Structure and filament dynamics of the pSK 41 actin-like ParM protein : implications for plasmid DNA segregation

David Popp , Weijun Xu, Akihiro Narita, Anthony J. Brzoska, Ronald A. Skurray, Neville Firth, Umesh Goshdastider, Yuichiro Maéda, Robert C. Robinson, Maria A. Schumacher From the, ERATO "Actin Filament Dynamics" Project, Japan Science and Technology Corporation, c/o RIKEN Harima Institute at Spring 8, 1-1-1 Kouto, Sayo, Hyogo, 679-5148, Japan, Institute of Molecular and Cell Biology, 61 Biopolis Drive, Proteos, 138673, Singapore, University of Texas M. D. Anderson Cancer Center, Department of Biochemistry and Molecular Biology, Houston, Texas, U.S.A. School of Biological Sciences, University of Sydney, Sydney, New South Wales 2006, Australia, Nagoya University Graduate School of Science, Structural Biology Research Center and Division of Biological Sciences, Furo-cho, Chikusa-ku, Nagoya, 464-8601, Japan. Running title: Structure of pSK41 ParM Address correspondence to: Dr. Maria A. Schumacher, Department of Biochemistry and Molecular Biology, Unit 1000, UT MDACC, 1515 Holcombe Boulevard, Houston, TX 77030, U.S.A., Tel: +1-713-834-6392; Fax: +1-713-834-6382; Dr. David Popp; ERATO "Actin Filament Dynamics" Project, Japan Science and Technology Corporation, c/o RIKEN Harima Institute at Spring 8, 1-1-1 Kouto, Sayo, Hyogo, 679-5148, Japan; Institute of Molecular and Cell Biology, 61 Biopolis Drive, Proteos, 138673, Singapore. Tel: +65-6586-9755; Fax: +65-779117 E-mail: maschuma@mdanderson.org; dpopp@imcb.A-star.edu.sg

Partition or segregation is the essential process whereby the genetic material is actively distributed into daughter cells (1)(2)(3). Prokaryotic plasmid partition systems represent excellent model systems to study DNA segregation at a structural level because they require only three elements: a cis-acting centromere DNA site, an NTPase, and a centromere-binding protein (1)(2)(3). In the first step of segregation, multiple centromerebinding proteins bind cooperatively to the centromere to form a partition complex. The partition complex then recruits the NTPase protein, which ultimately drives the physical movement of plasmids to opposite cell poles (1)(2)(3)(4). Plasmid-encoded partition (par) systems are typically divided into two main families based on the kind of NTPase present (3). Type I par systems contain NTPases with deviant Walker-A type folds, whereas the type II systems employ NTPases termed ParM (5)(6)(7)(8). More recently, type III and IV partition systems have been identified. The type III systems use NTPases with putative tubulin-like folds, whereas the type IV systems carry out partition using a single, non-NTPase protein with a helix-turn-helix motif and coiled-coil domain (9 -12).
How the partition proteins mediate the actual physical separation of plasmids has been a central question in partition. Studies on the type II R1 par system have provided key insights into this question. Specifically, experiments showing that the R1 ParM NTPase forms filamentous structures, led to the insertional polymerization model for partition (13)(14)(15)(16)(17). According to this model ParM filaments are first captured between paired partition complexes and continued filament growth, made possible by polymer stabilization, leads to plasmid movement to opposite cell poles (13). Remarkably, the R1 partition process can be reconstituted in vitro using only ParM, ParR, and the centromere, parC, which indicates that cellular factors are not required for the general separation mechanism (16). These studies also showed that the filaments grow by insertion of ParM-ATP molecules at the ParR-parC interface (16). Binding to ParR-parC stabilizes the filaments against catastrophic collapse as the filaments show dynamic instability (17). Structural studies showed that R1 ParM belongs to the actin/Hsp70 superfamily of ATPases, which not only includes actin and molecular chaperones but also FtsA, various sugar kinases, and the prokaryotic cell shape-determining protein, MreB (7, 18 -20). All members of this superfamily share the same basic architecture (19), but the details vary and additional domains are often present. At the amino acid sequence level the degree of similarity within the family is low. Several actin homologues have also recently been identified in Archaea. The structure of one such archaeal actin-like protein, Ta0583 from Thermoplasma acidophilum, has been solved and it has been suggested that Ta0583 may be an ancestral actin (21,22).
