Bacterial DNA segregation by dynamic SopA polymers

Abstract

Many bacterial plasmids and chromosomes rely on ParA ATPases for proper positioning within the cell and for efficient segregation to daughter cells. Here we demonstrate that the F-plasmid-partitioning protein SopA polymerizes into filaments in an ATP-dependent manner in vitro, and that the filaments elongate at a rate that is similar to that of plasmid separation in vivo. We show that SopA is a dynamic protein within the cell, undergoing cycles of polymerization and depolymerization, and shuttling back and forth between nucleoprotein complexes that are composed of the SopB protein bound to sopC-containing plasmids (SopB/sopC). The dynamic behavior of SopA is critical for Sop-mediated plasmid DNA segregation; mutations that lock SopA into a static polymer in the cell inhibit plasmid segregation. We show that SopA colocalizes with SopB/sopC in the cell and that SopB/sopC nucleates the assembly of SopA and is required for its dynamic behavior. When SopA is polymerized in vitro in the presence of SopB and sopC-containing DNA, SopA filaments emanate from the plasmid DNA in radial asters. We propose a mechanism in which plasmid separation is driven by the polymerization of SopA, and we speculate that the radial assembly of SopA polymers is responsible for positioning plasmids both before and after segregation.

Chromosome segregation in eukaryotes is mediated by dynamic proteins such as tubulin, which polymerize into filamentous structures that are capable of rapid reorganization by means of cycles of depolymerization and repolymerization (reviewed in refs. 1 and 2). Although much less is understood about the mechanism of DNA segregation in bacteria, dynamic bacterial actin homologues have recently been shown to play an important role in the process (3–7). The MreB proteins, for example, are members of the actin superfamily and have been implicated in chromosome segregation in Escherichia coli, Bacillus subtilis, and Caulobacter crescentus (8–11), although MreB is not essential for chromosome segregation in B. subtilis (12). The ParM protein encoded by the E. coli plasmid R1 represents another subfamily of bacterial actins and is required for the movement of replicated plasmids from midcell toward the cell poles (4). ParM assembles into two-stranded helical filaments in an ATP-dependent manner; these filaments are actin-like in structure but exhibit the dynamic instability of eukaryotic microtubules (13–15). Polymerization of ParM between plasmids is thought to drive plasmid separation (14–16).

One of the first DNA segregation systems to be identified in E. coli was the F plasmid Sop system (17), which consists of two protein components, SopA and SopB, and a cis-acting DNA sequence, sopC (17, 18). SopA is not a member of the actin superfamily but belongs instead to the unrelated ParA family of P-loop ATPases that have been implicated in plasmid and chromosome segregation and are highly conserved in bacteria and archaea (3–6, 19–25). SopB is a DNA-binding protein that binds cooperatively to the 12 SopB binding sites that constitute sopC (reviewed in ref. 19). Upon binding at sopC, SopB spreads out along the DNA beyond sopC and coats the DNA in an extended nucleoprotein complex (reviewed in ref. 19). In vitro, SopA interacts with SopB that is bound to sopC, and this interaction stimulates the ATPase activity of SopA (26–28).

There have been several efforts to determine the intracellular localization of ParA proteins. SopA was found to form foci near the cell poles (29); ParA from plasmid P1 was distributed uniformly throughout the whole cell (30); ParA from plasmid pB171 oscillated as a helical spiral (31, 32); and Soj, the chromosomally encoded ParA from B. subtilis, oscillated within cells when they entered stationary phase (33, 34). These different ParA proteins may indeed have very different localization patterns, but it is also possible that each of these studies captured specific features of a very complicated localization profile.

Plasmids containing a ParA system are positioned at the center of newly divided E. coli cells (31, 35–39). After plasmid replication, the two daughter plasmids separate and reposition from midcell to quarter-cell locations that correspond to the midcell of the future daughter cells. If their Par systems are deleted, these plasmids tend to be randomly distributed in the cell and are frequently lost upon cell division (31, 35, 40). Plasmids containing different (compatible) ParA systems localize to distinct midcell positions and separate at different times, suggesting that different Par systems function independently from each other for both positioning and separation (41, 42).

It is unclear how the ParA systems recognize the center of the cell or the quarter-cell positions and how they bring about the separation of replicated plasmids. A distant ParA relative, ParF of plasmid TP228, was recently shown to assemble into polymers in vitro, suggesting that, as in the case of the actin-like ParM, polymerization might provide the driving force for plasmid segregation (43).

