The level of DnaA, a key bacterial DNA replication initiation factor, increases during the Caulobacter swarmer-to-stalked transition just before the G1/S transition. We show that DnaA coordinates DNA replication initiation with cell cycle progression by acting as a global transcription factor. Using DnaA depletion and induction in synchronized cell populations, we have analysed global transcription patterns to identify the differential regulation of normally co-expressed genes. The DnaA regulon includes genes encoding several replisome components, the GcrA global cell cycle regulator, the PodJ polar localization protein, the FtsZ cell division protein, and nucleotide biosynthesis enzymes. In cells depleted of DnaA, the G1/S transition is temporally separated from the swarmer-to-stalked cell differentiation, which is normally coincident. In the absence of DnaA, the CtrA master regulator is cleared by proteolysis during the swarmer-to-stalked cell transition as usual, but DNA replication initiation is blocked. In this case, expression of gcrA, which is directly repressed by CtrA, does not increase in conjunction with the disappearance of CtrA until DnaA is subsequently induced, showing that gcrA expression requires DnaA. DnaA boxes are present upstream of many genes whose expression requires DnaA, and His6-DnaA binds to the promoters of gcrA, ftsZ and podJ in vitro. This redundant control of gcrA transcription by DnaA (activation) and CtrA (repression) forms a robust switch controlling the decision to proceed through the cell cycle or to remain in the G1 stage.
Two master regulators, CtrA and GcrA, have been shown to drive Caulobacter crescentus cell cycle progression (Holtzendorff et al., 2004). CtrA and GcrA, however, influence the transcription of only about 30% of the over 500 genes that are expressed at specific times in the cell cycle (Laub et al., 2000; Holtzendorff et al., 2004), suggesting that one or more additional factors also act as global cell cycle regulators. Here, we show that DnaA, a key replication initiation factor found in most Eubacteria (Gil et al., 2003), functions as a critical transcription factor required for the expression of the gene encoding the GcrA master regulator whose levels oscillate with CtrA during the cell cycle. Other genes in the DnaA regulon include those encoding nucleotide biosynthesis enzymes and components of the DNA replication machinery. DnaA has also been shown to act as a transcription factor in Escherichia coli (Messer and Weigel, 1997) and Bacillus subtilis (Burkholder et al., 2001).
Synchronized populations of swarmer-stage Caulobacter cells can easily be obtained and followed through the cell cycle (reviewed in Ryan and Shapiro, 2003). The motile swarmer cells have a single polar flagellum and several pili at one pole of the cell (Fig. 1A). CtrA, a DNA-binding response regulator that directly controls transcription of at least 95 genes in 55 operons (Laub et al., 2002), is present in swarmer cells where it binds to the Caulobacter origin of replication and blocks replication initiation (Quon et al., 1998). Swarmer cells undergo a swarmer-to-stalked cell differentiation during which the polar pili, flagellum and chemotaxis apparatus are lost and are replaced with a stalk, which grows at the pole previously occupied by the flagellum. Concurrent with the swarmer/stalked transition, Caulobacter cells degrade CtrA, initiate DNA replication, and progress from the G1 (pre-replication) to the S (DNA synthesis) phase of the cell cycle. Following the disappearance of CtrA, gcrA, whose promoter is directly repressed by CtrA, is transcribed. GcrA levels rise and affect the transcription of over 125 genes, including genes encoding components of the DNA replication and chromosome segregation machinery (Holtzendorff et al., 2004). Additionally, GcrA activates the P1 promoter of ctrA, the first of the two ctrA promoters to be consecutively activated during the cell cycle (Holtzendorff et al., 2004). During DNA replication, stalked cells elongate and develop into predivisional cells, which constrict at the future division plane and assemble a new flagellum at the pole opposite the stalk. Each predivisional cell ultimately divides asymmetrically into a sessile, replication competent stalked cell and a DNA replication quiescent swarmer cell.
