A novel nucleoid-associated protein coordinates chromosome replication and chromosome partition.

We searched for regulators of chromosome replication in the cell cycle model Caulobacter crescentus and found a novel DNA-binding protein (GapR) that selectively aids the initiation of chromosome replication and the initial steps of chromosome partitioning. The protein binds the chromosome origin of replication (Cori) and has higher-affinity binding to mutated Cori-DNA that increases Cori-plasmid replication in vivo. gapR gene expression is essential for normal rapid growth and sufficient GapR levels are required for the correct timing of chromosome replication. Whole genome ChIP-seq identified dynamic DNA-binding distributions for GapR, with the strongest associations at the partitioning (parABS) locus near Cori. Using molecular-genetic and fluorescence microscopy experiments, we showed that GapR also promotes the first steps of chromosome partitioning, the initial separation of the duplicated parS loci following replication from Cori. This separation occurs before the parABS-dependent partitioning phase. Therefore, this early separation, whose mechanisms is not known, coincides with the poorly defined mechanism(s) that establishes chromosome asymmetry: C. crescentus chromosomes are partitioned to distinct cell-poles which develop into replicating and non-replicating cell-types. We propose that GapR coordinates chromosome replication with asymmetry-establishing chromosome separation, noting that both roles are consistent with the phylogenetic restriction of GapR to asymmetrically dividing bacteria.

INTRODUCTION

Faithful genome duplication requires dedicated mechanisms to coordinate cell cycle events but for bacteria such coordination is especially problematic. For example, bacteria evolved cell cycle programs where chromosome replication and chromosome partitioning overlap (1). Once started, the cell is committed to chromosome replication that lasts most of the cell division cycle. Replication timing is critical since delays also reduce competitive fitness while commitment in unfavorable conditions risks genome damage and cell death (2). Since bacterial chromosomes use only one origin of replication, the bacterial strategy for regulating chromosome replication implies that much information is processed through this unique origin and this further implies complex protein-binding interactions (3,4). Our working hypothesis is that novel DNA-binding proteins evolved to coordinate chromosome replication and partitioning.

Caulobacter crescentus provides an excellent model to study the bacterial cell cycle (5). Its dimorphic growth presents distinct programs of chromosome replication and chromosome partitioning that are being exploited for detailed analysis. C. crescentus swarmer (Sw) cells present the motile and non-replicating cell stage. As the cell cycle proceeds, the Sw cells differentiate into stalked (St) cells that present the non-motile but replicating stage. Therefore, the initiation of chromosome replication is coordinated with the cell differentiation from the Sw to the St cell stages. Next, C. crescentus cell division proceeds asymmetrically as the elongating cell builds a new flagellum at the new Sw pole opposite the old St cell pole. Chromosome partitioning starts very soon after the initiation of chromosome replication and both cell cycle processes overlap the elaboration of asymmetric cell division that ultimately yields a Sw cell and a St cell (6). Therefore, this cell division program produces distinct cells, each with distinct non-replicating (Sw) and replicating (St) chromosomes.

We study the C. crescentus chromosome origin of replication (Cori; Figure 1A) and our experiments address how Cori directs the non-replicating Sw and the replicating St chromosome states. In this report, we identify a novel DNA-binding protein (GapR) and show that GapR is a dynamic-binding nucleoid-associated protein that selectively facilitates the initiation of chromosome replication and the earliest stage of chromosome partitioning. We propose that GapR connects the start of two overlapping yet otherwise mechanistically different cell cycle processes that ultimately create two functionally district chromosomes in an asymmetrically dividing cell.

Figure 1.

Identification of GapR (CCNA_03428) as a factor promoting chromosome replication. (A) A schematic of Cori including CtrA binding sites (black circles indicate half-sites), strong DnaA binding sites (two open arrows) and the most conserved region, studied in Taylor et al. (7) (underline). Flanking coding sequences (hemE and ccna_00001) are shown. The positions of three probes used for qChIP analysis in Figure 3B are marked by asterisks together with their identifying letters, for Cori L, Cori M and Cori R. (B) The ability of WT and mutant Cori DNA to drive autonomous plasmid replication was assessed by enzyme activity from the gusA gene reporting plasmid copy number. The higher CoriUP replication was assessed from five independent assays. The alignment shows WT sequence compared to up-replication mutant sequence (CoriUP; mutation underlined) together with the plasmid copy number (relative to WT). The position of a consensus CtrA binding site is indicated above the alignment. (C) EMSA experiments showing that a factor present in a cleared C. crescentus cell lysate binds more strongly to oligonucleotides carrying sequences from CoriUP than from CoriWT. 40 bp annealed oligonucleotides carrying either the CoriWT or CoriUP sequences were incubated with ten-fold dilutions of partially fractionated cleared cell lysate (left to right, dilutions indicated by a wedge) or in buffer alone (lanes marked ‘-’) and run on a non-denaturing gel. A greater fraction of the CoriUP oligonucleotides are shifted by the same lysate concentrations than the CoriWT oligonucleotides. (D) EMSA experiments comparing how equal concentrations of His-GapR protein (purified from E. coli) bind to the same annealed oligonucleotides as in (C). For each gel, in the leftmost lane oligonucleotides were incubated in buffer alone. Reactions in the right-hand lane were incubated with 500 nM His-GapR, and lanes to the left of this contained a two-fold dilution series of His-GapR. Preferential protein binding to the CoriUP oligonucleotide was replicated 4 time.

Figure 3.

Genome binding distributions of GapR. (A) The circular C. crescentus chromosome aligned with two whole-genome ChIP seq signals (175 bp resolution histogram bins representing % of total reads) using the GapR anti-serum first on the WT strain (NA1000, black outer signal) and then the ΔGapR suppressor strain (GM3875, grey inner signal). The standard bp positions of the WT NA1000 genome are displayed in the outer circle, with the small black bar above the 0-position indicating the region that is expanded in (C). (B) qChIP (GapR anti-serum) analysis performed on synchronized WT GM1609 cells grown in M2G media. GapR binding was measured by qPCR at the following positions: at parB (left panel; filled circles, dashed line), at ccna_02005 (left panel; open circles, solid line), and at three positions in Cori (middle panel, see Figure 1A for positions of probes: Cori L, crosses, solid line; Cori M, open diamonds, solid line; Cori R, filled diamonds, dashed line). The grey window shows the cell cycle period during which 95% of the chromosomes initiated replication, as measured by flow cytometry (right panel; see Figure 2A and Materials and Methods). (C) ChIP-seq signals over the region surrounding parB and Cori from untreated WT NA1000 cells (top trace) and from the same culture sampled and treated for 10 min with 30 ug/ml rifampicin (middle trace), or 100 ug/ml novobiocin (bottom trace). In the schematic below these three ChIP seq signals, the genetic organization in this region is shown with the parB gene indicated in black, and hemE and ccna_00001 which flank Cori indicated in white and dark grey respectively. The Cori region (the BamHI fragment that defines the autonomously replicating Cori) is indicated by vertical dashed lines (long dashes), the position of parS by one vertical dashed line (short dashes) and the position of the intergenic region downstream of parB by a vertical dotted line. The white vertical arrowhead indicates the position of the parB primers used for the qChIP analysis in (B). Note that (A) and (C) represent data from whole-genome ChIP-seq experiments on independent cultures and that both experiments position the most relevant GapR peaks (Supplementary Figure S7A) at the same locations.

