The Siderophore Piscibactin Is a Relevant Virulence Factor for Vibrio anguillarum Favored at Low Temperatures

Vibrio anguillarum causes vibriosis, a hemorrhagic septicaemia that affects many cultured marine fish species worldwide. Two catechol siderophores, vanchrobactin and anguibactin, were previously identified in this bacterium. While vanchrobactin is a chromosomally encoded system widespread in all pathogenic and environmental strains, anguibactin is a plasmid-encoded system restricted to serotype O1 strains. In this work, we have characterized, from a serotype O2 strain producing vanchrobactin, a novel genomic island containing a cluster of genes that would encode the synthesis of piscibactin, a siderophore firstly described in the fish pathogen Photobacterium damselae subsp. piscicida. The chemical characterization of this siderophore confirmed that some strains of V. anguillarum produce piscibactin. An in silico analysis of the available genomes showed that this genomic island is present in many of the highly pathogenic V. anguillarum strains lacking the anguibactin system. The construction of single and double biosynthetic mutants for vanchrobactin and piscibactin allowed us to study the contribution of each siderophore to iron uptake, cell fitness, and virulence. Although both siderophores are simultaneously produced, piscibactin constitute a key virulence factor to infect fish, while vanchrobactin seems to have a secondary role in virulence. In addition, a transcriptional analysis of the gene cluster encoding piscibactin in V. anguillarum showed that synthesis of this siderophore is favored at low temperatures, being the transcriptional activity of the biosynthetic genes three-times higher at 18°C than at 25°C. We also show that iron levels and temperature contribute to balance the synthesis of both siderophores.


INTRODUCTION
Vibrio anguillarum is a bacterium inhabitant of marine environments and also one of the most important pathogens in the aquaculture industry worldwide. It is the causative agent of vibriosis, a fatal hemorrhagic septicaemia affecting warm-and cold-water fish of highly economic importance (Toranzo et al., 2005). V. anguillarum isolates are classified into 23 different O-serogroups (O1 to O23) according to the European serotyping system (Pedersen et al., 1999), although most virulent strains belong to serotypes O1, O2 and, to a lesser extent, O3. The remaining V. anguillarum serotypes are mostly environmental strains isolated from seawater, marine animals or sediments (Frans et al., 2011;Toranzo et al., 2017).
It is well known that bacterial virulence is a multifactorial trait and that pathogenic bacteria must have a number of factors that enable them to colonize the host and establish an infection (Miller et al., 1989). The virulence-related factors identified in V. anguillarum so far include those related to chemotaxis and motility (O'Toole et al., 1996;Ormonde et al., 2000), adhesion (Wang et al., 1998), invasion (Croxatto et al., 2007), secretion of extracellular enzymes (Denkin and Nelson, 2004;Rodkhum et al., 2005;Li et al., 2008) and several iron uptake mechanisms (Li and Ma, 2017). Despite this knowledge, the pathogenesis of V. anguillarum is not completely understood (Hickey and Lee, 2017).
Iron restriction is an important host defense strategy, thus successful pathogens must possess mechanisms to acquire iron from host sources in order to cause disease. In V. anguillarum two catechol siderophores, vanchrobactin and anguibactin, have been described (Actis et al., 1986;Soengas et al., 2006;Lemos et al., 2010;Li and Ma, 2017). While anguibactin is encoded by pJM1-type plasmids and is restricted to some virulent strains of serotype O1, vanchrobactin is a chromosomally encoded siderophore that is widespread in all V. anguillarum isolates either environmental or pathogenic (Lemos et al., 2010;Li and Ma, 2017). It is noticeable that vanchrobactin is not produced by serotype O1 strains harboring a pJM1-type plasmid since the presence of this plasmid results in the inactivation of vanchrobactin synthesis by a transposition event from the plasmid to the vanchrobactin chromosomal gene cluster (Naka et al., 2008). Thus, to date the simultaneous production of more than one siderophore was not reported in V. anguillarum (Naka et al., 2008;Rønneseth et al., 2017). However, it was detected the simultaneous presence of anguibactin and vanchrobactin in a marine Vibrio sp. (Sandy et al., 2010). From previous reports, it is clear that anguibactin is a key virulence factor for V. anguillarum serotype O1 strains (Wolf and Crosa, 1986;Naka et al., 2011). However, the role of vanchrobactin in virulence remains unknown so far.
The in silico analysis of the V. anguillarum serotype O2 strain RV22 genome sequence (accession number AEZB00000000) showed the existence of a gene cluster whose closest homolog is the irp cluster encoding the siderophore piscibactin (Naka et al., 2011). This siderophore was firstly identified in the fish pathogen Photobacterium damselae subsp. piscicida (Souto et al., 2012) and it is one of the main virulence factors of this bacterium (Osorio et al., 2015). Piscibactin synthesis and transport are encoded by a pathogenicity island (PAI) located in the 69-kb plasmid pPHDP70 (Osorio et al., , 2015. The presence of irp genes in V. anguillarum RV22 suggests that this strain, besides producing vanchrobactin, likely produces also piscibactin or a piscibactin-like siderophore, which would imply the existence of a third siderophore in V. anguillarum.
In this work, we describe a novel genomic island present in many virulent strains of V. anguillarum that encodes the piscibactin system. We demonstrate that some serotype O2 strains produce in fact piscibactin and that this siderophore plays a relevant role in the cell fitness and contribute to a greater extent than vanchrobactin to the virulence for fish. We also show that the synthesis of both siderophores is balanced in response to iron levels and to the growth temperature.