DnaA is essential for replication initiation (Gorbatyuk and Marczynski, 2001), and DnaA binds specifically to DnaA boxes, nine base pair asymmetric DNA motifs found in the origin of replication of most Eubacteria (Mackiewicz et al., 2004). DnaA bound to the origin of replication causes the two strands of DNA to unwind and separate, allowing replication factors to enter and form an initiation complex (reviewed in Messer et al., 2001). The cellular level of DnaA is under cell cycle control, peaking at the G1/S transition (Gorbatyuk and Marczynski, 2005). To separate the effects of DnaA from other regulatory mechanisms active at the time of the G1/S transition, we used a dnaA-inducible strain (Gorbatyuk and Marczynski, 2001) to modulate dnaA expression and DnaA accumulation. Delaying dnaA expression delays processes that are largely DnaA-independent, such those involved in swarmer-to-stalked cell differentiation (Gorbatyuk and Marczynski, 2001), much less than it delays events dependent on DnaA, such as replication initiation and the transcription of some genes. Using this technique, we separated normally concurrent events in order to identify separate but overlapping regulatory pathways in complex cellular circuitry.
Delaying dnaA transcription separates the G1/S transition from the swarmer-to-stalked cell differentiation
At the time DnaA initiates DNA replication and causes Caulobacter cells to undergo the G1/S transition, many other cell cycle processes, including the swarmer/stalked morphological transition, are also happening (Fig. 1A). To distinguish DnaA-dependent events from other events that occur during the G1/S transition, we developed a DnaA depletion/induction cell cycle protocol that delays transcription of dnaA until after the swarmer/stalked transition and (presumably) other DnaA-independent events, that normally happen concurrently with the G1/S transition, have occurred (Fig. 1B). The protocol uses GM2471, a strain of C. crescentus where the only copy of the monocistronic dnaA gene is on the chromosome under the control of a xylose-inducible promoter (Gorbatyuk and Marczynski, 2001).
In the DnaA depletion/induction protocol, GM2471 swarmer cells are isolated from a culture grown in M2GX (minimal salts media with glucose and xylose) where dnaA is transcribed and placed in M2G (minimal salts media with glucose as the only carbon source) where dnaA is off. After 60 min in M2G, most GM2471 cells develop into stalked cells, and only about 10% of the cells, as judged by light microscopy, are still motile (Fig. 1C). During continued DnaA depletion in M2G, GM2471 cells fail to initiate DNA replication (Fig. 1D) or develop into pinched predivisional cells (Fig. 1C). In the DnaA depletion/induction protocol, xylose is added, inducing dnaA, after 75 min of DnaA depletion in M2G. Most cells then initiate DNA replication 25–45 min later (Fig. 1D), finish the cell cycle, and divide. The number of viable cells doubles within 3 h of DnaA induction (Fig. S1). By depleting DnaA for 75 min, we obtained the maximum separation between DnaA-independent and DnaA-dependent processes without decreasing the synchronicity of the subsequent recovery (data not shown).
Using gene expression microarrays, we obtained time courses of transcription during both the DnaA depletion/induction cell cycle and the wild-type cell cycle (Fig. 2A). Some genes previously identified as cell cycle-regulated (Laub et al., 2000) are not included in Fig. 2A because their expression profiles were too noisy, had insufficient variation during the cell cycle, or were affected by xylose, which was used to induce dnaA. We found that 267 genes had robust, reproducible, time-varying behaviour during both the wild-type and the DnaA depletion/induction cell cycles.
To facilitate comparisons between expression profiles from the wild-type cell cycle and the DnaA depletion/induction cell cycle, we defined a transition interval as the time during which a gene's expression profile increased from 25% to 75% of its maximum (Fig. 2B). We also defined the transition midpoint as the time when a profile had completed half of its increase (Fig. 2B). Transition times during the wild-type and the DnaA depletion/induction cell cycles were compared graphically (Fig. 2C). Based on their transition midpoint times, genes that ‘turned-on’ (went from low expression to high expression) during both the wild-type and DnaA depletion/induction cell cycles were partitioned into three groups (Fig. 2D). Genes in both the (i) Pre-DnaA and (ii) Post-DnaA sets are co-expressed early (before ctrA transcription) in the wild-type cell cycle. The Pre-DnaA genes, however, turn on before DnaA induction in the DnaA depletion/induction cell cycle, indicating that their expression does not require DnaA. In contrast, Post-DnaA genes turn on only after DnaA induction indicating that their expression requires DnaA and is not induced by the swarmer/stalked transition alone. The (iii) Late set contains genes transcribed with or later than ctrA in both the wild-type and the DnaA depletion/induction cell cycles. Exceptions are discussed in Supplementary material.
