In Vivo Imaging of the Segregation of the 2 Chromosomes and the Cell Division Proteins of Rhodobacter sphaeroides Reveals an Unexpected Role for MipZ

Cell division has to be coordinated with chromosome segregation to ensure the stable inheritance of genetic information. We investigated this coordination in the multichromosome bacterium Rhodobacter sphaeroides. By examining the origin and terminus regions of the two chromosomes, the ParA-like ATPase MipZ and FtsZ, we showed that chromosome 1 appears to be the “master” chromosome connecting DNA segregation and cell division, with MipZ being critical for coordination. MipZ shows an unexpected localization pattern, with MipZ monomers interacting with ParB of the chromosome 1 at the cell poles whereas MipZ dimers colocalize with FtsZ at midcell during constriction, both forming dynamic rings. These data suggest that MipZ has roles in R. sphaeroides in both controlling septation and coordinating chromosome segregation with cell division.

C ell viability relies on mechanisms ensuring accurate cell division. Not only does division usually have to take place at midcell, it has to happen after segregation of the chromosomes. The tubulin-like protein FtsZ is the major division effector recruited to the division plane, where it polymerizes as a dynamic ring intimately involved in the constriction process (1)(2)(3)(4).
Division control is based on a balance between positive and negative regulations of the stability of FtsZ polymers. Positive regulators are mainly involved in stabilizing the forming ring, whereas temporal and spatial control is performed by negative regulators, including the Min or MipZ system (5). Recently, additional proteins involved in FtsZ recruitment to midcell, including PomZ (6), SsgAB (7), and MapZ (8,9), have been identified, suggesting that the mechanism may, however, be more complex and diverse.
The Min system prevents the FtsZ ring forming at sites other than midcell in many species (10)(11)(12). For example, in Escherichia coli, ParA-like ATPase MinD recruits MinC, the inhibitor of FtsZ ring formation, and oscillates between poles, with the midcell becoming a MinCD-free zone as the cell grows, allowing the formation of the division complex. In contrast, in Bacillus subtilis, the MinCD proteins localize to the poles and to two rings flanking FtsZ at midcell (13,14), inhibiting FtsZ ring mispositioning. In both E. coli and B. subtilis, the Min system is complemented by a nucleoid occlusion system (15,16), allowing coordination of chromosome segregation and cell division.
In the alphaproteobacterium Caulobacter crescentus, a single ParA protein, MipZ, undertakes both of these levels of control (17,18). MipZ is a direct inhibitor of FtsZ polymerization. It binds the ParB-parS DNA complex at the origin (Ori) region of the chromosome and forms a gradient from the origin to the new pole. The segregation of the origin regions leads to the establishment of a bipolar gradient of MipZ, preventing FtsZ ring formation other than at midcell. Mechanistically, the gradient formation is based on ParB stimulating the turnover of the MipZ monomers into dimers which inhibit FtsZ polymerization.
The descriptions of the mechanisms presented above are based on species with one chromosome; however, around 10% of sequenced bacteria have a multipartite genome. While the physiology of these species has often been studied in detail, little is known about the behavior of the multipartite genomes throughout the cell cycle (19)(20)(21)(22). Rhodobacter sphaeroides is an alphaproteobacterium that is mainly studied for its physiological versatility (23,24), but it was also the first bacterium with a composite genome to be identified (25). Its genome has two circular chromosomes of ϳ3 (C1) and ϳ1 (C2) Mb. Here, using cell biology and fluorescence microscopy, we show that the segregation pattern for the origin (OriC) and terminal (Ter) loci of these chromosomes is different from the patterns seen with the multichromosomic alphaproteobacteria studied to date (19,22). We also show that a MipZ homologue not only is involved in coordinating the segregation of C1 to cell division but also localizes to midcell in a pattern very different from that characterized in C. crescentus, suggesting a previously unidentified role for this regulator in cell division and control of FtsZ.