MATERIALS AND METHODS

Bacterial strains and plasmids

Caulobacter crescentus strains and plasmids used in this study are described in Supplementary Tables S3 and S4, respectively. Construction of new plasmids and C. crescentus strains is described in the supplementary material. Escherichia coli strains were grown in LB media supplemented with ampicillin (100 μg/ml) or kanamycin (50 μg/ml) where noted. Caulobacter crescentus strains were grown in PYE or M2G media as noted, with xylose (0.5% w/v), glucose (0.2% w/v), ampicillin (20 μg/ml), chloramphenicol (1 μg/ml), spectinomycin (100 μg/ml) and streptomycin (2.5 μg/ml) added as described.

Cell fractionation and protein purification

The initial cell fractionation carried out to purify GapR from whole cell lysates as well as the purification buffer compositions are described in Supplementary Figure S1. Recombinant His-GapR and GST-CtrA were purified as described in supplementary methods.

EMSA reactions

EMSA reactions were performed with radio-labeled oligonucleotides as previously described (7) but in EMSA buffer consisting of 20 mM Tris–HCl pH 8, 100 mM KCl, 5 mM MgCl2, 1 mM CaCl2, 2 mM DTT, 50 μg ml−1 BSA. Briefly, radio-labeled annealed oligonucleotides were incubated with various protein fractions in the EMSA buffer on ice for 30 min before being loaded directly onto a 8% polyacrylamide gel in 1× TBE. The sequences of the oligonucleotides used for CoriWT and CoriUP are listed in Supplementary Table S5 (‘WT’ and ‘UP’ pairs respectively).

Autonomous replication assay

Autonomous replication assays were performed as previously described (7). Briefly, plasmids carrying the reporter gene gusA and relying on Cori or CoriUP for their replication were transformed by electroporation into GM1609 or derivatives, and cells were plated onto selective media (PYE + ampicillin). GM1609 can be selected for beta-lactamase resistance plasmids, because it has a deletion in its natural beta-lactamase gene. GM1609 is otherwise identical to the standard wild type strain NA1000. After four days, colonies were washed off into PYE media, and assayed for GusA activity to report plasmid copy number.

Immunoblotting

Antibodies were raised in rabbits against GapR (at the McGill Animal Facility) and CtrA (at MediMabs, Montreal), or have previously been described (7). All were used in immunoblotting at a 1:10,000 dilution. Secondary anti-rabbit antibodies conjugated to horse radish peroxidase (Sigma) were also used at a 1:10 000 dilution. Immunoblots from SDS-PAGE onto PVDF membrane were visualized using ‘Western Lightning’ reagents (Perkin-Elmer) and a VersaDoc system (BioRad).

Chromatin ImmunoPrecipitation coupled to deep sequencing (ChIP-Seq)

ChIP-seq was carried out essentially as previously described (8) using mid-log phase cells (O.D.660 nm ∼ 0.5), cultured in PYE or with an extra 10 min treatment with antibiotics (Rifampicin 30 μg/ml; Novobiocin 100 μg/ml, chloramphenicol 1 μg/ml). To form the immuno-complexes, the pre-cleared supernatants were incubated overnight at 4°C with GapR Polyclonal antibodies (1:1000 dilution). HiSeq 2000 runs of barcoded ChIP-Seq libraries yielded several million reads that were mapped to the Caulobacter crescentus NA1000 (NC_011916, circular form) and analyzed as described in supplementary methods. Briefly, the genome was subdivided into 1 bp (isolated regions) or 50 bp (full chromosome) probes, and we calculated the percentage of reads per probe as a function of the total number of reads. Analyzed data illustrated in Figure 3A using the Circos Software (9) are provided in Dataset S1 (50 bp resolution). Figure 3C focuses on the par and the Cori regions (4 026 155 to 1150 bp on the circular Caulobacter crescentus genome), analyzed datasets are provided in Dataset S2 (full chromosome at 50 bp resolution) and Dataset S3 (parABS and Cori regions at 1 bp resolution). Sequence data have been deposited to the Gene Expression Omnibus (GEO) database (GSE95535 series, accession nos. GSM2516003–GSM2516008).

Quantitative chromatin immunoprecipitation (qChIP)

Chromatin immunoprecipitation was performed as previously described but using 1:1000 anti-GapR (8). For mixed culture experiments, log phase cultures in PYE were treated with antibiotics (where described) for 10 min. qPCR was performed with oligonucleotide primers described in Supplementary Table S5 (primer pairs listed as ‘q[gene name] fwd/rev’) using the FastStart Universal SYBR green qPCR kit (Roche) on a Rotorgene 6000 thermocycler (Corbett) or a CFX 96 thermocycler (Bio-Rad) and analyzed using the Rotorgene (Corbett) or CFX Manager (Bio-Rad) software package. All qPCR reactions were validated by standard curves using template serial dilutions. Amplification efficiencies calculated from the slopes of these curves lay between 95% and 105%, with the R2 values for each curve being >0.98.

Flow cytometry to measure the fraction of cells that started chromosome replication

Flow cytometry was performed on synchronous cultures that had been isolated with a Percoll density gradient (10) using an replication run-out protocol that allowed cells to finish any rounds of chromosome replication that had already initiated, as previously described (2). Briefly, freshly isolated swarmer cells were transferred to PYE containing glucose or xylose and samples were taken from the synchronized culture at various times and transferred into media containing 60 μg/ml rifampicin and 35 μg/ml cephalexin for at least three hours. Ethanol was then added to 70% final concentration to fix cells, and samples washed once with 70% ethanol and were stored at 4°C prior to analysis. Staining and analysis were performed as described (2).

Microscopy

Fluorescence and DIC images were captured using a Zeiss Axiovert200M with a 100× objective at the McGill University Life Sciences Complex Advanced BioImaging Facility. For fluorescence microscopy experiments, samples were imaged on 1% agarose pads in water or PYE media for live cell time-lapse studies. For all experiments comparing GFP-ParB localization in synchronized cells, we also took control flow cytometry samples (see above) concurrently with their microscopic imaging to confirm that the fraction of cells that started chromosome replication were comparable between matched strains. This control distinguished replication timing defects from chromosome partitioning (GFP-ParB motion) defects in Figures 5 and 6. GFP-ParB localization was scored by inspecting images with large cell numbers (n-values are given in the respective figure legends).

Figure 5.

