Single-molecule Analysis of Protein·DNA Complexes Formed during Partition of Newly Replicated Plasmid Molecules in Streptococcus pyogenes*

The Streptococcus pyogenes pSM19035 partition locus is ubiquitous among plasmids from vancomycin- or methicillin-resistant bacteria. An increasing understanding of this segregation system may highlight novel protein targets that could be blocked to curb bacterial proliferation. pSM19035 segregation depends on two homodimeric (δ2 (ParA) and ω2 (ParB)) proteins and six cis-acting centromeric noncurved parS sites. In the presence of ATP·Mg2+, δ2 (δ·ATP·Mg2+)2 binds DNA in a sequence-independent manner. Protein ω2 binds with high affinity and cooperatively to B-form parS DNA. Atomic force microscopy experiments indicate that about 10 ω2 molecules bind parS, consisting of 10 contiguous iterons. Protein (δ·ATP·Mg2+)2, by interacting with the N terminus of ω2 bound to parS, loses its association with DNA and relocalizes with ω2·parS to form a ternary complex ((δ·ATP·Mg2+)2·ω2·parS) with the DNA remaining in straight B-form. Then, the interaction of two (δ·ATP·Mg2+)2·ω2·parS complexes via δ2 promotes pairing of a plasmid subfraction. (δD60A·ATP·Mg2+)2, which binds but does not hydrolyze ATP, leads to accumulation of pairing intermediates, suggesting that ATP hydrolysis induces plasmid separation. We propose that the molar ω2:δ2 ratio regulates the different stages of pSM19035 segregation, pairing, and δ2 polymerization, before cell division.

Accurate segregation of newly replicated plasmids or bacterial sister chromosomes is achieved by evolutionarily distinct stabilization systems (for review, see [1][2][3]. The mechanisms of stable maintenance of low copy number plasmids can be divided into those that function (i) by enhancing the resolution of oligomers into monomers, with the subsequent increase of the number of molecules to be segregated; (ii) by blocking proliferation of plasmid-free segregants; or (iii) by partitioning plasmid copies to daughter cells at cell division (4 -6). The majority of the plasmid partition systems include two trans-acting homodimeric proteins (one binds a nucleotide cofactor and the other the centromere) and one or more cis-acting centromere-like sequences (for review, see 2, 3,5). There are three types of nucleotide-binding proteins: type I or ParA ATPases, type II or ParM (actin-like) ATPases, and type III or TubZ (tubulin-like) GTPases. The mechanisms that underlie the accurate partitioning of plasmid DNA by the type I and type III motor proteins are still largely unknown (2,3,5). Centromere binding is performed by one of three well studied protein types. The large type Ia proteins recognize the cognate sequence via a helix-turn-helix domain and spread along the DNA up to several kilobases (7)(8)(9). The small type Ib and type II proteins recognize their cognate site via an antiparallel ␤-sheet of the ribbonhelix-helix fold and do not seem to spread from a cognate site into a nonspecific DNA sequence. The type Ib proteins (e.g. pSM19035-2 ) have an unstructured N-terminal domain required for activation of polymerization of ParA pSM19035-␦ 2 onto parS DNA (10), whereas the type II proteins have an extended C-terminal region involved in interaction with ParM (e.g. pSK41-ParR) (6). The presence of a centromere-binding protein and a centromere sequence in the type III partition system remains undetermined.
