Direct Quantitation of the Quorum Sensing Signal,Autoinducer-2, in Clinically Relevant Samples byLiquid Chromatography-Tandem MassSpectrometry

Shawn R. Campagna,* Jessica R. Gooding, and Amanda L. May

Department of Chemistry; The University of Tennessee, Knoxville, Tennessee 37996

Quorum Sensing is a type of bacterial cell-to-cell signalingthat allows for cell density dependent regulation of geneexpression. Many of the behaviors mediated by quorumsensing are critical for bacterial colonization or infection,and autoinducer-2 has been proposed as a universalinterspecies signaling molecule that allows multispeciescolonies of bacteria, e.g., biofilms or dental plaque, tobehave as pseudomulticellular organisms. However, thedirect detection of autoinducer-2 has been difficult, leav-ing the in vivo relevance of this signal in question. Hereinwe report a liquid chromatography-tandem mass spec-trometric technique that enables reproducible, quantita-tive, and sensitive measurement of the concentration ofautoinducer-2 from a variety of sources. This techniquewas applied to the detection of autoinducer-2 from Es-cherichia coli and Vibrio harveyi in proof-of-conceptstudies and was then used to directly measure theconcentration of the signal produced by oral bacteria inhuman saliva.

Quorum sensing is the cell density dependent regulation ofgene expression, and these signaling systems are thought to bemechanisms by which bacteria can enact group beneficial behav-iors only when enough members of the population are present tosuccessfully carry out the desired task.1-4 Indeed, quorum sensingnetworks have been implicated in behaviors including lumines-cence, host colonization, biofilm formation, and virulence. Utilizingproper interspecies signaling systems, bacterial communities canbehave as pseudomulticellular organisms. Quorum sensing hasalso gained much interest because of the role it plays in infectiousdisease processes;2,5-10 however, our understanding of intra- and

interspecies signaling in bacteria is still in its infancy. Controllingor attenuating biofilm production and/or colonization via quorumsensing could lead to fundamental new approaches for thetreatment of bacterial infections.

A single molecule, (S)-4,5-dihydroxy-2,3-pentanedione (DPD),may give rise to the universal bacterial signal,11,12 autoinducer-2(AI-2), as the gene encoding for the DPD signal synthase (luxS)13

has been found in the genomes of many sequenced bacteria, bothgram-positive and gram-negative.14 Further, cell-free supernatantsfrom many species for which the genome has not been sequencedhave also been shown to produce AI-2 activity via the Vibrioharveyi reporter assay that is based on the bioluminescence ofthis organism.15 Despite the interest in AI-2 mediated signalingsystems, the concentration of this molecule in complex, biologi-cally relevant samples has not, as of yet, been determined. Indeed,it may be that the concentration of AI-2 in the environment ismuch lower than that necessary to observe a phenotype underlaboratory conditions. Recent model studies utilizing human salivahave shown that mutualistic biofilm formation by two oral bacteria,Streptococcus oralis 34 and Actinomyces naeslundii T14V, ismediated by AI-2 and that the concentration of the signal in theeffluent from the biofilm ranged from 135-197 nM,16 concentra-tions much lower than the typical µM concentrations found forsingle species cultures grown in laboratory media.17-19 Further,evidence is mounting that some bacterial species, such asEscherichia coli and Salmonella enterica serovar typhimurium may

* To whom correspondence should be addressed. E-mail: campagna@ion.chem.utk.edu.

(1) Camilli, A.; Bassler, B. L. Science 2006, 311, 1113–1116.(2) Parker, C. T.; Sperandio, V. Cell. Microbiol. 2009, 11, 363–369.(3) Waters, C. M.; Bassler, B. L. Annu. Rev. Cell Dev. Biol. 2005, 21, 319–

346.(4) Williams, P.; Winzer, K.; Chan, W. C.; Camara, M. Philos. Trans. R. Soc.

London, Ser. B 2007, 362, 1119–1134.(5) Bjarnsholt, T.; Kirketerp-Moller, K.; Jensen, P. O.; Madsen, K. G.; Phipps,

R.; Krogfelt, K.; Hoiby, N.; Givskov, M. Wound Repair Regen. 2008, 16,2–10.

(6) Escaich, S. Curr. Opin. Chem. Biol. 2008, 12, 400–408.(7) Girard, G.; Bloemberg, G. V. Future Microbiol. 2008, 3, 97–106.(8) Kumari, A.; Pasini, P.; Daunert, S. Anal. Bioanal. Chem. 2008, 391, 1619–

1627.

(9) Kuramitsu, H. K.; He, X. S.; Lux, R.; Anderson, M. H.; Shi, W. Y. Microbiol.Mol. Biol. Rev. 2007, 71, 653-670.