RESULTS
A MipZ homologue controls cell division in R. sphaeroides. A search for division regulators in R. sphaeroides identified a protein with 44% sequence identity to C. crescentus MipZ (17) (EGJ22543.1; see Fig. S1 in the supplemental material). Deletion of the mipZ gene was possible only when an extra copy was present on the chromosome, indicating that the mipZ gene product is essential. Increased production of MipZ led to minicell formation, filamentation, and cell death, suggesting that MipZ is a negative regulator of cell division in R. sphaeroides (Fig. 1A).
We replaced mipZ in the genome with mipZ-RFP (mipZ-red fluorescent protein gene) expressed in the native locus behind the native promoter (Fig. 1B). MipZ-RFP is functional, with less than 0.5% of population forming minicells. MipZ-RFP exhibits a complex and dynamic localization pattern occupying three positions in the cell, including both poles and the septum. Classifying the cells by size, as a measure of cell age, revealed that MipZ forms a focus at the new pole in newly divided cells; as the cell grows, MipZ forms a focus at both poles and at midcell, with the levels of fluorescent intensity changing at all three sites; as division progresses, the MipZ present at the septum forms a clear ring which concentrates to a bright focus as septation continues (Fig. 1B).
The striking observation of MipZ at the septum as well as at the polar zones indicates a more complex role in cell division than has been shown for C. crescentus. We therefore investigated the possible roles of MipZ at both the poles and midcell.
Choreography of OriC1 and OriC2. In C. crescentus, MipZ links the inhibition of FtsZ polymerization at the poles to the segregation of chromosome by directly interacting with nucleoprotein complex ParB/parS at OriC. As MipZ is also at a pole in R. sphaeroides, we hypothesized that it might link cell division to the segregation of one or both chromosomes in R. sphaeroides. We therefore investigated the localization of the Ori sites of both C1 and C2 (OriC1 and OriC2) in living cells (Fig. 2).
We identified the origin region of each chromosome by cumulative GC skew analysis and searched for replication and partition loci (Fig. S2). To localize the origins of replication in live cells, we took advantage of the parABS system on each chromosome and of the findings that ParB proteins colocalize with OriC through their binding to parS sequences (26)(27)(28) and that each parABS system is specific to the replicon that encodes it (20).
Using ParB1-YFP (ParB1-yellow fluorescent protein), we observed either one or two defined foci in every cell, showing that the replication cycles did not overlap ( Fig. 2A, panel i). In newborn cells, OriC1 localized close to the old pole. OriC1 then duplicated, and one focus segregated to the opposite pole, resulting in two foci occupying the positions at about 15% and 85% of the cell length ( Fig. 2A, panel ii). Comparing OriC1 localization with that of a polar protein, membrane chemotaxis receptor McpJ (Fig. S3A), confirmed that OriC1 showed subpolar localization rather than being strictly polar, a finding also reflected by the lesser distances between the sister foci (Fig. S3B). Transient tight polar localization is seen only just after OriC1 duplication.
OriC1 positioning therefore showed 3 phases: in newly divided cells, OriC1 was loosely localized on one side of old pole; in small (1.7 to 2.5 m) cells with 2 foci, the OriC1 foci moved to be very close to the cell poles; in cells above 2.5 m in size, the positioning was again relaxed ( Fig. 2A, panel iii). This move from tightly polar to relaxed localization occurred in cells where the FtsZ ring was positioned at midcell (Fig. S4A) and began constriction (Fig. S4B). This suggests a transient attachment of OriC1 to a polar factor soon after segregation, which remains until after the reorganization of the chromosomes. Consistent with this model, simultaneous imaging of ParB1-YFP and mCh-ParA1 (Fig. S5A) revealed that ParB1-OriC1 complex duplication occurred in a manner concomitant with relocalization of ParA1. In newborn cells, ParA1 appears as a diffuse cloud at the new pole, a fraction of which then condenses into foci that colocalize with the newly segregated sister ParB1s (Fig. S5B).
