The Caulobacter crescentus Homolog of DnaA (HdaA) Also Regulates the Proteolysis of the Replication Initiator Protein DnaA

ABSTRACT

It is not known how diverse bacteria regulate chromosome replication. Based on Escherichia coli studies, DnaA initiates replication and the homolog of DnaA (Hda) inactivates DnaA using the RIDA (regulatory inactivation of DnaA) mechanism that thereby prevents extra chromosome replication cycles. RIDA may be widespread, because the distantly related Caulobacter crescentus homolog HdaA also prevents extra chromosome replication (J. Collier and L. Shapiro, J Bacteriol 191:5706–5715, 2009, http://dx.doi.org/10.1128/JB.00525-09). To further study the HdaA/RIDA mechanism, we created a C. crescentus strain that shuts off hdaA transcription and rapidly clears HdaA protein. We confirm that HdaA prevents extra replication, since cells lacking HdaA accumulate extra chromosome DNA. DnaA binds nucleotides ATP and ADP, and our results are consistent with the established E. coli mechanism whereby Hda converts active DnaA-ATP to inactive DnaA-ADP. However, unlike E. coli DnaA, C. crescentus DnaA is also regulated by selective proteolysis. C. crescentus cells lacking HdaA reduce DnaA proteolysis in logarithmically growing cells, thereby implicating HdaA in this selective DnaA turnover mechanism. Also, wild-type C. crescentus cells remove all DnaA protein when they enter stationary phase. However, cells lacking HdaA retain stable DnaA protein even when they stop growing in nutrient-depleted medium that induces complete DnaA proteolysis in wild-type cells. Additional experiments argue for a distinct HdaA-dependent mechanism that selectively removes DnaA prior to stationary phase. Related freshwater Caulobacter species also remove DnaA during entry to stationary phase, implying a wider role for HdaA as a novel component of programed proteolysis.

IMPORTANCE Bacteria must regulate chromosome replication, and yet the mechanisms are not completely understood and not fully exploited for antibiotic development. Based on Escherichia coli studies, DnaA initiates replication, and the homolog of DnaA (Hda) inactivates DnaA to prevent extra replication. The distantly related Caulobacter crescentus homolog HdaA also regulates chromosome replication. Here we unexpectedly discovered that unlike the E. coli Hda, the C. crescentus HdaA also regulates DnaA proteolysis. Furthermore, this HdaA proteolysis acts in logarithmically growing and in stationary-phase cells and therefore in two very different physiological states. We argue that HdaA acts to help time chromosome replications in logarithmically growing cells and that it is an unexpected component of the programed entry into stationary phase.

INTRODUCTION

While most bacteria use the DnaA protein to initiate chromosome replication (1–3), bacteria responding to diverse environmental pressures probably evolved many different mechanisms to regulate DnaA and thereby adjust replication control to their specific needs (3–5). The model Gram-negative bacterium Caulobacter crescentus is found in nutrient-poor freshwater lakes and streams (6). C. crescentus evolved to divide asymmetrically, and it produces a motile “swarmer cell” and a nonmotile “stalked cell” after each cell division (6–8). Chromosome replication is coupled to this dimorphic cell division since only the stalked cells initiate chromosome replication, and the swarmer cells differentiate into stalked cells before they replicate their chromosomes (9–13). CtrA is a global response regulator protein that binds to and represses the C. crescentus origin of chromosome replication in the swarmer cells (10, 14–16). Competitive binding between CtrA and DnaA is a key mechanism of replication control that blocks replication in swarmer cells while allowing replication in the stalked cells (10, 17). In addition to this dimorphic control, C. crescentus chromosome replication occurs once and only once per cell division cycle (18). However, CtrA activity does not block extra rounds of chromosome replication prior to cell division (16, 19).

Escherichia coli uses at least three mechanisms to block extra rounds of chromosome replication. As reviewed by Katayama and coworkers (20), E. coli possesses two mechanisms that restrict DnaA binding to the origin of replication (oriC) and one mechanism, termed RIDA (regulatory inactivation of DnaA), that inactivates DnaA. Of these three mechanisms, RIDA is the dominant mechanism in E. coli (20, 21). E. coli DnaA binds ATP or ADP, but only the active DnaA-ATP form can initiate oriC replication. RIDA prevents overreplication by producing the inactive DnaA-ADP form. This is essentially a negative-feedback mechanism that operates during the DNA synthesis phase as follows. Lagging-strand DNA synthesis requires the frequent loading—at least one loading per Okazaki fragment—of the β-subunit of DNA polymerase III (Pol III) (DnaN), also known as the “sliding clamp” or the “ring protein.” Since DNA polymerase III uses a new DnaN ring protein at each Okazaki fragment, surplus DnaN rings accumulate and start a negative feedback to inactivate DnaA. This RIDA step uses the Hda (homolog of DnaA) protein that displays strong homology to the AAA⁺ domain of DnaA. Hda binds to the surplus DnaN rings and slides along the DNA to make direct contacts with DnaA, and so Hda stimulates the hydrolysis of the ATP bound to DnaA. This Hda-mediated inactivation of DnaA is an important cell cycle control since Hda-deficient E. coli overinitiates replication from oriC, thereby causing chromosome instability and cell death (20, 21).

Do other bacteria possess similar RIDA mechanisms to block extra rounds of chromosome replication? It was reported that E. coli hda corresponds to C. crescentus hdaA and that C. crescentus uses both dnaN and hdaA in an apparent RIDA-like mechanism to limit chromosome replication (22–24).

In our present report, we confirm that the loss of C. crescentus HdaA causes chromosome overreplication in accordance with a RIDA-like mechanism operating in C. crescentus. We also identify HdaA as a novel regulator of DnaA protein stability. In contrast to the E. coli DnaA protein, the C. crescentus DnaA protein is unstable: C. crescentus DnaA is rapidly turned over during the logarithmic-growth phase, and it is completely removed when the cells enter stationary phase (25). Here we show that the loss of HdaA stabilizes the DnaA protein in both exponentially growing and stationary-phase cells. Our results also demonstrate that the HdaA-dependent pathway for degrading DnaA is separate from the severe starvation pathway that requires SpoT signaling to stimulate DnaA degradation (26). We will discuss how our results expand the regulatory role of HdaA and how they change our view of entry into stationary phase by Caulobacter cells.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