(10) Novick, R. P.; Geisinger, E. Annu. Rev. Genet. 2008, 42, 541–564.(11) Miller, S. T.; Xavier, K. B.; Campagna, S. R.; Taga, M. E.; Semmelhack,

M. F.; Bassler, B. L.; Hughson, F. M. Mol. Cell 2004, 15, 677–687.(12) Chen, X.; Schauder, S.; Potier, N.; Van Dorsselaer, A.; Pelczer, I.; Bassler,

B. L.; Hughson, F. M. Nature 2002, 415, 545–549.(13) Schauder, S.; Shokat, K.; Surette, M. G.; Bassler, B. L. Mol. Microbiol. 2001,

41, 463–476.(14) Xavier, K. B.; Bassler, B. L. Curr. Opin. Microbiol. 2003, 6, 191–197.(15) Bassler, B. L.; Greenberg, E. P.; Stevens, A. M. J. Bacteriol. 1997, 179,

4043–4045.(16) Rickard, A. H.; Campagna, S. R.; Kolenbrander, P. E. J. Appl. Microbiol.

2008, 105, 2096–2103.(17) Rajamani, S.; Zhu, J. G.; Pei, D. H.; Sayre, R. Biochemistry 2007, 46, 3990–

3997.(18) Zhu, J. G.; Pei, D. H. ACS Chem. Biol. 2008, 3, 110–119.(19) Thiel, V.; Vilchez, R.; Sztajer, H.; Wagner-Dobler, I.; Schulz, S. ChemBioChem

2009, 10, 479–485.

Anal. Chem. 2009, 81, 6374–6381

10.1021/ac900824j CCC: $40.75 2009 American Chemical Society6374 Analytical Chemistry, Vol. 81, No. 15, August 1, 2009Published on Web 07/13/2009

import and degrade AI-2 as a mechanism to interfere withsignaling in competitive species.19-22

Complete characterization of the active form of AI-2 for eachspecies producing DPD has been hindered since AI-2 is actuallya mixture of interconverting borated and unborated molecules.11,20

The chemical properties of DPD and molecules derived thereofare also responsible for the difficulties encountered duringattempts to quantitate AI-2. The major impediments are that nochromophore is contained in the molecule for fluorescent orsimilar detection, the low ionization potential of AI-2 renders massspectrometric (MS) based detection impossible, and the instabilityof DPD at high molarities does not allow concentration of themolecule. The inability to determine the concentrations of DPDpresent in the environment of bacteria living in complex com-munities has left open questions concerning the relevance of thismolecule as a signal in vivo.

To date, three classes of methods for the detection of AI-2have been utilized, the V. harveyi luminesence bioassay,15,16

biosensors derived from AI-2 receptor proteins,17,18 and LC/MSor GC/MS analysis of a DPD derivative.19 The V. harveyiluminescence bioassay has been the most utilized method for thedetection of AI-2 via the induction of light in a reporter strain bycell-free supernatants from cultures of interest. While this assayis sensitive over several orders magnitude with a limit of detection(LOD) in the low nM range, the culture to culture variability forV. harveyi has been shown to affect the reproducibility.16,21

Further, other molecules present in the cell-free supernatants ofthe studied species can alter the sensitivity of V. harveyi toAI-2.19,22 The small linear range of this assay (0.08-0.8 µM) alsolimits the quantitative ability.16 Recent work has constructed real-time biosensors for AI-2 derived from the V. harveyi receptorprotein, LuxP.17,18 These reporter proteins rely on a change influorescence from LuxP modified to contain chromophores thatchange emission characteristics upon AI-2 binding. This class ofsensors has a LOD of ∼100 nM, and the linear response rangesfrom ∼1-30 µM. However, these sensors are limited in sensitivityby the binding affinity (Kd ∼100 nM) of the ligand, and theyonly report on the fraction of DPD that has converted to theborate recognized by V. harveyi, making this detection methodsensitive to the borate concentration in the sample as well.Concurrent with our efforts to quantitate DPD, a method forGC/MS detection of this molecule was reported.19 Thistechnique utilized o-diaminobenzene as a derivatizing agent,and was able to detect the resulting quinoxaline with a LODof 5.3 nM and a limit of quantitation of 16 nM. However, a seriesof two derivatizations, a solid-phase extraction, and a 50 foldconcentration step were required to enable detection. Unfor-tunately, these methods for the detection of DPD are eithernot easily reproduced, not sufficiently quantitative, difficult toperform, only report on one of the forms of DPD, or acombination thereof.

We sought to develop a tool that would allow the facilemeasurement of [DPD] from biologically relevant environments.

For this technique, a derivative of DPD is detected via a selectedreaction monitoring event (SRM) on a tandem MS after separationvia LC. These events detect molecules by simultaneously monitor-ing for a specific parent ion mass-to-charge ratio (m/z) in the firstmass analyzer and for a characteristic and/or abundant fragmention m/z in a subsequent mass analyzer.23 Utilizing such tech-niques, compounds with identical nominal masses such as citrateand isocitrate can be distinguished by their unique fragmentationpatterns without the need for chromatography.24,25 The require-ment that the MS measure both the parent and the fragment ionpair of the analyte give SRM-based techniques good sensitivity,specificity, and a large linear dynamic range. Here, we report thedirect chemical detection and quantitation of DPD in both culturesof E. coli and V. harveyi and from the effluent of oral bacteria inhuman saliva. These results are the first to quantitate DPD incomplex, clinically relevant media and show that DPD concentra-tions in the human mouth are sufficient to mediate behaviorsobserved in model systems.