To localize OriC of C2 (OriC2), we used the ParB MT1 /parS system (Fig. S2B) developed in E. coli (29) as all attempts to conjugate a plasmid coding for a ParB2 fusion protein in the R. sphaeroides wild-type (WT) strain failed. OriC2 localized at midcell in 1-focus cells, and after duplication and symmetric segregation, the OriC2 foci localized at about 30% and 70% of the cell length (Fig. 2B). Interestingly, this pattern is more similar to that identified in the unrelated gammaproteobacterium Vibrio cholerae (19,30) than to that recently described in the alphaproteobacteria Sinorhizobium meliloti (22) and Brucella abortus (31).
Following the two loci simultaneously (Fig. 2C, panel i) and calculating the average interfocal distance (IFD) in the population of cells with 2 foci for each origin confirmed their different positions in the cell (Fig. 2C, panel ii). Interestingly, the comparison of the percentage of cells with 1 focus to the percentage of cells with 2 foci for each locus (25% to 75% versus 35% to 65%) revealed a difference between their spatial separation times (Fig. 2C, panel iii). By calculating the age of the cell at the time that the focus duplicated (31) to a 150-min cell cycle (Fig. 2C, panel iii), we estimated duplications of OriC1 foci at ϳ28.5 min of the cell cycle and of OriC2 foci at ϳ42 min. The earlier separation of OriC1 sister foci was observed in real time in the time-lapse experiments ( Fig. 2C, panel iv) and was further confirmed in cephalexin-treated filamentous cells ( Fig. S6A and B). It is noteworthy that in all multichromosomic bacteria studied so far, the origin of the biggest chromosome is always the first to be segregated.
Localization of Ter1 and Ter2. In order to better understand the process of segregation of the two chromosomes, we also characterized the segregation pattern of their terminus regions (Ter1 and Ter2). These were identified using cumulative GC sequence analysis and the positions of putative dif sites (32) (Fig. S2). We introduced a parS P1 sequence and a parS MT1 sequence into an intergenic region of convergent genes close to the dif site on C1 and C2, respectively, and expressed fluorescently labeled ParB P1 and ParB MT1 from pIND4 to visualize the two loci ( Fig. 2D to F).
Ter1 and Ter2 are at the new pole in newborn cells, and Ter2 migrates from the pole to midcell before Ter1. Both Ter sites remained as single foci until just before the end Applying the same method used for OriC analysis, we estimated that Ter2 duplicates at ϳ129 min and that Ter1 duplicates at ϳ146 min of a 150-min cycle, giving 117 min between the apparent duplications of OriC1 and Ter1 and 87 min between the apparent duplications of OriC2 and Ter2. Interestingly, the time between Ori duplication and Ter duplication does not reflect the 3-fold difference in size (DNA length) between C1 and C2 replicons. This suggests either a difference in the replication rates or, more likely, a difference in the control of cohesion of sister chromatids of these two chromosomes as suggested by the delay in Ter1 segregation observed in some cephalexin-induced filamentous cells (Fig. S6C). The latter mechanism would ensure the stringent control necessary for synchronizing the segregation of the 2 chromosomes with cell division.
MipZ and ParB1: localization and dynamic. Given the change in the subpolar position of OriC1, we examined whether OriC1 and MipZ might colocalize by analyzing ParB1-YFP in a strain expressing MipZ-RFP from the chromosome (Fig. 3A and B). Colocalization of ParB1 and MipZ was observed at the cell poles throughout the cell cycle, and their direct interaction was confirmed by a bacterial two-hybrid (B2H) assay (Fig. 3C).
This suggests interplay between ParB1, the MipZ position, and the segregation of OriC1 ensuring coordination of cell division and segregation of C1. As OriC1 duplicates and segregates before OriC2 whereas Ter1 duplicates before Ter2, C1 control of MipZ would be sufficient to coordinate duplication and segregation of the complete nucleoid with division.
