Essential Role of the Cytoplasmic Chemoreceptor TlpT in the De Novo Formation of Chemosensory Complexes in Rhodobacter sphaeroides

ABSTRACT

Bacterial chemosensory proteins form large hexagonal arrays. Several key features of chemotactic signaling depend on these large arrays, namely, cooperativity between receptors, sensitivity, integration of different signals, and adaptation. The best-studied arrays are the membrane-associated arrays found in most bacteria. Rhodobacter sphaeroides has two spatially distinct chemosensory arrays, one is transmembrane and the other is cytoplasmic. These two arrays work together to control a single flagellum. Deletion of one of the soluble chemoreceptors, TlpT, results in the loss of the formation of the cytoplasmic array. Here, we show the expression of TlpT in a tlpT deletion background results in the reformation of the cytoplasmic array. The number of arrays formed is dependent on the cell length, indicating spatial limitations on the number of arrays in a cell and stochastic assembly. Deletion of PpfA, a protein required for the positioning and segregation of the cytoplasmic array, results in slower array formation upon TlpT expression and fewer arrays, suggesting it accelerates cluster assembly.

IMPORTANCE Bacterial chemosensory arrays are usually membrane associated and consist of thousands of copies of receptors, adaptor proteins, kinases, and adaptation enzymes packed into large hexagonal structures. Rhodobacter sphaeroides also has cytoplasmic arrays, which divide and segregate using a chromosome-associated ATPase, PpfA. The expression of the soluble chemoreceptor TlpT is shown to drive the formation of the arrays, accelerated by PpfA. The positioning of these de novo arrays suggests their position is the result of stochastic assembly rather than active positioning.

INTRODUCTION

To survive in various and changing environments, bacteria must be adaptable. One of the clearest examples of this is the control of motility to ensure movement toward favorable conditions, a behavior known as chemotaxis. Most species that swim through aqueous media do so by the rotation of extracellular semirigid helical filaments, known as flagella (1). For example, Escherichia coli has multiple flagella (typically 4 to 6), which when all rotating in a counterclockwise direction, form a bundle pushing the cell forward. If one or more of the flagella change to a clockwise rotation, then this bundle will fly apart and the cell with cease to swim and tumble on the spot. It is the balance between these swimming and tumbling behaviors which allows bacteria to direct their movement. E. coli, as well as many other bacteria, employs a modified two-component signaling pathway to adjust the frequency of tumbling in response to environmental stimuli (2). Receptors spanning the inner membrane (methyl-accepting chemotaxis proteins [MCPs]) sense attractants and repellents in the periplasm and transmit these signals across the membrane to their cytoplasmic tips (3), which interact with an adaptor protein, CheW, and a histidine protein kinase, CheA. In response to worsening (i.e., high-repellent low-attractant) conditions, the MCPs stimulate the kinase activity of CheA, resulting in the phosphorylation of the response regulator CheY (4). Phosphorylated CheY (CheY-P) is able to interact with the cytoplasmic C ring of the flagellar motor, inducing a change in the direction of rotation and thus a cell tumble. To ensure a timely return to swimming, E. coli has a dedicated CheY-P phosphatase, CheZ. A second response regulator, the methylesterase CheB, is phosphorylated by CheA. CheB together with the methyltransferase CheR form the adaptation system. They act on specific glutamate residues of the MCPs, methylating (CheR) and demethylating (CheB-P) them, with the methylation state of these glutamates determining the sensitivity of the MCPs to stimuli (5).

The proteins in this pathway do not function alone but rather as large arrays. MCPs form homodimers that assemble into mixed trimers of dimers making up the hexagonal pattern seen in cryo-electron microscopy (cryo-EM) across many species (6). These arrays assemble in the inner membrane by the interaction of CheA and CheW at the base of the array. This array architecture has several important functions in chemotaxis signaling. First, the close spatial proximity of receptors within the array allows cooperativity for activating CheA; this is the case not just between similar receptors but also between those sensing different stimuli, allowing for the integration of multiple signals into a single output. In E. coli, there are four MCPs, two in high copy numbers and two in low copy numbers, and the oxygen sensor Aer. While only the high-copy-number MCPs can recruit CheR, all of them can be methylated/demethylated as the mixing of MCP dimers within the array brings the binding domains into proximity. Thus, the mixing of receptors within the array enables adaptation across all the receptors.