Construction of irp1 Defective Mutants by Allelic Exchange and Complementation
In-frame (non-polar) deletions of irp1 were constructed by allelic exchange in V. anguillarum RV22 wild type and vabF (vanchrobactin synthesis deficient) backgrounds using the suicide vector pNidKan as previously described . The oligonucleotides used are shown in Table 2. The mutagenesis process led to the generation of V. anguillarum single mutant strains RV22 irp1 (MB280) and the double mutant RV22 vabF irp1 (MB283). To ensure that all mutations were in frame, the deleted region was sequenced by the Sanger method using primers irp1_ang_1_XbaI and irp1_ang_4_EcoRI ( Table 2). pPHDP70 plasmid (from Pdp DI21 and containing all necessary genes for piscibactin synthesis) marked with a Km cassette (pPHDP70::kan) (Osorio et al., 2015) was mobilized into the V. anguillarum RV22 vabF irp1 double mutant to restore piscibactin synthesis.

Growth Promotion and Siderophore Production Assays
Growth promotion assays were performed using 96-well microtiter plates. Each well contained 200 µL of CM9 medium (Lemos et al., 1988) supplemented with 10 µM FeCl 3 to achieve iron excess conditions, or with the iron chelators EDDA [ethylenediamine-di(o-hydroxyphenyl-acetic acid)] 5 µM or 100 µM 2,2 -dipyridyl, to achieve iron restricted conditions. Each well was inoculated with a 1:50 dilution of a V. anguillarum overnight culture in TSB-1 adjusted to an OD 600 = 0.5. The plates were incubated at 25 • C or 18 • C with shaking at 150 rpm. Growth (OD 600 ) was recorded during 24 h in an iMACK Microplate reader (Bio-Rad). Bacterial cultures in CM9 with 50 µM 2,2 -dipyridyl and an OD 600 ≈ 0.8 (after 6 h of incubation) were used to obtain supernatants and measure siderophore production using the chrome azurol-S (CAS) liquid assay (Schwyn and Neilands, 1987). For this purpose, equal volumes of cell free supernatants and CAS reagent were mixed and, after 15 min of incubation at room temperature, A 630 was measured in a spectrophotometer (Hitachi).

Cross-Feeding Assays
To test whether V. anguillarum was able to use piscibactin as iron source, a cross-feeding assay was conducted using RV22 (wild type strain) and RV22 vabD mutant as indicator strains. These strains were inoculated into CM9 plates as follows: 0.5 mL of a TSB-1 culture at an OD 600 = 0.5 were included into 20 mL of CM9 medium containing 0.8% agarose and 120 µM of the iron chelator 2,2 -dipyridyl, a concentration close to the MIC and at which growth halos are easily visualized . V. anguillarum strains to be tested for piscibactin production and P. damselae subsp. piscicida DI21 (wild type strain, piscibactin producer) or a cured strain (DI21 lacking pPHDP70, unable to produce piscibactin and used as negative control) were cultured in TSA-1 plates and the cells were harvested with a sterile loop and placed on the surface of the plates inoculated with the V. anguillarum strains. The presence of growth halos of the V. anguillarum indicator strains around cells of P. damselae subsp. piscicida DI21 or V. anguillarum after overnight incubation at 25 • C was indicative of piscibactin utilization.

