Presence of acyl-homoserine lactones in 57 members of the Vibrionaceae family

Aims The aim of this study was to use a sensitive method to screen and quantify 57 Vibrionaceae strains for the production of acyl-homoserine lactones (AHLs) and map the resulting AHL profiles onto a host phylogeny. Methods and Results We used a high-performance liquid chromatography–tandem mass spectrometry (HPLC-MS/MS) protocol to measure AHLs in spent media after bacterial growth. First, the presence/absence of AHLs (qualitative analysis) was measured to choose internal standard for subsequent quantitative AHL measurements. We screened 57 strains from three genera (Aliivibrio,Photobacterium and Vibrio) of the same family (i.e. Vibrionaceae). Our results show that about half of the isolates produced multiple AHLs, typically at 25–5000 nmol l−1. Conclusions This work shows that production of AHL quorum sensing signals is found widespread among Vibrionaceae bacteria and that closely related strains typically produce similar AHL profiles. Significance and Impact of the Study The AHL detection protocol presented in this study can be applied to a broad range of bacterial samples and may contribute to a wider mapping of AHL production in bacteria, for example, in clinically relevant strains.


Introduction
Bacterial cells produce and excrete small molecules into their surroundings to control a number of cellular processes (Ng and Bassler 2009), such as bioluminescence (Meighen 1994), swarming (Eberl et al. 1996), biofilm formation, plasmid conjugal transfer (Fuqua et al. 1994), production of exo-enzyme virulence determinants in pathogens (Jones et al. 1993), antibiotic biosynthesis (Lowery et al. 2009), iron chelation (Kaufmann et al. 2005) and fouling (Cuadrado-Silva et al. 2013). There are currently at least two alternative hypotheses for this behaviour (West et al. 2012): (i) quorum sensing (QS), which assumes that the produced molecules are used as signals for cell-cell communication (i.e. a social trait), and (ii) diffusion sensing (DS), which assumes that the produced molecules are used to monitor the rate of diffusion (i.e. the relative speed of molecules that move away from the cell).
Several bioassays have been used to detect and quantify AHLs. In these assays, the presence of AHL compounds induces phenotypic changes (e.g. bioluminescence or b-galactosidase activity). Bioassays based on Agrobacterium tumefaciens and the b-galactosidase reporter system (Cha et al. 1998) are triggered by 3-oxo-HSL-induction of the b-galactocidase, and it was recently shown that A. tumefaciens was able to detect AHLs as low as 2-14 pmol l À1 (Baldrich et al. 2011). A second assay based on Chromobacterium violaceum produces the purple compound violacein in the presence of short AHLs with no substitutions on the C3 carbon on the acyl chain (McClean et al. 1997). Bacterial extracts separated by thin-layer chromatography (TLC) before subjection to a biosensor assay may further discriminate between different AHL types (Ravn et al. 2001).
Mass spectrometry (MS) in combination with either gas chromatography (GC) or liquid chromatography (LC) can both be used for detecting and quantifying AHLs. Although GC-MS has the advantage of being a rapid, simple and selective method, it is also associated with unwanted degradation of AHLs during the analysis, which results in low sensitivity (Osorno et al. 2012). LC-MS is particularly useful in identifying/quantifying particular molecules in a mixture of different chemicals (Wang et al. 2011). The technique has undergone significant development with respect to quantification levels with the utilization/refinement of, for example, nano-LC (Frommberger et al. 2004;Fekete et al. 2007) and Fourier transform ion cyclotron resonance MS (Cataldi et al. 2009(Cataldi et al. , 2013. Although the required volume of samples has declined, it still has remained relatively high. The first AHL study used 6-l sample volumes for extraction of AHLs (Eberhard et al. 1981); later it was reduced to 500 ml (Shaw et al. 1997), 20 ml (Yates et al. 2002) and 500 ll (Frommberger et al. 2004). In this work, we have used 75-ll sample volume, with a sample processing rate on the LC-MS machine at >200 samples per day. We also wanted to make the method affordable, so isotopically labelled internal standards were excluded.