The strains and plasmids used in this study are listed in Table 1. C. crescentus strains were grown in peptone-yeast extract (PYE) rich medium or M2G minimal glucose medium (27). The media used for strains requiring xylose were supplemented with 0.2% xylose (+X cultures) and 0.2% glucose (+G cultures) (PYEGX or M2GX). For selection in C. crescentus cultures, antibiotics were used at the following concentrations: spectinomycin (100 μg/ml), streptomycin (5 μg/ml), tetracycline (1.5 μg/ml), chloramphenicol (1.0 μg/ml), and kanamycin (20 μg/ml). When it was required to shut off the Pxyl promoter, xylose was removed from cultures by pelleting and suspending the cells twice in medium with 0.2% glucose. C. crescentus cells were severely starved for nutrients by shifting from complete minimal medium (M2G or M2GX) to minimal medium prepared without ammonium chloride (M2G−N or M2GX−N) or minimal medium prepared without glucose (M2). To specify the logarithmic- and stationary-phase conditions, an isolated colony was picked to start a fresh overnight (∼16 h, 28°C) culture. On the next day, this culture was diluted into the same fresh medium to an optical density at 660 nm (OD₆₆₀) of 0.05, and logarithmic growth resumed within 3 h, as judged by a linear semilog plot of time versus optical density. Logarithmic-growth samples were taken within this linear range and at optical densities of between 0.20 and 0.7. Stationary-phase samples were taken from the same cultures at ∼24 h, when the optical density had stopped changing for at least 8 h. All C. crescentus strains listed in Table 1 reached the same optical densities in the same media, specifically 1.1 OD₆₆₀ unit in PYE and 1.3 OD₆₆₀ units in M2G or in M2GX.

TABLE 1.

Bacterial strains and plasmids used in this study

Strain or plasmid	Description	Source or reference(s)
Strains
E. coli
DH5α	Cloning strain	Invitrogen
BL21(DE3)	Protein expression strain	Invitrogen
Caulobacter spp.
NA1000	WT Caulobacter crescentus, synchronizable	54, 55
GM3700	NA1000 Pxyl::hdaA ΔhdaA::Ω, xylose-regulated hdaA	This work
LS4427	NA1000 ΔspoT	26
GM2471	NA1000 Pxyl::dnaA ΔdnaA::Ω	25
FWC 14	Freshwater Caulobacter species	John Smit (UBC)a
FWC 21	Freshwater Caulobacter species	John Smit (UBC)
FCW 38	Freshwater Caulobacter species	41
Plasmids
pHP45	Ω DNA (spectinomycin and streptomycin resistance genes)	30
pNPTS138	DNA exchange vector, sacB nptI (Kan) oriT	29
pXTCYC-6	xylX locus integrating plasmid, tcyS cat (Cm)	31
pRXMCS-6	Replicating plasmid RK2 oriV, Pxyl promoter, cat (Cm)	31
pGM3891	pNPTS138-hdaA 5′ and 3′ DNA flanking Ω cassette	This work
pGM3892	xylX locus integrating plasmid, Pxyl::hdaA under xylose control	This work
pGM2876	Plasmid expressing pET19b-dnaA C. crescentus His-DnaA WT protein product	17
pGM3770	Plasmid expressing pET19b-dnaA C. crescentus His-DnaA R357A mutation	This work
pGM3771	Plasmid expressing pET19b-dnaA C. crescentus His-DnaA R357H mutation	This work
pGM3859	pRXMCS-6 Pxyl::dnaA WT under xylose control	This work
pGM3861	pRXMCS-6 Pxyl::dnaA R357A under xylose control	This work
pGM3862	pRXMCS-6 Pxyl::dnaA R357H under xylose control	This work

^aUBC, University of British Columbia, Vancouver, British Columbia, Canada.

Construction of the conditional hdaA expression strain GM3700.

GM3700 was constructed from the wild-type (WT) C. crescentus NA1000 strain by replacing the coding sequence of cc1711 (hdaA) with the Ω cassette (spectinomycin and streptomycin antibiotic resistance genes from pHP45) in the context of a trans-complementing hdaA gene integrated at the xylX locus (28). This construction was done stepwise as follows. DNA fragments for homologous DNA exchange, extending 1 kb upstream and 1 kb downstream of cc1711, were PCR amplified by the following primer pairs: upstream primer 1 (GAT ATC TCC TCA GGG TCC GGC GTG GTG AAC GTC) and upstream primer 2 (GGA TCC GCG AGG TTT TGC TTA TGT TGA CCA CGA GCG) or downstream primer 1 (GAT ATC GCC GTG ACG CTC CTT CTC CCC TTG CG) and downstream primer 2 (AAG CTT CGG CCT CTT CCT CGA TCT CGA G). The PCR-amplified DNA was cloned into vector pNPTS138 (29) between the BamHI and EcoRV restriction sites and between the EcoRV and HindIII restriction sites, as these complementary sites were included at the ends of the above primers. Next, the Ω cassette was removed from pHP45 (30) with SmaI and cloned into the EcoRV site of the plasmid described above to create pGM3891 (Table 1). A stable integration of pGM3891 into the cc1711 locus of NA1000 was confirmed by Southern blot analysis (see Fig. S1 in the supplemental material). The trans-complementing Pxyl::hdaA plasmid pGM3892 was constructed from pXTCYC-6 (31) as follows. First, the hdaA open reading frame was amplified by PCR primer pairs (hdaA primer 1 [GGT ACC TTG TCC ACC CAG TTC AAA CTG CCG C] and hdaA primer 2 [GAG CTC CTA CCC CTC ATC CCC CTC GAA CC]). Next, the PCR product was cloned between the KpnI and SacI sites of pXTCYC-6 to create the integrating plasmid pGM3892 (Table 1).

GM3700 was obtained from the WT NA1000 strain containing both integrated plasmids pGM3891 (sacB ΔhdaA::Ω) and pGM3892 (Pxyl::hdaA) as follows: These cells were grown in PYE plus 0.2% xylose (to express hdaA) and then plated on PYE with 0.2% xylose and 3% sucrose (to counterselect against the sacB gene on pGM3891). Colonies from cells that survived the sucrose counterselection were picked and tested for loss of pGM3891 by homologous recombination and for the exchange of the WT hdaA for the ΔhdaA::Ω allele. The loss of pGM3891 was initially scored by loss of nptI-encoded kanamycin resistance, and the exchange of the WT hdaA for the ΔhdaA::Ω allele was initially scored by the retention of Ω-encoded spectinomycin resistance and an acquired need for xylose to grow, since hdaA is essential for growth (22). The predicted DNA exchange was then confirmed by Southern blot analysis (see Fig. S1 in the supplemental material), and one cell line, GM3700, was selected for further studies.

Additional plasmid constructions and manipulations.

