Protein–protein interfaces in chemotaxis receptor complexes and their role in signaling

Abstract

Chemotaxis receptor complexes sense chemical gradient in the cellular environment to direct swimming towards favorable environments. The core signaling units of these complexes are made up of two trimers-of-dimers of chemoreceptors, two CheW and a CheA dimer, which further assemble into large hexagonal signaling arrays. Structural and biochemical studies have provided important information on the architecture and interfaces of these complexes. However, the signaling pathway of these complexes that controls the kinase is not fully understood. In this review, we highlight the highest resolution models of this system and examine the current consensus on the protein–protein interfaces based on models and interface experiments. We also highlight differences observed between signaling states for the individual proteins and the protein interfaces that are proposed to be part of the signaling mechanism. Overall, we conclude that there is strong structural consensus for the protein interfaces but, despite some intriguing results, more information is needed to understand how the interfaces change between signaling states and the role they play in signaling. An animated Interactive 3D Complement (I3DC) is available in Proteopedia at https://proteopedia.org/w/Journal:FEMS_Microbiology_Reviews:1.

There is strong consensus on the architecture and protein interfaces of chemotaxis receptor complexes, but changes in the protein interfaces between signaling states need to be further investigated to fully understand the signaling mechanism of the complexes.

Introduction

Chemotaxis is the process by which bacteria sense their environment and alter their swimming pattern in response to attractants and repellants. For a number of human diseases, chemotaxis is important to the initial infection of the host and the survival of the bacteria during infection (Matilla and Krell 2018). Since two-component signaling systems like bacterial chemotaxis are not found in mammals, chemotaxis proteins are an attractive target for novel antibiotics. Thus, it is important to understand the signaling mechanisms and protein–protein interactions of the chemotaxis signaling complexes as a foundation for the potential development of such therapies.

Chemoreceptor signaling complexes bind chemical ligands in the environment and initiate a signaling cascade that controls flagellar rotation and alters swimming direction (Fig. 1). The core signaling unit (CSU, enclosed in dashed red box in Fig. 1B) consists of two trimers-of-dimers (TODs) of chemoreceptors with a CheA dimer and two CheW. These assemble into large membrane-bound hexagonal arrays that predominantly localize to the poles of the cell (Zhang et al. 2007). In Escherichia coli, attractant molecule binding to the chemoreceptor leads to deactivation of the auto-kinase, CheA, decreasing phosphoryl transfer to the response regulator, CheY, which leads to counter-clockwise rotation of the flagella. The CCW-rotating flagella form a bundle that propels the cell forward in a “run” towards attractants. In the absence of attractant ligand, activation of the kinase leads to clockwise rotation of the flagella, disrupting the flagellar bundle and causing the cell to move in a random “tumble” motion. Different bacteria have different receptors with specificities for sensing the nutrients and toxins needed to enhance survival in their specific niche. This review will focus primarily on the structure and mechanism of E. coli chemoreceptor signaling complexes, the most well-studied of these systems, with some discussion of analogous systems from other bacterial species.

Figure 1.

Overview of chemotaxis signaling complexes and arrays. The kinase activity of CheA is regulated by both ligand binding to the receptor periplasmic domain and the methylation state of 4 Glu residues in the cytoplasmic domain. (A) A dimer of the E. coli aspartate chemoreceptor is shown from the PDB: 8C5V model. Binding of ligand (black diamonds) and demethylation by CheB shift the complex to the kinase-off state. Ligand dissociation and methylation (magenta spheres) by CheR shift the complex to the kinase-on state. When CheA is activated, it auto-phosphorylates and then transfers phosphoryl groups to the response regulators CheY and CheB. (B) Signaling array viewed from below the membrane. Core signaling units (CSU, highlighted by red dashed box) assemble to form hexagonal arrays connected by A/W rings (ring of alternating blue and cyan ovals) and W-only rings (three cyan ovals represent half of a W-only ring). Filled symbols highlight three types of protein interfaces within a core signaling unit: receptor/CheW (filled dark green box), receptor/CheA-P5 (filled light green box), and CheW/CheA-P5 (filled yellow circle, interface 1). Open symbols highlight interfaces connecting the array of CSUs: CheW/CheA-P5 (open yellow circle, interface 2) and CheW/receptor (open dark green box).

The E. coli transmembrane chemoreceptors, which include Tar, Tsr, Aer, Tap, and Trg, bind specific ligands to the periplasmic ligand binding domain and transmit a signal across the membrane (other than Aer, the aerotaxis receptor that mediates swimming towards oxygen). The ligand binding signal is thought to begin as a ∼2 Å piston displacement of the helix alpha1-TM2 that extends from the ligand binding site to the cytoplasmic edge of the membrane (Falke and Hazelbauer 2001). This signal then travels 200 Å across the cytoplasmic domain to control the histidine auto-kinase, CheA, bound at the membrane-distal tip of the receptor. The cytoplasmic domain contains the HAMP domain, methylation helix bundle, flexible bundle and protein interaction region (PIR), as shown in Fig. 1A.

