVision system (GE Healthcare), equipped with a pco.edge sCMOS camera, and an UPlan FL N 100×/1.30 oil objective (Olympus) using 0.3 s exposure time for phase-contrast, FITC and TRITC channels and a frame rate of 1 frame/15 min for different amounts of time.
For Fig. 4b, c, Supplementary Fig. 7, green Dendra2 was irreversibly photoconverted to red in the latest detectable predivisional stage 15 min before the first visible cell separation and birth of the swarmer cell using 2 s UV light (DAPI channel) leading to a conversion efficiency from green to red between 62-75%. Image processing and analysis was done using the Fiji software package 30 . From every predivisional cell and its offspring (swarmer and stalked cell) the total cell fluorescence and cell area was manually measured for every time point over the course of one cell cycle and if appropriate for up to three consecutive cell cycles and the integrated density (mean intensity normalised to cell area) determined. In case of time point 0 (predivisional cell after photoconversion), the mean intensity of the predivisional was normalised to the area of the swarmer or stalked cell. Integrated density values were then normalised for background fluorescence to obtain normalised fluorescence values over the cell cycle, plotted as arbitrary units (A.U.) in Supplementary Fig. 7.
In order to determine the fraction of cells with induced spmX promoter activity in Fig. 5c, for individual cells the mean of all normalised fluorescence values over one cell cycle was calculated. If the mean normalised fluorescence was above 150 A. U., cells were scored as showing induced PspmX activity; if below 150 A.U. then cells were scored as non-induced.
In Fig. 4c normalised fluorescence values over the cell cycle of each cell type were set relative to the value of the late predivisional cell after photoconversion (t 0 = 1). Green Dendra2 was irreversibly photoconverted to red only once in the latest detectable predivisional stage 15 min before the first visible cell separation and birth of the swarmer cell of generation 1. Image analysis was done as described above. Normalised fluorescence values of each cell type over each of the three cell cycles were set relative to the first late predivisional cell after photoconversion. The slope of relative fluorescence was determined over the course of one cell cycle and cells were scored as showing induced activity if the slope ≥0.19 or as not showing activity if the slope ≤0. 19. The fraction of cells with induced spmX promoter activity was then calculated and plotted.
The ON kinetics of P spmX activity in Fig. 4b was determined as following: Dendra2 was photoconverted once using 2 s UV light (DAPI channel) in the latest detectable predivisional stage 15 min before the first visible cell separation and birth of the swarmer cell. Cells were normalized and evaluated as above. The time of the first visible increase of cell fluorescence in the swarmer cell (green line) and the stalked cell (gray line) cells with slope ≥0.075 was plotted as fraction of cells with induced spmX promoter activity over time. For inferred OFF kinetics (blue dashed line), cells at different cell cycle stages were photoconverted once using 2 s UV light (DAPI channel), normalized as above and the increase of fluorescence within 15-min intervals was determined using the first time point after visible cell separation as reference. Fraction of cells scored as having induced spmX promoter activity over the cell cycle was plotted over time. Results were plotted using Prism 7 (GraphPad).
β-Galactosidase assays. For β-galactosidase assays strains harboring pAK502based lacZ reporter plasmids were grown in 2 ml PYE supplemented with chloramphenicol and additional antibiotics as appropriate overnight at 30°C in a drum roller and diluted the next day 20-fold in 2 ml of the same medium, followed by further incubation for an additional 4.5 h under the same conditions before sampling. β-Galactosidase assays were performed according to Miller 33 . For βgalactosidase assays strains harboring pRKlac290-based plasmids were grown overnight in 5 ml PYE supplemented with the appropriate antibiotics in a roller drum at 30°C. On the day of the experiment, these cultures were diluted 20-fold in 5 ml PYE supplemented with appropriate antibiotics and allowed to grow in a roller drum at 30°C until they reached an OD 660 of 0.3-0.54. 1.8 ml culture was spun and the pellet resuspended in 1.8 ml Z-buffer (0.06 M Na 2 HPO 4 , 0.04 M NaH 2 PO 4 , 0.01 M KCl, 0.001 M MgSO 4 , 0.05 M β-mercaptoethanol) of which 0.8 ml were used for OD 660 measurements and 1 ml transferred to a chloroformstable Eppendorf tube. Subsequently, 100 µl 0.1% SDS and 20 µl chloroform were added. Samples were vortexed for 10 s and incubated at room temperature for 30-60 min. From the top aqueous layer of the mixture three times 200 µl (three technical replicates) were transferred to a 96-well plate, 25 µl ONPG (4 mg ml −1 in Z-buffer) were added to each well and consumption of ONPG was followed at 405 nm using a BioTek Instruments EL800 plate reader (20 reads at the fastest interval). β-galactosidase activity was calculated as the initial slope of the increase of OD 405 over time at a linear range and corrected for the OD 660 and volume. Activities were normalized to the activity of a wild-type strain that was included in all assays.
