Global and Phylogenetic Distribution of Quorum Sensing Signals, Acyl Homoserine Lactones, in the Family of Vibrionaceae

Bacterial quorum sensing (QS) and the corresponding signals, acyl homoserine lactones (AHLs), were first described for a luminescent Vibrio species. Since then, detailed knowledge has been gained on the functional level of QS; however, the abundance of AHLs in the family of Vibrionaceae in the environment has remained unclear. Three hundred and one Vibrionaceae strains were collected on a global research cruise and the prevalence and profile of AHL signals in this global collection were determined. AHLs were detected in 32 of the 301 strains using Agrobacterium tumefaciens and Chromobacterium violaceum reporter strains. Ethyl acetate extracts of the cultures were analysed by ultra-high performance liquid chromatography-high resolution mass spectrometry (MS) with automated tandem MS confirmation for AHLs. N-(3-hydroxy-hexanoyl) (OH-C6) and N-(3-hydroxy-decanoyl) (OH-C10) homoserine lactones were the most common AHLs found in 17 and 12 strains, respectively. Several strains produced a diversity of different AHLs, including N-heptanoyl (C7) HL. AHL-producing Vibrionaceae were found in polar, temperate and tropical waters. The AHL profiles correlated with strain phylogeny based on gene sequence homology, however not with geographical location. In conclusion, a wide range of AHL signals are produced by a number of clades in the Vibrionaceae family and these results will allow future investigations of inter- and intra-species interactions within this cosmopolitan family of marine bacteria.