Methylation of receptors, at 4–5 glutamic acid residues in methylation helices 1 and 2 of the cytoplasmic domain, mediates adaptation to an ongoing stimulus: the extent of methylation serves as a “memory” of the recent attractant concentration and enables bacteria to direct their movement up a gradient of attractant molecules. Ligand binding suppresses tumbling; then adaptation (methylation) restores the basal tumbling rate. Methylation of the chemoreceptor activates the associated kinase, while demethylation deactivates the kinase. Due to this methylation, chemoreceptors are also known as methyl accepting chemotaxis proteins (MCPs). The methyltransferase, CheR, and methyl esterase, CheB, regulate the level of methylation of the receptor (Parkinson et al. 2015). The effects of ligand binding and methylation on signaling are illustrated in Fig. 1A. A major question in the field is how the signals from ligand binding and methylation are transmitted through the cytoplasmic domain and what changes occur in the PIR, where the receptor interacts with both CheA and CheW, to control CheA. Current proposals for differences between kinase-off and kinase-on signaling states include opposing changes in dynamics in different regions, either HAMP/methylation/PIR (Swain et al. 2009, Zhou et al. 2009, Reyes et al. 2024) or HAMP/PIR (Samanta et al. 2015). Other studies report widespread stabilization of the methylation through PIR regions of the cytoplasmic domain (Bartelli and Hazelbauer 2016, Kashefi and Thompson 2017, Li et al. 2019, Malik et al. 2020). Conformational changes, such as rotation of the receptor helices, are also proposed to influence the signaling state of the complexes (Gao et al. 2019, Flack and Parkinson 2022). However, observation of signaling-related changes in interactions of the cytoplasmic domain with CheA and CheW has been limited. Knowledge of changes in interactions is needed to understand how the signal is transmitted to control kinase activity.

CheA is a histidine auto-kinase made up of five domains that play roles in autophosphorylation, phospho-transfer, and regulation. The phosphorylation domain (P1) contains the histidine that undergoes autophosphorylation and phosphorylates response regulators, CheY and CheB, that bind to the docking domain (P2). The dimerization domain (P3) stabilizes the dimer that is structurally important for formation of core signaling units and mechanistically important for the trans autophosphorylation reaction. ATP binds to the catalytic domain (P4), which transfers the gamma phosphate from ATP to P1. Finally, the regulatory domain (P5) is the main domain to interact with the receptor (light green box in Fig. 1B) and with CheW (filled yellow circle in Fig. 1B). Figure. S1 shows the domains of the kinase core of CheA (P3, P4, and P5 domains), which will be shown in the models throughout this review.

The coupling protein, CheW, couples CheA to chemoreceptor control in CSUs and links multiple CSUs into hexagonal arrays (Liu and Parkinson 1989, Gegner et al. 1992, Ortega et al. 2013). CheW, a structural homolog of the CheA-P5 regulatory domain, interacts with CheA-P5 and with the membrane-distal tip of the receptor, the PIR. CheW forms CheA/CheW rings by binding two CheA-P5 domains, one within the same CSU [interface 1, filled yellow circle in Fig. 1B, as first reported in the crystal structure PDB:2CH4 (Park et al. 2006)] and one with CheA-P5 of another CSU (interface 2, open yellow circle in Fig. 1B). The A/W rings of the signaling arrays are presumably involved in transfer of the signal between neighboring CSUs. CheW-only rings also form within the arrays and increase the cooperativity of signaling (Piñas et al. 2022). Thus, CheW is involved in CheW/CheW interactions as well as CheW/receptor interactions (open green box in Fig. 1B) outside of the CSU. CheW has been proposed to play a role in control of the kinase, but there is limited evidence for a specific role in signal transduction. The variety of protein interfaces, both within the CSU and to neighboring proteins, are highlighted for one CSU in Fig. 1B.

The structure of individual chemotaxis proteins and complexes from different bacterial species were solved using crystallography and nuclear magnetic resonance (NMR) (Bilwes et al. 1999, Kim et al. 1999, Griswold et al. 2002, Park et al. 2006, Li et al. 2007, Briegel et al. 2012, 2013). Many of the x-ray crystallography and NMR structural studies are on proteins from the hyperthermophile T. maritima due to better solubility and thermal stability of the proteins. However, E. coli and S. typhimurium proteins were used in some structural and many mechanistic studies. There is significant sequence homology between E. coli, T. maritima, and S. typhimurium proteins and no current evidence that the structures differ. Cryo-electron tomography (cryo-ET) studies visualized the organization of the arrays at the poles of cells with the receptors inserted in the membrane and the CheA/CheW baseplate in the cytoplasm (Zhang et al. 2007, Briegel et al. 2009). Since then, cryo-ET has emerged as the main technique for studying full structures of functional CSUs and arrays. A variety of additional studies have provided information on the specific residues critical to the different interfaces.

In this review we summarize structural and mechanistic insights gained from models based on cryo-ET and crystallographic studies. We also compile the results from a multitude of interface studies to determine whether there is consensus on which residues are involved in interactions. Finally, we compare the interfaces of the receptor with the structurally homologous CheW and CheA-P5. Despite improvements in structural models, they still lack the resolution to reveal subtle changes at the protein interfaces that may affect signaling. There is some experimental evidence for differences at the interfaces between signaling states, but more knowledge is needed to fully understand the role of these interfaces in the signaling mechanism.

Note that the models and information presented throughout this review can be explored in more detail and manipulated in three dimensions in Proteopedia at https://proteopedia.org/w/Journal:FEMS_Microbiology_Reviews:1.