To create plasmids pGM3770 and pGM3771, the WT C. crescentus dnaA in pGM2876 (pET19b-dnaA) was converted to the dnaA R357A and dnaA R357H alleles using the QuikChange site-directed mutagenesis kit and protocols (Stratagene). All mutated and PCR-amplified DNA sequences were confirmed by the Sanger sequencing services provided by McGill University and Genome Quebec. Polyhistidine-tagged C. crescentus DnaA proteins (from pET19b-dnaA plasmids pGM2876, pGM3770, and pGM3771) were expressed in E. coli BL21(DE3), and these proteins were affinity purified as previously described (17, 32). The conditional dnaA-expressing plasmids (pGM3859 WT, pGM3861 R357A, and pGM3862 R357H) were constructed from the autonomously replicating plasmid pRXMCS-6 (broad-host-range RK2 oriV, Pxyl promoter, cat) as follows. The NdeI-to-BamHI dnaA-containing DNA bands (from pGM2876, pGM3770, and pGM3771) were gel isolated and ligated between the NdeI-to-BamHI sites of pRXMCS-6, thereby placing dnaA immediately downstream of the Pxyl promoter (31).

Immunoblot analysis.

For all immunoblots, the same cell mass was compared across all lanes equivalent to 1.0 ml of C. crescentus culture at an OD₆₆₀ of 0.1. The 1.0-ml culture samples were pelleted by spinning for 30 s at 12,000 rpm at 4°C in a microcentrifuge. The pellets were immediately suspended in 50 μl 2× Lammeli buffer, frozen at −20°C for storage, and/or boiled for 2 min prior to use. Whole-cell lysates were resolved using 8% SDS-PAGE. To ensure that equal amounts of protein were transferred, the membranes were stained with Ponceau S protein stain. The antiserum production and its strong specificity for DnaA were previously described (10). The immunoblot membranes were incubated with the DnaA antiserum diluted 1:10,000 followed by anti-rabbit immunoglobulin G conjugated to horseradish peroxidase diluted 1:10,000. The cross-reacting bands were visualized with the ECL enhanced chemiluminescence kit (PerkinElmer).

Fluorescence cytometry.

The chromosomal content of C. crescentus cells was determined using fluorescence cytometry (33). Cell samples were treated for 3 h with 60 μg/ml rifampin to block RNA synthesis (thereby preventing new rounds of chromosomal replication) and with 35 μg/ml cephalexin to block cell division. Therefore, under these conditions any chromosome that started replication would finish, and cells would contain multiples of whole chromosomes equal to the number of origins of replication present when the antibiotics were added. Cells were then washed in buffer (50 mM sodium citrate, 10 mM Tris-HCl, 1 mM EDTA [pH 7.2]) and fixed by adding ethanol to a final concentration of 70%. The RNA was removed by RNase A (10 μg/ml) treatment, and the DNA was stained with propidium iodide (20 μg/ml). The cell fluorescence data were acquired using the FL2 detector of a BD FACSCalibur (BD Biosciences) and the Cell Quest Pro acquisition software (BD Biosciences). Analysis of the data was performed using the FlowJo software (TreeStar, Inc.).

RESULTS

Conditional expression of HdaA in C. crescentus.

To explore the functions of HdaA in C. crescentus, we created a strain, GM3700, with the only copy of hdaA under the control of the xylose-requiring promoter Pxyl. The conditional expression of hdaA was evaluated and confirmed by immunoblot analysis using HdaA antiserum provided by Collier and Shapiro (22). Xylose is required to maintain HdaA protein in GM3700 cells, because after removal of xylose, the amount of HdaA protein in these cells decreased after 1 h, and HdaA was not detectable after 5 h (see Fig. S2 in the supplemental material). GM3700 requires xylose and therefore HdaA to form colonies on PYE plates (see Fig. S3 in the supplemental material). Microscopy showed that without xylose the GM3700 cells grow longer but without cell division (see Fig. S4A in the supplemental material). GM3700 cultures continue to increase cell mass and optical density without xylose (see Fig. S4B). However, after 3 h without xylose, the same GM3700 cells cannot be restored to normal growth and do not form colonies when plated on medium with xylose (see Fig. S4C). We conclude that HdaA is essential for C. crescentus growth and for normal cell cycle progression. We also conclude that GM3700 effectively shuts off hdaA expression, permitting a direct analysis of HdaA functions.

Loss of HdaA causes overreplication of the C. crescentus chromosome.

Both E. coli Hda (20, 21, 34, 35) and C. crescentus HdaA (22) restrict chromosome replication so that loss of Hda/HdaA produces cells with extra chromosome DNA. We confirmed this negative-feedback role for HdaA in GM3700 using two complementary techniques. First, we used fluorescence cytometry to monitor total chromosome DNA per cell resulting from an extra 3 h of incubation with rifampin and cephalexin to run out the established replication forks (33). Consistent with fluorescence cytometry results published by Collier and Shapiro, we similarly see that the loss of HdaA protein coincides with increasing DNA content per cell within 2 to 3 h following the removal of xylose (see Fig. S5 in the supplemental material). This timing roughly coincides with decreased CFU (see Fig. S4C in the supplemental material).

We also used a second technique that lets us monitor DNA methylation and chromosome replication specifically at the chromosome origin of replication (Cori). In C. crescentus, the CcrM methyltransferase is only expressed prior to cell division (13). Therefore, the first round of replication yields hemimethylated DNA (one unmethylated strand), and further rounds of chromosome replication yield unmethylated DNA that is uniquely sensitive to digestion by the restriction endonuclease HinfI (18). Cori DNA contains multiple HinfI sites (Fig. 1A), and their susceptibility to digestion is easily monitored by Southern blotting (18). Unlike fluorescence cytometry assays that measure bulk DNA, Southern blot assays monitor specific DNA sequences and do not require that DNA synthesis extend over long distances.

FIG 1.

(A) Locations of CcrM DNA methylation/HinfI target sites (GANTC) at the C. crescentus chromosome origin of replication (Cori). The adenines of these GANTC sites are methylated only within a narrow period, and this cell cycle property is used to detect extra chromosome replication. The restriction endonuclease HinfI only cuts GANTC DNA unmethylated on both strands that are produced by the passage of extra replication forks during the same cell cycle. Cori is located between the hemE and duf299 genes and the two BamHI sites used for Southern blot analysis below. (B) Southern blot analysis. C. crescentus GM3700 cells depleted of HdaA initiate extra chromosome DNA replication that is detected as new unmethylated DNA. GM3700 cells were grown in PYE medium supplemented with xylose and at 0 h, the culture was split into PYE medium with (+) or without (−) xylose. Samples were taken at the indicated times, and total chromosomal DNA was prepared, digested with the restriction endonucleases BamHI and HinfI, and used for Southern blot analysis. Southern blot membranes were hybridized with ³²P-radiolabeled DNA prepared from a 1.6-kb BamHI Cori fragment (shown above in panel A).

