Portal de Periódicos da CAPES

Spatial organization in bacterial chemotaxis

2010; Springer Nature; Volume: 29; Issue: 16 Linguagem: Inglês

10.1038/emboj.2010.178

1460-2075

Victor Sourjik, Judith P. Armitage,

Bacterial Genetics and Biotechnology

Focus Review18 August 2010free access Spatial organization in bacterial chemotaxis Victor Sourjik Corresponding Author Victor Sourjik Zentrum für Molekulare Biologie der Universität Heidelberg, DKFZ-ZMBH Alliance, Im Neuenheimer Feld, Heidelberg, Germany Search for more papers by this author Judith P Armitage Corresponding Author Judith P Armitage Department of Biochemistry, Oxford Centre for Integrative Systems Biology, University of Oxford, Oxford, UK Search for more papers by this author Victor Sourjik Corresponding Author Victor Sourjik Zentrum für Molekulare Biologie der Universität Heidelberg, DKFZ-ZMBH Alliance, Im Neuenheimer Feld, Heidelberg, Germany Search for more papers by this author Judith P Armitage Corresponding Author Judith P Armitage Department of Biochemistry, Oxford Centre for Integrative Systems Biology, University of Oxford, Oxford, UK Search for more papers by this author Author Information Victor Sourjik 1 and Judith P Armitage 2 1Zentrum für Molekulare Biologie der Universität Heidelberg, DKFZ-ZMBH Alliance, Im Neuenheimer Feld, Heidelberg, Germany 2Department of Biochemistry, Oxford Centre for Integrative Systems Biology, University of Oxford, Oxford, UK *Corresponding authors: Zentrum für Molekulare Biologie der Universität Heidelberg, Im Neuenheimer Feld 282, DKFZ-ZMBH Alliance, Heidelberg 69120, Germany. Tel.: +49 6221 54 6858; Fax: +49 6221 54 5894; E-mail: [email protected] of Biochemistry, Oxford Centre for Integrative Systems Biology, University of Oxford, Oxford OX1 3QU, UK. Tel.: +44 1865 613293; Fax: +44 1865 613338; E-mail: [email protected] The EMBO Journal (2010)29:2724-2733https://doi.org/10.1038/emboj.2010.178 PDFDownload PDF of article text and main figures. ToolsAdd to favoritesDownload CitationsTrack CitationsPermissions ShareFacebookTwitterLinked InMendeleyWechatReddit Figures & Info Spatial organization of signalling is not an exclusive property of eukaryotic cells. Despite the fact that bacterial signalling pathways are generally simpler than those in eukaryotes, there are several well-documented examples of higher-order intracellular signalling structures in bacteria. One of the most prominent and best-characterized structures is formed by proteins that control bacterial chemotaxis. Signals in chemotaxis are processed by ordered arrays, or clusters, of receptors and associated proteins, which amplify and integrate chemotactic stimuli in a highly cooperative manner. Receptor clusters further serve to scaffold protein interactions, enhancing the efficiency and specificity of the pathway reactions and preventing the formation of signalling gradients through the cell body. Moreover, clustering can also ensure spatial separation of multiple chemotaxis systems in one bacterium. Assembly of receptor clusters appears to be a stochastic process, but bacteria evolved mechanisms to ensure optimal cluster distribution along the cell body for partitioning to daughter cells at division. Bacterial two-component systems Most environmental sensing in bacteria is performed by the so-called two-component systems (see Bourret and Silversmith, 2010 and references therein for recent reviews on the two-component systems). A canonical bacterial two-component system is comparatively simple and consists of a sensory histidine kinase and a phosphorylatable response regulator (Figure 1A). The kinase typically spans the cytoplasmic membrane and senses the external stimuli using its periplasmic domain (Cheung and Hendrickson, 2010). In most cases, the signal is transmitted to the cytoplasmic part of the kinase through a conformational change in the stable kinase dimer, which causes changes in the kinase ability to autophosphorylate at a conserved histidine residue. In contrast to typical receptor-associated tyrosine or serine/threonine kinases in eukaryotes, the autophosphorylation does not stabilize the histidine kinase in an active state that is able to further phosphorylate substrate proteins. Rather, the phosphoryl group is directly transferred from the kinase histidine residue to the aspartate residue of the response regulator, dephosphorylating the kinase at each cycle of response regulator activation. This ensures not only fast signal termination, but also prevents the type of signal amplification that is common in eukaryotes, whereby a single molecule of stably activated kinase can phosphorylate multiple downstream target proteins before being dephosphorylated by a specific phosphatase. Figure 1.Two-component and chemotaxis signalling in bacteria. (A) Schematic representation of the canonical two-component system. Sensory histidine kinase (HK) is composed of the input (blue) and the autokinase (red) domains; the kinase is typically a dimer. The response