Recent work has started to shed light on the filamentous structures of actin family proteins (23)(24)(25). These studies have revealed the remarkable finding that although these actin-like proteins share similar overall folds, they often form very different filament structures that may be optimized for their specific cellular functions (23,24,26). For example, electron microscopy analyses on the R1 ParM filament ultimately showed that it forms a lefthanded two-start helix, which is very different from the righthanded F-actin filament (24). More recent work on the DNA segregation protein AlfA has shown that although it forms two-start, left-handed filaments, the filaments adopt structures quite different from those of R1 ParM (25). In these highly twisted filaments, the protomer nucleotide binding pockets are rotated away from the filament axis and the ribbon-like helix is much more open than those of R1 ParM and F-actin. Thus, more work is needed to understand the types of filament structures adopted by actin-like proteins and how these filaments are optimized to carry out their specific cellular functions. In the case of type II partition, only the R1 ParM protein has been studied in any detail in terms of its structure and polymer dynamics and thus it is not clear if all such type II partition NTPases form similar filaments. Here we report studies on the type II partition ParM from the Staphylococcus aureus pSK41 plasmid partition system. pSK41 is representative of a clinically important family of self-transmissible S. aureus multiple resistance plasmids, which confers resistance to several antibiotics, disinfectants, and antiseptics (27).
The pSK41 par operon is similar to the well characterized R1 par system in that it encodes ParR and ParM proteins (28, 29). Our previous studies revealed how multiple pSK41 ParR proteins bind cooperatively to the parC centromere to form a partition complex and also showed that this complex recruits ParM via interactions with the ParR C-terminal domain (29). However, the role of the pSK41 ParM protein in mediating partition once it is recruited to the partition complex has not been characterized. To address these questions we determined the 1.95-Å resolution crystal structure of pSK41 ParM and compared it to other filament forming actin-like proteins. The structure reveals that pSK41 ParM belongs to the actin/Hsp70 superfamily, which includes its functional homologue, R1 ParM. However, the pSK41 ParM structure shows the strongest similarity to the archaeal T. acidophilum Ta0583 protein, which is thought to be involved in cell shape determination (21,22). Furthermore, the regions shown to be involved in R1 ParM filament formation are not structurally conserved in pSK41 ParM. Mutagenesis experiments revealed that indeed, ParM does not use the same residues in filament formation as R1 ParM, consistent with our electron microscopy (EM) 3 and total internal reflection fluorescence microscopy (TIRF) studies, which show that the filament structure and dynamics of the pSK41 ParM protein are very different from that of R1 ParM. These results point to unexpected divergence in the cytoskeletal structures employed by type II plasmid partitioning systems from evolutionarily and morphologically distinct bacterial hosts.

EXPERIMENTAL PROCEDURES pSK41 ParM Purification and Crystallization
The pSK41 ParM overexpression plasmid, pSK9019, was constructed by cloning a 1-kb ParM PCR product into the pTTQ18-RGS6H vector, which places a C-terminal RGS6H tag on the protein. To obtain soluble ParM protein, the cells containing the overexpression plasmid were grown at 37°C and induced for 6 h at 22°C with 0.5 mM isopropyl 1-thio-␤-Dgalactopyranoside. The protein was purified using nickel-nitrilotriacetic acid chromatography and buffer exchanged into 25 mM Tris, pH 7.5, 300 mM NaCl, 5% glycerol, and 1 mM ␤-mercaptoethanol. For crystallization, the protein was concentrated to 5 mg/ml. Crystals were obtained by hanging drop vapor diffusion at 298 K by mixing equal volumes of ParM and 1.5 M citrate, pH 5.6, which acts as a crystallization reagent. The crystals were monoclinic, P2 1 , a ϭ 53.4 Å, b ϭ 79.2 Å, c ϭ 54.6 Å, and ␤ ϭ 110.2°and contain one ParM monomer in the crystallographic asymmetric unit.