We applied this DNA methylation assay as follows. Exponentially growing GM3700 cultures were supplemented with xylose and glucose and then shifted to medium supplemented with only glucose or both glucose and xylose. Samples were taken every 2 h for side-by-side (plus xylose versus minus xylose) analysis by Southern blotting. The chromosome DNA was digested with BamHI and HinfI, and the membranes were hybridized with ³²P-labeled DNA derived from the cloned Cori BamHI fragment (Fig. 1A). This analysis reveals a 1.6-kb Cori BamHI band and smaller bands if Cori DNA is cut by HinfI. In rich PYE medium, such unmethylated Cori DNA becomes detectable after 6 h post-xylose shutoff, and after 8 h, the lower unmethylated DNA bands are clearly visible (Fig. 1B). Unmethylated Cori DNA was never observed in the wild-type strain (NA1000) (data not shown), and it was never detected in GM3700 cultured with xylose (Fig. 1B). Very comparable results were seen in both rich (PYEGX and PYEG) (Fig. 1B) and synthetic (M2GX and M2G) (see Fig. S6 in the supplemental material) media. These results independently confirm that loss of HdaA causes overreplication at Cori.

Cells use HdaA to remove DnaA protein during stationary phase.

We reasoned that altered hdaA genetic expression in GM3700 might affect dnaA expression, because in addition to its role as the initiator of chromosomal replication, DnaA also regulates transcription (36–38) and the E. coli dnaA is autoregulated (39). Therefore, to determine if GM3700 has an altered abundance of DnaA, we performed immunoblot assays. However, compared with the WT NA1000 cells, no significant change in the quantity of the DnaA protein was observed in logarithmically growing GM3700 cells with xylose (Fig. 2A, lanes 1 and 2). This equalization of DnaA probably does not involve an adjustment to log-phase proteolysis rates, because DnaA protein decays with very similar rates (with half-lives of 30 to 40 min) in both WT and GM3700 cells grown under the same PYEX condition (data not shown).

FIG 2.

The clearing of DnaA from stationary-phase cells requires HdaA. (A) GM3700 and wild-type (NA1000) cells were grown in PYE medium supplemented with xylose, the cultures were split into PYE medium with (+) or without (−) xylose, and samples were taken at 0 h (lanes 1 to 4) and after 20 h (lanes 5 to 8). Sample loadings per lane were adjusted to the same cell mass equivalent to 1.0 ml at an OD₆₆₀ of 0.1. Ponceau S staining confirmed that equal amounts of total cell protein were transferred to the membrane, and immunoblot analysis was performed with anti-DnaA serum. The initial abundances of DnaA during the logarithmic phase were similar in the GM3700 and wild-type cells (lanes 1 to 4). Therefore, the manipulations that shifted the wild-type and GM3700 cells into medium without xylose did not affect the DnaA protein levels (lanes 1 to 2 versus 3 to 4). These cultures restarted logarithmic growth at OD₆₆₀ of 0.4 at 0 h, and their corresponding stationary-phase samples (lanes 5 to 8) were taken at 20 h. Wild-type and GM3700 cells grown in PYE medium with xylose efficiently cleared the DnaA protein (lanes 5 to 7). Only the GM3700 cells grown without xylose retained DnaA protein after 20 h (lane 8). (B) GM3700 cells were grown to the stationary phase in M2G minimal media with xylose (M2GX [lane 1]) and without xylose (M2G [lane 2]). The culture of lane 2 was sampled and incubated for an extra 2 h with 0.2% xylose (lane 3) and with 0.02% xylose (lane 4). Immunoblot analysis was performed with anti-DnaA serum, as shown above (A).

However, blocking HdaA synthesis during logarithmic growth did block the removal of DnaA protein as GM3700 cells entered the stationary phase. The abundance of DnaA in wild-type C. crescentus (NA1000) and in the GM3700 strain was determined by immunoblotting. In control wild-type and +X GM3700 cultures, DnaA protein was efficiently cleared from stationary-phase cells (Fig. 2A, lanes 5 to 7). However, very unexpectedly, DnaA protein was not cleared from the –X-cultured GM3700 stationary-phase cells (Fig. 2A, lane 8). As all of these cultures grew to the same final optical cell densities, and as the ±X-cultured cells for lanes 6 and 8 grew at the same rate and reached stationary phase at the same time, we concluded that DnaA protein is retained in the –X-cultured GM3700 cells because HdaA is not present to somehow promote DnaA proteolysis. This DnaA protein is persistently stable. The stationary-phase cultures in Fig. 2A were sampled after ∼24 h, but in similar independent experiments, the DnaA protein persisted at comparable levels in 2- and in 3-day-old –X GM3700 cultures (data not shown). The results in Fig. 2A were observed in 4 independent experiments, and they imply a new and unexpected role for HdaA.

If HdaA promotes DnaA proteolysis, then restoration of HdaA by adding xylose back to the stationary-phase culture should stimulate the degradation of the DnaA protein. This hypothesis was tested in Fig. 2B as follows. As in Fig. 2A, these immunoblots confirm that GM3700 cells that entered stationary phase also retained DnaA but only when xylose was removed (compare lanes 1 and 2 in Fig. 2B). As predicted, addition of only 0.02% xylose to the same stationary-phase cells completely removed DnaA (Fig. 2B, lane 4). This 0.02% xylose addition did not stimulate cell growth (data not shown), yet it was sufficient to activate the Pxyl promoter (28). Therefore, HdaA can operate to remove DnaA in the absence of cell growth. This experiment also indicates that while these cells suffer from division defects (see Fig. S4A in the supplemental material) and do not form colonies when xylose is restored (see Fig. S4C), they are nonetheless metabolically active, since DnaA proteolysis requires an ATP-dependent mechanism (25, 40). Addition of a higher concentration of xylose (0.2%) also reduced but did not completely remove all of the DnaA protein (Fig. 2B, lane 3). Since the 0.2% xylose addition did stimulate some growth, what appears to be incomplete proteolysis is probably obscured by some new DnaA protein synthesis.

Cells use HdaA for efficient DnaA proteolysis during logarithmic growth.

In wild-type C. crescentus cells, the DnaA protein is turned over during logarithmic growth and the normal cell cycle progression, so DnaA protein levels during logarithmic growth result from a balance of synthesis and targeted proteolysis (25). To determine if HdaA also affects DnaA proteolysis during logarithmic growth, we measured DnaA stability with the following hdaA shutoff experiment. First, a logarithmically growing GM3700 culture was split into “plus xylose” (+X) and “minus xylose” (−X) media (PYEGX and PYEG, respectively) for 3 h. This split created parallel growing cultures with and without HdaA synthesis. Next, both cultures were treated with the antibiotic tetracycline (to block protein synthesis) and sampled every 30 min to measure DnaA protein levels by immunoblot analysis. As expected from previous studies, DnaA protein rapidly disappeared in the +X-cultured cells with an approximate half-life of 30 min (Fig. 3A). In contrast, DnaA protein remained in the −X-cultured cells (Fig. 3A). For example, DnaA protein remained after 150 min in these −X-cultured cells compared to undetectable levels of DnaA protein after 150 min in the corresponding +X-cultured cells.