EXPERIMENTAL SECTIONGeneral Chromatographic Details. High performance liquid

chromatography (HPLC) was performed utilizing a quaternarypump to generate a gradient for the elution of compounds fromthe stationary phase. For all samples, 10 µL was injected onto thecolumn via an autosampler cooled to 4 °C. A flow rate of 150 µL/min was used, and the eluent was introduced directly into theMS for ion detection. The mobile phases were HPLC grade water(solvent A) and HPLC grade acetonitrile (solvent B), and thesewere used to construct the following 18 min gradient elutionprofile: (t ) 0 min, 15% solvent A, 85% solvent B; t ) 2 min, 15%solvent A, 85% solvent B; t ) 4 min, 95% solvent A, 5% solvent B;t ) 14, 95% solvent A, 5% solvent B; t ) 16, 15% solvent A, 85%solvent B; t ) 18, 15% solvent A, 85% solvent B). The stationaryphase used for these studies consisted of aminopropyl function-alized particles (5 µm pore size, 100 Å particle size) packed intoa 250 × 2 mm column (Phenomenex Luna NH2). All separationswere performed with a column temperature of 10 °C.

General Mass Spectrometric Detection Parameters. AllMS analyses were performed on a TSQ Quantum Discovery Maxtriple quadrupole MS (Thermo Electron Corporation). Sampleswere introduced into the electrospray ionization (ESI) chamberthrough a 0.1 mm internal diameter fused silica capillary afterdelivery by HPLC as described above, or by direct infusion via asyringe pump at a flow rate of 20 µL/min. The spray voltage forthe ESI source was set to 4500 V, and detection occurred inpositive ion mode. Nitrogen was used as the sheath gas (10 psi),and the inlet capillary temperature was 300 °C. The argon usedas the collision gas was set at 1.5 mTorr. The scan time for eachSRM was 0.05 s with a scan width of 1 m/z. All peaks from theMass Chromatograms were integrated manually using the Xcali-bur MS software package (Thermo Electron Corporation).

Preparation of Synthetic DPD (or (13C)DPD) for Biologi-cal Studies. This molecule was synthesized as describedpreviously,20 and only general handling protocols are described

(20) Semmelhack, M. F.; Campagna, S. R.; Federle, M. J.; Bassler, B. L. Org.Lett. 2005, 7, 569–572.

(21) Vilchez, R.; Lemme, A.; Thiel, V.; Schulz, S.; Sztajer, H.; Wagner-Dobler, I.Anal. Bioanal. Chem. 2007, 387, 489–496.

(22) DeKeersmaecker, S. C. J.; Vanderleyden, J. Microbiology-Sgm 2003, 149,1953–1956.

(23) Hopfgartner, G.; Varesio, E.; Tschappat, V.; Grivet, C.; Bourgogne, E.;Leuthold, L. A. J. Mass Spectrom. 2004, 39, 845–855.

(24) Bajad, S. U.; Lu, W. Y.; Kimball, E. H.; Yuan, J.; Peterson, C.; Rabinowitz,J. D. J. Chromatogr. A 2006, 1125, 76–88.

(25) Rabinowitz, J. D. Expert Rev. Proteomics 2007, 4, 187–198.

6375Analytical Chemistry, Vol. 81, No. 15, August 1, 2009

here. Fresh DPD was prepared daily from a stock solution of (S)-4,5-cyclohexylidenedioxy-2,3-pentadione (Pro-DPD) in water (1mg/mL, 4.7 mM). This solution was routinely stored at -80 °Cfor a period of up to 3 weeks, and no change in the data wereobserved. When needed, 100 µL aliquots of the Pro-DPD solutionwere prepared by addition of 1 µL 10% H2SO4. Deprotection wasallowed to proceed for 3 h at ambient temperature, and theresulting DPD solution (4.7 mM) was used without modifica-tion for further experimentation.

Bacterial Strains and Growth Conditions. The E. coli wildtype strain BW25113 and luxS- strain JW2662-1 were estab-lished as part of the Keio Collection26 and were purchased fromthe Coli Genetic Stock Center at Yale University. For allexperiments, E. coli cultures were grown in Luria-Bertanimedium (LB) at 37 °C with aeration. The V. harveyi wild typestrain BB12015 and luxS- strain MM3027 were a gift from B.L.Bassler (Princeton University and the Howard Hughes MedicalInstitute). Overnight cultures of V. harveyi BB120 were grownin Luria Marine medium (LM)28 at 37 °C with aeration.Overnight cultures of V. harveyi MM30 were cultured in LMsupplemented with 100 mg/L kanamycin. For experiments inwhich the [DPD] was to be determined, both V. harveyi strainswere cultured in LM at 30 °C with aeration.

Collection of Saliva Samples. Volunteers were each askedto deposit ∼1 mL of mucus free saliva into a specimen cup. Thesample was then agitated to remove bubbles, and the saliva wasthen used for further experimentation as described below.

