|Home | About | Journals | Submit | Contact Us | Français|
Until recently, a dedicated mitotic apparatus that segregates newly replicated chromosomes into daughter cells was believed to be unique to eukaryotic cells. Here we demonstrate that the bacterium Caulobacter crescentus segregates its chromosome using a partitioning (Par) apparatus that has surprising similarities to eukaryotic spindles. We show that the C. crescentus ATPase ParA forms linear polymers in vitro and assembles into a narrow linear structure in vivo. The centromere-binding protein ParB binds to and destabilizes ParA structures in vitro. We propose that this ParB-stimulated ParA depolymerization activity moves the centromere to the opposite cell pole through a burnt bridge Brownian ratchet mechanism. Finally, we identify the pole-specific TipN protein1,2 as a new component of the Par system that is required to maintain the directionality of DNA transfer towards the new cell pole. Our results elucidate a bacterial chromosome segregation mechanism that features basic operating principles similar to eukaryotic mitotic machines, including a multivalent protein complex at the centromere that stimulates the dynamic disassembly of polymers to move chromosomes into daughter compartments.
Recent evidence suggests that Caulobacter crescentus and other bacteria use DNA partitioning (Par) systems related to those found in plasmids to segregate chromosomal origin regions on DNA replication. Par systems are found throughout bacterial species3 and consist of three core components: 1) an origin-proximal centromeric DNA sequence, parS; 2) an ATPase ParA, hypothesized to provide the force for centromere segregation through dynamic polymerization; and 3) a mediator protein ParB, which binds to parS and is predicted to regulate and couple ParA-induced force to parS movement. In C. crescentus, ParA and ParB are essential4. Depletion of ParB, overexpression of ParA and/or ParB, extra parS sequences, or mutations in the ParA ATPase active site result in severe chromosome segregation defects4–6. Furthermore, the C. crescentus parS site has been identified as the functional centromere6, and blocking DNA replication initiation prevents translocation of the ParB–parS complex to the opposite cell pole7. In addition to the core Par components, C. crescentus uses a pole-specific protein PopZ to tether the parS region to the pole through direct interaction with ParB, which prevents reverse segregation of the ParB–parS complex8,9. Together, these data suggest that the C. crescentus Par system, in cooperation with the polar PopZ network, mediates the active segregation and subsequent tethering of the parS region to the cell pole to initiate chromosome partitioning.
Despite a clear role in DNA partitioning, the mechanisms proposed for Par systems are diverse and largely hypothetical10–16. However, Par systems have several common features. Various ParA homologues have been shown to polymerize in vitro10,11,16–20. Dynamic pole-to-pole oscillation of ParA localization has been observed in vivo, and in some cases has been shown to require ATPase activity and the presence of both ParB and parS10,12,13,15,19,21–25. Importantly, recent observations demonstrate a correlation between ParB movement and a retracting cloud-like localization of ParA during segregation12,15, suggesting that a ParA structure ‘pulls’ ParB–parS complexes. However, the architecture of ParA assemblies, the molecular mechanisms by which these structures form and generate chromosomal movement, and the cellular components required to impart directionality to ParA-mediated segregation have yet to be established.
To examine the role of ParA and ParB in chromosome segregation, we replaced the C. crescentus chromosomal parA and parB genes with parA-eyfp and cfp-parB, respectively, and used time-lapse microscopy to image synchronized cells. Initially CFP–ParB bound to parS formed a focus (red) at the old pole, as reported previously5, and ParA–eYFP (green) localized predominantly between the new pole and the CFP–ParB focus (Fig. 1a). Next, the CFP–ParB focus duplicated, and one focus followed the edge of a receding ParA–eYFP structure towards the opposite cell pole (Fig. 1a, top row; Supplementary Information, Fig. S1a), suggesting that a retracting ParA complex moves ParB–parS during segregation12,15.
To obtain higher resolution images of ParA in vivo, we performed two-colour single-molecule fluorescence imaging to extract super resolution images of ParA–eYFP and mCherry–ParB localizations during segregation in live cells. Figure 1b shows representative epifluorescence and super resolution images of ParA–eYFP (green) and mCherry–ParB (red) in cells at different stages of parS progression towards the distal pole. We observed that ParA–eYFP molecules localized to a discrete linear structure (Fig. 1b; Supplementary Information, Fig. S1a and b) with widths of 40.1 ± 9.5 nm. A cell imaged before replication initiation (Fig. 1b, cell A), shows a linear ParA–eYFP structure. Cells imaged during segregation (Fig. 1b, cell B) show linear ParA–eYFP assemblies that frequently have the highest density of ParA localizations between the new pole and the segregating ParB–parS complex, reflecting at super resolution the retracting cloud-like ParA localizations in the epifuorescence images in Fig. 1a (Supplementary Information, Fig. S1b). Finally, cells imaged after the completion of parS segregation (Fig. 1b, cell C) show linear ParA structures that stretch from pole to pole, suggesting reorganization of the ParA structure after segregation. No ordered assemblies were observed when we imaged cytoplasmic eYFP alone, but linear ParA–eYFP structures were observed in cells after fixing with formaldehyde (Supplementary Information, Fig. S1c) and when ParA was fused to mCherry (Supplementary Information, Fig. S1c). To further demonstrate the consistency between the epifluorescence and super resolution experiments, we reconstructed diffraction-limited images using the super resolution fitted localization data (Supplementary Information, Fig. S1d) that matched well with the epifluorescence images (Fig. 1a). We conclude that ParA–eYFP is assembled predominantly into a narrow linear structure oriented along the long axis of the cell, which could not be resolved with diffraction-limited microscopy.