FIG 3.

HdaA is required for DnaA proteolysis in logarithmically growing C. crescentus cells. (A) GM3700 cells were logarithmically grown in PYE medium supplemented with glucose and xylose (PYEGX), and the culture was split into media with xylose (PYEGX) and without xylose (PYEG) and grown for an extra 3 h to deplete HdaA. Tetracycline (1.5 μg/ml) was next added to halt translation, and samples of equal optical density were taken at 30-min intervals for immunoblot analysis with anti-DnaA serum. (B) The HdaA depletion period was changed to 0, 2, and 5 h. Tetracycline was added, and the cultures were sampled at 0 and 4 h. Otherwise these immunoblot experiments were performed as in Fig. 2A.

The preceding experiment established that a 3-h shutoff of hdaA expression was sufficient to inhibit proteolysis and stabilize DnaA protein in logarithmically growing cells. To more accurately establish the shutoff time required to reduce the concentration of HdaA to a nonfunctional level, we modified the preceding experiment (Fig. 3A) by varying the length of time (0, 2, and 5 h) that xylose was removed prior to our analysis of protein decay (Fig. 3B). Cultures were then sampled for DnaA immunoblot analysis at 0 h and at 4 h after tetracycline treatment. For all conditions, 4 h with tetracycline is sufficient to degrade all DnaA protein, except when the prior xylose/HdaA shutoff was 5 h (Fig. 3B, lane 12). In this case, about half the DnaA protein remains after 4 h with tetracycline. These experiments demonstrated that a shutoff period of 2 h is too short, while 3 h is sufficient to inhibit DnaA proteolysis. In summary, a critical level of HdaA protein also promotes DnaA proteolysis during logarithmic growth. This point is significant because in the E. coli/C. crescentus RIDA models Hda and HdaA serve as negative-feedback regulators to prevent extra rounds of chromosome replication by stimulating DnaA-ATP hydrolysis. This new data argues that DnaA proteolysis is also a part of the C. crescentus RIDA feedback loop.

HdaA-dependent proteolysis of DnaA is distinct from severe starvation proteolysis.

In wild-type C. crescentus cells, DnaA is degraded during the logarithmic growth phase, and the baseline degradation rate is accelerated when the cells are severely starved by removing carbon or nitrogen (25). To determine if C. crescentus HdaA is also required for starvation-accelerated DnaA proteolysis, we performed the following experiment: wild-type and GM3700 cells were grown in minimal medium with glucose and xylose (M2GX), then these cultures were split into the same medium with xylose (M2GX) and without xylose (M2G). These cells were grown for 4 h to reduce HdaA in the GM3700 M2G-cultured cells, and the levels of DnaA were measured as before by immunoblot analysis. A pilot HdaA depletion experiment, comparable to the experiment performed in Fig. 3A, but with M2G versus M2GX media showed that 4 h was sufficient to stabilize DnaA in –X-cultured GM3700 cells. The 4-h depletion increased the DnaA half-life from ∼30 min in M2GX to >120 min in M2G (data not shown). As shown in the left panel of Fig. 4A, all 4 cultures, wild type (±X) and GM3700 (±X) had comparable levels of DnaA protein (reproducing the controls of Fig. 2, lanes 1 to 4). Next, cells from all 4 cultures were placed in starvation media, i.e., the same media but lacking the nitrogen source ammonium chloride for 2 more hours. The immunoblot shown in the right panel of Fig. 4A demonstrates that for all 4 cultures, the DnaA protein was efficiently removed and barely detectable following this nitrogen starvation (−N). Therefore, GM3700 cells that were grown logarithmically and that lacked HdaA still had efficient starvation-accelerated DnaA proteolysis, as did the wild-type cells.

FIG 4.

C. crescentus cells depleted of HdaA can still degrade DnaA protein in response to sudden and severe nitrogen starvation both during logarithmic growth and during stationary phase. (A) Wild-type C. crescentus (NA1000) and GM3700 cells were grown logarithmically in M2G with xylose (+X), and their cultures were split into M2GX and M2G media. The cultures were grown for an extra 4 h to deplete HdaA from the M2G-grown GM3700 cells, and as before, DnaA abundance was measured by immunoblot analysis. These cells had equally large amounts of DnaA protein (lanes 1 to 4). Next these cells were transferred to the corresponding media lacking ammonium (−N), and therefore without the sole nitrogen source, for an additional 2 h, and DnaA abundance was measured by immunoblot analysis (lanes 5 to 8). DnaA protein was efficiently cleared, even in cells depleted of HdaA (lane 8, GM3700 M2G–N). (B) In lanes 1 and 2, GM3700 cells were grown to stationary phase in M2GX and M2G media. In the next 3 lanes, cells from the M2G culture were transferred to fresh medium lacking ammonium (M2G–N) or to fresh M2GX and fresh M2G media and then incubated for an additional 2 h. DnaA abundance was measured by immunoblot analysis of equal-cell-mass samples. Stable DnaA protein in stationary-phase M2G is efficiently removed by nitrogen starvation (lane 3, M2G–N).

To determine if HdaA is required for severe starvation DnaA proteolysis during stationary phase, we performed the following experiment: GM3700 cells were grown in minimal medium with glucose and xylose (M2GX), split into the same medium with xylose (M2GX) and without xylose (M2G), and then allowed to reach the stationary phase. As shown by immunoblot analysis, lanes 1 and 2 in Fig. 4B reproduce the expected result (as shown before in Fig. 2) that stationary-phase GM3700 cells retain DnaA protein only without xylose (M2G), i.e., without sufficient HdaA. This stationary-phase M2G culture (presented in Fig. 4B, lane 2) was then split 3 ways, and its cells were placed into fresh medium without ammonium and therefore lacking its sole nitrogen source (M2G−N) and as controls into fresh complete M2G and M2GX media. These split cultures were incubated for 2 more hours and analyzed on the same immunoblot. As shown in Fig. 4B, the DnaA protein that was stable in the stationary-phase cells (Fig. 4B, lane 2), became unstable and rapidly degraded when ammonium was removed (Fig. 4B, M2G−N, lane 3). The 2 adjacent control lanes show that these cells were metabolically active when returned to fresh M2GX and M2G media. The extra DnaA protein seen in these control lanes 4 and 5 is due to revived growth and new DnaA synthesis. Therefore, stationary-phase cells are not completely starved for nitrogen, and when the HdaA-dependent mechanism is blocked, a separate severe starvation-induced mechanism can still degrade the DnaA protein.