Here we show that SopA assembles into filaments both in vitro and in vivo, and that plasmid segregation requires that SopA assembly in vivo be dynamic. In vivo, SopA assembly is nucleated by SopB bound to sopC, and both proteins (SopA and SopB) colocalize with the plasmid DNA. In vitro, SopA assembles into filaments that in the presence of SopB and sopC DNA radiate from the SopB/sopC nucleoprotein complex to form spindle-like structures. Our results suggest a model for plasmid segregation that explains how a ParA system can accomplish both plasmid separation and plasmid positioning.

Materials and Methods

Strains and Growth Conditions. Strains were grown at 30°C, unless otherwise indicated, in LB supplemented with appropriate antibiotics: 100 μg/ml ampicillin, 30 μg/ml chloramphenicol, or 50 μg/ml kanamycin. Plasmids were electroporated into JP313 (44) or DHB4 (45) to yield JP5034 (JP313/pGEL11), JP5036 (DHB4/pGEL10), JP5037 (DHB4/pGEL11), JP5053 (JP313/pGEL10), JP5067 (DHB4/pGEL10/pGEL26), JP5081 (JP313/pGEL10/pGEL26/pGEL30), JP5119 (JP313/pGEL10/pGEL39), and JP5121 (DHB4/pGEL10/pGEL39). DHB4 contains F′sopABC⁺ lac⁺ and produces blue colonies on plates containing X-Gal, but produces white or sectored blue colonies if F′lac is destabilized.

Plasmids. Plasmids were constructed by cloning PCR fragments that were amplified with the primers listed in Table 1, which is published as supporting information on the PNAS web site, and described in Supporting Methods, which is also published as supporting information on the PNAS web site. All constructs were verified by DNA sequencing. Briefly, pGEL6, containing the gene that codes for GFP fused to SopA (SopA-GFP), was constructed by cloning a 1,326-bp fragment encoding SopA and its promoter from pXX167 (17) into the ClaI site of pMUTIN-GFP+ (46). This construction gave rise both to a wild-type SopA-GFP plasmid and to another plasmid that had two spontaneous mutations in SopA, M315I and Q351H. This mutant was designated SopA1-GFP. pGEL10 and pGEL11 express SopA-GFP or SopA1-GFP, respectively, at low levels from a weakened isopropyl β-d-thiogalactoside (IPTG)-controlled trc promoter of pDSW206, and were constructed by subcloning the wild-type sopA-gfp and sopA1-gfp into the XbaI site of pDSW206 (47). pGEL39 is a pACYC replicon that contains sopA1 (without GFP) under the control of the arabinose promoter and was constructed by subcloning sopA1 from pGEL11 into the XbaI site of pBAD33 (48). Plasmid pGEL26, containing the gene that codes for cyan fluorescent protein (CFP) fused to SopA (SopA–GFP), was constructed by first amplifying sopB (998 bp) from pXX167 and subcloning it into the ClaI site of pMUTIN-CFP (46) to yield plasmid pGEL24. Then sopB-cfp was amplified from pGEL24 and cloned into the XbaI site of pBAD33 to yield plasmid pGEL26. pGEL30, which contains the sopC region (518 bp), is a kanamycin-resistant miniRK2 plasmid derived from pRR10 (49) and was engineered by using the λ red recombination system as described in ref. 50. SopA–GFP and SopB–CFP were demonstrated to be functional on the basis of the ability to properly position plasmid pGEL30 within the cell. Neither protein interfered with the wild-type sopABC system on plasmid F.

pGEL16 (His₆-SopA) and pAID3037 (His₆-SopB) are pET28a derivatives. JP5038 is (BL21(DE3)/pGEL16), and JP3039 is (BL21(DE3)/pAID3037).