The Vibrionaceae family consists of Gram-negative marine bacteria, many of which are responsible for diseases in humans (Valiente et al. 2009) and marine animals (Milton 2006), and AHLs of QS systems likely play important roles in pathogenesis (Hentzer and Givskov 2003;Rasch et al. 2004;Nhan et al. 2010). In this study, we used a high-performance liquid chromatography-tandem mass spectrometry (HPLC-MS/MS) protocol in selective reaction monitoring (SRM) mode for determining the presence/absence (qualitative analysis) and concentrations of fifteen different AHLs and applied this method to spent media. We tested our method on A. fischeri ES114 to compare with previous data (Lupp and Ruby 2004;Stabb et al. 2008). We screened 7 Vibrio, 5 Photobacterium and 45 Aliivibrio isolates and included among others Vibrio anguillarum, Vibrio splendidus, Aliivibrio salmonicida, A. fischeri, Aliivibrio wodanis, Aliivibrio logei and Photobacterium phosphoreum. AHL measurements were finally mapped onto a host phylogeny to evaluate whether closely related species and strains produce similar AHL profiles.
When using SWT during the initial study, samples were collected at several optical densities, whereas when using LB25 during the screening, the culture was harvested after 50 h of growth and treated as described below.
Strains listed in Table 1 were grown in duplicates at 12°C in LB25 (as described above). The new strains reported in this study are from an in-house bacterial collection. Starting cultures were inoculated at OD 600 0Á001 and harvested after 50 h of growth. Samples were centrifuged at 1400 g in a Beckman Coulter Allegra X-15R tabletop centrifuge with SX4750A rotor for 20 min at 4°C. Four 75-ll aliquots (technical replicates) were acidified with 4 ll 1 mol l À1 hydrochloric acid to a final concentration of 51 mmol l À1 and stored at À20°C. One replicate was used for qualitative analysis, and the remaining three were used for quantitative analysis (described below).
Escherichia coli ArcticExpress (DE3) was grown overnight at 30°C in LB medium.

Sample preparation and HPLC-MS/MS analysis
The concentration of AHLs in spent media was measured as follows: AHLs were extracted from spent media and compared to commercial AHL standards with known concentrations which were extracted from fresh media using the same extraction protocol, as described below. Hence, it was the relative relationships between measured AHLs from spent media and the corresponding values from commercial AHLs that were used to calculate the final AHL concentrations in samples.
Four acidified 75-ll technical replicates, which were collected at each time point or OD, were each subjected to three consecutive extractions with three volumes (225 ll) of ethyl acetate. The ethyl acetate phase from the three extractions was pooled (675 ll) and dried using a rotary vacuum centrifuge (savant instruments Inc. model PH40-11), and samples were finally stored at À20°C for subsequent HPLC-MS/MS analysis. One technical replicate was redissolved in 150 ll of 20% acetonitrile with 0Á1% formic acid for qualitative HPLC-MS/MS analysis. Similarly, three technical replicates were redissolved in 20% acetonitrile with 0Á1% formic acid, which contained a fixed amount of an internal AHL standard [100 ng/ml (i.e. 330-400 nmol l À1 depending on the molecular weight of the AHL]. HPLC was performed by separating 20 ll of the 150 ll from each sample on a Hypersil GOLD C18 reverse-phase column (50 9 2Á1 mm, 1Á9 lm particle size; Thermo Fisher Scientific, Waltham, MA, USA) equilibrated with 5% acetonitrile, using a Accela autosampler and pump (Thermo Fisher Scientific). Mobile phase was 0Á1% formic acid in water and 0Á1% formic acid in HPLC-grade acetonitrile, and these were used to create the following gradient elution profile: 5% acetonitrile for 18 s, a linear increase in acetonitrile to 90% over 162 s and then isocratic 90% acetonitrile for 60 s. Then in next 60 s, the column was equilibrated with 5% acetonitrile. The column was then re-equilibrated for 60 s with 5% acetonitrile before the next sample injection. The flow rate was 500 ll min À1 for all steps, and the column was kept at 30°C for all separations.