It might be argued that GM3700 cells retain DnaA in stationary phase not because they lack HdaA but because they consume fewer nutrients and that they are therefore not starved as much as the wild-type cells. This interpretation is unlikely because both wild-type and GM3700 cells grow to the same optical cell densities in the same media and therefore consume the same amounts of nutrients to produce the same biomass (data not shown). To further show that both wild-type and GM3700 strains respond differently to exactly the same nutrient limitations and other stationary-phase medium conditions, we preformed the following used-medium experiments: First, to prepare the used media, GM3700 cells were grown to stationary phase in rich PYE medium without xylose, and the cells were removed by centrifugation. Next, wild-type NA1000 cells and GM3700 cells were grown logarithmically for 6 h in fresh PYE medium with or without xylose. These cells were centrifuged, resuspended in the prepared “used” media, and incubated for 2 h. The DnaA immunoblots in Fig. 5 demonstrate that wild-type and GM3700 cells respond very differently to the same “used” PYE media. After 2 h, the wild-type cells rapidly removed all DnaA protein, while the GM3700 cells retained DnaA protein (Fig. 5, bottom blot). This inability to remove DnaA results from a lack of HdaA since GM3700 cells grown without xylose retained about half of their DnaA protein (lane 4). That the Pxyl promoter driving hdaA is relatively weak (28, 31) also supports this conclusion and explains why the GM3700 cells grown with xylose did not completely remove DnaA (lane 3) compared to the wild-type cells (lanes 1 and 2).

FIG 5.

The used-medium (exhausted-medium) shift experiment. The top panel shows the DnaA immunoblot of logarithmically growing cells. The bottom panel shows the DnaA immunoblot of the same cells after they were shifted for 2 h to “used” PYE medium (the remaining liquid separated from a stationary-phase GM3700 PYE batch culture). This 2-h starvation condition completely removed DnaA protein from wild-type cells (lanes 1 and 2), while substantial DnaA protein was retained in HdaA-depleted cells (lane 4). The removal of most of the DnaA protein in the lane 3 control further argues that HdaA is a direct cause of DnaA proteolysis. The retention of some DnaA in lane 3 probably results from experimental manipulations: Specifically, this could result from weaker hdaA expression by the artificial Pxyl fusion and from the metabolism and exhaustion of xylose during the growth and the starvation periods.

Cells entering stationary phase do not remove DnaA with the starvation-induced proteolysis signaled by SpoT.

SpoT produces the starvation signal ppGpp that is somehow necessary for the accelerated degradation of DnaA during severe carbon starvation (26). Wild-type C. crescentus cells growing in M2G have a baseline DnaA degradation rate that is increased severalfold by removing the sole carbon source (25), and this increased degradation is not seen in ΔspoT C. crescentus strains (26). Starvation promotes entry into stationary phase, and therefore to test the contribution of SpoT signaling to DnaA degradation during entry into stationary phase, we compared DnaA degradation rates in a wild-type C. crescentus strain and in an isogenic ΔspoT strain. To confirm that our ΔspoT C. crescentus strain does not accelerate DnaA degradation when suddenly starved, these growing ΔspoT C. crescentus cells were shifted from M2G to the same medium lacking glucose (the sole carbon source). As previously reported (26), DnaA was rapidly cleared from wild-type cells but ∼3-fold more slowly from the ΔspoT C. crescentus cells (data not shown).

After establishing this control for SpoT activity, we tested if SpoT influences DnaA degradation as the cells transition into stationary phase. The abundance of DnaA was measured by immunoblot analysis from parallel wild-type and ΔspoT C. crescentus cultures. Samples were taken at 2-h intervals before and after these cells entered stationary phase (Fig. 6). The wild-type culture started at a lower cell density and therefore entered stationary phase ∼2 h later. Considering the equal band intensities at each time point, the ΔspoT cells show an ∼2-h delay in DnaA removal. Nonetheless, this is a very small influence. Contrary to expectations, DnaA was cleared completely and at the same rate from both cultures. Therefore, while HdaA activity is required for DnaA degradation during stationary phase (Fig. 2 and 4), SpoT activity is not required to eliminate DnaA protein during this period.

FIG 6.

The abundance of DnaA decreases with the same kinetics in both wild-type (NA1000) and in ΔspoT cells as they enter stationary phase. (A) Immunoblot analysis to determine the abundance of DnaA in wild-type and ΔspoT cells (LS4427) as they transition from growing cells to stationary-phase cells. Samples were taken at 2-h intervals, and as before, DnaA abundance was measured by immunoblot analysis of equal-cell-mass samples. (B) The quantitation of the DnaA protein bands from the immunoblot in panel A. (C) The graph shows the optical densities of the PYE cultures at the sampling times in panel A. Both cultures entered stationary phase at the same rate, and both reached the same saturating cell densities.

The activated DnaA-ATP form is stabilized in growing C. crescentus cells.

E. coli Hda contacts activated DnaA-ATP and stimulates its conversion to the DnaA-ADP form that does not support chromosome replication (20), and C. crescentus HdaA presumably works by a similar mechanism (13). We therefore asked whether DnaA-ADP is more unstable than DnaA-ATP. Previous studies showed that C. crescentus DnaA missense mutations at R357 cause chromosome overreplication. This result was attributed to higher levels of DnaA-ATP, because the corresponding missense mutations in E. coli DnaA maintain DnaA-ATP and suppress DnaA-ADP formation in vivo and in vitro (23).

The following experiments presented in Fig. 7 show that DnaA missense mutations R357H and R357A cause chromosome overreplication and that they increase DnaA protein stability. We used site-directed mutagenesis to create the R357H and R357A alleles of C. crescentus DnaA. The wild-type and R357H DnaA proteins were purified as N-terminal His-tagged proteins from E. coli pET19b expression plasmids (10). As expected, the R357H protein showed less intrinsic ATP hydrolysis than the wild-type protein when incubated in vitro with [α-³²P]ATP (32; data not shown). Therefore, assuming similar ATP binding affinity, cells expressing the DnaA R357H protein should also preferentially maintain the ATP-bound form. To test in vivo consequences, the wild-type (WT) dnaA and the R357H and the R357A dnaA alleles were placed under the control of the Pxyl promoter on the multicopy plasmid pRXMCS6. These plasmids were next introduced into wild-type C. crescentus strain NA1000 grown on PYE plus glucose (PYEG) medium so that the Pxyl promoter remained off and chromosome replication would be directed by the native DnaA protein. When these cells were shifted to PYE plus xylose (PYEX) medium, extra WT, R357H, or R357A protein should be produced from the plasmids, and this was confirmed by immunoblotting (Fig. 7A, compare lanes 1 and 4). While xylose induced comparable amounts of WT, R357H, or R357A DnaA protein, only the R357H and R357A proteins induced extra chromosome replication (Fig. 7B). As assayed by fluorescence cytometry of total chromosome DNA per cell, the cultures with extra (xylose-induced) R357H and R357A DnaA proteins contained many cells (26.3 and 34.7% of the cells) with more than 2 chromosomes. In contrast to these results, only 1.8% of the cells in the control culture (xylose-induced WT DnaA) contained more than 2 chromosomes. So, unlike the mutant DnaA proteins, the induced WT DnaA does not stimulate any chromosome replication, because a 1 to 2% score is the background noise in this analysis (Fig. 7B, PYEG results). Also, these xylose-induced (PYEX) R357H and R357A cultures showed long nondividing cells, while the WT dnaA plasmid culture did not (data not shown). Therefore, unlike the WT, only the R357H and R357A proteins have gained an activity to overreplicate the chromosome, and this is likely caused by higher levels of activated DnaA-ATP.

