Mechanism of agonism and antagonism of the Pseudomonas aeruginosa quorum sensing regulator QscR with non-native ligands

Summary

Pseudomonas aeruginosa is an opportunistic pathogen that uses the process of quorum sensing (QS) to coordinate the expression of many virulence genes. During quorum sensing, N-acyl-homoserine lactone (AHL) signaling molecules regulate the activity of three LuxR-type transcription factors, LasR, RhlR, and QscR. To better understand P. aeruginosa QS signal reception, we examined the mechanism underlying the response of QscR to synthetic agonists and antagonists using biophysical and structural approaches. The structure of QscR bound to a synthetic agonist reveals a novel mode of ligand binding supporting a general mechanism for agonist activity. In turn, antagonists of QscR with partial agonist activity were found to destabilize and greatly impair QscR dimerization and DNA binding. These results highlight the diversity of LuxR-type receptor responses to small molecule agonists and antagonists and demonstrate the potential for chemical strategies for the selective targeting of individual quorum-sensing systems.

Graphical abstract

LuxR-type transcription factors mediate acyl-homoserine lactone signaling during Pseudomonas aeruginosa quorum sensing. Biophysical and structural studies of the quorum sensing control receptor (QscR) with synthetic ligands reveals distinct characteristics of agonist and antagonist actions. Increased dimerization and stability of QscR induced by agonists is favorable for DNA recognition, but QscR with antagonists is more labile and less competent to recognize DNA. These findings reveal new strategies for antagonizing quorum sensing.

Introduction

Pseudomonas aeruginosa (P. a.) is an opportunistic human pathogen that uses a process of inter-cellular communication known as quorum sensing (QS) to promote virulence and biofilm formation (Smith and Iglewski, 2003, Passador and Iglewski, 1995). QS enables bacteria to sense their local bacterial population density through the synthesis, diffusion, and reception of small signaling molecules, which ultimately coordinates group behaviors (reviewed in (Fuqua et al., 1996, Miller and Bassler, 2001)). Since the discovery of the “Lux” quorum sensing system in Vibrio fischeri, dozens of species of Gram-negative bacteria have been found to utilize N-acyl L-homoserine lactones (AHLs) as their primary QS signal (Whitehead et al., 2001). AHLs are neutral, lipid-like molecules consisting of a conserved L -homoserine lactone head group and an acyl-chain tail that can vary both in length and substituents (Eberhard et al., 1981, Ruby, 1996, Eberl et al., 1996). AHLs are synthesized by LuxI-type AHL synthases and are recognized by intracellular LuxR-type AHL receptors that serve as transcription factors to regulate the activity of target genes once a threshold concentration of AHL, and thus a threshold cell density, is achieved (Schuster et al., 2003, Stevens et al., 2011, Churchill and Chen, 2011, Watson et al., 2002, Schuster and Greenberg, 2008).

The P. a. QS circuitry is relatively complex and consists of a hierarchy of AHL-mediated signaling circuits that includes two major lux-like signaling pathways (Wagner et al., 2007). In the rhl system, the AHL synthase RhlI produces a C4-homoserine lactone (C4-HSL) that is recognized by the cognate AHL receptor RhlR, and in the las system the AHL synthase LasI produces 3-oxo-C12-HSL (3OC12-HSL) that is detected by the cognate AHL receptor LasR (Latifi et al., 1996). Both systems regulate genes that promote P. a. virulence (Lindsay and Ahmer, 2005, Seed et al., 1995, Smith and Iglewski, 2003). The third AHL receptor, the quorum-sensing control receptor (QscR), is an orphan or LuxR “solo” receptor because it lacks a cognate AHL synthase (Chugani and Greenberg, 2014). Interestingly, QscR can respond to 3OC12-HSL produced via the las system to attenuate QS (Lequette et al., 2006, Chugani et al., 2001, Fuqua, 2006, Lee et al., 2006). However, QscR also exhibits promiscuity in its response by sensing and being strongly activated (in many instances) by native AHL signals produced by other species (Lee et al., 2006).

The development of synthetic small molecules designed to interfere with bacterial QS represents a potentially powerful approach to study fundamental aspects of QS mechanisms and modulate bacterial virulence phenotypes in bacterial pathogens, such as P. a.. Specifically, targeting the LasR, RhlR and QscR AHL receptors could attenuate pathogenicity (Pearson et al., 2000, Lee et al., 2006, Furiga et al., 2015, Weng et al., 2014, O’Loughlin et al., 2013, Kim et al., 2008). To this end, many groups (Smith et al., 2003, Cady et al., 2012, O’Loughlin et al., 2013, Weng et al., 2014), including ours (Eibergen et al., 2015, Mattmann et al., 2008, Borlee et al., 2010, Mattmann and Blackwell, 2010, Mattmann et al., 2011, Geske et al., 2007, Geske et al., 2008), have developed libraries of synthetic small molecules toward these three P. a. receptors. These compounds, the bulk of which are AHL-derived, are capable of agonizing or antagonizing each of these receptors. Preliminary studies indicate they act by competing with the native AHL ligand for its binding site on the LuxR-type receptor. The ongoing optimization of these ligands (specifically, the antagonists), for both potency and receptor selectivity has slowed recently, due at least in part to challenges manipulating these receptors in vitro and characterizing the biochemical mechanisms by which the ligands act. Understanding the molecular interactions small molecules have with LuxR-type receptors that engender receptor antagonism (and agonism) would significantly aid ongoing research efforts in chemical probe design, as well as supplement our understanding of LuxR-type function in general. The structure of the LasR ligand binding domain in complex with 3OC12-HSL and chemically distinct triphenyl mimics has provided insight into how unrelated ligand types can be accommodated by a highly selective receptor (Bottomley et al., 2007, Zou and Nair, 2009). However, full length LuxR-type protein structures are ideal to better capture the structural changes due to ligand binding.

To date, only four full-length structures of LuxR-type receptors have been reported: P. a. QscR (Lintz et al., 2011), Chromobacterium violaceum CviR (Chen et al., 2011), Agrobacterium tumefaciens TraR (Zhang et al., 2002, Vannini et al., 2002), and Escherichia coli SdiA (Ruby, 1996, Nguyen et al., 2015). These homodimeric AHL receptors have an N-terminal ligand-binding domain (LBD) that is connected through a flexible linker to the C-terminal DNA binding domain (DBD) (Stevens et al., 2011, Churchill and Chen, 2011). Despite the relative paucity of structural data, the available structures reveal surprisingly different AHL binding pockets and symmetry (Churchill and Chen, 2011, Lintz et al., 2011). First, AHLs with short acyl chains appear exposed to solvent (Nguyen et al., 2015, Chen et al., 2011, Vannini et al., 2002) (Zhang et al., 2002), but AHLs with long acyl chains are buried within the LBD away from the solvent (Zou and Nair, 2009, Lintz et al., 2011). Second, structures with bound agonists generally form a criss-cross symmetric homodimer that poises the DBD for DNA binding, whereas TraR forms an asymmetric dimer when bound to DNA (Zhang et al., 2002, Vannini et al., 2002). Only the CviR structure captures a receptor bound to an antagonist, specifically a non-native AHL-analog (CL, Figure 1A). Interestingly, this structure shows the CviR DBDs in an altered criss-cross configuration that would abrogate binding to the promoter region (Chen et al., 2011). In view of these limited data, it is unclear whether mechanistic observations based on a single LuxR-type receptor–ligand interaction are broadly applicable in this receptor class. Moreover, we lack any structural understanding of how a full-length LuxR-type receptor can respond to both agonist and antagonist ligand types.