To gain insight into the general mechanism that manages accurate ParAB-dependent partitioning of type Ib plasmids at cell division and to compare them with type II plasmids in bacteria of the Firmicutes phylum, the early stages of partitioning of the Streptococcus pyogenes plasmid pSM19035 were dissected in vitro. pSM19035, which has extraordinarily long inverted repeat sequences that comprise about 80% of the genome, has 2 Ϯ 1 copies/cell and is stably maintained (11)(12)(13) (Fig. 1A). The par loci of pSM19035 encode two trans-acting proteins, found as homodimers in solution (␦ 2 (ParA type I) and 2 (ParB type Ib)), and six cis-acting 2 target sites (parS centromeres) (14,15) (Fig. 1A). The parS centromeres overlap with the upstream region of the promoter (P) of ␦ (P ␦ or parS1), (P or parS2) and copS (P cop or parS3) genes and are present twice in the pSM19035 genome (14) (Fig. 1A). Each parS centromere (parS1, parS1Ј, parS2, parS2Ј, parS3, and parS3Ј) consists of 9, 7, and 10 contiguous iterons, respectively, in direct ‹ or Š inverse orientation with sequence 5Ј-WATCACW-3Ј. The centromeric-binding protein 2 binds one iteron or a non-parS control site with very low affinity (k D ϳ 1000 nM); however, it binds two iterons (k D ϳ 20 Ϯ 2 nM) or full-length parS sites (k D ϳ 5 Ϯ 1 nM) with high affinity and cooperatively. By . Genome organization of plasmid pSM19035 and proposed structure of the 2 ⅐parS complex. A, pSM19035 duplicated sequences, which comprise ϳ80% of the molecule, indicated by a thick line and unique nonrepeated sequences (NR1 and NR2) by a thin line. The replication origins (yellow boxes), direction of replication (denoted by arrows), the six resolution sites (orange boxes), and the parS sites (red boxes) are indicated. The outer thin arrows indicate the organization of the genes. The plasmid is divided in regions that direct (repS) and control (copS) replication (Rep region). Plasmid resolution (␣, ␤, and ␥ genes) (SegA) and postsegregation growth inhibition (⑀ and ; brown arrows) and segregation (␦ and ; red arrows) (SegB) are indicated. The resistance to erythromycin (erm1 and erm2) and the set of poorly characterized genes in blue and purple arrows are also indicated. The upstream region of the promoters of the copS, ␦, and genes constitute the six cis-acting centromere-like parS or parSЈ sites. A parS site consists of a variable number of contiguous 7-bp heptad repeats (symbolized by ‹, 5Ј-WATCACW-3Ј, where W is A or T). The number of repeats and their relative orientations (direct ‹ or invert Š repeats) are enlarged. B, structural model of 2 bound to parS1 DNA. The overall structure of 2 forming a left-handed matrix around straight parS1 DNA based on the crystal structure determined for (⌬19) 2 ⅐‹ ‹ DNA and (⌬19) 2 ⅐‹ Š complexes (10, 17) is shown. pSM19035 Partition Complex OCTOBER 30, 2009 • VOLUME 284 • NUMBER 44 binding to the centromeric sequences, 2 acts as a specific transcription factor, required for the correction of downward fluctuations of plasmid copy number control; the expression of the ⑀ and proteins needed for blocking proliferation of plasmidfree segregants (16) (see above); the accurate expression of ␦ 2 and 2 and ensuring plasmid partitioning (14,15,17) (Fig. 1A). Protein 2 forms a nucleoprotein complex at the centromere sites without apparent distortion (14) (Fig. 1B). In the presence of ATP⅐Mg 2ϩ , ␦ 2 (␦⅐ATP⅐Mg 2ϩ ) 2 is an ATPase with sequenceindependent DNA binding activity (k D of ϳ170 Ϯ 20 nM) (10). Protein (␦⅐ATP⅐Mg 2ϩ ) 2 does not appear to participate in regulating the par locus, leaving this function entirely to 2 (14).
The first 19 N-terminal residues of 2 ( 2 ⌬N19) are dispensable, both in vivo and in vitro, for binding to parS DNA (18). The structures of 2 ⌬N19 alone or in complex with two iterons in direct (‹ ‹) or inverted (‹ Š) orientations have been determined (17,19). Extrapolating from these structures, we predict that 2 molecules assemble as a left-handed protein helix to wrap around the parS sites on B-form DNA (see Fig. 1B). To test the crystal structure model (Fig. 1B) and to gain insight into the architectural requirement, AFM 3 studies were performed. Here, we report the visualization of nucleoprotein complexes formed by 2 on linear or circular parS DNA and its dynamic interaction with (␦⅐ATP⅐Mg 2ϩ ) 2 to form the partition complex. The (␦⅐ATP⅐Mg 2ϩ ) 2 -(␦⅐ATP⅐Mg 2ϩ ) 2 interaction from two straight partition complexes mediates site-specific pairing and formation of large superstructures on DNA. The topographies of those ParAB complexes were compared with the architecture of the ParRM partition system.