ΔGapR cells are specifically defective in the early phase of chromosome partitioning. (A–C) Quantification of cells at three stages of chromosome partitioning in synchronous cultures of WT cells carrying a chromosomal gfp-parB (GM3905; filled circles, dotted lines) and in a ΔGapR suppressor strain with the same chromosomal gfp-parB fusion (GM3920; empty circles, solid lines). Over the first 60 min of synchronous growth, these graphs show the changing fraction of cells that had a single GFP-ParB focus, i.e. ‘Unipolar’ stage (A), a partially partitioned pair of GFP-ParB foci, i.e. ‘Partitioning’ stage (B) and cells that had completed partitioning, i.e. ‘Bi-polar’ stage with a pair of polar GFP-ParB foci (C). n > 100 cells were scored for each time point. (D) Quantification of cells at each of the three stages of chromosome partitioning, as depicted in (A–C) and scored at the t = 20 min time-point. Data from three independent synchronous cultures of GM3905 (gray bars) and GM3920 (white bars). n > 300 cells were scored for each stage. (E) As in (D), quantification of cells at each of the three stages of chromosome partitioning depicted in (A–C) and scored at the t = 20 min time point of synchronous cultures. Strains GM3905 (gray bars) and GM3920 (white bars) carried a chromosomal integrated Pvan-parAK20R plasmid. ParAK20R protein expression was induced with vanillate two hours prior to synchronization and this treatment blocks ParA-mediated chromosome segregation, as shown by the absence of cells in the ‘Bi-polar’ stage. (F) Quantification of cells at the ‘Partitioning’ stage sampled at t = 25 min from synchronous cultures of GM3905 (gray bar), GM3920 (white bar) and GM3920 + pJT165 (plasmid carrying GapR under its native promoter; black bar). For D–F, n > 300 cells were scored for each condition. Significant differences (P < 0.01, z-test) are indicated by **.

Figure 6.

Time-lapse microscopy shows that GapR aids the early phase of chromosome partitioning. (A) Quantification of the partitioning events of GM3905 (WT, gray bars) and GM3920 cells (ΔGapR, white bars) over a 15 minute time-course assayed by time-lapse microscopy. Cells were counted and classified according to whether they started partitioning (‘start’; one polar GFP focus becomes two), completed partitioning (‘end’; one polar and one mid-cell GFP focus become two bi-polar foci), both started and completed partitioning (‘start and end’; one polar GFP focus becomes two bi-polar foci) or underwent a ‘remerge’ event (two foci combined to show one focus) over the 15 minutes of time-lapse microscopy. Error bars indicate the estimated standard deviation of a Poisson distribution (n = 350 and 190 for GM3905 and GM3920 respectively). Significant differences have probabilities (z-test) of P < 0.05 and P < 0.01, indicated by * and ** respectively. No significant difference (n.s.) P > 0.05. (B and C) Single-cell kymographs created from sliced and aligned time-lapse images, taken every minute over 15 minutes. Examples of GM3905 (B) and of GM3920 (C) cells undergoing a complete ‘start and end’ partitioning (B) and of a ‘remerge’ partitioning (C) of GFP-ParB foci. To the left of the DIC kymograph, the first whole cell image is indicated to show pole positions (note some drift in cell position during the experiment). White arrowheads in the GFP kymographs indicate separation events where two GFP-ParB foci became resolvable. The black arrowhead in (C) indicates a ‘remerge’ event where two GFP-foci fused into one foci. These images therefore illustrate the most significant ‘start and end’ (B) and ‘remerge’ (C) events quantified in (A).

RESULTS

Discovery of a novel DNA-binding protein

As part of our search for new regulators of the C. crescentus origin of chromosome replication (Cori) we previously reported a 5 bp mutation within the most conserved region of Cori (Figure 1A) that causes a ∼2 fold increase in Cori autonomous replication (7). Here, we introduced a 3 bp mutation into a similar position in Cori and observed that this mutation also caused a ∼2-fold increase in Cori replication (Figure 1B). We hypothesized that this mutation (CoriUP) altered the binding of a protein to Cori, and we sought to identify it by an electrophoretic mobility shift assay (EMSA). We probed C. crescentus cell lysates using radio-labeled annealed oligonucleotides carrying either the natural Cori sequence (CoriWT) or the mutated up-replication sequence (CoriUP). We found that the CoriUP sequence had a higher affinity for an unknown protein present in the cell lysate (Figure 1C). We purified this protein by biochemical fractionation (see Supplementary Figure S1 and methods) and identified it using liquid chromatography/tandem mass-spectrometry as the product of the ccna_03428 gene, previously annotated as a hypothetical cytosolic protein (11), that is highly conserved in alpha-proteobacteria, which was named GapR. We PCR cloned and purified N-terminally poly-histidine tagged GapR from E. coli and confirmed that this protein preparation also differentially shifts the CoriWT and CoriUP oligonucleotides (Figure 1D).

GapR is essential for normal rapid growth

We attempted to delete the GapR coding sequence and to replace it with the omega antibiotic resistance cassette through two-step homologous recombination. However, we could only delete GapR when we supplied an extra GapR gene on a replicating plasmid (see Supplementary Table S1). A high-throughput transposon mutagenesis screen also proposed that GapR is essential in C. crescentus (11). To study how removing GapR might affect C. crescentus, we replaced the plasmid covering the deletion with a plasmid carrying GapR under the control of a xylose inducible promoter (pJT157; see Supplementary Figure S2A). This strain (GM3817) now required xylose for growth and died when shifted to glucose (Supplementary Figure S2B). Interestingly, while most cells die during prolonged depletion of GapR, a few cells survive. These derived ‘ΔGapR suppressor strains’ can grow indefinitely without GapR and resemble wild type (WT) cells with key defects (described further below and in supplementary data). While these experiments were done under rapid growth conditions in rich (PYE) media, we also observed that these same GM3817 cells died under slow growth conditions in minimal (M2G/glucose) media. The derived ΔGapR suppressor strains grew slower than WT and their growth was stimulated by genetically restoring gapR on plasmid pJT165 (supplementary data). Therefore, GapR is essential for rapid growth and this requirement suggests that GapR regulates or optimizes essential cell cycle functions.

GapR is required for timing chromosome replication

Our analysis indicated that GapR positively promotes replication by directly binding to Cori (Figure 1), and therefore we next tested replication initiation in conditionally expressing Pxyl::GapR GM3817 cells. Chromosome replication initiates from Cori when swarmer (Sw) cells differentiate into stalked (St) cells. Therefore, synchronous C. crescentus Sw cell cultures of WT parental GM1609 and GM3817 were isolated (10) and placed in media supplemented with xylose or glucose. Samples of the synchronized cultures were removed at 0, 30, 60 and 90 minutes post-synchronization and treated with antibiotics to prevent new rounds of chromosome replication and cell division. Since this treatment allows ongoing chromosome replication to finish, this method distinguishes replicating and non-replicating cells. Cells that initiated chromosome replication when sampled will complete replication (but not cell division), and will have two complete chromosomes and cells which have not initiated replication will retain only one chromosome (see Figure 2A).

