Structural determinants driving homoserine lactoneligand selection in the Pseudomonas aeruginosaLasR quorum-sensing receptorAmelia R. McCreadya, Jon E. Paczkowskia, Brad R. Henkeb, and Bonnie L. Basslera,c,1

aDepartment of Molecular Biology, Princeton University, Princeton, NJ 08544; bOpti-Mol Consulting, LLC, Cary, NC 27513; and cHoward Hughes MedicalInstitute, Chevy Chase, MD 20815

Contributed by Bonnie L. Bassler, November 12, 2018 (sent for review October 9, 2018; reviewed by Helen E. Blackwell and Matthew B. Neiditch)

Quorum sensing is a cell–cell communication process that bacteriause to orchestrate group behaviors. Quorum sensing is mediatedby signal molecules called autoinducers. Autoinducers are oftenstructurally similar, raising questions concerning how bacteria dis-tinguish among them. Here, we use the Pseudomonas aeruginosaLasR quorum-sensing receptor to explore signal discrimination. Thecognate autoinducer, 3OC12 homoserine lactone (3OC12HSL), is a morepotent activator of LasR than other homoserine lactones. However,other homoserine lactones can elicit LasR-dependent quorum-sensingresponses, showing that LasR displays ligand promiscuity. We identifymutants that alter which homoserine lactones LasR detects. Substitu-tion at residue S129 decreases the LasR response to 3OC12HSL, whileenhancing discrimination against noncognate autoinducers. Con-versely, the LasR L130F mutation increases the potency of 3OC12HSLand other homoserine lactones. We solve crystal structures of LasRligand-binding domains complexed with noncognate autoinducers.Comparison with existing structures reveals that ligand selectivity/sensitivity is mediated by a flexible loop near the ligand-binding site.We show that LasR variants with modified ligand preferences exhibitaltered quorum-sensing responses to autoinducers in vivo. We sug-gest that possessing some ligand promiscuity endows LasR with theability to optimally regulate quorum-sensing traits.

crystal structure | LasR | Pseudomonas aeruginosa | quorum sensing |homoserine lactone

Quorum sensing is a cell–cell communication process thatenables bacteria to collectively control behavior (1). Quo-

rum sensing relies on the production, release, and detection ofextracellular signal molecules, called autoinducers (2–4). At lowcell density, autoinducer concentration is low, and bacteria act asindividuals. As cell density increases, autoinducer concentrationalso rises. Under this condition, autoinducers bind to cognatereceptors, initiating the population-wide regulation of genes un-derlying collective behaviors.Many species of Gram-negative bacteria use acylated homo-

serine lactones (HSLs) as autoinducers (5–9). HSL autoinducerspossess identical lactone head groups, but they vary in acyl chainlength and decoration. The chain modifications promote speci-ficity between particular HSL autoinducers and partner recep-tors (10). There are two kinds of HSL receptors. First, there areLuxR-type receptors, which are cytoplasmic HSL-binding tran-scription factors that possess variable ligand-binding domains(LBD) and well-conserved helix-turn-helix DNA-binding do-mains (DBD) (11, 12). There are also LuxN-type receptors,which are membrane-spanning two-component signaling pro-teins that bind HSLs in their periplasmic regions and transduceinformation regarding ligand occupancy internally by phos-phorylation/dephosphorylation cascades (1, 13, 14).Ligand specificity has been examined in the founding member

of the LuxN receptor family from Vibrio harveyi (15). LuxN isexquisitely selective for its cognate autoinducer 3OHC4HSL.Specific amino acids that confer selectivity for chain length andfor chain decoration were identified in the predicted LuxN

transmembrane-spanning region. Longer HSLs competitivelyinhibit LuxN, suggesting that, in mixed-species consortia, V.harveyi monitors the vicinity for competing species, and in re-sponse to their presence, exploits LuxN antagonism to delay thelaunch of its quorum-sensing behaviors, thus avoiding loss ofexpensive public goods to nonkin.Some analyses of HSL preference in LuxR-type receptors have

also been performed (16). TraR from Agrobacterium tumefaciensexcludes noncognate HSLs (11, 17–19), LasR from Pseudomonasaeruginosa detects several long-chain HSLs (20), and SdiA fromEscherichia coli is highly promiscuous and avidly responds toHSLs with variable chain lengths (21–23). Comparison ofstructures of the LasR and TraR LBDs suggests that increasedhydrogen bonding to the ligand in TraR, compared with LasR,accounts for the selectivity difference (20). Here, we systemati-cally explore the LasR response to 3OC12HSL and noncognateHSLs. We use mutagenesis to establish the amino acid deter-minants that enable LasR to discriminate between HSLs. Weidentify LasR S129 as an amino acid residue that, when altered,

Significance

Pseudomonas aeruginosa uses quorum-sensing–mediatedcommunication to coordinate group behaviors, including bio-film formation and virulence. Quorum sensing relies on extra-cellular signal molecules called autoinducers. In P. aeruginosa,the LasR receptor binds the cognate autoinducer 3OC12-homoserine lactone. Using genetic, biochemical, and structuralapproaches, we define a molecular basis underlying LasR li-gand specificity. Mutations in key residues alter LasR bindingpreferences for cognate and noncognate autoinducers, alteringP. aeruginosa quorum-sensing responses. We solve crystalstructures of the LasR ligand-binding domain complexed withdifferent autoinducers and show that LasR accommodatesthese ligands via a flexible loop near the ligand-binding region.These findings establish a mechanism a receptor can use todistinguish between related ligands, an issue fundamental tosystems beyond quorum sensing.

Author contributions: A.R.M., J.E.P., and B.L.B. designed research; A.R.M. and J.E.P. per-formed research; A.R.M., J.E.P., and B.R.H. contributed new reagents/analytic tools;A.R.M., J.E.P., B.R.H., and B.L.B. analyzed data; and A.R.M., J.E.P., B.R.H., and B.L.B. wrotethe paper.

Reviewers: H.E.B., University of Wisconsin–Madison; and M.B.N., Rutgers University–NewJersey Medical School.

The authors declare no conflict of interest.

Published under the PNAS license.

Data deposition: The atomic coordinates and structure factors have been deposited in theProtein Data Bank, www.pdb.org [PDB ID codes 6MVN (LasR LBD bound to 3OC10HSL) and6MVM (LasR LBD bound to 3OC14HSL)].1To whom correspondence should be addressed. Email: bbassler@princeton.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1817239116/-/DCSupplemental.

Published online December 17, 2018.

www.pnas.org/cgi/doi/10.1073/pnas.1817239116 PNAS | January 2, 2019 | vol. 116 | no. 1 | 245–254

MICRO

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Aug

ust 1

1, 2

021

decreases the LasR response to 3OC12HSL, and consequently re-stricts the set of HSLs capable of activation. In contrast, we find thatLasR L130F improves the LasR response to 3OC12HSL, and as aresult, allows LasR to respond to a broader set of HSLs than wild-type LasR. We also investigate how LasR ligand sensitivity andpromiscuity affect the timing and strength of quorum-sensing controlof P. aeruginosa behaviors. Finally, we solve crystal structures of theLasR LBD L130F bound to noncognate autoinducers to establish thestructural basis underlying ligand selectivity and sensitivity. We findthat a flexible loop located near the ligand-binding pocket promotesligand promiscuity in the wild-type protein. This loop exists in SdiA,which is promiscuous, but not in TraR, which is highly specific for itscognate ligand. We propose that evolution has established a balancebetween ligand discrimination and ligand sensitivity in LasR. BecauseLasR requires higher concentrations of noncognate autoinducersthan its cognate autoinducer for activation, this trade-off could allowLasR to robustly respond to its own signal molecule, even in thepresence of other bacteria that are producing HSLs. Nonetheless,because LasR can detect high concentrations of other HSLs, P.aeruginosa would be capable of reacting to the presence of theseother bacteria when it is outnumbered.