Figure 1. Agonists and antagonists of QscR.

(A) Structures of native and non-native homoserine lactone (HSL) molecules used in this study. Atom names and numbers are shown. (B) Dose response curves for the activity synthetic compounds C12-HSL, CL, R6, Q9, and S3 in E. coli using the QscR reporter. Agonist activity of the synthetic ligands was plotted assuming that 100% activity is equivalent to the activity of a natural agonist, 3-oxo-C12HSL (not shown). EC₅₀ values calculated using GraphPad Prism (v. 6.0). Error bars are the s.e. of the means of triplicate samples. (C) Dose responses and antagonism IC₅₀ values for AHLs R6 and Q9 in E. coli using the QscR reporter. Reporter activity was measured for varying concentrations of the synthetic ligands in the presence of a fixed concentration 15 nM of 3OC12-HSL. IC₅₀ values, indicating the extent antagonism, were calculated using GraphPad Prism (v. 6.0). Error bars are the s.e. of the means of triplicate samples.

We sought to define the mechanisms by which a P. a. LuxR-type receptor can respond to either AHL agonists or antagonists. As LasR was not suitable for study because of its low solubility in vitro (Kim et al., 2014, Wu et al., 2008, Lerat and Moran, 2004), we chose QscR because full-length QscR is amenable to structural studies, and like LasR, it recognizes 3OC12-HSL (Lintz et al., 2011). Reporter-based assays were used to define a small set of synthetic AHL agonists and antagonists developed in our laboratories (Mattmann et al., 2011) and elsewhere (Chen et al., 2011) for investigation. The molecular mechanisms underlying the responses of QscR to these agonists and antagonists were examined using a variety of biochemical, biophysical and structural approaches. Agonists were found to stabilize QscR, increase dimerization and DNA binding, whereas antagonists were greatly impaired in these functions. Notably, one AHL agonist was found to recognize QscR using both “short-acyl chain” and “long-acyl chain” modes of binding. These results support a model for QscR response to agonists and antagonists that is distinct from the one operating in the response of LuxR-type receptors to short-chain AHLs.

Results

Activity of QscR in response to AHL analogs that serve as agonists and antagonists

Our laboratory has previously reported responses of QscR to AHL analogs in cell-based reporter assays (Mattmann et al., 2011, Mattmann et al., 2008). Based on these findings, AHLs S3, Q9, and R6 (Figure 1A) were selected for further study as they displayed a range of activities in comparison to 3OC12-HSL. In addition, a chloroaryl ligand reported by Bassler and co-workers, CL (Figure 1A), was examined because it strongly antagonizes the LuxR-homolog CviR and is well understood from the [CviR:CL]₂ X-ray crystal structure (Chen et al., 2011). The C12-HSL, the 3OC12-HSL analog simply lacking the 3-oxo group, was also included to examine a closely related mimetic and naturally occurring AHL.

To determine the relative potencies of these compounds, each was evaluated in an optimized cell based reporter assay for QscR agonism and antagonism. We utilized an E. coli reporter strain (Lee et al., 2006, Mattmann et al., 2008) containing a QscR overexpression plasmid and a QscR-agonist activated β-galactosidase reporter gene (Table 1). Plots of the percent agonism as a function of AHL concentration allowed for the determination of the EC₅₀ for each compound (Figure 1B and Table 2). The C12-HSL, CL and the branched-chain S3 were found to be potent agonists of QscR activity, with EC₅₀ values comparable to the native ligand OdDHL (15 nM). This activity profile for CL was interesting, as CL is an antagonist of CviR activity (see above). In turn, phenyl HSL derivatives R6 and Q9 were very weak QscR agonists, with EC₅₀ values > 70 nM.

Table 1.

Bacterial strains and plasmids used for in vivo assays

Strain or plasmid	Description	Reference
Strain
E. coli JLD271	K-12 ΔlacX74 sdiA271::Cam; ClR	(Lindsay and Ahmer, 2005)
E. coli BL21 (DE3)pLysS	F− ompT hsdSB (rB−, mB−) gal dcm (DE3) pLysS(CamR)	Invitrogen
Plasmid
pET3a-qscR	Expression plasmid for full-length QscR	(Oinuma and Greenberg, 2011)
pJN105Q	arabinose-inducible QscR expression vector; GmR	(Lee et al., 2006)
pSC11-Q	PA1897’-lacZ transcriptional fusion; QscR reporter vector; ApR	(Eibergen et al., 2015)
pJN105L	arabinose-inducible LasR expression vector;GmR	(Lee et al., 2006)
pSC11	lasI’-lacZ transcriptional fusion; LasR reporter vector; ApR	(Chugani et al., 2001)

Table 2.

Activation and inhibition data for QscR with agonists and antagonists

Compound	IC50 [μM]	95% CIa [μM]	Inhibition (%)	EC50 μM]	95% CI [μM]	Activation (%)
3OC12-AHL	–	–	–	0.015	0.012 - 0.018	100
C12-AHL	–	–	–	0.015	0.011 - 0.020	112
S3	–	–	–	0.036	0.021 - 0.060	88
CL	–	–	–	0.016	0.010 - 0.027	86
Q9	0.026	0.015 - 0.042	80	0.12	0.0037 - 3.6	11
R6	0.042	0.017 - 0.10	58	0.074	0.025 - 0.22	23

^aCI = Confidence interval.

The antagonist activity of each compound was determined by examining the ability of the compound to compete with the naturally-occurring QscR ligand 3OC12-HSL at its EC₅₀ value. As a function of concentration, the percent activities for S3 and CL increased, as expected, due to their agonist activities. However, Q9 and R6 inhibited reporter activation by up to 80% with IC₅₀ values in the mid-nanomolar range (Figure 1C, Table 2). The maximal inhibitory activity of the compounds is equal to the activation in the agonism assay, suggesting a “classical” partial agonist behavior. Notably, not only are compounds Q9 and R6 the most potent known inhibitors of QscR, they also are ten-fold more potent than the best inhibitors of LasR (using an analogous E. coli reporter) and do not display the atypical partial agonist character seen at high concentrations for other inhibitors (Moore et al., 2015b).

