Aryl-homoserine lactone quorum sensing instem-nodulating photosynthetic bradyrhizobiaNathan A. Ahlgrena, Caroline S. Harwooda, Amy L. Schaefera, Eric Giraudb, and E. Peter Greenberga,1

aDepartment of Microbiology, University of Washington, Seattle, WA 98195; and bLaboratoire des Symbioses Tropicales et Méditerranéennes, Institut deRecherche pour le Développement, Université de Montpellier/SupAgro/Institut National de la Recherche Agronomique/Centre de Coopération Internationaleen Recherche Agronomique pour le Développement, F-34398 Montpellier, France

Contributed by E. Peter Greenberg, March 9, 2011 (sent for review February 2, 2011)

Many Proteobacteria possess LuxI-LuxR–type quorum-sensing sys-tems that produce and detect fatty acyl-homoserine lactone (HSL)signals. The photoheterotroph Rhodopseudomonas palustris isunusual in that it produces and detects an aryl-HSL, p-coumaroyl-HSL, and signal production requires an exogenous source of p-coumarate. A photosynthetic stem-nodulating member of thegenus Bradyrhizobium produces a small molecule signal that elicitsan R. palustris quorum-sensing response. Here, we show thatthis signal is cinnamoyl-HSL and that cinnamoyl-HSL is producedby the LuxI homolog BraI and detected by BraR. Cinnamoyl-HSLreaches concentrations on the order of 50 nM in cultures of stem-nodulating bradyrhizobia grown in the presence or absence ofcinnamate. Acyl-HSLs often reach concentrations of 0.1–30 μM inbacterial cultures, and generally, LuxR-type receptors respond tosignals in a concentration range from 5 to a few hundred nano-molar. Our stem-nodulating Bradyrhizobium strain responds topicomolar concentrations of cinnamoyl-HSL and thus, producescinnamoyl-HSL in excess of the levels required for a signal responsewithout an exogenous source of cinnamate. The ability of Brady-rhizobium to produce and respond to cinnamoyl-HSL shows thataryl-HSL production is not unique to R. palustris, that the aromaticacid substrate for aryl-HSL synthesis does not have to be suppliedexogenously, and that some acyl-HSL quorum-sensing systems mayfunction at very low signal production and response levels.

bacterial communication | LuxI-LuxR homologs | sociomicrobiology

Over 100 species ofProteobacteria use acyl-homoserine lactone(HSL) quorum-sensing signals to control expression of spe-

cific subsets of genes in a cell density-dependent manner (1–3).Signals are produced by members of the LuxI protein family anddetected by transcription factors of the LuxR family. The cognateluxI and luxR genes are often adjacent to each other (4). Becauseacyl-HSLs can diffuse in and out of cells, the environmental signalconcentration must reach a critical level to trigger a response.All but one of the acyl-HSL signals described to date are fatty

acyl-HSLs. A variety of different fatty acyl-HSLs have been de-scribed, and the acyl side chain confers signal specificity (1, 3, 5–7). The substrates for fatty acyl-HSL synthesis by LuxI homologsare S-adenosylmethionine (SAM) and the appropriate fatty acyl-acyl carrier protein (ACP) (8–10). The exception is RpaI fromthe photosynthetic bacterium Rhodopseudomonas palustris, whichproduces p-coumaroyl-HSL (pC-HSL) (11). Synthesis of pC-HSLby R. palustris requires an exogenous source of p-coumarate. As istrue for many fatty acyl-HSL quorum-sensing systems, tran-scription of rpaI is positively autoregulated by pC-HSL and thetranscription factor RpaR (11). Also similar to other acyl-HSLquorum-sensing systems, pC-HSL is found in fully grown culturesof R. palustris at a concentration on the order of 1–10 μM, and R.palustris shows a response to pC-HSL at concentrations as low asa few nanomolar, with saturation of the response at 20–50 nMpC-HSL (11).The report describing pC-HSL production by R. palustris in-

cluded a screen for other bacteria that, when grown in thepresence of p-coumarate, produced a signal that could substitute

for pC-HSL in bioassays for this material. Other than R. palustrisitself, the bacterium that produced the most biological activitywas Bradyrhizobium BTAi1, and the activity was about 1% of theactivity found in R. palustris cultures (11). Members of the genusBradyrhizobium are closely related to R. palustris, and the Bra-dyrhizobium BTAi1 LuxI homolog shows a high level of identity(58%) with the R. palustris pC-HSL synthase, RpaI. Bradyrhi-zobium BTAi1 is a curious member of the genus in that it isphotosynthetic and that it can form nitrogen-fixing nodules onthe stems of host plants, whereas with other bradyrhizobia, thenodules are restricted to plant roots (12, 13).It is not clear whether p-coumarate is required for production of

the R. palustris signaling molecule by Bradyrhizobium BTAi1, andit is not clear if the low level of activity was the result of low levels ofpC-HSL synthesis or whether BTAi1 might be producing a differ-ent molecule that could trigger a relatively weak response inR. palustris. Thus, we sought to identify the signal produced bystem-nodulating bradyrhizobia. We report here that the moleculeis cinnamoyl-HSL, that stem-nodulating bradyrhizobia do not re-quire added cinnamic acid to synthesize cinnamoyl-HSL, and thatthe signal-dependent transcription factor responds to extraordi-narily low cinnamoyl-HSL concentrations.

