ParA-like protein uses nonspecific chromosomal DNA binding to partition protein complexes

Abstract

Recent data have shown that plasmid partitioning Par-like systems are used by some bacterial cells to control localization of protein complexes. Here we demonstrate that one of these homologs, PpfA, uses nonspecific chromosome binding to separate cytoplasmic clusters of chemotaxis proteins upon division. Using fluorescent microscopy and point mutations, we show dynamic chromosome binding and Walker-type ATPase activity are essential for cluster segregation. The N-terminal domain of a cytoplasmic chemoreceptor encoded next to ppfA is also required for segregation, probably functioning as a ParB analog to control PpfA ATPase activity. An orphan ParA involved in segregating protein clusters therefore uses a similar mechanism to plasmid-segregating ParA/B systems and requires a partner protein for function. Given the large number of genomes that encode orphan ParAs, this may be a common mechanism regulating segregation of proteins and protein complexes.

Bacterial cell division is a carefully regulated process that ensures each daughter cell receives a copy of the genetic material. There are three main mechanisms for bacterial plasmid DNA segregation (1): tubulin-like GTPases, actin-like ATPases, and Walker A cytoskeletal P-loop ATPases. In all cases, the system requires three components: (i) an NTPase protein, which is thought to provide the movement required for partitioning; (ii) a partner protein, which binds to the DNA to be partitioned and regulates NTPase activity; and (iii) a centromere or region of the DNA, which binds the partner protein. The two proteins required for partitioning are normally found encoded next to each other on the genome and are found in a wide range of species (2). The Walker A cytoskeletal P-loop ATPase system is the common mechanism found in plasmid and chromosome segregation, and is typified by the ParA/B system.

Genes homologous to parA are also found outside of ParA/B operons, not adjacent to a parB, and are termed orphan ParA systems (2, 3). Several of these orphan parA genes are found in the middle of signaling and metabolic operons. Several of these orphan ParAs have been shown to be involved in the segregation of protein clusters upon division (3–5); however, their mechanisms of segregation are currently unclear. The proteins displaying ParA-dependent segregation form complexes where precise stoichiometry of the proteins in the complex may be required for accurate function and where each daughter cell needs to contain that protein complex immediately upon cell division. For example, cyanobacteria with a mutated parA are unable to segregate carboxysomes and take 9 extra hours to divide, presumably due to the time taken to synthesize a new carboxysome in the daughter cells (5). From our analysis of published genomes, most bacterial genomes encode an orphan parA, and species can contain between 1 and 10 orphan parA genes. Bioinformatic analysis suggests that some orphan Par systems are likely to work like plasmid partition systems (3).

Plasmid and chromosomal ParA/B systems have been the most widely studied and can be divided into type Ia and Ib systems, with type Ia ParA proteins containing a regulatory N-terminal region that binds to DNA to control par gene expression, whereas type Ib ParA proteins lack this regulatory N-terminal region. ParA acts as the Walker box ATPase in both type Ia and Ib systems, with ParB binding the plasmid and also activating the ParA ATPase in both types. ATP binds the Walker box of ParA and stimulates ParA dimer formation (1, 6). Both type Ia and Ib ParA dimers then bind DNA nonspecifically (7), and this DNA binding results in ParA polymerizing bidirectionally to form filaments along the DNA (8, 9). ParB binds to the plasmid cargo to be segregated by binding to a specific site (parS) on the plasmid. When ParB interacts with ParA, it stimulates the ATPase activity of ParA, resulting in depolymerization of the ParA filament with basic residues in the N-terminal region of ParB responsible for the stimulation of ParA ATPase activity (6, 10, 11). In some systems, polymerization and depolymerization result in the ParA oscillating over the chromosome (8). A related plasmid segregation system uses a ParA-like protein, ParF, which interacts with a protein unrelated to ParB, ParG. This, centromere-binding protein stimulates the ATPase activity of ParF via its N terminus (11). Proteins in other species show no sequence similarity to ParG but are able to act as functional equivalents (12).