The HPLC-separated compounds were introduced into the Ion Max electrospray ionization probe of the LTQ Orbitrap XL machine (Thermo Fisher Scientific). The probe was equipped with a 32-gauge stainless steel needle. Nitrogen was used for all type of gasses in the ion source. Electrospray settings were as follows: sheath gas flow rate 70, auxiliary gas flow rate 10, sweep gas flow rate 10, spray voltage +4Á50 kV, capillary temperature 330°C, capillary voltage 40 V and tube lens 90 V. The protonated parent ions from the electron spray were next selected and fragmentized with CID using LTQ (Linear Ion Trap Quadrupole) tandem MS in SRM mode under positive ion conditions. When subjected to tandem MS, all AHLs are broken down to the same fragment with the theoretical value m/z 102Á055. This fragment contains only the head (lactone) of the AHL (see Fig. 1). CID parameters during tandem MS were as follows: isolation width = 1Á7, normalized collision energy = 30, act Q = 0Á250 and act time = 30 ms for all components. Maximum injection time was 25 ms, and automatic gain control (AGC) target setting is 1Á00E+04 ions.
SRM mode involves selecting a parent ion and breaking it with a constant energy to give fragment ions. One such parent-fragment pair is called an SRM transition. To minimize the number of simultaneous SRM transitions and maximize the number of data points over each peak, the data collection window was divided into 6 segments (for details see Table S1). Data were collected from m/z 101-103. Ion chromatograms were extracted from the raw data files from m/z 101Á8-102Á6 for each compound (see Fig. S1 for typical ion chromatograms). The peaks were automatically integrated with optimized integration settings for each compound using Thermo Excalibur 2Á07. For the entire study, one-point calibration was used (instead of using standard curves), as previously described (Peters and Maurer 2007). One-point calibration was used for each purchased standard AHL. Calibration was performed in triplicates for each sample run, in same growth media as in samples.
In quantitative analyses, an internal standard was added to all biological samples and AHL standard mixture in equal amount to correct for deviations in injection volume of the autosampler and to correct for changes in the instrument sensitivity. 3-oxo-C12-HSL, 3-OH-C12-HSL, C10-HSL and C12-HSL were chosen as internal standards depending on the test sample. The AHL standard mixture was made fresh for each run. The limit of detection (LOD) and limit of quantification (LOQ) were determined (LOQs listed in Table 2) from the mix of synthetic AHLs serially diluted into growth medium to a minimum concentration of 0Á2 nmol l À1 , extracted and subjected to HPLC-MS/MS analysis. LOD was defined as the lowest AHL concentration where the signal-to-noise ratio was ≥ 3 : 1 (MacDougall et al. 1980), and LOQ was theoretically set to LOD multiplied by 3Á33 (Thomsen et al. 2003).

Phylogenetic analysis
The host phylogeny was based on 16S rDNA sequences from forty-five Vibrionaceae strains, and Photorhabdus luminescens and Grimontia hollisae were used as the outgroup. Available 16S rDNA sequences were obtained from GenBank, and for the remaining strains, 16S rDNA was PCR-amplified and sequenced. Accession numbers are provided in Table 1. The partial 16S rDNA sequences were aligned automatically by ClustalW (Larkin et al. 2007) and manually edited in BioEdit v.7Á0Á9 (Hall 1999). Sequence positions that could be reliably aligned were used, and the final alignment consisted of 852 nucleotide positions and 45 sequences (Dataset S1). The phylogenetic tree was constructed in MEGA v.5.03 (Tamura et al. 2011) using the neighbour-joining (NJ) method and the maximum composite likelihood substitution model. Stability of nodes was tested with a NJ bootstrap analysis (maximum composite likelihood model and 10 000 pseudo-replicates).