ResultsLasR Responds to Multiple HSL Autoinducers. To investigate thepreference LasR displays for different HSLs, we employed aplasmid reporter system in which transcription from a LasR-controlled promoter fused to luciferase (plasB-lux) was assessedin E. coli. Arabinose-inducible lasR was cloned on a second plasmid(24, 25). A set of HSLs differing in both carbon chain length and infunctionality at the C-3 position was synthesized using a modifica-tion of a previously reported method (SI Appendix) (26). Fig. 1Ashows reporter output following addition of 100 nM of the cognateautoinducer, 3OC12HSL, and four other HSLs (3OC14HSL,3OC10HSL, 3OC8HSL, and 3OC6HSL). All five HSLs activatedLasR, but to differing levels: 3OC12HSL, 3OC14HSL, and

3OC10HSL elicited maximal LasR activity, and 3OC8HSL and3OC6HSL stimulated 7-fold and 18-fold less activity, respectively.Using dose–response analyses, we examined these five compoundsand seven other HSLs harboring different functionalities on the C-3 carbon in combination with the various chain lengths (Table 1 andSI Appendix, Table S1). The most potent ligand was 3OC12HSL,with an EC50 of 2.8 nM; 3OC14HSL was half as potent, with anEC50 of 6.2 nM. From there, the EC50 values followed the order3OC10HSL < 3OC8HSL < 3OC6HSL. SI Appendix, Table S1 showsdata for the C-3 modifications, and that the ketone versions of themolecules are the most potent for every chain length. For thisreason, we used the five HSLs shown in Table 1 for much of theremainder of this work. In all assays, appropriate concentrations ofHSLs were chosen based on EC50 data from the dose–responseassays (Table 1 and SI Appendix, Table S1).We measured in vivo LasR activity in response to the test

HSLs using an elastase assay (27). Elastase is encoded by lasBand, thus, is driven by the same promoter as we employed in theabove recombinant E. coli lux reporter assay. For the elastaseanalyses, we used a ΔlasI P. aeruginosa strain that makes noendogenous 3OC12HSL. When supplied at 100 nM, all of the testcompounds elicited some elastase activity, but 3OC12HSL stim-ulated the highest elastase production (Fig. 1B). To ensure that100 nM was an appropriate ligand concentration for this as-say, we measured elastase activity at 3OC12HSL concentrationsranging from 1 nM to 1 μM (SI Appendix, Fig. S1). We do notethat in P. aeruginosa, 3OC14HSL and 3OC8HSL stimulated loweractivity than expected based on their EC50 values in E. coli. In-deed, as shown below, these two molecules had reduced activityin all assays in all P. aeruginosa strains used here. While we donot know the underlying molecular mechanism, we suspect thatefflux pumps present in P. aeruginosa but not in E. coli could beresponsible. For example, the P. aeruginosa MexAB-OprM effluxpump can eliminate homoserine lactones as evidenced by re-ductions in EC50 values in the ΔmexAB-oprM mutant (28).

Fig. 1. LasR is activated by multiple homoserine lactone autoinducers. (A) LasR-dependent bioluminescence was measured in E. coli. Arabinose-inducibleLasR was produced from one plasmid and the plasB-lux reporter construct was carried on a second plasmid; 0.1% arabinose was used for LasR induction. (B)Elastase activity was measured from ΔlasI P. aeruginosa using elastin-Congo red as the substrate. In A and B, 100 nM of the designated HSLs were tested. Twotechnical replicates were performed for each biological sample and three biological replicates were assessed. Error bars denote SDs of the mean. Paired two-tailed t tests were performed comparing each compound to the DMSO control. *P < 0.05, **P < 0.01, ***P < 0.0001. (C) Comparison of LasR LBD protein levelsin whole-cell lysates (W) and in the soluble fractions (S) of E. coli cells harboring DNA encoding the LasR LBD on a plasmid; 1 mM IPTG was used for LasR LBDinduction and either 1% DMSO or 10 μM of the indicated HSL was supplied. In all lanes, protein from 0.05 OD of cells was loaded. Results are representative ofthree trials. (D) Thermal-shift analyses of purified LasR LBD bound to 3OC10HSL (Left), 3OC12HSL (Center), and 3OC14HSL (Right) without (designated DMSO)and following supplementation with an additional 10 μM of the indicated HSLs. Normalized fluorescence represents the first derivative of the raw fluo-rescence data (59). Each line represents the average of three replicates.

246 | www.pnas.org/cgi/doi/10.1073/pnas.1817239116 McCready et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 1

1, 2

021

Nonetheless, our results indicate that with respect to theHSLs we tested, LasR is most responsive to its cognateautoinducer 3OC12HSL. However, noncognate HSLs can in-duce production of the quorum-sensing product elastase, andpresumably other quorum-sensing–regulated outputs.LasR and most other LuxR-type receptors fold around their

cognate HSL ligands. Thus, they are only soluble, capable ofdimerizing, and capable of binding DNA when ligand is present(18, 29). The results in Fig. 1 A and B suggest that LasR can foldaround HSLs in addition to 3OC12HSL. To verify this notion, wetested whether LasR could be solubilized by noncognate HSLs.To do this, we grew E. coli producing the LasR LBD in thepresence of each of the HSL compounds in our collection (Fig.1C shows the 5 test compounds and SI Appendix, Fig. S2A showsthe 11 other compounds in the collection). Consistent withprevious results, in the absence of any ligand (DMSO control),the LasR LBD is present in the whole-cell lysate, but not in thesoluble fraction, indicating that the protein is insoluble (30, 31).With respect to the main HSL test compounds, all except for3OC6HSL solubilized the LasR LBD (Fig. 1C). Nonetheless, wecould only purify to homogeneity the LasR LBD bound to theligands containing the longer acyl chains: 3OC12HSL, 3OC14HSL,and 3OC10HSL. Taken together, the results in Fig. 1 A–C suggestthat although 3OC8HSL and 3OC6HSL can bind to and activateLasR, their interactions must be less stable than ligands withlonger acyl chains. To test this hypothesis, we performed thermal-shift analyses on LasR LBD–ligand complexes without and withthe addition of either the same or a different HSL. The LasR LBDbound to 3OC10HSL, 3OC12HSL, and 3OC14HSL had apparentmelting temperatures of 42.3 °C, 49.1 °C, and 50.5 °C, respectively(Fig. 1D, black lines), showing that LasR stability increases withincreasing ligand chain length. Notably, the LasR LBD is slightlymore stable when bound to the noncognate ligand 3OC14HSL,than when bound to the cognate ligand 3OC12HSL. The dis-crepancy between the enhanced stability of purified LasRLBD:3OC14HSL relative to LasR LBD:3OC12HSL in the thermal-shift assay and the higher EC50 LasR displays for 3OC14HSLcompared with 3OC12HSL could result from increased hydrophobicinteractions with the C14 chain in the stably purified complex that donot drive LasR activation.Exogenously supplied autoinducers can shift the apparent

melting temperature of an existing receptor–ligand complex ifthe added ligand has the ability to stabilize the unfolding proteinas it releases the prebound ligand. Importantly, exogenouslysupplied HSLs can only stabilize the LasR LBD if they haveaffinities that are equal to or higher than the originally boundligand (25). For the LasR LBD:3OC10HSL complex, exoge-nously added 3OC10HSL, 3OC12HSL, and 3OC14HSL stabilizethe LasR LBD, increasing the apparent melting temperature3.4 °C, 6.4 °C, and 5.5 °C, respectively (Fig. 1D). In contrast, theLasR LBD:3OC12HSL and the LasR LBD:3OC14HSL com-plexes could be further stabilized by exogenously supplied

3OC12HSL and 3OC14HSL, but not by 3OC10HSL (Fig. 1D).While the short acyl chain HSLs could activate LasR with highmicromolar EC50 values in the plasB-lux reporter assay, 3OC6HSLand 3OC8HSL did not stabilize the LasR LBD protein sufficientlyin E. coli for purification, presumably due to their low affinities (Fig.1C and Table 1). Thus, they could not be tested in the traditionalthermal-shift assay. They also did not enhance the stabilization ofthe LasR LBD prebound with other ligands, analogous to the in-ability of 3OC10HSL to stabilize the LasR LBD:3OC12HSL com-plex as it melted (Fig. 1D). Taken together, our results indicate thatin an environment containing a mixture of HSL autoinducers, LasR,while capable of detecting several HSLs, will preferentially detectlong-chain HSLs, with superior preference for its own autoinducer3OC12HSL, followed closely by 3OC14HSL.

LasR S129 Mutations Alter Ligand Sensitivity and Ligand Discrimination.Our above results show that LasR detects multiple HSL ligandswith different preferences. To alter those preferences, we muta-genized the LasR ligand-binding pocket using the existing LasRLBD crystal structure as a guide (29). Among the set of mutantsthat we made, we identified one residue of interest: LasR S129.Previous work demonstrated that alteration of serine to alanine atthis position transformed some LasR activators into inhibitors,and vice versa (32, 33). Moreover, the LasR LBD structure sug-gests that S129 is part of the network that interacts with the ligandacyl chain (29). Finally, S129 is well conserved across LuxR-typereceptors, so we predicted that it could play a crucial role in ligandbinding (34). We constructed LasR S129C, S129W, S129F, S129T,and S129M and examined their activities in the E. coli plasB-luxreporter assay, using 1 μM of each compound (Fig. 2A). Wild-typeLasR and LasR S129C showed some response to every testcompound. Each of the other four mutants failed to respond toone or more of the HSL test compounds (Fig. 2A). We usedWestern blot analysis to confirm that all of the LasR S129 mutantproteins were produced at similar levels (SI Appendix, Fig. S2B),demonstrating that the phenotypes are due to alterations in re-sponses to ligands. Dose–response analyses with the complete setof HSLs enabled us to obtain EC50 values for each mutant. Thesevalues reveal that each LasR S129 mutant displays reduced activitywith 3OC12HSL compared with wild-type LasR (Table 1 and SIAppendix, Table S1). Moreover, there exists a consistent rela-tionship between the activity LasR mutants display with 3OC12HSLand the range of HSLs to which they respond. Mutants with thehighest activity with 3OC12HSL interact with a larger range of HSLsthan mutants with reduced activity with 3OC12HSL. For example,LasR S129M had the highest EC50 for 3OC12HSL, and it showedno response to any other HSL in our set of test compounds (Table 1and SI Appendix, Table S1). Conversely, LasR S129C respon-ded to all of the HSLs we tested except 3OHC8HSL, C8HSL,and 3OHC6HSL, and it has the lowest EC50/highest activitywith 3OC12HSL (Table 1 and SI Appendix, Table S1). Withrespect to EC50 for 3OC12HSL, our alleles followed the pattern:wild-type < LasR S129C < LasR S129W < LasR S129F < LasRS129T < LasR S129M. They followed the exact reverse order interms of breadth of response to different HSLs. These findings,and the location of S129 in the LasR ligand-binding pocket,support the proposal that substitutions disrupt a crucial hy-drogen bond between S129 and the ketone in the autoinduceracyl chain, thus negatively affecting ligand sensitivity (32).To explore the hydrogen-bonding role of the amino acid at

