Computational and Genetic Reduction of a Cell Cycle to Its Simplest, Primordial Components

Mathematical modelling and genetics in the bacterium Caulobacter crescentus identified redundancy in asymmetric cell cycle regulation through the dispensability of key transcription factor GcrA and methyltransferase CcrM, which together form a genetic module.


Introduction
Replicatively asymmetric cell cycles exist where the two distinct daughter cells resulting from cell division have distinct abilities to replicate their DNA. This is the case for the many Alphaproteobacteria that reproduce by asymmetric binary fission (e.g., Caulobacter and Brevundimonas species) or budding (e.g., Hyphomonas and Hyphomicrobium species) to produce a motile swarmer cell from a nonmotile stalked mother cell (see [1] and references therein) [2,3]. Swarmer cells do not replicate their DNA; they must first differentiate into stalked cells. During their motile juvenile phase, swarmer cells expend most of their energy on motility and little on growth [4,5]. Caulobacter crescentus [6], a species that is ubiquitous in water, has for many years been used as a model organism for the study of development and the cell cycle. There are also examples of nonstalked bacteria that exhibit the same asymmetry in replication and motility (e.g., Rhodopseudomonas palustris) [7,8]. Indeed, it has been proposed that morphological and functional asymmetry is much more widespread in the Alphaproteobacteria than previously thought [9]. This makes an understanding of asymmetric cell cycle regulation potentially even more relevant. However, the complexity of cell cycle control has made understanding the basic principles difficult. Here, we address this issue by using a minimal modelling approach to determine the core cell cycle regulatory circuit in Caulobacter crescentus.
Asymmetric division in C. crescentus yields a motile daughter swarmer (SW) cell and a sessile stalked (ST) cell. The ST cell immediately reinitiates replication, while the SW cell must differentiate into a ST cell before it can replicate and divide ( Figure 1A). These replicative and morphological asymmetries are, in part, controlled by the essential master regulator CtrA through its ability, when activated by phosphorylation (CtrA,P), to interact with DNA regulatory sequences in the origin of replication (Cori) and with many cell-cycle-regulated promoters [10,11]. A second regulator of Cori, DnaA, ubiquitous in bacteria as an essential replication initiator, also targets many cell-cycle-regulated promoters [11,12]. However, though DnaA levels are reduced in SW cells, they can support plasmid replication [13], indicating that it is likely not the regulator of replication asymmetry. This role is played by CtrA,P: high levels inhibit replication in SW cells, whereas low levels in ST cells allow replication to proceed [14]. Instead, DnaA appears to dictate the underlying frequency of replication [15]. Localization of the activator and stabiliser of CtrA, the essential membrane-bound hybrid histidine kinase CckA [16], specifically at the future SW cell pole, ensures a high level of CtrA,P in postdivisional SW cells and its removal in the ST compartment [17]. As C. crescentus regulates temporally both the abundance and activation of CtrA to control cell cycle progression [14], the cell cycle is very robust [18].
It has been proposed [19] that cell cycle progression in C. crescentus is controlled by a cyclical genetic circuit of four essential master cell cycle regulator proteins-DnaA, GcrA, CtrA, and CcrM-that are synthesised and degraded sequentially over the cell cycle. Here, we present a minimal mathematical modelling and experimental approach that challenges this assertion. Our model unexpectedly predicts that the ''essential'' cell cycle regulator GcrA is dispensable for core cell cycle progression. We experimentally test and verify this prediction. In addition, we experimentally uncover the dispensability of another cell cycle regulator, the methyltransferase CcrM, with simultaneous loss of the GcrA and CcrM module attenuating, rather than accentuating, cellular defects. Our conceptual approach resembles that applied to deciphering the minimal CDK control network in symmetrically dividing fission yeast [20], although here we study an inherently asymmetric cell cycle and also employ a mathematical modelling approach. We expect our results to hold in other Alphaproteobacteria, and our overall methodology should be useful in the dissection of more complex cell cycles.

Mathematical Model
Asymmetric cell cycles are dictated by the spatiotemporally varying concentration of a regulatory protein whose presence inhibits DNA replication initiation in the nonreplicating offspring, whereas its absence in the ''mother'' cell allows replication (or vice versa). This asymmetry must begin at or before the time of compartmentalisation. Since replication factors are generally cytoplasmic and hence likely diffuse and equipartitioned, this suggests the existence of an additional, localised protein that controls the first's activation and/or stability. With the above regulatory module, a diffuse regulatory protein, and its localised activator/stabiliser, basic regulatory control of an asymmetric cell cycle should be possible. However, previous mathematical models [21][22][23] of asymmetric cell cycles (e.g., in C. crescentus) have been much more complex and have not made experimentally verified predictions. Here, we therefore develop a simple, but strongly predictive, mathematical model of an asymmetric cell cycle, as applied to C. crescentus, constructed to include only minimal regulatory elements.
The model incorporates GcrA, CckA, and CtrA, but not DnaA or CcrM (see justifications below). The cell cycle regulator GcrA regulates genes involving DNA replication, division, and polar development [24]. Its synthesis is promoted by DnaA (see below) and repressed by CtrA,P. The ctrA gene has two promoters [25]: P 1 , activated by GcrA [24] but repressed by CtrA,P and silenced by full DNA methylation [26], and P 2 , a stronger promoter, activated by CtrA,P in a positive feedback loop. Halving of the P 1 methylation state (hemi-methylation), with associated subsequent P 1 activation, is due to movement of the DNA replication fork through the ctrA locus. This event is short in duration compared to other cell cycle timescales and is therefore modelled as a discrete event through the parameter S, which is switched from 0 to 1 at this time. We take the time at which CtrA,P levels drop below a low threshold as synchronous with the assembly of the replication machinery at Cori and take P 1 hemi-methylation to occur a fixed time later (the time required for replication initiation and subsequent movement of the replication fork past P 1 ). The DNA methyltransferase CcrM, which has been reported to be essential for viability, remethylates the P 1 promoter at the adenine within its GAnTC target site [27]. This remethylation (and as a result, silencing) occurs in late pre-divisional (PD) cells when P 1 is already repressed and after ccrM has been activated by CtrA,P. Therefore, for the model output, it is not relevant exactly when in late PD cells remethylation occurs, and we take it to be synchronous with compartmentalisation (for the ST compartment) or SW-ST differentiation (for the SW compartment), when the value of S is switched back to 0. The vital primary role of CcrM in resetting methylation-dependent promoters, as described above for the ctrA P 1 promoter, is essentially discrete. Hence, although switching of P 1 methylation is included, we do not explicitly include CcrM in our minimal model.
CckA initiates two phosphorelays leading to the phosphorylation (activation) and stabilisation of CtrA via a phosphotransferase, ChpT [28]. The active phosphorylated form, CckA,P, is localised primarily to the pole opposite the stalk ( Figure 1A) [29]. DivL, an essential noncanonical tyrosine kinase, activates, recruits, and colocalises with CckA [30,31], events for which replication initiation is a prerequisite (handled in the model by a dependence on the parameter S) [32]. Upon compartmentalisation, the phosphotransfer from CckA is cut-off in the ST compartment. As a result, CtrA,P is deactivated and removed from the ST cell progeny, while it remains stable and active in the SW cell ( Figure 1A). This distinction is the fundamental origin of the asymmetry between the ST and SW cells. Diffusive exchange of cytoplasmic molecules is largely unimpeded up until the last moments of constriction [33]. Hence we model compartmentalization as a discrete event again through the discrete parameter S, which is switched from 1 to 0 at compartmentalisation but only in the ST compartment. This ST compartment-specific switch is the origin of the ST/SW asymmetry in the model. In the SW compartment, S is also eventually switched back to 0, but only at a much later time corresponding to the unknown signals initiating SW to ST differentiation. Furthermore, because high CtrA,P levels activate the essential ftsQA cell division operon [34], we can take a high threshold level of CtrA,P as a proxy for compartmentalisation. In summary, our model consists of the minimal regulatory module (CtrA and CckA) suggested above, the cell cycle regulator GcrA, the experimentally described ctrA promoter regulation, and three discrete cell cycle events (replication, compartmentalization, and SW-ST differentiation).