The structure of QscR bound to S3 reveals a novel ligand binding mode

To date there are no structural details of a LuxR-type receptor that has been solved both in the presence of a small molecule antagonist or an agonist. Therefore, we made many attempts to co-crystallize QscR with agonists C12-HSL, S3, and CL or antagonists Q9 and R6, in order to investigate the structural basis underlying these activities. Full-length QscR (Fig. S1) together with agonist S3 or C12-HSL (abbreviated QscR–S3, QscR–C12) were crystallized and the structures were determined using x-ray crystallography (Figure 2A). The structures were solved at a resolution of 2.8 Å for QscR–S3 and 2.5 Å for QscR–C12 using molecular replacement with our previously reported QscR–3OC12-HSL structure as the search model (Lintz et al., 2011). The refined QscR models (Table 3) had continuous electron density in chain A from the N-terminal residue to the C-terminal residue Asn237 (Figure 2). For the QscR–S3 structure, chain B had regions of discontinuous electron density and B factors that were greatly elevated relative to chain A, which is consistent with the dearth of crystal packing contacts for chain B compared to chain A. There was clear electron density for the C12-HSL and S3 ligands (Figure 2B).

Figure 2. Structural analyses of QscR-ligand complexes.

(A) Chemical structures of 3OC12-HSL, C12-HSL, and S3. (B) Electron density of C12-HSL or S3 in the QscR-complex crystal structures (PDB IDs 6CC0 and 6CBQ, respectively). Each 2mFo-DFc map was contoured at 1.4 σ. (C) The overall structure of the QscR–S3 and QscR–C12 complexes. Chains A and B are colored in white or grey, respectively. S3 is in green and C12 is in cyan. The LBD of Chain A is connected to the DBD through a flexible linker. The dimerized protein has a criss-cross symmetric architecture, where each domain has contacts with all of the others. (D) Overlay of QscR-ligand complexes. S3 is in green, C12 in cyan and 3OC12 in blue. Spheres represent ordered solvent molecules within the structures. (E) Overlay of different AHL-receptor complexes. QscR–S3 (green), CviR–CL (purple), TraR–3OC8-HSL (beige), and SdiA–3OC8-HSL are superimposed to illustrate the buried and solvent exposed region of the HSLs.

Table 3.

Statistics for crystallographic data collection and refinement

Data statistics	QscR-S3 (PDB: 6CBQ)	QscR-C12 (PDB: 6CC0)
Spacegroup	P3121	P212121
Cell dimensions (Å)	a=94.12 b=94.12 c=105.68 α=90° β=90° γ=120°	a=57.59 b=91.91 c=94.00 α=90° β=90° γ =90°
Resolution range (Å)	47.06 - 2.80 (2.90 - 2.80)a	48.8 - 2.35 (2.48 - 2.35)a
Unique reflections	13,685 (1343)a	20,134 (2798)a
Redundancy	6.5 (6.3)a	4.0 (3.0)a
Rsymb (%)	9.0 (54.5)a	7.7 (55.9)a
Completeness (%)	99.3 (98.6)a	94.9 (91.6)a
Intensity (I/σ)	12.3 (2.8)a	11.8 (1.1)a
CC1/2		0.997 (0.5525)a
Wilson B factor (Å2)	77.5	61.2
Refinement statistics
Resolution Range (Å)	47.07-2.80 (3.02 - 2.80)a	48.8 - 2.50 (2.589 - 2.50)a
Unique reflections	14,344 (1379)	16,814 (1663)a
Rfreec (%)	26.6 (43.3)a	27.0 (36.5)a
Rworking (%)	21.2 (33.5)a	20.3 (30.3)a
Final Model
Number of protein atoms	3816	3950
Number of ligand atoms	48	40
Number of solvent atoms	10	51
Average B factor (Å2) (TLS groups)	94.6 (14)	34.1 (9)
R.m.s.d. bond lengths (Å)	0.002	0.007
R.m.s.d. bond angles (°)	0.42	0.87
Ramachandran Analysis	98% most favored; 1.7% allowed; 0.42% outlier	98% most favored; 1.7% allowed; 0.21% outlier

^aHigh resolution shell

^bR_sym = Σ|I − <I>|/ΣI

^cR_free calculated with an excluded set of 5%

Both QscR-ligand structures were virtually identical in overall architecture to QscR bound to the native agonist 3OC12-HSL (abbreviated QscR–3OC12). The subunits of QscR form a symmetrical criss-cross homodimer, where the LBD and DBD of one chain makes dimerization contacts with the adjacent chain LBD, poising the DBDs for DNA binding (Figure 2C and Table S1). The root mean square deviations (R.M.S.D.) of atomic positions for the individual domains of the three structures are within the coordinate error of the structures, and thus not significantly different. The R.M.S.D. values for the single subunits and the full-length proteins are higher, which indicates that the structures have slightly different interdomain and intersubunit configurations (Table S1). The R.M.S.D. of the Cα atoms for QscR-S3 compared to QscR-3OC12 is 0.67Å, but for QscR–C12 compared to either QscR–S3 or QscR–3OC12, the values are much higher, at 1.49 Å and 1.58 Å, respectively. The spacegroup and crystal packing of QscR–C12 differ from those of either QscR–S3 or QscR–3OC12, which appears to slightly alter the rotation and translation of the LBDs relative to each other. Even considering the difference in crystal packing, the dimerization interfaces of the LBDs are well conserved in all of the QscR-ligand complexes. Contacts that had previously been validated using activity assays in E. coli with QscR–3OC12 were generally conserved in these structures (Lintz et al., 2011).

The substituents within the ligands 3OC12-HSL, S3 and C12-HSL retain key agonist features. The C12-HSL acyl chain superimposes well with that of 3OC12-HSL. At atom 10 (Figure 2A), S3 branches into two aliphatic-chains of 6 and 8 carbons in length. The longer chain of S3 is buried in the ligand-binding pocket, where it extends to the same position as both C12-HSL and 3OC12-HSL (Figure 2D). Carbons C17-C19 of S3 superimpose with C19-C21 of 3OC12-HSL and C12-HSL at the distal end of the pocket and interact with the same residues. This finding provides additional independent evidence that there is a region deep in the binding pocket, in close proximity to the LBD-A-DBD-B dimerization interface that is important for agonist activity (Lintz et al., 2011).

C12-HSL is identical to 3OC12-HSL except that it lacks the carbonyl oxygen at the 3-position of the acyl chain (O12 in Figure 2A). S3 also lacks O12. The water molecules that form a hydrogen bond network within the 3OC12-HSL binding site near O12 are not seen in either the QscR–S3 or QscR–C12 structures, but the water bridging Ser38, Ser129, and Met127 and the network of hydrogen bonds linked to O9 are conserved (Figure 2D). Interestingly, for S3 and C12-HSL, which lack O12, the nearby atoms C8, O9, and C10 all superimpose well, surprisingly better than the overall lactone ring positions for all three ligands, and all three ligands appear to pivot about this point.

S3 bound to QscR mimics the features of AHLs with both long and short acyl chains found in the known LuxR-type receptor structures. One branch of S3 is a 6-carbon, aliphatic chain extends out of the binding pocket into the solvent (Figure 2 C and D), where it displaces the water molecules bound in QscR–3OC12 and Phe54 (Figure 2E). Phe54 adopts an alternate rotamer that would sterically clash with the acyl chains of C12-HSL and 3OC12-HSL. However, Phe54 forms numerous van der Waals interactions with S3 and also opens a channel to the solvent for the 6-carbon chain serving as a “gate keeper”. Therefore, one branch of S3 extends within the LBD, similarly to “long-acyl chain” AHLs bound to LasR and QscR (Bottomley et al., 2007, Lintz et al., 2011). The other shorter chain mimics the “short-acyl chain” AHLs and analogs bound to CviR (Chen et al., 2011), TraR (Vannini et al., 2002, Zhang et al., 2002), and SdiA (Kim et al., 2014) (Figure 2E), which all exit the LBD toward the solvent. Thus, S3 resembles both types of ligands due to its branched structure.