Basic chemical structure of AHLs and their fragmentation products produced during tandem MS in SRM mode. Chain lengths of AHLs screened in this study were from C4 to C12 with an increment of two carbons. Functional groups of the third carbon were 3-oxo, 3-OH or unsubstituted.
AHL, Acyl-homoserine lactones; LOQ, limit of quantification. Calculated as 3Á33 times of LOD values (signal-to-noise ratio better than 3:1), practically its values with signal-to-noise ratio are greater than 10:1.

AHL detection method
HPLC followed by tandem MS in SRM mode was used to measure concentrations of AHLs. As a standard protocol, HPLC was run with 5 min from injection to elution and 1-min column equilibration before next injection. To minimize the number of simultaneous SRM transitions and maximize the number of data points over each peak, the data collection window was divided into 6 segments (for details see Table S1). Before running biological samples, we first determined the LOD and LOQ. A mix of synthetic AHLs was serially diluted in water or growth medium (LB, LB25 or SWT) to a minimum concentration of 0Á2 nmol l À1 , extracted and subjected to our HPLC-MS/MS protocol (see Material and methods). Three replicates were performed for each test concentration. LOD was defined as the lowest AHL concentration where the signal-to-noise ratio was ≥ 3 : 1 (MacDougall et al. 1980), and LOQ was theoretically set to LOD multiplied by 3Á33 (Thomsen et al. 2003). Table 2 shows the LOQ values. For C4-chained AHLs and longer-chained AHLs, LOQ ranged between 6Á33-20Á06 and 0Á76-10Á35 nmol l À1 , respectively. In other words, to reliably quantify, for example, C4-chained AHLs, the AHL concentration in a sample must be ≥ 6Á33-20Á06 nmol l À1 depending on which medium is used.
Next, the linearity of AHL measurements was tested over a broad range of concentrations (15-3000 nmol l À1 ), in water or same media as described above. Measurements were plotted, and resulting R 2 values between 0Á9781 and 0Á9996 (for LB, LB25, SWT media and H 2 Oonly LB25 data are shown in Fig. S2) demonstrated linear relationships between expected AHL concentrations and experimental measurements. Accuracy (i.e. deviation from the true value) was determined by adding 14 synthetic AHLs (C10-HSL was arbitrarily chosen as the internal standard) at three different concentration ranges (15-30, 900-1800 and 1300-2700 nmol l À1 ) to a supernatant of Escherichia coli ArcticExpress (DE3) overnight culture (Table S2). Samples were subjected to HPLC-MS/MS. This strain was chosen because it does not carry AHL-producing enzymes (Van Houdt et al. 2006). The majority of values were within 10% of the expected concentrations, with three exceptions in lowest concentration range (15-30 nmol l À1 ), that is, C4-HSL (expected/detected=31 : 42 nmol l À1 ), 3-OH-C6-HSL (25 :22 nmol l À1 ) and 3-OH-C8-HSL (23 : 16 nmol l À1 ). Similarly, there were two deviations above 10% for medium and higher concentration ranges, that is, 3-OH-C4-HSL (1568 : 1410 and 2352 : 2028 nmol l À1 ) and 3-oxo-C6-HSL (2180 : 1860 nmol l À1 ). We suspect that major contributors to these variations are measurement errors (especially at low AHL concentrations) and degradation of the analytical standards. Measurement errors would include variations between technical replicates (random error) and lowered accuracy because all AHLs are compared to one single internal standard (systematic error). The most important systematic error in electrospray MS is the phenomenon known as ion suppression, where matrix components eluting at the same time as the analyte suppress the signal of the analyte. Random errors could be minimized by increasing the number of replicates. Ion suppression could be minimized by several approaches like better sample preparation, running standards in the exact same matrix as the analyte (performed as close as possible), altered chromatographic profile, diluting the sample, changing the mobile phase's organic solvent, spiking the sample with standards (timeconsuming) and so on. The best method is likely to use one stable isotopically labelled internal standard (SIL) for each AHL. The SIL should preferably have the exact same retention time as the analyte to be exposed to the same random spray variations and ion suppression effects. Deuterium labelling might give slightly different retention times, therefore 13 C-, 15 N-or 17 O-labelled standards are preferred. These are, however, usually unavailable or very expensive.

