A Bipolar Spindle of Antiparallel ParM Filaments Drives Bacterial Plasmid Segregation

See allHide authors and affiliations

Science  07 Dec 2012:
Vol. 338, Issue 6112, pp. 1334-1337
DOI: 10.1126/science.1229091


To ensure their stable inheritance by daughter cells during cell division, bacterial low-copy-number plasmids make simple DNA segregating machines that use an elongating protein filament between sister plasmids. In the ParMRC system of the Escherichia coli R1 plasmid, ParM, an actinlike protein, forms the spindle between ParRC complexes on sister plasmids. By using a combination of structural work and total internal reflection fluorescence microscopy, we show that ParRC bound and could accelerate growth at only one end of polar ParM filaments, mechanistically resembling eukaryotic formins. The architecture of ParM filaments enabled two ParRC-bound filaments to associate in an antiparallel orientation, forming a bipolar spindle. The spindle elongated as a bundle of at least two antiparallel filaments, thereby pushing two plasmid clusters toward the poles.

During bacterial cell division, an equal distribution of replicated plasmids to daughter cells ensures their stable inheritance. Low-copy-number plasmids encode the simplest known DNA segregation machines to perform this task. They comprise a nucleotide-driven (cytomotive) protein filament and a centromere-like DNA region, linked by an adaptor protein. The ParMRC segregation system of Escherichia coli R1 plasmid consists of ParM, an actinlike cytomotive protein (1) that forms polar, left-handed, double-helical filaments (2); ParR, an adaptor protein; and parC, a centromeric region (3, 4). Dynamic instability of ParM filaments enables plasmid segregation by a “search and capture” mechanism (5, 6), with ParRC (7, 8) stabilizing the filaments. It has been reported that ParRC binds to both ends of a single ParM filament (9, 10). This leads to a conundrum: How does ParRC bind to two different ends of a polar ParM filament?

Here, we provide a comprehensive description of ParM in the monomeric and filament states. An electron cryomicroscopy (cryo-EM) reconstruction (11) (Fig. 1A) provided a subnanometer-resolution map of the polar filament of ParM (resolution of 8.5 Å at FSC 0.5; fig. S1, A to D), polymerized in the presence of adenylyl-imidodiphosphate [AMPPNP, a nonhydrolyzable adenosine triphosphate (ATP) analog]. We determined the crystal structures of a nonpolymerizing mutant, ParM(Leu163→Arg163, L163R), bound to AMPPNP and ParM(Leu163→Ala163, L163A) bound to the C-terminal 17-residue peptide corresponding to the ParM-interacting region of ParR (8, 10) (ParRpept) and AMPPNP (11). Comparison of the crystal structures, including the previously reported apo- and adenosine diphosphate (ADP)–bound forms (1), revealed no large differences between the ATP and ADP states of ParM monomers (Fig. 1, B and C). In contrast, domains IB and IIB showed a rotation of 9.1° and 10.7° toward the nucleotide-binding pocket (Fig. 1D) in the ParM:ParRpept structure, compared with the free monomer. The domain rotations were reminiscent of the transition from G-actin to F-actin (12, 13). Fitting the monomeric structures of ParM into the cryo-EM reconstruction showed that the ParM:ParRpept structure fits best (Fig. 1, E and F, fig. S1E, and movie S1). Thus, the filament conformation of ParM is very similar to ParM:ParRpept. The best fit provided a quasi-atomic model of the ParM filament (Fig. 1G) and implies that binding of ParR or ParRC locks ParM monomers in the filament-like conformation.

Fig. 1

Conformational cycle of ParM. (A) Cryo-EM reconstruction of the ParM filament. Box indicates a monomeric segment. (B to D) Superposition (using Cα atoms of domain IIA) of the conformations of ParM with the ParM-AMPPNP state (blue; PDB ID: 4A61): (B) unliganded ParM [green (1); PDB ID: 1MWK], (C) ParM-ADP state [magenta; (1); PDB ID: 1MWM], and (D) ParM with ParRpept and AMPPNP (orange; ParRpept in pink cartoon representation; PDB ID: 4A62). Domain rotations are indicated. (E and F) Rigid-body fit of ParM monomeric states into the cryoEM reconstruction. Domain I residues were fitted by using Chimera (24). Real-space R-factors (RSR) against the map for ParM:AMPPNP (E) and ParM:ParRpept (F) quantify the best fit (movie S1 and fig. S1E). (G) A repeat unit of the ParM filament (EMDB: EMD-1980, PDB ID: 4A6).

ParRpept was bound in a hydrophobic pocket between domains IA and IIA of ParM (Fig. 2A and fig. S2). The ParM-ParRpept interaction resembled proteins that bind at the barbed end of actin (Fig. 2, A and B). Many actin-binding proteins, including formins (14) and Wiskott-Aldrich homology domain 2–containing proteins such as Spire (15), insert a helix between the corresponding subdomains 1 and 3 of actin (16). The ParM:ParRpept structure modeled onto the ParM filament highlighted a substantial clash between ParRpept and residues 37 to 46 of the next ParM monomer in the protofilament (Fig. 2C). Thus, the ParR-binding site lies within the polymerization interface of ParM, and bound ParR needs to be replaced during elongation. Overlap in the binding site of ParRpept and the polymerization interface occludes all the ParRC-binding sites on the ParM filament except those at the barbed end, implying that ParRC binds exclusively to this end.