position 129 on ligand selection, we focused on LasR S129F. Ourrationale was that a serine to phenylalanine substitution elimi-nates hydrogen bonding and, moreover, because the LasR S129Fmutant has an intermediate phenotype in our assays (i.e., not thestrongest and not the weakest), this allele would allow us toprobe activity in vitro and in vivo. First, we examined the con-sequences of the LasR S129F alteration in vivo by incorporatingthe lasR S129F mutation onto the chromosome of the ΔlasI

Table 1. EC50 values (nM) for LasR HSL compounds in the plasB-lux assay

LasR allele

Homoserinelactone

Wild-type S129C S129W S129F S129T S129M L130F

3OC12HSL 2.8 5.9 77 177 870 6,600 2.03OC14HSL 6.2 12 13 41 4,100 NR 5.33OC10HSL 8.0 32 2,600 4,980 4,500 NR 5.03OC8HSL 900 2,900 2,200 NR NR NR 813OC6HSL 2,400 4,400 NR 55,000 NR NR 1,400

NR, Nonresponsive; 95% Confidence intervals for EC50 values are providedin SI Appendix, Table S2.

McCready et al. PNAS | January 2, 2019 | vol. 116 | no. 1 | 247

MICRO

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Aug

ust 1

1, 2

021

P. aeruginosa strain and assaying elastase activity (Fig. 2B). Weused 10 μM of each test HSL compound because increased li-gand concentration is required to activate LasR S129F relative towild-type LasR (Table 1). LasR S129F was capable of promotingelastase production, but only in response to a subset of the HSLsthat were capable of stimulating wild-type LasR and, as noted, inevery case, at a higher ligand concentration than required forwild-type LasR.To investigate the mechanism underlying the reduction in

LasR S129F response to the HSL test compounds, we assessedwhether it could be solubilized by them. Fig. 2C shows that,similar to wild-type LasR, LasR S129F is not solubilized by ourshortest test compound 3OC6HSL, but unlike wild-type LasR,LasR S129F is also not solubilized by 3OC8HSL. These resultssupport our above finding regarding the inability of 3OC8HSL toelicit a response from LasR S129F and they underpin the lowsensitivity LasR S129F displays for 3OC6HSL (Fig. 2A and Table1). To evaluate the stability of LasR S129F, we used thermal-shift measurements. The LasR S129F:3OC12HSL complex hadan apparent melting temperature of 35 °C showing that it ismarkedly less stable than the wild-type LasR:3OC12HSL com-plex (Figs. 1D and 2D). Both 3OC12HSL and 3OC14HSL stabi-lized LasR S129F during melting and release of prebound3OC12HSL. Unlike wild-type LasR (Fig. 1D), LasR S129F pre-ferred 3OC14HSL over 3OC12HSL (Fig. 2D), consistent with itsfourfold lower EC50 for 3OC14HSL than for 3OC12HSL (Table1). These findings indicate that the loss of hydrogen bondingbetween the C-1 ketone and the S129 residue can be compen-sated by enhanced hydrophobic interactions with the C14 chain.

The LasR L130F Substitution Increases LasR Ligand Sensitivity. Basedon our results with the LasR S129 mutants, we suspected thatother LasR autoinducer binding-pocket residues located in thevicinity of the ligand acyl chain could participate in ligand se-lection. For this reason, we exchanged L130 and L128 for phe-nylalanine residues. As above, we tested their responses to theset of representative HSLs using the E. coli plasB-lux reporterassay (Fig. 3, Left, Table 1, and SI Appendix, Table S1). Table 1shows that, surprisingly, LasR L130F exhibited a lower EC50 forevery autoinducer tested. Shorter-chain HSLs showed the mostdramatic increases in activity. LasR L128F also conferred improvedsensitivity to some short-chain HSLs (SI Appendix, Table S1), but itsphenotype was less pronounced than that of LasR L130F, so wefurther analyzed the LasR L130F mutant here. Fig. 3, Left, shows thewild-type LasR and the LasR L130F responses to a low concentration(50 nM) of the test HSLs. At this concentration, LasR L130F wasroughly equivalent to wild-type LasR with respect to the response to3OC12HSL, 3OC14HSL, and 3OC10HSL. However, LasR L130F wasapproximately fivefold more responsive to 3OC8HSL than wild-typeLasR and it was modestly more responsive to 3OC6HSL. In vivoexamination of LasR L130F ligand responses support these findings.When we assayed low concentrations (50 nM) of the test compounds,wild-type LasR stimulated elastase production only in response to3OC12HSL, whereas LasR L130F responded to 3OC14HSL,3OC10HSL, and 3OC8HSL in addition to 3OC12HSL (Fig. 3, Right).We used thermal-shift analyses to explore the mechanism

underlying the increased sensitivity of the LasR L130F mutantfor HSLs. We used 3OC12HSL and 3OC14HSL as our test li-gands. Compared with the wild-type LasR LBD, the LasR LBDL130F is more stable when bound to each ligand (Fig. 4A),

Fig. 2. LasR S129 mutants alter LasR responses to HSL autoinducers. (A) Wild-type LasR and LasR S129 mutant bioluminescence from the E. coli plasB-luxreporter; 1 μM of the indicated HSLs were provided (see Fig. 1A for detail). (B) Wild-type LasR and LasR S129F-driven elastase activity in ΔlasI P. aeruginosa;10 μM of the indicated HSLs were provided (see Fig. 1B for detail). In A and B, two technical replicates were performed for each biological sample and threebiological replicates were assessed. Error bars denote SDs of the mean. Unpaired t tests were performed to compare each mutant LasR-HSL combination tothat of wild-type LasR with the same compound. Due to space constraints, statistics for A are provided in SI Appendix, Table S3. **P < 0.01, ***P < 0.0001. (C)Comparison of LasR S129F LBD protein levels in whole-cell lysates (W) and in the soluble fractions (S) of E. coli cells harboring DNA encoding the LasR S129FLBD on a plasmid; 1 mM IPTG was used for LasR S129F LBD induction and either 1% DMSO or 10 μM of the indicated HSL was supplied. In all lanes, proteinfrom 0.05 OD of cells was loaded. Results are representative of three trials. (D) Thermal-shift analyses of purified LasR S129F LBD bound to 3OC12HSL (Left) and3OC14HSL (Right) without (designated DMSO) and following supplementation with an additional 10 μM of the indicated HSLs. Normalized fluorescencerepresents the first derivative of the raw fluorescence data (59). Each line represents the average of three replicates.

248 | www.pnas.org/cgi/doi/10.1073/pnas.1817239116 McCready et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 1

1, 2

021

suggesting that the L130F alteration increases the overall sta-bility of the LasR protein. We exploited this feature to suc-cessfully purify the LasR LBD L130F bound to 3OC8HSL. Asmentioned above, we could not purify the wild-type LasR LBDbound to 3OC8HSL. We performed thermal-shift analyses on theLasR LBD L130F bound to 3OC8HSL, 3OC10HSL, 3OC12HSL,and 3OC14HSL, to which we added different HSLs. Similar tothe wild-type LasR LBD, exogenous addition of HSLs with longacyl chains further enhanced the stability of the LasR LBDL130F:HSL complexes compared with when HSLs with shorteracyl chains were added (Figs. 1C and 4B). For example, the LasRLBD L130F:3OC8HSL stability was enhanced by the addition of3OC8HSL, 3OC10HSL, 3OC12HSL, and 3OC14HSL, but not3OC6HSL (Fig. 4B). Protein solubility analyses track with thesefindings; long acyl chain ligands solubilize the LasR LBD L130F

protein, whereas short-chain ligands do not (Fig. 4C and SIAppendix, Fig. S2C). Consistent with our EC50 values, chainlength appears to be the most important factor driving proteinsolubility and stabilization for both the LasR LBD and the LasRLBD L130F (Fig. 4B and SI Appendix, Fig. S2C).