Author Summary
Cell cycle regulation is remarkably complex and the fundamental principles difficult to understand, even in simple cells. The bacterium Caulobacter crescentus is a popular model organism to study cell cycle regulation due to the two different daughter cells resulting from cell division: a mobile ''swarmer'' cell and a ''stalked'' cell that adheres to surfaces. Here, we use mathematical modelling and genetic experiments to identify the core components of the asymmetric cell cycle of these bacteria. Using our mathematical model we predicted and confirmed experimentally that the transcription factor and cell cycle regulator, GcrA, hitherto thought to be essential, is in fact dispensable. We also identified another master regulator, the methyltransferase, CcrM as dispensable. Furthermore, simultaneous deletion of both GcrA and CcrM removes the severe cell division defects observed on either single deletion, returning cells to near wild-type morphology. We found that GcrA and CcrM constitute an independent, dispensable, genetic module that regulates transcription of cytokinetic proteins during the cell cycle. Phylogenetically, the module is conserved in Alphaproteobacteria, the class of Caulobacter, but is not present in the tree root of the class, suggesting that we have identified the primordial core of the asymmetric cell cycle regulatory circuit in the Alphaproteobacteria. It has been proposed [19] that methylation of the dnaA promoter by the CcrM methyltransferase promotes dnaA transcription, thereby restarting the cascade of cell cycle regulators with a surge in DnaA synthesis. However, mutation of the putative methylation site in the dnaA promoter does not significantly alter promoter activity [35]. It is therefore unknown what leads to the burst in DnaA synthesis prior to replication initiation. It has recently been suggested that DnaA controls the initiation frequency rather than the acquisition of replication competence [15]. Due to this uncertainty and the focus of our minimal model on asymmetry, we do not explicitly include DnaA in our minimal model; we instead assume that DnaA levels rise as CtrA(,P) levels drop and take a low threshold in CtrA,P levels as being synchronous with replication initiation. This assumption is justified by the observation that CtrA and DnaA generally have alternating profiles-that is, when one is low, the other is high and vice versa [36], even under starvation conditions [37]-and CtrA proteolysis coincides with DnaA binding to Cori [38]. In vitro, CtrA binding displaces DnaA from Cori, indicating competitive binding [38]. The wiring diagram, equations, and parameters of the model are shown in Figure 1B,C. Further description and justification of the model are given in Text S1, and model SBML files are provided in Texts S2 and S3.
In order to select model parameter values, we used existing literature measurements (see Text S1). Out of 22 parameters, two are removed by normalisation, six are obtained from experiments, and 10 have constraints placed on their values by experiments. To constrain the model parameters further, we quantitated the relative amount of GcrA and CtrA during the same cell cycle at higher time resolution than previously (Figure 2A,B) by semiquantitative immunoblotting and fitted our model to the profiles. The small number of unknown parameters ensures that our model is constrained by the available data (see Text S1). The model recapitulated the known effects of various experimental perturbations in the methylation state of ctrA P 1 or constitutive expression of GcrA (see Text S1 and Figure S2B). Comparisons between the theory and experiment are shown in Figure 2B and Figure S1, with good agreement.

Mathematical Model Predicts and Experiments Confirm That GcrA Is Not Essential
Although the above model is simple and reproduces many features of the C. crescentus cell cycle, it is still more complex than our previous minimal two component module. In particular, we have included GcrA and intricate regulation of CtrA through two promoters. Consequently, we pursued a minimal two component model for C. crescentus by removing GcrA and its regulation of ctrA P 1 . Previously it was reported that GcrA-depleted cells die [24]. Accordingly, in previous models removal of GcrA led to failed cell cycle control. However, strikingly, our minimal in silico model predicted a functional cell cycle without GcrA, albeit with an extended period ( Figure S2C). The model predicts that P 2 promoter feedback, combined with the active (forward) CckA phosphorelay, is strong enough to raise CtrA levels without P 1 .
Prompted by these predictions, we revisited previous experimental approaches used to conclude that GcrA is indispensable. We inactivated gcrA by generalized transduction of a DgcrA::V allele (conferring spectinomycin resistance) into the NA1000 wildtype (WT) strain and found that colonies appeared after ,6 d in rich medium ( Figure 2C). By contrast, transduction into a WT strain harboring pMT335-gcrA (gcrA on multicopy plasmid pMT335) gave clones after ,2 d. This 4-d growth delay may explain why gcrA was first described as essential in complex (PYE) medium [24]. On minimal (M2G) medium this delay is reduced to ,1 d. However, the CCNA_01269 (CC_1211, henceforth gcrB) gene of C. crescentus encodes an uncharacterized gcrA paralog with 44% sequence identity to gcrA ( Figure S3). While ectopic expression of gcrB (from pMT335-gcrB) can partially compensate for the absence of gcrA, transduction of DgcrA::V into cells with an in-frame deletion in gcrB (DgcrB) gave similar results as transduction into WT cells ( Figure 2C). PCR analysis ( Figure S4A) and immunoblotting ( Figure 2G) confirmed that transductants had exchanged the gcrA gene with the DgcrA::V allele. Importantly, the pleiotropic phenotypic defects of DgcrA::V cells (see below) were corrected by pMT335-gcrA, partially corrected by pMT335-gcrB but not by the empty vector ( Figure S4B). Overall, these experiments confirm a key prediction of our model that GcrA is inessential for cell cycle progression.