AHLs in Aliivibrio fischeri ES114
Next we wanted to test our method using the model bacterium A. fischeri ES114 and compare AHL measurements with previous results (Stabb et al. 2008). For cultivation of A. fischeri ES114, we chose SWT medium and three different temperatures, that is, 12, 22 and 30°C. For each temperature, two cultures (biological replicates) were started at OD 600 0Á05, and samples were collected at different optical densities. Using 3-oxo-C12-HSL as internal standard [based on qualitative HPLC-MS/MS (SRM) analysis], the concentration of individual AHLs from the resulting growth culture supernatants was measured in triplicates. Table 3 shows the result which suggests that A. fischeri ES114 typically contains six AHLs (when grown in SWT medium) above LOQ, with highest concentrations of 3-oxo-C6-HSL and C8-HSL. Interestingly, the 3-oxo-C6-HSL profile was highly dependent on temperature. For example, at 12°C, 3-oxo-C6-HSL was found at 670 AE 120 and 1380 AE 320 nmol l À1 , at OD 600~0 Á5 and~1Á0, respectively, but was found below LOQ when cells were grown at 30°C (although below LOQ, our measurements suggest approx. 0Á15 nmol l À1 at OD 600~1 Á5, data not shown). In contrast, concentrations of C8-HSL appeared less dependent on temperature. Other AHLs that were detected above LOQ include C4-HSL, C6-HSL, 3-OH-C8-HSL and C10-HSL. Ruby and co-workers (Stabb et al. 2008) reported two AHLs in A. fischeri ES114, that is, 3-oxo-C6-HSL and C8-HSL. 3-oxo-C6-HSL was measured at 0Á01, 0Á1 and 0Á2 nmol l À1 at OD 595 0Á5, 1Á5 and 2Á8, respectively, which is very similar to what we found at 30°C, but several orders of magnitude lower than our measurements at 12 and 22°C. Furthermore, Stabb et al., reported C8-HSL at 35 nmol l À1 at OD 595 0Á5, which is in close agreement with our measurement (80 AE 20 and 53 AE 3 at 12 and 22°C, respectively), but at OD 595 = 1Á5, C8-HSl was measured at 1100 nmol l À1 . Lupp and Ruby (Lupp and Ruby 2004) measured C8-HSL using a bioassay, also using A. fischeri ES114. After growth in SWT medium at 28°C, they reported approximately 2, 20 and 80 nmol l À1 at OD 600~0 Á5,~1Á5 and~3Á5. In summary, we conclude that A. fischeri ES114 may contain six different AHLs, typically with 3-oxo-C6-HSL and C8-HSL at highest concentrations. Measurements of 3-oxo-C6-HSL and C8-HSL appear to be in general agreement with previous data (Lupp and Ruby 2004;Stabb et al. 2008), although the concentration of AHLs can greatly vary depending on growth parameters.