LasR Ligand Selectivity and Sensitivity Influence the Timing and Strengthof in Vivo Quorum-Sensing–Controlled Behaviors.LasR L130F respondsto HSLs at lower concentrations than does wild-type LasR, whereasLasR S129F requires higher concentrations (Table 1). Thus, wepredicted that introducing these alleles into P. aeruginosa shouldinfluence its quorum-sensing responses in opposing manners. Totest this prediction, we assayed pyocyanin production over time asthe quorum-sensing readout in response to 3OC12HSL, in a ΔlasI P.aeruginosa strain containing wild-type lasR, lasR S129F, or lasRL130F. For comparison, we also evaluated pyocyanin productionover time in wild-type P. aeruginosa under the same conditions (SIAppendix, Fig. S3). Guided by the EC50 value of LasR with3OC12HSL, we tested an appropriate low (50 nM) and high con-centrations (1 μM) of 3OC12HSL to enable us to observe responsesfrom the wild-type and each mutant. Consistent with their relativeEC50 values, when a low concentration of 3OC12HSL (50 nM) wasadded, the strain with LasR L130F made significantly more pyo-cyanin than the strain with wild-type LasR at every time point (Fig.5, Left). At this HSL concentration, the strain carrying LasR S129Fnever activated pyocyanin production, presumably because theconcentration of 3OC12HSL was far below the EC50 for LasRS129F (Fig. 5, Left). When the same assay was performed with 1 μM3OC12HSL, rapid and maximal pyocyanin output occurred for P.aeruginosa carrying both wild-type LasR and LasR L130F. Fur-thermore, this high ligand concentration reveals that LasR S129Fcan drive pyocyanin production in response to 3OC12HSL, albeitweakly and only after 5 h (Fig. 5, Right). These results suggest thataltering the sensitivity of LasR can change the production ofquorum-sensing products such as pyocyanin.

Fig. 3. LasR L130F displays enhanced responses to HSL autoinducers. (Left)Wild-type LasR and LasR L130F-driven bioluminescence from the E. coliplasB-lux reporter; 50 nM of the indicated HSLs were provided (see Fig. 1Afor detail). (Right) Wild-type LasR and LasR L130F-driven elastase activity inΔlasI P. aeruginosa; 50 nM of the indicated HSLs were provided (see Fig. 1Bfor detail). In both A and B, two technical replicates were performed foreach biological sample and three biological replicates were assessed. Errorbars denote SDs of the mean. Unpaired t tests were performed to compareeach LasR L130F-HSL combination to that of wild-type LasR with the samecompound. *P < 0.05, ***P < 0.0001.

Fig. 4. The LasR LBD L130F is more stable than the wild-type LasR LBD. (A) Thermal-shift analyses of purified LasR LBD (solid lines) and LasR LBD L130F(dotted lines) bound to 3OC12HSL (orange) or 3OC14HSL (green). Each line represents the average of three replicates. (B) Thermal-shift analyses of purifiedLasR LBD L130F bound to 3OC8HSL (Upper Left), 3OC10HSL (Lower Left), 3OC12HSL (Upper Right), and 3OC14HSL (Lower Right) without (designated DMSO)and following supplementation with an additional 10 μM of the indicated HSLs. In A and B, normalized fluorescence represents the first derivative of the rawfluorescence data (59). (C) Comparison of LasR L130F levels in the whole-cell lysates (W) and the soluble fractions (S) of E. coli cells harboring DNA encodingLasR LBD L130F on a plasmid (see Fig. 1C for details). Either 1% DMSO or 10 μM of the indicated HSL molecule was added.

McCready et al. PNAS | January 2, 2019 | vol. 116 | no. 1 | 249

MICRO

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Aug

ust 1

1, 2

021

Structural Basis Underlying LasR Ligand Preferences. To define themolecular basis enabling LasR to accommodate noncognateautoinducers, we determined the structures of the LasR LBDL130F bound to 3OC10HSL and 3OC14HSL. As mentioned, thestructure of the wild-type LasR LBD bound to 3OC12HSL wasreported previously (29). We used LasR LBD L130F for ourstudies, reasoning that its enhanced stability might allow crys-tallization with noncognate ligands. Indeed, we were able todetermine the structures of LasR LBD L130F:3OC10HSL andLasR LBD L130F:3OC14HSL to 2.1 and 1.9 Å, respectively (SI

Appendix, Table S4). This resolution shows unambiguous liganddensity enabling us to compare our structures with the LasRLBD:3OC12HSL structure (Fig. 6A and SI Appendix, Fig. S4 Aand B). The side chain of F130 is buried in a hydrophobic pocketdistal to the ligand-binding pocket (SI Appendix, Fig. S4 C andD). Compared with leucine, phenylalanine provides increasedhydrophobic interactions with amino acid residues L23, L30,F32, I35, L114, L118, L128, L151, and L154. These interactionsmay stabilize LasR, in turn allowing it to accommodate an ex-panded set of HSLs compared with wild-type LasR.To understand how LasR can accommodate multiple long-

chain HSL ligands in its binding site, we used the crystal struc-tures to examine the binding pocket in detail. In all of thestructures, the lactone head groups have the exact same place-ment, likely due to hydrogen bonding between the lactone ringketone moiety and residue W60 (Fig. 6 B and C). The headgroup ketone and the two ketones on the acyl chain are furtherstabilized by an extensive hydrogen-bonding network consistingof residues Y56, R61, D73, T75, W88, Y93, and S129 (Fig. 6 Band C). In the three structures, the C10, C12, and C14 chains takesimilar paths until carbon 6, where C10 and C14 diverge from thepath taken by the C12 chain. As a result, C14 and C12 fill the samevolume (SI Appendix, Fig. S4A). However, the absence of twocarbons in the C10 chain compared with the C12 chain couldaccount for the increase in EC50 and lower stability of the LasRLBD L130F:3OC10HSL complex compared with the LasRLBD:3OC12HSL and LasR LBD L130F:3OC14HSL complexes(SI Appendix, Fig. S4B).We next investigated whether there were any structural rear-

rangements in the crystals that could account for the expandedHSL binding capabilities of LasR L130F. An ∼2-Å shift occurs in

Fig. 5. Wild-type LasR, LasR S129F, and LasR L130F display distinct pyocya-nin production phenotypes in response to high and low concentrations of3OC12HSL. Pyocyanin production was measured in ΔlasI P. aeruginosa strainsover the growth curve. The y axis “Pyocyanin” is the amount of pyocyaninpigment (OD695 nm) over cell density (OD600 nm). Designations are: red,DMSO control; blue, wild-type LasR; green, LasR S129F; pink, LasR L130F.Concentrations used are: (Left) 50 nM 3OC12HSL, (Right) 1 μM 3OC12HSL.Data show the mean of three biological replicates. Error bars denote SDs ofthe mean. Two-way ANOVAs were performed to compare LasR S129F andLasR L130F to wild-type LasR under each condition. A two-way ANOVA wasalso performed to compare wild-type LasR with 3OC12HSL and with DMSOalone. *P < 0.05, ***P < 0.0001.

Fig. 6. Crystal structures of LasR LBD L130F bound to 3OC10HSL and 3OC14HSL. (A) Crystal structures of LasR LBD L130F:3OC10HSL (gold) and LasR LBDL130F:3OC14HSL (magenta) compared with the wild-type LasR LBD:3OC12HSL structure [blue, modified from PDB: 2UV0 (29)]. The Lower images show 90°rotations of the crystal structures relative to the images above. In the two Upper rightmost structures, the asterisks highlight the LasR loop region thatincludes residues 40–51 and that undergoes a conformational shift when 3OC14HSL is bound. (B) Structural comparison of LasR LBD:3OC12HSL (protein: blue,ligand: orange; modified from PDB ID code 2UVO) (29) and LasR LBD L130F:3OC10HSL crystal structures (protein: gold, ligand: magenta). Amino acids drawn instick format show important residues for lactone head binding (Y56, W60, R61, D73, T75, W88, Y93, and S129) and acyl chain binding (G38, L40, A50, I52, A70,V76, L125, and A127). (C) Structural comparison of LasR LBD:3OC12HSL (protein: blue, ligand: orange; modified from PDB ID code 2UVO) (29) and LasR LBDL130F:3OC14HSL crystal structures (protein: magenta, ligand: green). Amino acids drawn in stick format are the same as in B. Note the shift in the position ofthe loop region corresponding to residues 40–51.