Figure 2.

Cells with insufficient GapR fail to initiate chromosome replication. (A) Presentation of the flow cytometry-based assay to measure initiation times of chromosome replication. A synchronized culture of WT strain GM1609 proceeds through the cell cycle (cartoon on left) developing from swarmer cells (top) to stalked cells, predivisional cells, and finally asymmetrically divided cells (bottom). Flow cytometry was used to distinguish cells with replication versus non-replicating chromosomes. Cultures were sampled at progressive times and transferred into media with antibiotics that block further replication initiation and cell division while allowing ongoing replication to continue and produce two chromosomes (see Materials and Methods). Flow cytometry assess fluorescent chromosome DNA versus cell numbers (traces on right). Prior to initiation, t = 0 cells contain one chromosome, while after initiation t = 15 min cells contain two chromosomes. (B and C) Initiation time experiments in synchronized cultures of the parental WT strain (GM1609; B) and the GapR Pxyl conditional-expression strain (GM3817; C) assessed using flow cytometry as outlined in (A) and grown in media containing xylose (left histograms; GapR expression induced) or glucose (right histograms; GapR expression not induced). Samples were taken at the indicated times (minutes). (D) Immunoblots for CtrA master regulator protein. Samples were taken at time points indicated above the lanes (minutes) from synchronized cultures in (B) and (C). CtrA patterns confirm microscopic observations and show that the cell cycles, with respect to early cell differentiation, proceeded normally and were not influenced by the xylose versus glucose media conditions. The delayed replication response seen here in 2C Glucose was replicated in two additional experiments not shown.

As expected, cells from WT GM1609 differentiated rapidly and initiated chromosome replication by 30 min and GM1609 showed no replication timing differences between xylose or glucose media (Figure 2B). In contrast, even with xylose to induce GapR, GM3817 cells displayed a pronounced delay in replication initiation, and most of these cells failed to initiate replication in glucose media without xylose to induce GapR expression (Figure 2C). Immunoblots show that GapR levels in GM3817 grown in xylose were ∼10% of the WT cells (not shown). Therefore, both replication delays (WT versus GM3817 Figure 2BC and xylose versus glucose Figure 2C) are attributable to lower GapR levels.

We also assayed the synchronized cells in Figure 2 by tracking the abundance of two marker proteins that are degraded at the Sw to St cell transition. Immunoblot analysis demonstrated that both the master cell cycle regulator CtrA (Figure 2D) and the chemotaxis protein McpA (Supplementary Figure S3) were degraded at the appropriate cell cycle times. For example, CtrA was completely degraded in all cultures by 30 min when all WT GM1609 cells had initiated chromosome replication. Interestingly, both marker proteins show a delayed return late in the cell cycle of the GM3817 cells suggesting that later stalked cell programs (e.g. cell division) are delayed with reduced GapR levels. In summary, while GapR probably aids later cell cycle progression, it does not significantly influence the earlier Sw to St cell differentiation and yet lower GapR levels selectively retard the timely start of chromosome replication, implying the selective activity at Cori.

Direct GapR binding to Cori promotes chromosome replication in vivo

If GapR is promoting chromosome replication by directly binding to Cori, then increasing Cori affinity for GapR should ameliorate the delayed replication seen above during GapR depletion. We therefore created the strain GM3880 with the CoriUP mutation placed at the natural Cori chromosome position but which is otherwise identical to GM3817 used above in Figure 2. By using GM3880 in the otherwise same replication timing experiment, we now see that the CoriUP mutation advances replication and the severe replication timing lag is no longer seen in glucose media (compare flow cytometry results in Figure 2 with Supplementary Figure S4A).

Similarly, if increased GapR binding to Cori is driving the extra Cori-plasmid replication of CoriUP seen in Figure 1B, then genetically removing GapR should remove the differential replication of the CoriWT and CoriUP sequences. To test this prediction, we isolated a spontaneous suppressor strain which retained viability despite lacking GapR (GM3875; see Supplementary Methods). Repeating the autonomous replication assays confirmed that the ∼2-fold differential replication between the CoriWT and CoriUPCori-plasmids seen in WT GM1609 is no longer seen in GM3875 (Supplementary Figure S4B).

In vitro GapR has broad DNA-binding specificity like nucleoid-binding proteins

EMSA experiments show that GapR has high affinity (Kd ∼ 10 nM) for Cori DNA and that GapR probably binds DNA through very specific peptide contacts, because an GapR mutation (Y82A) at its conserved tyrosine abolishes all high affinity EMSA binding to Cori (Supplementary Figure S6). GapR clearly has DNA sequence preferences (Figure 1) and we sought to identify GapR binding sites inside Cori, as we did for other Cori regulatory proteins (7), through in vitro DNase I protection (‘footprint’) experiments. Therefore, purified histidine tagged GapR protein (as in Figure 1D) was added to 32-P end-labeled WT Cori DNA. However, we did not see DNase I protection until a threshold concentration of GapR protein produced complete protection over the Cori DNA shown in Figure 1A (data not shown). Such results were repeated under a range of in vitro conditions and they imply that the intrinsic DNA-binding specificity of GapR is broad and relatively non-selective. However, since GapR selectively affects Cori replication in vivo (Figures 1B, and 2, Supplementary Figure S4A, S4B), we reasoned that additional ‘extrinsic’ factors must direct GapR to Cori and that clues can be found through whole genome binding studies.

In vivo GapR has broad genomic DNA-binding specificity but with specific peak distributions

We used chromatin-immuno-precipitation and Illumina sequencing (ChIP-seq) to determine how GapR is distributed on the C. crescentus chromosome in growing WT cells. To show that ChIP-seq specifically identifies true GapR binding positions, we performed the same analysis on the ΔGapR GM3875 ‘suppressor strain’ that survives without GapR. This comparison confirmed the high specificity of our GapR anti-serum, because none of the prominent WT peaks are present in the ΔGapR GM3875 control (Figure 3A; compare outer Circos black trace with inner grey trace). This ChIP-seq analysis also argued for weak DNA sequence specificity since these GapR peaks are broad, often spanning whole genes (e.g. see Figure 3C) and they lack distinct boundaries.

Interestingly, the first ranked GapR peak was found over the chromosome partitioning gene parB (end of the parABS operon) in two independent ChIP seq experiments (Figure 3A and in C ‘Untreated’) and we confirmed this abundant GapR binding at parB and other key positions using ChIP and qPCR (Supplementary Figure S5A). We attempted to identify DNA sequences under this and other top ranked peaks using the MEME software suite (12). While this sequence alignment approach has identified DNA binding sites with similar ChIP-seq data (13), we did not find DNA motifs that accounted for the GapR distributions on the chromosome. Nonetheless, GapR does show a weak ‘intrinsic DNA specificity’ based on its association with degenerate DNA motifs which we address further below.