Additionally, we identified an Early set which contains genes whose observed expression variations are likely due to the stress of the synchrony process (carbon starvation and cold) rather than the normal cell cycle. These Early genes were expressed at the beginning of both the wild-type and DnaA depletion/induction cell cycles, but, unlike genes expressed in swarmer cells, were not expressed at the end of the wild-type or DnaA depletion/induction cell cycle experiments.
The 59 Pre-DnaA genes in Table 1 and the 40 Post-DnaA genes in Table 2 belong to multiple functional classes, with the Post-DnaA set being particularly enriched for genes encoding nucleotide biosynthesis enzymes and components of the DNA replication, recombination and repair machinery. The Post-DnaA set contains genes encoding PodJ, a polarity determinant (Viollier et al., 2002); PleC, a dynamically localized histidine kinase required for multiple polar events (Wheeler and Shapiro, 1999); and FtsZ, a GTPase that forms a ring at the division plane and is required for cell division (Quardokus et al., 2001). Of particular interest, the expression of gcrA, which encodes one of the two oscillating global cell cycle master regulators, waits for DnaA induction. Furthermore, 10 of the 40 genes in the Post-DnaA set are GcrA-induced (Holtzendorff et al., 2004).
Table 1. Early cell cycle genes whose transcription does not require DnaA (Pre-DnaA set).
Chromosome partitioning and cell division
intracellular septation protein A, putative
DNA replication, recombination and repair
ATPase related to the helicase subunit of the Holliday junction resolvase
Although DnaA depletion does not prevent the expression of the 59 Pre-DnaA genes, DnaA depletion does delay the expression of 23 Pre-DnaA genes by at least 20 min, based on the start time of the corresponding transition intervals. Additionally, transition intervals for 30 Pre-DnaA genes are at least 20 min longer for the DnaA depletion/induction cell cycle than for the wild-type cell cycle. The slow turn-on of many Pre-DnaA genes during DnaA depletion suggests that DnaA directly or indirectly affects their expression.
The Pre-DnaA set includes the hemE gene, adjacent to the Caulobacter origin of replication. The hemE gene has two promoters: the weak Pw promoter, whose transcripts are translated into HemE protein, and the CtrA-repressed, strong Ps promoter, whose transcription during the stalked stage may help unwind the origin of replication, promoting DNA replication initiation (Marczynski et al., 1995). Consistent with the results of Gorbatyuk and Marczynski (Gorbatyuk and Marczynski, 2001), we find that hemE is transcribed during DnaA depletion after the swarmer/stalked transition, presumably in response to the absence of CtrA (see below), but that hemE transcription is insufficient for DNA replication initiation.
Using transition midpoint times, we found that while the 139 Late genes turn on in essentially wild-type order during the DnaA depletion/induction cell cycle, transcription of Late genes in the DnaA depletion/induction cell cycle is delayed ∼60 min compared with the wild-type cell cycle. In both the wild-type and DnaA depletion/induction cell cycles, ctrA is one of the first genes in the Late set to be expressed. Many CtrA-activated genes are in the Late set, including genes encoding the DivK and PleD response regulators, the CcrM methyltransferase, and flagellum, pili, holdfast and chemotaxis proteins. The comparable activation order of genes in the Late class during both the wild-type and the DnaA depletion/induction cell cycle suggests that once ctrA transcription has been turned on, the cells have largely recovered from the DnaA depletion. The widespread dependence on CtrA of transcription that occurs after ctrA activation appears to buffer the timing of expression of genes needed near end of the cell cycle from perturbations that happen earlier in the cell cycle.
The expression profiles for CC2234 (bottom of Fig. 2A), which is expressed early in the DnaA depletion/induction cell cycle and late in the wild-type cell cycle, do not fit into any of the four main classes. CC2234 encodes a protein of unknown function, and the cause of CC2234's anomalous behaviour is unknown.