250 | www.pnas.org/cgi/doi/10.1073/pnas.1817239116 McCready et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 1

1, 2

021

the loop corresponding to residues 40–51 (highlighted by aster-isks in Fig. 6A and displayed in SI Appendix, Fig. S4 D and E) inthe LasR LBD L130F:3OC14HSL structure compared with the LasRLBD L130F:3OC10HSL and the LasR LBD:3OC12HSL structures(Fig. 6 B and C), and we propose the change in the placement of thisloop underpins the altered binding capabilities of LasR L130F. Ourstructures with the different HSLs are consistent with recentlyreported data showing that the LasR loop conformation can be al-tered by synthetic agonists and, moreover, potency of the nonnativeligands correlates with the thermal stability of the complexes (35).In each of our described structures, there are two LasR mono-

mers in the asymmetric unit. In the LasR LBD L130F:3OC14HSLstructure, the crystallographic contacts made by the loop regions inmonomer A and monomer B differ slightly, while in the LasR LBDL130F:3OC10HSL structure, the monomer A and monomer B loopregions are similar, and each makes crystal contacts similar to thosein monomer B in the LasR LBD L130F:3OC14HSL structure. It waspossible that crystal contacts, rather than ligand binding, drove thedifferent loop conformations in the LasR LBD L130F:3OC14HSLstructure. Because we know that the conformations of the loopregions in the previously published wild-type LasR LBD:3OC12HSLstructure are not influenced by crystallographic contacts, we canreliably compare the positions of the loop regions between LasRLBD L130F:3OC14HSL monomer A and monomer A from theLasR LBD:3OC12HSL structure. Both are free from crystallo-graphic contacts, allowing us to conclude that the loop shifts in ourcrystal structures are ligand-induced (Fig. 6 A and C).The crystal structures show that the majority of the LasR LBD

exists as an interacting core that binds the lactone head and aportion of the acyl chain, whereas the loop consisting of residues40–51 appears to act as a deformable region that enables LasR toaccept ligands with different length acyl chains (Fig. 6 B and C).In the LasR LBD L130F:3OC14HSL structure, Y47 lies parallelto the ligand (Fig. 6C). We contrast this arrangement to theorientation of Y47 in the LasR LBD L130F:3OC10HSL andLasR LBD:3OC12HSL structures (Fig. 6C and SI Appendix, Fig.S4 E and F). In those cases, Y47 lies perpendicular to the li-gands. We suggest that the ability of Y47 to adopt multipleconformations helps accommodate the different ligands. To testthis possibility, we mutated LasR residue Y47. LasR Y47R andLasR Y47S displayed decreased sensitivity to both 3OC12HSLand 3OC14HSL in the E. coli reporter assay (SI Appendix, Fig.S5), suggesting that the loop provides interactions with the li-gand chains that foster increased protein stability, allowingLasR to be activated. As previously noted, the packing ofY47 against the ligand acyl chain shields the ligand-bindingpocket from the bulk solvent (36). We infer that alterationsat this site, and perhaps in the loop region generally, lead toprotein instability.To understand the structural basis underlying specificity and

promiscuity in this family of proteins, we compared all reportedstructures of the LBDs of LuxR-type receptors to our structureof LasR LBD L130F:3OC14HSL. The structures are SdiALBD:3OC8HSL, CviR LBD:C6HSL, TraR LBD:3OC8HSL, andQscR:3OC12HSL (Fig. 7A and SI Appendix, Fig. S6). Similar towhat we show here for LasR, SdiA is promiscuous with respect toligand selectivity (21–23). A loop similar to the one we pinpointin Fig. 6A as critical for LasR to accommodate different ligandsexists in the SdiA LBD (Fig. 7A). Similarly, the orphan P. aer-uginosa LuxR-type receptor QscR contains such a loop (SI Ap-pendix, Fig. S6), and QscR displays ligand promiscuity (37, 38).Conversely, CviR and TraR display strict specificity for theircognate autoinducers (11, 18, 39, 40). No such large loop exists inthe CviR and TraR LBD structures (Fig. 7A). Compared withthe 12-residue loop in LasR and the 10-residue loop in SdiA, thecorresponding CviR and TraR loops are eight and four residueslong, respectively. These findings are consistent with the ideathat this flexible loop confers ligand promiscuity.

In addition to the differences in the overall structures of thereceptors, we noted different conformations for the acyl chainsin the LasR LBD structures compared with those in the otherLuxR-type receptors. The autoinducer acyl chains in the CviR,TraR, and SdiA LBD structures have similar conformationswithin the ligand binding pockets terminating between residuesY88 and M89 for CviR, Y61 and F62 for TraR, and Y71 andQ72 for SdiA. In contrast, the acyl chains of the different ligandsin the LasR LBD L130F structures orient their terminal carbonstoward the opposing face of the ligand binding pocket (Fig. 7B).These different paths appear to be driven by the hydrophobicinteractions we described above for the different LasR ligands(Fig. 6 B and C). We identified eight hydrophobic residues (G38,L40, A50, I52, A70, V76, L125, and A127) that dictate the shapeof the ligand binding pocket in LasR. Indeed, these residueshave generally hydrophobic characteristics in all LuxR-typeproteins, but the size of each side chain varies (Fig. 7B). Forexample, A127 in LasR corresponds to F132 in SdiA. Thesmaller residue in LasR accommodates the altered path takenby its autoinducer, allowing the terminal carbon of the acyl

Fig. 7. A conserved flexible loop region confers promiscuity to LuxR-typereceptors. (A) Upper images: structural comparison of the LuxR-type recep-tors LasR LBD L130F:3OC14HSL (magenta) and SdiA LBD:3OC8HSL [green,modified from PDB: 4Y17 (21) PDB ID code AY17] that exhibit promiscuitywith respect to ligand binding. Lower images: structural comparison of theLuxR-type receptors CviR LBD:C6HSL [pink, modified from PDB: 3QP1 (39)]and TraR LBD:3OC8HSL [silver, modified from PDB: 1L3L (40)] that displaystrict ligand specificity. (B) Structural comparison of the protein:ligand in-terfaces for LasR LBD L130F:3OC14HSL (Upper Left, magenta), SdiALBD:3OC8HSL (Upper Right, green), CviR LBD:C6HSL (Lower Left, pink), andTraR LBD:3OC8HSL (Lower Right, silver). Residues that make important hy-drophobic side chain interactions in each protein:ligand complex are shownin stick format and named.

McCready et al. PNAS | January 2, 2019 | vol. 116 | no. 1 | 251

MICRO

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Aug

ust 1

1, 2

021

chain to bind proximal to residue A127. A bulkier residue atthis position, as in SdiA, sterically hinders this path for the li-gand, forcing the acyl chain to bend in the opposite direction.Thus, in SdiA, F132 forces the terminal carbon of 3OC8HSL tobind distally. CviR I153 and TraR M127 appear to performanalogous roles.To investigate whether a bulkier residue at position 127 would

prevent LasR from binding to HSL ligands with long acyl chains,we constructed LasR A127W and assayed it in our E. coli plasB-luxreporter assay with 3OC12HSL, 3OC10HSL, 3OC8HSL, and3OC6HSL. Fig. 8 shows that LasR A127W did not respond well to3OC12HSL; therefore, we could not determine an EC50 value.However, the LasR A127Wmutant did exhibit responses to shorter-chain HSLs. The EC50 value of LasR A127W for 3OC10HSL was3,530 nM, ∼500 times higher than wild-type LasR for 3OC10HSL(8 nM). LasR A127W had EC50 values similar to wild-type LasR for3OC8HSL and 3OC6HSL. Indeed, in the case of 3OC8HSL, theLasR A127W EC50 was somewhat lower than wild-type LasR(244 nM and 885 nM, respectively) (Fig. 8). We suggest that thelarge hydrophobic tryptophan side chain at position 127 cannotaccommodate the long-chain 3OC12HSL ligand, but the protein isresponsive to 3OC8HSL and 3OC6HSL because the enlarged aminoacid side chain in the protein compensates for the missing carbonson the ligands. This finding suggests that larger amino acid residuesat A127 improve the LasR response to shorter chain HSLs by en-abling more stable protein–ligand complexes to form.

DiscussionP. aeruginosa often occupies niches containing other bacterialspecies that produce HSL autoinducers (41). Our results show thatLasR can, with reduced sensitivity, detect noncognate HSLs and inresponse, activate transcription of quorum-sensing target genes.We suggest that determining the mechanisms that promote orrestrict ligand access to LasR is important for understanding howthe P. aeruginosa quorum-sensing response could be naturally orsynthetically manipulated. Here, we identified mutations that alterhow effectively LasR interacts with particular HSLs. ChangingS129 increases LasR specificity for long-chain ligands over short-chain ligands but at a cost of reduced ligand potency, the conse-quence of which is delayed and dampened quorum-sensing outputby P. aeruginosa (Fig. 5). In contrast, the LasR L130F alterationincreases LasR response to 3OC12HSL, but again at a cost; thistime, the penalty is a diminished ability to discriminate againstother HSL ligands. We presume, based on the pyocyanin data inFig. 5, such a change would lead to premature and increasedquorum-sensing activity. Such mistimed and misregulated releaseof public goods may not benefit P. aeruginosa.