Consequences of GcrA Deletion
We first focused on the cell doubling time during exponential growth, as determined by the optical density of a liquid culture. This time is dependent on the type of growth medium used. We found that the doubling time of DgcrA::V cells was 75% longer than for the WT in PYE, and 40% longer in M2G, qualitatively consistent with the model ( Figure 2D,E). The cells also exhibited a lengthened lag phase (unpublished data). However, DgcrA::V cells have an origin-to-terminus ratio similar to WT cells, demonstrating that this lengthened doubling time is not due to a defect in DNA replication initiation ( Figure 2F, see below). Consistent with these results, fluorescence-activated cell sorting (FACS, Figure S5) revealed an increase in cell length ( Figure S5A) and in chromosome number ( Figure S5B) in DgcrA::V versus WT cells. The increase in chromosome number scales with the increase in cell length, which arises from perturbed cytokinesis ( Figure 2D,E, see below). Immunoblotting showed a reduction in levels of the MipZ division regulator and the essential late cell division protein FtsN ( Figure 2G). Finally, the difference in cell length in DgcrA::V versus WT cells was attenuated in M2G (176% of WT) compared to PYE (264% of WT), consistent with the respective increases in doubling time described above.
We also observed that DgcrA::V cells do not undergo the cellcycle-regulated switch in buoyancy that is exploited to separate swarmer and stalked cells ( Figure S4B) and are poorly motile on soft (0.3% PYE) agar ( Figure S4C) and in broth (unpublished data). Additionally, DgcrA::V cells are resistant to the S-layerspecific bacteriophage WCr30 and the pilus-specific bacteriophage WCbK ( Figure S4B) consistent with reduced levels of the pilin subunit PilA, the polarity factor PodJ (both required for pilus assembly), and of the S-layer subunit RsaA in DgcrA::V versus WT cells ( Figure 2G).
The abundance of CtrA, DnaA, and CcrM was also altered in DgcrA::V cells: while CtrA was diminished (as expected from our model), the abundances of CcrM and DnaA were elevated ( Figure 2G). Importantly, immunoblotting revealed identical defects on protein abundance seen in DgcrA::V already after 5 h of GcrA depletion using the xylose-inducible promoter (P xyl ). These defects were still present 24 h after depletion but were reversed following re-instatement of gcrA expression for 16 h ( Figure S6A,C).

Lack of FtsN Is the Major Cause of Defects in DgcrA::V Cells
To test if one or a combination of these abnormalities impairs growth of DgcrA::V cells, we screened for DgcrB DgcrA::V cells mutagenized with an himar1 transposon (Tn) that form colonies faster than the parent. Backcrossing and mapping identified nine distinct Tn insertions, eight of which were in either the 59 end of ftsN or in the 39 end of the upstream gene, CCNA_02087 (CC_2008), that reads in the same direction as ftsN ( Figure 3A). These eight P ftsN ::Tn insertions attenuate the growth and division defect of DgcrA::V cells as demonstrated by DIC imaging (cf., Figure 3D with Figure 2D) and by FACS ( Figure S5A,B). Moreover, immunoblotting revealed that these insertions restore FtsN to near WT levels ( Figure 3F), presumably because of an outwardly facing promoter of the Tn directing ftsN transcription. As before, the origin-to-terminus ratios were similar to the WT ( Figure 3E). We also confirmed that transduction of DgcrA::V into WT cells expressing FtsN from P van on pMT335 (pMT335-ftsN) or from P xyl at the chromosomal xylX locus (DftsN xylX::P xyl -ftsN) gave colonies after ,3 d on rich medium (unpublished data), compared to ,6 d for the WT expressing FtsN from the endogenous promoter. Consistent with functional interaction of ftsN and gcrA, accumulation of ftsN mRNA was shown before to be GcrAdependent [24]. Using a lacZ-based transcriptional reporter plasmid, we confirmed that P ftsN indeed requires GcrA for full activity. After 5 h and 24 h of GcrA depletion, P ftsN -lacZ is reduced   to 52% and 46% of WT activity, respectively ( Figure 3G). Chromatin immunoprecipitation using antibodies to GcrA followed by deep-sequencing (ChIP-Seq) revealed that GcrA binds the ftsN promoter (P ftsN ) in vivo ( Figure 3C), suggesting that ftsN is a direct target of GcrA. In sum, activation of P ftsN by GcrA is critical for efficient growth and division, and reduced FtsN levels cause growth defects in DgcrA::V cells.
ccrM Is Dispensable for Viability Surprisingly, one Tn of the nine insertions was found in the middle of the ccrM gene (ccrM::Tn, Figure 4A). We performed complementation experiments and found that WT cells expressing ccrM from P van on pMT335 (pMT335-RBS-ccrM) formed colonies on PYE ,3 d after transduction of ccrM::Tn compared to ,5 d for WT cells harbouring the empty vector (unpublished data). We also found that ccrM::Tn could be transduced into WT cells on PYE and that genomic DNA extracted from the ccrM::Tn mutant is susceptible to cleavage by the methylation-sensitive restriction enzyme Hinf1, as is the case for genomic DNA extracted from DccrM::V cells ( Figure S7A) [39]. Immunoblotting with antibodies to CcrM provided further confirmation that ccrM::Tn is a null allele ( Figure 4A). Therefore, ccrM, like gcrA, is dispensable for viability, consistent with the recent report by Gonzalez and Collier [40]. However, ccrM::Tn colonies take 4 d to form on PYE, similar to DccrM::V, and present a lengthened doubling time ( Figure 4B). This slow growth rate may explain why ccrM was previously reported to be essential [27]. As before, neither the DgcrA::V mutation nor the ccrM::Tn mutation affect the origin-to-terminus ratio compared to the WT ( Figure 4C).