General Procedure for the Determination of the [DPD].For the detection of DPD in samples of E. coli or V. harveyi, 300µL aliquots were taken at regular intervals from duplicate cellcultures of both the wild type and luxS- strain grown as describedabove. Single 300 µL aliquots for each sample of saliva weregenerated as previously described. These aliquots were thenadded to a 1.5 mL microcentrifuge tube containing a solutionof the internal standard, (13C)DPD (341 µM in 10 µL H2O forbacterial samples, 34.1 µM in 10 µL H2O for saliva samples).The resulting solution was thoroughly mixed and then centri-fuged (13,200 rpm, 16,100 rcf, 1 min for bacterial samples, 5min in for saliva samples) to remove cells and other particu-lates. A portion of the resulting supernatant (200 µL) was thentransferred to a 300 µL screwcap autosampler vial containinga solution of tag 7 (14 mM in 20 µL H2O), and the two liquidswere thoroughly mixed. After 1 h incubation at ambienttemperature, the samples were quickly frozen and then storedat -80 °C until no more than 1.5 h prior to MS analysis.Duplicate injections of each saliva sample were performed. TwoSRMs, 381-202 and 382-203, were used to determine therelative signal arising from detection of DPD-M1CQ to thatof (13C)DPD-M1CQ. After manual peak integration, the datawere corrected for isotopic impurities as described in theSupporting Information, Discussion and Table S4, and thefollowing equation was used to calculate the [DPD]: [DPD] )

DPD-M1CQ signal/(13C)DPD-M1CQ signal × (13C)DPDconcentration.

RESULTS AND DISCUSSIONDetermination of the Optimal DPD Derivatizing Reagent

for LC-MS/MS Analyses. The facile reaction of DPD witho-diaminobenzene, 1, (Figure 1A) has been used to qualitativelydetect AI-2 as the resulting quinoxaline derivative, DPD-Q.29,30

In our hands, 1, was not useful for the LC-MS/MS quantitationof DPD; therefore, a set of functionalized o-diaminobenzenetagging reagents (Figure 1B) were produced as described in theSupporting Information. All of the tagging reagents showed someutility for the measurement of [DPD] via LC-MS/MS. Todetermine which quinoxaline was optimal to quantitate AI-2 whenintroduced via an ESI source, standard solutions of each weregenerated and analyzed on a triple quadrupole MS. Briefly,synthetic DPD (2.35 mM, pH ) 1.8) was reacted with each ofthe o-diaminobenzene tags, 2-7, in a 1:3 molar ratio. After thereaction was allowed to proceed for 2 h, the solutions were dilutedin water and introduced into the MS via direct infusion. In allcases, a peak was detected corresponding to the mass of theprotonated quinoxaline in positive ionization mode, and the mostsensitive SRMs for each quinoxaline were determined via a MS2

experiment (Supporting Information, Table S1). After suitabledetection parameters were established, each quinoxaline solutionwas injected onto an aminopropyl column and eluted using achromatographic method modified from standard metabolomicprocedures.24 Although a peak was detected for all compounds,the quinoxalines generated from 4-6 (DPD-1CQ, DPD-2CQ,and DPD-3CQ, respectively) gave no more than 10% of the signalmeasured for the other quinoxalines (Figure 2). The quinoxalinesderived from 3, collectively DPD-EBAQ, gave the most intensesignal, however, the peak shape for this molecule was notGaussian. Asymmetry was also seen in the peak from thequinoxalines derived from 2, collectively DPD-BAQ, presumablybecause of partial separation of the two regioisomeric quinoxalinesgenerated from the reaction with DPD (Figure 2 A-B). Althoughthe absolute peak intensity for the quinoxaline generated viareaction with 7 (DPD-M1CQ) was only 40% that of DPD-EBAQ,the fact that DPD-M1CQ was a single molecule with well behaved

(26) Baba, T.; Ara, T.; Hasegawa, M.; Takai, Y.; Okumura, Y.; Baba, M.;Datsenko, K. A.; Tomita, M.; Wanner, B. L.; Mori, H. Mol. Syst. Biol. 2006,2, 2006.008.

(27) Surette, M. G.; Miller, M. B.; Bassler, B. L. Proc. Natl. Acad. Sci. U.S.A.1999, 96, 1639–1644.

(28) Bassler, B. L.; Wright, M.; Silverman, M. R. Mol. Microbiol. 1994, 13, 273–286.

(29) Hauck, T.; Hubner, Y.; Bruhlmann, F.; Schwab, W. Biochim. Biophys. Acta2003, 1623, 109–119.

(30) Zhu, J. G.; Dizin, E.; Hu, X. B.; Wavreille, A. S.; Park, J.; Pei, D. H.Biochemistry 2003, 42, 4717–4726.

Figure 1. Quinoxaline formation and DPD tagging reagents used(A) Stable quinoxlines, such as DPD-Q, can be formed via thereaction of o-diaminobenzenes with R-diketones. (B) Molecularstructures of the DPD tagging reagents used in this study.