Structural and mechanistic information gained from crystallography and cryo-ET based models

Crystal structures for individual proteins and ternary complexes are available, but do not represent functional complexes in defined signaling states. Cryo-ET was performed on samples ranging from native arrays with full length receptors in lysed E. coli cells [resolution ∼20 Å (Yang et al. 2019) and ∼12 Å (PDB: 8C5V)(Cassidy et al. 2023)] and minicells (Liu et al. 2012, Burt et al. 2020, 2021) to arrays of receptor cytoplasmic fragments reconstituted on lipid monolayers (resolution ∼13 and 8 Å, PDB: 3JA6 and 6S1K, respectively)(Cassidy et al. 2015, 2020). The difference in resolution between these cryo-ET studies highlights the tradeoff between resolution and sample thickness, as discussed in a recent review (Hadjidemetriou et al. 2022). Models of CSUs made using combinations of crystal structures and cryo-ET provide information on protein–protein interfaces, structural mobility of the proteins, and some mechanistic insights. The models are a good representation of the current knowledge of the structure of CSUs, the make-up of CheA/CheW rings, and the protein interfaces involved in both. However, the resolution of the cryo-ET data is thus far too low to resolve any structural changes between signaling states in the baseplate region where the receptor, CheA-P5, and CheW interact. A critical step for understanding the signaling mechanism is to determine what changes occur at the protein interfaces of the complex to convey the information from the receptor (its ligand occupancy and methylation state) to regulate kinase activity.

The highest resolution model of a CSU was made by fitting crystal structures to cryo-ET densities of in vitro arrays reconstituted with the cytoplasmic domain of the chemoreceptor. The in vitro arrays consist of a cytoplasmic fragment (CF) of the E. coli aspartate chemoreceptor (Tar), CheW, and CheA assembled onto a lipid monolayer. MD simulations optimized the conformations of protein crystal structures from T. maritima [PDB: 2CH7 (Park et al. 2006), 1B3Q (Bilwes et al. 1999), 4XIV (Greenswag et al. 2015), and 4JPB] and E. coli [PDB: 1QU7 (Kim et al. 1999)] to best fit the cryo-ET density maps, resulting in an all-atom model of the CSU with a resolution of 8.4 Å (PDB: 6S1K) (Cassidy et al. 2020). This improved the resolution over the previous model, which had a resolution of 11.3 Å (PDB: 3JA6) for a similar sample (Cassidy et al. 2015). The higher resolution density map includes more detailed structural information for each protein, such as visualization of the helices in the receptor TODs and the beta structures of CheA and CheW. Since it has the highest resolution of all the cryo-ET based models, 6S1K is the main model of the CSU shown in this review.

Although many of the crystal structures used to fit the cryo-ET densities are of individual proteins and protein fragments, two crystal structures of ternary complexes of T. maritima proteins are foundational to current models of the CSU. The first is a 4.5 Å crystal structure (PDB 3UR1) of ternary complexes of the MCP fragment Tm14s, CheAΔ354 (P4-P5 only), and CheW that shows potentially native contacts between CheW and the membrane-distal tip of the receptor, but CheA-P5 contacts the wrong end of the receptor (Briegel et al. 2012). Cryo-ET conducted in this study provides additional information on the CheA/CheW rings observed in the crystals, yielding a structural model of the array. A similar architecture for the arrays was deduced by another group from their cryo-ET data on E coli minicells (Liu et al. 2012). A second crystal structure (PDB: 4JPB) of a ternary complex of similar T. maritima proteins (Li et al. 2013) has an improved resolution of 3.2 Å and includes contacts between CheA-P5 and a chemoreceptor fragment with an “unzipped” structure (Li et al. 2013). A companion biochemical study in native arrays used disulfide mapping and mutations to demonstrate that the native CheA-P5 to receptor interface is consistent with the contacts observed in the Li et al. model (Piasta et al. 2013). The arrangement of proteins in these crystals is only partly consistent with the architecture found in native arrays. Both crystals contain a non-native receptor contact with CheW or CheA, and both include non-native structural elements: 3UR1 has receptors extending from the CheA/CheW rings in alternating directions and 4JPB has an “unzipped” structure of the receptor that formed tetramers instead of dimers, resulting in some non-native contacts. Despite these issues, the crystal structures contain valuable information on the protein arrangements and contacts in the ternary complexes that have been used to fit cryo-ET densities and to select residues to test in protein interaction studies.

The only model of CSUs that includes full-length receptors was created based on cryo-ET density maps of arrays from phage E-protein lysed E. coli cells. This study achieved a resolution of 12 Å with sub-tomogram averaging (Cassidy et al. 2023). AlphaFold2 was used to predict the structures of each of the E. coli CSU proteins to fit to the density maps. It should be noted that AlphaFold2 is trained on existing crystal structures of proteins, which includes all the structures that were used in the previous models. The resulting model (PDB 8C5V) includes full-length E. coli serine chemoreceptor, Tsr (Cassidy et al. 2023). Additional cryo-ET studies with molecular modeling that have not resulted in PDB-deposited models also provide valuable information on the structure of the CSUs and arrays, as discussed in other reviews (Cassidy et al. 2018, Hadjidemetriou et al. 2022, Riechmann and Zhang 2023).