regulator (RR) consists of the receiver (purple) and output (green) domains. The phosphate group is transferred from the histidine residue on the kinase to the asparate residue on the response regulator, activating the output domain, which typically regulates gene expression. The response regulator can be dephosphorylated by the phosphatase activity of the kinase. (B) Molecular composition of the chemotaxis pathway in E. coli. Receptors sense and transmit signals to regulate the activity of the cytoplasmic histidine kinase CheA. Receptors form trimers of dimers, where different types of receptors (light or dark blue) are mixed. CheA binding and regulation by receptors are aided by CheW. CheA transfers phosphate group to CheY, the single-domain response regulator controlling flagellar motor, and to CheB, composed of the regulatory receiver domain and the output methylesterase domain. Receptors are methylated on glutamate residues by the methyltransferase CheR. CheY is dephosphorylated by the phosphatase CheZ. Receptors, CheW, CheA and CheZ form a stable signalling core, to which CheR, CheB and CheY dynamically localize. Download figure Download PowerPoint Two-component response regulators typically consist of two domains—a regulatory, or receiver, domain phosphorylated by the sensory kinase and an output domain that performs a physiological function inside the cell (Bourret, 2010). In most cases, the output domain is a transcription factor that is inhibited by the unphosphorylated regulatory domain. The phosphorylated state of the response regulator is often intrinsically unstable, decaying in a matter of minutes or even seconds. Dephosphorylation of the response regulator is frequently further enhanced by the respective sensory kinase, which may, therefore, have a dual kinase–phosphatase function depending on its activity state (Kenney, 2010). Although most of the two-component kinases appear to be evenly distributed over the cell membrane (VS, unpublished), several cases of specific spatial organization have been established, for example for the pathway that controls the asymmetric division of Caulobacter crescentus (Thanbichler, 2009). This review will focus on the spatial organization of the pathways that control chemotaxis, which form the most prominent signalling structures known so far in bacteria. We will consider examples of Escherichia coli, in which chemotactic signalling has been most thoroughly studied, and Rhodobacter sphaeroides, which represents a higher level of spatial organization complexity. Strategy of bacterial chemotaxis Chemotaxis pathways mediate navigation of motile bacterial cells in chemical gradients, enabling bacteria to migrate towards higher concentrations of attractants, while avoiding repellents (Sourjik, 2004; Wadhams and Armitage, 2004; Vladimirov and Sourjik, 2009). The small size of a bacterial cell would make direct spatial detection of gradients inefficient (Berg and Purcell, 1977), and bacteria consequently evolved a chemotaxis strategy that relies on temporal—rather than spatial—comparisons of chemoeffector concentrations (Berg and Brown, 1972; Macnab and Koshland, 1972). Bacteria moving in a gradient initially choose a swimming direction at random, but can then rapidly sense whether the chosen direction is favourable or not, and prolong the movement in a favourable swimming direction. Typically, the time for such temporal comparisons is limited to a few seconds, because on a longer time scale the cell would become reoriented by Brownian motion. The sensory pathway thus has to perform several tasks: (i) react to changes in chemoeffector concentration on a sub-second time scale; (ii) compare the level of stimulation at a given time point with that of 1–2 s earlier, which requires a short-term memory and (iii) gradually refresh the memory as the cell moves up or down the gradient. In E. coli, the general chemotaxis strategy is implemented by controlling the relative frequency of two swimming states: a smooth swimming 'run', which corresponds to the counterclockwise rotation of the flagellar motors and propels the cell forward, and re-orienting 'tumbles', which are produced by the clockwise (CW) motor rotation. In the adapted state with no gradients present, runs last ∼2 s and are interrupted by ∼0.1 s tumbles (Berg and Brown, 1972). As a result, cell performs a random walk that allows it to efficiently explore its environment. In the presence of a gradient, cells bias their random walk by suppressing tumbles, whereas swimming in a favourable direction, which results in an efficient net movement of a population up the gradient. Bacteria with single flagella achieve a similar random walk by either switching motor direction, resulting in a brief period of pulling rather than pushing the cell, or simply stopping flagellar rotation and letting Brownian motion reorient the cell for the next swimming period. The frequency regulation of reversals or stops produces the biased random walk. This