The narrow linear structures of ParA–eYFP observed in vivo suggest that these structures consist of ParA polymers. We therefore purified ParA and measured multimerization using light scattering (Supplementary Information, Fig. S2a). ParA combined with ATP produced a rapid increase in light scattering, indicating polymerization (green). No increase in light scattering was observed in the absence of nucleotide, and ADP stimulated a slow increase (blue and red, respectively). We imaged ParA structures directly using negative-stain electron microscopy. When incubated without ATP, no ParA polymers were observed (Supplementary Information, Fig. S2b). However, in the presence of ATP, ParA formed linear polymers that were laterally bundled (Fig. 1c, upper and lower panels), as observed for other ParA homologues10,11,16,17,19.
We performed a mutational analysis to determine the roles of ParA biochemical interactions in ParA localization. The proposed ParA biochemical pathway18 is shown in Fig. 2a. Apo–ParA binds to ATP (Fig. 2a, top), stimulating ParA homodimerization18,19. The ATP-bound ParA dimer interacts with ParB, binds to DNA, or polymerizes18,19. ParB stimulates ParA ATP hydrolysis11,19,26 or nucleotide exchange27, releasing ParA as monomers (Fig. 2a, bottom). We mutated conserved ParA residues to abrogate specific biochemical interactions (Fig. 2a; Supplementary Information, Fig. 2c–e) and observed the localizations in C. crescentus using fluorescence microscopy (Fig. 2b). Wild-type ParA–eYFP localized as a retracting ‘comet’-like structure (Figs 1a, ,2b).2b). An ATP-binding mutant, ParAK20Q (ParAbinding)12,13,18,22,23,28 localized diffusely with puncta at the new pole (Fig. 2b). A ParA dimerization mutant, ParAG16V (ParAdimer)18,23,29, localized diffusely and in bipolar foci (Fig. 2b), and an ATP hydrolysis mutant, ParAD44A (ParAhydrolysis)18,29, colocalized with ParB foci and in patches throughout the cell (Fig. 2b). Localization of ParA proteins that contained a ParAbinding mutation, combined with a ParAdimer or a ParAhydrolysis mutation, was identical to that of the ParAbinding mutant alone (Supplementary Information, Fig. S3f). Similarly, localization of a ParA protein that contained a ParAdimer mutation, combined with a ParAhydrolysis mutation, was indistinguishable from that of the single ParAdimer mutant (Supplementary Information, Fig. S3f), consistent with the proposed hierarchy.
We assessed the role of nucleoid binding in ParA localization. We created a DNA-binding mutant, ParAR195E (ParADNA)11,25,30, and found that it localized exclusively in foci at the cell poles (Fig. 2b), suggesting a role for DNA binding in ParA localization. To further examine ParA DNA binding, we observed the localizations of ParA–eYFP mutants in Escherichia coli (Fig. 2c), which does not contain a Par system3 but has prominent nucleoid masses. In E. coli, ParAbinding–eYFP, ParAdimer–eYFP and ParADNA–eYFP all localized diffusely (Fig. 2c). By contrast, wild-type ParA–eYFP and ParAhydrolysis–eYFP localized in patches along the nucleoid (Fig. 2c and data not shown), supporting the requirements of ATP binding and dimerization for ParA interaction with DNA.
To directly examine the biochemical requirements for ParA interaction with ParB and with DNA, we used surface plasmon resonance (SPR). When we immobilized ParB and added ParA and ATP, we observed a rapid increase in response (Fig. 2d). ParA injected with ADP or without nucleotide produced a minimal response (Fig. 2d). We next immobilized the non-specific DNA duplex, parS-scr8, and assessed ParA association. ParA produced an increase in response when combined with ATP (Fig. 2e). On its own, or when combined with ADP, ParA produced a minimal response (Fig. 2e), suggesting that ATP is required for ParA polymerization and its interaction with ParB and with DNA.