Relative to agonists, antagonists destabilize QscR

The structural analyses of LuxR-type receptors so far, including those described above for QscR, suggest the hypothesis that the mechanism of agonism and antagonism might be different between the receptors that recognize long-acyl chain AHLs (i.e., LasR and QscR) compared to receptors that respond to short-acyl chain AHLs (i.e., TraR, SdiA, and CviR). The mechanism of antagonism for CviR by CL is well understood and involves the stabilization of CviR in the antagonist bound form (Chen et al., 2011). From studies in cells or in vitro comparing free QscR and LasR to agonist bound forms, the agonist bound form might be physically stabilized (Oinuma and Greenberg, 2011, Suneby et al., 2017). These results suggest that there may be an alternative model for AHL receptor agonism and antagonism than that observed for CviR. Unfortunately, we have been unable to obtain crystals of QscR bound to the antagonists.

To understand the underlying basis for the antagonism of QscR, we instead compared the physiochemical stability of QscR with agonists or antagonists. Like some other LuxR-type receptors, such as TraR (Zhu and Winans, 1999), LasR (Schuster et al., 2004) and LuxR (Urbanowski et al., 2004), QscR is soluble and stable in cells with overexpression in the presence of agonists, but is largely insoluble upon expression with antagonists (Fig. S2). This observation suggests that QscR might be less stable in a complex with an antagonist than an agonist, and limited proteolytic digestion studies were used to examine this hypothesis (Figure 3 A and B, Fig. S3). Tryptic cleavage of QscR at an arginine in the linker between the LBD and DBD produced a relatively stable LBD that was quantitated as a function of time. Proteolytic digestion revealed that the QscR-agonist complexes had a nearly 2-fold greater half life (t_1/2) than the QscR-antagonist complexes (Figure 3 A and B, Table 4, Fig. S4), supporting our hypothesis.

Figure 3. Stability of QscR in the presence of agonists or antagonists.

(A) SDS-PAGE showing proteolytic digestion of QscR in the presence of trypsin for a representative agonist and antagonist. (B) Quantitation of the time-course of proteolysis for 3OC12-HSL, S3, CL, Q9, and R6. (C) Circular dichroism thermal denaturation plots of the QscR LBD in the presence of agonist 3OC12-HSL (blue) or antagonist R6 (orange). D) Circular dichroism thermal denaturation plots of the QscR 3OC12-HSL (blue) or antagonist R6 (orange).

Table 4.

Half lives of QscR-ligand complexes in the presence of Trypsin

Ligand	3OC12-HSL	S3	CL	Q9	R6
t1/2 (min)	11.3 ± 0.9	12.3 ± 1.0	19.3 ± 3.3	7.8 ± 0.4	8.1 ± 0.3

As another measure of the ability of ligands to stabilize QscR, we examined the thermal denaturation properties of one agonist complex, QscR-3OC12-HSL, and one antagonist complex, QscR-R6. Circular dichroism (CD) spectroscopy provides a sensitive measurement of alterations in protein secondary structure that occur as the temperature is raised. The LBD-ligand CD denaturation curves were similar regardless of which ligand was bound (Figure 3C). However, for the full-length QscR-ligand complexes, a different pattern in the thermal denaturation curves was observed for each ligand. QscR-R6 exhibited a pre-melting transition with the appearance of a second transition, whereas the denaturation curve for QscR-3OC12-HSL showed cooperative behavior consistent with the denaturation of one main species (Figure 3D). These denaturation patterns support the model that the agonist stabilizes the formation of a cooperatively folded complex, which denatures as a single unit, whereas the domains or subunits appear to unfold separately for the antagonist complex.

Dimerization of QscR is altered by antagonist binding

Differences in the conformation of proteins can often be detected using nondenaturing electrophoretic mobility shift assays (EMSA). Using such EMSAs, we observed that QscR migrated differently when agonists or antagonists were present. QscR bound to the antagonists Q9 or R6 migrates as two species, whereas the agonists favor the formation of the more slowly migrating species (Figure 4A). These EMSA results, together with both the increased exposure of the linker of QscR-antagonist complex to proteolysis and loss of cooperative unfolding in thermal denaturation experiments, suggested that the dimerization properties of QscR might be different with bound agonist or antagonist.

Figure 4. Oligomerization of QscR bound to agonists or antagonists.

(A) Electrophoretic mobility shift assay assessing the mobility of QscR in the presence of the agonist 3OC12-HSL, agonist S3, antagonist Q9 or antagonist R6. (B) Size-exclusion chromatography (SEC) of QscR at different concentrations with different compounds. The peaks of the traces were normalized to 1. Lines at ‘a’, ‘b’ and ‘c’ indicate the elution volumes observed for ‘a’ agonists at higher QscR concentrations, ‘b’ agonists at lower QscR concentration and antagonists at higher QscR concentrations, and ‘c’ antagonists at lower QscR concentration. (C) Graphical representation of all of the SEC elution volumes as a function of QscR concentration with agonists (closed circles) and antagonists (open squares). *** p-value <.001

In previously reported analytical size-exclusion chromatography experiments (Oinuma and Greenberg, 2011), the elution volume of QscR decreased as the concentration of 3OC12-HSL increased from 1.4 μM – 160 μM, indicating the appearance of a larger complex, which was interpreted to be a QscR homodimer. We used similar analytical size-exclusion chromatography experiments here to compare QscR with different concentrations of agonist or antagonist. The five ligands were tested with the following concentrations of QscR: 20 μM, 120 μM, 140.5 μM (and an additional concentration for the antagonists at 146.8 μM) (Figure 4B). A plot of the elution volume as a function of complex concentration for the agonists compared to antagonists showed that QscR-agonist complexes consistently have a lower elution volume at all concentrations compared to the QscR-antagonist samples (Figure 4C). These data suggest that agonist binding promotes a larger, dimeric form of QscR much more efficiently than antagonists.

Given that agonists compared to antagonists alter the QscR dimerization equilibrium and potentially the mode of dimerization, we examined the QscR-ligand complexes using protein crosslinking. Disuccinimidyl suberate (DSS) can crosslink lysines for which the Cα to Cα distance is up to approximately 21 Å. A representative SDS-PAGE showed that a QscR complex with an agonist 3OC12-HSL or S3 in the presence of DSS forms a greater proportion of crosslinked dimers than the QscR–Q9 antagonist complexes (Figure 5 A and B). Crosslinked monomers were observed in all cases. To identify the positions of the crosslinks, we performed crosslinking mass spectrometry (XL-MS) analyses of the specific monomer and dimer complexes. Although only a few crosslinked peptides were observed (Figure 5C), there were notable differences between the agonist and antagonist samples. Agonists promoted amino acid 63-223 intrasubunit crosslinks, whereas the 1-191 intrasubunit crosslink was observed only for the antagonist. The 1-1 and 208-208 intersubunit crosslinks were observed for the agonist complexes, but not for the antagonist complexes, and instead 121-223 was observed.