GcrA, FtsZ and PodJ accumulation requires DnaA
Using immunoblots, we investigated the accumulation of GcrA, FtsZ and PodJ, which are encoded by genes in the Post-DnaA set, during the DnaA depletion/induction cell cycle. We also tracked the levels of CtrA, which disappears during the swarmer/stalked transition in both wild-type cells (Fig. 1A) and cells undergoing DnaA depletion (Fig. 3A). In the wild-type cell cycle, depicted in Fig. 1A, GcrA appears concurrent with CtrA degradation (Holtzendorff et al., 2004). In contrast, during DnaA depletion, GcrA levels remain low following CtrA degradation, leading to a situation where neither CtrA nor GcrA is present (Figs 1B and 3A). Subsequent induction of DnaA results in the accumulation of GcrA, consistent with the observed increase in gcrA mRNA levels following DnaA induction. These results suggest that the gcrA promoter, in addition to being repressed by CtrA (Holtzendorff et al., 2004), is activated by a DnaA-dependent factor or by DnaA itself.
Like the gcrA promoter, the ftsZ and podJ promoters are directly repressed by CtrA (Kelly et al., 1998; Crymes et al., 1999; Holtzendorff et al., 2004). During the DnaA depletion/induction cell cycle, ftsZ and podJ mRNA levels, like gcrA mRNA levels, increase little in response to CtrA degradation and instead wait for DnaA induction. During DnaA depletion, FtsZ levels increase slightly following the swarmer/stalked transition and CtrA degradation, and induction of DnaA causes an additional increase in FtsZ levels (Fig. 3B).
The podJ gene is specifically transcribed following the swarmer/stalked cell transition and PodJ in its long form, PodJL, is positioned at the incipient swarmer cell pole of predivisional cells (Viollier et al., 2002). At cell division, PodJL is processed to a short form, PodJS, which remains at the pole of the progeny swarmer cell until it is cleared by a regulated proteolytic event during the swarmer/stalked transition (Chen et al., 2005). While the clearance of PodJS at the swarmer/stalked transition does not require DnaA, PodJL accumulates only after DnaA induction (Fig. 3C). The failure of cells to accumulate GcrA, FtsZ and PodJ while the G1/S transition is blocked by DnaA depletion demonstrates the high degree of coordination between DNA replication, polar development and cell division in Caulobacter.
DnaA boxes are present upstream of some DnaA-inducible genes
Of the 40 genes in the Post-DnaA set, 21 contain at least one match (seven of nine or better) to TTATNCACA, the E. coli consensus DnaA box (Messer and Weigel, 1997), in the 200 bases upstream of their translation start site (either strand). Neither random sets of genes nor the Pre-DnaA, Early, or Late sets contain similar enrichments of upstream DnaA boxes. Thus, the Post-DnaA set possesses a unique feature, and the DnaA boxes upstream of Post-DnaA genes are most likely functional rather than a chance occurrence, suggesting that Caulobacter DnaA acts as a transcription factor.
We created an improved, more specific model for Caulobacter DnaA boxes from the candidate DnaA boxes upstream of the most strongly induced genes in the Post-DnaA set and the four best matches to the E. coli DnaA box present in the Caulobacter origin of replication (Marczynski and Shapiro, 1992) using MEME (Bailey and Elkan, 1994). Thirteen genes that belong to the Post-DnaA set and contain a DnaA box (MEME model) in the 200 bases upstream of the translational start site were identified as candidates for the DnaA regulon (Table 3). Expression of all 13 genes increases following DnaA induction. Of these 13 genes, six have DnaA boxes upstream of the −10 to −35 promoter region and two have DnaA boxes downstream of their +1 sites. The +1 sites for five of the 13 genes have not been identified. The DnaA boxes upstream of podJ and gcrA both overlap DNA sites recognized by the CcrM methyltransferase, so the methylation state of the DNA, which reflects the replication state of the chromosome (Marczynski, 1999), may affect DnaA binding.
Table 3. Caulobacter DnaA regulon identified by delaying dnaA transcription.