ResultsStem-Nodulating Photosynthetic Bradyrhizobia Produce a pC-HSLSubstitute. We previously showed that an R. palustris reportersensitive to pC-HSL responded to something produced by Bra-dyrhizobium strain BTAi1 grown with p-coumarate. Because theresponse was quite low compared with responses to materialproduced by R. palustris, we hypothesized that BradyrhizobiumBTAi1 either makes very low levels of pC-HSL or makes a differ-ent but related aryl-HSL with a limited ability to serve as a sub-stitute for pC-HSL (11). To begin to discriminate between thesetwo possibilities, we separated the active material extracted withethyl acetate from cultures ofBradyrhizobiumBTAi1 and a secondphototrophic stem-nodulating Bradyrhizobium strain, ORS278, byHPLC and assessed individual fractions for activity in the R. pal-ustris pC-HSL reporter system (Fig. 1A).We found active fractionsin extracts of culture fluid from both strains ofBradyrhizobium, butthe majority of activity was found in fractions 34 and 35, whicheluted later than the fraction containing pC-HSL (fraction 25).This indicated that the activematerial was less polar than pC-HSL.Furthermore, similar levels of active material were obtained fromcultures grown with or without added p-coumarate.It seemed likely that the substance that could substitute for

pC-HSL was a related aryl-HSL. If true, it is also possible thatthe active material might represent a small percent of the totalamount of aryl-HSL synthesized, with the rest being material that

Author contributions: N.A.A., C.S.H., A.L.S., and E.P.G. designed research; N.A.A. andA.L.S. performed research; N.A.A., A.L.S., and E.G. contributed new reagents/analytictools; N.A.A., C.S.H., A.L.S., and E.P.G. analyzed data; and N.A.A., C.S.H., A.L.S., andE.P.G. wrote the paper.

The authors declare no conflict of interest.1To whom correspondence should be addressed. E-mail: epgreen@u.washington.edu.

www.pnas.org/cgi/doi/10.1073/pnas.1103821108 PNAS | April 26, 2011 | vol. 108 | no. 17 | 7183–7188

MICRO

BIOLO

GY

is not active in the pC-HSL bioassay. To test this possibility, weused a radiotracer assay for acyl-HSLs that does not discriminatebetween bioactive and inactive material (14, 15). We used Brady-rhizobiumORS278 for the radiotracer experiment and most otherexperiments. The genomeof strainORS278 contains a single set ofluxR-luxI homologs. Cultures of Bradyrhizobium ORS278 wereincubated with a mixture of 13 different aromatic acids (as po-tential aryl substrates for the LuxI homolog) and L-[1-14C] me-thionine. Any acyl-HSL produced by the bacteria will haveradiolabel in the HSL ring. The radiolabel extracted in acidifiedethyl acetate was found almost exclusively in fractions in whichmost of the biological activity was detected (fractions 34 and 35)(Fig. 1B). Thus, it seems likely that Bradyrhizobium ORS278produces primarily a single aryl-HSL that is less polar than pC-HSL but can serve as a pC-HSL substitute in bioassays.

Identity of the Aryl-HSL Produced by Bradyrhizobium ORS278. Todetermine the chemical composition of the putative Bradyrhi-zobium ORS278 acyl-HSL, we extracted culture fluid with ethylacetate, purified the active material in the ethyl acetate extract bymethanol gradient and isocratic HPLC, and analyzed the purifiedmaterial by high-resolution MS (Materials and Methods). The pri-mary parent ion peak of the purified material had a mass of232.0981, and there was an ion corresponding to a mass of254.0797. These masses are within 5 ppm of the predicted M+Hand M+Na masses (232.0974 and 254.0793, respectively) ofa compound with the chemical composition of C13H13NO3. This isthe chemical composition of cinnamoyl-HSL, which differs frompC-HSL only in that it lacks the para-hydroxy group on the aro-matic ring (Fig. 2A). Lack of the hydroxyl reduces the polarity ofcinnamoyl-HSLwith respect to pC-HSL, and this is consistent withthe HPLC analysis. To gain evidence in support of the idea thatthe molecule produced by Bradyrhizobium ORS278 is cinnamoyl-HSL, we synthesized this compound chemically. The syntheticcompound showed an HPLC elution profile that was identical to

the purified biological material, and the MS spectra were in-distinguishable (Fig. 2B). Furthermore, as predicted, cinnamoyl-HSL was active in the bioassay for pC-HSL; however, the amountof cinnamoyl-HSL required to give a half-maximal response wasabout 100 times the amount of pC-HSL needed to give a half-maximal response (Fig. 2C).

Cinnamoyl-HSL–Specific Bioassay. The genome of BradyrhizobiumORS278 contains a luxI homolog and a luxR homolog thatare adjacent to each other, braI (BRADO0941) and braR(BRADO0942) (Fig. 3A). Because many luxI homologs arepositively autoregulated by the acyl-HSL that they produce andtheir cognate LuxR homolog (1, 16), we suspected that braI

A

B

Fig. 1. HPLC analyses of acyl-HSLs produced by stem-nodulating bradyrhi-zobia. (A) Bradyrhizobium ORS278 and BTAi1 were grown with 500 μM p-coumarate, and ethyl acetate extracts of cultures were fractionated by HPLC.HPLC fractions were screened with the R. palustris CGA814 pC-HSL–sensitivebioassay. Extracts of strain ORS278 (■), and strain BTAi1 (○) are shown.Background β-galactosidase without added extract was 23,000 and is sub-tracted from the data shown in the figure. Synthetic pC-HSL was eluted infraction 25. (B) Radiotracer assay for acyl-HSL produced by BradyrhizobiumORS278. Cells were grown with 14C-methionine and a mixture of 13 aromaticcompounds (Materials and Methods). The cultures were extracted, andextracts were separated by HPLC as above. The dashed lines show the per-cent methanol in the gradient.

