Bacterial TCSs play crucial roles in environmental adaptation. Although many model TCSs are well established, how their activities are controlled by stimuli has not, in general, been rigorously investigated, primarily due to the profound lack of identified ligands and the difficulties encountered in biochemical and structural studies of membrane spanning proteins (Lee et al., 1999
; Timmen et al., 2006
; Cheung and Hendrickson, 2010
). In addition, TCSs containing many components are difficult to reconstitute in vitro
. Even the paradigmatic bacterial chemotaxis TCS has not been examined in a completely reconstituted system in vitro
. This is due to the fact that the membrane bound chemotaxis receptors (e.g. Tar) are not histidine kinases; rather, these receptors regulate the activity of the soluble histidine kinase CheA (Hazelbauer and Lai, 2010
). Second, the chemotaxis system relies on the specific organization of the receptors in arrays (Webre et al., 2004
). Third, chemotaxis signalling is subject to multiple types of regulation such as methylation of the receptor by CheR, demethylation by CheB, and dephosphorylation of the response regulator CheY by the phosphatase CheZ (Ninfa et al., 1991
). In V. cholerae
, by contrast, QS relies on a three protein phosphorelay system to detect the major autoinducer CAI-1 (Miller et al., 2002
). Moreover, the structure of the ligand CAI-1 is known, and CAI-1 and analogues are readily available (Kelly et al., 2009
; Ng et al., 2010
; Bolitho et al., 2011
; Wei et al., 2011
). Finally, a panel of CqsS ‘gatekeeper’ mutants, with altered ligand specificities, has been engineered (Ng et al., 2010
). Thus, the CqsS system is tractable to mechanistic dissection in vitro
. Using inverted E. coli
membranes containing the full-length CqsS receptor together with purified downstream protein components, we were able to reconstitute the V. cholerae
QS phosphorylation pathway in vitro
. The complete phosphorylation cascade was established from CqsS, the hybrid sensor histidine kinase, through LuxU, the HPt protein, to LuxO, the response regulator. Our set of sensing mutants and ligand analogues allowed us to discover that ligand-receptor specificities are maintained in vitro
, and that CAI-1 binding regulates His194 auto-phosphorylation, whereas phosphotransfer and phosphatase activity are not affected by ligand binding.
It is interesting that QS relies on a phosphorelay system rather than a simple TCS system. We speculate that having the HPt protein LuxU provides a hub for signal integration. Indeed, V. cholerae
QS relies on two parallel sensory pathways. In addition to the CAI-1/CqsS pathway, input from the AI-2/LuxPQ sensory pathway converges with that from CAI-1/CqsS at LuxU (Miller et al., 2002
). Likewise, in the related bacterium V. harveyi
, information from three hybrid sensor histidine kinases, CqsS, LuxPQ and LuxN all converge at V. harveyi
LuxU (Henke and Bassler, 2004
). The particular domain architecture of the Vibrio
QS pathway ensures that the enzymatic activities required for phosphotransfer to and dephosphorylation of LuxU are colocalized within the same protein (CqsS, LuxQ and LuxN). Therefore, modulation of receptor levels potentially regulates phospho-flow in both directions. Feed-back regulation of QS receptor levels has been demonstrated in V. harveyi
(Teng et al., 2011
). We speculate that V. cholerae
CqsS could also be subject to similar feed-back regulation.
Proteins like CqsS that contain both the H1 and D1 modules are common in bacteria. Some other bacterial hybrid histidine kinases have domain arrangements similar to ArcB and BvgS, which contain the H1, D1 and the H2 domains (Iuchi et al., 1990
; Uhl and Miller, 1996
). ArcB responds to the redox state of the membrane and anaerobic metabolites, while BvgS responds to temperature, SO42–
, and nicotinic acid. CqsS, interestingly, resembles the domain architectures of sensory proteins present in eukaryotic organisms. In Arabidopsis thaliana
, over 10 sensor histidine kinases have been identified, all of which contain the histidine kinase domain (H1) and an attached receiver domain (D1) (Mizuno, 2005
). While most of these sensor histidine kinases have unknown functions, interestingly, in A. thaliana
, detection of ethylene, a plant hormone that controls growth and development, relies on the hybrid sensor histidine kinase ETR1 that resembles CqsS (Voet-van-Vormizeele and Groth, 2008
; Kim et al., 2011
). Response to another A. thaliana
master growth hormone, cytokinin, requires a phosphorelay system consisting of several hybrid sensor histidine kinases CRE1, AHK2 and AHK3 (H1, D1), multiple HPt proteins called APHs (H2), and multiple response regulators called ARRs (D2) (Inoue et al., 2001
; Sheen, 2002
). It is striking to us that ‘hormone-like’ signalling molecules in both bacteria and plants are detected by similarly arranged sensory systems. We wonder if a fundamental advantage of these systems is in signal integration. HPt proteins linking sensor histidine kinases and response regulators might serve as ideal merging points for inputs from multiple receptors.