In many bacterial species, including E. coli, these transmembrane arrays are predominantly found at the poles of the cells (6, 7), with some laterally positioned arrays seen in longer cells (8). This polar preference presents an elegant solution for ensuring each daughter cell inherits at least one array upon cell division. This requires cells that have inherited just one array at their old pole to create and position a second one at the other pole before another round of division. Interestingly, in E. coli, the lateral clusters were found to be positioned periodically along the length of the cell (9). This spatial organization of arrays suggests an underlying mechanism, with the current model for how this pattern is achieved being stochastic self-assembly (10). In this model, MCP molecules newly inserted into the membrane can have two fates, either to join an existing array or to nucleate the formation of a new one. In small cells, MCPs are more likely to join an existing array by random diffusion; as the cell grows, the probability swings toward nucleation at the point furthest from the existing array, i.e., the other pole. Further growth then allows for lateral arrays to form at midcell, equidistant from both poles. Superresolution microscopy has revealed a range of array sizes is present in cells and they are spatially distributed as predicted by the stochastic self-assembly model, with large clusters spaced as far apart as possible and small clusters and single molecules between (11).

Not all species position their chemosensory arrays in this manner (12). Caulobacter crescentus divides asymmetrically into a swimmer and stalked cell and only needs to form an array in the swimmer daughter cell. This is done via TipN, a marker of the new cell pole, indicating a cellular landmark mechanism for array positioning (13, 14). The positioning of the chemosensory array in Vibrio spp. requires a landmark protein but is also dependent on a ParA homologue, ATPase proteins that are involved in positioning and segregating DNA and protein assemblies in many bacterial species (15). HubP marks the poles of the cell and is required for the capturing and positioning of ParA homologues required for positioning of the ori of chromosome I (ParA1), the flagellar motor (FlhG), and the chemosensory array (ParC) (16). As the cell cycle progresses, ParC transitions from a single focus at the old pole to foci at both poles in a HubP-dependent manner. After a second ParC focus has formed, the chemotaxis proteins are recruited, forming the second array (17). This recruitment is dependent on a second protein, ParP, encoded immediately upstream of ParC. It has been shown to interact with both ParC and CheA and to be able to stabilize the array, thus possibly aiding the nucleation of the second array (18).

Many bacterial species have more than one chemosensory pathway (19). Rhodobacter sphaeroides is a well-studied species with multiple pathways. It expresses two spatially distinct chemosensory arrays, namely, a transmembrane array and a cytoplasmic array (20). The positioning of the transmembrane array appears similar to that of E. coli, primarily polar but with some smaller lateral clusters, although there are a few notable differences, including their absence from midcell when the Z ring is present (21). While the transmembrane arrays are stabilized by both the membrane and the CheA/W baseplate, it was previously unclear how the cytoplasmic arrays are arranged without a membrane. Cryo-EM studies revealed that the array forms a double layer, with the sensing domains of the receptors facing inwards and the CheA/W baseplates on the outside (22). Interestingly, a similar structure was seen in vitro for the cytoplasmic fragment of the E. coli receptor Tar along with CheA, CheW, and molecular crowding agents. Earlier work had shown that CheW and CheA are required to bring receptors together, and the similarity of the hexagonal arrangements of both membrane and cytoplasmic structures suggests spontaneous cross-linking. The cytoplasmic arrays show a positioning that differs from that of the transmembrane arrays, with a single array positioned midcell of a newly divided cell. As the cell progresses through the cell cycle, a second array forms, and the two arrays are positioned at roughly one-quarter and three-quarter positions. This duplication and segregation is dependent on a ParA ATPase homologue, PpfA, encoded in the same operon as the components of the array (23). In ΔppfA cells, only one cluster is formed, so that on division, only one daughter cell inherits an array. Interestingly, the cells lacking clusters after division form new arrays as they grow, although they are nonchemotactic until clusters have formed. Previous work suggested one of the two soluble chemoreceptors in this array, TlpT, has a key role in the formation and segregation of the array. Without TlpT, the array does not form and the other array components are diffuse in the cytoplasm (24). TlpT has also been implicated in interactions with PpfA, with an N-terminal truncation allowing cluster formation but with the same loss of cluster duplication and segregation phenotype as with the ΔppfA mutant (25).