EM Studies
ParM Mutant Studies-ParM filament formation was assayed by using EM. Specifically, negative stain EM studies were carried out using stock solutions of 5-10 M ParM protein (either wild type or mutant), 1 mM AMP-PCP, and 5 mM MgCl 2 in polymerization buffer (25 mM Tris, pH 7.5, 150 mM NaCl, and 1 mM dithiothreitol). Samples at various dilutions were applied to glow discharged, nickel-coated, copper grids, blotted, negatively stained with 1% uranyl acetate, air dried, and examined at the M.D. Anderson High Resolution Electron Microscopy Facility on a Jeol 1010 microscope operating at 80 kV.
Structural Analysis of pSK41 ParM Filaments-Wild type samples in Par buffer (80 mM KCl, 2 mM MgCl 2 , 25 mM Hepes, pH 7.4, 1 mM dithiothreitol) were applied to carbon-coated copper grids, blot stained with 1% uranyl acetate, and visualized under a Jeol JEM-2010 HC microscope operated at 100 keV at a nominal magnification of 10,000 to 40,000. Films were digitized with Photo Scan 2000 (Z/I Imaging) at 7-m steps. Optical transforms and filtered images were calculated using the EM software package EOS (36). Otherwise, electron micrographs were scanned at 1200 disintegrations/min on an Epson scanner and displayed by Photoshop.

Total Internal Reflection Fluorescence Microscopy
TIRF experiments were conducted in a similar fashion as previously described (3,24). pSK41 ParM was diluted to the desired concentration, using Par buffer in the presence of 10 mM dithiothreitol and crowding agents (1% methylcellulose or 8% polyvinyl alcohol). ATP or GTP (2 mM final concentration) were added, directly mixed on the glass slide, covered with a glass slip, and sealed with nail polish. For labeling, ParM was dialyzed against appropriate buffers. Proteins were briefly centrifuged at high speed before use. Lysines were labeled with Alexa 488 carboxylic acid succinimidyl ester 6-isomer or 5-(and-6) carboxyl tetramethylrhodamine succinimidyl ester at a molar ratio of about 1:4 (dye molecule:pSK41 ParM molecule) overnight at 4°C as previously described (3,24). Labeled pSK41 ParM filaments were observed with an inverted Nikon microscope as previously described (3,24). Filament length was analyzed and error estimation determined using the programs Aquacosmos, ImageJ, and KaleidaGraph 3.6, as previously described (37).

Kinetics, P1 Release, and Critical Concentration Studies
pSK41 ParM polymerization was initiated by the addition of 2 mM nucleotide in Par buffer. The amount of pSK41 ParM polymer formed was monitored by 90 degree perpendicular light scattering using a Hitachi F-4500 fluorescence spectrophotometer at 300 nm and 24°C. The data were fitted using the program Dynafit (38). k 1 and k Ϫ1 describe concentration independent steps of the monomer activation when assembly is initiated by the addition of nucleotide, whereas k 2 and k Ϫ2 are related to the dimerization step. All subsequent rate constant pairs are equal to k e and k Ϫe and describe the steps of elongation. Without a dimerization step, the fit to the kinetic data were substantially worse.
The release of inorganic phosphate upon nucleotide hydrolysis during pSK41 ParM polymerization was measured at 24°C using a phosphate assay kit (E-6646) from Molecular Probes (39). To determine the critical concentration of ParM for polymerization, various protein concentrations were tested. Reactions were equilibrated at 24°C for 1 h. Ninety degree perpendicular light scattering experiments were carried out using a Hitachi F-4500 fluorescence spectrophotometer at 300 nm and 24°C.