Tn-Insertions in ccrM Are Greatly Overrepresented in
DgcrA::V Cells as Compared to WT To quantitatively evaluate the relationship between gcrA and ccrM, we determined the relative frequency of Tn insertions in WT and DgcrA::V mutant cells by Tn-Seq following himar1 Tn mutagenesis. Tn insertions were hugely overrepresented along the ccrM coding sequence for DgcrA::V cells compared to the WT ( Figure S7B), being ,115 times more abundant than the average of insertions over other coding sequences ( Figure 4D). By contrast, Tn insertions in scpAB and ftsE, both with promoters bound by GcrA in vivo based on ChIP-Seq (Table S1), are underrepresented in DgcrA::V compared to the WT ( Figure 4E). Insertions in the region upstream of ftsN were also found to be greatly overrepresented in DgcrA::V cells compared to the WT ( Figure S7C) and were even more frequent than in the ccrM sequence ( Figure S7B) consistent with the number and location of the nine insertions found in the Tn suppressor screen (Figures 3A and 4A). We also observed an increased bias in insertions in the gcrB promoter region (Table S3), Returning to the ccrM::Tn insertion found in the screen, we discovered that it greatly improves the cytokinetic defect of DgcrB DgcrA::V cells ( Figure 4B). This was quantified with FACS analyses ( Figure S5), which showed a substantial reduction in cell length and chromosome number, and with a variance much closer to WT than that of DgcrB DgcrA::V P ftsN ::Tn2 cells. We also observed a reduced lag phase (unpublished data), though there was little or no improvement in the doubling time. The presence of stalked cells in DIC images indicted that morphological asymmetry is at least partially maintained in DgcrB DgcrA::V ccrM::Tn cells ( Figure 4B). To investigate this further, we examined the localisation pattern of the stalked pole-specific marker SpmX [41] in DgcrA::V ccrM::Tn cells and found a unipolar focus of SpmX-mCherry at the same site as the stalk (Figure S8A), confirming that morphological asymmetry is maintained. Consistent with elevated steady-state levels of PilA ( Figure 4A), we also observed that the resistance of DgcrA::V cells to WCr30 and WCbK is diminished by the ccrM::Tn mutation ( Figure  S4B). Lastly, we addressed replicative asymmetry by studying the localisation pattern of the centromere-binding protein ParB, which binds to the parS site near the origin of replication. Localisation of GFP-ParB [42] revealed an uneven number of foci in DgcrA::V cells, consistent with the presence of replicative asymmetry ( Figure S8B). Most importantly, time-lapse imaging of GFP-ParB in DgcrA::V ccrM::Tn cells showed asymmetric duplication and segregation of GFP-ParB ( Figure S8C), strongly suggesting that replicative asymmetry is maintained in these cells.

ftsN Is Regulated by CcrM Methylation
We next explored the regulatory basis for the Tn insertion enrichment. The results of the suppressor screen and Tn-seq suggest that lack of FtsN is a significant limiting growth factor in DgcrB DgcrA::V cells. Therefore, we reasoned that the recovery due to the ccrM::Tn insertion might, at least partially, be mediated through ftsN. Accordingly we noted that P ftsN also harbors a GAnTC methylation site. We found in ChIP-seq experiments that a polyclonal antibody to N6-methyladenosine (m6A) [39] precipitated P ftsN efficiently from WT but not from DccrM::V cells ( Figure 3B), indicating that P ftsN carries a CcrM-dependent m6A mark overlapping the GcrA target site. To evaluate the importance of this methylation site, we mutated the GAnTC site to GTnTC and found in lacZ-promoter probe assays ( Figure 3H) that the mutant (P ftsN* ) promoter fires only at 14% of the WT rate in PYE (25% in M2G), consistent with reduced FtsN levels when ccrM is disrupted ( Figure 4A, lanes 1 and 6). Moreover, P ftsN* activity doubles by 24 h after depletion of GcrA from DgcrB cells ( Figure 3G), consistent with increased FtsN levels in the DgcrB DgcrA::V ccrM::Tn mutant compared to ccrM::Tn ( Figure 4A, lanes 5 and 6), while, as noted earlier, the activity of WT P ftsN responds in the opposite fashion ( Figure 3G). However, the activities, in absolute units, of P ftsN and P ftsN* 24 h after depletion of GcrA in the DgcrB background are very similar (63768 and 60866 Miller units, respectively). This would suggest that the methylation state of P ftsN has an effect only in the presence of GcrA and that therefore the recovery in FtsN levels observed in DgcrB DgcrA::V ccrM::Tn cells compared to DgcrB DgcrA::V ( Figure 4A, lanes 4 and 5) is probably not mediated through the ftsN promoter (assuming the mutation does not have unwanted side effects on the firing of the core promoter), at least on a low-copy lacZ-reporter plasmid. Because methylation is transient, the net effect on P ftsN may result in the same measured activity as P ftsN* , but the timing of methylation could be important or the results may be skewed due to a low level of GcrA expression being maintained from P xyl even under repressive conditions that is absent in the DgcrB DgcrA::V ccrM::Tn mutant. However, taken together, these data suggest that ftsN is positively regulated by GcrA in the presence of methylation but negatively regulated in its absence.
FtsN abundance in DgcrB DgcrA::V ccrM::Tn is still considerably lower than in WT or DgcrB DgcrA::V P ftsN ::Tn cells, even though the aberrant division of DgcrB DgcrA::V cells is largely repaired. The improved variance in cell length and chromosome number and reduced lag in growth of DgcrB DgcrA::V cells carrying the ccrM::Tn versus the P ftsN ::Tn mutation (see above) is likely due to the pleiotropic nature of the mutation, elevating the expression of many division genes to some extent. This is consistent with the somewhat raised steady state levels of CtrA and MipZ in DgcrB DgcrA::V ccrM::Tn cells compared to DgcrB DgcrA::V ( Figure 4A,  lanes 4 and 5).
Because ccrM mutants also exhibit a reduction in FtsN abundance compared to WT cells ( Figure 4A), we tested if transduction of DccrM::V into DftsN xylX::P xyl -ftsN yielded colonies on PYE (with 0.3% xylose) similar to the DgcrA::V mutation. Transduced DftsN xylX::P xyl -ftsN colonies appeared after ,4 d, whereas transduced WT colonies expressing FtsN from the endogenous promoter only appeared after 5-6 d (unpublished data). Thus, the growth defect of ccrM mutants can be improved by expression of extra FtsN (in addition to extra FtsZ [40]).