Fig. 2

ParRC binds at the barbed end of ParM filaments. (A and B) The ParRpept binding site corresponds to that of Spire on actin. The helices of the interacting proteins are shown in pink, with the rest in gray. (A) ParM:ParRpept (PDB ID: 4A62), (B) actin:Spire (PDB ID: 3MMV). (C) ParRpept binds at the interprotofilament interface of ParM. Two subunits of the ParM filament with hypothetical ParRpept at the binding sites are shown. In that position, ParRpept clashes with loop 37-46 from domain IB of the adjacent subunit.

Our ParM-ParRpept structure supports a forminlike mechanism for the processive movement of ParRC (fig. S3). Ten ParR dimers bind parC repeats (7, 8), forming a scaffold that allows a forminlike stair-stepping mechanism (fig. S3, B and C) (17). The C-terminal helices from the 20 ParRs are localized in a confined area. This probably facilitates ParM filament nucleation by ParRC (18). It also ensures ParRC-bound ParM monomer recruitment upon ParR displacement from the filament (fig. S3D), explaining end tracking of ParM filaments by ParRC through insertional polymerization (6, 19). The ParRC cap locks the terminal monomers in the filament conformation, thus protecting the filament from dynamic instability.

To confirm the proposed single-end binding of ParRC, we examined the effect of ParRC on ParM filament elongation by using total internal reflection fluorescence (TIRF) microscopy (11). The experiments were performed with nonhydrolyzable AMPPNP to prevent dynamic instability. ParM-AMPPNP elongated symmetrically at both ends from an initial seed of the ParM filament (Fig. 3A and movie S2) (5). In the presence of unlabeled ParRC, one end of the filaments grew faster, resulting in asymmetric growth (Fig. 3B, fig. S4A, and movie S3). The rates of growth were 9.4 ± 4.1 and 2.5 ± 1.9 monomers per s (numbers after the ± symbol indicate standard deviation from the mean), for fast- and slow-growing ends (n = 32; table S3). Experiments with labeled ParRC showed that ParRC accelerated growth and recruited ParM monomers at the ParRC-bound filament end only (Fig. 3C, fig. S4B, and movie S4), reconfirming insertional polymerization (6). The unidirectional elongation by ParRC leads us to the key question: What enables bipolar plasmid segregation?

Fig. 3

ParRC accelerates growth at one end of ParM filaments. (A and B) TIRF microscopy kymographs of filaments in dual-label experiments (11) (A) without ParRC and (B) with ParRC. Growth of ParM filaments without ParRC is bidirectional and symmetric, with equal slopes on both ends of the kymograph, whereas addition of ParRC results in asymmetric growth, with unequal slopes. The relevant boundaries are highlighted. (C) Monomers are recruited to the ParRC-bound end of ParM filament. Kymograph from a filament labeled with Alexa-568 (magenta) and YOYO-1–labeled ParRC (green) are shown.

Frequent condensation events were observed between ParM filaments, both with ATP and AMPPNP (Fig. 4A and movies S5 to S8). Furthermore, continuous motion because of interfilament sliding occurred within filament bundles (Fig. 4B, fig. S4C, and movies S9 to S11). A molecular model for the filament sliding and condensation was generated on the basis of two monomers in the crystal packing of ParM:ParRpept (fig. S5A). ParM filament subunits, when superposed sequentially onto the ParM monomers, produced sliding and an antiparallel packing of filaments (Fig. 4C and fig. S5A). Mutation of residues within loop 18-21 [Ser19→Arg19 and Gly21→Arg21 (S19R and G21R)] at the proposed antiparallel filament interface (Fig. 4, C and D, and fig. S5B) prevented interfilament sliding (movie S12, Fig. 4E, and fig. S4D), because of stronger interactions between filaments caused by alternating charges.

Fig. 4

A bipolar spindle comprises at least two antiparallel ParM filaments. (A) ParM-AMPPNP filaments condensing into a bundle (movie S7). (B) Kymograph of a bundle of ParM. Static filaments yield straight lines, whereas concerted zigzag motion shows sliding filament movement. (C) Atomic model of an antiparallel filament pair. (Inset) Interface involving loop 18-21 in the antiparallel arrangement. D, Asp. (D) Interface mutations and their effect on sliding and spindle stability. (E) Kymograph of a bundle of ParM(S19R,G21R) filaments that are static because of stronger interfilament interactions. (F) Kymographs of disassembling spindles of ParM and ParM(S19E,G21E). Arrows highlight the slopes of disassembly. (G) Montage of spindle disassembly events in ParM(S19E,G21E) mutant (movie S17). The spindle splits apart because of repulsive surface charges, demonstrating that the bipolar spindle of ParM (magenta) is formed of more than one filament. The constituent filaments with ParRC (green) bound at one end begin disassembly from the pointed ends. The pointed ends are highlighted with arrowheads. (H) Schematic representation of the bipolar spindle model (see also fig. S6).