By following MipZ and ParB1 localization throughout the cell cycle using time-lapse analysis (Fig. 3B), we identified major differences in the positioning pattern compared with C. crescentus (17). We found no evidence for a "tail" or bipolar gradient of MipZ from the cell poles (17) but rather saw symmetric fluorescence intensity profiles for both ParB1 and MipZ foci. Again, unlike C. crescentus, MipZ localized at the old pole Simultaneous imaging of FtsZ-YFP and MipZ-YFP using three-dimensional structured illumination microscopy (3D-SIM) (Fig. 4D) showed FtsZ rings similar to those previously observed using other high-resolution microscopy techniques in R. sphaeroides and other bacteria (33)(34)(35). In addition, we observe that MipZ-YFP also formed a ring at midcell (Fig. 4D). Measurements of the diameter of FtsZ and MipZ rings showed that, on average, the diameter of the FtsZ rings was larger than that of the  (Fig. 4E). As MipZ is recruited to the septum after FtsZ ring formation, this suggests colocalization during the septation process, possibly regulating FtsZ ring constriction.
MipZ dimers localize at the septum and inhibit cell division. As MipZ appears to show at least two different behaviors during the cell cycle of R. sphaeroides and because it is related to the ParA family of ATPases, we tested whether the ATP-dependent monomer/dimer switch described for ParA proteins has a role in function (36) by generating a range of mutants in described conserved sites (18,37,38). We expressed in the WT strain the predicted monomer locked forms MipZ G12V (mutation in the P-loop, which interferes with dimer formation) and MipZ K16Q (mutation in the nucleotide-binding site, which reduces affinity for nucleotides) and the dimer locked form D40A (ATP hydrolysis-defective mutant), each fused to RFP (Fig. 5A). In parallel, we measured their interaction with ParA1 and ParB1 by B2H assay (Fig. 5B). A clear association of localization and function was observed. The monomer forms were mainly found at the poles, where they interacted with ParB1. They were inactive in division control, as overexpression had no effect on division (Fig. 5A). Interestingly, the MipZG12V monomer, which retains reduced localization at midcell (Fig. 5A), showed a direct interaction with ParA1 (Fig. 5B). This reveals that, in addition to their interaction with ParB, the two ATPases MipZ and ParA1 interact in a nucleotide-binding-dependent manner.
The MipZ dimeric form was exclusively observed at midcell and even low-level expression had a dominant-negative effect on division over the WT proteins, leading to extensive cell filamentation under conditions of overproduction. These data support the hypothesis that MipZ cycles between an ADP-bound monomer and an ATP-bound monomer which interacts with ParAB1 and an ATP-bound dimer which interacts with FtsZ at the septum and controls constriction.

DISCUSSION
By following the origins and termini of the two chromosomes and the associated controlling proteins ParB and MipZ throughout the cell cycle (Fig. 6), we showed that the dynamic pattern of movement and the roles of the proteins, particularly that of MipZ, are different from those described in related species (Fig. 7).
Segregation of the R. sphaeroides chromosomes. The segregation mechanism for the two R. sphaeroides chromosomes appears to be closer to that of the unrelated Vibrio cholerae (30) than to the pattern described for the alphaproteobacteria S. meliloti and A. tumefaciens (19). Unlike the pattern described for V. cholerae or C. crescentus, OriC1 is only transiently strictly polar, with the predominant subpolar localization more reminiscent of that of Pseudomonas aeruginosa (39). This suggests that the mechanisms of chromosome management and the patterns of segregation are not necessarily confined to closely related species (Fig. 7A).
Analysis of the spatiotemporal patterns of segregation of the OriC and Ter loci showed no correspondence between the size of the replicons and the timing of their segregation. While the origins replicated and segregated early in the cell cycle, both Ter sites showed delays until just before a late stage of septation before segregation occurred, possibly because the last steps of DNA segregation for both chromosomes are controlled by the division process, with cohesion being maintained until late constriction. The data clearly show that the first event and the last event in R. sphaeroides genome segregation correspond to the segregation of the C1, suggesting that this is the "master" chromosome, coordinating genome segregation and cell division.