Our work shows that while histidine auto-phosphorylation is inhibited by CAI-1, the CqsS phosphatase activity is not regulated by the ligand. Having a constant CqsS phosphatase activity could be crucial for properly timed QS transitions. During the low cell density to high cell density transition, although CAI-1 inhibition of the CqsS kinase ensures that little new LuxO~P is generated, existing LuxO~P could still activate transcription, due to its high stability (Fig. S2
). However, because CqsS possesses a constant phosphatase activity, it ensures rapid dephosphorylation of any existing LuxO~P. During the high cell density to low cell density transition, the constant CqsS phosphatase activity would not significantly affect LuxO~P generation due to the much stronger CqsS kinase activity. Thus, rapid phosphorylation of LuxU and LuxO occurs when CAI-1 disappears.
His194 auto-phosphorylation is inhibited by CAI-1. This result is consistent with previously proposed models suggesting that TCS binding of ligands triggers downstream conformational changes in sensor histidine kinases that affect interactions between the catalytic domains and the histidines in the DHp domains (Borkovich and Simon, 1990
; Neiditch et al., 2006
). We verified this idea by showing that dephosphorylation of His194~P by ADP is slower in the presence of CAI-1, suggesting that binding of CAI-1 in the transmembrane domain causes a conformational change that repositions the His194 on the DHp domain away from the catalytic domain. Therefore, in the presence of CAI-1, new rounds of phosphorylation are inhibited. Additionally, already phosphorylated His194 also becomes less accessible to the catalytic domain, resulting in its higher stability compared with in the absence of the ligand.
Our previous structural study of an analogous QS receptor, LuxPQ, suggests that upon ligand binding, a symmetry-breaking conformational change occurs in the periplasm that is transduced across the membrane to alter the relative positions of the histidine in the DHp domain and the catalytic domain active site (Neiditch et al., 2006
). However, in the LuxPQ case, two proteins are required for transducing the AI-2 autoinducer signal. LuxP, the periplasmic autoinducer binding protein, binds AI-2 and induces the conformational change in the partner hybrid sensor kinase LuxQ through protein–protein interactions (Neiditch et al., 2005
). Extensive studies with CheA have led to a similar model (Borkovich and Simon, 1990
). In the CheA case, the chemotaxis receptor Tar is membrane associated and it regulates the cytoplasmic kinase CheA. When the ligand binds to the Tar receptor, it is proposed to induce a ‘closed’ confirmation of the CheA kinase. It is curious to us that three distinct TCSs with three different modular domain architectures use a related mechanism for regulating kinase activity. Evolution has produced a variety of domain arrangements for histidine kinases in TCSs, yet; at least in those cases examined, the underlying regulation mechanism seems to be conserved.
In our current study, P-CAI-1, a CqsS antagonist, also antagonizes CqsS in the presence of CAI-1 in vitro. In addition, C8-CAI-1, a weak agonist of CqsS, also acts as an antagonist of CqsS in the presence of CAI-1 in vitro. This result is consistent with our two-state model (see Results). Competition between a weak agonist and a strong agonist for the ligand binding pocket results in an intermediate effect rather than a synergistic effect. Antagonism can also be replicated in vitro with the CqsSC170Y mutant receptor, which possesses altered specificity for ligands. In vitro, C8-CAI-1 is a strong agonist and CAI-1, which does not affect CqsSC170Y when acting alone, is an antagonist of the CqsSC170Y receptor. Therefore, CAI-1 likely binds and stabilizes the ‘kinase on’ and ‘kinase off’ states of CqsSC170Y receptor with equal preference, and thus appears neutral in the absence of other molecules. Alternatively, to function as an antagonist, CAI-1 could prefer to bind and stabilize the ‘kinase on’ state of CqsSC170Y. In either case, CAI-1 certainly competes for the binding site with C8-CAI-1 and therefore acts functionally as an antagonist. Thus, predictions of the two-state theoretical model are consistent with the results of the experiments using our in vitro phosphorylation system.
Bacteria exist in niches containing complex microbial-species compositions. Thus, the chemical environments they encounter contain numerous signalling cues, including many classes of autoinducers, the concentrations of which change in time and in space. In a simplified system, we have examined receptor histidine kinase regulation by a set of agonists and antagonists. To the best of our knowledge, this study shows, for the first time, that histidine auto-phosphorylation of the sensor is the only step that is regulated by the ligands. Structural studies of apo-CqsS and the CqsS receptor bound to agonists and antagonists should reveal the conformational changes induced by these ligands. Additionally, future work with reconstituted circuits that contain both the CqsS and LuxPQ sensors will be useful to understand QS signal integration.