A

B

C

Fig. 2. (A) Chemical identity of the Bradyrhizobium ORS278 acyl-HSL:cinnamoyl-HSL (C13H13NO3). (B) Mass spectra of chemically synthesizedcinnamoyl-HSL (Upper) and material purified from Bradyrhizobium ORS278(Lower). (C) Induction of β-galactosidase in R. palustris CGA814 (rpaI-lacZ)by synthetic pC-HSL (◆) and synthetic cinnamoyl-HSL (□).

7184 | www.pnas.org/cgi/doi/10.1073/pnas.1103821108 Ahlgren et al.

might be responsive to cinnamoyl-HSL. It is often but not alwaysthe case that a lux box-like sequence, an 18- to 20-bp invertedrepeat, can be identified in the promoter region of an acyl-HSL–dependent gene. This is the case for the pC-HSL–responsive rpaIpromoter in R. palustris, for example (11). Although we wereunable to identify such a sequence in the braI promoter region,we nevertheless constructed a strain with a promoterless lacZinserted in the braI ORF. Because this strain has an un-interrupted braR and an inactive acyl-HSL synthase gene, tran-scription of lacZ should be activated by the addition of thecognate signal if there is positive autoregulation. We monitoredlacZ expression by measuring β-galactosidase and found that thebraI promoter was activated by as little as 2–5 pM cinnamoyl-HSL, and the response showed saturation at <100 pM cinnamoyl-HSL. The braI promoter also responded to pC-HSL, but it wasmuch less sensitive to pC-HSL (Fig. 3B). Although the braI pro-moter response required about 50 times more pC-HSL thancinnamoyl-HSL, it was roughly equal in sensitivity to pC-HSLas was R. palustris, which shows specificity for pC-HSL (11).We conclude that the braI gene is positively autoregulated bycinnamoyl-HSL. Furthermore, the lacZ insertion strain can serveas a very sensitive bioassay of cinnamoyl-HSL.

Production of Cinnamoyl-HSL Does Not Require an Exogenous Sourceof Cinnamate. We determined the concentration of cinnamoyl-HSL in late logarithmic phase cultures of BradyrhizobiumORS278 by extracting the signal with ethyl acetate and using thebraI-lacZ mutant as a reporter in a bioassay. Fluid from culturesgrown with or without cinnamate (5 μM) or a mixture of aro-matic compounds contained about 10 nM cinnamoyl-HSL (Fig.4). This level of acyl-HSL production is at the low end of thespectrum compared with many other acyl-HSL–producing bac-teria. For example, R. palustris produces about 1–10 μM pC-HSL, but because Bradyrhizobium ORS278 is so exquisitelysensitive to cinnamoyl-HSL, the level of production is actually ingreat excess of that required for a signal response.

BraI and BraR Define a Cinnamoyl-HSL–Dependent Quorum-SensingCircuit.Wetestedourhypothesis thatbraI is required for cinnamoyl-HSL synthesis by monitoring cinnamoyl-HSL production by abraI mutant and by monitoring cinnamoyl-HSL production bya recombinant R. palustris pC-HSL mutant (RpaI−) containing abraI expression plasmid. As predicted, the braI mutant did notproduce any detectable cinnamoyl-HSL as measured by usingthe bioassay described above. The braI-containing recombinantR. palustris produced levels of cinnamoyl-HSL comparable withthose producedbyBradyrhizobiumORS278, and it producedmuchhigher levels (over 100 nM) when grown in the presence of cin-namate (Fig. 4). Thus, braI codes for cinnamoyl-HSL production.To determine whether BraR is involved in cinnamoyl-HSL–

dependent activation of braI, we measured cinnamoyl-HSLproduction in a braR mutant and found that there was abouta 90% reduction in the level of cinnamoyl-HSL produced by themutant compared with the parent (Fig. 4). This finding supports

B

A

Fig. 3. The cinnamoyl-HSL–specific bioassay. (A) The BradyrhizobiumORS278 braI-braR region and flanking genes. Hypothetical ORFs are shadedin gray, and genes of putative function are filled in black. The gene upstreamof braR (shaded black) is annotated as a putative enoyl-CoA hydratase, andthe gene downstream of braI (lined) shows some sequence similarity to pu-tative nitroreductases; the genes further downstream of braI (shaded black)are annotated as putative acyl-CoA dehydrogenases. (B) We used Bradyrhi-zobium NA1, which contains a braI::lacZ insertion, as a bioassay, and wemeasured responses to synthetic pC-HSL (□) and cinnamoyl-HSL (●) at theconcentrations indicated.

