The data presented in this study demonstrate the dynamic nature of chemosensory arrays. We find that ligand binding loosens the overall packing of the receptors (Figs 1 and 2) and correspondingly decreases the effective cooperativity of the kinase responses (Figs 4 and 5). The emerging picture is summarized in Fig. 6. Time-lapse fluorescence imaging demonstrated that these changes do not, in most cases, lead to dissociation of the receptor arrays (Fig. 3).
Ligand binding changes the packing in receptor arrays
The observation that addition of α-methylaspartate affects the Tar clusters but not the Tsr clusters (Fig. S3) indicates that the effect of α-methylaspartate on the Tar receptors is most likely caused by its known binding interaction with the Tar receptor. Indeed, the concentration of α-methylaspartate that triggers the modulation of the receptor clusters (Figs 1 and 2) was well within the dynamic range of the normal kinase responses and, as previously documented (Mesibov and Adler, 1972), was approximately 10-fold higher than the corresponding aspartate concentration, presumably because of the lower affinity of α-methylaspartate to the Tar receptor. Thus, the slow modulation of Tar clusters is most likely triggered by the primary conformational response of the Tar receptor to ligand binding.
Several lines of evidence indicate that the dynamics of ligand binding do not play a dominant role in the observed anisotropy response. First, and most importantly, if the dynamics of ligand binding were important, the observed changes in fluorescence anisotropy would strongly depend on the concentration of the added ligand, with faster dynamics expected as the ligand concentration increased. However, the observed anisotropy response was insensitive to an increase in the concentration of the ligand to more than 10-fold higher than the saturation value (Fig. 2). Second, despite the fact that the ability of the cell to metabolize and internalize α-methylaspartate is very different than that for aspartate (Kay, 1971; Mesibov and Adler, 1972), both ligands triggered very similar dynamics (Fig. S5). Finally, the rate of transport of α-methylaspartate into the cytoplasm is expected to be very different from the rate of its secretion out of the cytoplasm. Thus, if the binding dynamics of the ligand were dominant, strongly asymmetrical response would be expected upon the addition or removal of the ligand. However, the dynamics were generally symmetrical (Fig. 2). We therefore conclude that the dynamics of the anisotropy responses following a stimulus represent the physical response of the receptor clusters to ligand binding.
Because of limitations in the stimulation time (Fig. S1), the data shown in Figs 1 and 2B do not allow us to separate the initial anisotropy change, which could potentially reflect the primary conformational response of the receptors to ligand binding (Vaknin and Berg, 2007; Sferdean et al., 2012), from the subsequent, more gradual increase in anisotropy. The gradual component of the response most likely reflects more complex molecular processes within the clusters. The non-exponential dynamics observed in Fig. 2 are consistent with such complex molecular dynamics. Given that the baseline anisotropy measured for receptors clustered with CheA/W is significantly lower (homo-FRET higher) than that measured in the absence of CheA/W (Fig. 1), it is expected that changes in receptor packing would affect the measured anisotropy by modulating the efficiency of inter-trimer, and possibly also intra-trimer, homo-FRET. To account for the increase in anisotropy following a stimulus, the cluster modulation ought to involve an overall decrease in the packing efficiency of the receptors. This scenario is generally consistent with the effect of the receptors modification state on their packing efficiency in vesicles (Besschetnova et al., 2008).
Receptor trimers of dimers tend to form extended receptor arrays supported by rings of CheA and CheW proteins that bind to the cytoplasmic tips of receptor trimers and induce a basic hexagonal order (Briegel et al., 2012; Liu et al., 2012). Upon ligand binding, a conformational change is propagating along the receptor to induce a conformational change in the receptor trimer tip, which ultimately induces a conformational change in the associated kinase. It is therefore conceivable that stimulus-induced conformational changes in the trimer tip, CheA and CheW junction might also have a subtle effect on the stability of the CheA/W rings, and thereby on receptor packing. Such an effect might be related, for example, to changes in the flexibility of the receptor trimer/CheA/CheW junctions (Kim et al., 2002; Swain et al., 2009; Parkinson, 2010). This scenario is consistent with the enhanced thermal stability of the receptors upon clustering (Frank et al., 2011) and with the observation that the dynamics of CheA association with the array can be altered by stimuli (Schulmeister et al., 2011).