Screening of 57 Vibrionaceae strains for AHL production
Following the results from A. fischeri ES114, we screened an additional 56 Vibrionaceae strains, which were grown in two biological replicates in LB25 at 12°C and harvested after 50 h of growth. LB25 was used for convenience and because this medium could support growth of all tested strains. (The majority of tested strains are of Arctic or sub-Arctic marine origin.) AHLs were measured in three technical replicates per biological replicate using our HPLC-MS/MS protocol, and AHL concentrations were compared to known standards. Variations between biological replicates were generally low (Table S3). For simplicity, the final AHL measurements were organized into three categories, that is, 'high' (>5000 nmol l À1 ), 'medium' (25-5000 nmol l À1 ) and 'low' (<25 nmol l À1 ) and mapped accordingly onto a 16S rDNA host phylogeny (Fig. 2). Data from 24 strains are presented, whereas the remaining 33 were excluded from the figure because of redundancy in the data (listed in Table S4). Twenty-one nontested strains were included in the tree as reference strains to increase the phylogenetic diversity in the data set. Figure 2 shows that (i) tested strains (except Aliivibrio wodanis SA12) group into one of six phylogenetically distinct clades (named Clades I-VI), and (ii) half (12 strains) produce multiple AHLs, two strains produce two AHLs, four strains produce one AHL, and we were unable to detect AHL in six (nine in total of 57; see Table S4).  (2004) The tandem MS SRM method was used for quantitative analysis with internal standard (3-oxo-C12-HSL). Values are given as nmol l À1 of AHLs calculated against one-point calibration of known standards. Two parallel colonies were grown and harvested at different optical densities. Tested samples were grown at 12, 22 and 30°C in SWT medium. Cultures were grown from OD 600 0Á05.
All representatives of Clade I produce 8-11 different AHLs, with 3-oxo-C6-HSL in 'high' quantities. A. wodanis SA12 is placed immediately outside of Clade I, but shows an AHL profile that resembles those of Clade I. In Clade II, A. fischeri ES114 was the only representative tested, and detectable levels of 5 AHLs were identified. This contrasts the earlier finding when using SWT medium where we found six AHLs (C4-HSL was below LOQ when grown in LB25). Clade III represents a group of A. wodanis strains that produce 'medium' amount of 3-OH-C10-HSL, except for A. wodanis 01/401, which also produces 'medium' amount of C8-HSL. Clade IV represents two Aliivibrio strains (Aliivibrio R8-63 and Aliivibrio R8-67) that are phylogenetically distinct from other Aliivibrio species and have no apparent AHL production. Several representatives from the Vibrio genus were grouped into Clade V and show somewhat different AHL profiles. Specifically, no AHLs were detected in two members, that is, Vibrio sp. B9-25 and V. tapetis 99/196. However, it should be noted that the B9-25 strain grew slowly and had only Figure 2 Detected AHLs mapped onto a 16S rDNA ML phylogeny. Bootstrap values above internal nodes were calculated using the NJ method and the maximum composite likelihood model. Clades containing strains that were tested for AHL production are named Clades I-VI. GenBank accession numbers are shown in brackets for species that were added to increase the phylogenetic diversity in the tree. Numbers in parentheses to the left of Clades I-VI denote the number of strains in each clade that were tested, but excluded from the final tree because of redundancy in the data. Asterisks denote strains that grew slowly with a final OD 600 =0Á1 when harvested after 50 h of growth. Complete list of results is shown in Table S3. Concentrations of AHLs are grouped into four categories: 'high' >5 lmol l À1 (+++), 'medium' 25-5000 nmol l À1 (++), 'low'<25 nmol l À1 (+) and values below LOQ (-).

Discussion
We have used a sensitive HPLC-MS/MS method and AHL standards to first detect (qualitative analysis) and then quantify (with an appropriate internal standard) the presence of fifteen different AHLs in 57 different Vibrionaceae strains, five, seven and 45 of which belong to the Photobacterium, Vibrio and Aliivibrio genera, respectively. AHL quantification was mapped onto a host phylogeny tree. Typically, closely related strains produced similar AHL profiles, but different profiles were also found among closely related strains (e.g. Vibrio splendidus).