seemingly simple strategy has been greatly optimized through the course of evolution. Detailed quantitative analyses of E. coli chemotactic behaviour have shown that the response resulting in motor switching is extremely sensitive, with cells being able to detect minute changes in the level of stimulation (Segall et al, 1986), close to the physical limit set by the noise of ligand binding (Berg and Purcell, 1977). Cells can integrate multiple stimuli, enabling navigation in mixed gradients (Adler and Tso, 1974; Kalinin et al, 2010). Moreover, signalling by the pathway is robust against perturbations, such as variations in ambient ligands concentrations or in protein levels (Barkai and Leibler, 1997; Kollmann et al, 2005), which allows cells to maintain efficient chemotaxis in varying environments. Signalling in E. coli chemotaxis Given the complexity of its functions, it is not surprising that the chemosensory pathway in E. coli is much more sophisticated than a canonical two-component system (Figure 1B). Although signalling in chemotaxis similarly relies on a regulated phosphotransfer between a histidine kinase and a response regulator, the kinase CheA does not possess a sensory domain, but instead—together with the 'adaptor' protein CheW—associates with dedicated membrane-spanning receptors. E. coli has five types of such receptors, with four of them (Tar, Tsr, Tap and Trg) sensing not only a range of amino acids, sugars and dipeptides, but also pH and temperature, and the fifth (Aer) sensing redox potential. Changes in ligand binding to a periplasmic domain of receptors modulate conformation of their cytoplasmic parts (Hazelbauer et al, 2008; Hazelbauer and Lai, 2010) and consequently the autophosphorylation activity of receptor-associated CheA, whereby attractant binding decreases CheA activity and repellent binding enhances it. CheW is apparently required for this regulation, although its exact function remains unclear. As in other two-component systems, the CheA phosphoryl group is subsequently transferred to the response regulator CheY, which can diffuse through the cytoplasm and transmit the signal, in this case to flagellar motors. CheY is an example of a single-domain response regulator, in which the output function—binding to the motor—is encoded in the receiver domain. Binding of phosphorylated CheY to the motors enhances the probability of CW rotation and thus causes the cell to tumble. When the cell swims up an attractant gradient, increased attractant stimulation results in CheA inactivation and suppression of tumbles, which promotes continued swimming in this direction. Rapid suppression of tumbles is further promoted by the sub-second dephosphorylation of CheY-P, which—in contrast to the canonical two-component sensors—is ensured by the dedicated phosphatase CheZ rather than by the phosphatase activity of the kinase itself. Another unique feature of chemotaxis compared with the other two-component systems is its adaptability, which is mediated by methylation and demethylation of chemoreceptors on four specific glutamate residues by the methyltransferase CheR and the methylesterase CheB, respectively. Methylation tunes the ability of receptors to activate CheA, with higher modification increasing CheA activity and decreasing receptor sensitivity to attractants. Methylation thus resets the pathway activity upon an initial stimulation, counteracting the effects of ligand binding. Adaptation is achieved by at least two feedback mechanisms linking the activity of the ternary complex of receptors, kinase and CheW to the adaptation system. At a primary level, the feedback is provided by the substrate specificity of the adaptation enzymes, whereby CheR preferentially methylates inactive receptors and CheB preferentially demethylates active receptors (Barkai and Leibler, 1997; Alon et al, 1999; Amin and Hazelbauer, 2010). An additional negative feedback comes from a CheA-dependent phosphorylation of CheB, which possesses an inhibitory regulatory receiver domain. The kinetics of methylation and demethylation are relatively slow (Goy et al, 1977), because of the enzymatic properties of CheR and CheB and their low copy numbers. Depending on the initial stimulation strength, the adaptation time can take tens or hundreds of seconds, but for cells swimming in a typical gradient the methylation level lags behind changes in receptor activity by about a second (Brown and Berg, 1974; Berg and Tedesco, 1975). This lag confers the short-term memory to the chemotaxis system, which is required for temporal comparisons of attractant concentrations and is thus essential for chemotaxis. Spatial organization of E. coli chemotaxis pathway Chemoreceptor clusters Although in vitro experiments suggest that the transmembrane signal transduction and kinase regulation in chemotaxis could be performed by small receptor–kinase complexes that consist of two to three receptor dimers, several CheWs and one CheA (Boldog et al, 2006; Amin and Hazelbauer, 