RESULTS AND DISCUSSION
Overall Structure of Apo-pSK41 ParM-The 346-residue pSK41 ParM structure in its apo form was solved using MAD phasing techniques (Table 1 and Fig. 1A) (30,31). Specifically, MAD data were collected on a selenomethionine-substituted ParM crystal and used to obtain the phases for structure determination. Once the structure was solved, a high resolution (1.95-Å resolution) native data set was used to carry out the final refinement, which resulted in an R work /R free of 19.5/24.6%. The final model includes ParM residues 2-230, 239 -248, 262-343, and 294 water molecules. The structure shows that pSK41 ParM, which contains an ␣/␤-fold consisting of 13 ␤-strands and 10 ␣-helices, is a member of the actin/Hsp70 superfamily of ATP-binding proteins (Fig. 1B). pSK41 ParM, like other members of actin/Hsp70 superfamily, contains two domains, termed domains I and II, which are further subdivided into subdomains Ia, Ib and IIa, IIb (Fig. 1B). Characteristic of the actin/Hsp70 superfamily is the presence of two similar structural motifs in subdomains Ia and IIa, which are comprised of a 4 -5-stranded ␤-sheet surrounded by three ␣-helices. Interestingly, construction of this motif in subdomain Ia requires crossover of an ␣ helix, ␣10 in the pSK41 ParM structure, from domain II into domain I (Fig. 1B). It has been postulated that the actin-like fold evolved through gene dupli-cation of this ␣/␤ motif and the acquisition of additional domains modulates the protein function, as members of this superfamily play many diverse roles from filament formation, glycolysis, and chaperone activity to DNA segregation. Movement between the two domains is mediated by a hinge region, which in pSK41 ParM is composed of ␣4 and ␣10. The pSK41 ParM nucleotide-binding site is predicted to be located in a cleft between the two domains based on structures of other actin/ Hsp70 proteins in complex with nucleotides.
Structural Homology Searches Reveal Strong Homology with Archaeal Actin Ancestor-Besides pSK41 ParM, only one other structure of a ParM protein, that of R1 ParM, is available (6). Because both proteins function in plasmid DNA segregation, we anticipated that they would be structurally homologous. Interestingly, however, data base searches reveal that the pSK41 ParM structure shows the strongest structural homology to the archaeal T. acidophilum Ta0583 protein. pSK41 ParM shares 22% sequence identity with Ta0583, which although slightly higher than the sequence identity between the pSK41 and R1 ParM proteins (18%), is still in the "twilight zone" of protein sequence homology. However, the structural correspondence between the pSK41 ParM protein and Ta0583 is notable; the structures can be maximally superimposed (using 76 corresponding C␣ atoms) with an r.m.s. deviation of 2.4 Å, resulting in a Z-score of 7.7 ( Fig. 2A) (21). The Ta0583-ADP structure shows slightly lower structural homology and superimposes onto pSK41 ParM with an r.m.s. deviation of 2.6 Å for 71 corresponding C␣ atoms and a Z-score of 5.1. By contrast, the R1 ParM-GMP-PNP structure (6, 24) can be superimposed onto the pSK41 ParM structure with an r.m.s. deviation of 4.0 Å for 74 corresponding C␣ atoms, which provides a Z-score of only 3.1 ("Experimental Procedures") (Fig. 2B).