Figure 5. QscR mass spectrometry crosslinking (XL-MS) in the presence of agonists or antagonists.

A) SDS-PAGE showing QscR complexes with either agonists or antagonists that have been treated without or with DSS. (B) Quantitation of the ratio of crosslinked dimer to monomer from experiments represented in panel A. C) Crosslinked amino acids identified by XL-MS. The crosslinks for the monomer are shown below the dimer crosslinks in the panel, and colored blue for 3OC12, green for S3 and red for Q9.

To determine which AHL-receptor model(s) would satisfy the crosslinking constraints, the crosslinks were mapped onto the QscR-agonist and the reported CviR-antagonist structures (Chen et al., 2011) (Figure S5 and Table 5). The Cα to Cα distances between the crosslinked amino acid residues were used as a measure of similarity, where a distance of 21 Å is approximately a maximum distance for a DSS crosslink (Table 5). Comparison of the models of the QscR monomer, QscR dimer to the CviR dimer revealed that the QscR-agonist crosslinks are consistent with the QscR dimer crystal structure, and not the CviR structure. For the QscR-Q9 antagonist monomer and dimer complexes, more than one crosslink is not consistent with the QscR structure, which indicates that it samples alternative conformations that bring the DBD closer to the LBD compared to QscR bound to agonists. Therefore, the QscR dimer crosslinks are consistent with the QscR dimer crystal structure, but not the CviR model, and the differences between the QscR-agonist versus QscR-antagonist complexes point to alternative forms of QscR when bound to an antagonist.

Table 5.

Crosslinks and inter-Cα distances

	QscR crosslink	QscR Inter Cα distance (Å)	CviR equivalent crosslink	CviR equivalent distance (Å)
Monomer
C12	63-223	38	85-245	52
C12	217-229	11	239-251	16
C12	121-208	29	147-230	32
Q9	1-191 (4-191)	53	7-213	38
Q9	63-229	30	85-251	50
Dimer
C12	1-1 (4-5)	20	7-7	27
C12	208-208	26	230-230	51
Both	223-223	13	245-245	50
Both	208-223	20	230-245	51
Q9	121-223	27*	147-245	35

DNA recognition by QscR is severely impaired in the presence of antagonists

The degree to which antagonists compared to agonists alter the DNA binding affinity of QscR is unknown. Therefore, we used EMSA to obtain the K_D values for the two types of complexes with DNA. A fluorescently labeled palindromic 31 base pair DNA duplex that containing the QscR binding site shifts up in the gel to give a well defined complex in the presence of the agonist ligands. For antagonists, in contrast, we observed a broad smeared band, which indicates that the complexes are dissociating in the gel, or that there are multiple forms of complexes. (Figure 6A and Fig. S6). The calculated K_D values from (Figure 6B) for the QscR with agonists are nearly 2-orders of magnitude lower than for QscR with the antagonists (compare the average 1.2 nM to 77 nM, Table 6). Moreover, the Hill coefficients differ (compare the average of 1.83 for agonists to 0.8 for antagonists). These results indicate that agonists promote high affinity cooperative DNA binding, in contrast to the antagonists that bind more weakly and do not show cooperative DNA binding. These trends are analogous to those recently reported for LasR-DNA binding in the presence of agonist and antagonist ligands (Suneby et al., 2017).

Figure 6. DNA binding of QscR in the presence of agonists or antagonists.

(A) Representative electrophoretic mobility shift assays of DNA and QscR bound to 3OC12-HSL, S3, Q9, and R6, with concentrations of QscR and bound and free bands indicated. B) Quantitative analyses of all EMSAs. Data were plotted and fit with the binding Eqn. 4, which accounts for ligand depletion.

Table 6.

DNA binding affinity of QscR with agonists and antagonists.

Compound	a KD [nM]	b KD [nM]	b Hill coefficient
3OC12-HSL	0.53 ± 0.23	1.0 ± 0.2	1.7 ± 0.4
S3	0.76 ± 0.29	1.3 ± 0.2	2.0 ± 0.4
CL	0.72 ± 0.21	1.2 ± 0.1	1.9 ± 0.3
Q9	56 ± 14	62 ± 27	0.9 ± 0.2
R6	54 ± 22	92 ± 30	0.7 ± 0.3

^aValues calculated using Eqn 4.

^bValues calculated using Eqn 5.

Discussion

The results presented here provide new insights into how both agonists and antagonists influence the structure and activity of QscR and support a new model for antagonism of AHL-mediated quorum sensing.

A mechanism for QscR antagonism

Currently there are several models for LuxR-type receptor responses to native AHLs (reviewed in (Stevens et al., 2011)). TraR and LasR represent LuxR-type receptors for which agonists stabilize and dimerize otherwise unfolded protein (Figure 7). Receptors such as LuxR are nonfunctional dimers that require AHL for DNA binding, whereas MrtR-type receptors are non-functional monomers that dimerize to bind DNA. In contrast, SdiA (Yao et al., 2006) and others function as monomers in the absence or presence of agonists (Almeida et al., 2016, Nguyen et al., 2015), and dimerization is thought to be induced by DNA binding. The model for QscR activation resembles that of TraR and LasR, where agonists stabilize QscR to aggregation and proteolysis in the cell and bound AHL stabilizes the protein in vivo and in vitro (Oinuma and Greenberg, 2011, Corral Lugo et al., 2017, Zhu and Winans, 2001).

Figure 7. Model for antagonism of QscR.

The well studied model systems QscR (green), CviR (purple) and TraR (beige) highlight different responses of AHL receptors toward agonists (blue) and antagonists (red). Thin arrows indicate that the pathway is minor.

Less is known about AHL receptor antagonism. The model put forth for antagonism of CviR shows stabilization of dimeric CviR in a non-DNA binding form (Chen et al., 2011). CviR together with the antagonist CL was less resistant to limited proteolysis than with an agonist (Chen et al., 2011). Interestingly, it is the antagonist bound form of CviR that is more stable and is dimeric in solution than the agonist form (Figure 7). In contrast, the biophysical studies reported here reveal that in vitro QscR is less stable with bound antagonists than agonists (Figure 7). Further, CL behaves as a very strong agonist of QscR, in contrast to its antagonistic activity profile reported with CviR (Chen et al., 2011).

AHL receptor dimerization on DNA is critical for transcriptional activation. In the presence of agonists QscR forms a relatively stable cooperatively folded dimer. In contrast, in the presence of even high concentrations of QscR and antagonists, QscR is largely monomeric and less stable. Moreover, QscR with antagonists binds to DNA approximately 100-fold weaker, and the low fractional saturation of the DNA indicates that the dimeric complexes that do form are less stable and have a higher off-rate than DNA complexes formed with QscR and agonists (Hoopes et al., 1992). This behavior of QscR presents a model for the antagonism of AHL receptors that is distinct from the model established for CviR (Figure 7).