Live Cell Microscopy. Strains were grown overnight in LB at 30°C, diluted 1/100 into fresh medium, and grown to an OD₆₀₀ of 0.1–0.2. Expression of SopA-GFP and SopA1-GFP was induced for 1–2 h with IPTG concentrations ranging from 2 to 100 μM, depending on the particular experiment. SopB-CFP was induced for 40 min with 0.2% arabinose, repressed with 0.2% glucose to prevent overexpression, and imaged 20 min later. Cells were stained with 0.5 μg/ml of FM 4-64 (51) and affixed to a poly(l-lysine)-treated coverslip. Images were captured as described in ref. 37 by using an optical sectioning microscope (Applied Precision, Issaquah, WA) equipped with a QLM laser module for fluorescence recovery after photobleaching (FRAP) experiments. Six to eight optical sections, 0.15 μm apart, were collected; medial focal planes are shown. Fluorescence micrographs of wild-type SopA-GFP fluorescence show the undeconvolved image. SopA-GFP gave rise to complex patterns of fluorescence whose intensities varied over a 10-fold range within a single cell and which were difficult to represent faithfully as images. To display this data quantitatively, we exported the pixel intensity data for a given field to Microsoft Excel to produce graphs that display the intensity values for the raw and unadjusted data. Time-lapse microscopy was performed on strains JP5036 and JP5037 as described in ref. 37.

For FRAP experiments, part of a cell was photobleached with a 0.5-sec pulse at 100% power from an argon laser (488 nm) focused in an ≈2 μm beam. Images were collected at various time intervals immediately afterward. Control experiments demonstrated that, under these conditions, GFP was irreversibly bleached, with no fluorescence recovery after 20 min.

SopA and SopB Purification and Labeling. His₆-SopA and His₆-SopB were purified by nickel affinity chromatography using standard methods (see Supporting Methods). Purified undialyzed SopB (100 μg) was labeled with Alexa Fluor 488 by reaction with a tetrafluorophenyl ester conjugate (Alexa Fluor 488 monoclonal antibody labeling kit, Molecular Probes), and unbound label was removed according to the manufacturer's instructions.

In Vitro Polymerization Assay. Twenty-micromolar SopA was incubated with 2 mM ATP, 20 mM MgCl₂, and 1 μM Nile red at room temperature in a 5-μl reaction volume for 5 min and mounted on a slide. After 10–30 min of incubation, images of SopA filaments were collected. Filaments were not observed unless SopA was present. Polymer formation occurred more rapidly on a glass slide than in a polypropylene tube, as has been reported for FtsZ (52). To monitor SopA polymerization in the presence of SopB and sopC, 10 μM SopA, 2 μM SopB, 3.5 mg/ml pGEL30 (contains sopC region), 2 mM ATP, 20 mM MgCl₂, and 1 μM Nile red in a 9-μl reaction volume were incubated at room temperature for 1 h and mounted on a slide. To monitor SopA polymerization in the presence of labeled SopB and sopC, 10 μM SopA, 2 μM Alexa 488-SopB, 34 mg/ml pXX167, 3 μg/ml DAPI, 2 mM ATP, 20 mM MgCl₂, and 1 μM Nile red in an 8-μl reaction volume were incubated for 5 min at room temperature and mounted on a slide.

Results

Purified SopA Assembles into Polymers in Vitro. SopA was tagged at its N terminus with hexahistidine and purified over a nickel affinity gel column. When incubated with ATP, SopA polymerized into long filaments, a process that could be visualized in the fluorescence microscope when the protein was stained with the nonspecific stain Nile red (Fig. 1A) (53, 54). The lengths of 28 filaments were measured and ranged from 4.1 to 15.9 μm, with a mean length of 10 ± 3 μm. Polymerization did not occur unless ATP was present, indicating a requirement for ATP binding or hydrolysis.

Fig. 1.

Filament formation by SopA in vitro.(A) Filaments were grown in the presence of ATP and stained with Nile red as described in Materials and Methods. (Scale bar, 1 μm.) (B) The distance between the ends of a growing filament were measured and the lengths were plotted vs. time. The average elongation rate for the six filaments is 0.18 μm/min. (C and D) Time-lapse microscopy of filament formation. The elongation of a single filament is followed in each row. Numbers at the bottom of each micrograph indicate filament length in μm.

Time-lapse fluorescence microscopy enabled us to follow the kinetics of assembly of individual filaments. SopA polymers elongated at an average rate of ≈0.18 μm/min (Fig. 1 B and C) (n = 6; SD ± 0.05). These rates are similar to those at which bacterial plasmids and chromosomal DNA have been observed to separate in vivo during segregation. For example, the average rate of separation for plasmids F, RK2, and P1 is 0.43, 0.3, and 0.41 μm/min, respectively, and for the chromosomal oriC DNA of C. crescentus, B. subtilis, and E. coli, the average rate of separation is 0.27, 0.17, and 0.07 μm/min, respectively (10, 37, 55–57). These findings suggest that SopA polymerization might be an integral part of the mechanism by which plasmid DNA is segregated within the cell.