His6-DnaA binds to the gcrA, ftsZ and podJ promoters in vitro
We used an in vitro gel-shift assay to test the binding of His6-DnaA to the gcrA, ftsZ and podJ promoters. γ[32P]ATP-labelled, 29 base pair, double-stranded DNA probes containing the DnaA boxes centred 64 (ftsZ-1) and 47 (ftsZ-2) bases upstream of the ftsZ+1 site (Kelly et al., 1998), 55 bases upstream of the gcrA+1 site (Holtzendorff et al., 2004), and 84 bases upstream of the podJ+1 site (Crymes et al., 1999) formed concentration-dependent, specific complexes with His6-DnaA (Fig. 4A). The gcrA probe had the strongest binding while the ftsZ-2 probe had the weakest. The close proximity of two of the three DnaA boxes in the ftsZ promoter suggests that DnaA may bind cooperatively to the full-length region. We used a region of the origin of replication containing a DnaA box as a positive control. As a negative control, we used the same region with a two base pair mutation in the DnaA box known to be incompatible with replication initiation (Marczynski and Shapiro, 1992). The wild-type origin region formed a complex with His6-DnaA while the mutated origin did not. An excess of unlabelled probe in the reactions disrupted the binding of the [32P]-labelled probe, but excesses of either of two arbitrary competitors did not (Fig. 4B). The presence of DnaA boxes upstream of genes that are induced rapidly following DnaA induction (the Post-DnaA set) coupled with the binding of His6-DnaA to promoters of selected DnaA-inducible genes suggests that DnaA acts as a transcription factor in Caulobacter.
The essential DnaA replication initiation factor is present at peak concentrations during the Caulobacter G1/S transition coincident with swarmer-to-stalked cell differentiation (Gorbatyuk and Marczynski, 2001; 2005). Furthermore, in both E. coli (reviewed in Messer and Weigel, 1997) and B. subtilis (Burkholder et al., 2001), DnaA has been shown to function as a transcription factor. We have shown here that DnaA plays a central role in the top-level genetic circuit that controls Caulobacter cell cycle progression.
In the normal Caulobacter cell cycle, many regulatory events occur coincidently during the swarmer/stalked cell and G1/S transitions, obscuring the role of DnaA. By delaying dnaA expression, we separated DnaA-dependent events including the G1/S transition from the swarmer/stalked cell transition and other early cell cycle pathways that normally execute concurrently (Fig. 5). Of 99 genes expressed during the swarmer/stalked cell transition, we identified 59 (Table 1) whose expression does not require DnaA and 40 whose expression is DnaA-dependent (Table 2). Thirteen of the DnaA-dependent genes have a DnaA box in their promoter regions and likely belong to the DnaA regulon (Table 3). Because delaying dnaA expression delays DNA replication, some genes whose expression appears to be indirectly dependent on DnaA may actually be dependent on DNA replication (Humphery-Smith, 1999). For example, expression of some genes may depend on the methylation state of the DNA, which reflects the replication state of the chromosome (Marczynski, 1999).
The DnaA regulon includes genes encoding nucleotide biosynthesis enzymes, components of the DNA replication machinery, and GcrA, a global regulator of the Caulobacter cell cycle. The DnaA regulon also contains podJ, which encodes a polarity determinant that is asymmetrically localized in predivisional cells (Viollier et al., 2002), and ftsZ, which encodes an essential cell division protein (Quardokus et al., 2001). His6-DnaA directly binds the promoters of gcrA, ftsZ and podJ.
Multiple regulatory factors control a subset of the DnaA-regulated genes. CtrA directly regulates four genes (Laub et al., 2002) and GcrA directly or indirectly regulates seven genes in the DnaA regulon (Holtzendorff et al., 2004). For four genes in the DnaA regulon, DnaA is the only known regulator. While DnaA boxes are upstream of the −10 to −35 promoter regions of most genes in the DnaA regulon, DnaA boxes are downstream of the +1 sites for hu and nrdA, similar to the case in E. coli where DnaA positively regulates the bacteriophage λpR promoter by binding downstream of the +1 site (Glinkowska et al., 2003).