6376 Analytical Chemistry, Vol. 81, No. 15, August 1, 2009

chromatographic characteristics was promising (Figure 2C).Because of the ease of detection for DPD-BAQ, DPD-EBAQ,and DPD-M1CQ in the preliminary experiments, each of thesemolecules were subjected to further investigation.

While the LC-MS/MS detection of a molecule with an ap-propriate m/z for each quinoxaline gave strong evidence that thedesired tagging reactions had occurred, further confirmation forthe formation of these molecules was desired. To obtain thesedata as well as to probe the efficacy and product distribution foreach reaction, proton nuclear magnetic resonance (1H NMR)spectroscopy was employed. For these experiments, the reac-tions yielding DPD-BAQ, DPD-EBAQ, and DPD-M1CQwere studied by the addition of 2-3 mol equivalents of theappropriate tag in 1 mol equivalent portions to a solution ofsynthetic DPD (4.7 mM) in D2O, and the formation of productwas observed by 1H NMR. For DPD-BAQ, 1 equivalent of tagwas added to the DPD solution at 0, 1, and 2 h (SupportingInformation, Figure S2). Upon the addition 1 equivalent of tag2, the [DPD] decreased and the concentration of DPD-BAQincreased over the course of 1 h. However, the reaction did notproceed to completion. Further addition of 2 more equivalents of2 was also not sufficient to drive the reaction to completion after2.5 h. Beyond revealing that the reaction leading to DPD-BAQwas not facile, this experiment also confirmed the formation ofan ∼1:1 mixture of regioisomeric quinoxalines. Monitoring theformation of DPD-EBAQ was then carried out as described forDPD-BAQ, except 2 equivalents of tag 3 were sufficient to drivethe reaction to completion over a 2 h period (Supporting Informa-tion, Figure S3). Addition of 1 equivalent of 3 to DPD led to theconsumption of 50% of the tag over 1 h, and another equivalent of3 was added to the reaction. This addition was sufficient to drivethe reaction to completion after another hour. Although we werehopeful that the added steric bulk of the ethyl ester would favor

formation of one of the regioisomers of DPD-EBAQ, this wasnot the case as both were observed in an ∼1:1 ratio. Because ofthe sluggish reaction and undesirable mixtures of productsgenerated during the formation of DPD-BAQ and DPD-EBAQ,we refocused our attention on 7. When 1 equivalent of this tagwas added to synthetic DPD, reaction proceeded quickly, and 7was ∼90% consumed after 30 min. Addition of a second equivalentof tag led to complete conversion of the remaining 10% DPD toDPD-M1CQ after further reaction for 30 min (SupportingInformation, Figure S4). The efficiency of reaction coupled withthe generation of only a single product, led to the selection of 7as the tagging reagent for further experimentation.

Validation of the Sensitivity and Selectivity for DPDDetection. To determine the LOD for DPD-M1CQ, a standardsolution of the quinoxaline was used to generate a calibration plotof signal vs concentration (Supporting Information, Figure S4).Serial dilutions of a 2.35 mM solution of DPD-M1CQ into wateryielded calibration standards ranging in concentration from 74pM to 23.5 µM. The resulting standards were analyzed via LC-MS/MS by monitoring SRMs 381-363, 381-231, 381-202, and381-201 (reported as parent m/z-product m/z). From theseexperiments, DPD-M1CQ could be detected at all concentrationsof 230 pM or greater, and the signal was linear over the entiredetectable concentration range. The SRM, 381-363, correspond-ing to the loss of water had a poor signal-to-noise ratio (S/N) andwas not used for LOD determination. Monitoring the SRM381-202 was more sensitive than monitoring either the 381-231or 381-201 SRMs. The S/N of only ∼3:1 for SRM 381-202 at230 pM DPD was not sufficient for quantitation, although tentativedetection of DPD-M1CQ is still possible. The signal from SRM381-202 at the next highest [DPD], 740 pM, had a S/N of ∼5:1,which is sufficient for detection and quantitation of the molecule

Figure 2. Mass Chromatograms for each DPD derived quinoxaline. A 10 µL injection of a 10 µM solution for each was detected by an appropriateSRM after gradient elution from an aminopropyl LC column. (A) Chromatogram of DPD-BAQ as detected via SRM 249-231. (B) Chromatogramof DPD-EBAQ as detected via SRM 277-231. (C) Chromatogram of DPD-M1CQ as detected via SRM 381-202. (D) Chromatogram ofDPD-1CQ as detected via SRM 353-231. (E) Chromatogram of DPD-2CQ as detected via SRM 381-219. (F) Chromatogram of DPD-3CQas detected via SRM 409-219.

6377Analytical Chemistry, Vol. 81, No. 15, August 1, 2009

(Supporting Information, Figure S5). Further determination of theexact LOD was not carried out as the quantitation of the [DPD]in subsequent experiments was to be performed via comparisonof the signal for DPD-M1CQ to that observed for an isotope-labeled internal standard as described below.