To further understand the roles of both TlpT and PpfA in the formation and segregation of the cytoplasmic chemosensory array of R. sphaeroides, we induced TlpT in a ΔtlpT strain in a background with a fluorescently tagged cytoplasmic CheW (CheW4-yellow fluorescent protein [YFP]). As the arrays reformed, their numbers and positioning were evaluated using time-lapse fluorescence microscopy. These experiments were repeated with a strain also lacking ppfA to determine the impact of PpfA on the formation of the array.

RESULTS

Clusters reform on the reintroduction of TlpT.

To test whether reintroduction of TlpT is sufficient for formation of the cytoplasmic chemosensory array, TlpT expression from the pIND4 plasmid was induced by the addition of 50 μM IPTG (isopropyl-β-d-thiogalactopyranoside) to both ΔtlpT cheW4-yfp and ΔtlpT ΔppfA cheW4-yfp cultures. The cells were immediately placed on the microscope and the cellular distribution of CheW4-YFP was followed over time. These time courses (Fig. 1) show that, upon expression of TlpT, arrays reform in cells both with and without ppfA. The addition of IPTG to ΔtlpT cheW4-yfp cells with empty pIND results in no change in the distribution of CheW4-YFP within the cell (see Fig. S1 in the supplemental material). To quantify these results, the images were analyzed using MicrobeTracker (26) to determine the number of detectable fluorescent foci within individual cells at each time point (Fig. 2). For ΔtlpT cells, there was a sharp increase in the number of cells with 1 or 2 arrays and a corresponding drop in those with none after the addition of IPTG. The percentages of cells with 0, 1, and 2 arrays appear to stabilize at approximately 50 min, indicating that the continued expression of TlpT does not result in ever greater numbers of arrays. This suggests that a fixed number of arrays can form in a cell. This may be due to limiting amounts of the other proteins that make up the array, or the size of the cell spatially limits the number of arrays able to form, akin to the stochastic self-assembly model of E. coli array formation.

FIG 1.

Fluorescence time course images of ΔtlpT cheW4-yfp pIND4-tlpT (top) and ΔtlpT ΔppfA cheW4-yfp pIND4-tlpT (bottom) cells showing the formation of cytoplasmic chemosensory arrays following the reexpression of TlpT. Time is minutes from the addition of IPTG. Bars, 1 μm.

FIG 2.

Percentages of cells with 0, 1, and 2 fluorescent spots over time following the addition of IPTG for ΔtlpT cheW4-yfp pIND4-tlpT (A) and ΔtlpT ΔppfA cheW4-yfp pIND4-tlpT (B) cells. The average numbers of cells analyzed at each time point are 620 (A) and 235 (B).

Interestingly, cells lacking both tlpT and ppfA showed a slower response to the expression of TlpT (Fig. 2B). The rate of increase in cells with 1 array was slower than that of those with PpfA, and the number of cells in each category did not stabilize during the long time period of the experiment. This suggests that while PpfA is not necessary for array formation, its presence accelerates the process. It is also notable that 6% of ΔtlpT ΔppfA cells did form 2 arrays after expression of TlpT, while cells just lacking ppfA never formed more than a single array.

Positioning of newly formed clusters.

Wild-type cells have either a single array roughly at midcell or two arrays near the one-quarter and three-quarter positions. Therefore, we compared this positioning pattern with that of the newly formed arrays in the experiments above. Kymographs of individual cells (Fig. 3) show that cells adopt patterns similar to those seen in cells with preexisting arrays, and that the arrays appear reasonably static after formation both in the presence and absence of PpfA.

FIG 3.

Kymographs of array formation in ΔtlpT cheW4-yfp pIND4-tlpT (A) and ΔtlpT ΔppfA cheW4-yfp pIND4-tlpT (B) cells. Time is from the addition of IPTG; each pixel represents 5 min on the x axis.