Introduction
Quorum sensing (QS) is a process induced by cell population density and allows bacteria to sense and act on their local environment, as well as, communicate both within and between species [1].The essence underlying the mechanism of QS is based on the production and accumulation of signalling molecules called autoinducers.When the threshold concentration of the signals is reached, they bind to receptor proteins that then act as either transcriptional activators or repressors [2].The QS signals are small molecules such as acyl homoserine lactones (AHLs) produced by Gram-negative bacteria, autoinducer-2 (AI-2) used by Gram-negative and Gram-positive bacteria and oligopeptides which are utilized by Gram-positive bacteria [1,3].QS systems are involved in the regulation of several different bacterial phenotypes such as biofilm formation, bioluminescence, virulence and production of bioactive compounds [1,[4][5][6].
Vibrionaceae is a large family of Gram-negative marine, facultative anaerobic bacteria belonging to the Gammaproteobacteria.This family includes several genera of which the largest are Vibrio and Photobacterium that include human and fish pathogens such as Vibrio cholerae, V. anguillarum, and V. vulnificus.Also included are the algal and squid symbionts such as V. pomeroyi, V. aestuarianus, and Aliivibrio fischeri (formerly known as Vibrio fischeri) [6][7][8][9].Quorum sensing was discovered by the Hastings lab working on luminescence of A. fischeri [10,11] and subsequently the genetic basis was unravelled leading to the characterization of the two major protein components AHL synthase (LuxI) and the AHL receptor (LuxR) [12,13].
Vibrionaceae are not only important as symbiotic or pathogenic bacteria, but have more recently also been hailed as a potential source of novel bioactive secondary metabolites [14][15][16][17] such as the antibiotic molecules andrimid and holomycin [18], the antifungal compound vibrindole A [19] and the anticancer-active pelagiomicin C [20].Furthermore, Photobacterium halotolerans produces solonamides and ngercheumicins that interfere with quorum sensing and virulence in Staphylococcus aureus [21,22].The QS-interfering Photobacterium sp. and the antibiotic producing Vibrio spp.were isolated during a screening of 500 bacterial strains collected on the global marine research cruise Galathea 3 [23].Three hundred and one of the 500 strains were identified as Vibrionaceae by 16S rRNA gene analysis and are the subject of this study.
In this study, we investigated the abundance and diversity of AHLs in 301 Vibrionaceae strains from a global collection [23].The purpose of this survey was to determine how widespread AHL signalling is in environmental Vibrionaceae spanning most climate zones using biological monitors as well as LC-MS identification.The study is part of a larger work in which we aim to determine to what degree QS is involved in regulating production of bioactive secondary metabolites in Vibrionaceae.
Limit of detection (LOD, after re-extraction from growth media) for the different assays demonstrates that UHPLC-HRMS (most abundant ion ± m/z 0.01 [M + H] + or [M + Na] + with signal/noise 1:10) was the most sensitive method for most compounds except for the long chained oxo AHLs (Table 1).Compared to Purohit et al. [33], the LC-MS LODs reported here are 3-70 times higher which is to be expected when comparing a Time of Flight MS against the more sensitive triple quadrupole MS.The method herein presented can identify unexpected AHLs, enable retrospective data-analysis, and tentatively identify AHLs.Furthermore, it is more specific due to more fragment ions being detected, high resolution determination of the fragment ions, and high resolution determination of several pseudomolecular ions ([M + H] + , [M + Na] + and [M + NH4] + ).
At concentrations 2-3 above LOD (if only using one ion), qualifier ions could additionally be detected by giving increasing strength to the identification (Figure 1B-J).In Figure 1B-F,K, the default tuning was used, which results in [M + H] + being less predominant than [M + Na] + as [M + H] + are lost due to fragmentation, while [M + Na] + is more stable (and cannot be used for MS/MS experiments).In Figure 1G-J,L, the data from small molecule tuning is shown, providing much more [M + H] + , which can be MS/MS fragmented into a very specific fingerprint (Figure 1N) that is identical with the library spectrum seen (Figure 1M).Thus, it provides an additional identification method for the AHLs via their tandem HRMS spectra [36].These MS/MS spectra are less noisy than full scan spectra, as all ions except the parent ion (±m/z 0.5) are filtered away; therefore, all fragment ions are generated from the parent ion.In this case, by using a QTOF MS, the fragment ions are additionally measured with high mass accuracy (MS/HRMS).Due to the losses during the MS/MS fragmentation and that the [M + H] + and/or [M + NH4] + needs to be auto-selected for MS/MS, (this analysis is not as sensitive as the full scan shown in Figure 1B-J).Nonetheless, the MS/HRMS sensitivity was in almost all cases sensitive enough for the detection of AHLs.
More importantly, the MS/HRMS spectra of the reference standard AHLs showed the known fragment ions m/z 102.05495 (homoserine ring) and C3H5N, m/z 56.04948 (further loss of CH2O2 from the homoserine ring) as well as m/z 120.06552 derived from the open homoserine ring moiety.Thus, data files could be mined for open and closed form AHLs by extracted ion chromatograms of m/z 102.05495, 56.04948, 120.06552 (±m/z 0.01) from all MS/HRMS precursors.This is similar to parent ion scanning on triple quadruple instruments, and can identify unexpected AHLs as long as they contain the homoserine ring moiety.).Samples were analysed with a time difference of two months and thus retention was altered, but constant in the sequence and identical (±0.02 min) to reference standard.