SopA-GFP Assembles into Polymers Within the Cell. To investigate whether SopA also forms filaments in vivo, we constructed a functional fusion of GFP to the C terminus of wild-type SopA. When SopA-GFP was expressed by itself, without the F sopABC⁺ plasmid, a diffuse fluorescence of uniform intensity appeared to coat the chromosome (Fig. 2A). This pattern was most likely due to the nonspecific binding of SopA-GFP to DNA (29). When expressed in the presence of all three sop components, i.e., when an F (sopABC⁺) plasmid was present in the cell, SopA-GFP gave rise to a complex pattern of intracellular fluorescence. Most cells contained discrete foci (60%; Fig. 2B) or a nonuniform haze of fluorescence (35%; Fig. 2 C and D). The haze typically took the form of a symmetric midcell peak that dropped off in intensity toward the two poles (Fig. 2C) or a quarter-cell peak with an asymmetric haze that extended toward the opposite pole (Fig. 2D). Some cells contained two foci connected by a central haze (Fig. 2E). A few cells contained either uniform diffuse fluorescence (4%), which suggested that they had lost the F plasmid, or short filaments (<1%; Fig. 7, which is published as supporting information on the PNAS web site). Given our finding that purified SopA assembles into polymers in vitro, we considered the possibility that the haze of fluorescence might correspond to polymers that were too numerous, too close together, or too dynamic to be resolved by fluorescence microscopy.

Fig. 2.

Localization of SopA-GFP and SopB-CFP. (A–E) SopA-GFP (green) was expressed from plasmid pGEL10 by induction with 2 μM IPTG, as described in Materials and Methods. Cell membranes were stained with FM 4-64 (red). (A) Strain JP5053, which contains no other Sop system components. (B–E) Strain JP5036, which contains an intact Sop system. Foci (B) were observed in 60% of the cells, and hazes (C and D) were observed in 35% of the cells. In the fluorescence intensity plots, colors indicate pixel intensity on a linear scale in arbitrary units: blue, 0–200; red, 200–400; yellow, 400–600; green, 600–800; purple, 800–1,000; and orange, 1,000–1,200. (F) Colocalization of SopA-GFP foci with SopB-CFP foci in strain JP5067. SopA-GFP (green) was expressed from plasmid pGEL10 by induction with 2 μM IPTG, and SopB-CFP (blue) was expressed from plasmid pGEL26 by induction with arabinose as described in Materials and Methods. (Scale bars, 1 μm.)

To explore this possibility further, we monitored these F⁺ cells with time-lapse microscopy. In nearly all of the cells, SopA-GFP oscillated between the quarter-cell positions with a period of ≈20 min (Fig. 3). SopA-GFP appeared as a focus at the quarter-cell position (0 min) that then redistributed into a haze around the three-quarter-cell position (5 min). This haze then coalesced into a focus at the three-quarter-cell position (10 min) and subsequently redistributed into a haze around the quarter-cell position (15 min). This haze then coalesced again into a focus at the quarter-cell position (20 min), completing one cycle of oscillation. The foci and hazes observed in the still images (Fig. 2 B–E) therefore represent individual steps within this oscillation cycle. This oscillation might correspond to cycles of SopA polymerization and depolymerization.

Fig. 3.

Time-lapse microscopy of SopA-GFP. SopA-GFP (green) was expressed from plasmid pGEL10 in strain JP5036 by induction with 2 μM IPTG, as described in Materials and Methods. (Upper) Cell membranes were stained with FM 4-64 (red). Time in minutes is shown in white text in the top right corner of each panel. (Scale bar, 1 μm.) (Lower) Intensity plot of GFP fluorescence. Colors indicate pixel intensity, as in Fig. 2. The oscillation was observed in 98% of the cells in the field (n = 196).

SopA Colocalizes with SopB in the Cell. The presence of SopA-GFP foci at the quarter-cell positions, where the Sop system positions the F plasmid, suggested that SopA-GFP was assembling at the plasmid. We therefore coexpressed SopA-GFP together with SopB-CFP, which binds to sopC on the plasmid and forms bright fluorescent foci that mark the position of the plasmid within the cell (Fig. 7) (29, 58). We found that 94% of the SopA-GFP foci were coincident with SopB-CFP foci (n = 100; Fig. 2F). This finding demonstrated that SopA-GFP does assemble at the plasmid and that SopA and SopB likely interact at the plasmid in vivo. It is also in keeping with previous studies demonstrating that SopA interacts with SopB (26–28).