As ParA readily binds DNA in vitro and in vivo, we hypothesized that nucleoid-immobilized ParA structures move the ParB-bound centromere complex through ParB-stimulated dissociation of ParA subunits from the DNA. We examined the role of ParB in ParA dynamics by localizing ParA–eYFP in ParB-depleted cells. After ParB depletion, ParA localized uniformly throughout the cell, whereas dynamic ParA–eYFP structures were observed in cells not depleted of ParB (Fig. 3a). In cells depleted of wild-type ParB, but expressing mCherry–ParB, ParA–eYFP localization was dynamic and led mCherry–ParB foci poleward (Fig. 3a). However, expression of a ParA interaction-deficient mutant, ParBL12A (ref. 32; Supplementary Information, Fig. S3a) produced static mCherry–ParB foci and diffuse ParA–eYFP localization (Fig. 3a). To dissect the role of parS, we localized ParA and ParB in E. coli cells with and without a parS-containing plasmid. ParA–eYFP expressed with or without the parS plasmid localized to the nucleoid (Fig. 3b). CFP–ParB expressed alone localized diffusely without parS, but formed foci in the presence of the parS plasmid (Fig. 3b). Co-expressed ParA–eYFP and CFP–ParB localized similarly to the single expression strains without parS, but in the presence of parS, CFP–ParB formed foci and ParA–eYFP occasionally oscillated between nucleoids (Fig. 3b, c). These results suggest that, in vivo, ParB clustered on parS stimulates the dynamic localization of ParA structures over the nucleoid.
We tested the effect of ParB on the stability of ParA–DNA complexes in vitro using SPR. When associated with a nonspecific DNA surface, ParA with ATP produced a rapid increase in response, followed by a slow dissociation with buffer only (Fig. 3d). When ParB was injected during ParA dissociation, we observed an abrupt increase in response, indicating the formation of a ParB complex with DNA-bound ParA. Subsequently, the signal rapidly decreased to well below the ParA dissociation curve, indicating the dissociation of ParA from the DNA (Fig. 3d, red). Similar results were observed using gel shifts (Supplementary Information, Fig. S3b). These data suggest that the ParB–parS complex moves relative to the ParA-bound nucleoid through simultaneous binding to and removal of ParA from the structure.
The C. crescentus ParA dynamics observed in E. coli suggest that ParA, ParB and parS are sufficient to assemble a dynamic machine. However, the polar localization of ParA mutants in C. crescentus (Fig. 2b) suggests that additional factors contribute to ParA localization. To identify polar interaction partners of ParA, we expressed the bipolar-localized ParADNA–eYFP in strains with deletions in proteins known to localize to the new cell pole. In cells lacking the new pole protein TipN1,2, we observed a decrease in the frequency of new-pole ParADNA–eYFP foci (data not shown), suggesting that TipN is required to position ParADNA. To examine the role of TipN in segregation, we visualized ParB–parS segregation in synchronized wild-type (JP138) and ΔtipN (JP141) strains. The JP138 strain had a similar efficiency of chromosome segregation as that observed for the parB::cfp-parB strain (Fig. 4a). However, the ΔtipN strain showed predominantly partial parS segregation events (Fig. 4a). Time-lapse imaging of ParA–eYFP and mCherry–ParB in ΔtipN showed that ParB–parS translocation paused frequently and reversed direction (Fig. 4b; Supplementary Information, Fig. S3c). Reversal correlated with ParA redistribution to the opposite side of the ParB–parS complex (Fig. 4b; Supplementary Information, Fig. S3c). Therefore, TipN is required to maintain ParA-mediated parS translocation directionality towards the new pole.
To determine whether ParA and TipN interact directly, we developed an assay to screen for protein–protein interactions in E. coli. This assay used a peptide from the Shigella protein IcsA (IcsA507–620, hereafter referred to as IcsA) to localize proteins to the E. coli cell pole33, allowing colocalization studies with other fluorescent proteins. Full-length C. crescentus TipN fused to IcsA localized to the E. coli pole and recruited ParADNA–eYFP (Fig. 4c), whereas IcsA alone did not (data not shown). IcsA fusions to both the TipN N-terminal domain (TipNNTD, residues 1–207) and the C-terminal domain (TipNCTD, residues 205–888) also localized to the cell pole, but only the TipNCTD recruited ParADNA–eYFP (Fig. 4c). We assayed the direct interaction of ParA with immobilized using TipNCTD in vitro SPR. On addition of ParA and ATP, we observed an increase in signal corresponding to ParA binding that was specific for TipNCTD (Fig. 4d). ParA and ADP, or no nucleotide, produced a lower signal than that observed with ATP (Fig. 4d), suggesting that apo-ParA interacts directly with the C-terminal region of TipN, and that ATP augments the interaction.