2010), these complexes are organized in the cell into much larger macromolecular clusters that can contain thousands of receptors and associated chemotaxis proteins (Figure 2A). Receptor clusters are one of the largest structures in bacterial cells and can be easily observed at cell poles and along the cell body with immuno-electron, cryo-electron and fluorescence microscopy (Maddock and Shapiro, 1993; Sourjik and Berg, 2000; Zhang et al, 2007; Briegel et al, 2008). In these clusters, receptors are arranged in roughly hexagonal arrays (Figure 2B and C) that are presumably formed by trimers of receptor homodimers (Kim et al, 1999; Briegel et al, 2008, 2009; Khursigara et al, 2008). Clusters are further stabilized by the association of CheA and/or CheW (Maddock and Shapiro, 1993; Skidmore et al, 2000; Sourjik and Berg, 2000; Studdert and Parkinson, 2005; Kentner et al, 2006). Receptors of different ligand specificities appear to be uniformly mixed in the arrays, interacting with each other both directly and indirectly through CheA and CheW. Receptor arrays are not perfectly regular structures: the hexagonal order appears to be frequently distorted (Khursigara et al, 2008) and the stoichiometry of the receptors to CheW and CheA can vary (Levit et al, 2002; Sourjik and Berg, 2004). Figure 2.Receptor clusters in bacteria. (A) Fluorescence images of receptor clusters in E. coli cells. Images were obtained by expressing CheY fusion to yellow fluorescent protein (YFP) at the native location of the cheY gene on the chromosome. (B) Electron cryo-tomography images of receptor clusters in Vibrio cholerae, viewed from the 'top'. Cartoon illustrates fitting of trimers of dimers into the hexagonal lattice of a receptor array. Six trimers of dimers enclose one hexagon (red). Image is the courtesy of Ariane Briegel and Grant J Jensen, based on Briegel et al (2009). (C) Corresponding schematic representation of the receptor array, illustrating its functions in signal processing. According to the MWC model, receptors function in cooperatively switching signalling teams of 10–20 receptor dimers (yellow shading), whereby one team may correspond to a hexagon of the lattice (red). Adaptation enzymes tethered by binding to the pentapeptide sequence of receptors (CheR, green pentagon) can methylate or demethylate glutamates on ∼6 receptors, creating an adaptation neighbourhood (light blue shading). Download figure Download PowerPoint Interactions between receptors, CheW and CheA that are involved in the formation of the cluster core primarily occur at the cytoplasmic tip of the receptors (Miller et al, 2006; Park et al, 2006), although the exact arrangement of these proteins continues to be debated. All other chemotaxis proteins localize to the clusters by interaction with either receptors or CheA. CheR and CheB both bind to the NWETF pentapeptide sequence at the C-terminus of the major receptors, Tsr and Tar (Wu et al, 1996; Barnakov et al, 1999; Shiomi et al, 2002; Banno et al, 2004). CheZ interacts specifically with the N-terminus of the short form of CheA, CheAs, that lacks the first 97 amino acids including the phosphorylation site (Sourjik and Berg, 2000; Cantwell et al, 2003; Cantwell and Manson, 2009; Hao et al, 2009; O'Connor et al, 2009). CheY associates with the cluster through binding to the P2 domain of CheA (Sourjik and Berg, 2000), which also binds CheB (Li et al, 1995). Cluster stability Fluorescence recovery after photobleaching experiments in vivo (Schulmeister et al, 2008) and biochemical measurements in vitro (Gegner et al, 1992; Erbse and Falke, 2009) showed that the sensory core of the cluster is stable on the time scale of chemotactic signalling. In vivo, receptors show no exchange between clusters even after 30 min, whereas CheW and CheA exchange very slowly, with characteristic times of 12 min, suggesting that signalling in chemotaxis is mediated by conformational changes in the cluster and not by its assembly or disassembly. Nevertheless, slow equilibration of the core components on the time scale of cell division could be beneficial, ensuring that the stoichiometry of sensory complexes is uniform throughout the cluster and between clusters, and that receptors of different specificities are well mixed. Yet another component of the stable cluster core is CheZ, with the exchange time of CheZ being ∼8 min (Schulmeister et al, 2008). Other proteins associate with the cluster in a more dynamic way, in an apparent correspondence with their physiological function. The adaptation enzymes exchange between clusters on the time scale of ∼15 s, which is much longer than the excitation time of the chemotaxis system (∼0.1 s), but is comparable with the time required for adaptation to saturating stimuli. Equilibration on this time scale may be essential to ensure a uniform distribution of adaptation enzymes—and thereby of receptor methylation states—in a cluster, as well as between multiple clusters in the same cell, particularly