SopB and sopC Nucleate SopA Assembly in Vivo. To determine whether SopB bound to sopC-containing DNA (SopB/sopC) serves as a nucleation site for SopA assembly within the cell, we asked whether the induction of SopB could cause SopA to relocalize to the plasmid. We expressed SopA-GFP in a strain carrying another plasmid that contained SopB-CFP under the control of the arabinose promoter and a third plasmid that contained sopC. In the absence of arabinose, where SopB-CFP was not produced, SopA-GFP gave rise to the diffuse chromosomal fluorescence that we had observed when SopA-GFP is expressed in the absence of the other Sop components (Fig. 4D and Fig. 2 A). After 30 min of SopB-CFP induction with arabinose, SopA-GFP relocalized in 99% of cells (n = 133), into an asymmetric haze (3%), short filaments (15%), or bright foci (81%) (Fig. 4E), and 96% of these SopA-GFP foci colocalized with SopB-CFP foci. These results demonstrate that SopA-GFP assembly is nucleated by SopB/sopC within the cell.

Fig. 4.

Recruitment and nucleation of SopA by SopB/sopC.(A–C) Radial aster formation by SopA in vitro when the other Sop components are present. Assembly reactions with SopA, SopB, plasmid pGEL30, and ATP were carried out as described in Materials and Methods. (A and B) Staining was with Nile red. (C) SopB was covalently labeled with Alexa 488 (green), and pGEL30 DNA was stained with DAPI (blue). Nile red stains both SopA and SopB. No asters are observed if any of the components are omitted from the assembly reactions (data not shown). (D and E) SopB recruits SopA. In strain JP5081, SopA-GFP (green) was expressed from plasmid pGEL10 by induction with IPTG, and SopB-CFP (blue) was expressed from plasmid pGEL26 by induction with arabinose, as described in Materials and Methods and Results.(D) Before induction of SopB-CFP, SopA-GFP gives rise to diffuse fluorescence. (E) After induction of SopB-CFP, SopA-GFP colocalizes with SopB-CFP at the plasmid. Cell membranes were stained with FM 4-64 (red). (Scale bars, 1 μm.) In the absence of SopA-GFP, SopB-CFP fluorescence tracks with the chromosome, owing to its ability to bind to DNA nonspecifically in the absence of sopC DNA (Fig. 7).

SopB Nucleates Assembly of SopA into Radial Asters in Vitro. In light of these in vivo findings, we examined the effect of including purified SopB and a plasmid containing sopC in our in vitro SopA assembly reactions. In the presence of ATP, SopB, and sopC-containing DNA, SopA polymers appeared to form radial asters originating from a single point, with long thin strands of fluorescence projecting in all directions (Fig. 4 A and B). No asters were observed if any of the components were omitted from the assembly reactions. The appearance of these asters suggested that SopB/sopC might be functioning as an organizing center for SopA filaments. The assembly reaction was therefore repeated with SopB that had been labeled with the fluorophore Alexa 488. The SopA filaments that formed radiated from a centrally located SopB that coincided with the DAPI-stained DNA (Fig. 4C). Thus, in vitro, SopB/sopC organizes SopA filaments into a radial aster.