Together, our data support a burnt-bridge Brownian ratchet model for Par-mediated chromosome segregation in C. crescentus (Fig. 5a, b). In vitro, ParA formed linear polymers, but also interacted readily with DNA in vitro and in vivo, suggesting that ParA polymers may form either along the nucleoid or freely in the cytoplasm, or both, and bundle into a linear structure (Fig. 5a, vi). In vitro, ParB removes ParA from DNA, consistent with our observations in vivo that ParB depletion or mutation quenches ParA dynamics, and that wild-type ParB complexes ‘follow’ a receding ParA structure. Thus, we propose that ParB stimulates the dissociation of ParA subunits from the ends of a ParA structure while remaining attached, moving the ParB-parS complex along a retracting ParA structure (Fig. 5a, vii). The simultaneous interaction with, and dissociation of, the ParA structure may be explained by the association of multiple ParB proteins with the parS region34,35. Thermal motion of the ParB-parS complex may be trapped by ParB binding to the ParA structure as the structure shortens, explaining the rectified diffusional motion observed for ParB complexes in Vibrio cholerae36. Finally, our data suggest that ParB-parS complexes move along a subset of fibres within the ParA bundle, as a less intense structure is often left behind the translocating ParB complex. Thus, ParA may be available for ParB-stimulated removal only when located at protofilament termini.
The C. crescentus Par system mobilizes the parS locus unidirectionally from the old pole to the new pole37, in contrast to the bidirectional movement observed for plasmid segregation15. One contributor to unidirectionality in C. crescentus is the polar protein PopZ, which tethers ParB-parS to the cell pole8,9 (Fig. 5b, i) to prevent reversals. Here we identify a new directionality factor for the C. crescentus Par system: the new pole-specific protein TipN1,2. Without TipN, ParA localizes aberrantly, causing pauses and reversals in ParB–parS segregation. These defects observed in the absence of tipN may reflect secondary effects, such as on the MreB-associated cytoskeleton1. However, ParA and TipN interact directly in vitro (Fig. 4d), suggesting a functional interaction in vivo. TipN might nucleate or stabilize ParA structures at the new pole (Fig. 5b, i). Alternatively, TipN might simply provide a binding site for ParA to increase the local concentration and bias the insertion of free ParA molecules into the structure at the new pole. After segregation, the translocated ParB–parS complex is anchored to PopZ at the new pole (Fig. 5b, v), while TipN is recruited to the division plane to remain at the new poles of the daughter cells to reset the cycle.
Overall, the basic operating principles that drive DNA segregation seem to be shared between prokaryotic and eukaryotic mitotic machineries. The bacterial ParB–parS complex shares functional and architectural similarities with the eukaryotic kinetochore complexes, as both associate with, and spread along, the centromere DNA region38. Both C. crescentus and eukaryotic kinetochores seem to use multivalent attachments to allow the simultaneous binding to, and depolymerization of, the polymers that guide their movement, reminiscent of the eukaryotic DamI–Ndc80 complex proposed to follow along depolymerizing microtubule ends38. Finally, polar TipN may function as a centrosome-like organization centre to bias the movement of retracting polymers towards the cell pole.
Methods and any associated references are available in the online version of the paper at http://www.nature.com/naturecellbiology/
We thank Jimmy Blair for assistance with modelling of ParA mutants, and critical reading of the manuscript; and Grant Bowman, Erin Goley and Julie Biteen for technical advice. We thank Jian Zhu and Thomas Earnest for providing purified 6His–ParB. This work is supported by National Institutes of Health grants R01 GM51426 R24 and GM073011-04d to L.S., NIH/NIGMS fellowship F32GM088966-1 to J.P., NIH/NIGMS award R01GM086196-2 to W.E.M., the Smith Stanford Graduate Fellowship to E.T., and a Helen Hay Whitney postdoctoral fellowship to E.G. This work was also supported by the Director, Office of Science, Office of Biological and Environmental Research, of the U.S. Department of Energy under contract no. DE-AC02-05CH11231.
Supplementary Information is available on the Nature Cell Biology website.
AUTHOR CONTRIBUTIONSJ.P., S.L., W.E.M. and L.S. designed the research; J.P. performed C. crescentus genetic, epifluorescence microscopy and biochemical experiments; S.L. performed single molecule imaging and data analysis; E.G. purified native ParA and performed ParA light-scattering experiments; E.T. designed ParA/DNA SPR experiments and performed time-lapse microscopy experiments on ΔtipN strains; M.E. performed SPR experiments and analysis; L.C. performed ParA negative-stain electron microscopy imaging; W.E.M. and L.S. supervised the study; J.P., S.L., W.E.M. and L.S. wrote the paper.
The authors declare no competing financial interests.
Reprints and permissions information is available online at http://npg.nature.com/reprintsandpermissions/