The Mathematical Model Provides an Explanation for the Lack of Recovery in Doubling Time
Disruption of ccrM ameliorates the cytokinetic and morphological defects of DgcrB DgcrA::V cells but not the doubling time ( Figures 2D and 4B). Returning to our model, we asked if these experimental findings could be recapitulated. We found in our simulations that maintaining ctrA P 1 in the unmethylated state (mimicking the loss of ccrM) in DgcrA cells caused only a slight change in cell cycle timing compared to the loss of GcrA alone ( Figure 4F), with a ,3% decrease in the swarmer cell cycle period, consistent with the experimentally observed trend ( Figure 4B). Without methylation to suppress early activation, our model suggests that basal ctrA P 1 transcription in DgcrA cells results in premature synthesis of CtrA and so has a negative effect on cell cycle progression, which cancels the positive effect during CtrA reaccumulation. The same neutral effect on doubling time was also seen in a WT background ( Figure S2B) consistent with previous results (Text S1) [26]. Our model therefore provides a possible explanation for why the decrease in cell doubling time is small, even though DgcrB DgcrA::V ccrM::Tn cells otherwise show substantial phenotypic recovery versus DgcrB DgcrA::V cells.

Discussion
Our combined minimal modelling and forward genetics approach in C. crescentus has unexpectedly uncovered gcrA and ccrM as a dispensable genetic module. Strikingly, the core C. crescentus asymmetric cell cycle network can therefore function without two of the four ''master'' regulators, revealing a high level of robustness. Though both gcrA and ccrM are highly conserved, it is much more common for both to be present or absent in the alphabacterial lineages rather than either gene being present alone ( Figure 5A) [43], which supports our findings. Furthermore, gcrA and ccrM are not present in the tree root of the Alphaproteobacteria, whereas ctrA and cckA are [43], suggesting that our minimal model describes the cell cycle of the primordial Alphaproteobacterium. Within our minimal cell cycle network, we find that asymmetry can be controlled with just two fundamental components: an inhibitor of DNA replication initiation (CtrA) and a localised activator that controls the former's cell-type-specific activation (CckA). In Figure 5B we present a schematic of the current biological model of cell cycle regulation in C. crescentus with the dispensable GcrA/CcrM module highlighted. Several elements have yet to be fully understood (indicated by question marks): What triggers the pulse in DnaA concentration at the beginning of the cycle? Is this mediated by the Lon protease [44] and/or transcriptional control of dnaA [19] The co-evolution of GcrA and CcrM and our results on ftsN regulation are consistent with the recent discovery that promoter binding and transcriptional activation by GcrA is methylationdependent [39]. Connecting transcriptional activation to chromosome position and cell cycle timing (via the hemi-methylation caused by the passing of the replication fork) potentially contributes greatly to cell cycle robustness, an important property especially for oligotrophic bacteria. From the point of view of the circuit shown in Figure 5B the GcrA/CcrM module could contribute to robustness in two ways: (1) Methylation-regulated transcription of ctrA P 1 controls the timing and prevents the early accumulation of CtrA, which would hinder cell cycle progression; (2) GcrA provides an additional connection between DnaA and CtrA, resulting in more reliable accumulation and activation of CtrA after DNA replication initiation, in addition to the link between DNA replication initiation and CckA localisation/ activation.
The potential role of the GcrA/CcrM module in the robustness of the timing of cell cycle events may also explain why its absence is primarily observed in obligate endosymbionts and pathogens, as are found in the Rickettsiales. These bacteria have a long generation time as an adaption to the slower growth of the host cell. For example, Rickettsia prowazekii has a doubling time of about 10 h. This is likely reflected in a more relaxed coordination of cell cycle events with replication fork progression, coordination for which the GcrA/CcrM module is of great importance. Thus, without strong selection, it is perhaps not surprising that this module was lost during the evolution of this branch.
The success of our approach in reducing an asymmetric cell cycle to its simplest, primordial components illustrates the potential to define the core regulation of sophisticated cell cycles using this strategy. This methodology would be especially useful for eukaryotic cell cycles. The complexity of their regulatory networks has made understanding the basic principles involved a challenging task. Mathematical models of eukaryotic cell cycles, such as those of fission and budding yeast [45][46][47], have made limited progress in simplifying the regulatory networks or identifying essential, core components. Indeed, progress on identifying redundancy in the cell cycle circuitry has been led by experiments rather than by modelling [20,48]. We therefore expect that our minimal modelling approach could play an important role in dissecting these more complex cell cycles.

Bacterial Strains, Plasmids, and Oligonucleotides
Bacterial strains, plasmids, and oligonucleotides used in this study are listed and described in tables below.

Plasmid Constructions
pNPTS138-DgcrB-KO. The plasmid construct used for gcrB (CCNA_01269 or CC_1211) deletion was made by PCR amplification of two fragments. The first, a 633 bp fragment flanked by an EcoR1 site at the 59 end and a BamHI at the 39 end (amplified using primers delgcrB_1-EcoRI and delgcrB_1-BamHI), encompasses the upstream region of gcrB and extends to 15 bp downstream of the predicted start codon. The second fragment, flanked by a BamHI site at the 59 end and a HindIII site at the 39 end (amplified using primers delgcrB_2-BamHI and delgcrB_2-HindIII), harbors the last 6 bp of the gcrB coding sequence and extends 629 bp downstream of the gene. These two fragments were first digested with appropriate restriction enzymes and then triple ligated into pNTPS138 (M.R.K. Alley, unpublished) that had been previously restricted with EcoRI and HindIII.
pNPTS138-DrsaA. The plasmid construct used for rsaA (CCNA_01059 or CC_1007) disruption was made by PCR amplification of a 691 bp fragment that overlaps the rsaA coding sequence (+1,500 to +2,191 relative to the TTG start codon, amplified using primers delrsaA-EcoRI and delrsaA-HindIII). This fragment is flanked by a HindIII site at the 59 end and an EcoRI at the 39 end, and then cloned into corresponding sites of the pNTPS138 after hydrolysis with the appropriate restriction enzymes.
placZ290-pftsN. The ftsN promoter region (2469 to +228 relative to the ATG) was PCR-amplified using pftsN-EcoRI and pftsN-XbaI primers using NA1000 chromosomal DNA as a template. This fragment was digested by appropriate enzymes and cloned into a EcoRI-XbaI-digested placZ290 promoter probe vector.
placZ290-pftsN*. Same as previous, using the same pftsN-EcoRI and pftsN-XbaI primers and synthesized pftsN* DNA fragment (DNA2.0 Inc, Menlo Park, CA) as a template where the  Figure 2 and Table S2 of [43]). One species of each genus analysed is shown (37 out of a total of 65 genomes). (B) Current biological model of cell cycle regulation in Caulobacter crescentus, with the dispensable GcrA/CcrM module highlighted. Question marks indicate elements of the circuit that have yet to be fully elucidated. doi:10.1371/journal.pbio.1001749.g005 A position (252 relative to the ATG) of the GATTC site is replaced by a T.