We propose that LasR has evolved to finely balance ligandsensitivity with ligand promiscuity. We investigated the publishedsequences of hundreds of clinical isolates of P. aeruginosa andS129 and L130 are strictly conserved (42), suggesting that theyare crucial for P. aeruginosa fitness. Striking the ideal balancebetween LasR ligand-sensitivity and ligand-specificity may meanthat detection of some noncognate HSLs must be tolerated, al-though importantly, only at higher concentrations relative to thecognate autoinducer.We propose that LasR promiscuity could serve an important

function in the environment (i.e., soil) and in eukaryotic hosts whereP. aeruginosa encounters other species of bacteria. For example, inthe soil, Pseudomonas putida produces 3OC12HSL, 3OC10HSL,3OC8HSL, and 3OC6HSL (43). Klebsiella pneumoniae, which re-sides in human airways, produces C12HSL (44). LasR promiscuitymay allow P. aeruginosa to “eavesdrop” on its competitors. None-theless, wild-type LasR detects 3OC12HSL more efficiently thanother HSLs. Therefore, when P. aeruginosa is at high cell density, itis likely that 3OC12HSL outcompetes all other autoinducers. Incontrast, when P. aeruginosa is at low cell density, provision ofnoncognate ligands made by other bacterial species that are athigher cell density could induce P. aeruginosa to activate its quorum-sensing behaviors prematurely. It could be beneficial for P. aeruginosato engage in such behavior, for example, to enable the synthesisof toxic defensive products, such as pyocyanin and rhamnoli-pids that endow P. aeruginosa with an advantage over competingspecies (45, 46). Beyond defensive products, perhaps some of thequorum-sensing–controlled products produced under such condi-tions can be used exclusively by P. aeruginosa and not by com-peting species. If so, LasR promiscuity could grant P. aeruginosa a“last ditch” opportunity to survive in environments in which it isvastly outnumbered by competing species.Unlike LasR, SdiA, and QscR, other quorum-sensing recep-

tors are exquisitely specific (e.g., TraR and CviR). Thus, evolu-tion can build receptors to possess or to lack ligand specificity.Presumably, the particular niche in which a quorum-sensingbacterium resides drives the optimal receptor ligand-detectionpreference strategy. Promiscuous quorum-sensing receptors couldbe superior when quorum sensing stimulates the production ofdefensive products that aid in competition with other bacterialspecies. Conversely, a receptor possessing strict ligand speci-ficity could be optimal in cases in which quorum sensing ac-tivates production of vital public goods that require protectionfrom cheaters.Our combined genetic, biochemical, and structural work

revealed the molecular basis for noncognate autoinducer rec-ognition by LasR. A key flexible loop present in LasR, SdiA, andQscR appears to endow these receptors with the ability to bindto multiple HSL ligands. Conversely, the shorter and apparentlyless flexible loop present in TraR and CviR confers high speci-ficity. Indeed, this structure–function analysis may help explainwhy a competitive inhibitor of CviR, chlorolactone, behaves asan agonist for LasR (31). These findings are particularly en-lightening when considering attempts to design inhibitors thatspecifically target different LuxR-type receptors. The flexible loopand hydrophobic residues in LasR near the acyl chain bindingsite will need to be taken into account when developing small-molecule inhibitors that target LasR or other LuxR-type proteinsthat possess this feature. Designing molecules that target theflexibility of the loop region and that destabilize the protein couldbe explored for promiscuous receptors, such as LasR. In contrast,focused screening around molecules that resemble chlorolactonecould yield inhibitors for LuxR-type proteins that display strictspecificity for their cognate autoinducers, such as CviR.

Materials and MethodsSite-Directed Mutagenesis. Mutations in lasR were constructed on the pBAD-A-lasR plasmid (25). Primers were designed using the Agilent Quikchange

Fig. 8. LasR A127W occludes long chain HSLs from the binding pocket. E.coli plasB-lux reporter assays were performed with wild-type LasR and LasRA127W in response to 3OC12HSL, 3OC10HSL, 3OC8HSL, and 3OC6HSL. EC50

values were obtained by performing dose–response assays for each HSL andanalyzing the resulting data using nonlinear regression analysis. Two tech-nical replicates were performed for each biological sample and at least threebiological replicates were assessed. In the case of wild-type LasR, three bi-ological replicates were performed for each experiment and then combinedacross all assays.

252 | www.pnas.org/cgi/doi/10.1073/pnas.1817239116 McCready et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 1

1, 2

021

primer design tool and PCR with pFUltra polymerase (Agilent). PCR reactionswere treated with DpnI to eliminate parental plasmid DNA and the plasmidswith the mutant lasR genes were transformed into One Shot TOP10 chem-ically competent E. coli cells (Invitrogen). Reactions were plated on LB agarplates containing ampicillin (50 μg/mL) and individual mutants were verifiedvia sequencing with primers for the lasR gene (ARM203 and ARM204). Primersand strains used in this work are listed in SI Appendix, Tables S5 andS6, respectively.

P. aeruginosa Strain Construction. In-frame, marker-less lasR mutations wereengineered onto the chromosome of P. aeruginosa PA14 using pEXG2-suicide constructs with gentamicin selection and sacB counter selection(47, 48). The lasR gene and 500 bp (base pairs) of flanking regions werecloned into pUCP18 (49). Site-directed mutagenesis was performed as de-scribed above to construct point mutations in plasmid-borne lasR. The DNAcarrying the mutant lasR genes was obtained from pUCP18 by restrictionenzyme digestion with BamHI and EcoRI (New England Biolabs), and sub-sequently the fragments were ligated into pEXG2. The recombinant pEXG2plasmids were transformed into E. coli SM10λpir and, from there, the plas-mids were mobilized into ΔlasI P. aeruginosa PA14 via biparental mating (50,51). Exconjugants were selected on LB plates containing 30 μg/mL genta-micin and 100 μg/mL irgasan after overnight growth at 37° C. After recovery,5% sucrose was used to select for loss of the plasmid. Candidate mutantswere patched onto LB plates and LB plates containing 30 μg/mL gentamicinto select against the resistance marker. Colony PCR was performed on gen-tamicin sensitive patches with primers that annealed 500-bp upstream anddownstream of lasR (ARM455 and ARM456). These PCR products were se-quenced with lasR forward and reverse primers (ARM203 and ARM204).

E. coli plasB-lux Reporter Assay for LasR Activity. The development of an assaythat reports on LasR activity in response to exogenous ligands using luciferaseas the readout has been described previously (25). In brief, 2 μL of overnightcultures containing plasB-luxCDABE and pBAD-A with either wild-type lasRor mutant lasR alleles were back-diluted into 200 μL LB medium and placedinto clear-bottom 96-well plates (Corning). The plates were shaken at 30 °Cfor 4 h and 0.1% arabinose was added to each well along with a test HSL atthe concentrations designated in the text and figures. To perform dose–response analyses, 1 mM of each HSL was serially diluted threefold 10 times,and 2 μL of each dilution was added to the wells. Higher or lower HSLconcentrations were assayed when EC50 values did not fall into this range.Plates were shaken at 30 °C for 4 h. Bioluminescence and OD600 weremeasured using a Perkin-Elmer Envision Multimode plate reader. Relativelight units were calculated by dividing the bioluminescence measurement bythe OD600-nm measurement. Nonlinear regression was performed in GraphpadPrism6 to obtain EC50 values.

Elastase Assay. The P. aeruginosa PA14 ΔlasI strains carrying either wild-typeor mutant lasR genes were grown overnight with shaking at 37 °C in LBmedium. Cultures were back diluted 1:50 in 3 mL of LB and grown for anadditional 8 h with shaking at 37 °C. Strains were back-diluted 1:1,000 into3 mL of LB medium and test HSLs or an equivalent volume of DMSO wereadded to each culture. These cultures were grown overnight at 37 °C withshaking. One milliliter of each culture was removed and the cells werepelleted by centrifugation at 16,100 × g. The supernatant was removed andfiltered through a 0.22-μm filter (Millipore) and 100 μL of supernatant wasadded to 900 μL of 10 mM Na2HPO4 containing 10 mg of elastin-Congo redsubstrate (Sigma-Aldrich). These preparations were incubated at 37 °C for2 h. The mixtures were subjected to centrifugation at 16,100 × g for 10 min.The resulting supernatants were removed and OD495 nm measured with aBeckman Coulter DU730 spectrophotometer against a blank of H2O.

Thermal-Shift Assay. Thermal-shift analyses of 6xHis-LasR LBD and 6xHisLasRL130F LBD bound to HSLs were performed as previously described (25). Inshort, ligand-bound protein was diluted to 5 μM in reaction buffer (20 mMTris·HCl pH 8, 200 mM NaCl, and 1 mM DTT) containing DMSO or 10 μM HSLtest compound in 18-μL total volume. The mixtures were incubated at roomtemperature for 15 min. Next, 5000× SYPRO Orange (Thermo-Fisher) inDMSO was diluted to 200× in reaction buffer and used at 20× final con-centration. Two microliters of 200× SYPRO Orange was added to the 18-μLsample immediately before analysis. Samples were subjected to a linear heatgradient of 0.05 °C/s, from 25 °C to 99 °C in a Quant Studio 6 Flex System(Applied Biosystems) using the melting-curve setting. Fluorescence wasmeasured using the ROX reporter setting.