EXPERIMENTAL PROCEDURES
Proteins and DNA-Proteins 2 , 2 ⌬N19, 2 T29A, ␦ 2 , ␦ 2 K36A, and ␦ 2 D60A and plasmid-borne parS3 DNA (pUC57) were purified as described (10,14,18). Only experiments with parS3 DNA, herein parS DNA, are described because similar results were obtained with the three parS sites. Wild-type or mutant variants of 2 , or ␦ 2 were incubated with linear (EcoRIcleaved) or supercoiled parS DNA (25 nM in DNA molecules) for 20 min at 30°C in buffer A (20 mM HEPES, pH 7.5, 5 mM MgCl 2 ,50 mM KCl) containing 1 mM ATP. When indicated, ␦ 2 variants were incubated with preformed 2 (variants)⅐parS DNA complexes. The solution was diluted in buffer A and spotted onto freshly cleaved mica treated with 10 mM spermidine for 20 min at room temperature. The mica was washed several times with Milli-Q water and dried with nitrogen.
AFM Measurements and Image Analysis-AFM was performed using a Nanoscope IIIa or IV microscope (Digital Instruments) in air using the tapping mode. The cantilever (OMCL-AC160TS-W2; Olympus) was 160 mm in length with a spring constant of 33-62 N/m. The scanning frequency was 1-2 Hz, and images were captured with the height mode in a 512 ϫ 512-pixel format. The images obtained were plane-fitted and flattened by the computer program accompanying the imaging module. In the geometrical analyses of AFM images of DNA and protein, the "tip effect" was removed by using the apparent size of DNA as a reference (20,21). Briefly, the appar-ent dimensions of the molecules obtained by AFM depend on the radius of the tip curvature and are apparently larger than the real dimensions. The relationship among the width of the globular molecule in the AFM image (W), the radius of the tip curvature (R c ), and the radius of the molecule (R m ) is given by W ϭ 4(R c R m ) 1/2 . When two different molecules with R m1 and R m2 radii are imaged with the same tip, the relationship between the measured widths (W 1 and W 2 ) can be given by W 1 ϭ W 2 (R m1 /R m2 ) 1/2 . Because the diameter of a DNA molecule is known (2 nm), we used the established parameter in our AFM image calculations as W 1 . We could then calculate the radius of the particles (R m2 ) from the apparent width of the DNA (W 2 ). The image processing and analysis was carried out using Image SXM (22).
The length of the complexes was measured at half-maximal height to compensate for distortions introduced by the tip geometry (see 23). The theoretical volume of the proteins was calculated assuming that they were globular (24), and the volume measurements were done as described previously (25)(26)(27).

RESULTS
2 Forms a Nucleoprotein Complex without Apparent Distortion of parS DNA-To study the type of nucleoprotein complexes formed by wild-type 2 with full-length parS site on linear or supercoiled DNA, AFM experiments were performed. The 70-bp parS DNA consists of 10 contiguous iterons (heptads) organized in three blocks of two 7-bp direct repeats and one 7-bp inverted repeat, plus one direct repeat (‹ ‹ Š ‹ ‹ Š ‹ ‹ Š ‹) (14) (Fig. 1A). The measured length of the linear DNA agreed with the expected value (ϳ966 nm) for a 2927-bp DNA with a 0.33-nm rise/bp. The parS site was located 120 bp from one end. No end showed any apparent curvature.
Increasing concentrations of 2 incubated with a fixed concentration of linear parS (1-20 2 molecules/DNA molecule) gave rise to an increased number of similar protein⅐DNA complexes, with a concomitant decrease in the number of proteinfree DNA molecules (data not shown). The optimal ratio for limiting 2 concentrations and lower amount of protein-free DNA molecules was observed in the presence of ϳ7 2 molecules/parS site (7.2 2 /DNA molecule) ( Fig. 2A). A visual inspection of the images revealed that DNA molecules with bound 2 contained only one discrete straight complex at the parS site (Fig. 2, A and B). The distance between the 2 ⅐parS complex and the nearest end had a mean value of 31.2 Ϯ 5.0 nm (Fig. 2C). This was consistent with a length of 120 bp (theoretical value of 39.6 nm) if a minor 2 spread (Ͻ15 bp toward the 5Ј end of the repeats) was taken into consideration (see 15). Similar results were observed at ϳ20 2 /DNA molecule (data not shown).