The cytoplasmic fragments of the receptor and full-length receptors have been used interchangeably in both structural and mechanistic studies of the arrays. Figure 2 shows a comparison of the protein interface residues of the highest resolution models of E. coli CSUs with the CF (PDB 6S1K) and full-length receptor (PDB 8C5V). Interface residues were defined as residues with any atom within 4 Å of the other protein, as identified with PyMOL; Fig. 2 shows the 6S1K model, with spheres representing the alpha carbons of interface residues. The green spheres, signifying residues within 4 Å in both models, show that the core of the protein interfaces in these models agree, indicating the complexes formed by just the CF and full-length receptor have very similar architecture. The differences in the interfaces (yellow and orange in Fig. 2) are mainly at the edges of the interaction surfaces, which may be due to the difference in resolution or due to fitting the density with crystal structures (6S1K) compared to fitting with structures predicted by AlphaFold2 (8C5V). The similarities and differences in the residues at these interfaces in 6S1K and 8C5V are also listed in Table 1. The overall agreement between the protein interaction surfaces demonstrates that samples made with the CF represent the cytoplasmic part of the CSU structure seen in native complexes. However, mechanistic studies of samples containing only portions of the cytoplasmic domain cannot investigate the effects of periplasmic ligand binding; instead signaling state is controlled by mutating the adaptation sites to the unmethylated (E) or methyl-mimic (Q) residue. Table S1 lists the residues found at the interfaces in all the models of ternary complexes, to highlight the residues with the most structural evidence for being at the interfaces.

Figure 2.

Overlap of interface residues predicted by 6S1K and 8C5V models. All predicted residues for each interface (residues with any atom within 4 Å of any atom in residues of the other protein in the 6S1K model) are highlighted by alpha carbon spheres on the 6S1K model. In each pair of panels, the structure is rotated to highlight an interface, and colored to indicate which predictions are in common: green represents residues predicted to be at interfaces by both models, yellow shows residues predicted by only 6S1K, and orange shows residues predicted by only 8C5V. MCP/CheA interface residues are colored on (A) MCP and (B) CheA-P5. MCP/CheW interface residues are colored on (C) MCP and (D) CheW. CheA/CheW interface residues on are colored (E) CheA and (F) CheW. Only the receptor monomer, CheW, or CheA-P5 domain participating in the interaction are shown in cartoon representation, with the rest shown as transparent surfaces. Residues highlighted in this figure are listed in Table 1.

Table 1.

Comparison of interface residues,¹ found within 4 Å of other protein in PDB files 6S1K vs 8C5V.

	MCP/CheA		MCP/CheW		CheA/CheW
	MCP	CheA	MCP	CheW	CheA	CheW
Both (green spheres, Fig. 2)	Q374	M515	I377	I33	M287	T46
	I377	L528	L380	V36	F568	A49
	L380	V531	N381	I39	N569	N50
	N381	M532	V384	V105	S612	T51
	V384	S534	R388	S106	N613	I55
	R388	Q536	R404	D107	Y614	V58
		I581			R615	N60
		L599			I625	G63
					G629	I65
					V631	P67
						R96
						E155
6S1K only (yellow spheres, Fig. 2)	A383	Q602	A387	Q37	V570	I48
				V108	K616	M156
					D628	L158
8C5V only (orange, Fig. 2)	F373	E533	Q374	L20	R291	R47
	L378	L535	E385	L34	V567	P52
	E385	R538		E38	A623	Q95
		Q579		N85	L633	V97
						V98

¹Residues with any atom within 4 Å of any atom in residues of the other protein. All residue numbers are for E. coli and MCP numbers are for E. coli Tsr.

Cryo-ET studies provide some evidence of changes between signaling states that may be involved in the signaling mechanism. Changes in the compactness of the TOD and degree of bend for the cytoplasmic domain of MCPs, specifically at the HAMP domain and glycine hinge, are seen upon ligand binding or change in methylation state (Khursigara et al. 2008, Yang et al. 2019). Change in the bend of the receptor has also been observed by crystallography and transmission electron microscopy (Pollard et al. 2009, Akkaladevi et al. 2018). The unmethylated state of the receptor has lower resolution in cryo-ET studies than the methylated-mimic, specifically in the cytoplasmic region just below the HAMP domain. Lower resolution indicates that the unmethylated receptor is less structurally homogeneous, which suggests that the unmethylated receptor is more dynamic and less well-structured in this region (Yang et al. 2019, Cassidy et al. 2020). These differences may reflect changes in the conformation and stability of the TODs, which could then impact the signaling state of the associated kinase. Observed changes in the conformation and interdomain interactions of CheA include a proposed “dipping” motion of the CheA-P4 domain observed in MD simulations (Cassidy et al. 2015, 2020, 2023) and changes in the “keel” density observed by cryo-ET (Briegel et al. 2013), which is thought to be the P1 and P2 domains. Both of these changes are proposed to play roles in the regulation of kinase activity. However, none of these studies resolved changes in the baseplate region of the CSU, where the receptor PIR interacts with CheW and CheA-P5. While the models have yielded a consensus on the overall structure of CSUs, other techniques are needed to confirm which residues play a role in protein interactions and how these interactions change between signaling states.

Biochemical and biophysical experiments, including mutagenesis, crosslinking, and NMR studies, have identified residues potentially involved in protein–protein interactions in the CSU. This evidence informed construction of the models, which in turn predicted additional interface residues to be tested by such experiments. NMR studies of pairs of T. maritima protein fragments investigated each binary interaction to provide insights into the structure of the different interfaces of the ternary complex. Crosslinking and mutagenesis studies also investigated the protein interactions in the context of functional complexes of proteins from E. coli, T. maritima, and S. Typhimurium. Evidence for interface locations of residues includes chemical shift perturbations, loss of complex formation in mutants, and high yields of disulfide bond formation.