Pyocyanin Time Course. Overnight cultures of the P. aeruginosa wild-type,ΔlasI, ΔlasI lasR S129F, and ΔlasI lasR L130F strains were grown in LB mediumwith shaking at 37 °C. 1.5 mL of each culture was diluted into 50 mL of freshLB medium. Next, 3OC12HSL was added at the concentrations described inthe text and figures and the cultures were shaken at 37 °C for an additional3 h. From there forward, 1-mL aliquots were removed every 30 min for300 min and cell density (OD600 nm) was measured immediately using aBeckman Coulter DU730 Spectrophotometer. The aliquots were subjected tocentrifugation at 16,100 × g for 2 min and the clarified supernatants wereremoved. The OD695 nm of the supernatants were measured. Pyocyaninactivity was determined by plotting the OD695 nm/OD600 nm over time foreach strain.

Protein Production, Purification, and Crystallography. Recombinant 6xHis-LasRLBD and 6xHis-LasR LBD L130F proteins bound to 3OC8HSL, 3OC10HSL,3OC12HSL, or 3OC14HSL were purified as previously described for LasRLBD:3OC12HSL using Ni-NTA affinity columns followed by size-exclusionchromatography (25). The 6xHis-LasR LBD bound to 3OC10HSL and 6xHis-LasR LBD bound to 3OC14HSL were crystallized by the hanging-drop diffu-sion method in a solution of 200 mM Mg(NO3)2 and 20% PEG 3350. Dif-fraction data were processed using the HKL-3000 software package (52). Thestructures were solved using Phaser in Phenix (53, 54) by molecular re-placement, with the structure of LasR LBD:3OC12HSL used as the searchmodel (29). The space group was P1211. Model building was performedusing Coot (55, 56) and further refinement was accomplished using Phenix(53). Images of the structures were generated with PyMOL (57). We notethat the Rfree is higher in the higher-resolution structure of LasR LBDL130F:3OC14HSL than it is for the lower-resolution structure of LasR LBDL130F:3OC10HSL. The data from the LasR LBD L130F:3OC14HSL structure wereof lower quality than that for the LasR LBD L130:3OC10HSL structure,resulting in a higher Rfree value in the highest-resolution bin for the LasRLBD L130F:3OC14HSL structure, and therefore an overall higher Rfree value.Lowering the resolution cut-off for LasR LBD L130F:3OC14HSL and rerefine-ment in Phenix did not lead to a demonstrable change in the Rfree value.

Protein Solubility Assay. E. coli BL21 DE3 (Invitrogen) containing plasmid-borne 6xHis-LasR LBD or 6xHis-LasR LBD L130F were grown overnight andback-diluted 1:500 in 20 mL of LB medium containing ampicillin (100 μg/mL).Cultures were grown to OD600 of 0.5 and protein production was inducedwith 1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG). Upon addition ofIPTG, the desired test HSL was also added at a final concentration of 10 μM,and the cultures were incubated at 25 °C with shaking for 4 h. Cells wereharvested at 3,000 × g and resuspended in lysis buffer (500 mM NaCl, 20 mMTris·HCl pH 8, 20 mM imidazole, 5% glycerol, 1 mM EDTA, and 1 mM DTT).The cells were lysed using sonication (1-s pulses for 15 s with a 50% dutycycle). The fraction we call the whole-cell lysate was harvested after soni-cation. The soluble fraction was isolated by centrifugation at 32,000 × g.Samples were subjected to electrophoresis on SDS/PAGE gels (Bio-Rad) andimaged with an Image Quant LAS4000 gel dock using the chemiluminescentsetting (GE Healthcare).

Western Blotting to Assess LasR Protein Production Levels. Whole-cell lysatesand soluble fractions from E. coli carrying wild-type 6xHis-LasR and 6xHis-LasR S129C, S129F, S129M, S129T, S129W, and L130F were prepared as de-scribed in the preceding section, and then quantified using a Pierce BCAProtein Assay Kit (Thermo-Fisher). Samples were subjected to immunoblot-ting using a procedure adapted from Hurley and Bassler (58). Ten micro-grams of each sample were loaded into each well, and samples were probedwith a monoclonal antibody to the 6xHis tag (Thermo-Fisher) followed bygoat anti-mouse IgG2b cross adsorbed secondary antibody HRP (Thermo-Fisher). Both antibodies were used at 3 μg/mL. The blot was imaged withan Image Quant LAS4000 gel dock using the transillumination setting (GEHealthcare). Samples were tested in triplicate.

Statistical Methods. In all experiments, values are the average of three bi-ological replicates, each of which was assessed in two or three technicalreplicates as noted. For EC50 analyses, wild-type LasR was used as the control.EC50 values that appear in multiple experiments represent the mean from allexperiments (Fig. 8, Table 1, and SI Appendix, Fig. S5 and Table S1). In thesecases, two technical replicates of three biological replicates were assayed forevery protein/molecule in each experiment. The 95% confidence intervalswere calculated for each EC50 value in Table 1, and are provided in SI Ap-pendix, Table S2. Error bars represent SDs of the means. Two-tailed t testswere performed to compare experimental groups as noted in the figures.Two-way ANOVAs were performed to compare LasR variants to wild-type

McCready et al. PNAS | January 2, 2019 | vol. 116 | no. 1 | 253

MICRO

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Aug

ust 1

1, 2

021

LasR as noted in Fig. 5. P values are noted with asterisks: *P < 0.05, **P <0.01, ***P < 0.0001.

ACKNOWLEDGMENTS. We thank Dr. Fred Hughson and Dr. Philip Jeffereyfor assistance with crystallography; and Dr. Chari Smith and the entire B.L.B.group for insightful ideas about this research. This work was supported by

the Howard Hughes Medical Institute, National Institutes of Health Grant5R37GM065859, and National Science Foundation Grant MCB-1713731 (to B.L.B.);National Institute of General Medical Sciences T32GM007388 (to A.R.M.); and aJane Coffin Childs Memorial Fund for Biomedical Research Postdoctoral Fellow-ship (to J.E.P.). The content is solely the responsibility of the authors and does notnecessarily represent the official views of the National Institutes of Health.

1. Papenfort K, Bassler BL (2016) Quorum sensing signal-response systems in Gram-negative bacteria. Nat Rev Microbiol 14:576–588.

2. Albus AM, Pesci EC, Runyen-Janecky LJ, West SE, Iglewski BH (1997) Vfr controlsquorum sensing in Pseudomonas aeruginosa. J Bacteriol 179:3928–3935.

3. Latifi A, et al. (1995) Multiple homologues of LuxR and LuxI control expression ofvirulence determinants and secondary metabolites through quorum sensing inPseudomonas aeruginosa PAO1. Mol Microbiol 17:333–343.

4. Engebrecht J, Nealson K, Silverman M (1983) Bacterial bioluminescence: Isolation andgenetic analysis of functions from Vibrio fischeri. Cell 32:773–781.

5. Hanzelka BL, et al. (1999) Acylhomoserine lactone synthase activity of the Vibrio fi-scheri AinS protein. J Bacteriol 181:5766–5770.

6. Pearson JP, et al. (1994) Structure of the autoinducer required for expressionof Pseudomonas aeruginosa virulence genes. Proc Natl Acad Sci USA 91:197–201.

7. Brint JM, Ohman DE (1995) Synthesis of multiple exoproducts in Pseudomonas aer-uginosa is under the control of RhlR-RhlI, another set of regulators in strainPAO1 with homology to the autoinducer-responsive LuxR-LuxI family. J Bacteriol 177:7155–7163.

8. Eberhard A, et al. (1981) Structural identification of autoinducer of Photobacteriumfischeri luciferase. Biochemistry 20:2444–2449.

9. Cao JG, Meighen EA (1989) Purification and structural identification of an auto-inducer for the luminescence system of Vibrio harveyi. J Biol Chem 264:21670–21676.

10. Churchill MEA, Chen L (2011) Structural basis of acyl-homoserine lactone-dependentsignaling. Chem Rev 111:68–85.

11. Vannini A, et al. (2002) The crystal structure of the quorum sensing protein TraRbound to its autoinducer and target DNA. EMBO J 21:4393–4401.

12. Nasser W, Reverchon S (2007) New insights into the regulatory mechanisms of theLuxR family of quorum sensing regulators. Anal Bioanal Chem 387:381–390.

13. Freeman JA, Lilley BN, Bassler BL (2000) A genetic analysis of the functions of LuxN: Atwo-component hybrid sensor kinase that regulates quorum sensing in Vibrio harveyi.Mol Microbiol 35:139–149.

14. Bassler BL, Wright M, Showalter RE, Silverman MR (1993) Intercellular signalling inVibrio harveyi: Sequence and function of genes regulating expression of lumines-cence. Mol Microbiol 9:773–786.

15. Ke X, Miller LC, Bassler BL (2015) Determinants governing ligand specificity of theVibrio harveyi LuxN quorum-sensing receptor. Mol Microbiol 95:127–142.

16. Galloway WR, Hodgkinson JT, Bowden SD, Welch M, Spring DR (2011) Quorumsensing in Gram-negative bacteria: Small-molecule modulation of AHL and AI-2 quorum sensing pathways. Chem Rev 111:28–67.

17. You YS, et al. (2006) Use of bacterial quorum-sensing components to regulate geneexpression in plants. Plant Physiol 140:1205–1212.