The mechanism by which polymerization/depolymerization results in partitioning of plasmids is unclear, although there are a number of mathematical models that explain partitioning based on the biological data. These include a pulling model, with length-dependent detachment (8), and a diffusion ratchet model (9). Both of these mechanisms require the ParA protein to undergo a dynamic cycle of ATP hydrolysis, release, and nonspecific rebinding to the DNA. In all cases, the ATP hydrolysis of the ParA controls both its cellular localization and ParB binding activity. In plasmid segregation, the result of these systems is one plasmid focus at midcell in a newly divided cell, and once copied this changes to two plasmid foci at 1/4 and 3/4 positions, at what will be new midcell positions after division (13).

In some bacterial species, chromosome segregation uses related proteins, but the mechanism in this case moves the chromosome origins to the poles of the dividing cell. An example is the dynamic ParA-dependent segregation of the Caulobacter cresentus chromosome. In this case, ParA does not oscillate but instead uses filaments of ParA, which are guided to the poles by the action of TipN and PopZ (14, 15). Thus, despite the clear similarity between ParA/B systems, evidence suggests there are a variety of different mechanisms used depending on the species and the cargo to be segregated.

The ParA/B system also shows strong similarity to the oscillating MinCDE system required for midcell positioning of the FtsZ ring in Escherichia coli, with MinD being the Walker ATPase and MinE acting to stimulate the ATPase activity (7). In both the Min and Par systems, the activity of the ATPase is regulated by a partner protein to control localization of complexes within the cell. Both systems also require a surface on which to polymerize, DNA in the case of ParA and the cell membrane in the case of MinD (1, 7).

An orphan ParA was shown to control the segregation of a cytoplasmic protein cluster involved in chemotactic signaling in Rhodobacter sphaeroides (4). The chemotaxis system in R. sphaeroides is encoded in three operons, of which two are expressed under laboratory conditions. These operons encode two pathways that form independent protein clusters: one located in the cytoplasm, with soluble chemoreceptors, and one at the pole of the cell, with membrane-spanning chemoreceptors (16). The formation of the polar cluster probably depends on stochastic self-assembly, as is the case in E. coli (17). The formation of the cytoplasmic cluster has been shown to be dependent on the cytoplasmic chemoreceptor TlpT and the linker protein CheW₄, both components of the chemotaxis cluster (18). The single cytoplasmic cluster normally localizes to midcell in a new cell and, before cell division, two clusters move to 1/4 and 3/4 positions such that each daughter cell receives a cluster upon division. This segregation and movement is reminiscent of ParA/B plasmid segregation and is dependent on a type Ib orphan ParA homolog, PpfA, encoded in the third chemotaxis operon (4). Homologous orphan parA genes have been identified in over 53% of chemotaxis operons. Orphan parA genes have also been shown to control the localization of polar chemotaxis proteins in Vibrio cholera, in this case probably using a system related to the PopZ system of chromosome segregation (3). Bioinformatic analysis has suggested that there may be a number of distinct clades of Par-like orphan systems (3), probably using different mechanisms to segregate membrane and cytoplasmic complexes.

We have investigated the mechanism of PpfA-dependent segregation of the cytoplasmic chemotaxis cluster in R. sphaeroides and show that, as with plasmid-segregating ParA, it depends upon dynamic localization to the chromosome, using nonspecific DNA binding and driven by its ATPase activity. The mechanism also requires interaction with a partner protein, in this case the N terminus of the soluble chemoreceptor TlpT.

Results

Designing Mutants.