The interaction of 2 with supercoiled parS DNA was examined (Fig. 2D). Similar results were observed when 2 was bound to supercoiled DNA or when the DNA was nicked prior to examination; the latter results, which simplified the interpretation, are shown. In the presence of 7.2 2 /DNA molecule, one nucleoprotein complex/circular DNA molecule could be clearly identified (Fig. 2D). Inspection of the images revealed that the DNA molecule with bound protein contained one discrete 2 ⅐parS complex without apparent distortion of parS (Fig. 2D).
Structure of the 2 ⅐parS Complex-The size of the 2 ⅐parS complex was analyzed by measuring the length and width of 2 bound to linear or circular DNA (n ϭ 70 and 100, respectively). The mean height of the DNA was 0.23 nm (data not shown), which deviated from the theoretical height for double-stranded DNA (2 nm), confirming that DNA is usually seen smaller than the sample itself in AFM images (25)(26)(27). The apparent length (26.3 Ϯ 3.3 nm on linear and 25.8 Ϯ 3.4 nm on circular DNA) of the complexes observed in AFM images is consistent with the length of the parS centromere (70 bp) covered by 2 (Fig. 3, A  and B). To calculate the width of the complexes, the width of the DNA was plotted against the width of the protein⅐DNA complexes in the same image. On average, the analyzed complexes were 4.05 Ϯ 0.83 nm and 4.89 Ϯ 1.18 nm wide for linear and circular DNA, respectively (n ϭ 39). The structure of 2 bound to mini-parS DNA shows that the protein wraps around the DNA (17), forming a cylindrical structure with similar height and width. Therefore, the volume was also calculated assuming a cylindrical shape of the complex and using the width value obtained as described above. The volume of 2 ⅐parS DNA complexes calculated in this way was 180 -320 nm 3 (n ϭ 100), consistent to the expected (ϳ235 nm 3 assuming that the theoretical volume of 2 was 19.8 nm 3 plus the DNA contribution).
To validate the results of volume analysis further with a second methodology, Image SXM 169 software was used. The average height and area of a manually defined nucleoprotein complex and the adjacent region containing only DNA were subtracted to determine the volume of the complex. The volume obtained by this method was 212 Ϯ 42 nm 3 for the 2 ⅐parS complex on linear and 235 Ϯ 56 nm 3 on circular parS DNA (n ϭ 36). Hence, it is likely that 2 binds to full parS DNA with a stoichiometry of ϳ1 Ϯ 0.2 2 /iteron. This is consistent with stoichiometry experiments showing (i) minor 2 spreading (Ͻ15 bp) toward the 5Ј end of the contiguous iterons (see 15); (ii) 2.2, 3.1, and 4.5 2 molecules binding DNA segments containing two, three, and four heptads, respectively (15); and (iii) the co-crystal structure of the 2 ⌬N19⅐mini-parS complexes (17). 2 Bound to parS Promotes ␦ 2 Recruitment and Plasmid Pairing-At a ratio of 100 2 /DNA molecule, 2 binds to parS regions on linear DNA and mediates site-specific pairing with very low efficiency (ϳ1% of the total molecules analyzed by electron microscopy in the presence of glutaraldehyde fixation) (10). At 7.2 2 /parS DNA molecule, 2 bound to linear or circular parS DNA but failed to promote any site-specific plasmid pairing (Fig. 2).
Protein (␦⅐ATP⅐Mg 2ϩ ) 2 (at 7 ␦ 2 /DNA molecule) bound DNA in a sequence-independent manner (supplemental Fig. S1), but such protein⅐DNA complexes were not observed when ATP was omitted (data not shown). We failed to detect any sitespecific intermolecular pairing (supplemental Fig. S1).
At 7.2 2 /DNA molecule, 2 promoted relocalization of (␦⅐ATP⅐Mg 2ϩ ) 2 from any position on the supercoiled DNA toward the 2 ⅐parS complex in more than 90% of the molecules with subsequent increase in the protein⅐DNA volume. A ternary complex ((␦⅐ATP⅐Mg 2ϩ ) 2 ⅐ 2 ⅐parS) was observed in ϳ67% of the molecules with site-specific intermolecular pairing in the remaining fraction (Fig. 4A). When linear parS DNA was used, 2 bound to parS promoted redistribution of (␦⅐ATP⅐Mg 2ϩ ) 2 to relocalize with 2 ⅐parS to form the ternary complex. Similar results were observed when limiting protein concentrations were used (data not shown).