Identification of a SopA Mutant Affecting Polymerization and Segregation. In the course of constructing SopA-GFP, we isolated a mutant (SopA1-GFP) that formed long filaments in most cells even in the absence of the other sop genes (Fig. 5A). This mutant, which contained two mutations in SopA, did not give rise to the complex localization profile of the wild-type SopA-GFP, producing neither foci nor hazes. In time-lapse microscopy experiments, the filaments appeared stable with no change in localization over time. We therefore considered the possibility that the primary defect of SopA1 was that it had lost the dynamic properties of the wild-type SopA. FRAP was therefore used to determine the rate at which subunits within the SopA1-GFP filaments exchanged with subunits from the cytoplasmic pool. When a small region of a cell containing a section of the filament was bleached by exposure to intense laser light, no recovery of fluorescence was observed even after 5 min, indicating that the SopA1-GFP filaments were static (Fig. 5B). When the same experiment was attempted with SopA-GFP expressed from pGEL10 in strain JP5036, the dynamic behavior of SopA-GFP made the outcome uninterpretable. In strain JP5053, where SopA-GFP is expressed from pGEL10 in the absence of the other Sop components, most of the fluorescence was recovered 4 min after bleaching (Fig. 7). The untagged SopA1 was also able to coassemble with the wild-type SopA-GFP. SopA1 was expressed from the inducible arabinose promoter in a strain expressing the wild-type SopA-GFP fusion. Before induction of SopA1, SopA-GFP localized as expected as a uniform distribution along the chromosome if SopB and sopC were absent (Fig. 5E) and as foci or hazes when SopB and sopC were present (Fig. 5G). In both cases, after induction of SopA1, wild-type SopA-GFP assembled into filaments that extended the length of the cell. These filaments appeared to be extraordinarily stable, because cell division could not be completed across them. After 1 h of SopA1 induction, the cells formed long chains that appeared to be held together by the SopA1/SopA-GFP filaments (Fig. 5 F and H). The mutant and wild-type SopA proteins therefore interact within the cell and assemble mixed polymers or bundles of polymers. That SopA1 can incorporate the wild-type SopA into filaments illustrates that the ability to assemble into filaments is a shared property of SopA1 and the wild-type SopA.

Fig. 5.

Properties of the SopA1 mutant. (A) SopA1-GFP (green) was expressed from plasmid pGEL11 in strain JP5034 by induction with 100 μM IPTG as described in Materials and Methods. Cell membranes were stained with FM 4-64 (red). (B) FRAP analysis of a SopA1-GFP (green) filament in strain JP5034. No fluorescence recovery is apparent at 5 min after bleaching. Cell membranes were stained with FM 4-64 (red). (C and D) Strains JP5036 (C) and JP5037 (D) were plated on LB agar plates containing 40 μg/ml X-Gal and incubated overnight at 30°C. The blue (Lac⁺) colonies of JP5036 indicate that F′lac is stably inherited. The white colonies and sectored blue colonies of JP5037 indicate that F′lac is unstable. (E–H) SopA-GFP was expressed from plasmid pGEL10 by induction with 100 μM IPTG, and SopA1 was expressed from plasmid pGEL39 by induction with 0.2% arabinose for 30 min. (E and F) Strain JP5119. (G and H) Strain JP5121. Before induction of SopA1, SopA-GFP tracks the chromosome (E) or forms foci and hazes (G); after induction, SopA-GFP is polymerized into filaments that link the cells in chains (F and H). (Top) Cell membranes stained with FM 4-64 (red). (Middle) SopA-GFP (green). (Bottom) Overlay. (Scale bars, 1 μm.)

Having demonstrated that SopA1 was able to trap the ordinarily dynamic wild-type SopA into static filaments, we determined the effect of a shift from dynamic to static on Sop-mediated plasmid segregation. The wild-type SopA-GFP fusion and the SopA1-GFP fusion were expressed at similar levels in a strain carrying an F′lac plasmid with an intact wild-type Sop system. Expression of SopA1-GFP but not of wild-type SopA-GFP destabilized the F′lac plasmid dramatically. When plated on LB X-Gal plates, only 1% of cells expressing SopA-GFP (n = 637) and only 2% of cells with the vector alone (n = 439) lost the F′lac and produced white colonies (Fig. 5C). In contrast, 76% of cells expressing SopA1-GFP (n = 372) produced colonies that were either white or blue with white sectors, indicating that the F′lac was rapidly lost from the population (Fig. 5D). Thus, a shift from dynamic to static SopA assembly disrupts the function of the Sop plasmid segregation system.

Discussion

Plasmid-encoded ParA proteins are required both to position DNA within the cell and to rapidly separate replicated plasmid DNA. Here we show that SopA polymerizes into filaments nucleated by SopB/sopC and that SopA polymerization plays a central role in plasmid segregation. The SopA1 mutant protein traps the wild-type SopA in static polymers and disrupts plasmid segregation. It is therefore not merely the ability of SopA to polymerize that is essential for Sop system function but the ability to do so dynamically, to undergo reversible cycles of polymerization and depolymerization.