To investigate whether PpfA uses a mechanism similar to classical ParA proteins, PpfA was aligned with ParA proteins. PpfA aligns well with ParA proteins, including Soj (ParA) from Bacillus subtilis and Thermus thermophilus (Fig. S1). Three residues previously shown to be functionally important in ParA proteins (6) are G16, with G16V in Soj preventing dimerization; K20, with K20A in Soj preventing ATP binding; and D44, with D44A locking Soj in an ATP-bound state. All these residues are conserved in PpfA (Fig. S1). A subset of ParA proteins has also been shown to nonspecifically bind DNA as part of their mechanism (3, 19). The alignment of PpfA shows suitable residues at the correct positions for predicted DNA binding, and threading PpfA through the Soj structure (6) suggests that these residues would be in the correct orientation to allow for DNA binding. These residues were therefore mutated to identify whether nonspecific chromosome binding might be involved in segregation of the chemosensory protein complex in vivo.

In classical ParA/B plasmid-segregating systems, ParB is encoded immediately downstream of parA and, with parS, forms the par operon. Basic residues in the N terminus of ParB stimulate the ATPase action of ParA, helping drive plasmid segregation. TlpT, a soluble chemoreceptor, is encoded immediately downstream of ppfA, an arrangement found in 80% of chemotaxis operons encoding PpfA-like proteins (20). The presence of TlpT has previously been shown to be essential for cluster formation, with deletion abolishing cluster formation (18), suggesting that it forms the scaffold for other chemosensory proteins. TlpT has a chemoreceptor domain at the C terminus, immediately preceded by two HAMP (present in Histidine kinase, Adenyl cyclase, Methyl-accepting chemotaxis and Phosphatase proteins) domains. The N-terminal 120 amino acids, however, do not have an identifiable domain but do contain 33 basic residues, suggesting it may act as a ParB-like domain. We therefore deleted this region to test this hypothesis, in vivo.

Effects on Cytoplasmic Chemotaxis Cluster Localization.

It has previously been shown that deletion of PpfA prevents the segregation of the cytoplasmic chemotaxis cluster in R. sphaeroides (4). To test whether mutations in the Walker ATPase regions and putative DNA-binding residues of ParA proteins also prevented cluster segregation, the corresponding mutants (G10V, K14A, D39A and R167E, K196E) were made in ppfA, and the mutant genes were introduced into the genome of R. sphaeroides to replace ppfA.

Wild-type cells have one cytoplasmic chemotaxis cluster at midcell just after division. As the cell grows this becomes two cytoplasmic chemotaxis clusters, which move to future midcell (1/4 and 3/4) positions to ensure each daughter cell receives one cluster upon division. R. sphaeroides has a doubling time of about 170 min under the conditions used, and dividing cells represent about 10% of the population. In ΔppfA cells, one cluster is visible throughout the cell cycle. This is not in any fixed position and increases in intensity throughout the cell cycle, suggesting all new cluster protein incorporates into the cluster (4). Cell division results in one cell with a cluster and one without a cluster, a new cluster becoming visible about 20 min after cell division, suggesting newly synthesized protein now nucleates to form a new cluster (4). It is currently impossible to determine whether PpfA duplicates or segregates the cluster. All four point mutations result in phenotypes similar to ΔppfA, with only one cluster per cell and cells dividing to produce one daughter with a cluster and one lacking a cluster, a new cluster becoming visible after about 20 min. This suggests that each of the residues mutated is essential for PpfA-dependent segregation but not formation of the cytoplasmic chemotaxis cluster (Fig. 1 and Table 1).

Fig. 1.

Effect of PpfA point mutations on cytoplasmic chemotaxis cluster partitioning. Fluorescence images of TlpT-YFP as a measure of cytoplasmic chemosensory cluster localization in wild-type and PpfA mutant backgrounds. Population measurements are shown in Table 1.

Table 1.

Number of chemotaxis clusters per cell in different strains

Strain	Clusters per cell (% of population)
	0	1	2
Wild type	0	84.2	15.8
PpfA deletion	33.3	66.7	0
PpfAG10V	29.9	70.1	0
PpFAK14A	19.9	80.1	0
PpfAD39A	22.1	77.9	0
PpfAR167E, K196E	30.5	69.5	0
TlpTΔNterm	20.1	79.9	0

This is an average of the entire population.