Fig. 4. Cinnamoyl-HSL production by WT and mutant strains of Bradyrhi-zobium ORS278 and the R. palustris rpaI::lacZ bioassay strain containingpBraI. Cultures were extracted at an OD of 0.1. Cinnamoyl-HSL in extractswas measured by using the Bradyrhizobium braI::lacZ reporter strain NA1.Cinnamate (cinn) or a mixture of aromatic salts (aromatics) was added to thegrowth medium as indicated. BraI− and BraR− indicate extracts from the braIand braR mutants. Extracts of the R. palustris rpaI mutant (R. pal) containingpBraI grown with 100 μM or without added cinnamate were also tested. Thedata are means of six biological replicates, with each replicate assayed intriplicate. The error bars are SDs from the means.

Fig. 5. Cinnamoyl-HSL production by Bradyrhizobium ORS278 as a functionof growth: CFU per mL (○) and cinnamoyl-HSL concentration (●). We usedthe cinnamoyl-HSL bioassay to measure signal concentrations in culturesextracted at the indicated times. A standard curve was constructed by usingsynthetic cinnamoyl-HSL.

Ahlgren et al. PNAS | April 26, 2011 | vol. 108 | no. 17 | 7185

MICRO

BIOLO

GY

the hypothesis that BraR is a cinnamoyl-HSL–dependent tran-scription factor.One characteristic of acyl-HSL production in positively auto-

regulated LuxI-LuxR–type circuits is that, in early logarithmicphase cultures, the signal remains low and increases rapidly whenthe culture reaches a sufficiently high density. This is precisely theresult that we obtained when we monitored cinnamoyl-HSL pro-duction by BradyrhizobiumORS278 with our bioassay (Fig. 5). Ofnote, the maximal amount of cinnamoyl-HSL was accumulated inthe late logarithmic phase, and levels decreased in stationaryphase. The level of cinnamoyl-HSL, which accumulated in the latelogarithmic growth phase, was similar in strain BTAi1.

The braI Promoter Responds Relatively Poorly to Fatty Acyl-HSLs. TherpaI promoter in R. palustris is activated by RpaR and pC-HSL.We show here that it also responds to relatively high concen-trations of cinnamoyl-HSL (Fig. 2C). The rpaI promoter was notresponsive to fatty acyl-HSLs (11). We screened 11 differentfatty acyl-HSLs with fatty acyl groups ranging from 4 to 12 car-bons for the ability to activate expression of the braI-lacZ fusionin the Bradyrhizobium reporter strain. We used a range of con-centrations for each acyl-HSL and determined a maximum re-sponse as well as the acyl-HSL concentration required to givea half-maximal response. The maximum responses for all of thefatty acyl-HSLs were similar to that for cinnamoyl-HSL, but theconcentrations of these acyl-HSLs required to give a half-maximalresponse were anywhere from 3 to 6 orders of magnitude higherthan the cinnamoyl-HSL half-maximal response (Fig. 6). Al-though the Bradyrhizobium ORS278 quorum-sensing systemshows specificity for the acyl-HSL that it produces, cinnamoyl-HSL, the sensitivity is so great that, although fatty acyl-HSLsfunction as very poor substitutes for cinnamoyl-HSL, they func-tion in physiologically relevant ranges from 10 nM to 10 μM.

DiscussionWe previously reported that R. palustris produced the aryl-HSLquorum-sensing signal pC-HSL and that production of pC-HSLwas dependent on exogenously supplied p-coumarate. The genusBradyrhizobium and the genus Rhodopseudomonas are closelyrelated α-Proteobacteria (13). Because p-coumarate is producedby plants but not animals, we and others have speculated thatquorum sensing in R. palustris might involve an intimate re-lationship with plants (11, 17). Thus, it was of particular interestthat stem-nodulating bradyrhizobia produced a molecule thatsubstituted, albeit poorly, for pC-HSL as an R. palustris quorum-sensing signal. We have determined that this molecule differs

from pC-HSL only in that there is not a hydroxyl group on thearomatic ring. The compound is cinnamoyl-HSL, and perhapseven more interesting is that Bradyrhizobium ORS278 producesroughly equal amounts of this acyl-HSL when grown with orwithout aromatic acids added to the culture medium (Fig. 4).Furthermore, when given a choice of 13 different aromatic acids inthe growth medium, the predominant Bradyrhizobium ORS278acyl-HSL remained cinnamoyl-HSL (Fig. 1B). Thus, it is not uni-versally true that aryl-HSL quorum-sensing systems are coupled tothe presence of an exogenous supply of an aromatic compound.Although Bradyrhizobium strains ORS278 and BTAi1 produce

cinnamoyl-HSL in the absence of added cinnamate, the amountproduced, on the order of 20 nM (Figs. 4 and 5), is very lowcompared with the amount of pC-HSL produced by R. palustris(11). We have shown that cinnamoyl-HSL production is de-pendent on the Bradyrhizobium luxI homolog braI (Fig. 4). Byanalogy to other LuxI homologs (8–11), we presume that the braI-encoded protein catalyzes the synthesis of cinnamoyl-HSL fromSAM and an activated form of cinnamic acid. Some bacteria arecapable of producing cinnamic acid by using phenylalanine am-monia lyase (PAL) to deaminate phenylalanine. BradyrhizobiumORS278 contains a gene (BRADO1604) coding for a polypeptide,which shows similarity to theStreptomycesmaritimusPAL (18), butthe polypeptide contains conserved amino acid residues charac-teristic of histidine ammonia lyases (19, 20). This putative aminoacid lyase might be involved in cinnamate production by Bradyrhi-zobium ORS278, but we note that recombinant braI-containingR. palustris also produced nanomolar concentrations of cinnamoyl-HSL without added cinnamate and that the R. palustris genomedoes not possess any obvious PAL homologs.With the fatty acyl-HSL synthases, it has not been possible to