18. Zhu J, Winans SC (2001) The quorum-sensing transcriptional regulator TraR requiresits cognate signaling ligand for protein folding, protease resistance, and di-merization. Proc Natl Acad Sci USA 98:1507–1512.

19. Hawver LA, Jung SA, Ng WL (2016) Specificity and complexity in bacterial quorum-sensing systems. FEMS Microbiol Rev 40:738–752.

20. Gerdt JP, et al. (2017) Chemical interrogation of LuxR-type quorum sensing receptorsreveals new insights into receptor selectivity and the potential for interspecies bac-terial signaling. ACS Chem Biol 12:2457–2464.

21. Nguyen Y, et al. (2015) Structural and mechanistic roles of novel chemical ligands onthe SdiA quorum-sensing transcription regulator. MBio 6:e02429-14.

22. Michael B, Smith JN, Swift S, Heffron F, Ahmer BM (2001) SdiA of Salmonella entericais a LuxR homolog that detects mixed microbial communities. J Bacteriol 183:5733–5742.

23. Sitnikov DM, Schineller JB, Baldwin TO (1996) Control of cell division in Escherichiacoli: Regulation of transcription of ftsQA involves both rpoS and SdiA-mediated au-toinduction. Proc Natl Acad Sci USA 93:336–341.

24. Pearson JP, Pesci EC, Iglewski BH (1997) Roles of Pseudomonas aeruginosa las and rhlquorum-sensing systems in control of elastase and rhamnolipid biosynthesis genes.J Bacteriol 179:5756–5767.

25. Paczkowski JE, et al. (2017) Flavonoids suppress Pseudomonas aeruginosa virulencethrough allosteric inhibition of quorum-sensing receptors. J Biol Chem 292:4064–4076.

26. Chhabra SR, et al. (2003) Synthetic analogues of the bacterial signal (quorum sensing)molecule N-(3-oxododecanoyl)-L-homoserine lactone as immune modulators. J MedChem 46:97–104.

27. Gambello MJ, Iglewski BH (1991) Cloning and characterization of the Pseudomonasaeruginosa lasR gene, a transcriptional activator of elastase expression. J Bacteriol173:3000–3009.

28. Moore JD, Gerdt JP, Eibergen NR, Blackwell HE (2014) Active efflux influences thepotency of quorum sensing inhibitors in Pseudomonas aeruginosa. ChemBioChem 15:435–442.

29. Bottomley MJ, Muraglia E, Bazzo R, Carfì A (2007) Molecular insights into quorumsensing in the human pathogen Pseudomonas aeruginosa from the structure of thevirulence regulator LasR bound to its autoinducer. J Biol Chem 282:13592–13600.

30. Schuster M, Urbanowski ML, Greenberg EP (2004) Promoter specificity in Pseudo-monas aeruginosa quorum sensing revealed by DNA binding of purified LasR. ProcNatl Acad Sci USA 101:15833–15839.

31. O’Loughlin CT, et al. (2013) A quorum-sensing inhibitor blocks Pseudomonas aeru-ginosa virulence and biofilm formation. Proc Natl Acad Sci USA 110:17981–17986.

32. Gerdt JP, McInnis CE, Schell TL, Rossi FM, Blackwell HE (2014) Mutational analysis ofthe quorum-sensing receptor LasR reveals interactions that govern activation andinhibition by nonlactone ligands. Chem Biol 21:1361–1369.

33. Gerdt JP, McInnis CE, Schell TL, Blackwell HE (2015) Unraveling the contributions ofhydrogen-bonding interactions to the activity of native and non-native ligands in thequorum-sensing receptor LasR. Org Biomol Chem 13:1453–1462.

34. Fuqua C, Winans SC, Greenberg EP (1996) Census and consensus in bacterial ecosys-tems: The LuxR-LuxI family of quorum-sensing transcriptional regulators. Annu RevMicrobiol 50:727–751.

35. O’Reilly MC, et al. (2018) Structural and biochemical studies of non-native agonists ofthe LasR quorum-sensing receptor reveal an L3 loop “out” conformation for LasR. CellChem Biol 25:1128–1139.e3.

36. Zou Y, Nair SK (2009) Molecular basis for the recognition of structurally distinct au-toinducer mimics by the Pseudomonas aeruginosa LasR quorum-sensing signalingreceptor. Chem Biol 16:961–970.

37. Wysoczynski-Horita CL, et al. (2018) Mechanism of agonism and antagonism of thePseudomonas aeruginosa quorum sensing regulator QscR with non-native ligands.Mol Microbiol 108:240–257.

38. Lintz MJ, Oinuma K, Wysoczynski CL, Greenberg EP, Churchill MEA (2011) Crystalstructure of QscR, a Pseudomonas aeruginosa quorum sensing signal receptor. ProcNatl Acad Sci USA 108:15763–15768.

39. Chen G, et al. (2011) A strategy for antagonizing quorum sensing. Mol Cell 42:199–209.40. Zhang RG, et al. (2002) Structure of a bacterial quorum-sensing transcription factor

complexed with pheromone and DNA. Nature 417:971–974, and erratum (2011) 476:240.41. Tashiro Y, Yawata Y, Toyofuku M, Uchiyama H, Nomura N (2013) Interspecies in-

teraction between Pseudomonas aeruginosa and other microorganisms. MicrobesEnviron 28:13–24.

42. Feltner JB, et al. (2016) LasR variant cystic fibrosis isolates reveal an adaptablequorum-sensing hierarchy in Pseudomonas aeruginosa. MBio 7:9.

43. Steidle A, et al. (2002) Identification and characterization of an N-acylhomoserinelactone-dependent quorum-sensing system in Pseudomonas putida strain IsoF. ApplEnviron Microbiol 68:6371–6382.

44. Yin WF, et al. (2012) N-acyl homoserine lactone production by Klebsiella pneumoniaeisolated from human tongue surface. Sensors (Basel) 12:3472–3483.

45. Baron SS, Rowe JJ (1981) Antibiotic action of pyocyanin. Antimicrob AgentsChemother 20:814–820.

46. Smalley NE, An D, Parsek MR, Chandler JR, Dandekar AA (2015) Quorum sensingprotects Pseudomonas aeruginosa against cheating by other species in a laboratorycoculture model. J Bacteriol 197:3154–3159.

47. Borlee BR, Geske GD, Blackwell HE, Handelsman J (2010) Identification of syntheticinducers and inhibitors of the quorum-sensing regulator LasR in Pseudomonas aer-uginosa by high-throughput screening. Appl Environ Microbiol 76:8255–8258.

48. Kukavica-Ibrulj I, et al. (2008) In vivo growth of Pseudomonas aeruginosa strainsPAO1 and PA14 and the hypervirulent strain LESB58 in a rat model of chronic lunginfection. J Bacteriol 190:2804–2813.

49. Schweizer HP (1991) Escherichia-Pseudomonas shuttle vectors derived from pUC18/19.Gene 97:109–121.

50. Mukherjee S, Moustafa D, Smith CD, Goldberg JB, Bassler BL (2017) The RhlR quorum-sensing receptor controls Pseudomonas aeruginosa pathogenesis and biofilm devel-opment independently of its canonical homoserine lactone autoinducer. PLoS Pathog13:e1006504.

51. Simon R, Priefer U, Puhler A (1983) A broad host range mobilization system for in vivogenetic-engineering—Transposon mutagenesis in Gram negative bacteria. Bio/Technology 1:784–791.

52. Minor W, Cymborowski M, Otwinowski Z, Chruszcz M (2006) HKL-3000: The in-tegration of data reduction and structure solution—From diffraction images to aninitial model in minutes. Acta Crystallogr D Biol Crystallogr 62:859–866.

53. Adams PD, et al. (2011) The Phenix software for automated determination of mac-romolecular structures. Methods 55:94–106.

54. Afonine PV, et al. (2012) Towards automated crystallographic structure refinementwith phenix.refine. Acta Crystallogr D Biol Crystallogr 68:352–367.

55. Emsley P, Cowtan K (2004) Coot: Model-building tools for molecular graphics. ActaCrystallogr D Biol Crystallogr 60:2126–2132.

56. Emsley P, Lohkamp B, Scott WG, Cowtan K (2010) Features and development of Coot.Acta Crystallogr D Biol Crystallogr 66:486–501.

57. DeLano WL (2009) PyMOL molecular viewer: Updates and refinements. Abstr Pap AmChem S 238:1.

58. Hurley A, Bassler BL (2017) Asymmetric regulation of quorum-sensing receptors drivesautoinducer-specific gene expression programs in Vibrio cholerae. PLoS Genet 13:e1006826.

59. Grøftehauge MK, Hajizadeh NR, Swann MJ, Pohl E (2015) Protein-ligand interactionsinvestigated by thermal shift assays (TSA) and dual polarization interferometry (DPI).Acta Crystallogr D Biol Crystallogr 71:36–44.

254 | www.pnas.org/cgi/doi/10.1073/pnas.1817239116 McCready et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 1

1, 2

021