In the presence of subsaturating amounts of 2 and (␦⅐ATP⅐Mg 2ϩ ) 2 at 1:1 molar ratio, site-specific intermolecular pairing in a subfraction (ϳ22%, n ϭ 160) of the molecules was observed (Fig. 4A). Similar results were reported previously in the presence of an ϳ14-fold excess of both 2 and (␦⅐ATP⅐Mg 2ϩ ) 2 (at 1:1 ratios) proteins per linear DNA molecule when analyzed by electron microscopy in the presence of glutaraldehyde fixation (10).
The parS DNA region on the pairing complex was not significantly distorted by 2 binding or (␦⅐ATP⅐Mg 2ϩ ) 2 interaction with 2 (Fig. 4B). We could hypothesize that 2 -mediated centromere pairing but the pairing complex is unstable, and (␦⅐ATP⅐Mg 2ϩ ) 2 was required to "activate" 2 to promote stable plasmid pairing. Because the presence of higher (␦⅐ATP⅐Mg 2ϩ ) 2 concentrations (14 or 28 ␦ 2 /DNA molecule) under limiting 2 concentrations led to the formation of high order superstructures, with three or more 2 ⅐parS complexes paired by (␦⅐ATP⅐Mg 2ϩ ) 2 (Fig. 4A), we do not favor this hypothesis. Indeed, at 2 :(␦⅐ATP⅐Mg 2ϩ ) 2 ratios of 1:3.9, ϳ33% of 162 analyzed molecules had three or more plasmid paired molecules (Fig. 4A).
The (␦D60A⅐ATP⅐Mg 2ϩ ) 2 variant, which binds but does not hydrolyze ATP, at 7 (␦D60A⅐ATP⅐Mg 2ϩ ) 2 /DNA molecule, binds DNA in a sequence-independent manner (10). At a 1:1 2 :(␦D60A⅐ATP⅐Mg 2ϩ ) 2 molar ratio, 2 recruited (␦D60A⅐ATP⅐ Mg 2ϩ ) 2 toward the 2 ⅐parS complex in ϳ90% of molecules (Fig.  5), suggesting that ATP hydrolysis is not required for (␦D60A⅐ATP⅐Mg 2ϩ ) 2 redistribution toward the 2 ⅐parS complex. A ternary complex ((␦D60A⅐ATP⅐Mg 2ϩ ) 2 ⅐ 2 ⅐parS) was observed in ϳ64% of the molecules with site-specific intermolecular pairing in the remaining fraction (Fig. 5A). Site-specific intermolecular pairing of (␦D60A⅐ATP⅐Mg 2ϩ ) 2 bound to 2 ⅐parS with a second 2 ⅐parS was reduced ϳ3-fold compared with wild-type (␦⅐ATP⅐Mg 2ϩ ) 2 . Here, ϳ26% of the analyzed molecules had three or more plasmid molecules paired (n ϭ 157). At 1:2 2 :(␦D60A⅐ATP⅐Mg 2ϩ ) 2 molar ratio, the accumulation of ternary partition complexes ((␦D60A⅐ ATP⅐Mg 2ϩ ) 2 ⅐ 2 ⅐parS) was reduced (ϳ39% of the molecules), and pairing of three or more plasmids accounted for ϳ62% of the analyzed particles (n ϭ 169) (Fig. 5A). The data suggest that (␦D60A⅐ATP⅐Mg 2ϩ ) 2 was redistributed toward the 2 ⅐parS complex and the dynamic assembly of plasmid pairing requires ATP binding, but, in the absence of its hydrolysis, disassembly of paired molecules is slowed down or abolished. On the other hand, when the (␦K36A) 2 mutant, which does not bind ATP or DNA, was used, neither ternary complex formation nor plasmid pairing was observed in the presence of nucleotide cofactor and metal ion (data not shown).