relate primary polypeptide sequence to the preferred acyl-HSLproduct.We have now defined a family of two aryl-HSL synthases,RpaI from R. palustris and BraI from Bradyrhizobium ORS278.These two aryl-HSL synthases aremuchmore similar to each other(56% identity) than they are to any known fatty acyl-HSL syn-thases (<32% identity). In fact, Bradyrhizobium japonicum pos-sesses a LuxI homolog that is related to RpaI and BraI (38% and34% identity, respectively). We are interested to determine thechemical nature of the putativeB. japonicum acyl-HSL.R. palustrisand Bradyrhizobium ORS278 are very closely related at the ge-nomic level. Thus, we cannot discern whether the LuxI homologsimilarity is because of constraints on enzymes that produce aryl-HSLs or a consequence of the close evolutionary relationshipbetween these species. We hope that it will be possible to identifyan aryl-HSL–producing bacterium that is not so closely related tobradyrhizobia and R. palustris. This would aid in discriminating be-tween constraints on enzyme activity and evolutionary relatedness.We were able to develop an assay for cinnamoyl-HSL by

inserting a lacZ reporter into theBradyrhizobiumORS278 braI. Byusing this reporter, we showed that a BraR mutant produced verylittle cinnamoyl-HSL (Fig. 4). This supports the conclusion thatthe LuxR homolog BraR is the receptor for cinnamoyl-HSL. Wealso were able to measure the amounts of cinnamoyl-HSL thataccumulated in the culture fluid of WT stem-nodulating bra-dyrhizobia, and we found that the two strains that we examinedmade on the order of 30–65 nM cinnamoyl-HSL (Fig. 5). This is atthe low end of the range of acyl-HSL concentrations found inculture fluid of other bacteria. For example, R. palustris produces1–10 μM pC-HSL (11), andPseudomonas aeruginosa produces twofatty acyl-HSLs, both in the range of 10 μM (21, 22). There isconsiderable strain variability in Vibrio fischeri acyl-HSL levels,with some strains producing very little 3-oxo-hexanoyl-HSL. InV. fischeri, the strain variation in levels of the acyl-HSL signalcorrelates with levels of quorum-controlled luminescence (23, 24).Culture fluid from brightly luminescent strains contains micro-molar levels of signal, and culture fluid from dim strains containsnanomolar levels of signal. Furthermore, in vivo responses of

Fig. 6. Concentrations of various HSLs required for half-saturationresponses in the cinnamoyl-HSL bioassay. Each acyl-HSL was tested overa range of concentrations, and the concentrations at which the response washalf-saturated were determined from the response curves. Data are themeans of duplicate experiments, and error bars show the ranges.

7186 | www.pnas.org/cgi/doi/10.1073/pnas.1103821108 Ahlgren et al.

LuxR homologs generally require nanomolar signal concen-trations. However, our Bradyrhizobium braI-lacZ reporter strainresponds to picomolar levels of cinnamoyl-HSL (Fig. 3B). We areaware of one other example of a response to picomolar levels of anacyl-HSL signal. This example involves engineered strains ofAgrobacterium tumefaciens. Strains of A. tumefaciens that expressnormal levels of the acyl-HSL receptor TraR respond to nano-molar concentrations of the cognate signal 3OC8-HSL, and atrelatively high concentrations, certain other fatty acyl-HSLsfunction as antagonists of 3OC8-HSL (25). Strains engineered tooverexpress functional TraR respond to picomolar levels of 3OC8-HSL, and signals that function as antagonists with normallyexpressed TraR function as activators in the TraR overexpressionconstructs (25, 26).There are several possible explanations for the ultrasensitive

response of ourBradyrhizobium quorum-sensing reporter strain topicomolar levels of cinnamoyl-HSL. For example, there could bean active uptake system for cinnamoyl-HSL. However, we favora simpler hypothesis that derives from the sensitivity of TraRoverexpression strains ofA. tumefaciens. It is plausible that BraR isa high-affinity cinnamoyl-HSL receptor that is produced at par-ticularly high levels compared with WT levels of TraR, for exam-ple. If this is true, it might provide an explanation as to why thebasal level of braI-lacZ transcription is relatively high and whyexogenous signal affects a relatively meager two- to threefold in-duction of the β-galactosidase reporter. There are data indicatingthat overexpression of the V. fischeri LuxR results in an increasedexpression of the quorum-sensing, controlled luminescence genesin an acyl-HSL–independent manner (27). Expression can befurther stimulated by acyl-HSL signal addition. Furthermore, theBradyrhizobium cinnamoyl-HSL reporter is activated by relative-ly high concentrations of a wide variety of aryl-HSLs and fattyacyl-HSLs. This is consistent with findings for TraR overexpres-sion strains of A. tumefaciens, which show activation of quorum-controlled genes by a range of acyl-HSLs that interfere with3OC8-HSL gene activation in strains expressing normal levels ofTraR. Of course, our favored hypothesis awaits direct testing.Although our data indicate that Bradyrhizobium ORS278