AHL-Producing Vibrionaceae Strains
Classification of the bacterial isolates as Vibrionaceae was confirmed by 16S rRNA gene sequencing (GenBank acc.no. in Table 2).Additionally, all strains grew well on TCBS-agar confirming their Vibrio-specific metabolism.Of the 301 strains, 32 (corresponding to 11%) induced a response in either A. tumefaciens or C. violaceum when Vibrio biomass from marine agar was spotted directly onto reporter plates (Table 3).Nine strains induced both monitors, while 15 induced only the A. tumefaciens and eight only the C. violaceum.Using five different reporter strains in a double-layer microplate high-throughput assay, 85% of Vibrionaceae isolates gave a positive response in a study by Garcí a-Aljaro et al. [32].Without using a biomonitor prescreen, Purohit et al. [33] analysed 57 Vibrionaceae isolates for AHLs by HPLC-MS/MS and did not recover AHLs from nine of the strains, also demonstrating a hit rate of about 85% [33].Our comparably low number of AHL-positive strains could be correlated with the lower number of biomonitor strains, our pre-screening method or the strain diversity.However, an increased number of biomonitor strains has not yielded an improved detection of AHLs in Vibrionaceae previously [32].The media used in the bioassays have low salt concentrations, which could have caused stress to the spotted Vibrionaceae biomass and thus could lead to an underestimate of the total number of AHL-positive strains.Hence, the herein AHL-negative strains may still be AHL-producers in the environment or under different culturing conditions including variation in nutrients, temperature, salinity or pH.Also, they may require specific biological cues from other organisms to trigger the production of QS signals.These variations and optimizations might be included in future studies to gain more knowledge on the physiological background of AHL production in Vibrionaceae.The strain collections used in the two studies described above [32,33] were selected strains, namely fish-derived strains.Our strain collection represents a broader range of environmental strains and this difference in strain profile could also be the cause of the differences in AHL-positive strains.
For extract preparation, the 32 AHL-positive strains were grown in LB10.Nine strains (S0188, S0203, S0209, S0273, S0843, S1073, S2606, S4497 and S4738) did not grow in LB10 and required higher salinity and thus, they were grown in LB20 (Table 3).All ethyl acetate extracts were re-tested for induction of the monitors A. tumefaciens and C. violaceum in a well assay.Extracts of 15 strains demonstrated a different response compared to the initial biomass screen, either by gaining or loosing activity for one of the biomonitors.This could be attributed to the change of bacterial growth condition (plate versus liquid medium) or the final concentration of AHLs in the extracts, which may or may not have crossed the limit of detection [34,35].In V. anguillarum 90-11-287, a positive reaction for C6-AHL was observed in the biomonitor strain as evaluated with biomass, while LC-MS did not detect C6-AHL [26].False positives might potentially occur when a strain is spotted too close to another strain on a plate.In this second screen, 10 extracts induced both monitors, while 18 induced only the A. tumefaciens and interestingly, none of the extracts induced C. violaceum alone.The higher hit rate for A. tumefaciens has been described before and is probably due to the wider range of response to various AHLs of this strain [32] (see Section 2.1).
Three extracts (of strains S0344, S0845, S1194) were negative in the bioassays.However, in all three extracts, AHLs were detected by LC-MS, namely either N-butanoyl (C4), N-(3-hydroxy-butanoyl) (OH-C4) or N-(3-hydroxy-hexanoyl) (OH-C6) HL (Table 3). A. tumefaciens is not sensitive to these short chain AHLs, however C. violaceum does react at least to C4 homoserine lactone (see Section 2.1).Comparing the chemical analyses of all extracts in standard and small molecule tune, the sizes of the peak areas indicate that the concentration of AHL in these three samples might not have been sufficient for detection by the biomonitor.