Deletion of TlpT causes the loss of clusters, but the deletion of just the N-terminal region of TlpT alone allows the cytoplasmic chemotaxis clusters to still form; however, these clusters are no longer segregated (Table 1), resulting in a phenotype similar to ppfA deletion. Cephalexin-treated filamentous cells have multiple cytoplasmic chemosensory clusters, but deletion of the N terminus of TlpT results in a single, randomly positioned cluster, similar to a ppfA deletion phenotype (Fig. S2). These data suggest that the N terminus of TlpT, which contains a number of basic residues, is required for correct segregation and may interact with PpfA in the partitioning process. These data also demonstrate that the N-terminal region of TlpT is not required for cluster formation, only for positioning.

Effect of Mutations on the Location of PpfA in the Cell.

PpfA fused to CFP was expressed from an inducible plasmid in a TlpT-YFP background strain to determine the localization of PpfA within R. sphaeroides relative to the cytoplasmic chemotaxis cluster. PpfA was diffuse throughout the cell but also formed distinct foci colocalized with the chemotaxis cluster, suggesting a greater concentration of PpfA coincident with the cytoplasmic chemotaxis protein cluster (Fig. 2A). This was independent of the level of PpfA-CFP induction, and the same localization pattern was also seen in a background when TlpT was not fused to YFP, indicating it is not an effect of the fusion proteins. Overexpression of PpfA-CFP partially complements a ppfA deletion, with the percentage of cells containing two cytoplasmic chemotaxis clusters increasing from 0% in ΔppfA to 11% when complemented compared with 25% in wild type. Oscillation has been observed in some plasmid partition systems. We were not able to observe any oscillations of PpfA-CFP using observation intervals of 0.1 s, 1 min, 5 min, or 10 min.

Fig. 2.

Localization of PpfA in R. sphaeroides cells. (A–E) PpfA-CFP (A) and point mutants of PpfA-CFP (B–E) were expressed from a plasmid in a TlpT-YFP background. TlpT-YFP is used as a marker for cytoplasmic chemosensory cluster positioning. Each image shows the merged (Upper Left) Differential Interference Contrast (Upper Right), YFP channel showing TlpT localization (Lower Left), and CFP channel showing PpfA localization (Lower Right). The following point mutations were visualized: (B) PpfAG10V-CFP; (C) PpfAK14A-CFP; (D) PpfAD39A-CFP; (E) PpfAR167E, K196E-CFP. (F and G) Fluorescence images of PpfA-CFP and TlpTΔN-terminal-YFP.

The mutations that resulted in loss of function of PpfA were made in PpfA-CFP to identify whether PpfA localization patterns depend on the residues identified as essential for cluster segregation. Two distinct patterns of localization were observed with the mutants: G10V, K14A, and R167E, K196E each resulted in diffuse PpfA with no observable foci (Fig. 2 B, C, and E), which is consistent with their predicted phenotypes (loss of dimerization, ATP binding, and DNA binding), resulting in the loss of TlpT binding. However, PpfA D39A, which, if equivalent to ParA, should lock PpfA in an ATP-bound state, resulted in PpfA foci that were brighter than wild-type cells and clearly coincident with chemotaxis cluster foci (Fig. 2D). These data are consistent with PpfA acting analogously to ParA and strongly suggest that PpfA interaction with the chemotaxis cluster requires both ATP binding and dimerization. These data also show that the Walker box is essential both for localization of PpfA to the cluster and for cluster segregation.

To test whether the N terminus of TlpT does interact with PpfA and hence control its localization, PpfA-CFP was expressed from an inducible plasmid in a TlpTΔNterm-YFP background (Fig. 2F). In a wild-type background, PpfA is diffuse, with foci colocalized with the cytoplasmic chemosensory cluster; however, when the N-terminal region of TlpT was deleted, no PpfA-CFP foci were observed. In a wild-type TlpT background, PpfAD39A formed bright foci coinciding with the cytoplasmic chemosensory cluster, but when expressed in the TlpTΔNterm-YFP background, no foci were observed (Fig. 2G). These data are consistent with the hypothesis that the N terminus of TlpT interacts with PpfA and thus may act as a ParB. These data also support the hypothesis that PpfA needs to be ATP-bound to interact with the chemosensory cluster.