Experimental evidence for protein interfaces in the hexagonal rings of the array

The membrane-bound chemotaxis signaling array is connected by two types of hexagonal rings of proteins. A/W rings are made by alternating CheA-P5 and CheW of three CSUs, with two types of CheA-P5/CheW interfaces. Interface 1 is within the CSU and interface 2 is between CheW and CheA-P5 of neighboring CSUs. These A/W rings are sufficient to connect the CSUs into an extended hexagonal array, but additional CheW-only rings, which connect six CSU’s, have been observed (Cassidy et al. 2015) and shown to increase the cooperativity of signaling (Piñas et al. 2022). The 6S1K model of a core signaling unit includes only interface 1. The additional interfaces, interface 2 of the A/W ring and the interfaces of the CheW-only ring, are represented in an expanded 6S1K model (Fig. 3B) that was created by using PyMOL to align 6S1K and three additional CheW to the p6_Ec_array.pdb model (which lacks sidechains) available at zenodo.org/records/4302473.

Figure 3.

Agreement between residues of A/W and W-only ring interfaces predicted by models and identified in protein–protein interaction studies. In panels A, C, and D, predicted interface residues (within 4 Å of other protein) are represented by an alpha carbon sphere on the expanded 6S1K model that was modified to include additional CheW (see text). (A) CheA residues in interface 1 and 2 of A/W ring. (B) Top-down view of one CSU with a connected A/W and W-only ring (part of p6_Ec_array.pdb). (C) CheW residues in interface 1 and 2 of A/W ring. (D) CheW interfaces in W-only ring, with different shades of cyan used to distinguish the two CheW. Green represents residues shown to be a part of interactions in protein–protein interaction studies, as listed in Tables S2, S3, S4. Residues at the interfaces in the model but not tested by other interface studies are shown in blue for CheA and cyan/light cyan for CheW. Only CheA-P5 and CheW involved in the interactions are shown as cartoons.

Residues involved in interface 1 between CheA-P5 and CheW within a CSU were identified by multiple studies. Cysteine cross-linking studies of in vitro reconstituted arrays were done on residues considered to be at the periphery of the CheA-P5/CheW interface based on the 4JPB model. As listed in Table S2, three S. typhimurium CheA-P5/CheW residue pairs have fast disulfide formation under mild oxidation conditions, D586C-N50C, S647C-Q44C, and S647C-T46C, indicating proximity to each other at the CheA-P5/CheW interface. In addition, TAM-IDS (tryptophan and alanine mutations to identify docking sites) of CheA identified residues N630, V634 and L650 to be a part of this interface due to loss of array incorporation (Natale et al. 2013). Fluorescence anisotropy binding assays also identified residues at the CheA/CheW interface. In one study, incorporation of a bulky probe at S. typhimurium CheA residues E539, L627, and G636 reduced CheW binding (Miller et al. 2006). Additional studies identified E. coli CheW residues V36, G41, G57, R62, G133 (Boukhvalova et al. 2002) and V45, T46, K56, L158 (Boukhvalova et al. 2002), where mutations decreased the binding affinity for CheA. Pull-down assays of CheA/CheW binding identified E. coli CheA-P5 residues likely involved in binding: incorporation of a bulky probe at residues V607, N609, R615, A622, L633, and I634 (Zhao and Parkinson 2006a), and mutations at residues K616, A622, G629, and V631 decreased binding (Zhao and Parkinson 2006b). Methyl-focused NMR chemical shift perturbation studies of a binary complex of truncated CheA (P3P4P5) and full-length CheW from T. maritima identified that CheW residues I30, V42, I59, V91, I134, and I142 are part of the CheW/CheA-P5 interface (Hamel and Dahlquist 2005). Further studies of the CheW/CheA-P5 interface used cysteine mutations and alkylation by an isotope-coded affinity tag. Cysteine mutations at E. coli CheW residue I65 diminished signaling of the complexes, likely due to perturbation of the CheW/CheA interface; cysteine residues at T46, I48, T59, and N60 all had decreased alkylation in the presence of E. coli CheA (Underbakke et al. 2011). All of these studies for interface 1 show good agreement with the predicted interface from the models. Figure. 3A and C and Table S2 demonstrate the agreement between the interface studies and the 6S1K model.

Interface 2, between CheA-P5 and CheW of different CSUs, is integral to the formation of CheA/CheW rings that form the baseplate of the array structure. Residues involved in interface 2 were identified by mutating CheA-P5 and CheW residues predicted to be at this interface (Piñas et al. 2016), based on CheA/CheW rings observed in crystals (PDB code 4JPB) (Li et al. 2013). Efficient receptor-dependent in vivo disulfide crosslinking of E coli CheA A546C/CheW E27C suggests that these residues are at CheA/CheW interface 2. Mutations of E coli CheA-P5 residues L545, V551, and Y558 and CheW residues R117, E121, and F122 disrupted interface 2 and significantly decreased signaling cooperativity as well as array formation in vivo (Piñas et al. 2016). Fewer studies have been done on interface 2 compared to interface 1, as illustrated by the smaller number of residues highlighted in Fig. 3A and C, and listed in Table S3. Despite this, there is strong agreement between the interfaces in the expanded model and residues identified in interface studies (green spheres at interface 2).