The Caulobacter DnaA regulon likely contains more than the 13 genes conservatively identified in this work. For example, genes whose expression is strongly affected by the carbon source used to turn on and off the inducible promoter employed to delay dnaA expression (Meisenzahl et al., 1997; Gorbatyuk and Marczynski, 2001) were not considered for membership in the DnaA regulon. In addition, genes induced by both DnaA and other factors in a redundant fashion may not have been identified as DnaA-dependent. Somewhat surprisingly, repression by DnaA, which can be caused by DnaA binding within a promoter region or DnaA binding within a transcribed region leading to transcriptional termination (reviewed in Messer and Weigel, 1997), was not definitively observed, suggesting that any DnaA-repressed genes are also affected by other factors. Additionally, if DnaA regulates genes at times in the cell cycle other than the G1/S transition, these genes would not have been identified.
DnaA is a master regulator of the Caulobacter cell cycle
As an activator of gcrA transcription, DnaA, along with CtrA and GcrA, is an integral component of the top-level genetic circuit controlling Caulobacter cell cycle progression (Fig. 6). In swarmer cells, CtrA actively inhibits replication initiation by binding to the chromosome origin of replication (Quon et al., 1998). At the same time, CtrA directly represses transcription of gcrA (Holtzendorff et al., 2004), ftsZ (Kelly et al., 1998) and podJ (Crymes et al., 1999), blocking the early steps in cell division and polar development. The concurrent proteolysis of DnaA in swarmer cells (Gorbatyuk and Marczynski, 2005) removes DnaA's activating effect on the same components. During the swarmer/stalked transition, CtrA is degraded (Quon et al., 1996) while DnaA levels increase (Gorbatyuk and Marczynski, 2005). The presence of DnaA, and not just the absence of CtrA, is required to trigger an increase in GcrA levels and start the next wave of cell cycle transcription, which includes the expression of genes encoding nucleotide biosynthesis and DNA replication enzymes. Thus, DnaA not only initiates DNA replication but also promotes the transcription of the components necessary for successful chromosome duplication. DnaA also activates transcription of ftsZ and podJ, starting the cell division and polar organelle development processes that, in addition to DNA replication, will prepare the cell for asymmetric division. Ultimately, the DnaA-induced increase in GcrA levels coupled with DNA replication turns ctrA transcription back on (Holtzendorff et al., 2004).
CtrA and DnaA levels exhibit inverse behaviour throughout the cell cycle (Gorbatyuk and Marczynski, 2005). For example, swarmer cells contain CtrA and only low levels of DnaA while stalked cells contain DnaA but not CtrA. Caulobacter cells use the reciprocal behaviour of CtrA and DnaA in a regulatory logic motif involving (i) CtrA negative regulation and (ii) DnaA positive regulation to form a robust switch to implement the decision to stay in the G1 stage or to undertake a round of cell division.
We surmise that this reliable, redundant control of entry into the cell division pathway would be advantageous for an organism like Caulobacter that normally lives in an environment that permits only infrequent divisions (Poindexter et al., 2000). Having to change the level of two master regulators, CtrA and DnaA, rather than just one, in order to cause key cell cycle events like replication initiation or gcrA, ftsZ and podJ transcription reduces the chance of erroneously triggering cell cycle progression.
Liquid cultures of C. crescentus CB15N (wild-type) and GM2471 (Gorbatyuk and Marczynski, 2001) were grown at 30°C in peptone yeast extract (PYE) or M2 minimal media containing either 0.2% glucose (M2G) or 0.2% glucose and 0.3% xylose (M2GX) (Ely, 1991). Cell cultures were synchronized using Ludox density centrifugation (Quon et al., 1996) or Percoll density centrifugation (Tsai and Alley, 2001).
Samples for microarray analysis were collected by centrifugation and were frozen in liquid nitrogen. RNA was isolated using Trizol (Invitrogen) (Holtzendorff et al., 2004).
For one wild-type cell cycle data set, a synchronized culture of CB15N swarmer cells grown in M2G was sampled every 15 min during the 150 min cell cycle. For a second wild-type cell cycle data set, a synchronized culture of CB15N swarmer cells grown in PYE (rich media) was sampled every 10 min during the faster 100 min cell cycle. We also used the wild-type cell cycle data from (Laub et al., 2000), which provided a 15 min sampling of the CB15N cell cycle in M2G. All times shown for the wild-type cell cycle refer to the 150 min M2G cell cycle, not the faster PYE cell cycle.