HPLC-MS/MS as a method to detect and quantify AHLs in bacterial cultures
Methods to detect AHLs in biological samples are becoming increasingly more sensitive, and different types of MS protocols have recently been developed with the purpose to find these signalling molecules in low concentrations, for example, picomoles of AHLs (Makemson et al. 2006). However, the use of such protocols for screening purposes has been hampered by several factors. MS methods typically suffer from the demand of relatively large sample sizes in order to detect low concentrations of AHLs. In addition, extraction and analysis procedures are timeconsuming and relatively costly. Some protocols include solid-phase extraction to isolate AHLs (Fekete et al. 2010), and this further adds to the handling time. In this study, we set out to facilitate small sample sizes and minimum handling times, along with scalability and automated analysis of samples. In our extraction protocol, we used 75 ll spent growth media and ethyl acetate as the extraction solvent. AHLs in spent growth media were stabilized with hydrochloric acid before storage at À20°C and subsequent HPLC-MS/MS analysis. Samples could be safely stored for 2 months without affecting the quality of the HPLC-MS/MS analysis. The ethyl acetate extracts were dried and redissolved in 20% acetonitrile, which was run on HPLC-MS/MS. During LC, acetonitrile was used as the eluent for the reverse-phase chromatography. Methanol has been used in the majority of earlier work (Hanzelka et al. 1999;Yates et al. 2002), but it was recently shown that methanol forms more adducts compared to acetonitrile (Cataldi et al. 2008). The extracted AHLs were typically run in sets of 200 samples per day. MS/MS in SRM mode was regarded as the suitable method, based on previous reports (Morin et al. 2003;Gould et al. 2006;Ortori et al. 2011;May et al. 2012), because the SRM method provides better selectivity for C4 AHLs (3-OH-C4-HSL, C4-HSL, 3-oxo-C4-HSL) than full-scan Orbitrap. Our results also showed that C4-chained AHLs were more difficult to detect and quantify compared to the longer-chained AHLs. There are two main reasons for this. First, C4-chained AHLs are detected as relatively broad peaks near the injection peak where the mass baseline (noise level) is high (Fig. S1). Second, the C4-chained AHLs elute when the mobilephase solvent consists mostly of water (approximately 90% water and 10% acetonitrile), which then typically also contain more metal ions such as sodium and potassium. The AHLs are, therefore, more likely to form, for example, sodium adducts, which create additional problems during the interpretations of collected data. Chances of running into problems with sodium and other unwanted adducts are, however, reduced by using acetonitrile and water as mobile-phase solvents (Cataldi et al. 2008).
Perhaps the most important limitation to experiments as described here is that the AHL production in bacteria is greatly dependent on growth parameters. Differences in, for example, medium composition, growth temperatures, starting optical densities and shaker speed can greatly affect the results. Hence, as we have experienced during this study, reproducing the growth conditions and results from previous studies can be very difficult. Moreover, different species and/or isolates can have very different growth requirement; hence, direct comparison of results from different species/isolates can also be challenging.

AHL production in Vibrionaceae
In Vibrionaceae, the production of specific AHLs has been carefully studied for several model bacteria, for example, Aliivibrio fischeri (Kuo et al. 1994), Vibrio anguillarum (Buchholtz et al. 2006) and Vibrio harveyi (Cao and Meighen 1989), and their AHL-producing enzymes have been characterized in detail (Schaefer et al. 1996;Milton et al. 1997Milton et al. , 2001Hanzelka et al. 1999). In A. fischeri, AinS is responsible for the production of C8-HSL (Hanzelka et al. 1999), whereas LuxI is responsible for the production of 3-oxo-C6-HSL and C6-HSL (Kuo et al. 1994). Our results suggest the existence of at least six AHLs in A. fischeri ES114. No previous study has reported this many AHLs in A. fischeri. One explanation for this could be that previous studies have used techniques, for example, field desorption (FD) ionization with double-focusing (magnetic and electric sectors in series) MS (Eberhard et al. 1981), desorption chemical ionization probe and electron impact ionization with magnetic sector MS (Kuo et al. 1994), that differ from our LTQ MS with SRM approach. AHL concentrations reported by Stabb et al. (2008) are based on a method described earlier (Gray and Greenberg 1992) where radioisotopes were used. Hence, they could detect very low levels of AHLs, but their analysis was restricted to specific AHLs (3-oxo-C6-HSL and C8-HSL). In our approach, we searched for 15 different AHLs by using their specific retention times during HPLC, their molecular mass to charge ratio and their fragmentation product (lactone ring; m/z 102Á055). There is significantly more AHLs than has been searched for earlier, for example, Kuo et al. (Kuo et al. 1994) searched for 6 and May et al. (May et al. 2012) searched for 8 different AHLs. May et al., also searched for C4-HSL, but they used LM medium (LB media with 2% salt). In a similar medium (LB25), we also found C4-HSL below LOQ in A. fischeri ES114.