Examination of the superimpositions of pSK41 ParM with R1 ParM and Ta0583 reveals that region Ib is the most divergent between all the structures (Fig. 2, A and B). This region contains an extended loop in these proteins and in the pSK41 ParM structure it contains an extra, short helical segment, ␣1, which is not present in either Ta0583 or R1 ParM. The other major structural difference between pSK41 ParM and Ta0583 is in subdomain IIb. Residues 235 to 248 in this region in the Ta0583 structure forms a short antiparallel ␤-sheet, whereas in the pSK41 ParM structure the corresponding residues are disordered (Figs. 1B and 2A). Notably, in the R1 ParM structure, the region is helical. However, aside from these two areas in regions Ib and IIb, pSK41 ParM and Ta0583 show strong correspondence in their secondary structural elements. This is not the case in comparing the pSK41 and R1 ParM structures. In addi-  Rigid body rotations of domains I and II have been shown to accompany nucleotide binding in both R1 ParM and Ta0583. However, the magnitude of the structural changes is different. In R1 ParM, the domains rotate ϳ25°relative to each other upon nucleotide binding, whereas in Ta0583 the rotation is much smaller, ϳ8°(20 -21). This structural change facilitates the optimal positioning of residues in the pocket to bind nucleotide. Interestingly, the apo structure of pSK41 ParM shows stronger similarity to the nucleotide bound form of R1 ParM rather than its apo form. This does not seem to be caused by molecules in the crystallization solution mimicking nucleotides, as no such molecules are observed in the binding pocket (supplemental Fig. S1A). As discussed below, we find by EM that pSK41 ParM readily form filaments upon ATP-Mg 2ϩ addition (supplemental Fig. S1B). Thus, to address whether the crystallization solution stabilizes the closed, filamentous state in the absence of nucleotide, we carried out EM on pSK41 ParM in 1.5 M citrate, pH 5.6 (supplemental Fig. S1C). These studies clearly show that the crystallization solution does not force pSK41 ParM into the closed state. Indeed, the EM images are identical to those of the apoprotein.
Examination of the crystal packing shows that there are no crystal contacts to residues in or near the nucleotide binding pocket that might affect the protein conformation. The crystal contacts that are present are fairly equally distributed around the surface of the molecule and are made to both lobes (supplemental Fig. S2). It is possible that these contacts could somehow affect the closure state of pSK41 ParM, but it is not immediately obvious how. Indeed, our EM experiments support the contention that ParM adopts the apo state under these conditions, arguing against this idea. Thus, pSK41 ParM may be sim-ilar to Ta0583 in requiring little conformational change to bind nucleotides, but this remains to be determined.

R1 Filament Forming Residues Mapped onto pSK41 ParM: Implications for Polymer Formation-Re-
cently, a three-dimensional model for the R1 ParM filament was deduced using the R1 ParM structure for image reconstruction (24). An important outcome of this work was the determination of the residues involved in intra-strand filament contacts. Three primary regions were found and include residues 33-44, 107-114, and 237-242. In addition, two shorter segments, residues 161-164 and residues 297-299, were also involved in intra-strand interactions. The corresponding regions in the pSK41 ParM protein were 30 -50, 120 -125, and 256 -261 and the shorter segments correspond to residues 181-184 and 312-314.
Analysis of the R1 filament model reveals that the regions containing R1 ParM residues 33-44 and 107-114 provide the most numerous intra-strand contacts and thus, are crucial for filament formation. Interestingly, recent data shows that ParM filaments can contain a small percentage of "open" ParM molecules. Residues 33-44 were shown to be involved in filament formation in the open as well as closed forms of the polymer (40). Consistent with the central role played by these residues in filament formation, it was found that point mutation of residues 33, 34, 36, and 40 completely abrogate R1 ParM filament formation (14). Therefore, if pSK41 ParM forms polymers that are similar to the R1 ParM protein, then residues 30 -50 should also be essential for its polymerization. However, when the regions known to be critical for filament formation in R1 are examined in the pSK41 structure, we find that they appear structurally different, in particular pSK41 ParM residues 30 -50. As noted, this region forms an extended loop, which adopts very different conformations in the two proteins (Fig. 3,  A and B). Indeed, in pSK41 ParM this loop contains an insertion and there are more residues present than in R1 ParM. When a pSK41 ParM filament is modeled based on that of the R1 ParM filament by overlaying the pSK41 structures onto the R1 structures in the polymer, we find that pSK41 ParM residues 30 -50 clash with the next subunit within the strand (Fig. 3C). As was mentioned, this region in subdomain Ib also shows the most divergence between the pSK41 ParM and Ta0583 structures. The large differences in the filament forming regions of R1 ParM, pSK41 ParM, and Ta0583 suggest that these proteins may not form the same types of filamentous structures.