A possible complication for the selectivity of this method isthat the Lobry de Bruin-van Ekenstein rearrangement or theMaillard reaction of carbohydrates can be promoted via theaddition of o-diaminobenzene at pHs higher than 7.5.19,31 Thesereactions can lead to the formation of quinoxalines identical orsimilar in structure to those derived from DPD. In this work, carewas taken to ensure that the pH remained at or below 7.2 duringsample preparation, handling, and analysis. To validate theselectivity of this derivatization and detection approach in complexmedia, tag 7 was added to Luria-Bertani Broth (LB) and allowedto incubate for 1 h. These samples were subjected to LC-MS/MSanalysis, and monitoring of the 3 most sensitive SRMs wasperformed to determine whether undesired side-reactions of 7with carbohydrates had occurred. For SRMs 381-201 and381-231, peaks with very low intensity were observed. Fortu-nately, SRM 381-202 did not have a peak derived from impurities.The samples were reanalyzed after sitting at 4 °C for 11 h, andno changes in the data were observed (Supporting Information,Figure S6). These data showed that SRM 381-202 is selectivefor detection of DPD-M1CQ.

Implementation of an Internal Standard for [DPD] Quan-titation. The addition of stable isotope-labeled internal standardshas become a powerful technique that allows absolute quantitationof analytes during MS detection without the need to construct anexternal calibration plot.32-34 Two options for the addition of aninternal standard during these experiments were possible. Thefirst option would rely on the addition of an isotopically labeledversion of the quinoxaline to provide internal calibration, similarto that reported in a recent GC-MS DPD detection protocol;19

and the second option would utilize the addition of isotope-labeledDPD as the internal standard (Figure 3). We were concernedthat the use of (13C)DPD-M1CQ would not give accuratequantitation because of differential loss of DPD and thequinoxaline during sample handling or because of incompletereaction of the tag with DPD. Both options were explored todetermine the most accurate method. For these experiments,the internal standard, either (13C)DPD-M1CQ or (13C)DPD,was added at a final concentration of 10 µM to aliquots of a V.harveyi BB120 culture grown for ∼8.5 h in Luria-Marine

Medium (LM). The samples were then centrifuged to removecells, and 130 equivalents of tag 7 were added. After reactionfor 1 h, LC-MS/MS was used to detect both the DPD-M1CQfrom V. harveyi via SRM 381-202 and the internal standardvia SRM 382-203. The [DPD] in each sample was thendetermined by taking the ratio of the integrated peak intensityfor DPD-M1CQ and (13C)DPD-M1CQ (corrected as de-scribed below) and multiplying by the concentration of addedinternal standard. For samples utilizing (13C)DPD-M1CQ asthe internal standard, the [DPD] from V. harveyi was measuredto be 5.6 µM. However, the use of (13C)DPD as an internalstandard led to a concentration measurement for the biologi-cally derived DPD of 12.3 µM (Supporting Information, TableS2). This confirmed our suspicion that DPD is either lost in thebiological matrix during handling or that the tagging reaction doesnot go to completion in complex media. However, the latterexplanation seems less plausible because of the large excess ofderivatizing reagent (>130 molar equiv tag 7) used in thesestudies. On the basis of these results, (13C)DPD was chosen asthe internal standard for further experimentation. A furtherdiscussion of the selection of proper SRMs and error correctionfor accurate internal quantitation is provided in the SupportingInformation, Discussion, Figure S7, Scheme S1, and Tables S3and S4.

Detection and Quantitation of DPD from BiologicalSources. To validate our AI-2 detection method in vivo, the[DPD] was measured in cell free supernatants for both WTand luxS- strains of two species, E. coli and V. harveyi. Thesespecies were chosen as the relative levels of DPD producedduring growth has been studied for both via severalmethods.15,17-19,28,35 Two aliquots from each of an E. coliBW25113 (wild type, WT) and E. coli JW2662-1 (luxS- strain)culture grown to stationary phase were diluted 50 fold infresh LB, and samples of the cultures were collected atregular intervals until the cells entered late exponential/early stationary phase. A final concentration of 10 µM(13C)DPD was added to each aliquot. After centrifugationto remove cells and other particulates had occurred, tag 7was also added to the samples at a final concentration of 1.3mM. Following 1 h incubation, the samples were subjectedto LC-MS/MS analysis. This led to facile detection of DPDin the WT E. coli supernatants, but not in those from theluxS- strain (Figure 4A). The [DPD] peaked at 3.5 h, and themolecule was then quickly consumed over the course of anhour as expected. Further, the maximum [DPD] was measuredto be 20.6 ± 0.1 µM. Measurement of the [DPD] in supernatantsharvested from V. harveyi proceeded analogously to thosedescribed for E. coli, except that LM was used. In V. harveyiBB120 (WT) the maximum [DPD] of 22.4 ± 0.6 µM wasreached at 6.5 h, before slowly decreasing and then beginningto level. Again, no signal was observed in the supernatants ofthe luxS- strain, V. harveyi MM30 (Figure 4B). To determinethe reproducibility of these results, the V. harveyi experimentswere performed on a separate day, and the maximum [DPD]in the WT strain was determined to be 22.1 ± 1.6 µM. Further,

(31) Glomb, M. A.; Tschirnich, R. J. Agric. Food Chem. 2001, 49, 5543–5550.(32) Wu, L.; Mashego, M. R.; van Dam, J. C.; Proell, A. M.; Vinke, J. L.; Ras, C.;

van Winden, W. A.; van Gulik, W. M.; Heijnen, J. J. Anal. Biochem. 2005,336, 164–171.