The 2 ⌬N19⅐parS complexes were indistinguishable from the 2 ⅐parS complexes. However, in the presence of both 2 ⌬N19 and (␦⅐ATP⅐Mg 2ϩ ) 2 , (␦⅐ATP⅐Mg 2ϩ ) 2 relocalization and site-specific pairing of supercoiled DNA (Ͻ1% of the analyzed particles, n ϭ 138) were not observed (supplemental Fig.  S2A). This is consistent with data showing that 2 ⌬N19, which lacks the first 19 residues, binds to parS DNA and represses promoter utilization with affinity similar to wild type (18) but fails to interact with (␦⅐ATP⅐Mg 2ϩ ) 2 and to promote pairing of linear DNA by electron microscopic examination upon glutaraldehyde fixation at saturating concentrations of the protein (100 2 ⌬N19/DNA molecule) (10). The 2 T29A mutant binds DNA with low affinity and without sequence specificity (18). In the presence of a 10-fold excess of 2 T29A/DNA molecule, 2 T29A formed discrete com-plexes on DNA. At 2.5:1 2 T29A:␦ 2 molar ratios, 2 T29A promoted relocalization of (␦⅐ATP⅐Mg 2ϩ ) 2 from any position on the DNA toward the 2 T29A⅐DNA complex in ϳ75% of the molecules, and pairing of multiple plasmid in only ϳ8% (n ϭ 136) of the analyzed particles was observed (supplemental Fig. S2B). Protein (␦⅐ATP⅐Mg 2ϩ ) 2 seems to interact with all 2 T29A⅐DNA complexes, forming large protein⅐DNA complexes.
Relative Stoichiometry of the Partition Complex-To study the apparent stoichiometry of all three components of the partition complex, we compared the volumes of 2 ⅐parS complex with the ternary (␦⅐ATP⅐Mg 2ϩ ) 2 ⅐ 2 ⅐parS partition complex and the paired plasmid complexes. Because of the heterogeneous shape of the paired complex (Fig. 4B), Image SXM 169 software was used to measure the volumes and to compare them with the volumes obtained for the 2 ⅐parS complex on circular DNA (see above).
Based on the theoretical volume of (␦⅐ATP␥S⅐Mg 2ϩ ) 2 (ϳ85 nm 3 ) and considering that the stoichiometry of 2 to parS remains unchanged by varying 2 concentration (between 1 and 20 2 molecules/parS DNA), we can estimate the apparent number of (␦⅐ATP⅐Mg 2ϩ ) 2 molecules in the ternary and pairing complexes to be ϳ4 in the partition complex. In the paired complex, the results show great variability, depending on the number of paired molecules. The number of (␦⅐ATP⅐Mg 2ϩ ) 2 molecules in two paired DNA plasmids was ϳ20, whereas in higher order superstructures with five or more plasmid molecules paired, the mean value was ϳ80.

Protein 2 Forms a Nucleoprotein Complex on parS DNA-
During pSM19035 proliferation, DNA synthesis and genome segregation are coordinated by 2 to ensure its stable inheritance (see the Introduction). The overall structure of the 2 ⅐parS complex, in linear or supercoiled DNA, revealed the formation of a discrete structure without significant spreading or compaction, shortening, or distortion of the DNA. This is consistent with (i) cooperative binding of 2 to 7-10 contiguous iterons (parS sites) with a marginal spreading onto non-parS DNA, as shown in DNase I protection studies where 2 spreads Ͻ15 bp, if at all; (ii) DNA titration of increasing numbers of heptads, indicating that 2 binds with a 2 :heptad stoichiometry of 1:1 (15); (iii) each iteron recruiting one 2 molecule, implying that the fully bound parS site is cooperatively coated by ϳ10 2 molecules (15,17); and (iv) the homotetrameric structure of 2 binding to mini-parS, which forms a left-handed wrap around ideal B-form DNA (17). The topography of 2 (ParB Ib) bound to parS differs from P1-ParB or F-SopB (ParB Ia) bound to parS or sopC. Both P1-ParB and F-SopB bind and spread up to several kilobases of DNA in a centromere-dependent manner. In the case of P1, the DNA might wrap around a multimeric protein core (28), whereas F-SopB wraps DNA in a righthanded manner (29).