Structural Abundance of AHLs in Vibrio Strains
UHPLC-DAD-QTOFMS was performed on extracts of the 32 strains and control strain V. anguillarum 90-11-287 to identify the AHL molecules produced.V. anguillarum 90-11-287 produced three AHLs, N-(3-hydroxy-hexanoyl) (OH-C6), N-(3-hydroxy-decanoyl) (OH-C10) and N-(3-oxo-decanoyl) (O-C10) homoserine lactone, agreeing with previous studies [26,33].Across all strains, a total number of 21 different AHLs was detected (Table 3) of the 33 closed ring forms searched for.A few open ring forms were also detected, however, only in very small amounts, since open ring forms are generally closed by the extraction procedure.The maximum number of nine different AHLs was found in V. splendidus S0188.It is known that several Vibrionaceae strains, e.g., V. anguillarum, produce a number of AHLs, however, production of nine AHLs appears to be unique for a single strain.
It is important to mention that we have used a very sensitive chemical detection procedure, thus this type of diversity might be found in other organisms, but goes currently undetected.The high number of AHLs might result from biochemical processes inside the cells and might not be genetically encoded or biologically relevant in environmental conditions.However, they could potentially be part of differentiated signaling at the species level.The genome of this strain will be subjected to sequence analysis in the future.Eighteen strains produced three AHLs or more, and in five strains only one type of AHL was identified.In two strains (S0787, 1073), no AHLs could be identified from the full scan MS data, nor from the MS/MS data (e.g., m/z 102 fragment), although for both, monitor response was observed.This might indicate that novel unknown AHLs (with modifications on the homoserine ring that would not create the m/z 102 fragment) might be produced by these strains or that other types of molecules, such as diketopiperazines, are being produced [37].This will be studied in more detail in the future including accurate concentration measurements of the detected AHLs.
The most abundant AHL was OH-C6 homoserine lactone, which was identified in 17 extracts, followed by OH-C10, OH-C4 and O-C10 homoserine lactone, which were identified in 12, 11 and 10 extracts, respectively.Rare AHLs were C12, OH-C9, OH-C11 and O-C4.The Vibrionaceae extracts studied by Purohit et al. [33] were dominated by O-C6, OH-C10 and OH-C4, demonstrating a certain overlap.As indicated above, variations might be due to the different species composition in our strain collection.For the strains belonging to the Splendidus cluster, they detected C4, OH-C4, OH-C6, OH-C8 and O-C6, while we measured OH-C4, OH-C6, OH-C7, OH-C8, OH-C10 and OH-C12, a pattern conserved through four different isolates (Figure 2).Variations could be due to the differences in growth conditions or chemical analysis applied; also for example, Purohit et al. [33] did not scan for odd-numbered AHLs.Generally, AHLs with odd numbers of carbons are rare in nature [38], however, they are known to be produced by other gammaproteobacteria such as Pseudomonas aeruginosa [39], Yersinia ruckeri [40] or Edwardsiella tarda [41].Furthermore, Aliivibrio fischeri has been to shown to be capable of detecting odd-numbered AHLs such as C5 [42].To the authors' knowledge, this is the first study demonstrating the production of odd-numbered AHLs by strains of the Vibrionaceae family.

Figure 2.
Cluster analysis of AHL diversity per strain and phylogenetic affiliation using CIMminer [43].Dark grey: AHL present; light grey: AHL absent; colours distinguish different clades, clusters marked in bold.

MS/HRMS Screening of AHLs
The upper base peak chromatogram (BPC, m/z 50-1700) (Figure 3A) of the ethyl acetate extract from strain S1162 demonstrates the complexity of the sample, with B-D showing the extracted ion chromatograms of the major AHL fragment ions (Figure 3); clearly indicating the presence of two unassigned peaks at 4.7 and 3.36 min.The first peak was also present in sample blanks, while the second peak corresponded to an open lactone AHL.The peaks at m/z 120 and 137 correspond to fragmentation between the homoserine and alkyl chains with the charge residing on one or the other side.With the fragment at m/z 137 containing a C=O group (where the charge is believed to reside), the accurate mass left exactly C6H5O2 requiring four unsaturations.This is most likely possible using an aromatic ring containing two presumed phenolic groups, although a larger ring with ketones cannot be excluded either.Nonetheless, this compound seems to be novel AHL, a finding that needs to be confirmed by preparative purification and NMR techniques or total synthesis.Bioinformatic analysis of the PKS genes in the AHL gene cluster may also aid in this tentative identification of the molecule and genome sequencing is in progress.