(33) Bennett, B. D.; Yuan, J.; Kimball, E. H.; Rabinowitz, J. D. Nat. Protoc. 2008,3, 1299–1311.

(34) Yuan, J.; Bennett, B. D.; Rabinowitz, J. D. Nat. Protoc. 2008, 3, 1328–1340. (35) Xavier, K. B.; Bassler, B. L. J. Bacteriol. 2005, 187, 238–248.

Figure 3. Structures of isotope labeled standards used for internalcalibration during measurement of [DPD].

6378 Analytical Chemistry, Vol. 81, No. 15, August 1, 2009

the intra- and interday variability among the cultures wasgreater than the error seen during multiple analyses of a singlesample (Supporting Information, Table S5).

After the successful detection and quantitation of the [DPD]from cultures of both E. coli and V. harveyi, we sought to applythis technology to the measurement of the [DPD] in a complex,

Figure 4. Production of DPD in relation to cell growth during the exponential growth phase. (A) Autoinducer production by E. coli. (B) Autoinducerproduction by V. harveyi. Measurement of the [DPD] for each WT species was performed in duplicate and the average data are reported.Although experiments for the luxS- strains were performed in duplicate, [DPD] measurement for every time point was only performed once. Forselect time points, measurement was conducted in duplicate to confirm that no DPD was detected in the luxS- strains. The average OD600 forall four cultures, two WT, and two luxS-, was used as no growth differences were observed between the strains. Error bars in the graphsindicate the range of the data.

6379Analytical Chemistry, Vol. 81, No. 15, August 1, 2009

clinically relevant medium. Human saliva was chosen as measure-ment of the [DPD] in bacterial model systems mimicking theoral cavity has been performed via the V. harveyi bioluminescenceassay.16 Saliva from eight volunteers was collected in a specimencup and then immediately spiked with 1 µM (13C)DPD. Aftercentrifugation to remove any particulates, an aliquot of thesupernatant was taken, and tag 7 was added to a concentrationof 1.3 mM. The LC-MS/MS analysis was then performed induplicate for each sample after the reaction with DPD had beenallowed to proceed for 1 h. In all cases, DPD was detected,and the average [DPD] was measured to be 526 nM withvalues ranging from 244-965 nM (Table 1). It was also notedthat the concentration in two of the samples was considerablyhigher than for the others, although the exact cause of this is asof yet undetermined.

CONCLUSIONWhile DPD has been implicated as a universal signal that may

be responsible for mediating many bacterial behaviors necessaryfor colonization or infection, the lack of methods for detectionhas left open questions as to whether the signal is relevant invivo. The instability of DPD makes quantitation of this moleculedifficult; however, we have developed a selective derivatizationand LC-MS/MS analysis protocol that allows facile detection froma variety of sources.

Several features of our derivatizing reagent facilitate detectionof DPD. Although measurement of [DPD] via the quinoxalineDPD-Q is known,29,30 this molecule does not have sufficientchemical characteristics for optimal detection by tandem MSmethods. Specifically, this molecule does not undergo a largeenough number of collision induced fragmentations to allow forsuitable SRM selection. However, a new fragmentation manifoldbecomes available for DPD-M1CQ because of the addition ofether appended ester functionalities to the DPD-Q core. Thisallowed for sensitive and selective SRMs to be determined.Electron donation from the ether oxygens is also responsible for

the enhanced nucleophilicity of diamine 7 and is likely responsiblefor improving the basicity of the DPD-M1CQ nitrogens. To-gether, these characteristics not only speed quinoxaline formationbut also increase the fraction of molecules protonated during MSdetection.

The chromatographic and detection methods chosen for thiswork were carefully selected for both utility and sensitivity. Forthe chromatography, a method derived from published metabo-lomic techniques was utilized as we also desire to detect othermetabolites during future experimentation. For this report, the18 min run time used for separation was to ensure that othermolecules with identical parent m/z-fragment m/z combinationsdid not co-elute with DPD-M1CQ. However, the lack of peaks inSRM 381-202 attributable to molecules other than DPD will allowshortening the LC method, and work is underway to optimize thethroughput of our procedure. Having a molecule capable ofmultiple fragmentation events was critical to allow the selectivemeasurement of the [DPD-M1CQ]. Other matrix componentswere observed to react with tag 7 to generate molecules with thesame parent mass; however, the inability of these molecules toproduce a fragment with m/z 202 facilitated separation via theMS. Further, the choice of (13C)DPD as the internal standardalso proved critical as addition of preformed (13C)DPD-M1CQled to underestimation of the [DPD].