Effect on Localization in E. coli.

The R. sphaeroides chromosome occupies the full cell volume and, as with other α-subgroup proteobacteria, it is not possible to condense the R. sphaeroides chromosome. This makes it very difficult to identify whether the diffuse PpfA is bound to the chromosome of R. sphaeroides. To test whether PpfA binds DNA nonspecifically, we expressed PpfA-YFP in E. coli and stained the chromosome with DAPI. PpfA was shown to colocalize with the chromosomal DNA in E. coli (Fig. 3A). Mutation of the putative DNA-binding residues (R167E and K196E) resulted in PpfA becoming diffuse throughout the E. coli cell and no longer being colocalized with the condensed chromosome (Fig. 3B). This suggests that these residues are required for nonspecific DNA binding. Mutations in the PpfA Walker box (G10V and K14A) also result in loss of colocalization with the chromosome, and resulted in PpfA being diffuse throughout the cell (Fig. 3 C and D). PpfAD39A, which is expected to be locked in the ATP-bound form, still colocalized with the E. coli chromosome (Fig. 3E). All these data suggest that PpfA binds DNA nonspecifically, and binds only in the dimeric ATP-bound state.

Fig. 3.

Localization of PpfA in E. coli demonstrates it binds DNA nonspecifically. Point mutations of PpfA-YFP expressed from a plasmid in E. coli whose chromosomal DNA has been labeled with DAPI. Each image shows the merged (Upper Left) DIC (Upper Right), YFP channel showing PpfA localization (Lower Left), and DAPI channel (Lower Right). The following point mutations are shown: (A) PpfA-CFP; (B) PpfAR167E, K196E-YFP; (C) PpfAG10V-YFP; (D) PpfAD39A-YFP; (E) PpfAD39A-YFP.

The dynamics of PpfA dimerized and associated with the chromosomal DNA in E. coli is expected to be very different from the dynamics of protein in monomeric form, free in the cytoplasm. The dynamics of the chromosome-bound and diffuse PpfA were therefore measured using fluorescence recovery after photobleaching (FRAP). PpfA-CFP has a half-life of around 10 s to recover to the bleached region, and this was very similar to the time taken by the D39A mutant, suggesting these proteins are not freely diffusing but that their rate of movement is restricted (Table 2 and Fig. S3). However, the G10V, K14A and R167E, K196E mutant PpfA-CFP proteins each has half-lives that are too short to be measured from the experimental data, suggesting these proteins are freely diffusing in the cytoplasm.

Table 2.

FRAP analysis of PpfA in E. coli

Mutation	FRAP recovery t1/2 (s)
Wild type (PpfA-YFP)	16 ± 2
G10V	<1
K14A	<1
D39A	19 ± 3
R167E, K196E	<1

FRAP recovery half-lives of PpfA-YFP and point mutations of PpfA-YFP expressed from a plasmid in E. coli. Sample images are in Fig. S3. The range represents SEM.

FRAP Measurements of PpfA in R. sphaeroides.

If the hypothesis that PpfA is acting like a plasmid-partitioning ParA protein is correct, then it would be expected that different dynamics of PpfA would depend on whether it is bound to the N terminus of TlpT in the cytoplasmic chemotaxis cluster, bound to the chromosome, or free in the cytoplasm of R. sphaeroides. FRAP was used to measure the mobility of PpfA and the mutant PpfA proteins in vivo (Table 3 and Figs. S3 and S4). The rate of return of PpfA-CFP to a bleached spot was faster, compared with the D39A mutant, than that measured in E. coli, suggesting E. coli lacks a protein capable of stimulating PpfA. The rate of return was similar regardless of whether the bleached region contained a cytoplasmic cluster, suggesting PpfA is constantly being turned over in the wild-type state. No oscillation was also observed in the recovery curves. PpfAD39A, which mimics the ATP-bound form and forms more stable foci with TlpT, showed a slow rate compared with wild type of recovery to the bleached foci, suggesting PpfA is strongly bound to TlpT and the cytoplasmic chemotaxis cluster. The mutations G10V, K14A and R167E, K196E in the Walker ATPase and nonspecific DNA-binding residues, respectively, each showed very rapid recovery, suggesting that these mutants do not bind to either the cluster or DNA but are free in the cytoplasm.