Strain Constructions
NA1000 DgcrB. pNPTS138-DgcrB-KO was first introduced into NA1000 (WT) by intergeneric conjugation and then plated on PYE harboring kanamycin (to select for recombinants) and nalidixic acid to counter select E. coli donor cells [50]. A single homologous recombination event at the CCNA_01269 locus of kanamycin-resistant colonies was verified by PCR. The resulting strain was grown to stationary phase in PYE medium lacking kanamycin. Cells were then plated on PYE supplemented with 3% sucrose and incubated at 30uC. Single colonies were picked and transferred in parallel onto plain PYE plates and PYE plates containing kanamycin. Kanamycin-sensitive cells, which had lost the integrated plasmid due to a second recombination event, leaving a deleted version of gcrB behind (DgcrB), were then identified for disruption of the gcrB locus by PCR.
NA1000 DrsaA::pNPTS138. As above, the pNPTS138-DrsaA plasmid was introduced into NA1000 by intergeneric conjugation. Kanamycin-resistant ex-conjugants having undergone a single homologous recombination event were isolated and the integration was verified by PCR.
QCr30 Transduction of the DgcrA Mutation DgcrA::V (Spc R ) transducing phage stock is a QCr30 lysate of LS3707 [24]. Overnight cultures of NA1000 and NA1000 DgcrB strains harboring or not pMT335, pMT335gcrA, or pMT335gcrB plasmids were first washed with fresh liquid medium (PYE or M2G) and resuspended at a concentration of 10 9 cfu/ml. We infected 0.5 ml of cells with 50 mL of QCr30 phage stock (,10 10 pfu/ml), incubated them for 2 h at room temperature, and then plated them on solid PYE or M2G containing spectinomycin/streptomycin antibiotics. Plates were incubated at 30uC, and visible colonies were counted each day. Experimental values represent the average of three biological replicates. Addition of 50 mM vanillate to the plates, to further induce GcrA or GcrB synthesis from pMT335 plasmids (that harbor gcrA or gcrB under the control of the vanillateinducible promoter, Pvan [54]), did not change the transduction values, presumably due to leaky expression from Pvan. Integration of the DgcrA::V construction was checked by PCR and confirmed by immunoblot analysis. PCR was done using pro-gcrA and gcrA-EcoRI primers that allow amplification of the gcrA gene only in strains that do not carry the DgcrA::V allele. Strains harboring the pMT335gcrA complementation plasmid served as a positive control of QCr30 transduction efficiency. The fact that the ratio of NA1000+pMT335 versus NA1000+pMT335gcrA colonies counted on PYE or M2G plates are 0.38 and 1.05, respectively, indicates that spontaneous suppressors are not frequent. In support of this, genome sequencing of several DgcrA::V derivatives failed to reveal suppressor mutations.

Cori/ter Ratio Determination
Cells were grown to exponential phase in PYE or M2G medium and then harvested. Chromosomal DNA was extracted using Ready-Lyse Lysozyme solution (Epicentre Biotechnologies) and DNAzol Reagent (Invitrogen) and then precipitated in 100% ethanol. The DNA pellet was washed 3 times with 70% ethanol and incubated at room temperature in 8 mM NaOH solution for 4 h. HEPES 1 M (pH 7) was added to neutralize the pH. Corifwd/Cori-rev and Ter-fwd/Ter-rev primers were used to amplify ,170 bp DNA close to the origin (Cori) or close to the terminus (ter), respectively, of the Caulobacter crescentus chromosome [55]. Real-time PCR was performed using the Step-One Real-Time PCR system (Applied Biosystems) at different DNA dilutions (5 mL), with 12.5 mL of SYBR green PCR master mix (Quanta Biosciences), 0.5 mL of each primer (10 mM), and 6.5 mL of water per reaction. PCR assay parameters were one cycle at 95uC for 5 min followed by 40 cycles at 95uC for 15 s, 55uC for 20 s, and 60uC for 15 s. Dilutions of NA1000 extracted DNA were used to generate Cori and ter standard curves, involving all Cori/ter ratios being normalized to the WT value. Average values are from triplicate measurements from two independent DNA extractions.
b-Galactosidase Assays b-Galactosidase assays were performed at 30uC as described previously [52,56]. We lysed 50 mL of washed cells at OD 660 nm = 0.1-0.6 with chloroform and mixed them with 750 mL of Z buffer (60 mM Na 2 HPO 4 , 40 mM NaH 2 PO 4 , 10 mM KCl, and 1 mM MgSO 4 heptahydrate). We added 200 mL of ONPG (4 mg/ml o-nitrophenyl-b-D-galactopyranoside in 0.1 M KPO 4 pH 7.0), and the reaction was timed. When a medium-yellow color developed, the reaction was stopped with 400 mL of 1 M Na 2 CO 3 . The OD 420 nm of the supernatant was determined and the units were calculated with the equation: U = (OD 420 nm * 1000)/(OD 660 nm * time (in min) * volume of culture (in ml)). For GcrA depletion experiments in M2G using strain NA1000 DgcrB DgcrA::V xylX::Pxyl-gcrA, M2G supplemented with 0.3% xylose overnight cultures were harvested and washed 3 times with M2 minimal salt solution, and then restarted in appropriate M2G or M2GX medium for 5 or 24 h at 30uC. For the 24 h time point, culture dilutions were done to maintain cells in exponential growth phase. Experimental values represent the averages of four independent experiments.

GcrA Depletion Experiment
For GcrA depletion experiments, overnight cultures of strain NA1000 DgcrB DgcrA::V xylX::Pxyl-gcrA grown in M2G supplemented with 0.3% xylose were harvested and washed 3 times with M2 minimal salt solution, and then resuspended in M2G (GcrA depletion) or M2GX (GcrA expression) medium for 2, 5, or 24 h at 30uC ( Figure S6A). Then, the 24 h M2G culture was supplemented with 0.3% xylose and incubated with the 24 h M2GX culture for an additional 16 h at 30uC. For the 24 h and 40 h time points, culture dilutions were done to maintain cells in exponential growth throughout the experiment.