A hot spot for agonists in QscR

In vivo, QscR responds with greatest activity to AHLs that have acyl chains that are between 10 and 14 carbons in length (Oinuma and Greenberg, 2011). Consistent with this, our structural analyses of QscR bound to three different agonists (3OC12-HSL (Lintz et al., 2011), C12-HSL and S3) highlight conserved contacts in the distal end of the AHL binding pocket. Structure-function studies of AHL libraries using reporter strains have also demonstrated that AHL agonists of QscR typically (but not always) possess aliphatic tails, and AHL antagonists usually have benzoyl type tails (Mattmann et al., 2011). Despite the difference in the chemical nature of the “acyl chain substituent”, the length of the substituents from the end to the 1-oxo-position of the HSL headgroup was similar, and compounds that were either too long or too short fail to activate QscR as potently (Mattmann et al., 2011). S3 has one branch of 8 carbons that, like C12-HSL and 3OC12-HSL, reaches the same position in the distal end of the QscR binding pocket. Interestingly, its agonistic activity was not abolished even when at the same time the S3 C6 branch adopts a position much like an antagonist bound to CviR (Chen et al., 2011).

In contrast to our expectations, based on the antagonism of CviR by CL (Chen et al., 2011), we found that CL was a potent agonist of QscR with an EC₅₀ value comparable to that of 3OC12-HSL. CL is bulkier at the distal end (away from the HSL) than any of the ligands studied here, and is well suited to the relatively large QscR binding pocket (Lintz et al., 2011). In fact, we previously found that C14-HSLs as well as biaryl and distal cyclohexane HSLs of similar length to CL have some agonistic activity in QscR (Mattmann et al., 2011). Furthermore, the distance between the 1-oxo-position of the acyl-chain and the terminal atom of CL is nearly identical to that for C12-HSL, 3OC-12-HSL and S3. Therefore, we predict that CL would similarly bind in the acyl-chain pocket and stabilize the cooperatively folded structure.

There are now multiple lines of evidence that the distal region of the ligand-binding pocket is a critical feature of QscR agonism, and we now refer to it as the “agonist hot spot”. We previously reported support for this model because a substitution of Gly 40 to Phe in QscR, near the distal end of the AHL pocket, resulted in increased response to 3OC6-HSL and a decreased response to 3OC12-HSL (Lintz et al., 2011). The Phe substitution is predicted to reduce the space in the pocket, which would allow AHLs with shorter acyl-chains to have better agonist activity. Like QscR, LasR has room to accommodate longer ligands (Zou and Nair, 2009, Bottomley et al., 2007), and these ligands superimpose in a hotspot in LasR created by different residues. In contrast, the SdiA (Kim et al., 2014, Nguyen et al., 2015), CviR (Chen et al., 2011) and TraR (Zhang et al., 2002, Vannini et al., 2002) structures show larger residues blocking this distal site, which indicates that agonists with short acyl chains may have a different mechanism of activation.

Implications for the design of new agonists or antagonists

For AHL receptors in general, it has been difficult to produce synthetic compounds with EC₅₀ values that are lower than natural agonists, or with IC₅₀ values that are lower than 100-1000x the EC₅₀ of the natural agonists ((Eibergen et al., 2015, Mattmann et al., 2008, Borlee et al., 2010, Mattmann and Blackwell, 2010, Mattmann et al., 2011) and reviewed in (Stevens et al., 2011)). S3 is one of the most potent synthetic QscR agonists reported to date (along with CL) (Chen et al., 2011). S3 is unusual because it is branched and exhibits two biologically relevant modes of binding. The long chain is buried in the LBD and is important for agonism. The short chain adopts the mode of binding that has been observed for antagonists (CviR) and for agonists with short acyl chains (TraR and SdiA). This dual mode of binding of S3 is possible because of a change in the rotamer of the key gatekeeper residue, Phe54, from a position that blocks the exit channel to one that opens it. The residues equivalent in TraR (Ala) and CviR (Val), are shorter and rather than closing the pocket, they appear to orient the exiting ligands. Interestingly, the structure of the equivalent Phe in SdiA is much like Phe54 QscR bound to C12-HSL and 3OC12-HSL. The equivalent Phe in LasR adopts a completely different position in the reported structures of LasR, as the entire loop is oriented differently, perhaps as a result of crystal packing (Zou and Nair, 2009).

This difference between the QscR and LasR structures can help explain the differences in compound activity observed in cell-based reporter screening assays (Table S2). Indeed, S3 shows much less activation of LasR (Mattmann et al., 2011), highlighting that the short acyl-chain mode of binding is likely less applicable to LasR. Q9 and R6 are partial agonists with the most potent known IC₅₀ values in the QscR cell based reporter, but with LasR, these compounds have reduced potency in both agonism and antagonism assays (Table S2). Such activity further supports the observation that LasR has a slightly smaller binding pocket than QscR. Interestingly, CL contains an aromatic ring and is an extremely selective QscR agonist. It activates QscR as potently as the “native” QscR ligand 3OC12-HSL. In contrast, LasR, is inhibited by CL (Table S2) (Moore et al., 2015a). This may be due to the ability of QscR to accommodate ligands that are longer and bulkier than 3OC12-HSL as compared to LasR (Mattmann et al., 2008). Because compounds that selectively modulate specific LuxR-type P. aeruginosa receptors allow for more precise phenotype modulation, future QscR agonist development may benefit from added distal bulk that mimics the length and packing of a C12 ligand.

Although no crystal structures with any antagonists were obtained, our improved knowledge of the molecular mechanism of QscR agonism points to new design approaches to improve antagonist activity. The distance between the 1-oxo-position and the distal end of the AHL binding pocket is rather strict. R6 is approximately 25% longer than S3, and similar in length to C12-HSL and 3OC12-HSL, but it appears too bulky to fit in the restricted entrance to the binding pocket, and likely extends to the solvent. Q9 exhibits lower partial agonist and higher antagonist activity than R6. However, Q9 is even longer than R6, which also may favor exiting the binding pocket. One approach to the design of new antagonists would be to develop compounds with a hydrophobic moiety that cannot reach the agonist hotspot, but will contribute binding energy from other interactions. Additional interactions could come from a second branch that has hydrophobic character in the exit channel and hydrophilic character at the terminus to improve solubility and affinity, and concomitantly lower partial agonist activity. These types of compounds could be improved antagonists of QscR. Our results and model suggest that with further design modifications antagonists could be developed that have even less partial agonist activity, and notably, more potent IC₅₀ values. More broadly, these design principles could be applied to other LuxR-type receptors that respond to AHLs with longer acyl chains, for the development of improved chemical probes and as strategies to further study QS as an anti-infective target.

Experimental Procedures

Chemicals

Non-native AHLs were synthesized as described previously (Mattmann et al., 2011). N-(3-oxododecanoyl) L-homoserine lactone (OdDHL) was purchased from Sigma–Aldrich. Chlorophenol red-β-D-galactopyranoside (CPRG) was purchased from Roche.

Strains and plasmids

Strains and plasmids used for this study are summarized in Table 1. Media and reagents were obtained from commercial sources and were used according to manufacturer’s instructions. Strains were grown in Luria broth (LB) at 37°C unless specified otherwise.