Phylogeny and Geographical Distribution of AHL-Producing Strains
The 32 AHL-producing strains were identified using MLSA of the 16S rRNA, recA, toxR and rpoA genes in combination with a novel technique using the fur gene for species differentiation in the family of Vibrionaceae [44].16S rRNA gene sequences do not allow sufficient discrimination in the Vibrionaceae family due to the presence of multiple alleles and MLSA requires the amplification of several genes.Thus, Machado et al. [44] demonstrated that the fur gene, encoding a ferric uptake regulator, would have the power to replace the previously used techniques as efficient and accurate phylogenetic marker in this family.The neighbor joining tree of concatenated partial rpoA and recA gene sequences   4).Additionally, we analysed one Photobacterium strain, P. angustum S1192.Within a clade, species tend to form certain phylogenetic pairs that are very difficult to distinguish, such as V. fluvialis and V. furnissii, V. brasiliensis and V. tubiashii and V. splendidus and V. lentus [46,47].Those pairs could not be properly resolved by our MLSA analysis; however, the fur gene analysis was a very robust discriminator in these situations.For instance, S2719, S2757, S3857 and S4497 could not be identified using MLSA of the four marker genes, however, the fur analysis revealed their association with the Orientalis and Harveyi clades (Figure 4).The 32 strains did not cluster taxonomically according to their isolation sites (Supplementary Figure S1); e.g., the seven Anguillarum strains were isolated from opposite sides of the globe: the west coast of Africa and the west coast of Australia.In addition, there is no tendency for strains to group together according to their climate zone; moreover, all Harveyi and all Splendidus clade strains were isolated from tropical waters.Finally, no correlation between isolation site and AHL profile, climate zone and abundance or diversity of AHLs was observed.
When comparing phylogeny to the AHL profiles of the strains (Figure 2), two distinct clusters formed: as with the control strain V. anguillarum 90-11-287, most other Anguillarum clade isolates produced OH-C6, OH-C10 and O-C10 homoserine lactone, building a fingerprint for the Anguillarum clade.Four out of the six strains of the Splendidus clade produce OH-C4, OH-C6, OH-C7, OH-C8, OH-C10 and OH-C12.These results, where not every strain of the clade fell in one fingerprint, resemble the taxonomic pattern.This could mean that a fingerprint possibly represents a distinct subgroup of a clade rather than a complete clade ("subclade").The Anguillarum and Splendidus strains were taxonomically very similar (Figure 4).
However, the Anguillarum clade strains were isolated from four different isolation sites from both temperate and tropical climates.The Splendidus clade strains were isolated from the same climate zone, but different locations.This demonstrates the differences between the isolates making them to unique strains besides taxonomic similarity.Combining the individual AHLs produced by the individual strains of a clade as identified by previous studies build similar profile to the herein described complete AHL profiles for the Anguillarum and Splendidus clade [26,33].
Previous studies on AHLs in Vibrionaceae have either not addressed the chemistry of the signals or lacked a high throughput setup [9,32,33].All demonstrated a limited phylogenetic analysis using only the 16S rRNA gene.Because they possess several alleles of the 16S rRNA gene, Vibrionaceae require a more stringent phylogenetic analysis [46,47].This, however, might be greatly facilitated by utilization of the fur gene as marker in the future [44].Our findings will add a crucial amount of information to the current knowledge on AHLs in the family of Vibrionaceae and hereby, we propose an AHL fingerprint correlated to specific phylogenetic subclades.

Bacterial Strains and Growth Conditions
Three hundred and one Vibrionaceae strains were isolated on the global Galathea 3 expedition [23] on the basis of their ability to inhibit Vibrio anguillarum 90-11-287 [49].Vibrionaceae strains were grown on Marine Agar plates (MA; Difco 212185) for 1-3 days at 25 °C for the initial AHL screening based on biomonitors.For chemical analyses of AHLs, the Vibrionaceae strains (Table 2) were grown in Luria-Bertani broth (LB; Difco 244620) under aerated conditions (200 rpm) at 20 °C for one day assuming stationary phase, which has been demonstrated for a selected number of strains.Strains that did not grow in LB were grown in LB with extra NaCl (2% final concentration).Agrobacterium tumefaciens NT1 (pZLR4) [34] and Chromobacterium violaceum CV026 [35] were used as reporter strains in bioassays to detect AHLs. A. tumefaciens was grown on ABt agar or in LB broth with 0.5% NaCl (LB5) with 20 µg mL −1 gentamycin and 50 mg mL −1 X-Gal.C. violaceum was grown on LB5 agar or in LB with 20 µg mL −1 kanamycin.V. anguillarum 90-11-287 was used as reference strain in the UHPLC-DAD-QTOFMS analysis [26].All strains were stored at −80 °C in a freeze medium (30.0 g tryptone soy broth (TSB; Oxoid CM129B), 5.0 g glucose, 20.0 g skim milk powder, 40.0 g glycerol, 1000 mL H2O) [50].