Analyzing the positions of newly formed arrays across the population was performed using MicrobeTracker. Spot positions were normalized to cell length, and the distribution of array positions was obtained for each time point during the experiment (Fig. 4). This shows that for ΔtlpT cells expressing TlpT, the distribution is similar to the distribution of 1 and 2 arrays seen in wild-type (wt) tlpT-yfp cells. These data suggest that once arrays are of a detectable size, they remain positioned rather than form first and then subsequently position.

FIG 4.

Histograms showing the positions of arrays along the long axes of the cells at each time point for ΔtlpT cheW4-yfp pIND4-tlpT (A) and ΔtlpT ΔppfA cheW4-yfp pIND4-tlpT (B) cells. For both, data from cells with one array (i) and two arrays (ii) are shown. The number of arrays at each position is given by the color scale.

The ΔtlpT ΔppfA cells expressing TlpT showed a distribution of arrays similar to that of ΔppfA tlpT-yfp cells, with the distribution centered around the midcell but broader than that measured in wt tlpT-yfp cells. The spot positions in the very few ΔppfA ΔtlpT cells which displayed 2 arrays showed one-quarter and three-quarter positioning, suggesting that they are both complete arrays and that the 2nd array is not an artifact of the analysis.

Length dependency of the number of clusters formed.

Continued expression of TlpT does not result in ever-increasing numbers of arrays, indicating a limiting factor for array numbers. This could be either the amount of the other cytoplasmic chemosensory proteins or cell size. To assess the influence of cell size on array numbers, cells at the end of the time course were binned by cell length and the percentages of cells with different numbers of arrays were determined (Fig. 5Ai and Bi). This shows that for cells both with and without ppfA, an increasing cell length results in an increasing propensity to have higher numbers of arrays. For a given cell length, cells with PpfA formed a greater number of arrays than those without.

FIG 5.

The proportions of cells in each length bin with different numbers of fluorescent spots for ΔtlpT cheW4-yfp pIND4-tlpT (A) and ΔtlpT ΔppfA cheW4-yfp pIND4-tlpT (B) cells, both untreated (i) and cephalexin treated (ii).

To further assess the numbers and positioning of these newly formed arrays, the experiments were repeated with filamentous cells elongated by cephalexin treatment (see Fig. S2 and S3). As cephalexin prevents septation, cells continue to grow without dividing, resulting in filamentous cells. These elongated cells allow for testing of the reformation of arrays in a range of cell sizes that extends beyond those of untreated cells.

These elongated cells display the same trend as the untreated cells, whereby increasing cell length allows the formation of a greater number of arrays. Cells lacking ppfA showed a reduced number of arrays formed at a given cell length compared with cells with PpfA.

DISCUSSION

It was previously shown that arrays did not form in cells with a tlpT deletion, while single arrays formed in cells with ppfA deletions. Here, we show that by simply reintroducing TlpT into a background where the other chemosensory proteins are diffuse, these proteins come together to form arrays. This, along with work on the E. coli receptor Tar showing that arrays form in vitro in the presence of CheW, CheA, and molecular crowding agents (22), highlights the inherent tendency for these proteins to form hexagonal arrays, whether transmembrane or cytoplasmic. This tendency may be an evolutionary response to the need for cooperativity for these bacterial chemosensory systems to function, both in signaling and adaptation. While the second soluble chemoreceptor, TlpC, is not sufficient to form arrays in the absence of TlpT, cells lacking TlpC retain arrays, indicating the key structural role of TlpT in these arrays.

It was previously shown that PpfA is required for the segregation of cytoplasmic arrays on cell division but not for their formation. This is confirmed here, with de novo array formation upon the expression of TlpT, independent of PpfA.

Interestingly, although PpfA is not required for array formation, it does appear to influence it. A quantitative analysis of the time course for array formation showed that cells lacking PpfA form on average fewer arrays, and they do so more slowly than those with PpfA. The cell length dependency of the number of arrays formed with PpfA suggests a mechanism driving the formation of the chemosensory clusters similar to that seen for the E. coli chemoreceptor array, that is, stochastic assembly. These results also suggest that PpfA may both lower the spatial requirement for a given number of arrays, stabilizing the formation of arrays such that they form faster, and shift the system toward the nucleation of arrays. This stabilization may be due to TlpT interacting with PpfA, which itself interacts with DNA, providing a potential scaffold to promote array formation. If the nucleoid does act as scaffold, it would have been interesting to correlate the numbers of arrays and nucleoids; however, the nucleoids in R. sphaeroides do not condense, and therefore this was not possible.