FIG 7.

DnaA alleles R357H and DnaA R357A increase DnaA protein stability and chromosome replication. (A) Immunoblot measurements of DnaA protein stability in cells expressing the WT DnaA and ATPase-deficient mutants DnaA R357H and DnaA R357A. Replicating Pxyl::dnaA plasmids pGM3859 (WT), pGM3861 (R357H), and pGM3862 (R357A) were introduced into wild-type C. crescentus strain NA1000 grown logarithmically in PYEG (glucose) medium. Half of each culture was shifted to PYEX (xylose) medium for 2 h. Next, tetracycline (1.5 μg/ml) was added, and these cultures were sampled for immunoblot analysis at 0, 30, and 60 min, as shown. (B) Chromosome DNA fluorescence cytometry of the 0-min cultures in panel A. This table presents the percentage of cells with threshold DNA fluorescence above the 2-chromosome peak. These samples were taken immediately before tetracycline was added and analyzed by fluorescence cytometry as described in Materials and Methods: i.e., rifampin and cephalexin were added to complete chromosome replication without cell division. The inserted picture shows the plot of DNA fluorescence versus cell numbers from the WT DnaA PYEX culture sample, which was used to set the DNA fluorescence threshold for the other culture samples.

To test the stability of the WT, R357H, and R357A DnaA proteins, tetracycline was added to the same cultures, and equal samples (cell densities) were assayed by immunoblotting for DnaA (Fig. 7A). As before, DnaA was unstable and decayed with a half-life of ∼20 min in the three PYEG cultures and in the PYEX (WT DnaA) culture. In contrast, the two PYEX (R357H and R357A) cultures maintained strong DnaA protein bands long after the WT protein disappeared. Since both wild-type DnaA protein (from the chromosome) and R357 mutant DnaA proteins (from the plasmid) are present in these cells, the immediate decay of DnaA protein (Fig. 7A, compare lanes 4 and 5) is probably due to the wild-type protein. The R357 mutant proteins decay with much longer half-lives (2 to 3 h [data not shown]) and presumably predominate at 60 min in lane 6. Since the experiments in Fig. 7 show that DnaA missense mutations R357H and R357A are sufficient to promote DnaA protein stability and extra chromosome replication, these data also imply that DnaA-ATP is more stable than DnaA-ADP. The simplest interpretation implies that HdaA can promote DnaA proteolysis by creating DnaA-ADP.

Stationary-phase proteolysis of DnaA is conserved among Caulobacter strains.

To test whether regulated DnaA proteolysis is unique to C. crescentus or if it is more generally used by other bacteria, we measured DnaA protein levels in three additional Caulobacter species (41) under both logarithmic PYE growth and stationary-phase conditions. The serum raised against C. crescentus DnaA detects a protein band of comparable molecular weight in the related freshwater Caulobacter sp. strains FWC 14, FWC 21, and FWC 31 (Fig. 8). As with C. crescentus (NA1000), all of the freshwater Caulobacter strains in Fig. 8 show reduced levels of DnaA in stationary phase (S) compare to logarithmic growth (L). These results argue that proteolysis of DnaA is not incidental to our standard laboratory strains. In other words, it is not a trait that was artificially acquired by more than 40 years of laboratory subculture. Instead, proteolysis of DnaA is a generally conserved mechanism that evolved to regulate this essential replication protein.

FIG 8.

DnaA is cleared from wild-type C. crescentus (NA1000) and from related freshwater Caulobacter strains (FWC 14, 21, and 38) during stationary phase. DnaA abundance was measured by immunoblot analysis, as before. Samples were taken from logarithmic-growth-phase (L) PYE cultures and after the same cultures reached stationary phase (S), ∼20 h later. As before, all samples were adjusted to an equal cell mass equivalent to 1.0 ml at an optical density of 0.1. Control lanes 1 and 2 confirm antiserum specificity and show that C. crescentus strain GM2471 (Pxyl::dnaA) produces DnaA in xylose (X) but not glucose (G) PYE medium. Lanes 3 to 10 show that each of the Caulobacter strains decreases DnaA abundance in stationary phase relative to the logarithmic growth phase.

DISCUSSION

RIDA (regulatory inactivation of DnaA) is the predominant feedback mechanism that restricts DnaA activity in E. coli. RIDA uses Hda to inactivate DnaA-ATP so as to limit chromosome replication and thereby to coordinate replication with cell growth (20, 21). In agreement with Collier (13), we confirmed this major role for C. crescentus hdaA in limiting chromosome replication. In this report, we showed that blocking hdaA expression causes overreplication (i.e., more than once per cell cycle initiation of chromosome replication), as evidenced by the accumulation of multiple chromosomes per cell and the production of unmethylated DNA at the chromosome origin of replication. Intriguingly, we also discovered a novel role for HdaA in DnaA protein turnover. We showed that HdaA affects not only the activity of DnaA but also the stability of the DnaA protein.

Proteolysis is a major means for remodeling cells and for regulating cellular processes. C. crescentus has proved to be an outstanding model for bacterial proteolysis, and it has been especially informative about how proteolysis drives a dimorphic cell cycle (42–44). Regarding chromosome replication, both chromosome origin binding proteins CtrA and DnaA (10) are selectively targeted by proteolysis (25, 42). During the usual mode of genetic regulation, a protein is made when it is needed and proteolysis acts when it is not needed. For example, C. crescentus DnaA is made in growing cells and is removed when they enter stationary phase (25). In this first mode, proteolysis acts as the final and most secure suppressor of DnaA protein activity.