Detection of AHL Compounds
Reporter plates were prepared as described previously [53,54].The Vibrionaceae strains were grown on MA and a loop of bacterial biomass was placed on bioassay plates embedded with reporter strain.3 µL 1 µM OHHL (OH-C6) was spotted on the A. tumefaciens plates as a control and 3 µL 1 mM BHL (C4) was spotted on the C. violaceum plates.After 1 day at 25 °C, C. violaceum plates were inspected for purple zones due to AHL-induced violacein production.After 2 days at 25 °C, A. tumefaciens plates were inspected for formation of blue colour due to AHL-induced β-galactosidase activity.Extracts were prepared for all strains being positive in the pre-screen.To evaluate the extracts for AHL activity, 6 mm wells were punched in the solid bioassay plates and 30 µL extract were pipetted into the wells.Plates were read as described above.
To confirm the AHL detection range of each biomonitor, 16 AHL standards (Supplementary Table S1) were tested against C. violaceum and A. tumefaciens in a well assay, using three tenfold diluted concentrations in acetonitrile.Plates were read as described above.

Extracts for Bioassays and UHPLC-DAD-QTOFMS
Ten mL of liquid LB10 or LB20 culture was mixed with 10 mL ethyl acetate (EtOAc) containing 0.5% formic acid (FA), and incubated at 20 °C and 200 rpm for 30 min.The upper EtOAc phase was collected, and evaporated to dryness under nitrogen.Samples were resuspended in 0.5 mL EtOAc with 0.5% formic acid and stored at −20 °C until use.30 µL of extracts (i.e., corresponding to 1.6 mL of original culture) were re-tested for AHL activity as described above.

Figure 3 .
Figure 3.Chemical analysis of strain S1162.(A) base peak chromatogram; (B) m/z 56.04948 ± 0.02; (C) m/z 102.05495 ± 0.02; and (D) m/z 120.06552 ± 0.02, showing the extracted ion chromatograms of the three major diagnostic fragment ions with labels of six known long chain AHLs.At 3.36 min a novel open chain AHL is eluting with its MS/HRSM spectrum at 10 eV showed in E with a tentatively identified compound.

Figure 4 .
Figure 4. Neighbor joining tree of concatenated partial rpoA and recA gene sequences of 32 AHL-positive Vibrionaceae strains using the Jukes-Cantor method and Shewanella oneidensis MR-1 as outgroup.Bootstrap values are based on 100 replicates.Square brackets indicate clades [45,48].

Table 1 .
Limit of detection (nmol L −1 ) of a subset of acyl homoserine lactones (AHLs) using the three different assays.ND = not detected.

Table 2 .
32 AHL-producing Vibrionaceae strains and their taxonomic identification by multilocus sequences analysis (MLSA) of the 16S rRNA, recA, toxR and rpoA gene sequences and geographic data of isolation site.
a MLSA was uncertain; b sequenced contig BLASTed against NCBI, not the fur gene database; ND = not defined.

Table 3 .
AHLs in 32 Vibrionaceae strains tested against C. violaceum (Cv) and A. tumefaciens (At) using biomass or acidified EtOAc extracts and AHLs detected by UHPLC-DAD-QTOFMS; numbers demonstrate the peak area of the AHL in the chromatogram of the first run; numbers in brackets demonstrate the peak area of the AHL in the chromatogram of the second run; the total no. of occurrences/AHL does not include the reference strain V. anguillarum 90-11-287.