The helical geometry of ParM filament, with 12 subunits per turn, is compatible with a hexagonal or square packing of ParM filaments in a bundle (fig. S5C) (4), in contrast to nonbundling actin filaments with a 13-monomer repeat. Bundles of ParM filaments have been observed in E. coli cells expressing ParM at wild-type levels (20) and during plasmid partitioning in vivo (21). Also, bundles of antiparallel ParM filaments have been described in vitro (22).

Antiparallel pairing explains observations of ParRC at both ends of ParM filaments (9, 10). Of the 826 filaments we counted, 104, 540, and 182 filaments were observed with ParRC at zero, one, and two ends, respectively, consistent with single-end binding and pairing. Bipolar spindles of ParM were observed by TIRF microscopy using ATP. Stable filaments were seen only as elongating spindles with ParRC at both ends, pushing plasmids apart at a rate of 22.6 ± 4.8 monomers per s (n = 40; movies S13 and S14 and table S3). Upon loss of ParRC at limiting concentrations of ParM (350 to 500 nM), dynamic instability caused spindle disassembly at a rate of 100.3 ± 18.7 monomers per s (n = 61; Fig. 4F, fig. S4E, movies S15 and S16, and table S3).

To demonstrate that the spindles comprised at least two antiparallel filaments, we introduced negatively charged residues within loop 18-21 (S19E, G21E, where E indicates Glu) (Fig. 4D) to weaken the interfilament interaction through repulsive electrostatic forces. We observed spindles (n = 65) that split into the constituent filaments (movie S17 and Fig. 4G). This confirms that spindles are not formed by a single ParM filament, with ParRC attached at both ends (9, 10) (Fig. 4H). ParM filaments in a spindle were protected from dynamic instability by binding of ParRC at the barbed end and pairing with another ParRC-bound filament at the pointed end (movies S15 to S19; Fig. 4, F to H; and fig. S4, E and F). The trigger for disassembly was either the loss of ParRC (movies S15 and S16) or exposure of pointed ends because of unwinding of the antiparallel bundle (movies S17 and S19).

These observations and previously published work (3, 4) provide a comprehensive model of ParMRC-mediated plasmid segregation (fig. S6): a critical concentration of ATP-bound monomers in the cell nucleates ParM filaments (or nucleation is ParRC-mediated) (18). ParRC binding rescues the dynamic filaments from disassembly at the barbed end only. ParRC speeds up the growth at the barbed end by a forminlike mechanism. The free pointed end remains prone to disassembly unless it pairs up antiparallel with another ParM filament bound to ParRC, probably shortly after plasmid replication. Thus, a bipolar spindle of two antiparallel filaments is the minimum unit in plasmid segregation. R1 plasmid has a copy number of about six, which is about the number of filaments within bundles in plasmid-segregating cells (20). ParM bundles are stronger than single filaments, which is advantageous when pushing plasmids through the cytoplasm. A single bundle will also ensure concerted segregation of all sister plasmids, as observed in vivo (21).

The lateral interaction among dynamic filaments, as in ParMRC, may also facilitate contraction in other cytomotive filament systems such as FtsZ in bacterial cell division (23). Our model of antiparallel actinlike ParM filaments, without the necessity of bundling factors or motor proteins, provides an attractive conceptual precursor for actin contractile systems, such as muscle.

Supplementary Materials

Materials and Methods

Figs. S1 to S6

Tables S1 to S3

References (2441)

Movies S1 to S19

References and Notes

  1. See supplementary materials available on Science Online.
  2. Acknowledgments: We acknowledge N. Barry and C. Johnson (MRC-LMB, help with TIRF microscopy and isothermal titration calorimetry); T. Kato (Osaka University, technical support with cryo-EM); beamline support at ID29, ID14-2 (European Synchrotron Radiation Facility, France), and I02 (Diamond Light Source, UK); L. Amos, S. Bullock, A. Carter, and R. Gasper-Schönenbrücher (MRC-LMB, critical reading of the manuscript); and K. Gerdes (Newcastle University) and A. Toste Rêgo (MRC-LMB) (provision of reagents). Support by the MRC (U105184326 to J.L.), Grants-in-Aid for Scientific Research, Ministry of Education, Culture, Sports, Science and Technology (MEXT), Japan (21227006 to K.N.), and FP7-IIF-2008 fellowship (P.G.) is acknowledged. Protein Data Bank (PDB) accession codes: ParM-AMPPNP, 4A61; ParM-AMPPNP-ParRpept, 4A62. The EM reconstruction has been deposited in the Electron Microscopy Data Bank (EMD-1980), fitted coordinates in the PDB (4A6J).
View Abstract

Navigate This Article