The SopA1 protein forms stable filaments within the cell that were easily visualized by microscopy, but the dynamic wild-type SopA protein was more difficult to capture visually, assembling either a focus at a SopB/sopC nucleoprotein complex or a haze that emerges from one. The haze does not resemble the uniform fluorescence generated by soluble GFP, nor does it track with the chromosome, as it does when SopA-GFP is expressed in the absence of the other Sop components. The haze is either symmetric and positioned at midcell or asymmetric and extended from one of two SopB/sopC complexes toward the other. The haze likely arises from a cellular structure that is fundamentally similar to our in vitro reconstructions, in which multiple filaments of SopA project outward radially from a single SopB/sopC complex. Time-lapse microscopy underscores the dynamic nature of SopA and suggests that the haze might correspond to an intermediate stage in the relocalization of SopA from one SopB/sopC complex to the other. Relocalization might occur either through depolymerization of SopA at one SopB/sopC complex, diffusion within the cell, and repolymerization at the other SopB/sopC complex, or possibly through the capture of the ends of polymers by the other SopB/sopC complex.

Similar hazes are often produced by eukaryotic cytoskeletal proteins whose abilities to polymerize into filaments within cells is very well established. In Caenorhabditis elegans, for example, only a haze of fluorescence can be observed in the vicinity of the centrosome when tubulin is localized by immunofluorescence with anti-α-tubulin antibodies, although distinct microtubules are visible several micrometers away (see figure 1 A in ref. 59). We suspect that the SopA filaments are too short (the haze extends for only 500 nm), too numerous, or too dynamic to be resolved in our experiments. The ParA protein of plasmid pB171 and Soj of B. subtilis give rise to similar asymmetric hazes and oscillate from one end of the cell to the other (31, 33, 34). Although the pB171 ParA was resolved into spiraling helices after deconvolution (31), we observed no SopA-GFP helices in raw or deconvolved images from either still or time-lapse experiments.

SopA filaments elongate in vitro at the same rate as plasmids separate from each other in the cell, suggesting that SopA polymerization is likely to be coupled to the physical separation of daughter plasmids. Because SopA filaments extend outward from the SopB/sopC complex, they most likely separate plasmids by polymerizing between them and pushing them apart, much as ParM has been proposed to separate R1 plasmids (4). But unlike the R1 system, in which a single bundle of ParM filaments is proposed to push plasmids apart, multiple SopA filaments are propagated from the SopB/sopC complex, and these filaments project not just linearly along the axis between the plasmids, but radially outward from the SopB/sopC hub. This structure of SopA polymers resembles the mitotic spindle, and it is noteworthy that the mitotic spindle has the capacity not only to separate chromosomes but also to move itself to the center of the cell (1, 2, 60). This centering ability is accomplished through contact of microtubules with the cell cortex. As forces applied by or to the spindle are propagated in a radial array along microtubules, the spindle is moved to the center of the cell, where the applied forces are equal. The forces may result from either pushing or pulling, and they can be generated by motor proteins or directly by microtubule polymerization and depolymerization (60). We propose that the network of SopA filaments projecting from the SopB/sopC nucleoprotein complex could function in a similar manner to position plasmids within the cell. Before plasmid replication, the radial network would center the plasmid within the cell (Fig. 6A). After plasmid replication, polymerization of SopA between the two SopB/sopC complexes would move them apart, with the movement ceasing once equal force was projected in all directions from the two equivalent radial networks (Fig. 6 B and C). After replication and separation, the two daughter plasmids would be positioned equidistant from each other and from the cell poles (Fig. 6C). This model, if correct, would explain many of the functions of a Par system: the positioning of plasmids at midcell before replication, the separation of daughter plasmids after replication, and the repositioning of daughter plasmids once they are separated.

Fig. 6.

Proposed mechanism for plasmid segregation as mediated by the Sop system. (A) Before plasmid replication, a radial array of SopA polymers (red) assembled on a nucleoprotein complex of SopB (green) bound to sopC-containing plasmid (blue) centers the plasmid at midcell through a force-balancing mechanism. (B) After replication, polymerization of SopA between the SopB/sopC complexes separates the daughter plasmids. (C) The radial arrays of SopA polymers emanating from the two SopB/sopC complexes center the plasmids with respect to each other and with respect to the cell boundaries by a force-balancing mechanism.

SopA shares the ability to undergo dynamic polymerization with other members of the ParA family. Both MinD and ParF assemble into polymers in vitro, and MinD-GFP fusion proteins are highly dynamic in vivo (61–64). The mechanism that we propose here might therefore also apply to other members of this well conserved family.