For each of two DnaA depletion/induction cell cycles, swarmer cells were isolated from GM2471 cultures grown in M2GX and placed in M2G, which turned off dnaA transcription. After 74 min in M2G, xylose was added to induce dnaA transcription. Samples were taken every 15 min from 0 to 195 min. The delay in dnaA transcription causes the DnaA depletion/induction cell cycle to take longer than the M2G CB15N cell cycle even though both experiments use similar growth conditions.
Microarrays containing a 50mer oligo for each Caulobacter gene were printed, hybridized, and washed as in Holtzendorff et al. (2004), except that arrays for the wild-type M2G cell cycle were hybridized at 45°C; all other microarrays were hybridized at 44°C. Each microarray compared the relative abundance of mRNA species from one time during the cell cycle to a common reference of RNA from a log-phase CB15N culture that contained all cell cycle stages. For example, the M2G wild-type cell cycle data set consists of arrays comparing the 0 min sample to the reference, the 15 min sample to the reference, the 30 min sample to the reference, . . . , and the 150 min sample to the reference. All samples from each cell cycle experiment used the same reference RNA. Reference RNA for the PYE, M2G and DnaA depletion/induction cell cycles was isolated from cells grown in PYE, M2G and M2GX respectively. Reference RNA for the DnaA depletion/induction series was additionally treated with DNase I (Hottes et al., 2004).
Arrays were scanned using a GenePix 4000B scanner (Axon Instruments), and spots were located and quantified using GenePix Pro 4.0 (Axon Instruments) software. Ratios were filtered and normalized as in (Hottes et al., 2004). Microarray data were visualized using Treeview (Eisen et al., 1998).
Identifying the effect of delaying dnaA expression on cell cycle transcription
Based on pair-wise correlation coefficients between expression profiles from different repetitions and the fold change in expression during each experiment, we identified 1482 genes that changed reproducibly during the DnaA depletion/induction cell cycle and 921 genes that behaved reproducibly during the wild-type cell cycle. Overall, 569 genes belonged to both sets. Using simulations with randomized profiles, the false discovery rate [(false positives)/(total discovered)] was estimated to be 1% (5.6/569) (Benjamini and Hochberg, 1995). The number of repeatable DnaA depletion/induction profiles is larger than the number of reproducible wild-type profiles because the preliminary selection did not discriminate between reproducible responses to the addition of xylose and reproducible responses to cell cycle progression. Expression profiles for genes that are neither cell cycle-regulated nor xylose-sensitive do not contain large, reproducible variations and were removed.
To characterize the transcriptional response to the addition of xylose, mixed cultures of wild-type (CB15N) cells grown in M2GX were subjected to a mock synchrony, which did not separate out the different cell types, and then were placed in M2G. The mock synchrony was similar to a Ludox density centrifugation synchrony (Quon et al., 1996) except that ice-cold M2 was used instead of Ludox. One RNA sample was taken after 70 min in M2G. After 75 min in M2G, xylose was added, making M2GX. Thirty minutes later (105 min from the beginning) a second RNA sample was taken. Four paired samples from independent cultures were compared using oligo microarrays.
A DnaA depletion/induction cell cycle expression profile's xylose response component was considered negligible compared with its cell cycle response component if one of two conditions were met. First, a profile was considered mainly cell cycle-responsive if the gene's expression changes during a part of the DnaA depletion/induction cell cycle when the cells were in a single media (i.e. M2G or M2GX) exceeded the gene's expression change in response to xylose addition. Second, a profile was considered cell cycle-responsive if the gene's immediate response to DnaA induction (xylose addition) was greater than its response to xylose addition alone. Genes with a rapid response to the synchronization process (i.e. a large expression change between 0 and 15 min) and no additional variation over the course of the experiment, are neither xylose nor cell cycle-responsive and did not meet the above criteria. Overall 267 genes had reproducible wild-type and reproducible, non-xylose DnaA depletion/induction expression profiles. Microarray data (Table S1) and additional details about data processing are available in Supplementary material and in the NCBI's GEO database under Accession number GSE3171.