However, C. crescentus DnaA is made and also degraded in growing cells. In fact, DnaA protein is synthesized and degraded throughout the swarmer cell-to-stalked cell period of differentiation and during the dimorphic cell division period of the C. crescentus cell cycle (25). Is this a futile cycle of synthesis and degradation, or what might be the advantage of this second mode of proteolysis in growing cells? At present we have two hypotheses for this apparent paradox. Our first hypothesis is based on observations that new E. coli DnaA synthesis is required for oriC replication (20, 45). Newly synthesized apo-DnaA protein binds the more abundant ATP nucleotide, and this simple mechanism preferentially produces the active DnaA-ATP. Applying these considerations to the C. crescentus cell cycle implies that cycles of synthesis and degradation will shift the pool of active and inactive DnaA molecules toward the active state. We also proposed that this mode of proteolysis is an advantageous level of control, linking the cell's capacity for protein synthesis to DnaA activity: i.e., actively growing cells would be more likely to initiate chromosome replication (25).

A second hypothesis to explain the cycle of DnaA synthesis and degradation comes from recent observations that proteotoxic stress, i.e., the accumulation of denatured proteins triggers the C. crescentus lon protease to degrade DnaA (40). These results suggest that proteolysis could serve as a quality control to remove dysfunctional DnaA molecules during cell cycle progression. Some amount of proteotoxic stress occurs in all growing cells, but when it becomes too great and denatured proteins (bulk proteins and not just denatured DnaA) saturate the protein chaperones, Jonas et al. propose that Lon protease is activated to selectively remove DnaA and thereby to stop chromosome replication (40). These two hypotheses are not mutually exclusive but compatible positive and negative regulatory inputs, since the first proposes an advantageous stimulation, while the second proposes an advantageous override of DnaA activity.

Severe starvation triggers a third mode of C. crescentus DnaA proteolysis. When the sole carbon or nitrogen source is washed out of the cultures, C. crescentus cells respond by increasing the rate of DnaA proteolysis severalfold over the rate seen in complete media (25), and this response requires the SpoT signal ppGpp (26). Our results show that this more severe starvation response overrides HdaA-dependent modes of DnaA proteolysis in logarithmically growing cells (Fig. 4A) and in stationary-phase cells (Fig. 4B). Therefore, while past studies suggested that the SpoT signal simply accelerates an ongoing process, our present results suggest instead that SpoT and presumably ppGpp activate a new mechanism for DnaA proteolysis, one that does not use HdaA.

How might HdaA promote DnaA proteolysis? Two potential mechanisms are suggested by our work and by the literature. First, since HdaA probably makes a direct contact with DnaA (13), it is possible that HdaA acts as a protein chaperone and HdaA helps escort DnaA to the ClpP (25) or to the Lon (40) protease. As another possibility, HdaA may indirectly expose DnaA to proteolysis by removing DnaA from the DNA, by promoting DnaA unfolding, or simply by forming DnaA-ADP. Our data in Fig. 7 in fact argue that DnaA-ATP is more stable than DnaA-ADP. Therefore, during logarithmic growth, HdaA probably promotes DnaA proteolysis by forming DnaA-ADP. However, RIDA as presently understood only operates in actively replicating cells, so it does not account for HdaA-dependent proteolysis in stationary-phase cells. For example, DnaA is stable during stationary phase when HdaA is depleted, yet DnaA is removed when hdaA is subsequently expressed without restoring cell growth (Fig. 2B). Therefore, to explain HdaA-dependent proteolysis, the contributions of other mechanisms cannot be excluded at this time.

Our results also change our view of how C. crescentus cells enter stationary phase. We previously believed that cells entering stationary phase remove DnaA simply in response to starvation (25). However, this view is too simplistic. For example, the starvation SpoT/ppGpp signal stimulates DnaA proteolysis (26), but when WT and ΔspoT strains enter stationary phase, they remove DnaA with identical kinetics (Fig. 6), indicating that they are not using the SpoT/ppGpp signal. As discussed above, we now distinguish between the two HdaA-dependent modes of DnaA proteolysis—in logarithmic growth and in stationary phase—and the third HdaA-independent mode of DnaA proteolysis in response to severe and sudden starvation.

Bacterial entry into stationary phase is a response to gradual starvation plus the accumulation of metabolites and metabolic stresses (46, 47). Furthermore, the entry into stationary phase is a specific genetic program that evolved to anticipate severe conditions and to prepare the cells before these dangers are encountered. For example in Fig. 6, C. crescentus cells start to remove DnaA protein well before they stop growing. C. crescentus clearly has a complex stationary-phase program(s) (48–52), and DnaA is not just a replication factor: DnaA is also a global regulator of many genes (36). In light of our results, we propose that HdaA is not only a RIDA component but also a key component of the programs that prepare cells for survival in the stationary phase.

In summary, we have argued that HdaA has multiple roles, but which role is essential for cell viability? When HdaA is shut off, C. crescentus GM3700 cells drop their viability (CFU per milliliter in Fig. S4 in the supplemental material) around 4 to 5 h before they produce unmethylated chromosome DNA (Fig. 1), so the absence of natural DNA methylation does not cause this drop. Since the observed amount of unmethylated DNA results from a balance between new DNA synthesis and the methyltransferase activity of the CcrM enzyme (12, 18), the CcrM activity hides much DNA synthesis, and the assay in Fig. 1 detects severe overreplication. In contrast, DNA fluorescence cytometry detected overreplication in HdaA shutoff GM3700 cells only after 2 to 3 h (see Fig. S5 in the supplemental material), and this timing is roughly coincident with the viability drop (see Fig. S4 in the supplemental material). However, these are in fact 2- and 3-h sample times followed by an extra 3-h incubation with antibiotics (rifampin and cephalexin) that are needed to run out the established replication forks and so amplify the DNA. Therefore, while the cytometry assay seems to detect overreplication much sooner, it is not clear how to factor the 3-h incubation plus the consequences of the antibiotic treatments. In the wild-type and most C. crescentus cells, this antibiotic treatment blocks the initiation of replication and allows complete chromosome replication that is revealed as distinct peaks (10, 33). However, the broad fluorescent traces in Fig. S5 suggest that GM3700 cells are defective in both the initiation and the elongation phases of DNA synthesis. Interestingly, Baxter and Sutton also proposed extra roles for E. coli Hda (53). For example, in contrast to its role in RIDA, Hda may also promote the elongation of chromosome replication by stabilizing Pol III (the main replicative DNA polymerase) at the replication forks. Interestingly, these authors also isolated hda alleles that are defective for RIDA and yet remain viable, and they argue for separate essential roles, including DNA repair by Pol II and IV and DNA synthesis by Pol III (53). Therefore, while the exact cause of cell death remains controversial, it is clear that Hda and HdaA are regulators with many critical roles that deserve further exploration.