CheW-only rings are also a part of the extended array structure. Crosslinking identified disulfide pairs that support creation of a hexameric CheW, which includes A49C-F75C, V58C-R117C, I65C-R117C, I65C-E121C, R96C-E121C, R96C-F122C, I65C-A123C, and R96C-A123C, consistent with these E coli residues being proximal in the CheW/CheW interface (Piñas et al. 2022). These residues are listed in Table S4 and highlighted in Fig. 3D. These studies identified residues I65, R117, E121, and F122 as part of both the CheW/CheW interface and CheW/CheA-P5 interface 1 or 2. There is also significant overlap in the interaction surfaces of CheW in both the A/W and W-only rings (spheres in Fig. 3C and D). This suggests that the orientation of CheW is similar in the CheA/CheW and CheW-only rings. Like interface 2, the few studies of the CheW/CheW interfaces show good agreement with the model. To understand how signals move through the array, more information is needed on the ring interfaces.

Experimental evidence for MCP interfaces in the CSU

MCPs form an interface with CheA-P5 that may play a part in coupling the receptor signaling information (ligand occupancy and methylation state) to the activity of the kinase. NMR and crosslinking identify residues on both sides of this interface. An NMR study of T. maritima proteins in various combinations of receptor fragment (TM0014_90-206), CheAΔ354 (P4-P5 domains only), and CheW identified the residues listed in Table S5 as part of the interface: T. maritima MCP residues I135, L136, L138, and I142 along with CheA residues I563 and I566 exhibit significant chemical shift changes of methyl resonances upon binding. Furthermore, mutagenesis of two of these (MCP I142 and CheA-P5 I566) abolishes binding (Wang et al. 2012). In a disulfide crosslinking study, E. coli Tsr MCP and S. typhimurium CheA in functional reconstituted signaling arrays with CheW show fast disulfide formation rates between Tsr-CheA pairs A383C-L545C, A387C-L545C, V398C-E550C, A383C-Q619C, A387C-Q619C, and G401C-Q619C (Piasta et al. 2013). TAM-IDS studies further probe the importance of the identified residues to in vivo incorporation of CheA into complexes, indicating that V548 is essential for CheA-receptor binding, while L545 and S551 are part of the interface but not essential (Piasta et al. 2013). Additional interface studies of the CheA side include the protein-interactions-by-cysteine-modifications (PICM) method (Miller et al. 2006), in which 70 cysteine mutations placed throughout all domains of S. typhimurium CheA are coupled to 5-fluorescein-maleimide. The bulky probe inhibits kinase activity and incorporation of CheA into membrane-bound complexes when incorporated at residues E311, S340, L521, E539, L545, L627, and G636. A fluorescence anisotropy assay showed that labeling three of these (E539, L627, G636) interfered with CheA/CheW binding, and the others were presumed to be a part of the CheA/MCP interface (Miller et al. 2006). Residues E311 and S340 are in CheA-P3 and L521 is in CheA-P4; the locations of these residues are shown in Fig. S2 as they are not part of the primary interface of CheA-P5 with CF. Returning to the primary interface, another study identified similar residues in E. coli (Piñas et al. 2018). In vivo disulfide crosslinking experiments demonstrated CheW-dependent disulfide formation between cysteines introduced at CheA residues L528, V531, S534, L599 and Tsr residues F373, N376, L380, A383, V384, A387. Furthermore, mutations at each of these CheA sites reduced chemotaxis (Piñas et al. 2018). The residues identified by these studies are mapped onto the 6S1K model in Fig. 3 A and B (green color), with interface residues of the model (within 4 Å of the other proteins) shown as alpha carbon spheres. This experimental evidence is consistent with the MCP/CheA-P5 interface in the CSU model 6S1K, as summarized in Table S5 and illustrated by the number of green spheres in Fig. 4A and B. Green residues in Fig. 4 without spheres represent residues identified in the above studies that do not have atoms within 4 Å of the other protein in the 6S1K model.

Figure 4.

Agreement between residues of MCP interfaces in the CSU model 6S1K and residues identified in protein–protein interaction studies. One MCP monomer and its interacting partner, CheA-P5 in A and B, or CheW in C and D, are shown in cartoon representation. Interface residues in the model (within 4 Å of partner protein) are represented by an alpha carbon sphere, and residues identified in interface experiments are green, for (A) MCP interface with CheA-P5, (B) CheA-P5 interface with MCP, (C) MCP interface with CheW, and (D) CheW interface with MCP.

Experimental evidence has shown the interaction between the MCP and CheW is necessary for the formation of active core signaling units and that these proteins can form a complex in the absence of CheA. NMR chemical shift perturbation studies of the T. maritima receptor fragment (TM0014_90-206) combined with T. maritima CheW identified the following residues of the MCP/CheW interface (also listed in Table S6): MCP residues E132, I135, L136, A137, L138, N139, A140, T141, I142, E143, A145, R146, I156; CheW residues L14, V27, I30, V98, and L99 (Vu et al. 2012, Wang et al. 2012). Disulfide cross-linking of E. coli MCPs Tsr and Tar with CheW in vivo, with and without CheA present, identified MCP-CheW residue pairs V384C-S15C, R388C-I33C, E391C-S15C, E391C-G41, V398C-S15, and V398C-R62 as part of the interface (Pedetta et al. 2014). Interestingly, they saw a higher rate of reaction for the V398C-R62C receptor-CheW pair without CheA, and observed the opposite for R388C-I33C, which they suggested might indicate a slight difference in the MCP/CheW interface between binary and full ternary complexes. Such a difference suggests NMR studies of the receptor/CheW binary complex must be interpreted with caution. An early study used hydroxylamine random mutagenesis to identify mutations in E. coli CheW and Tsr that suppressed the chemotaxis defects caused by certain mutations of the other protein. The CheW residues E38, R62, T86, V88, G99, V105, and V108 and the Tsr residues S357, Q374, T375, A400, E402, and A413 suppressed the defects of the partner protein, suggesting they could be part of the interface (Liu and Parkinson 1991). Additional mutational studies identified residues important to just the CheW side of the CheW/MCP interface. Decreased accessibility for alkylation of cysteine mutants of E. coli CheW residues in the presence of E. coli Tsr identified residues Q37 and R62 as a part of the CheW-Tsr interface (Underbakke et al. 2011). Pull-down assays of mutated CheW with E. coli Tar-containing inner membrane preparations identified residues V36, G41, R62, and G133 as important to the CheW/Tar interface (Boukhvalova et al. 2002). The same pull-down assay was used in another study to identify residues E38 and V87 as part of the CheW/receptor interface (Boukhvalova et al. 2002). The large overlap in the residues indicated to be part of this interface in both interface experiments and the 6S1K model is shown in Fig. 4C and D (green spheres) and listed in Table S6. This demonstrates that there is a strong structural understanding of the architecture of the MCP interfaces within a CSU.