Table 3.

FRAP analysis of PpfA in R. sphaeroides

Mutation	FRAP recovery t1/2 (s)
PpfA-CFP (foci)	10 ± 2
PpfA-CFP (diffuse)	8 ± 1
PpfAG10V	<1
PpfAK14A	<1
PpfAD39A (foci)	22 ± 3
PpfAR167E, K196E	<1
PpfA-CFP in TlpTΔN strain	16 ± 1

FRAP recovery half-lives for PpfA-CFP and point mutations of PpfA-CFP expressed from a plasmid in R. sphaeroides. Sample images are in Figs. S4 and S5. The range represents SEM.

If the N terminus of TlpT were to be acting as a ParB analog, by analogy to the ParA/B systems a change in PpfA dynamics upon the deletion of the TlpT N terminus would be expected. To test this, PpfA-CFP recovery rates were tested by FRAP in a TlpTΔNterm strain (Fig. S5). In this strain, PpfA-CFP shows slower recovery than the wild-type rate, and this rate is similar to the slow recovery rate shown by the PpfAD39A mutant in wild-type cells. These data suggest that the N terminus of TlpT is acting to stimulate the ATPase activity of PpfA, increasing its dynamics in the cell in a manner analogous to ParB in the classical ParA/ParB systems.

Discussion

Orphan ParA proteins are found encoded in many genomes (2), and some of these have been shown to be involved in protein complex segregation (1). Although these orphan ParAs have been shown to be required to segregate protein complexes upon cell division and position them correctly in the cell, the mechanism of action of these orphan ParAs is currently unknown, but recent studies of the ParA homolog in V. cholerae that localized the chemosensory proteins to the cell pole suggest there are at least two different families of protein-partitioning ParA homologs.

PpfA, which ensures each R. sphaeroides cell has a cluster of chemosensory proteins upon cell division, appears to use a mechanism related to that suggested for plasmid segregation. Bioinformatic-led mutagenesis and in vivo measurements of subsequent protein localization in both R. sphaeroides and E. coli show that PpfA-dependent localization and segregation depend on the nonspecific interaction of PpfA with the chromosome. PpfA clearly binds to the E. coli chromosome, indicating that the binding of PpfA to DNA is nonspecific and does not require the activity of any other R. sphaeroides proteins. The slow recovery shown by the FRAP data also suggests that this binding is either relatively tight or that the dynamic turnover when associated with the chromosome is slow when no R. sphaeroides proteins are present. Point mutations that prevent nonspecific DNA (and hence chromosomal DNA) binding also prevent cytoplasmic chemotaxis cluster segregation in R. sphaeroides, again demonstrating the importance of nonspecific DNA binding for the segregation process and suggesting PpfA is using the chromosome as a scaffold for segregation.

The dependence of both segregation and PpfA localization on the Walker ATPase is indicated by equivalent mutations to those identified in other ParA systems. These data suggest that PpfA uses a mechanism very similar to that of ParA and thus requires ATPase activity and dimerization to function. Active PpfA protein has proved impossible to purify to date; however, the different localization phenotypes and dynamics of the range of mutants analyzed are consistent with those that would be expected for a ParA protein and strongly suggest that PpfA undergoes the same cycle involving ATP binding, dimerization, DNA binding, and ATP hydrolysis described for classical ParA proteins. For example, as has been seen with ParAD44A from C. cresentus (14), the equivalent PpfAD39A ATP-locked mutant forms tight foci which coincide with the cargo to be partitioned, in the case of PpfA the cytoplasmic chemotaxis cluster.