Activity assays

Assays for QscR and LasR activity were performed as previously described utilizing the E. coli strain JLD271 harboring pJN105-type expression plasmids and pSC11-type reporter plasmids (pSC11Q/pJN105Q and pSC11/pJN105L for QscR and LasR, respectively). For antagonism assays, increasing concentrations of compound were screened against OdDHL at approximately its EC₅₀ value in the QscR (15 nM) or LasR (2 nM) bacterial reporter strains. For both QscR and LasR agonism assays, increasing concentrations of compound were screened and compared to 100 μM 3OC12-HSL to define maximum activity. A modified Miller assay was performed using chlorophenol red-β-D-galactopyranoside (CPRG) as a β-galactosidase substrate. The amount of product was measured from the OD₅₇₀ nm using a Synergy 2 plate reader. Enzymatic activity was calculated using the following equations (Eqs. 2a and 2b):

$$\mathrm{Y}=\text{Bottom}+\frac{\left(\text{Top}-\text{Bottom}\right)}{\left(1+{10}^{\left(\left({\text{LogEC}}_{50}-\mathrm{X}\right)\ast \text{Hill}\text{Slope}\right)}\right)}$$ Eq. 2a

$$\mathrm{Y}=\text{Bottom}+\frac{\left(\text{Top}-\text{Bottom}\right)}{\left(1+{10}^{\left(\left({\text{LogIC}}_{50}-\mathrm{X}\right)\ast \text{Hill}\text{Slope}\right)}\right)}$$ Eq. 2b

Where Top and Bottom are the plateaus in the dose response curve. The EC₅₀ or IC₅₀ values are the concentrations resulting in a response half way between the Top and Bottom plateaus. Hill Slope describes the steepness of the curve.

QscR solubility tests

E. coli strain BL21 (DE3) pLysS (Invitrogen) containing pET3a-qscR (Oinuma and Greenberg, 2011)(gift from the Peter E. Greenberg, University of Washington) was precultured in LB broth with 100 μg/mL ampicillin and 34 μg/mL chloramphenicol at 37°C overnight. Three 500 mL flasks with 100 mL LB and 100 μg/mL ampicillin were inoculated with 100 μL of this 6 mL overnight starter culture. The cells were grown at 37°C to an OD₆₀₀ of 0.89. Either 3OC12-HSL, agonist or antagonist was added to each flask at a final concentration of at 50 μM and the flasks were cooled on ice for 2 min. Expression was induced with a final concentration of 0.5 mM IPTG (GoldBio), and the cells grown for 17 hr at 17°C. Cells were harvested and lysed in 25 mM sodium phosphate, pH 7.0, 1 mM DTT, 1 mM EDTA, 10% glycerol, 150 mM NaCl, 0.01% Tween20, with a protease inhibitor cocktail tablet (Roche). In addition, each compound was added to a final concentration of 10 μM to the lysis buffer. The insoluble pellets were resuspended in 5 mL of lysis buffer. SDS-PAGE was used to evaluate 10 μL of each sample of the cleared lysate and resuspended pellet.

Expression and purification of QscR

The same E. coli strain BL21 (DE3) pLysS (Invitrogen) containing pET3a-qscR (Oinuma and Greenberg, 2011) was precultured in LB broth with 100 μg/mL ampicillin and 34 μg/mL chloramphenicol at 37°C overnight. 1% of the pre-culture was added to 0.5 L LB containing 100 μg/mL ampicillin in each of 8 flasks. Cultures were grown at 37°C to an OD₆₀₀ of 0.4-0.6, and the flasks were cooled on ice for 2 min. 3OC6-HSL was added to 50 μM and expression was induced using 0.5 mM IPTG (GoldBio). Cells were grown for 17 hr at 17°C, and harvested and stored at −80°C.

Buffers included lysis buffer (LYS), low salt buffer (LS), and LS with 1 M NaCl. Lysis buffer: 25 mM Tris, pH 7.8 at 4°C, 100 mM NaCl, 1 mM EDTA, 1 mM DTT, 10% glycerol, 0.01% Tween20, with a protease inhibitor cocktail tablet (Roche) and 100 μM of the desired compound (either 3OC12-HSL, agonist, or antagonist). Low salt buffer: 25 mM Tris, pH 7.8 at 4°C, 100 mM NaCl, 1 mM EDTA, 1 mM DTT, 10% glycerol, 0.01% Tween 20, 10 μM final concentration of the desired compound.

QscR was purified by resuspending cells in 100 mL of LYS buffer and lysing them using sonication and centrifugation at 17,000 RPM for 30 min at 4°C. Ammonium sulfate (AS) fractionation of the cleared lysate was conducted at 45% AS saturation with stirring slowly for 30 min and incubation on ice for at least 4 hours. The precipitate was collected by centrifugation at 7,000 RPM for 10 min. It was resuspended in 30 mL of LYS buffer (with compound) and dialyzed against 2L of LS buffer. The sample was sterile filtered.

For chromatography, a tandem-connected GE HiTrap Q 5 mL column and GE Heparin 5 mL column was used. It was pre-equilibrated with LS buffer, loaded with the sample and washed to remove unbound proteins. The Q-column was disconnected, and QscR was eluted from the heparin column with a NaCl gradient from 0.1 M to 1.0 M. Fractions containing QscR were pooled and concentrated to 4 mL using a stirred-cell concentrator. Size exclusion chromatography (SEC) was performed (Superdex 75, GE Healthcare) in LS buffer, and pure QscR fractions were combined.

The QscR ligand binding domain (LBD) was obtained from full-length QscR (as above) but without protease inhibitors. 2 mg of trypsin was added to each 20 mL of lysate and incubated for 1.5 hr. The cleaved protein was purified by anion-exchange chromatography (HiTrap Q, GE Healthcare), and the LBD fractions were pooled and concentrated for further purification using SEC (Superdex 30, GE Healthcare). MALDI mass spectrometry was used to confirm the cleavage site in QscR at residue Arg167.

QscR-ligand structure determination and analyses

Each QscR–ligand complex was concentrated to approximately 3 mg/mL. Using vapor diffusion at 4 C° QscR–S3 crystals were grown from added 0.2 M potassium citrate, 20 w/v PEG 3350 and 5% glycerol, and QscR–C12 from added 0.2 M sodium formate, 20 w/v PEG 3350. Diffraction data were collected at beamline 4.4.2 (Advanced Light Source, Berkeley), and processed using d*trek (Pflugrath, 1999). The structures were solved at a resolution of 2.5 Å by molecular replacement using QscR as a model (Lintz et al., 2011) with the PHASER module of the CCP4 or PHENIX software suite (Bailey, 1994, Adams et al., 2010). The model was built using COOT (Emsley and Cowtan, 2004) and refinement was conducted using PHENIX. Group TLS refinement was used in the refinement as there were large regions of chain B in QscR–S3, with much higher than average B-factors. Several sections of chain B are poorly defined due to this disorder.