Antibiotic treatments rapidly redistribute GapR peaks

We considered what ‘extrinsic’ factors might direct GapR into prominent peaks, thereby overcoming GapR’s weak ‘intrinsic specificity’ for DNA. Growing cells are engaged in replication, transcription and coupled translation and such macromolecular syntheses, each driven by large molecular machines, might move and reposition GapR. We therefore tested ChIP-seq signals from WT growing cultures and from cultures split and treated with rifampicin or with novobiocin. We saw that most GapR binding peaks were eliminated after 10 min with either treatment, as for example the peak at parB (Figure 3C).

Novobiocin and rifampicin redistribute GapR to the hemE/Cori replication locus

Interestingly, a new prominent peak over the left side of Cori and within the hemE gene, was created after either novobiocin or rifampicin treatments (Figure 3C). This observation is very significant, because hemE transcription and Cori replication are interrelated and this DNA uniquely supports autonomous replication (14). To confirm this result, we performed qChIP experiments with these same antibiotics. While ChIP-seq reports relative amounts of binding, our qChIP experiments demonstrated that the absolute amount of GapR binding to Cori at hemE increases following either novobiocin or rifampicin treatments (Supplementary Figure S5B).

The significance and uniqueness of GapR peaks at parB, hemE/Cori versus other loci

The preceding observations and especially GapR binding at parB and hemE/Cori are potentially very significant for cell cycle studies. This consideration prompted us to more rigorously test the significance of these peaks in our ChIP seq experiments. Here we summarize the details provided in our supplementary data: We analyzed the whole genome ChIP seq data set (partially presented in Figure 3C). For the ‘untreated’ growing cells we identified 36 GapR relevant peaks (according to the cutoff criteria described in Supplementary Figure S7). These same 36 peaks are reproducible since they are all present among the top 80 GapR relevant peaks from the independent ChIP seq experiment presented in Figure 3A, and most significantly, parB is the first GapR relevant peak in both experiments.

The short 10 min antibiotic treatments with rifampicin (Rif) or with novobiocin (Novo) leveled the genome-wide distribution of GapR. While 36 GapR relevant peaks were seen in the untreated culture, only three GapR +Rif relevant peaks and only seven GapR +Novo relevant peaks satisfied the same cutoff criteria (Supplementary Figure S7ABC). None of the 36 GapR relevant peaks were found among the GapR +Rif and GapR +Novo peaks (Supplementary Figure S7DE). Therefore, these GapR peaks that characterize growing cells, including parB, result from activities inhibited by either antibiotic, suggesting that active replication and transcription and their consequences, e.g. molecular collisions and/or DNA supercoiling work to actively maintain these 36 peaks. Interestingly, despite the overall more even GapR distribution created by Rif and Novo, the one common relevant peak produced by both Rif and Novo treatments is at the hemE/Cori locus (Figure 3C, Supplementary Figure S7F). Therefore, this last result suggests a relationship between Cori and GapR that is unique among the whole genome.

GapR binding correlates with the initiation of chromosome replication and partitioning

Since parB is located within 10 kb of Cori and since both loci have essential cell cycle roles, we further examined GapR binding during the cell cycle with qChIP experiments on synchronized WT cells (Figure 3B). Using qChIP primers at parB this signal peaked early in the cell cycle immediately preceding the Sw to St cell transition and the initiation window for chromosome replication (20–60 min Figure 3B). The first ranked parB timing contrasted with the second ranked GapR binding region, at ccna_02005 close to the terminus, which peaked much later at ∼80 min in to the cell cycle. We also examined GapR binding at three positions spanning Cori (Figure 1A; asterisks). Cori L binding proximal to hemE was weak and showed no cell cycle variation. In contrast, at Cori M binding to the middle of Cori was much stronger and this signal doubled during the initiation window for replication and again late in the cell cycle prior to the next round of replication initiation that starts after cell division. In contrast to these patterns, GapR binding to Cori R, close to the in vitro EMSA target in Figure 1CD, peaked most discretely at 60 min, during the latter half of the initiation window for replication (Figure 3B). Therefore, GapR binding changes significantly during the cell cycle: GapR binding inside Cori correlates with the initiation of chromosome replication and the distinct Cori L M R binding patterns during replication initiation suggest specialized roles and interactions with different replication proteins. Interestingly, peak GapR binding at parB precedes chromosome duplication and separation and yet remains high, suggesting early roles before and during chromosome partitioning that we discuss further below.

Microscopy shows that GapR is dynamically repositioned during chromosome partitioning

C. crescentus chromosome partitioning has been studied in gfp-parB strains where the GFP-ParB fusion protein binds the parS (centromere-like) locus near parB and Cori to form fluorescent foci at the cell poles (15). We fused the mCherry protein to the N-terminus of GapR and integrated this construct at Pxyl in strain GM3905 (gfp-parB) to create strain GM3921 (gfp-parB Pxyl::mCherry-GapR). This mCherry-GapR protein is induced by xylose and it retains important in vivo functions, as GM3921 derived strains remained viable when the native GapR gene was deleted (see Supplementary Table S1). Therefore mCherry-GapR localization probably reports the native GapR localization. We synchronized GM3921 cells and observed how the mCherry-GapR distribution changed (Figure 4 and Supplementary Figure S8). In t = 0 Sw cells, mCherry-GapR initially distributed as a gradient with its peak very near but not overlapping the flagellated cell pole marked by the single GFP-ParB focus. As chromosome replication and partitioning initiated, in t = 15 and t = 30 St cells, one GFP-focus remained polar while one GFP-focus moved away and the mCherry-GapR peak moved with this partitioning GFP-ParB focus (peaks of GFP and mCherry fluorescence co-localized in 76% of partitioning cells at t = 15). This movement created a wider mCherry-GapR free gap between the polar GFP-focus and the partitioning GFP-ParB focus (seen in 94% of partitioning cells at t = 15). The GFP-ParB focus then passed through the peak of the mCherry-GapR gradient as it moved further away to the opposite pole. Finally in t = 90 pre-divisional cells, mCherry-GapR established a separate gradient in each nascent cell compartment, with a peak of mCherry-GapR close to, but separated from, each pole marked by its GFP-ParB focus.

Figure 4.

GapR distribution changes during cell cycle progression and correlates with chromosome parS movements. (A) Microscopy images from a synchronized culture of GM3921 (gfp-parB Pxyl::mCherry-GapR) taken at t = 0, t = 15, t = 30 and t = 90 (minutes). A larger set of images is presented in Supplementary Figure S8. At each time-point we imaged >400 cells. Representative fields viewed with DIC, mCherry and GFP channels are shown in black and white and the merged mCherry/GFP channels are colored (pink and green respectively). The arrowheads in the merged t = 15 and t = 30 pictures mark cell regions between the two GFP-ParB foci that (compared with t = 0 cells) become exceptionally free of mCherry-GapR. In these cells the peak mCherry-GapR fluorescence often co-localizes with the partitioning GFP-ParB focus. (B) A cartoon summary of inferred GFP-ParB and mCherry-GapR movements over the cell cycle from Sw cell (t = 0, left) to pre-divisional cell (t = 90, right). As in (A) the arrowhead marks the sub-cell region between two GFP-ParB foci that becomes cleared of mCherry-GapR as both fluorescence signals move together at this early phase of the cell cycle.