The N terminus of TlpT is also clearly required for the segregation of the cytoplasmic chemotaxis cluster. Without this region the chemosensory cluster still forms, but does not segregate. The deletion of this region also results in a loss of PpfA foci coincident with the chemosensory cluster, demonstrating that the N terminus of TlpT is interacting with PpfA at these foci. PpfAD39A forms strong foci with the cytoplasmic chemotaxis cluster and, when the N terminus of TlpT is deleted, PpfAD39A no longer forms these foci, adding further support to TlpT interacting with PpfA and suggesting the ATP-bound form of PpfA is the state that binds to TlpT. The N terminus of TlpT contains a number of basic residues, and it has previously been shown that basic residues in ParB proteins are used for ParA ATPase activation (6). The reduction in the dynamics of the wild-type PpfA to the similar rate as the PpfAD39A ATPase-locked protein in the TlpT N-terminal deletion supports a role for the N terminus of TlpT in stimulating PpfA ATPase activity. As there is no significant difference in the dynamics of the wild-type PpfA and PpfAD39A when expressed in E. coli, it is likely that E. coli does not contain a protein capable of stimulating the ATPase activity of PpfA. Therefore, despite the lack of an identifiable ParB homolog encoded next to orphan parA genes, the corresponding activity may be found encoded in other neighboring genes. In the case of PpfA, ParB activity is found in the N terminus of TlpT, a chemoreceptor. As with ParG/F systems, TlpT shows no sequence homology to ParB but is acting like a ParB. Bacterial genomes with multiple chemosensory operons often have a PpfA protein encoded within that operon, and in over 80% of cases they are encoded next to a TlpT homolog, suggesting that this mechanism of segregating chemotaxis proteins may be common (20).

PpfA, an orphan ParA encoded in the cytoplasmic chemosensory cluster operon of R. sphaeroides, uses a similar mechanism to that described for plasmid-segregating ParA proteins, requiring a dynamic cycle of ATP binding, dimerization, nonspecific DNA binding, and ATP hydrolysis to ensure that each daughter cell receives a cytoplasmic chemotaxis protein cluster. It is probable that PpfA can only bind DNA in the dimeric ATP-bound state and in this state is also able to bind the N terminus of TlpT, forming local foci. If ATP-bound PpfA forms filaments over the chromosome, as suggested for ParA, interaction with the N terminus of TlpT would result in ATP hydrolysis and movement, resulting in the chemosensory cluster remaining roughly centered over the chromosome. The cycle appears similar to that observed for plasmid-partitioning ParA proteins, with the ATP-bound form able to bind the cargo to be partitioned. Previous work has suggested two mechanisms may exist for Par-mediated protein segregation (3), and this paper demonstrates that PpfA’s mechanisms are similar to that of plasmid segregation. A possible stepwise sequence for PpfA action consistent with these data is shown in Fig. S6. It should be noted that this sequence is consistent with either the diffusion ratchet or the filament pulling models previously suggested for ParA proteins (21). A possible mechanism would be that if the chemosensory clusters roughly center on the chromosome as a result of this type of interaction between PpfA polymerized on the chromosome surface and TlpT in the chemosensory cluster, the separation of the chromosomes before cell division would increase the PpfA surface area to an extent that may allow a second cluster to nucleate and segregate. Using a range of optical systems and time courses, we were unable to observe oscillation of PpfA. If PpfA does not oscillate, it suggests its mechanism might be more like the P1 Par system, where gradients of the portioning protein drive the partitioning (13). The N-terminal region of TlpT appears to stimulate the ATPase activity of ParA, suggesting that orphan ParAs may not be orphan in the mechanistic sense, but may function with domains of neighboring, associated proteins. The exact mechanism of how the cluster is split remains unclear, but TlpT is also essential for cluster formation. As the proteins are constantly expressed through the cell cycle, there may be a point after chromosome segregation when the level of PpfA in the cell interacting with TlpT is sufficient to nucleate a second cluster on the new chromosome. We suspect that, given the number of orphan ParAs in bacterial genomes, this system may be used in other species to partition other, very different, protein complexes in bacteria before division.