responds relatively poorly to acyl-HSLs other than cinnamoyl-HSL, the level of activation at sufficiently high acyl-HSL con-centrations is comparable with the level of cinnamoyl-HSL ac-tivation. The BraR response to noncognate acyl-HSLs occurs inthe range of nanomolar to micromolar. This is within the rangeto which the cognate LuxR homologs respond and well withinthe range of signals produced by cultures of other bacteria. Thisinformation raises the possibility that, in certain soil habitats,a Bradyrhizobium ORS278 cell might respond to signals pro-duced by groups of other species. We do not have any idea aboutthe advantage that might be conferred by the potential ability todetect and respond to quorums of other species. We have not yetidentified the functions other than BraI that are regulated bycinnamoyl-HSL. Obviously, such knowledge might inform usabout the role of quorum sensing and potentially, eavesdroppingon other species in stem-nodulating bradyrhizobia. Also, quorumsensing has been shown to affect interactions of other membersof the Rhizobiales with their plant hosts (28–32). It will be ofinterest to compare plant interactions of BraI and BraR mutantswith the parent. Finally, we point out that it seems a real pos-sibility that we have overlooked acyl-HSL signaling systems inother bacteria, because we have focused on systems that requirenanomolar concentrations of signals for a response. That wehave now identified one system that functions below the nano-molar range suggests to us that there are other such systems.

Materials and MethodsBacterial Strains and Growth Conditions. Bradyrhizobium ORS278 (13) andBTAi1 (12) were grown in Arabinose-Gluconate (AG) medium (33) at 30 °Cwith shaking. Strain NA1 (described below) was grown in AG medium plus

kanamycin (50 μg/mL) unless otherwise specified. The BraR mutant of strainORS278 contains a Tn5-Gm insertion at base pair 392 of braR (locus tagBRADO0942). The mutant was selected by PCR from a Tn5 mutant library(34) by using a Tn5 and a braR-specific primer. For growth of R. palustrisCGA814, an rpaI::lacZ derivative of strain CGA009 (11), we used Photosyn-thesis Minimal (PM) medium in the light with 10 mM succinate as a carbonsource (35). For growth of R. palustris CGA814 containing the braI expressionvector pBraI, we added gentamicin (100 μg/mL) to the PM medium forplasmid maintenance. Individual aromatic salts were added at the indicatedconcentrations. Where indicated, we used a pool of 13 aromatic salts, includingbenzoate, p-coumarate, m-coumarate, o-coumarate, cinnamate, caffeate, fer-ulate, methoxycinnamate, sinapate, p-hydroxybenzoate, vanillate, phenylala-nine, and tryptophan, each at a concentration of 100 μM.

Mutant and Plasmid Construction. To construct the braI-lacZ mutant Brady-rhizobium NA1, we replaced all of the Bradyrhizobium ORS278 braI gene(locus tag BRADO0941), except the translation start and stop codons, with thepromoterless lacZ-kanamycin resistance cassette from pHRP314 (36). First, wePCR-amplified thebraIflanking regionswith primers containing 5′ EcoRI and 3′PstI restriction sites on the upstream flanking region of braI and 5′ PstI and3′ XbaI sites on the downstream flanking region. These two fragments wereligated with EcoRI- and XbaI-digested pSUP202pol6k (Tetr) (37) to form anintermediate cloning vehicle. The intermediate plasmid was digested withPstI, and this was ligated with an lacZ::Kmr cassette containing the PstI frag-ment of pHRP314. The resultant lacZ::Kmr suicide plasmid was introduced intoBradyrhizobium ORS278 by conjugation with Escherichia coli S17-1. Recombi-nants were selected on AG-kanamycin agar plates containing naladixic acid(35 μg/mL), and double cross-over recombinants were identified as those thatdid not grow when screened on AG tetracycline plates (50 μg/mL).

To construct pBraI, we PCR-amplified a braI DNA fragment from Bra-dyrhizobium ORS278 DNA. The PCR product extended from 21 bp upstreamof the braI start codon through the braI stop codon. The primers weredesigned to yield a product with a 5′ HindIII restriction site and a 3′ XbaIrestriction site. The PCR fragment was cloned into pCR-TOPO 5.1 (InvitrogenCorp.), subsequently excised with HindIII and XbaI, and inserted into HindIII-and XbaI-digested pBBR1MCS-5, a broad-host range expression vector (38),to yield a Plac-braI fusion vector.

Acyl-HSL Bioassays.We used an R. palustris rpaI-lacZ reporter strain (CGA814),which responds best to pC-HSL, and a Bradyrhizobium ORS278 braI-lacZderivative (NA1) as a reporter, which responds best to cinnamoyl-HSL inbioassays. Bioassays were either with extracts of whole bacterial cultures orwith cell-free culture fluid extracted with two equal volumes of acidifiedethyl acetate (0.1 mL glacial acetic acid per 1 L ethyl acetate). Separation ofcells and culture fluid was by centrifugation followed by filtration througha 0.22-μm membrane. Ethyl acetate extracts were fractionated as describedbelow. For bioassays with R. palustris CGA814, we used the method ofSchaefer et al. (11). For bioassays with Bradyrhizobium NA1, ethyl acetatesolutions containing acyl-HSLs were dried in 1.5-mL Eppendorf tubes or 2-mLwells in 96-well microtiter plates. We added 500 μL culture that had reachedan OD of 0.01 at 600 nm to the dried samples, and after 16–20 h at 30 °Cwith shaking, we measured β-galactosidase activity as described elsewhere(39). Standard curves were generated by using commercially supplied pC-L-HSL and cinnamoyl-L-HSL (Syntech).