Strategies to test if chromosome partitioning is a dominant cell cycle role for GapR

Since chromosome partitioning requires parS site duplication and GapR depletion blocks replication (Figure 2), we could not use the same GapR depletion strategy and track GFP-ParB foci during the cell cycle. As our first strategy, since we could not study partition during the short ∼40 min Sw to St cell cycle window, we used prolonged GapR depletion times instead. Accordingly, we constructed gfp-parB strains GM3905 gfp-parB and GM3918 gfp-parB ΔGapR Pxyl::GapR corresponding to the WT and regulated GapR expression strains used in Figure 2. Prolonged GapR depletion experiments, e.g. GM3918 cells 16 h PYE with glucose, allowed many cells to finally initiate chromosome replication and as expected, these cells had extra misplaced GFP-ParB foci (data not shown). However, we could not draw firm conclusions, because such GM3918 cells also showed cell division defects. As a broad-binding protein, it is also likely that GapR depletion impacts many functions that cumulatively produce unpredictable consequences, especially since such cells lose viability during the prolonged shut-off period (Supplementary Figure S2B).

As our second strategy, we took advantage of the viable ΔGapR suppressor strains (described in supplementary data). As we noted, while the initial removal of GapR is lethal, the ΔGapR suppressor strains that survive resemble WT cells. We reasoned that the most dominant roles, such as perhaps chromosome partitioning would be the most difficult to compensate and therefore only partially suppressed in these ΔGapR strains. Also, we selected a suppressor strain with normal replication kinetics during the most relevant Sw to St cell differentiation period (Supplementary Figure S4C). This strain allowed us to observe GFP-ParB/partitioning independently from chromosome replication.

Early-phase partitioning is assisted by GapR

According to the preceding strategy, we compared synchronous cultures of the gfp-parB ΔGapR suppressor strain (GM3920) with the gfp-parB parental strain (GM3905) and we quantified 3 types of GFP-ParB localization patterns (‘unipolar’, ‘partitioning’ and ‘bi-polar’ patterns, Figure 5ABC) at progressive cell cycle times. GM3920 consistently showed more unipolar (Figure 5A) and fewer partitioning GFP foci (Figure 5B) during the early 0–40 min cell cycle period. These differences reflect early partitioning defects, because they are too large to be caused by delayed chromosome replication in ΔGapR GM3920 versus WT GM3905. As controls, these same cultures when spotted for microscopy were also sampled for the replication run-out assay (as done in Figure 2). For all samples the fraction of replicating cells was practically the same and the differences between GM3920 and GM3905 were always <10%. Interestingly, the bi-polar GFP-ParB foci appeared at the same rate in both synchronized cultures (Figure 5C), suggesting that despite early differences both ΔGapR the WT strains have equally efficient means to pull the GFP-ParB foci to their ends. Therefore, to confirm these results we next focused on the early partitioning period; we repeated these synchronized culture experiments in triplicate and examined cells at the 20 min time point. This fixed time analysis (Figure 5D) confirmed the time-course analysis (Figure 5A–C) and demonstrated that GM3920 cells without GapR have a deficit of ‘partitioning’, i.e. double-foci cells during the early phase of partitioning.

Genetic analysis also implicates GapR in early phase partitioning without ParA activity

The late phase of chromosome portioning requires ParA activity while the early phase does not (16). To confirm that the defect in GFP-ParB/parS separation of GM3920 occurs during the early ParA-independent phase of chromosome segregation, we blocked ParA-mediated chromosome partitioning by expressing the dominant negative allele ParAK20R in synchronized cultures prior to the microscopic analysis at the 20 min time point. As expected (16), expression of ParAK20R abolished bi-polar localization of GFP-ParB foci (Figure 5E) that requires ParA. However, the deficit of ‘partitioning’ GFP-ParB foci remained in GM3920 when ParA was blocked (Figure 5E), thereby placing the cause of this deficit in the early ParA-independent phase of chromosome partitioning.

Because we examined chromosome partitioning in a suppressor strain rather than in a simple ΔGapR strain, we sought to confirm that the early partitioning deficit in GM3920 results from the lack of GapR rather than from a suppressor mutation. We therefore transformed GM3920 with pJT165, carrying GapR under its native promoter to complement the GapR deletion, and we assayed chromosome partitioning as in Figure 5DE. pJT165 ameliorated the partitioning deficit of GM3920 (Figure 5F), further showing that GapR directly aids the early phase of chromosome partitioning.

Time-lapse microscopy further implicates GapR during the earliest parS/centromere separation

To investigate what caused the deficit of ‘partitioning’, i.e. fewer double GFP-ParB/parS foci in the ΔGapR cells, we performed time-lapse microscopy on single cells of synchronized cultures. Swarmer cells were isolated and after 10 minutes of synchronous growth in PYE liquid, cells were transferred to nutrient PYE/agar pads on microscope slides for a 15 min time-lapse analysis of GFP-ParB foci. During this cell cycle window, we observed and quantified 4 categories of foci movements: First, we observed that fewer "start’ events (one polar focus becoming two close foci) for GM3920 cells compared to control GM3905 cells (Figure 6A). Second, we observed fewer ‘start and end’ events (one polar focus becoming two complete bipolar foci) for GM3920 cells (Figure 6A, with a specific example in Figure 6B). Third, we also observed fewer ‘end’ events (two close foci becoming completely bipolar) for GM3920 cells. Therefore, all of these observations independently support defective partitioning ΔGapR cells.

The fourth category of foci movements was most informative. We observed significantly more ‘remerge’ events (two close GFP-ParB foci combining into one focus) for ΔGapR GM3920 cells (Figure 6A, with a specific example in 6C). These ‘remerge’ events probably account for the deficit of separated GFP-ParB foci in GM3920 that we scored as ‘partitioning’ foci in Figure 5BDE and they would also give a false impression of defective replication scored as ‘unipolar’ foci in Figure 5A. We interpret a ‘remerge’ event as a failure to maintain the first separation step of chromosome partitioning. This defective event coincides with the symmetry-splitting ‘discrimination’ function of the par system that keeps one parS locus at the St pole and that sends the other parS locus to the emerging Sw pole. To the best of our knowledge, such parS remerging events have not been reported before, probably because they are rare in WT cells (Figure 6A) and the small separation is difficult to observe (Figure 6C). ‘Remerge’ events only become apparent in our ΔGapR cells when the early partitioning phase is closely examined.