Flow cytometry. Cultures were inoculated from a fresh colony and strains were grown in 2 ml PYE at 30°C in a drum roller 20 h (stationary phase) or diluted back 20-fold in 2 ml of the same medium after overnight incubation, followed by incubation for an additional 4.5 h (exponential phase). Rifampicin was added to cultures at a final concentration of 25 µg ml −1 and cultures were incubated for 2 h under the same conditions before cells were being fixed in cold 70% ethanol. Cells were collected by centrifugation, resuspended in 0.5 ml FACS buffer (10 mM Tris-HCl pH 7.5, 1 mM EDTA, 50 mM sodium citrate, 0.01% [v/v] TritonX-100) containing 2.5 µl RNaseA solution (Sigma) and incubated at room temperature for 30 min, after which 50 µl were transferred to 1 ml FACS buffer containing 1.5 µl YO-PRO-1 iodide (Thermo Fisher Scientific) and incubated for 2 h at room temperature in the dark. Data were acquired using a FACS Canto II (BD Biosciences) with >50,000 events recorded and analyzed with FlowJo software (FlowJo LLC). The gating strategy is outlined in Supplementary Fig. 10.
Mass spectrometry-based proteome analysis. 10 9 C. crescentus cells grown in PYE to exponential phase were collected, washed twice with phosphate buffer, resuspended in 50 µl lysis buffer (1% sodium deoxycholate, 0.1 M ammoniumbicarbonate), reduced with 5 mM TCEP for 15 min at 95°C and alkylated with 10 mM chloroacetamide for 30 min at 37°C. Samples were digested with trypsin (Promega) at 37°C overnight (protein to trypsin ratio: 50:1) and desalted on C18 reversed phase spin columns according to the manufacturer's instructions (Microspin, Harvard Apparatus).
One microgram of peptides of each sample were subjected to LC-MS analysis using a dual pressure LTQ-Orbitrap Elite mass spectrometer connected to an electrospray ion source (both Thermo Fisher Scientific). Peptide separation was carried out using an EASY nLC-1000 system (Thermo Fisher Scientific) equipped with a RP-HPLC column (75 μm × 30 cm) packed in-house with C18 resin (ReproSil-Pur C18-AQ, 1.9 μm resin; Dr Maisch GmbH, Ammerbuch-Entringen, Germany) using a linear gradient from 95% solvent A (0.15% formic acid, 2% acetonitrile) and 5% solvent B (98% acetonitrile, 0.15% formic acid) to 28% solvent B over 75 min at a flow rate of 0.2 μl/min. The data acquisition mode was set to obtain one high resolution MS scan in the FT part of the mass spectrometer at a resolution of 240,000 full width at half-maximum (at m/z 400) followed by MS/MS scans in the linear ion trap of the 20 most intense ions. The charged state screening modus was enabled to exclude unassigned and singly charged ions and the dynamic exclusion duration was set to 20 s. The ion accumulation time was set to 300 ms (MS) and 50 ms (MS/MS). The collision energy was set to 35%, and one microscan was acquired for each spectrum. For all LC-MS measurements, singly charged ions and ions with unassigned charge state were excluded from triggering MS2 events.