The mechanisms that have evolved to segregate genomes of Firmicutes differ between plasmids that use the ParAB (e.g. pSM19035-ParAB) and the ParMR (pSK41-ParRM) partition systems. The centromeres of both systems (parS and parC) are in B-form DNA, and both proteins, 2 and ParR, bind mini-parS and mini-parC, respectively, as homooligomers and form superhelical structures (17,30). However, the architectures of these complexes differ markedly. Protein 2 binds to parS and assembles as a left-handed matrix with DNA binding sites facing inward (Fig. 1B), and homooligomeric 2 bound to parS fails to promote site-specific pairing of supercoiled DNA molecules (Fig. 2). Each 2 molecule is displaced relative to its neighbor by 7 bp and left-handed rotated by 252°around the straight DNA helix (17). In contrast, ParR assembles into a right-handed solenoid structure with the DNA binding sites facing outward. This assembly induces a pre-bent architecture on noncurved parC DNA (30,31). Then parC wraps the ParR scaffold, and this spiral structure mediates pairing of DNA molecules (segrosome structure) in the absence of ParM (30 -32).
(␦⅐ATP⅐Mg 2ϩ ) 2 Promotes the Formation of Ternary and Pairing Complexes-pSM19035 pairing requires the N terminus of 2 and its ability to bind parS DNA, the binding of ␦ 2 to ATP⅐ Mg 2ϩ , and the interaction between 2 and (␦⅐ATP⅐ Mg 2ϩ ) 2 . First, 2 binds parS to form a 2 ⅐parS complex on  Fig. 4. B, two representative images showing two and three plasmids paired.
straight DNA, and (␦⅐ATP⅐Mg 2ϩ ) 2 interacts with DNA in a sequence-independent manner or might interact with 2 . However, such protein-protein interaction could be "unstable" because we failed to detect the accumulation of such intermediates. Second, (␦⅐ATP⅐Mg 2ϩ ) 2 bound to DNA interacts with the unstructured N-terminal domain of 2 bound to parS DNA. Indeed, 2 ⌬N19 bound to parS DNA failed to interact with (␦⅐ATP⅐Mg 2ϩ ) 2 . Third, the interaction of 2 bound to parS with (␦⅐ATP⅐Mg 2ϩ ) 2 promotes its relocalization from any DNA site to the 2 ⅐parS complex. This is consistent with data showing that addition of 2 to preformed (␦⅐ATP⅐Mg 2ϩ ) 2 ⅐DNA led only to the formation of discrete ternary complexes (parS⅐ 2 (␦⅐ATP⅐ Mg 2ϩ ) 2 ) in the majority of cases (ϳ67% of total molecules) (see Fig. 4A). Fourth, (␦⅐ATP⅐Mg 2ϩ ) 2 mediates site-specific pairing, quaternary parS⅐ 2 (␦⅐ATP⅐Mg 2ϩ ) 2 ⅐ 2 ⅐parS complexes, in ϳ22% of the cases without apparent distortions in the DNA. However, when (␦⅐ATP␥S⅐Mg 2ϩ ) 2 was used, the amount of quaternary complexes markedly increased to levels comparable with (parS⅐ 2 ⅐(␦D60A⅐ATP⅐Mg 2ϩ ) 2 ⅐ 2 ⅐parS complexes (see below). Fifth, the formation of quaternary parS⅐ 2 ⅐(␦⅐ATP⅐ Mg 2ϩ ) 2 ⅐ 2 ⅐parS complexes between the six different parS sites should condense the plasmid molecule. Like protein ␦ 2 (ParAlike) in vitro, P1-ParA facilitates pairing (blocked supercoil diffusion) of P1-ParB bound to a parS site in two plasmid monomers in vivo (33). Finally, (␦D60A⅐ATP⅐Mg 2ϩ ) 2 was sufficient to induce pairing, although dislodging the paired molecules might require ATP hydrolysis (compare Figs. 4 and 5). The topography of the ternary complex differs markedly from the ParM⅐ParR⅐parC complex, where the oligomeric right-handed ParR ring, with parC wrapped around, acts through its C-terminal domain as an anchor on the ends of the elongating ParM filaments (31, 32, 34 -37). Mechanistic Implications of the Pairing Complex-The differences in the structures of 2 ⅐parS (type Ib) and ParR⅐parC (type II) complexes might