Comparison of the MCP interfaces with CheA and CheW

The membrane-distal tip of MCPs, specifically the protein interaction region, has interfaces with the other dimers in the TOD, CheA, and CheW. CheA and CheW compete for binding to the MCP (Asinas and Weis 2006) and excess CheW out-competes CheA for the interaction (Gegner et al. 1992). Excess levels of CheW can also prevent formation of receptor TODs by competing with binding of dimers to each other at the membrane-distal tip, indicating a strong binding affinity of CheW to the receptor tip (Cardozo et al. 2010). The models and interface experiments also show significant overlap in the interaction surfaces of the MCP for CheA-P5 and CheW (Fig. 5). Interface experiments on MCPs have identified eight residues that interact with both CheA and CheW, and most residues shown to interact with just one binding partner are found in the same area (Fig. 5A and Tables S5-S6). More MCP residues have been found to interact with CheW (red and cyan residues in Fig. 5A) than with CheA (red and blue residues in Fig. 5A); perhaps this is because more random mutagenesis studies have been done on CheW than on CheA. This highlights the need for clarification of the details of these interfaces, and studies that directly compare the CheW and CheA-P5 interfaces on MCPs. The residues above the dotted line (S357 and A413) demonstrate that mutagenesis studies can identify residues that are not a part of the interface but still distally affect the interaction. In the 6S1K model there is a large overlap in the MCP residues that are 4 Å away from CheW and CheA-P5 (green spheres Fig. 5B and C). Only two residues at each predicted interface (Q374 and A383 for the MCP/CheA, A387 and R404 for MCP/CheW) do not overlap (yellow spheres in Fig. 5B and C). This illustrates that there is significant overlap in the region of CF involved in the interfaces with CheA-P5 and CheW, but the finer details of the specific residue-level interaction may be different.

Figure 5.

Overlap of MCP interfaces with CheA-P5 and CheW. (A) MCP residues experimentally identified to interact with either CheW (cyan), CheA-P5 (blue), or both CheW and CheA-P5 (red). All of these residues are listed in Tables S5-S6; selected residues are labeled to orient the sequence on the model. Many of these residues also participate in the dimer-dimer contacts in the trimer-of-dimers of chemoreceptors. The dashed line represents the top of the region of CF that interacts with the partner proteins in 6S1K. Spheres in (B) and (C) indicate alpha carbons of MCP residues 4 Å away from CheA-P5 (B) or CheW (C) in 6S1K model. Green spheres indicate residues found at both interfaces in the model and yellow spheres indicates residues found at only one of the interfaces. Only one subunit of the MCP is shown as a cartoon for clarity.

The coupling protein CheW, and the regulatory domain, CheA-P5, are structurally homologous, and the same region interacts with the MCP, with subtle differences. CheW and CheA-P5 each have two subdomains that are made from two intertwining β-barrels (Bilwes et al. 1999, Griswold et al. 2002). The central hydrophobic groove between these subdomains is where the tip of the MCP interacts with both proteins (Briegel et al. 2012). The structural similarity, particularly for this groove region, is shown in Fig. 6A. However, there is much lower sequence homology between the two proteins. Consequently, the residues in this groove region differ in identity and orientation, which likely fine-tunes the binding affinity and specificity for the MCP (Fig. 6B and C). Similar amino acids are present in the groove, including hydrophobic residues like isoleucine, valine and leucine, as well as some charged or polar residues like serine, glutamate and glutamine. However, the arrangement of these residues differs between P5 and CheW, which likely changes the direct interactions with the receptor. Figure 6B and C shows the residues of the MCP interaction groove of both CheA-P5 and CheW, respectively, with spheres indicating residues within 4 Å of the MCP in 6S1K. The number and arrangement of hydrophobic groups (dark pink and green spheres) are similar, but the arrangement of the polar residues (light pink and green spheres) differ, and the CheW groove includes a charged residue (D107). Since the orientation is important for the formation of hydrogen bonds, this difference in the arrangement of polar and charged residues in the groove could affect the strength of the interfaces. Figure 6D and E shows the groove residues identified in interface studies with MCPs. More hydrophobic residues that interact with MCPs have been identified in CheW than in CheA-P5; if there really are more hydrophobic groups involved in the CheW/MCP interface, this interface could be stronger than the CheA-P5/MCP interface. This suggests that despite the structural homology, the interfaces might differ in the types of interactions present, which could affect the binding affinity and the role they play in the signaling of the complex.

Figure 6.