An analysis of the positioning of the newly formed arrays showed that once formed, they move very little, indicating that either they remained positioned where they form or they are positioned before they reach a detectable size. Short cells form one array at midcell, while longer cells form two arrays at one-quarter and three-quarter positions, typical of cells in a wild-type situation. In these experiments, the arrays were formed de novo, and the pattern of positioning could be the result of simultaneous array formation, and thus a stochastic self-assembly-type mechanism, or due to PpfA. In the few cases (6%) where ΔppfA cells formed two arrays, they were at one-quarter and three-quarter positions. In cephalexin-treated cells, arrays were distributed along the cell length, suggesting that in this artificial situation, arrays can position independent of ppfA, indicating stochastic assembly aided by the arrangement of PpfA on the chromosome.

Altogether, these data suggest that where arrays are formed de novo and simultaneously, a stochastic self-assembly mechanism drives array formation. The presence of PpfA is shown to influence this, as it results in faster array formation and a greater number of arrays. Therefore, PpfA not only segregates and positions the arrays but helps drive their formation.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

Strains used in this study are listed in Table 1. R. sphaeroides strains were grown in succinate medium (29) aerobically at 30°C with shaking. The ΔtlpT ΔppfA cheW4-yfp strain was created by a single deletion of both tlpT and ppfA, which are adjacent on the genome, from a cheW4-yfp background (20). The ΔtlpT cheW4-yfp strain was created previously (24). Elongated cells were prepared by adding 1 μg/ml cephalexin to cultures at an optical density at 700 nm (OD₇₀₀) of 0.2 and allowing them to grow for a further 2 h.

TABLE 1.

Strains used in this study

Strain	Description
ΔtlpT cheW4-yfp pIND strain	WS8N (27) with a deletion of tlpT, cheW4-yfp in place of the wild-type cheW4, and empty pIND plasmid (28)
ΔtlpT cheW4-yfp pIND-tlpT strain	WS8N with a deletion of tlpT, cheW4-yfp in place of the wild-type cheW4, and pIND plasmid containing tlpT
ΔtlpT ΔppfA cheW4-yfp pIND strain	WS8N with deletions of tlpT and ppfA, cheW4-yfp in place of the wild-type cheW4, and empty pIND plasmid
ΔtlpT ΔppfA cheW4-yfp pIND-tlpT strain	WS8N with deletions of tlpT and ppfA, cheW4-yfp in place of the wild-type cheW4, and pIND plasmid containing tlpT

Fluorescence microscopy.

Cells were immobilized on a poly-l-lysine (0.01% [wt/vol] in water)-coated tunnel slides. After washing with fresh medium, the cells were washed with medium containing 50 μM IPTG, and then the ends of the tunnel were sealed with nail varnish and the slide was placed immediately on the microscope for imaging. A Nikon Ti microscope was used with a 100× differential interference contrast (DIC) objective (Nikon). The Perfect Focus system was employed to prevent drift in the z axis during time courses. A YFP filter set (Chroma) was used along with an Andor iXon charge-coupled-device (CCD) camera to capture images. Images were captured every 5 min for 180 min. For each strain and condition, 3 biological replicates were imaged and analyzed except for cephalexin-treated ΔtlpT ΔppfA cheW4-yfp pIND4-tlpT cells, where 7 biological replicates were used to generate an adequate number of cells for analysis.

Image analysis.

DIC images were converted to pseudophase contrast images using custom-written software. This used a Hilbert transformation adapted from previously developed software (30). The cells in these images were then segmented using MicrobeTracker (26). The fluorescence images were analyzed using the SpotFinderZ function. Spot-finding parameters were trained on images of cells with existing arrays (cheW4-yfp) and those with completely diffuse fluorescence (ΔtlpT cheW4-yfp) to reduce instances of false positives and negatives.