Mutagenesis Studies-Although the R1 ParM and pSK41 ParM structures show significant differences in the key filament forming region (comprised of residues 33-44 in R1 ParM), it was still possible that this region is important in pSK41 ParM polymerization. For example, structural changes  (24). To produce the model, pSK41 structures were maximally docked onto the R1 structures. This figure underscores the large structural differences in the regions involved in filament contacts between R1 and pSK41 ParM proteins. Especially note that the large loop in pSK41 ParM subdomain Ib would clash with its interacting neighbor in such a filament, suggesting that pSK41 ParM might not form filaments similar in structure to R1 ParM (indicated by asterisk). might occur during polymerization and/or nucleotide binding that could influence its structure and hence its ability to engage in polymer forming contacts. Indeed, it has been demonstrated that this region, region Ib, in actin adopts various conformations that are dependent on its nucleotide bound state (41)(42)(43).
Thus, we next directly tested the importance of this region in pSK41 ParM filamentation. Because it has been shown that point mutations of residues in this region (residues 33, 34, 36, and 40) ablate filamentation by R1 ParM, we created corresponding point mutations in the pSK41 ParM protein and assayed the affect of these substitutions on the ability of ParM to polymerize. Specifically, we made four amino acid substitutions, E37I, N38D, N47T, and K48S (Fig. 4A).
pSK41 single mutant proteins, pSK41 ParM(E37I), ParM(N38D), ParM(N47T), and ParM(K48S), were made first. Subsequently, a protein containing all four substitutions, ParM(E37I,N38D,N47T,K48S) was constructed. All the mutant ParM proteins were expressed and purified similarly to the wild type protein. The proteins were then tested for their ability to form filaments using negative staining, electron microscopy techniques. In the absence of ATP or GTP the wild type protein did not form polymers (Fig. 4B). However, addition of ATP, GTP, or ATP analogs to the wild type protein resulted in the clear formation of filaments as visualized by negative stain EM (Fig.  4C). The proteins with single mutations, E37I, N38D, N47T, and K48S all formed filaments indistinguishable from those formed by the wild type protein. Notably, the protein containing all four mutations also formed filaments nearly identical to those produced by the wild type protein (Fig. 4D).
Filament Structures-Our structural and biochemical data indicate that pSK41 ParM filaments are different from those formed by R1 ParM. To examine the pSK41 ParM filament structure in detail we carried out EM studies. These experiments showed that the pSK41 ParM protein assembled preferentially into well ordered bundles, which were typically several micrometers long, in the presence of ATP, GTP, or the non-hydrolysable nucleotide, AMP-PNP (Fig. 5, A-C). Such filaments were not observed upon addition of ADP and GDP (Fig. 5D). Like R1 ParM, pSK41 ParM does not form filaments in the absence of divalent cations (supplemental Fig. S3, A and B).
In addition to bundles, individual filaments (Fig. 4C) and paracrystalline nets were also observed (Fig. 6,  A-E). The filaments were found to assemble under a wide range of ionic strengths (supplemental Fig.  S3A). Optical diffraction patterns from bundles, in which individual filaments were arranged in parallel, showed a set of layer lines extending  to 25 Å, which are typical for a helical structure. For pSK41 ParM-ATP the observed layer lines were spaced at 270, 130, 75, 58, 47, and 40 Å and a meridional reflection arising from the subunit spacing at 26 Å was observed (Fig. 7A). For pSK41 ParM-GTP and pSK41 ParM-AMP-PNP the set of reflections observed was similar: 240, 120, 73, 56, 39, and 25 Å for pSK41 ParM-GTP and 250, 120, 74, 57, and 39 Å for pSK41 ParM-AMP-PNP (Fig. 7, B and C). The slight differences in the observed layer line spacings may indicate a nucleotide-dependent variability in helical twist as observed for F-actin (44). The helical parameters of pSK41 ParM were determined from the Fourier transforms. In general, the symmetry of a helical structure can be defined in terms of three parameters, the subunit repeat (h), the repeat of the helix (C), and the radius of the helix (r). If one is dealing with a helix with S subunits in T turns of the helix, S equals C/h and T corresponds to a strong off-meridional reflection observed in the "first helix" cross (44,45). One also expects to see reflections at T layer lines up and down from the meridional reflection in the "second helix" cross (44,45). From both pSK41 ParM-ATP and pSK41 ParM-GTP we observed strong meridional reflections at ϳ25 Å (Fig. 7, A  and C), arising from the subunit repeat, h. The first layer line at about 250 Å arises from helical repeat C (Fig. 7A). pSK41 ParM filaments can therefore be described as 10/4 helices (10 subunits in 4 turns of the helix) with a repeat of about 250 Å. This assignment is consistent with the strong reflections at ϳ57 Å (4th layer line), and on the 6th (10-4) layer line at ϳ39 Å in the first helix cross (Fig. 7B).