HPLC Separation and Purification of Acyl-HSLs from Spent Culture Fluid. Fluidfrom cultures of stem-nodulating bradyrhizobia or recombinant R. palustriswas extractedwith two equal volumes of acidified ethyl acetate after removalof cells by centrifugation and filtration through a 0.22-μm filter. Ethyl acetateextracts of culture fluid were concentrated by rotary evaporation, and theconcentrated extracts were applied to a C18 reverse-phase HPLC column. Theflow rate was 1mL/min, and 1-mL fractions were collected and tested for acyl-HSLs as described in the text. We used a 10–100%methanol in water gradientspanning 70 min or a 35% isocratic methanol separation as indicated.

To purify the active compound produced by Bradyrhizobium ORS278, weextracted fluid from a 1-L culture grown to an OD of 1. The ethyl acetateextract was concentrated by rotary evaporation, and acyl-HSLs were sepa-rated by HPLC in a 10–100% methanol in water gradient; 1-mL fractionswere collected, and activity in each fraction was assessed by using the Bra-dyrhizobium NA1 bioassay. The three 1-mL fractions showing greatest ac-tivity were pooled, and the pooled fractions were concentrated and thenseparated by HPLC in 35% methanol. The Bradyrhizobium NA1 bioassayshowed that activity was eluted in fractions 26–29. These 1-mL fractionswere pooled and evaporated to dryness. We used a Waters Micromass High-Definition Time of Flight MS System (collision energy = 30 eV, cone = 35 V) to

Ahlgren et al. PNAS | April 26, 2011 | vol. 108 | no. 17 | 7187

MICRO

BIOLO

GY

determine an accurate molecular mass for the dried material and to com-pare the mass spectrum of this material to synthetic cinnamoyl-HSL.

Radiotracer Analysis of Acyl-HSLs. We used a 14C-tracer method (14) to detectacyl-HSLs produced by Bradyrhizobium ORS278 as follows. Cells were grownto mid-logarithmic phase (OD of 0.5) and diluted in 5 mL fresh medium to anOD of 0.1 in 15-mL plastic tubes. We then added L-[1-14C] methionine(American Radiolabeled Chemicals) as described elsewhere (14) and the ar-omatic acid mixture described above. Cultures were incubated for 20 h at

30 °C with shaking. Radioactive acyl-HSLs were extracted in acidified ethylacetate, and the extract was separated by methanol-gradient HPLC. Frac-tions were mixed with 4 mL Complete Counting Mixture (Research ProductsInternational), and radioactivity was measured with a Beckman LS 1800liquid scintillation counter.

ACKNOWLEDGMENTS. We thank Ross Lawrence for assistance with massspectrometry. This work was supported by US Army Research Office GrantW911NF0910350.

1. Fuqua C, Parsek MR, Greenberg EP (2001) Regulation of gene expression by cell-to-cellcommunication: Acyl-homoserine lactone quorum sensing. Annu Rev Genet 35:439–468.

2. Waters CM, Bassler BL (2005) Quorum sensing: Cell-to-cell communication in bacteria.Annu Rev Cell Dev Biol 21:319–346.

3. Whitehead NA, Barnard AM, Slater H, Simpson NJ, Salmond GP (2001) Quorum-sensing in Gram-negative bacteria. FEMS Microbiol Rev 25:365–404.

4. Fuqua C, Winans SC, Greenberg EP (1996) Census and consensus in bacterialecosystems: The LuxR-LuxI family of quorum-sensing transcriptional regulators. AnnuRev Microbiol 50:727–751.

5. Eberhard A, Widrig CA, McBath P, Schineller JB (1986) Analogs of the autoinducer ofbioluminescence in Vibrio fischeri. Arch Microbiol 146:35–40.

6. Schaefer AL, Hanzelka BL, Eberhard A, Greenberg EP (1996) Quorum sensing in Vibriofischeri: Probing autoinducer-LuxR interactions with autoinducer analogs. J Bacteriol178:2897–2901.

7. Ng WL, Bassler BL (2009) Bacterial quorum-sensing network architectures. Annu RevGenet 43:197–222.

8. Schaefer AL, Val DL, Hanzelka BL, Cronan JE, Jr., Greenberg EP (1996) Generationof cell-to-cell signals in quorum sensing: Acyl homoserine lactone synthase activityof a purified Vibrio fischeri LuxI protein. Proc Natl Acad Sci USA 93:9505–9509.

9. Moré MI, et al. (1996) Enzymatic synthesis of a quorum-sensing autoinducer throughuse of defined substrates. Science 272:1655–1658.

10. Parsek MR, Val DL, Hanzelka BL, Cronan JE, Jr., Greenberg EP (1999) Acylhomoserine-lactone quorum-sensing signal generation. Proc Natl Acad Sci USA 96:4360–4365.

11. Schaefer AL, et al. (2008) A new class of homoserine lactone quorum-sensing signals.Nature 454:595–599.

12. Eaglesham ARJ, Szalay AA (1983) Aerial stem nodules on Aeschynomene spp. Plant SciLett 29:265–272.

13. Molouba F, et al. (1999) Photosynthetic bradyrhizobia from Aeschynomene spp. arespecific to stem-nodulated species and form a separate 16S ribosomal DNA restrictionfragment length polymorphism group. Appl Environ Microbiol 65:3084–3094.