DISCUSSION

We first proposed that GapR was a DNA sequence-specific regulator of Cori but further studies caused us to modify this hypothesis. For example, the broad DNA-binding specificities and dynamic peak distributions, argue that GapR should instead be classified as a nucleoid associated protein (NAP) that potentially impacts the whole genome. Interestingly, genome-wide surveys confirmed that GapR has unique associations with the hemE/Cori and the nearby parABS loci and we gathered evidence for GapR’s expanded roles. While GapR lacks strong DNA sequence-specificity, it does act as a regulator. We propose that GapR is best described as a facilitator of complex regulatory processes and specifically of the initiation of chromosome replication and of the initiation of chromosome partitioning. Regulators optimize essential processes and need not be essential themselves. Accordingly, GapR is essential for normal rapid growth and while we could isolate ΔGapR strains, these strains always grew slower and progressed less efficiently through the cell cycle. We specifically isolated one ΔGapR suppressor strain that appeared to sufficiently restore the cell cycle program during the Sw to St cell transition so that, at least under standard rich media conditions, these cells looked normal and initiated chromosome replication with normal timing. However, closer examinations revealed that this ΔGapR strain had less efficient chromosome separation and partitioning, emphasizing GapR’s dominance in this process.

During our studies on GapR (http://biorxiv.org/content/early/2016/12/04/091496), two independent groups also began studying the same C. crescentus gene/protein (CCNA_03428), reported as GapR (17,18). Accordingly, we changed our earlier designation of opaA (origin and partition binding protein A) to match this recent literature. Unlike our replication function-based approach (Figure 1), both groups started from a bioinformatics survey of essential genes (11). In agreement with our conclusions, both groups presented GapR/OpaA as a NAP that participates more directly in the dimorphic cell cycle of C. crescentus than previously studied NAPs. For example, both groups reported that GapR/OpaA has subtle yet global impacts on genetic transcription, although they differed on the identity of the genes and their significance for the cell cycle. Most pertinent to our results, while the first report by Ricci et al. attributes GapR/OpaA distributions on the chromosome primarily to preferential DNA-binding to AT-rich DNA (17), the second report by Arias-Cartin et al. (18) deemphasizes the importance of specific DNA-binding sequences. Instead, Arias-Cartin et al. proposed a dynamic binding model whereby replication forks dislodge DNA-bound GapR/OpaA proteins which then preferentially rebind to the replicated DNA.

How does GapR in fact bind genomic DNA and how can we explain these contradictions?

We observed a dynamic repositioning of GapR/OpaA by DNA-protein crosslinking (Figure 3) and fluorescence microscopy (Figure 4). GapR/OpaA binding distributions change rapidly after antibiotic treatments (Figure 3AC, Supplementary Figures S7, S9, S10) and during the course of normal cell cycle progression (Figures 3 and 4, Supplementary Figure S8). According to Arias-Cartin et al., replication forks starting from Cori create an GapR/OpaA gradient peaking at Cori, because this region has more time to rebind the displaced protein (18). Their observations are potentially consistent with our microscopic observations (Figure 4), because in t = 0 Sw cells we also see mCherry-GapR gradients decreasing from the parS/ParB-GFP polar foci, and in t = 15, t = 30 cells we see mCherry-GapR receding from these poles, producing a fluorescence gap that could be caused by replication forks displacing mCherry-GapR. However, replication fork interactions alone cannot explain local GapR/OpaA peaks, nor can they explain their rapid and local cell cycle variations (Figure 3B). We see that transcription and translation also reposition GapR/OpaA. Regarding our ChIP seq experiment from Figure 3C, we see that blocking RNA polymerase evens the genomic distribution of GapR/OpaA, since a 10 min +Rif treatment leaves only three relevant peaks (above the cutoff) compared with the ‘untreated’ 36 relevant genome-wide peaks (Supplementary Figure S7). However surprisingly, a 10 min chloramphenicol (+Chl) treatment of this same culture sharpens the genomic distribution of GapR/OpaA and now produces 81 relevant peaks (Circos Supplementary Figures S9B and S10B). This is a very significant redistribution considering that only 2 of these +Chl relevant peaks overlap the ‘untreated’ 36 GapR peaks and interestingly one of these overlap peaks is parB, yet another unexpected connection between GapR/OpaA and chromosome partitioning.

Further regarding the genetic contexts of GapR/OpaA peaks, we see that like the peak over the end of parB, ∼80% of the ‘untreated’ 36 peaks position where two genes/operons converge (Supplementary Figure S10E). In contrast, over 80% of the GapR +Chl 81 relevant peaks position over upstream genetic regions which are often AT-rich. Plotting genome-wide peak ranking versus GC content confirms that +Chl creates the most relevant peaks that are AT-rich (compare Supplementary Figure S10F and G). These observations probably have regulatory significance as antibiotics or similar stresses that block translation would rapidly direct GapR/OpaA to AT-rich promoters and thereby alter genetic transcription.

While our ChIP seq experiments analyzed natural GapR/OpaA protein, both recent reports used genetic/protein fusions for ChIP seq and other analyses (17,18). We argue that such gene/protein fusions cause artifacts and misleading conclusions. For example, Ricci et al. (17) proposed that GapR/OpaA binds preferentially to AT-rich promoters thereby ‘fine-tuning’ genetic transcription of cell cycle genes. However, Ricci et al. used a FLAG-GapR fusion and we therefore reanalyzed their primary ChIP seq data with our criteria. Accordingly, we see 474 FLAG-GapR relevant peaks and therefore a much sharper and focused genome-wide distribution than our natural GapR/OpaA with only 36 relevant peaks (compare Supplementary Figure S10C and S10A). Noting that the chloramphenicol treatment (+Chl) also sharpened and focused GapR/OpaA distributions, we see that 73 out of our 81 relevant +Chl peaks overlap with the 474 FLAG-GapR peaks, while only 8 out of our 36 ‘untreated’ GapR relevant peaks show an overlap (Supplementary Figure S10D). Furthermore, like the +Chl peaks, the highest ranking FLAG-GapR peaks show a very similar bias towards AT-rich DNA (compare Supplementary Figure S10G and S10H). We conclude that the FLAG-GapR fusion mimics chloramphenicol stress and that all experiments with GapR fusions should be interpreted with care. Nonetheless, the redistribution of GapR to AT-rich DNA, including to parB and Cori loci (Supplementary Figure S10D) is significant and probably reflects a weak intrinsic DNA sequence bias suggested by specific AT-rich motifs (Supplementary Figure S10IJK). While our studies on chromosome replication and partitioning focused on non-stressed cells growing under optimal conditions, it is likely that GapR also responds to stresses and that GapR helps regulate stress-response genes, a hypothesis also suggested by Arias-Cartin et al. (18). We also speculate that GapR binds and perhaps protects AT-rich DNA, because we observe that the CoriUP mutant DNA is more AT-rich and this promotes what appears to be GapR protein oligomerization (higher multimeric EMSA shifts in Figure 1D).

How is GapR involved in chromosome replication?