The radius of pSK41 ParM filaments was ϳ50 Å as measured from filtered images, where the bundle width was only a single layer thick (Fig. 6, D and E), consistent with the interfilament distance in bundles determined from the equatorial reflections of Fourier transforms (Fig. 7B). This radius appears smaller than the radii observed for F-actin or R1 ParM filaments, indicating that pSK41 ParM filaments may, in fact, be singlestranded (24,46). Indeed, filtered images of thin bundles clearly showed that pSK41 ParM formed single-stranded helices unlike the two-start double helical polymers formed by R1 ParM (24).
Filament Dynamic Properties-Next we investigated the polymerization kinetics of pSK41 ParM using light scattering. Assembly was substantially slower, ϳ10 times, than R1 ParM and ϳ3 times faster than F-actin under similar protein concentrations (24,47). Polymerization kinetics were nucleotide dependent and fastest for ATP, slightly slower for GTP, although AMP-PNP formed polymers substantially slower (Fig. 8, A and  B). At physiological protein concentrations of ϳ7 M, steady state was reached about 300 s after adding either ATP or GTP. Phosphate release lagged behind polymerization, as observed for R1 ParM and F-actin (Fig. 8A) (24,47). The critical concentration of polymerization was about 0.4 M and similar for both pSK41 ParM-ATP and pSK41 ParM-GTP (Fig. 8C). This value is comparable with the critical concentration of actin but lower than R1 ParM, which was ϳ2 M (16,47,48). Polymerization of F-actin required a nucleation step prior to elongation (49). In actin polymerization experiments, a lag phase prior to the intensity increase of light scattering was characteristic for the period of nucleation. In the case of pSK41 ParM no lag phase for pSK41 ParM-ATP or ParM-GTP was detected, whereas for pSK41 ParM-AMP-PNP, due to the slow kinetics, a considerable lag phase was observed (Fig. 8B). We therefore estimated if and how large the initial nucleus size may be from the slope of double logarithmic plots of the maximal rate of apparent elongation versus the protein concentration, as applied before for F-actin (49,50). The slopes of pSK41 ParM-ATP and pSK41 ParM-GTP were similar and the size of the nucleus was calculated to be 2-2.2 consistent with a dimer nucleus. By contrast, the fil-  ament nucleus of actin is a trimer (47,49,50). Assembly kinetic data could be fit using a minimal kinetic model similar as described for the linear protofilaments of Escherichia coli FtsZ (51) involving three steps: monomer activation, dimer formation, and elongation (supplemental Fig. S4). The rate constants are summarized in Table 2.
To investigate the dynamics of pSK41 ParM filaments at steady state, we analyzed fluorescently labeled pSK41 ParM filaments using TIRF microscopy (Fig. 9A). Labeled pSK41 ParM filaments were identical to those formed by unlabeled protein as ascertained from EM images and their Fourier transforms (supplemental Fig. S5, A and B). Interestingly, time lapse TIRF images revealed no significant dynamic instability for pSK41 ParM filaments (Fig. 9B). Indeed, pSK41 ParM filaments displayed a form of treadmilling more like F-actin rather than the marked dynamic instability exhibited by R1 ParM filaments.