Immunoblot Analysis
Protein samples were separated by SDS-PAGE and blotted on PVDF (polyvinylidenfluoride) membranes (Merck Millipore).

RsaA Acid Extraction
Cell cultures of 5 ml were grown to exponential phase in M2G medium and then harvested. Cells were washed twice with 5 ml of 100 mM HEPES (pH 7.2) and then resuspended in 200 mL of 100 mM HEPES (pH 2). After 10 min of incubation at room temperature, cells were pelleted and removed. The supernatant pH was neutralized by adding 3 mL of 5N NaOH solution. Samples were separated on 7.5% acrylamide SDS-PAGE gel followed by Coomassie Blue staining.

Microscopy
PYE or M2G cultivated cells in exponential growth phase were immobilized using a thin layer of 1% agarose. For time-lapse experiments, LT494 cells in exponential growth phase were immobilized using a thin layer of PYE supplemented with 1% agarose. Fluorescence and contrast microscopy images were taken with an Alpha Plan-Apochromatic 1006/1.46 DIC(UV) VIS-IR oil objective on an Axio Imager M2 microscope (Zeiss) with 405 and 488 nm lasers (Visitron Systems GmbH, Puchheim, Germany) and a Photometrics Evolve camera (Photometrics) controlled through Metamorph V7.5 (Universal Imaging). Images were processed using Metamorph V7.5.

FACS
Cells in exponential growth phase (OD 660 nm = 0.3-0.6), cultivated in PYE or M2G, were fixed in ice cold 70% Ethanol solution. Fixed cells were resuspended in FACS staining buffer pH 7.2 (10 mM Tris-HCl, 1 mM EDTA, 50 mM NaCitrate, 0.01% TritonX-100) and then treated with RNase A (Roche) at 0.1 mg/ml for 30 min at room temperature. Cells were stained in FACS staining buffer containing 0.5 mM of SYTOX Green nucleic acid stain solution (Invitrogen) and then analyzed using a BD Accuri C6 flow cytometer instrument (BD Biosciences, San Jose, CA). Flow cytometry data were acquired and analyzed using the CFlow Plus V1.0.264.15 software (Accuri Cytometers Inc.). We analyzed 20,000 cells from each biological sample. The forward scattering (FSC-A) and Green fluorescence (FL1-A) parameters were used to estimate cell sizes and cell chromosome contents, respectively. Relative chromosome number was directly estimated from the FL1-A value of NA1000 cells treated with 20 mg/ml Rifampicin for 3 h at 30uC. Rifampicin treatment of cells blocks the initiation of chromosomal replication, but allows ongoing rounds of replication to finish.

Cell Generation Time Determination
Cell growth in PYE or M2G medium was done in an incubator at 30uC under agitation (190 rpm) and monitored at OD 660 nm . Generation time values were extracted from the curves using the Doubling Time application (http://www.doubling-time.com). Values represent the averages of at least three independent clones.
HiSeq 2000 runs of barcoded ChIP-Seq libraries yielded several million reads that were mapped to Caulobacter crescentus NA1000 (NC_011916) according to the ELAND alignment algorithm (services provided by Fasteris SA, Switzerland). The standard genomic position format files (BAM) were imported into SeqMonk (Braham http://www.bioinformatics.babraham.ac.uk/projects/ seqmonk/, version 0.21.0) to build sequence read profiles. The initial quantification of the sequencing data was done in SeqMonk to allow the comparison of different conditions and to isolate regions of interest. To this end, the genome was subdivided into 50 bp probes, and for every probe an associated value was calculated, a value that derives from the pattern of reads that occurs within the probe region used for the quantitation using the Red Count Quantitation option. To discern between background signal (modeled with a Poisson or negative binomial distribution) and candidate peaks, we calculated the ratio of reads per probe as a function of the total number of reads. The overall average read count (for all probes) plus twice the standard deviation was used to establish the lower cut-off that separates the background from candidate peaks. Analyzed data are provided in Table S1 with selected peak values highlighted in yellow. Figure 3C focuses this analysis on the ftsN and CCNA_02087 region (2,235,000 to 2,238,000 bp on the Caulobacter crescentus genome). Figure 3B used global m6A ChIP-Seq analysis data obtained in the laboratory [39].

Transposon Suppressor Screen and Mapping
An overnight DgcrB DgcrA::V cell culture was grown in PYE and mutagenized with a himar1 transposon (Tn) [62]. To create the himar1 strain collection, transposition was induced by mobilizing the himar1 transposon (Kan R ) from plasmid pHPV414 in Escherichia coli S17-1 into the NA1000 DgcrB DgcrA::V strain and selecting for kanamycin-nalidixic acid-resistant Caulobacter clones that form colonies faster than the parent at 30uC (,5/6 d) on PYE. Nine distinct Tn clones appearing after ,3 d were selected and already showed, under the microscope, less filamentous defects than the parent strain. Using QCr30-mediated generalized transduction, the Tn insertions were backcrossed in NA1000 DgcrB followed by transduction of the DgcrA::V strain to verify the phenotypes. Chromosomal DNA of the nine selected suppressors was extracted and partially digested for 4 min with HinP1 restriction. Digested DNA was recircularized by T4 DNA ligase (Roche) treatment and then electroporated in E. coli EC100D pir+ 116 (Epicentre Biotechnologies). Plasmids of kanamycin-resistant clones were extracted and mapped using the himar-Seq2 primer that allowed the sequencing of the DNA region on the Caulobacter crescentus chromosome adjacent to the Tn insertion region.