indicate how they interact with their partners (␦ 2 and ParM) and how they segregate plasmid molecules. Previous cytological, biochemical, and structural studies together with this work provide detailed sequential information about the mechanism of ParAB active plasmid partitioning (10,15,17). (␦⅐ATP⅐Mg 2ϩ ) 2 , which shows a slow rate of ATP hydrolysis, binds DNA and interacts with 2 bound to parS; however, neither ␦ 2 K36A, which does not bind ATP, nor (␦⅐ADP⅐Mg 2ϩ ) 2 interacts with DNA (10). parS-bound protein 2 interacts with (␦⅐ATP⅐Mg 2ϩ ) 2 bound to DNA and recruits it to the parS⅐ 2 complex. Limiting concentrations of (␦⅐ATP⅐Mg 2ϩ ) 2 promotes pairing of two supercoiled plasmid molecules in a subfraction of the molecules (Fig. 4). Once the plasmids are paired, the local intracellular concentration of 2 increases, the ATPase activity of ␦ 2 increases, and plasmid pairing is lost. On the other hand, when (␦D60A⅐ATP⅐Mg 2ϩ ) 2 was used, multiple pairing molecules were observed, suggesting that in the absence of ATP hydrolysis, paired intermediates accumulate (Fig. 5). In the nanomolar range, both proteins are needed for plasmid pairing (Fig. 4), but in the millimolar range, both proteins are needed for (␦⅐ATP⅐Mg 2ϩ ) 2 polymerization onto DNA (10). The mechanic role of ␦ 2 in promoting plasmid segregation by polymerization and depolymerization from the DNA, prior to cell division segregation, is poorly understood. After Leonard and co-workers (2), we proposed that 2 bound to parS assists treadmilling of ␦ 2 . The processive disassembly of ␦ 2 filaments could contract the (␦⅐ATP⅐Mg 2ϩ ) 2 spiral-like structure, observed both in vivo and in vitro, and move the cargo (individual 2 ⅐parS) stepwise outward along the cell axis (10). Recently, it has been shown that decreasing local intracellular concentration of 2 leads to a decrease on ATP hydrolysis and a stimulation of polymerization of (␦⅐ATP⅐Mg 2ϩ ) 2 or ␦ 2 ⅐ATP⅐ADP⅐Mg 2ϩ onto plasmid DNA, thereby generating one (minus) end of the nucleoprotein filament (parS⅐( 2 ) 10 ⅐((␦⅐ATP⅐Mg 2ϩ ) 2 or ␦ 2 ⅐ATP⅐ ADP⅐Mg 2ϩ ) n . From the initial assembly site (␦⅐ATP⅐Mg 2ϩ ) 2 or ␦ 2 ⅐ATP⅐ADP⅐Mg 2ϩ , polymerization leads to the formation of a nascent ␦ 2 filament that depends on the presence of ATP, but is independent of ATP hydrolysis, because ␦ 2 polymerization was observed in the presence of ATP␥S⅐Mg 2ϩ (see 10). When the plus end of the filament reaches another ternary parS⅐ 2 ⅐(␦⅐ATP⅐Mg 2ϩ ) 2 complex (see Fig. 1B), the 2 :␦ 2 molar ratio decreases. This change in the 2 :␦ 2 ratio stimulates the ATPase activity of ␦ 2 . (␦⅐ADP⅐Mg 2ϩ ) 2 dissociates from the nucleoprotein complex, proximal to the 2 ⅐parS region. Alternatively, (␦⅐ ATP⅐Mg 2ϩ ) 2 , by promoting intramolecular pairing of the six different 2 ⅐parS complexes, condenses the plasmid DNA and plays an essential role in the early stages of plasmid segregation. Upon interaction with 2 ⅐parS, (␦⅐ATP⅐Mg 2ϩ ) 2 binds and polymerizes onto plasmid and/or chromosomal DNA. Protein (␦⅐ ATP⅐Mg 2ϩ ) 2 polymerized onto chromosomal DNA moves the ( 2 ⅐parS) cargo passively toward the cell poles by an 2 -␦ 2 interaction. The proposed models for how 2 wrapped around parS promoters, ␦ 2 -mediated pairing and subsequent polymerization of (␦⅐ATP⅐Mg 2ϩ ) 2 or ␦ 2 ⅐ATP⅐ADP⅐Mg 2ϩ and depolymerization of (␦⅐ADP⅐Mg 2ϩ ) 2 differ markedly from the type II partition system. In the ParRM system, each pair of plasmids, anchored ParM at ParR⅐parC, drives apart the plasmid copies by ParM⅐ADP⅐Mg 2ϩ polymerizing bidirectionally (31, 34 -37).