Materials and Methods

Bacterial Strains and Growth Conditions.

R. sphaeroides strains were grown aerobically in succinate medium (22) at 30 °C with shaking at 225 rpm. E. coli strains were grown aerobically in LB broth at 37 °C with shaking at 225 rpm. The bacterial strains and plasmids used in this study are outlined in Table S1.

Sequence Alignment and Threading.

Sequences were aligned using Clustal X (23), and a homolog model with Soj (6) was made using SWISS-MODEL (24).

PpfA Mutants.

Mutations in ppfA were created using the QuikChange XL Kit (Agilent) (primers in Table S2). PpfA chromosomal mutants were created by chromosomal replacement (25) into JPA1328, which contains an in-frame deletion of ppfA, or JPA1553, which contains TlpT-YFP and an in-frame deletion of ppfA in the genome.

Overexpression of PpfA.

ppfA and mutant ppfA were cloned into pIND containing CFP or YFP in-frame downstream of the insertion site, producing PpfA tagged at the C terminus. These plasmids were propagated in E. coli and conjugated into R. sphaeroides strains as described previously (26). Isopropyl β-d-1 thiogalactopyranoside was then used to express PpfA in both R. sphaeroides and E. coli, and the resultant strains were observed using fluorescence microscopy.

Fluorescence Microscopy and Analysis.

Cells were immobilized on a thin layer of 0.8% agarose in succinate medium on microscope slides (4). Differential interference contrast microscopy and fluorescence images were acquired with a Nikon TE200 microscope and CFP and YFP filter set (Chroma) and recorded with a cooled charge-coupled device camera (Hamamatsu Photonics).

For short-time interval imaging (0.1-s and 1-min periods), live wide-field fluorescence microscopy was performed using an OMX v2 (Applied Precision) microscope, exciting at 488 nm, and recorded on a Roper Scientific Cascade II cooled CCD camera with a 525/50 filter.

DAPI Staining of E. coli.

Cells were grown to an OD₆₀₀ of 0.4. One microliter of DAPI was added to 1 mL of cells and incubated for 5 min at room temperature. Cells were then washed twice in LB and visualized using fluorescence microscopy as above using a DAPI filter (Chroma).

Analysis of Filamentous Cells.

Cells were grown aerobically to an OD₇₀₀ of 0.2 and treated with 2.5 μg/mL cephalexin for 4 h before being viewed by fluorescence microscopy (as described above).

FRAP Acquisition and Analysis.

FRAP was performed on an A1 confocal system (Nikon) using a 100× objective with the pinhole open. Slides of E. coli and filamentous R. sphaeroides were prepared per the fluorescent microscopy protocol. Cells were excited and observed using a 40-mW argon laser (457.9 nm for CFP and 514.5 nm for YFP). Cells were imaged prebleach every 250 ms for 14 s (E. coli) or 2 s (R. sphaeroides). An ∼1-μm–diameter region of the cell was then bleached by stimulating with the excitation laser for 300 ms at a single point in the field of view and then observing the field of view every 250 ms (E. coli) or 750 ms (R. sphaeroides) for 2 min. To determine FRAP times, circular regions of interest of equal size were defined around bleached and unbleached foci and diffuse areas of PpfA using NIS elements software (Nikon). The effect of acquisition photobleaching was corrected for by measuring the percentage loss of total cell fluorescence at each time point, correcting back to the point of bleaching and correcting the loss in the regions of interest by this percentage. The time taken for the fluorescence intensity in the bleached region to reach half the plateau level was calculated. At least five cells were analyzed per strain.