Transposon Suppressor Screen Coupled to Deep Sequencing (Tn-Seq)
Tn collections of .100,000 kanamycin-nalidixic acid-resistant clones were collected for NA1000 (WT) and DgcrA::V strains, with the same protocol as previously described [62]. For each collection, all Tn clones were mixed and chromosomal DNA extracted. This DNA was used to generate a barcoded ChIP-Seq library and submitted to Illumina HiSeq 2000 sequencing. Tn insertion-specific reads (50 bp long) were sequenced using the himar-Tnseq primer and yielded several million reads that were mapped to Caulobacter crescentus NA1000 (NC_011916) [63] according to the ELAND alignment algorithm (Johann Mignolet, Methods, manuscript in preparation). The Tn insertion coordinate format files (BED) were generated and imported into SeqMonk V0.21.0 for analyses. Table S3 is an excel table and includes NA1000-TnSeq and DgcrA::V-TnSeq data, where values are assigned for each Tn integration position on the Caulobacter crescentus chromosome. Both datasets were normalized to the total number of reads of the largest dataset. This file was used to generate Figure S7B and Figure S7C that are respectively focused on the ccrM and ftsN regions. Table S2 is also an excel table and includes DgcrA::V-TnSeq/NA1000-TnSeq ratio analyses. To this end, the annotated genome was subdivided into coding sequence (CDS) probes (precluding the analysis of noncoding sequences such as promoter regions), and for every probe, an associated Tn insertion value was calculated. Both datasets were normalized to the total number of reads of the largest dataset. An average value of all CDS-Tn insertions normalized to the gene size was calculated, and 1% of this normalized value was used to correct each CDS-Tn insertion value. This correction prevents, during ratio calculations, a CDS-Tn insertion value of 0 and excludes also from the analyses CDS that do not share sufficient statistical Tn insertions. This file was used to generate Figure 4D and 4E.  ) plates (supplemented with spectinomycin 30 mg/ml and streptomycin 5 mg/ml) were screened by PCR using primers Pro-gcrA and gcrA-EcoRI, revealing that the majority of the 17 tested colonies had exchanged endogenous gcrA with DgcrA::V. This primer pair amplifies a 594 bp fragment when endogenous gcrA is present, while only a primer dimer band is seen for the DgcrA::V background. As positive (C+) and negative (C2) controls, PCR amplifications were also carried out with NA1000 and DgcrA::V, xylX::P xyl -gcrA cells. (B) Sensitivity of WT and mutant cells to the Slayer specific phage QCr30 and the pilus-specific phage QCbK. Serial dilutions of QCr30 and QCbK were spotted on lawns of cells embedded in the top-agar on PYE plates. Spot tests on WT and gcrA mutants carrying various plasmids (as indicated) are shown in the frame. The first column shows controls with WT, flagellin (Dfljx6), pilus (DpilA), and rsaA mutants (DrsaA). The schematics to the right of the spot test represent test tubes showing the buoyancy of the strain. WT cells give two buoyancies (high and low). The mutants are affected in this property. The thickness of the bands reflects the number of cells obtained with a given buoyancy. (C) Motility assays of WT and various mutants on PYE soft (0.3%) agar plates. Complementation experiments of WT and gcrA mutants carrying various plasmids (as indicated) are shown in the frame. The first column shows negative control of swarming using a flagellin (Dfljx6) mutant. (EPS) Figure S5 Relative cell size and chromosome number distributions analysed by FACS for NA1000 and various mutants. (A) Relative cell size distributions in NA1000 and various mutants cultivated to exponential phase in PYE or M2G medium were analysed by flow cytometry. Forward scattering values (FSC-A) were used to estimate the cell size. A total of 20,000 cells were analysed for each population, and a density heat map was used to represent the cell population distribution as a function of the FSC-A parameter. Median values are indicated (white lines) and normalized to the median value of NA1000 cultivated in PYE. (B) Same FACS samples acquired in (A) were analysed using the green fluorescence parameter (FL1-A) to estimate the relative chromosome number distribution. Additional NA1000 sample treated with Rifampicin (+Rif) was used to obtain a correlation between the FL1-A parameter and the relative chromosome number. A total of 20,000 cells were analysed for each population, and a density heat map was used to display the cell population distribution as a function of the FL1-A parameter. Median values are indicated (white lines). (EPS) Figure S6 Depletion experiment supports the finding that GcrA is not essential. (A) Scheme summarizes the experimental approach used to prepare GcrA depleted samples in (C). NA1000 DgcrB DgcrA::V xylX::Pxyl-gcrA grown in M2G supplemented with 0.3% xylose were harvested and washed 3 times with M2 minimal salt solution, and then resuspended in M2G (GcrA depletion) or M2GX (GcrA expression) medium for 2, 5, or 24 h at 30uC. Then, the 24 h M2G culture was supplemented with 0.3% xylose (G+X) and incubated with the 24 h M2GX culture (X) for an additional 16 h at 30uC. For the 24 h and 40 h time points, culture dilutions were done to maintain cells in exponential growth throughout the experiment. (B) Immunoblots showing steady-state levels of various proteins in WT and mutant cells in M2G (reproduced from Figure 2G). (C) Immunoblots showing steady-state levels of various proteins after 2, 5, or 24 h of GcrA depletion in M2G. Red rectangle highlights that 5 h of GcrA depletion are sufficient to reconstruct DgcrA::V protein steady-state profiles shown in (B). After 24 h of GcrA depletion in M2G, culture was supplemented with 0.3% xylose (G+X) and incubated with the 24 h M2GX culture (X) for an additional 16 h at 30uC to obtain, respectively, X and G+X 40 h samples. Blue rectangle indicates that 16 h of GcrA reinduction restores WT protein steady-state profile, confirming that prolonged depletion of GcrA phenocopies the effect of the DgcrB DgcrA::V strain and that WT protein levels can be restored by reinstatement of GcrA expression. This argues against the possibility that suppressive mutations accumulate in the DgcrB DgcrA::V strain under these conditions. (EPS) Figure S7 Tn-insertion bias at the ccrM and ftsN loci and confirmation that the ccrM::Tn insertion is a null allele. (A) Chromosomal DNA of NA1000 and of some DgcrA::V and/or ccrM::Tn variants were purified and submitted (+) or not (2) to Hinf1 restriction analysis. Hinf1 can only cleave unmethylated GAnTC sites. The fact that the ccrM::Tn (as with the DccrM::V) extracted DNA is sensitive to restriction confirms that the ccrM::Tn mutation is a null allele. (B, C) Tn-insertion bias of DgcrA::V (blue) and NA1000 (red) strains determined by Tn-seq at the ccrM (B) and the ftsN (C) loci. Abscissa shows position as function of genome position, and ordinate gives Tn-insertion value. This Tn-Seq approach confirmed the Tn-suppressor screen, Tn-integration accumulating specifically all along the ccrM coding sequence and the ftsN promoter region in DgcrA::V strain compared to the WT. (EPS) Figure S8 The DgcrA::V ccrM::Tn strain retains morphological and replicative asymmetry. (A) Fluorescence and DIC micrographs of NA1000, DgcrA::V, and DgcrA::V ccrM::Tn, spmX-mCherry cells after growth in PYE. In all three strains, when a stalk structure is visible on the DIC micrograph, the stalked-pole-specific marker SpmX reveals unipolar SpmX-mCherry localization at this site, confirming that elongated