The basic lattice constant of the receptor hexagonal array appears to be insensitive to stimuli (Khursigara et al., 2008; Briegel et al., 2011). Thus, it seems more likely that changes in the trimer junctions affect more global features of the array. In principle, changes in the trimer junctions can affect the long-range order of the receptor arrays and thus the overall packing of the receptors. Structural variations in receptor arrays might include molecular variations within each cluster, which can either be random or reflect changes between the core of the cluster and its edges. In particular, in addition to the primary CheA/W-mediated clustering, which fits the hexagonal order, direct associations between receptors have also been reported (Kentner et al., 2006). To the extent that such associations also occur in the presence of CheA/W, they have the potential to cause local distortions in the receptor array. Also, as mentioned above, fluorescence recovery after photobleaching (FRAP) experiments suggest that the association of the scaffold protein CheA with the array is dynamic and depends on ligand binding (Schulmeister et al., 2011). Such dynamics would imply a constant introduction of ‘defects’ to the array structure that could lead to local distortion of the array. Recent cryo-EM tomography studies have reported varying levels of disorder in receptor arrays that correlated with their signal-transduction properties (Khursigara et al., 2011). In addition, variability in the size of the arrays or in the stoichiometry of different components within the arrays might be expected both between different clusters within each cell or between cells. High-resolution fluorescence-imaging studies suggest that a continuous distribution of cluster sizes exist within each cell, an observation that is consistent with a dynamic model for cluster assembly (Greenfield et al., 2009).
Consistent with previous studies (Homma et al., 2004; Liberman et al., 2004; Vaknin and Berg, 2004; Studdert and Parkinson, 2005; Erbse and Falke, 2009; Briegel et al., 2011; Khursigara et al., 2011), we find that the overall integrity of the clusters is generally preserved in the presence of a stimulus (Fig. 3). Changes in the global clustering pattern was found in only a small fraction of the cells, which exhibited low-contrast localization prior to stimulation (Fig. 3B), possibly reflecting unfavourable conditions for clustering in these cells. The changes observed in the immunolabelling patterns of chemoreceptors upon stimulus (Borrok et al., 2008) might reflect changes in either receptor packing or overall clustering. More complex behaviour was observed in B. subtilis cells (Wu et al., 2011).
Ligand binding induces changes in signal transduction
In addition to ligand-induced modifications in the packing of receptor arrays, we find evidence for corresponding ligand-induced evolution in the intrinsic signal-transduction properties of these arrays. In general, we find that receptor-stimulated kinase activity depends not only on the current ligand concentration but also on the preceding equilibrium ligand concentration (Fig. 4). Moreover, the effect of prior exposure to a ligand is clearly dynamic and becomes more dominant as the time of exposure increases (Fig. 4A and C, lower plot). The timescale of this modulation of kinase activity is comparable to that of the physical modulation of clusters. These stimulus-induced changes in kinase activity can be viewed as a slow modulation in the effective cooperativity of the receptor-mediated response (Fig. 5). Such a correlation between the cooperativity of the kinase responses and the packing efficiency of the receptors is predicted by an understanding that the kinase cooperativity is, at least in large part, caused by receptor packing within clusters (Vaknin and Berg, 2007; Amin and Hazelbauer, 2010; Khursigara et al., 2011). Thus, the data presented in Figs 4 and 5 suggest that the intrinsic signal-transduction properties of the cluster, i.e. the dose–response function, can be dynamically modulated by the ligand. This point is depicted schematically in Fig. 6.
Potential implications of cluster dynamics
How can the dynamics of clusters affect signalling under native conditions? We note first that the cluster modulation observed in this study occurs over timescales that are comparable to those measured for receptor methylation (Sourjik and Berg, 2002; Lazova et al., 2011). Because the change in receptor packing is coupled to the primary change in receptor conformation, we expect that receptor packing will also be coupled to the adaptation process of the receptors. Moreover, because the adaptation of each receptor dimer also depends on neighbouring receptors (Li and Hazelbauer, 2005), changes in the packing of the cluster can also affect the intrinsic rates of adaptation. Other factors, such as the accessibility of the adaptation proteins, can also be considered in this context. If the overall effects of clustering on the rates of methylation and demethylation are precisely the same, changes in receptor packing can alter the dynamics of adaptation but not the final equilibrium level of kinase activity. However, if the effects of clustering on the rates of methylation and demethylation are not precisely the same, changes in receptor packing can alter the final steady-state level of kinase activity, leading to deviations from exact adaptation. Such deviations from exact adaptation have been observed in cells that express Tar as the sole chemoreceptor (Meir et al., 2010).
Given the observed changes in the cooperativity of the kinase responses, it is plausible that cluster dynamics could also alter the coupling between receptors of different types. Integration of different signals by the chemoreceptor clusters is an important aspect of chemosensing. Dynamic coupling has recently been invoked in models of mixed clusters to explain the lack of coupling in the steady-state adaptation level of different receptor types (Hansen et al., 2010). In fact, the transient coupling in adaptation observed in such mixed clusters (Lan et al., 2011) occurs over timescales that are comparable to those observed in this study for cluster modulation.
Finally, it has been suggested that Aer-mediated aerotaxis does not involve methylation of Aer and thus does not require the adaptation enzymes (Bibikov et al., 2004; Gosink et al., 2006). Such taxis behaviour could potentially rely on an alternative adaptation mechanism or a process that does not require receptor adaptation (Mazzag et al., 2003). Under such condition in which the primary adaptation mechanism is not effective, dynamical changes in receptor arrays might be a way to modify the transduction properties of the arrays according to external stimuli.