considering the low copy numbers (200–400) of CheR and CheB compared with ∼15 000 receptors (Li and Hazelbauer, 2004). On the other hand, a finite dwell time of the adaptation enzymes at the cluster appears to be important to allow fluctuations in the adapted level of kinase activity, optimizing bacterial search behaviour (Korobkova et al, 2004; Emonet and Cluzel, 2008; Matthaus et al, 2009). In contrast, CheY shows rapid cluster exchange kinetics on the time scale of several seconds, in concert with its function as messenger between the spatially localized sensory clusters and flagellar motors. Cluster assembly and positioning Recent experiments that determined the distribution of the size and number of receptor clusters and their dependence on the expression levels of chemotaxis proteins and on the cell length (Thiem and Sourjik, 2008; Greenfield et al, 2009) suggest that receptor clusters are formed by stochastic self-assembly, whereby individual receptors are inserted along the entire cell membrane (Shiomi et al, 2006) and subsequently either join existing clusters or nucleate new ones. Whereas the former process is primarily limited by receptor diffusion in the membrane and thus depends on the distance of the insertion site from the existing clusters, the latter is governed by the concentration of free receptors (Thiem and Sourjik, 2008). This interplay between concentration-dependent growth of old clusters and distance-dependent nucleation of new clusters results in the regular distribution of nucleated clusters along the cell body (Thiem et al, 2007; Thiem and Sourjik, 2008; Wang et al, 2008). Stochastic self-assembly can further account for the observed exponential distribution of receptor cluster sizes in photoactivated localization microscopy experiments (Greenfield et al, 2009). Such mode of receptor cluster assembly couples cluster nucleation to cell body extension, such that additional clusters are nucleated away from the existing ones before a cell doubles its length and divides. This provides a simple and efficient mechanism to ensure an approximately uniform segregation of receptor complexes during cell division. Receptor segregation appears to be additionally aided by the anchoring of larger lateral clusters to a putative periodic structure at future cell division sites (Thiem et al, 2007), possibly preventing cluster movement and accumulation on one side of the cell. As another consequence of anchoring at the division sites, clusters become positioned at a cell pole after a subsequent cell division. After division, clusters are apparently free to move, but remain restricted to the pole by some other mechanism, possibly the membrane curvature or lipid composition of the polar membrane. Membrane curvature is further likely to limit cluster size, destabilizing larger clusters because of the mismatch with the intrinsic curvature of receptor oligomers (Kim et al, 1999). Such mismatch may also explain the observed partial disorder in receptor arrays (Khursigara et al, 2008), which apparently consist of loosely connected small groups of well-arranged receptors. The uniform cluster distribution along the cell body is not only important for cluster segregation, but it also benefits signalling. Uniform cluster distribution ensures that, even in longer cells, sites of CheY phosphorylation are not too remote from the laterally distributed flagellar motors, and the kinetics of chemotactic signalling is not limited by the diffusion of phosphorylated CheY (Sourjik and Berg, 2002a; Lipkow et al, 2004). Function of spatial organization in E. coli signal processing Scaffolding and regulation What are the benefits of clustering for chemotaxis signalling? One obvious function—as in many other pathways—is to increase the efficiency of pathway reactions by raising local concentrations of all pathway components. Cluster assembly tends to position chemotaxis proteins favourably for their catalytic activities, and mutations that specifically affect protein localization can be compensated by overexpression of the affected proteins (Okumura et al, 1998; Jahreis et al, 2004). Localization seems to be particularly important for the function of the adaptation enzymes. Both CheR and CheB are able to modify several adjacent receptors while remaining tethered to one pentapeptide site, which is connected to the rest of the receptor by a flexible linker (Li and Hazelbauer, 2006), creating the so-called adaptational assistance neighbourhoods (Figure 2C) with an estimated size of about six receptor dimers (Li and Hazelbauer, 2005). Such neighbourhoods appear to be important to ensure high precision of adaptation to a wide range of stimulus strengths (Endres and Wingreen, 2006), by providing each enzyme with a much larger number of available methylation sites—48 for a neighbourhood of six dimers instead of 8 for an individual dimer—and thereby allowing a much more gradual adjustment of the overall receptor activity and preventing enzyme saturation. Protein