ParM Filament Formation and Implications for Partition-Our structural, mutagenesis, and filament data demonstrate that pSK41 ParM forms filaments that are distinct from those formed by R1 ParM. Specifically, pSK41 ParM forms helical filaments that contain a single strand. This differs significantly from the twostranded helical filaments of F-actin (13/6) (3), R1 ParM (37/17) (46), and from the linear protofilaments of MreB (20) (Fig. 10). In addition, pSK41 ParM filaments are more stable than R1 ParM filaments, which display markedly dynamic instability (17,24). These large distinctions between pSK41 and R1 filament struc-  tures and dynamics have important ramifications for the multistep partition processes mediated by these proteins. The first step of partition is mediated by the centromerebinding protein, ParR. This protein binds multiple tandem repeats that comprise the DNA centromere site, giving rise to a superhelical partition complex. Our recent pSK41 ParR-centromere structure showed that the partition complex has a large diameter of ϳ180 Å (29). We found that the highly acidic C-terminal domains of ParR, which are flexible, fill the central pore of the superhelix, which ranges from 50 to 90 Å. The structure of pB171 ParR, which also revealed that, at high protein concentrations the ParR protein alone forms a superhelix similar to that of pSK41 ParR-centromere complex (52), suggests that all type II ParR proteins form similar partition complex superstructures, which function to capture the requisite ParM filament. Indeed, the next step in partition is recruitment of the ParM filament to the paired partition complexes of two plasmids. In the case of R1, data indicate that the filament binds the center of the pore of each partition complex (14). ParR binding to ParM activates its NTPase activity. The next step is NTP hydrolysis of bound nucleotide by ParM. Once ParM hydrolyzes the NTP, it no longer binds ParR with high affinity. As a result the filament disengages briefly from the partition complex. The presence of a two-start helix in R1 ParM allows the second strand to remain bound, whereas the first has been disengaged allowing time for insertion of a new ParM-NTP. This continued binding, NTP hydrolysis, release, and ParM-NTP binding at the interface ultimately leads to growth of the filament between the bound partition complexes and hence separation of the plasmids bound at each end.
We have shown that several differences exist between pSK41 and R1 ParM filaments. Instead of a two-stranded ParM helix, the pSK41 ParM helix consists of a single strand, composed of ParM molecules ϳ50 Å in diameter. This single-stranded helix must either bind to the partition complex pore as does the R1 filament or at the edges of the filaments that are undergoing treadmilling motion. The latter could ultimately move plasmids to distinct locations in the cell. However, the former model, binding at the center of the pore, is also plausible as, unlike the R1 ParM filaments, the pSK41 filaments appear more stable. Hence, after NTP hydrolysis, even if the pSK41 ParM filament is momentarily released from its contact from the partition complex, the polymer remains stable long enough to permit insertion of a fresh ParM-NTP subunit at the growing interface.
Examination of the electrostatic potential of ParM reveals that its outside surface faces or "sides" are quite positively charged (Fig. 11). This is interesting in light of the data showing that pSK41 ParM forms a single-stranded filament because it  suggests that the sides of the filament would contain a positive charge, whereas the region of ParR known to bind and recruit ParM is highly negatively charged (Fig. 11). Hence, charge complementarity between the surfaces of ParR and ParM could play some role in binding between the ParM filament to the center of the ParR-centromere pore. Specifically, each end of a ParM filament could engage or interact with a partition complex. Alternatively, the partition complexes could bind on the outside surface of filament bundles undergoing treadmilling.
In conclusion, our combined data show that despite their commonality of ancestry and function, the pSK41 and R1 type II plasmid segregation actin-like proteins have diverged to adopt distinct filament properties in disparate bacterial hosts. Existing models of type II segregation rely on filament assembly along the long axis of a rod-shaped cell as a basis for effective plasmid distribution into future daughter cells (53). However, such a simple explanation seems inadequate to account for efficient plasmid partition in spherical S. aureus cells. The extent to which such a fundamental difference in bacterial host morphology may have contributed to the evolution of the observed filament diversity is an open question at this time. Nonetheless, our study highlights and further strengthens the hypothesis that the only common features between all actin-like polymers are their ability to form lateral protofilament contacts (54).