Our efforts to detect and quantitate AI-2 from biologicalsources produced good results, and the [DPD] was easilymeasured in two well studied species of bacteria, E. coli and V.harveyi. We determined that a maximum [DPD] of 20.1 µM wasreached at 3.5 h in E. coli BW25113 cultures after which theautoinducer concentration rapidly decreased to ∼450 nM. Similarbehavior has been qualitatively observed for E. coli MG1655 when[AI-2] is monitored via the V. harveyi luminescence bioassay.35

For V. harveyi BB120 grown in LM, a maximum [DPD] of 22.4µM was observed at 6.5 h. Similar dynamics for the productionof AI-2 by V. harveyi have been observed previously,17 andquantitation of the [AI-2] in closely related strains of this species

Table 1. Quantitation of the DPD Concentration in Saliva Collected from Eight Volunteersa

volunteerobserved signal

for DPD-M1CQ (ion counts)observed signal

for (13C)DPD-M1CQ (ion counts)corrected signalb for

(13C)DPD-M1CQ (ion counts)[DPD](nM)

average[DPD] (nM)

1 injection 1 9.32 × 104 2.37 × 105 2.23 × 105 417 406injection 2 1.50 × 105 4.02 × 105 3.81 × 105 3952 injection 1 1.12 × 105 2.51 × 105 2.35 × 105 477 474injection 2 1.58 × 105 3.59 × 105 3.36 × 105 4713 injection 1 2.08 × 104 5.40 × 104 5.10 × 104 407 426injection 2 6.72 × 104 1.61 × 105 1.51 × 105 4444 injection 1 1.32 × 105 2.97 × 105 2.78 × 105 473 464injection 2 1.63 × 105 3.83 × 105 3.60 × 105 4545 injection 1 1.66 × 104 4.42 × 104 4.18 × 104 398 400injection 2 1.15 × 105 3.01 × 105 2.85 × 105 4036 injection 1 3.27 × 105 4.46 × 105 3.99 × 105 819 832injection 2 2.68 × 105 3.55 × 105 3.17 × 105 8457 injection 1 1.11 × 105 4.35 × 105 256 244injection 2 7.09 × 104 3.06 × 105 2328 injection 1 3.41 × 105 3.98 × 105 3.49 × 105 977 965injection 2 4.09 × 105 4.89 × 105 4.30 × 105 952

average 526

a Except for volunteer 7, data for all samples were corrected using the following calculation: (13C)DPD-M1CQ Signal ) Obs. (13C)DPD-M1CQ Signal - 0.144 × Obs. DPD-M1CQ Signal. b For a discussion of data correction see the Supporting Information, Discussion, Figure S7,Scheme S1, and Tables S3 and S4.

6380 Analytical Chemistry, Vol. 81, No. 15, August 1, 2009

has led to measurement of 1.95 µM DPD in V. harveyi BB15219

and of ∼40 µM DPD in V. harveyi BB12017 when grown inminimal media. Our measured [DPD] for V. harveyi was inbetween those previously reported, and some of this variation isdue to the use of rich media. In no cases was signal arising fromDPD observed during studies involving luxS- strains. Takentogether, these results validate the usefulness of our DPDquantitation methodology.

Once validated, we used this method to perform an initialdetermination of the [DPD] from complex, clinically relevantsamples. As AI-2 mediated quorum sensing has been postulatedto be critical for colonization of the human mouth by oralbacteria,9,36 we sought to determine the [DPD] in saliva. Theresults of these efforts have shown that the average [DPD] insaliva is 526 nM, a concentration sufficient to mediate thecolonization of oral bacteria during model studies. Indeed, thisconcentration is within 2-fold of that observed for studies of S.oralis 34 and A. naeslundii T14V cultured in 25% saliva.16 Furtherwork is underway to determine the changes in [DPD] in response

to differing environmental stimuli on the oral bacteria and to detectthis molecule from other complex environments.

ACKNOWLEDGMENTStrains of V. harveyi used in this study were a kind gift from

Bonnie L. Bassler (Princeton University and the Howard HughesMedical Institute). The authors thank Alex H. Rickard (Bingham-ton University, SUNY) and Joshua D. Rabinowitz (PrincetonUniversity) for helpful discussion concerning the need for andimplementation of the detection method, and David C. Baker(University of Tennessee, Knoxville) and Michael D. Best (Uni-versity of Tennessee, Knoxville) for input on the chemicalsyntheses and general discussion. This work was supported bystart-up funds provided to S.R.C. by the University of Tennessee.

SUPPORTING INFORMATION AVAILABLEAdditional information as noted in the text. This material is

available free of charge via the Internet at http://pubs.acs.org.

Received for review April 16, 2009. Accepted June 29,2009.

AC900824J

(36) Rickard, A. H.; Palmer, R. J.; Blehert, D. S.; Campagna, S. R.; Semmelhack,M. F.; Egland, P. G.; Bassler, B. L.; Kolenbrander, P. E. Mol. Microbiol.2006, 60, 1446–1456.

6381Analytical Chemistry, Vol. 81, No. 15, August 1, 2009