MipZ: a new role in cell division control. R. sphaeroides MipZ shares several characteristics with all the ParA-like proteins, including differential protein-protein interactions and functions controlled by monomer-dimer cycling. Moreover, we ob-  Fig. 2). The rightmost panel shows the coordination between the segregation of the chromosomes, represented as a noncompacted nucleoid, and the division process through MipZ. In a newborn cell, MipZ localizes as a focus at the new pole where FtsZ resides and with the ParB1-parS-OriC1 complex at the old pole. After OriC1 duplication and segregation, ParB relocalizes the MipZ focus from a very polar to a subpolar localization at the new pole and MipZ is observed at ParB at the old pole, finalizing the "ParB catching MipZ" phase. After FtsZ ring initiation at midcell, MipZ colocalizes as a single ring with the division protein, where it stays until the end of the division process. MipZ localization is dynamic and oscillates between the poles, where it interacts with ParB1 as monomers, and the septum, where it acts on the division process as dimers.
Unexpected Role for MipZ ® served that, similarly to other ParA-like proteins, MipZ exhibited nonspecific DNA binding activity (see Fig. S8 in the supplemental material). It also shares some of C. crescentus MipZ specificities with MipZ binding to ParB as a monomer and apparently negatively regulates FtsZ polymerization as a dimer. However, the localization and the mechanism of action of the monomer and dimers are very different from those observed in C. crescentus, suggesting that the monomers and dimers regulate chromosome segregation and septation using a mechanism different from those described to date.
The difference in behavior of the two MipZ monomer mutants, with both binding ParB1 but with the nucleotide-binding, nondimer mutant also binding ParA1 and showing localization at midcell (Fig. 5), implies that structural and interaction differences occur as the monomer binds and releases nucleotide, with DNA segregation machine ParABS of C1 probably regulating the dynamics to control cell division. As overproduction of the MipZ dimer inhibits septation, this suggests inhibition of septum formation. However, unlike C. crescentus, there was no observable bipolar gradient, and the dimer never localized to the poles; rather, MipZ dimers formed a ring which colocalized with the FtsZ constriction ring (Fig. 7B). MipZ catalytic domains are well conserved in R. sphaeroides and C. crescentus, while more-variable regions appear notably located on the external surface of the protein (Fig. S1A and B). The differences in the residues exposed at the protein surface could potentially modulate interactions with other factors and explain the singular behavior of MipZ.
We suggest that MipZ monomers and FtsZ monomers in a newly divided cell localize at positions that are very close to the new pole, probably targeted by a polar localizing protein, whereas the OriC-parS complex is at the other pole with ParB1. OriC1 duplicates and segregates and ParB1 interacts with MipZ monomers, releasing them and FtsZ from the tight polar position. FtsZ, closely followed by MipZ ATP-bound monomers and dimers, moves to midcell and forms single rings. As both nucleotidebound monomers and dimers are found at midcell, it is possible that there is a dynamic exchange between the monomers and ring-forming dimers. MipZ clearly oscillates between the poles but does not form a gradient. It seems likely that the movement between the poles and nucleotide binding and exchange with the dimer ring ensure that FtsZ-dependent septation does not happen until the two chromosomes have duplicated and segregated, C2 being controlled by C1, as duplication of both OriC2 and Ter2 happens within the time frame of C1 duplication. This unusual pattern of protein movement could represent an adaptation to specific constraints imposed by the multichromosomic architecture of R. sphaeroides genome, or the C. crescentus mechanism may have evolved from a species that buds daughter cells from one pole. The MipZ cycling between two monomer and one dimer forms regulates MipZ dynamics from the poles to the septum and fine-tunes the constriction process in response to segregation of master chromosome 1 through the key coordinator ParAB system.