To determine changes in protein expressions across samples, a MS1-based label-free quantification was carried out. Therefore, the generated raw files were imported into the Progenesis QI software (Nonlinear Dynamics, Version 2.0) and analyzed using the default parameter settings. MS/MS-data were exported directly from Progenesis QI in mgf format and searched against a decoy database of the forward and reverse sequences of the predicted proteome from Caulobacter crescentus (strain NA1000/CB15N) (Uniprot, download date: 08/09/2015, total of (8234 entries) using MASCOT. The search criteria were set as following: full tryptic specificity was required (cleavage after lysine or arginine residues); three missed cleavages were allowed; carbamidomethylation (C) was set as fixed modification; oxidation (M) as variable modification. The mass tolerance was set to 10 ppm for precursor ions and to 0.6 Da for fragment ions. Results from the database search were imported into Progenesis QI and the final peptide measurement list containing the peak areas of all identified peptides, respectively, was exported. This list was further processed and statically analyzed using our in-house developed SafeQuant R script (SafeQuant, https://github.com/eahrne/SafeQuant 34 ). The peptide and protein false discovery rate was set to 1% using the number of reverse hits in the dataset. All quantitative analyses were performed in biological triplicates. All raw data and results associated with the manuscript have been deposited to the ProteomeXchange Consortium via the PRIDE 35 partner repository with the dataset identifier PXD012739.
Genetic selection for c-di-GMP-independent mutations in shkA. Six independent colonies of strain AKS1 (CB15 rcdG 0 ΔlacA::Ω ampG::pNPTStet-ampG) harboring in addition plasmid pAK503-spmX were inoculated into 5 ml PYE containing chloramphenicol (1 mg l −1 ) and oxytetracycline (10 mg l −1 ) and grown overnight. For UV mutagenesis, 2 ml of each culture was distributed in six-well plates and irradiated with 10,000 mJ UV light using a Stratalinker, after which 5 ml PYE containing chloramphenicol (1 mg l −1 ) and oxytetracycline (10 mg l −1 ) were added and cultures grown for 7 h at 30°C with shaking. Five hundred microliters of each culture were plated on PYE plates containing chloramphenicol (1 mg l −1 ), oxytetracycline (10 mg l −1 ), and kanamycin (6.25 mg l −1 ) and plates incubated at 30°C for 4 days. All colonies from one plate were pooled, diluted to an OD 660 of 0.1 and grown for 7 h at 30°C, and phage lysates were prepared for each independent pool. Following transduction of strain AKS17 (CB15 rcdG 0 ΔlacA::Ω /pAK502-spmX) with individual pool lysates, cells were plated on PYE containing chloramphenicol (1 mg/l), oxytetracycline (10 mg l −1 ), and X-Gal (40 mg l −1 ) and incubated at 30°C for 3 days. Blue colonies were once restreaked to confirm their blue colony phenotype, followed by colony-PCR using primers 9765 and 9766 and sequencing (Microsynth, Balgach, Switzerland) of the PCR product using primers 9765, 9766, 6011, and 6498.
Genetic screen for mutations abolishing spmX'-'lacZ activity. Transposon mutagenesis was performed by delivery of pSAM-Ec in strain UJ6168 carrying pAK502-spmX by conjugation and cells were plated on PYE containing chloramphenicol (1 mg l −1 ), kanamycin (50 mg l −1 ), nalidixic acid (20 mg l −1 ), and X-Gal (40 mg l −1 ) and incubated at 30°C for 3 days. White colonies were pooled in 5 ml PYE supplemented with appropriate antibiotics and grown overnight to prepare φCR30 lysates. From this pool lysate transposon insertions were transduced in a clean background (UJ6168/pAK502-spmX), followed by blue/white screening as described above. Lysates were prepared from single white colonies and mutations transduced in a NA1000 wild-type background. Transposon insertion sites were mapped using a two-step semi-arbitrary PCR approach 36 . Primers 9058, 9003 and 9004 were used in the first PCR and primers 9058 and 9005 in the second PCR. PCR products were sequenced (Microsynth, Balgach, Switzerland) with