The structures were analyzed for stereochemical and geometrical quality with the validation RCSB server (Rose et al., 2017). The root mean squared deviation (R.M.S.D.) values were calculated using PyMol and COOT (Emsley and Cowtan, 2004) and contacts were identified using CCP4. Figures were made using PyMol and Photoshop (Adobe).

Limited Proteolysis of QscR

QscR samples, purified with each desired compound, were diluted in LS buffer to a concentration of 17.98 μM and incubated with 250 μM of the compound for one hr at room temperature (RT). A 1 μM stock of trypsin was added at a 1:125 protein:protease ratio, and digestion performed at RT. Samples were drawn every four minutes and the reaction was quenched with the addition of 2X SDS BME and boiling for 5 min. Reactions were resolved using SDS-PAGE. After Coomassie blue staining, the gels were imaged using the Li-Cor Odyssey imager at 700 nm, and quantified using ImageJ software. The rate and half-life (t_1/2) values were obtained using a one-phase decay curve-fitting model implemented in the GraphPad Prism program (Eq. 3):

$$\mathrm{Y}=\left(\mathrm{Y}0–\text{Plateau}\right)\ast \text{exp}\left(-\mathrm{k}\ast \mathrm{X}\right)+\text{Plateau}.$$ Eq. 3

EMSA analysis of QscR with agonists and antagonists

QscR that was purified with compounds was concentrated to 0.55 mg/mL in sodium phosphate buffer. 100 μL aliquots were incubated at RT with either 100 μM 3OC12-HSL, 100 μM S3, or 100 μM Q9 for one hr. Samples were loaded onto a 6% non-denaturing TBE polyacrylamide gel (37.5:1 acrylamide:bis-acrylamide ratio) and electrophoresed for 3 hr at 75 V at RT in 0.2 X TBE. Gels were stained using Coomassie blue.

Circular Dichroism

All of the CD spectra were recorded on a JASCO-J815 spectrophotometer with Peltier temperature control. Spectral scans were performed at 4°C and thermal denaturation data were collected at 222 nm using a temperature range of 4 to 95°C with a ramp rate of 1.5 °C/min. For the QscR-LBD complexes with each bound compound, samples of the QscR LBD that had been purified with 3OC6-HSL were first concentrated to 10 μM, and then dialyzed into 25 mM sodium phosphate buffer, pH 7.5 and 100 mM NaCl with 5 μM of the desired compound. The full-length QscR samples were prepared similarly. QscR–3OC6-HSL at 140.5 μM was incubated for 1 hr on ice with the desired compounds (3OC12-HSL, S3, CL, Q9, or R6) at a final concentration of 500 μM. Prior to acquisition of the spectra, the samples were diluted to 8 μM in 25 mM sodium phosphate, pH 7.5 and 100 mM NaCl.

Analytical size exclusion chromatography

QscR was purified with 3OC6-HSL and concentrated to either 20 μM or 140.5 μM. Samples were soaked with either 3OC12-HSL, S3, CL, Q9, or R6 at a final concentration of 500 μM at 4°C for 2 hours. Samples were loaded onto an analytical Superdex 75 (10/300) column equilibrated in LS buffer at 4°C and the resulting traces normalized at the peak.

Crosslinking mass spectrometry of QscR-ligand complexes

Samples of 10 μM QscR, purified with 3OC6-HSL in buffer (50 mM sodium phosphate, pH 7.5, 100 mM NaCl, 1 mM EDTA, 0.5 mM TCEP, 0.01% Tween20, 10% glycerol), were soaked with 500 μM of each desired compound for 3 hr at 4°C. A 50/50 sample of ²H-labeled and unlabeled DSS was added to a 10-fold excess. After 15 min at RT, the reactions were quenched using Tris, pH 7.4 at 50 mM. Samples were resolved using a 4-12% gradient Nupage™ gel in a Bis-Tris buffer and stained using Coomassie blue. Bands of approximately 50 kDa, 26 kDa, and 24 kDa were excised and processed as described (Dzieciatkowska et al., 2014). Proteins were then digested overnight at RT with sequencing grade modified trypsin (Promega).

LC-MS/MS analyses were performed in the Proteomic Mass Spectrometry Facility at the U. Colorado-Anschutz Medical Campus. Each sample was analyzed in triplicate with technical duplicates. Nanoflow reversed-phase LC-MS/MS was performed using an Eksigent nanoLC-2D system (Eksigent) coupled to LTQ Orbitrap-Velos mass spectrometer (Thermo Fisher) as described previously (Hill et al., 2015). Data acquisition was performed using Xcalibur™ (Version 2.1) software. Peak lists were generated from RAW files using PAVA (UCSF). Initial searches were performed on an in-house Mascot server (Version 2.3, Matrix Science), against both the SwissProt Database, and a custom database containing QscR, common lab contaminants, and randomized decoy sequences.

Crosslink searches were performed with Protein Prospector. Search conditions required trypsin specificity with up to 4 missed cleavages. Variable modifications included incorrect monoisotopic peak assignments, and the light (¹H) and heavy (²H) full length and dead-end DSS mass additions. MS and MS/MS search tolerance was set to 10 and 25 ppm, respectively. Results were filtered according to parameters defined previously (Trnka et al., 2014). Crosslinks with a score difference above 0, and an expect value below 1 were considered, resulting in a false discovery rate of 0.48%.

QscR-DNA EMSA

Electrophoretic mobility shift assays (EMSA) were used to measure the affinity of QscR-ligand complexes for the PA1897 promoter region. A fluoresceine containing palindromic 31 base-pair DNA duplex (Oinuma and Greenberg, 2011) of sequence: 5’ 6FAM- TGGACAACCTGCCCGATCGGGCAGGTTGTCC- 3’ was purified using DEAE ion-exchange chromatography. QscR, purified in the presence of either 3OC12-HSL, S3, CL, Q9, or R6, was added to 1 nM DNA in increasing concentrations in LS buffer. After incubation for 1 hr at 4°C, the samples were electrophoresed for 45 min at 4°C using nondenaturing PAGE (as above). Gels were imaged using the Typhoon PhosphorImager (488 nm excitation, 526 BP filter) and quantified using ImageQuant software. Under ligand depletion conditions, binding curves were fit with Eq. 4:

$$Y=\left(\mathit{Bmax}\right)\ast \left(\frac{\left(\mathit{DNA}+x+\text{KD}\right)-\surd \left(\left({\left(\mathit{DNA}+x+\text{KD}\right)}^2\right)-\left(4\ast x\ast \mathit{DNA}\right)\right)}{2\ast \mathit{DNA}}\right)$$ Eq. 4

where B_max is the maximum binding, DNA is the concentration of the DNA and [P] is the varying concentration of protein added. To compare potential cooperativity, the binding curves were fit to a single-site cooperative binding isotherm (Eq. 5),

$$\mathrm{Y}=\left({\left[\mathrm{P}\right]}^{\mathrm{n}}/{\mathrm{K}}_{\mathrm{D}}\right)/\left(1+{\left[\mathrm{P}\right]}^{\mathrm{n}}/{\text{ K}}_{\mathrm{D}}\right)$$ Eq. 5

where [P] is the total protein concentration, n is the Hill coefficient and Y is the fraction bound.

The K_D values are the mean of at least three independent experiments.