localization to the cluster can also have a subtler function. Although CheZ localization to the site of CheY phosphorylation seems counterintuitive, its importance seems to be in flattening the concentration gradient of phosphorylated CheY through the cell, so that flagellar motors at different distances from receptor clusters 'see' similar levels of phosphorylated CheY (Vaknin and Berg, 2004). Moreover, it is apparent that binding sites to the cluster for several proteins—CheA and CheW, CheR and CheB, and CheY and CheB—overlap, and binding competition between individual proteins might have a regulatory function. For example, the competition between CheY and CheB (Li et al, 1995; Kentner and Sourjik, 2009) seems to strengthen the negative feedback from receptor activity to the methylation system, and may be important for robustness of the chemotaxis pathway (M Kollmann, personal communication). Signal amplification and integration by receptor arrays As mentioned above, the chemotactic response in E. coli is extremely sensitive, with cells responding to as little as 10 nM steps in the concentration of aspartate (Mao et al, 2003), that is <10 molecules of aspartate in the 1.4-fl volume of an E. coli cell. An increase in receptor occupancy by just 0.2% has been estimated to result in a 23% change in the bias of motor rotation (Segall et al, 1986), indicating signal amplification (or gain) by a factor of ∼100. This amplification is achieved by highly cooperative behaviour of two macromolecular complexes that are involved in chemotaxis signalling: receptor arrays and flagellar motors. Interestingly, allosteric signal amplification by receptor arrays was initially proposed as a likely explanation of high sensitivity in chemotaxis based on purely theoretical considerations (Bray et al, 1998). Subsequent experimental studies confirmed the existence of cooperative interactions between receptors in vitro (Li and Weis, 2000; Lai et al, 2005) and in vivo (Gestwicki and Kiessling, 2002; Sourjik and Berg, 2004; Vaknin and Berg, 2008), and also showed that these interactions largely amplify chemotactic signals in vivo (Sourjik and Berg, 2002b, 2004). These findings have provided a basis for the development of a number of detailed mathematical models describing cooperative behaviour of receptor arrays (Duke et al, 2001; Sourjik and Berg, 2004; Mello and Tu, 2005; Keymer et al, 2006), in which receptors in arrays switch cooperatively between active and inactive states, so that the activity of each individual receptor depends not only on the presence of a bound ligand molecule and on the level of receptor methylation, but also on the activities of neighbouring receptors. Most commonly used are the Monod–Wyman–Changeux-type (MWC) models, which further assume that a cluster consists of many independent tightly coupled cooperative units, or signalling teams, of 10–20 receptor dimers (Figure 2C), with all receptors in one unit switching synchronously between inactive and active states (Sourjik and Berg, 2004; Mello and Tu, 2005; Keymer et al, 2006). Other, Ising-type, models describe the cluster as one lattice in which the activities of adjacent receptors are coupled with a finite strength of interaction, leading to an effective distance of the conformational spread (Duke and Bray, 1999; Mello and Tu, 2003). The number of effectively interacting receptors corresponds to the size of the cooperative unit in the case of the MWC model and to the extent of the conformational spread in the Ising model. It is likely that real receptor arrays may be better described by a hybrid model, in which groups of tightly coupled receptors are more loosely coupled to one another, consistent with the experimental analyses of the array structure (Khursigara et al, 2008). In the MWC and Ising models, cooperative interactions between receptors in clusters can largely amplify changes in the kinase activity relative to ligand binding. If receptors have similar probability of being active or inactive, then additional inactivation of only a few receptors can stabilize the entire array in the inactive state. Moreover, as receptor homodimers with different ligand specificities appear to be randomly mixed in clusters (Studdert and Parkinson, 2004), cooperative interactions between them can ensure integration of different chemotactic signals. In the simplest case, the output of such a mixed array simply depends upon the total number of receptors occupied by ligand, allowing cells to regulate their chemoeffector preference by adjusting the numbers of receptors of particular type. Dynamics of signalling arrays Clustered organization of signalling proteins allows an additional level of regulation, through the composition and structure of the complex. Although the chemoreceptor arrays as a whole are stable on the time scale of signalling, local stimulation-induced rearrangements can occur, and the degree of cluster compactness indeed appears to be modulated by ligand stimula