Although we have detailed knowledge on a few model bacteria, the distribution of AHL-based QS systems in environmental Vibrionaceae isolates has been unclear. It has also been largely unknown whether the majority of Vibrionaceae members produce similar or very different AHL profiles. This was recently addressed by two independent studies, in which they screened for QS signals in 25 (Yang et al. 2011) and 106 (Garc ıa-Aljaro et al. 2012) Vibrionaceae strains. Using different biosensors and TLC, they found a variety of AHLs. Although these semi-quantitative methods are efficient methods for screening many samples for the presence of QS signal molecules, they are also inherently limited by (i) their abilities to discriminate between the many different AHLs (chain lengths, hydroxy/oxo modifications); (ii) their abilities to discriminate between AHLs and AHL mimics (chemically different molecules that have similar activities as AHLs in biosensor assays); and (iii) that the methods are typically less reliable when detecting AHLs at very low concentrations. In this study, we screened 57 Vibrionaceae members for AHL production using the HPLC-MS/MS and AHL standards and provided quantitative measurements of fifteen different AHLs (C4-C12-chained AHLs with unsubstituted 3′ carbons or 3′ Oxo or OH modifications; Fig. 1). Thirteen of the fifteen tested AHLs were found in the bacterial samples surveyed in this study, C12-HSL and 3-oxo-C12-HSL were never detected, and C10-HSL and 3-OH-C12-HSL were rarely found. As in previous studies, we conclude that AHLs are found widespread in Vibrio and Aliivibrio, with only a handful of isolates completely lacking AHL production (under our defined growth conditions, and when using our detection method). Furthermore, we show that closely related species typically produce similar AHL profiles, although interesting exceptions do exist.

Why multiple AHLs in bacteria?
The phylogenetic tree in Fig. 2 shows that half of the tested strains produce multiple AHLs, whereas others produce one, two or even none. Likely explanations for this AHL diversity are for example, that one AHL can bind to and regulate the activity of different receivers or that several AHLs bind to and regulate same receiver(s). Both of these cases can be seen in A. fischeri where the AinS product C8-HSL binds to both AinR and LuxR at intermediate cell densities. In addition to binding C8-HSL, LuxR also binds 3-oxo-C6-HSL at high cell densities (Lupp et al. 2003). Moreover, LuxR type of regulators can be classified into five categories based on their activation mechanisms (Stevens et al. 2011). The first type (TraR type) is activated by tight AHL binding; the second type (LuxR type) represents dimers that are activated after AHL binding; the third type (MrtR type) represents nonfunctional monomers that dimerize and are activated upon AHL binding; the fourth type (EsaR type) represents functional dimers that are inactivated upon AHL binding, and finally the fifth type (SdiA type) represents monomers that are activated only upon binding to AHL, but still remain as monomers. This illustrates some of the diversity of activation mechanisms that AHLs are involved in.

Supporting Information
Additional Supporting Information may be found in the online version of this article: Figure S1 Ion chromatograms of AHLs with retention time (RT). Figure S2 Linearity of AHL measurement over a broad concentration range.
Table S1 AHLs standards with their retention time and fragments generated by the SRM method used in this study for analysis. Table S2 Validation of precision and accuracy of Escherichia coli ArcticExpress (DE3) supernatant spiked with AHLs. Table S3 AHL concentrations in spent media from A. fischeri ES114. Measurements (average of three technical replicates with standard deviation) from two biological replicates are shown.
Table S4 Quantitative analysis of fifty-seven bacteria from Vibrionaceae family.
Dataset S1 16S rRNA gene sequences from 45 strains used to make the phylogenetic tree in Fig. 2.