Locating DnaA boxes
In searching for DnaA boxes using the MEME-generated position-specific scoring matrix (Bailey and Elkan, 1994), the likelihood that each sequence contained a DnaA box was compared with the likelihood that it was a typical Caulobacter intergenic sequence. Caulobacter intergenic sequence was represented using a third-order Markov model (Liu et al., 2001). Only high scoring motifs that were at least a 7/9 match to TTATNCACA, the E. coli consensus DnaA box (Messer and Weigel, 1997) were accepted. The middle base, which often is cytosine in E. coli, was not constrained because DnaA does not interact with it in a base-specific manner (Fujikawa et al., 2003). The cut-off for the presence of a DnaA box was set by trading off between the number of DnaA boxes found in random sequences and the number of DnaA boxes found upstream of the most strongly induced genes in the Post-DnaA set. The cut-off used yielded 7.7 false positives for every 100 promoters of random sequence (200 base pairs) generated using the background model. Similarly, the motif is present in the 200 bases upstream of the translational start sites of 6.7% of Caulobacter genes. As 40 genes in the Post-DnaA set were searched for DnaA boxes, we estimate that the 13 gene DnaA regulon identified contains ∼3 false positives. In addition to the one DnaA box upstream of ftsZ that met the cut-off, two other nearby possible DnaA boxes are also shown. Sequences were manipulated using Genome-tools (Lee and Chen, 2002).
We used an E. coli-derived model for DnaA boxes (seven of nine or better match to TTATNCACA) to determine that DnaA boxes are enriched upstream of the genes in the Post-DnaA set to avoid biasing the analysis by training the model on the same Caulobacter data that would later be tested. As the E. coli model lacks specificity (∼35% of Caulobacter genes have a seven of nine or better match in their upstream 200 bases), we used MEME and the best Caulobacter candidate DnaA boxes to refine the model before predicting the Caulobacter DnaA regulon.
Immunoblotting was carried out as described in (Chen et al., 2005). Each pair of blots (induced and non-induced) in Fig. 3 includes the same samples for 0, 20, 40 and 60 min. Data between corresponding blots were normalized using a linear model (Weisberg, 1985) based on the data for the shared time points. Experiments were performed at least twice.
Cells were fixed in 2.5% formaldehyde and 30 mM sodium phosphate buffer, pH 7.5 at the indicated times. Cells were either immobilized on agarose pads or placed directly on glass slides. Nomarski differential interference contrast (DIC) images were taken with a 100 × DIC objective on a Nikon E800 microscope with a 5 MHz Micromax 5600 cooled CCD camera controlled through Metamorph (Universal Imaging). Images were prepared using Adobe Photoshop and Metamorph version 4.5.
His6-DnaA gel shift
The complete coding sequence of the dnaA gene from the CB15 wild-type Caulobacter strain was amplified using primers 5′-CGTTAGCCCCGCAGCTTGCGCG-3′ and 5′-GGCG GTGGACATATGACCATGAAGGGC-3′. The polymerase chain reaction (PCR) product was digested with NdeI and HindIII and cloned into NdeI-HindIII-digested pET-28a(+) (Novagen). The resulting plasmid, pHis6DnaA, expresses an N-terminal His6-tagged DnaA protein. Plasmid preparation, DNA cloning and PCR amplifications were carried out according to manufacturers’ instructions.
His6-DnaA was over-expressed in the BL21 E. coli strain by inducing for 4 h at 30°C with isopropyl-β-D-thiogalactopyranosid and was purified using native conditions (Qiagen, 2003). Gel shift reactions were carried out as in (Ausubel et al., 2001). Details and oligos used are in Supplementary material.
We thank Swain Chen for constructing the His6-DnaA over-expression strain, Greg Marczynski (McGill University) for providing strain GM2471, Michael Laub and Melanie Prasol for advice on oligo microarray protocols, Maliwan Meewan for microarray printing, Justine Collier for critical reading of the manuscript, and Ji Park for computer support. This work was supported by ONR Grant N00014-02-1-053 to H.M., DOE Grants DE-FG03-01ER63219-A001 and DE-FG02-01ER63219 to H.M., and NIH Grants GM51426 and GM32506 to L.S.