14. Schaefer AL, Greenberg EP, Parsek MR (2001) Acylated homoserine lactone detectionin Pseudomonas aeruginosa biofilms by radiolabel assay. Methods Enzymol 336:41–47.

15. Singh PK, et al. (2000) Quorum-sensing signals indicate that cystic fibrosis lungs areinfected with bacterial biofilms. Nature 407:762–764.

16. Goryachev AB (2009) Design principles of the bacterial quorum sensing genenetworks. Interdisciplinary Rev:Syst Biol Med 1:46–60.

17. Palmer AG, Blackwell HE (2008) Deciphering a protolanguage for bacteria-hostcommunication. Nat Chem Biol 4:452–454.

18. Xiang L, Moore BS (2005) Biochemical characterization of a prokaryotic phenylalanineammonia lyase. J Bacteriol 187:4286–4289.

19. Röther D, Poppe L, Viergutz S, Langer B, Rétey J (2001) Characterization of the activesite of histidine ammonia-lyase from Pseudomonas putida. Eur J Biochem 268:6011–6019.

20. Williams JS, Thomas M, Clarke DJ (2005) The gene stlA encodes a phenylalanineammonia-lyase that is involved in the production of a stilbene antibiotic in Photorhabdusluminescens TT01. Microbiology 151:2543–2550.

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

22. Pearson JP, Passador L, Iglewski BH, Greenberg EP (1995) A second N-acylhomoserinelactone signal produced by Pseudomonas aeruginosa. Proc Natl Acad Sci USA 92:

1490–1494.23. Boettcher KJ, Ruby EG (1995) Detection and quantification of Vibrio fischeri

autoinducer from symbiotic squid light organs. J Bacteriol 177:1053–1058.24. Nealson KH (1977) Autoinduction of bacterial luciferase. Occurrence, mechanism and

significance. Arch Microbiol 112:73–79.25. Zhu J, et al. (1998) Analogs of the autoinducer 3-oxooctanoyl-homoserine lactone

strongly inhibit activity of the TraR protein of Agrobacterium tumefaciens. J Bacteriol180:5398–5405.

26. Zhu J, Chai Y, Zhong Z, Li S, Winans SC (2003) Agrobacterium bioassay strain for

ultrasensitive detection of N-acylhomoserine lactone-type quorum-sensing molecules:Detection of autoinducers in Mesorhizobium huakuii. Appl Environ Microbiol 69:

6949–6953.27. Choi SH, Greenberg EP (1991) The C-terminal region of the Vibrio fischeri LuxR

protein contains an inducer-independent lux gene activating domain. Proc Natl Acad

Sci USA 88:11115–11119.28. Gao Y, Zhong Z, Sun K, Wang H, Zhu J (2006) The quorum-sensing system in a plant

bacterium Mesorhizobium huakuii affects growth rate and symbiotic nodulation.Plant Soil 286:53–60.

29. Marketon MM, Glenn SA, Eberhard A, González JE (2003) Quorum sensing controls

exopolysaccharide production in Sinorhizobium meliloti. J Bacteriol 185:325–331.30. Rosemeyer V, Michiels J, Verreth C, Vanderleyden J (1998) luxI- and luxR-homologous

genes of Rhizobium etli CNPAF512 contribute to synthesis of autoinducer moleculesand nodulation of Phaseolus vulgaris. J Bacteriol 180:815–821.

31. Zheng H, et al. (2006) A LuxR/LuxI-type quorum-sensing system in a plant bacterium,

Mesorhizobium tianshanense, controls symbiotic nodulation. J Bacteriol 188:1943–1949.32. Daniels R, et al. (2002) The cin quorum sensing locus of Rhizobium etli CNPAF512

affects growth and symbiotic nitrogen fixation. J Biol Chem 277:462–468.33. Sadowsky MJ, Tully RE, Cregan PB, Keyser HH (1987) Genetic diversity in

Bradyrhizobium japonicum serogroup 123 and its relation to genotype-specific

nodulation of soybean. Appl Environ Microbiol 53:2624–2630.34. Giraud E, et al. (2007) Legumes symbioses: Absence of Nod genes in photosynthetic

bradyrhizobia. Science 316:1307–1312.35. Kim M, Harwood C (1991) Regulation of benzoate-CoA ligase in Rhodopseudomonas

palustris. FEMS Microbiol Lett 83:199–203.36. Parales RE, Harwood CS (1993) Construction and use of a new broad-host-range lacZ

transcriptional fusion vector, pHRP309, for gram- bacteria. Gene 133:23–30.37. Zufferey R, Preisig O, Hennecke H, Thöny-Meyer L (1996) Assembly and function of

the cytochrome cbb3 oxidase subunits in Bradyrhizobium japonicum. J Biol Chem 271:9114–9119.

38. Kovach ME, et al. (1995) Four new derivatives of the broad-host-range cloning vectorpBBR1MCS, carrying different antibiotic-resistance cassettes. Gene 166:175–176.

39. Whiteley M, Lee KM, Greenberg EP (1999) Identification of genes controlled by

quorum sensing in Pseudomonas aeruginosa. Proc Natl Acad Sci USA 96:13904–13909.

7188 | www.pnas.org/cgi/doi/10.1073/pnas.1103821108 Ahlgren et al.