MATERIALS AND METHODS
Bacterial strains, plasmids, and growth conditions. R. sphaeroides was grown aerobically with shaking at 225 rpm in succinate medium (SUX) (40) or on LB agar plates at 30°C. E. coli was grown aerobically with shaking at 225 rpm in LB medium. Kanamycin (Km), when required, was used at 25 g/ml. Cephalexin was used at 1 g/ml and at 10 g/ml for R. sphaeroides and E. coli, respectively. All expression plasmids derived from pIND4 were conjugated into R. sphaeroides strain WS8N as described previously (41). Expression relied on either promoter leakage or isopropyl ␤-D-1-thiogalactopyranoside (IPTG) addition. Overproduction experiments were done by spotting of 10 l of exponential cultures of strain WS8N carrying pIND4 and derivatives on LB agar supplemented of 0, 50, or 100 M IPTG. Morphological changes induced by MipZ overproduction were visualized after 7 h 30 min (ϳ5 generations) by microscopy.
Construction of DNA deletions and insertions and mutants. DNA insertion or deletion has been performed previously using plasmid pKT18mobsac (42). For mipZ deletion, 0.7-kb fragments of upstream and downstream regions of mipZ amplified with oligonucleotides delmipZUPa and delmipZUPb and delmipZDWa and delmipZDWb were cloned together into HindIII-EcoR1 in pKT18mobsac (pKTdelmipZ), introducing a new Kpn1 site at the junction. This construct allows deletion of 722 of the 810 bp of the gene. parS sequences were amplified from pGBKD3-parSP1 and pGBKD3-parSpMT1 and cloned in a unique preexisting NotI or XhoI site or created by overlap PCR in an OriC2, Ter1, or Ter2 DNA fragment cloned in pKT18mobsac. mipZ mutants were constructed by PCR mutagenesis. Oligonucleotides with a single point mutation corresponding to the single base change were designed to amplify pINDMipZ-RFP (43) with Phusion DNA polymerase (NEB). PCR were treated with DpnI (NEB), and DH5␣ cells were transformed. Plasmid DNA was extracted from the resulting clones (MiniPrep kit; Qiagen) and sequenced. Oligonucleotides, plasmids, and strains are listed in Table S1A to C in the supplemental material. Sequence analysis. Proteins, genes, or sequences of interest were analyzed using pBLAST, BLAST, or Clone Manager software. Sequences of ParA/MinD/MipZ proteins were aligned using ClustalW, and the phylogenetic tree was produced using Phylogeny.fr (http://phylogeny.lirmm.fr/phylo_cgi/index.cgi). The GC skew of each chromosome was calculated and visualized with GenSkew, and the positions of Ori and DnaA boxes were verified with OriFinder.
Fluorescence microscopy and analysis. Cells were grown to optical density at 550 nm (OD 550 ) of 0.2 to 0.4 in SUX medium, washed in SUX, and spread on a layer of SUX medium-1% agarose on a microscope slide.
Differential interference contrast (DIC) and fluorescence images were acquired using a Nikon TE200 microscope equipped with a cooled charge-coupled-device (CCD) camera (Hamamatsu) or a Nikon Unexpected Role for MipZ ® Eclipse TE2000-U microscope equipped with a Photometrics Cool-Snap HQ CCD camera. Image analysis used ImageJ or MicrobeTracker together with custom programs run in Matlab.
3D structured illumination microscopy (3D-SIM) imaging was performed as described by Lesterlin et al. (44) on an OMX V3 Blaze microscope (Applied Precision/GE Healthcare) equipped with a 60ϫ/1.42 oil UPlanS Apo objective (Olympus) and 488-nm-wavelength-diode lasers and scientific complementary metal-oxide-semiconductor (sCMOS) cameras (PCO). Three-dimensional stacks of FtsZ-YFP and MipZ-YFP were obtained using 8 to 12 125-nm z-sections (resulting from a striped illumination pattern) (angles of Ϫ60°, 0°, and ϩ60°) and were shifted in five phase steps. Acquisition settings were 10 ms of exposure for FtsZ-YFP and 35 ms of exposure for MipZ-YFP with 10% and 100% transmission of a 488-nm laser, respectively. Reconstruction was performed using SoftWoRx 6.0 (Applied Precision), and 3D rendering of fluorescent signal together with volume surfacing was done using IMARIS analysis BITPLANE software.