MCP interaction groove between the subdomains of CheA and CheW. (A) Superposition of CheA-P5 and CheW (both from PDB 6S1K) with the two subdomains indicated and the inter-subdomain grove that interacts with the MCP highlighted by a red circle. (B) CheA-P5 with residues within 4 Å of CF in the 6S1K model shown with alpha carbon spheres. Light pink indicates charged or polar residues and dark pink indicates non-polar hydrophobic residues. (C) CheW with residues within 4 Å of CF in the 6S1K model shown with alpha carbon spheres. Light green indicates charged or polar residues and dark green indicates non-polar hydrophobic residues. (D) CheA-P5 regulatory domain with residues in the interaction groove indicated to interact with the MCP by interface experiments shown with sidechain sticks, with the same color scheme as B. (E) CheW with residues in the interaction groove indicated to interact with the MCP by interface experiments shown with sidechain sticks, with the same color scheme as C.

The relative importance and role of CheA-P5 and CheW interfaces with MCPs has been explored by several studies. CheA, unlike CheW, does not form binary complexes with MCPs (Gegner et al. 1992), though truncated versions of CheA have been shown to have weak association to chemoreceptor clusters in the absence of CheW (Kentner et al. 2006, Schulmeister et al. 2008). Also, single and even double/triple mutations at CheA-P5/MCP interface can still be incorporated into complexes and support some levels of chemotaxis, indicating that none of the CheA/MCP interface residues are essential for incorporation into functional complexes (Piñas et al. 2018). However, mutations at the CheW/CheA interface 1 have detrimental effects on proper incorporation and activation of CheA into complexes (Khursigara et al. 2008, Cassidy et al. 2018, Riechmann and Zhang 2023). The 6S1K structural model contains three salt bridges at the CheW/MCP interface and none at the CheA/MCP interface, consistent with a stronger CheW/MCP interaction (Cassidy et al. 2020). This information has led to the proposal that the interactions of CheW with both the MCP and CheA are more critical for kinase control than the direct interaction of the MCP with CheA-P5 (Piñas et al. 2018).

Biochemical evidence for signaling-related changes of the protein interfaces

Multiple studies have investigated changes in the CSU protein interfaces between different signaling states. One such study looked at the effect of the serine chemoattractant on the rate of disulfide bond formation between E. coli Tsr and CheA interface residues (Piasta et al. 2013). Using in vitro reconstituted complexes of E. coli Tsr with S. typhimurium CheA and CheW, this study found that Tsr-CheA residues V398C-E550C had increased disulfide formation rate with ligand, indicating the residues are closer together in the kinase-off state; the opposite was observed for G401C-Q619C, indicating these residues are closer in the kinase-on state (Piasta et al. 2013). A similar study of the S. typhimurium CheA-CheW interface investigated the effects of serine addition on disulfide formation rate between residues. Addition of serine reduced the disulfide formation for all CheA-CheW pairs tested: D586C-N50C, S647C-Q44C, and S647C-T46C, indicating these residues are farther apart in the kinase-off state (Natale et al. 2013). An in vivo study investigating changes in the E. coli Tsr-CheW interface reported that serine enhanced crosslinking for the V398C-R62C Tsr-CheW pair, but reduced crosslinking for the R388C-I33C pair, indicating that the distances between these pairs change with signaling (Pedetta et al. 2014). The reaction rate of the V398C-R62C Tsr-CheW pair was also higher when tested using a Tsr locked-off mutant than with a Tsr locked-on mutant, consistent with these residues being closer together in kinase-off complexes (Pedetta et al. 2014). Together, these results provide intriguing evidence of signaling-related changes involving a few of the contact residues at each interface of the CSU: two MCP residues and two CheA residues of the MCP/CheA interface (Piasta et al. 2013), twp MCP and two CheW residues of the MCP/CheW interface (Pedetta et al. 2014), two CheA and three CheW residues of the CheA/CheW interface 1 (Natale et al. 2013). This evidence has been used to propose that the signaling mechanism includes a shift or rotation of these interfaces that causes different protein contacts (Natale et al. 2013, Piasta et al. 2013, Pedetta et al. 2014).

CSU structural model and future prospects: in pursuit of the signaling mechanism

As summarized above, extensive structural and biochemical studies have yielded an excellent structural model for the chemotaxis signaling array. Agreement between the different models and additional protein interface experiments shows that there is good consensus in the field for the locations of the interaction surfaces and the overall structure of the CSU. Structural and interface studies have provided vital information on the residues that are essential to interaction of the proteins. Disulfide crosslinking studies have begun to show that there are differences in the interfaces between signaling states, leading to proposals that signaling involves shifts or rotations of proteins within the CSU, but observed changes have been limited to just a few pairs of residues at each interface. Current cryo-ET techniques lack the resolution to see subtle changes in the protein contacts in the base plate. Additional experiments are needed to test the proposed changes and develop further understanding of how the signal is propagated to control kinase activity. Such experiments must be done in functional complexes and would ideally survey a larger number of residues to gain a more complete picture of the changes occurring at the interfaces. Understanding how the protein interfaces change during signaling would address questions such as how CheA activity is regulated and what role CheW plays in that regulation, to reveal the signaling mechanism of the remarkable bacterial chemotaxis signaling array.

Conflicts of interest

The authors declare no conflict of interest.

Funding

This work was supported by the National Institutes of Health R01-GM120195 to LKT. JJA was partially supported by National Research Service Award T32 GM139789 from the National Institutes of Health.