We initially purified GapR by its EMSA binding to specific Cori DNA and we proposed that GapR regulates chromosome replication, because increased GapR binding to CoriUP mutant DNA in vitro correlates with increased autonomous replication of CoriUP mutant plasmids in vivo (Figure 1). Subsequent experiments support a chromosome replication hypothesis. For example, our genetic shut-off experiments show that lack of GapR selectively prevents the initiation of chromosome replication in vivo (Figure 2). We scrutinized the consequences of gapR shut-off over the short-term 90 min period of the C. crescentus cell cycle. These are in vivo protein depletion experiments, because GapR is an abundant and stable protein (data not shown). Therefore, we actually observed the consequences of reduced GapR protein levels. Under these conditions and as in WT cells, swarmer (Sw) cells differentiated into stalked cells (St) which then divided asymmetrically. The CtrA master regulator marks the Sw cells and its disappearance marks the development of St cells while its reappearance (new synthesis) marks the onset of asymmetric cell division (19,20). Therefore, it is very significant that CtrA degradation kinetics are not changed by GapR depletion while the initiation of chromosome replication is severely retarded in the GapR shut-off cells (Figure 2). We also presented two supplementary experiments (Supplementary Figure S4) demonstrating that GapR acts directly at Cori. Therefore, lack of GapR selectively inhibits chromosome replication, while the general program of Sw to St cell differentiation proceeds normally.

Interestingly, GapR protein levels increase during the cell cycle, and compared to Sw cells GapR concentrations are at least 2-fold higher in late dividing St cells (Supplementary Figure S11). Therefore, bulk GapR protein concentrations are asymmetrically distributed by cell division and Sw cells receive lower concentrations of GapR than St cells. However, GapR protein concentrations alone do not determine DNA-binding. When GapR concentrations are low, substantially more GapR binds to the parABS DNA of synchronized Sw cells (Figure 3B). Also, GapR binds to Cori with very specific spatial and temporal patterns (Figure 3B, compare Cori L, M and R patterns). Cori is a platform for many protein interactions. Therefore, these GapR binding kinetics presumably reflect the many replication protein interactions that occur during synchronized chromosome replication.

How is GapR involved in chromosome partitioning?

C. crescentus chromosome partitioning is bi-phasic, with an initial slow separation of ParB-parS complexes (16) followed by a rapid movement of one ParB-parS complex to the new cell pole (21). Only the rapid second phase of partitioning requires ParA ATPase. Blocking ParA activity by expressing the dominant-negative allele ParAK20R abolishes final parS placement at the distant cell pole (16,21). Nevertheless, these ParAK20R cells still undergo the initial slow and partial separation of ParB-parS. Our microscopy studies showed that significantly fewer cells separate the ParB-parS complexes in the ΔGapR suppressor strain and our experiments placed the separation defect in the initial ParA-independent phase of chromosome partitioning. Additionally, time lapse microscopy showed that fewer ΔGapR suppressor cells separate their ParB-parS complexes and that a greater proportion of them remerge. Therefore, the initial separation of the par-Cori region is not maintained and GapR may supply or help direct a driving force for this separation.

The dynamic cellular mCherry-GapR distributions (Figure 4, Supplementary Figure S8) also argue that GapR acts during the initial slow separation phase. Interestingly, the strongest mCherry-fluorescence stops next to the GFP-ParB/parS focus in the t = 0 Sw cells (e.g. Supplementary Figure S8C). Subsequently, two GFP-ParB/parS foci form and one focus follows the mCherry-fluorescence as it recedes away from the cell pole still occupied by the remaining focus (t = 15, t = 30 cells). These mCherry-GapR movements do not use ParA and they suggest a quasi-mitotic process. At later times, the GFP-ParB/parS focus stops following and instead moves through the main mCherry-GapR fluorescence on its way to the opposite cell pole. This movement into the mCherry zone probably reflects a transfer from the early/slow GapR-system to the later/fast ParA-system.

The motor for the early phase of chromosome partitioning is unknown (16) and previous studies have suggested that RNA polymerase, DNA replication and bulk polymer properties may drive these movements (22–24). Inhibitors of these polymerases affect GapR binding at parB and noting that the receding edge of mCherry-GapR coincides with the moving GFP-ParB/parS focus (Figure 4 and Supplementary Figure S8t = 15, t = 30), we propose that replication forks may not only displace GapR/OpaA (18) they may also selectively move one of the two GFP-ParB/parS foci. Accordingly, we propose that GapR/OpaA displacement is a byproduct of its primary interactions with early partitioning motors. Replication is a likely candidate but we also propose that coupled transcription and translation have motive functions. In our Supplementary data (in Supplementary Figure S12), we present experiments that implicate both RNA polymerase and ribosomes. How GapR/OpaA might aid and guide the motors will be an important subject for future studies.

How is GapR involved in establishing chromosome asymmetry?

Bacterial chromosome replication and partition overlap temporally and this regulatory link is also implied by chromosome organization, because origins of replication often lie close to par loci. For example, in C. crescentus Cori and parS are separated by only 10 kb. These temporal and spatial linkages probably evolved to meet the special needs of bacteria and it should also be advantageous for both parS and Cori to share regulatory proteins. Interestingly, a recent study also implicated the replication initiator DnaA during chromosome partitioning in C. crescentus (25). Therefore, replication and partitioning both utilize DnaA and GapR in C. crescentus. Similarly, in Vibrio cholerae and Bacillus subtillis, ParA interacts with DnaA to control replication (26,27).

Since chromosome replication precedes its partitioning, one might expect regulators to act sequentially, first at Cori and then at parS, but counter intuitively a recent study shows that as the activity of C. crescentus DnaA rises, it first acts at parS and then at Cori, because the threshold for DnaA action is lower at parS than at Cori (25). In a like manner, we see peak binding of GapR at parB immediately preceding the GapR binding interactions at Cori (Cori M and R) during the initiation of replication (Figure 3B). Therefore, while earlier cell cycle binding of GapR to parS at first appeared to be a paradox, it now agrees with an emerging view that both GapR and DnaA must respond to cell cycle signals and by as yet unknown mechanisms they must prepare parS in anticipation of replication initiation at Cori.

The early ParA-independent separation of parS loci is very important for asymmetric cell division, because this separation enables ParA, which moves from the distal pole, to bind just one chromosome for transport. Otherwise, if two ParB-parS loci are near when they encounter ParA, they risk being co-transported and not segregated (16,28,29). Therefore, while the ParA-system is ‘blind’ and chooses indiscriminately, it is the earlier GapR-utilizing system that ‘sees’ and discriminates between two alternative chromosomes. This is a key symmetry-splitting phase that channels the chromosomes into distinct Sw-pole versus St-pole programs (including alternative chromosome replication programs). Therefore, we propose that GapR facilities C. crescentus asymmetry. This role may also explain why GapR homologs are found exclusively in the α-proteobacteria (Supplementary Figure S13) (30) which typically divide asymmetrically and localize their origins of replication at the cell poles (31–33).

Click here for additional data file.