Piscibactin Detection by Mass Spectrometry
Siderophore piscibactin detection was carried out in cell free supernatants from cultures of V. anguillarum grown at 25 • C in flasks containing 1 L of CM9 minimal medium with 40 µM 2,2 -dipyridyl with continuous shaking at 150 rpm for ca. 24 h (Souto et al., 2012). Flasks were inoculated with 20 mL of a fresh culture of V. anguillarum RV22wt, RV22 vabF or RV22 vabD mutants in TSB-1. When bacterial cultures achieved an OD 600 = 1.0, 10 mg of GaBr 3 were added to each 1.0-L batch. After 12 h of incubation at 4 • C, bacterial cells were removed by centrifugation at 10,000 × g for 10 min (Beckman J-21 High Speed Centrifuge) and supernatants were filtered through a 0.45-µm pore size membrane by a continuous filtration cartridge (Millipore). Subsequently, ca. 900 mL of each cell-free supernatant were concentrated under vacuum to 410 mL. Fifty milliliters of the solution were passed through an Oasis hydrophilic lipophilic (HLB) cartridge (Waters) (35 cm 3 , 6 g) (Espada et al., 2011), which was previously conditioned and equilibrated with 60 mL of acetonitrile (solvent B) and 60 mL of water (solvent A). They were eluted with 30 mL of the following mixtures of water and acetonitrile: 1:0, 7:3, 1:1, 3:7, and 0:1. The fraction eluted with a mixture of water and acetonitrile (1:1), named RV22WTF3, RV22 vabFF3, and RV22 vabDF3, were subjected to high-performance liquid chromatography (HPLC)-high-resolution electrospray ionization mass spectrometry (HRESIMS) analysis using an Atlantis dC18 column (100 mm × 4.6 mm, 5 µm) with a 35 min gradient from 10 to 100% CH 3 CN-H 2 O, then 5 min at 100% CH 3 CN, and finally a 10 min gradient from 100 to 10% H 2 O-CH 3 CN, at a flow rate of 1 mL min −1 . Piscibactin-Ga(III) complex was detected with a t R of 9.56 min from RV22WTF3 and with a t R of 9.71 min from RV22 vabFF3 by comparison of the t R , HRESIMS and UV spectral data to those reported for that compound isolated from P. damselae subsp. piscicida DI21 (Souto et al., 2012). Vanchrobactin was detected with a t R of 2.74 min from RV22WTF3.

RNA Purification and RT-PCR
The organization of piscibactin genes into operon(s) was tested by reverse transcription PCR. V. anguillarum RV22 was grown until exponential phase (ca. OD 600 = 1) in 10 mL CM9 medium containing 5 µM EDDA or 10 µM Fe 2 (SO 4 ) 3 . Cells were pelleted by centrifugation at 10,000 × g for 10 min and total RNA was isolated with RNAwiz (Ambion) following the manufacturer's recommendations. RT-PCR was performed with 1 µg RNA pre-treated with RQ1 RNase-Free DNase (Promega) by using the M-MLV reverse transcriptase (Invitrogen). An appropriate primer ( Table 2), located at the 3 -end of irp5 gene, was used to obtain a cDNA that was then used as template for three PCR reactions targeted into araC1 (PCR1), frpA (PCR2) and between irp2 3 -end and irp1 5 -end (PCR3) (Figure 1). All primers used are listed in Table 2. A negative control reaction was performed with total RNA without M-MLV reverse transcriptase to confirm the lack of genomic DNA contamination in each reaction mixture. The PCR positive control reaction was done using 100 ng of genomic DNA as template.
lacZ Transcriptional Fusions and β-Galactosidase Assays DNA fragments corresponding to V. anguillarum araC1, araC2 and frpA promoter regions were amplified by PCR. The PCR-amplified promoter regions were fragments of about 700 bp, including the region upstream of the start codon and the first nucleotides of the araC1, araC2 or frpA coding sequences (ca. 50 bp). These putative promoter regions were fused to a promoterless lacZ gene in the low-copy-number  Table 2), located at the 3 -end of irp5 gene, was used to obtain a cDNA (denoted as RT in the figure) that was then used as template for three PCR reactions targeted into araC1 (PCR1), frpA (PCR2) and between irp2 3 -end and irp1 5 -end (PCR3). M, size marker from 100 to 1,000 bp; +, positive control PCR using genomic DNA as template; −, negative control PCR using total RNA without reverse transcriptase; RT-PCR1-3, results of the three different RT-PCRs.

In Silico Analysis of irp ang Genomic Island Structure and Analysis of Its Distribution
The NCBI services were used to consult DNA and protein sequence databases using BLAST algorithms. All V. anguillarum genome sequences (43 genomes) deposited in GenBank were screened for the presence of the irp ang genomic island. For this purpose, the whole nucleotide sequence of irp ang genomic island of V. anguillarum RV22 and the 3 and 5 -ends surrounding regions were used as query in a BLASTN search. Nucleotide sequences producing significant alignments were downloaded and a multiple alignment of the conserved genomic sequences was obtained by using MAUVE (Darling et al., 2010). The alignment was used to compare the genomic island structure and to identify the insertion point. The Virtual Footprint Promoter Matches from the online database PRODORIC Release 8.9 1 (Münch et al., 2003) was used to identify putative Fur-box sequences. BLASTP algorithm was also used to analyze homology of each protein encoded by the irp ang gene cluster from V. anguillarum with its counterpart in P. damselae subsp. piscicida. All this information was used to create Figure 1 and Supplementary Figure S1.

Fish Virulence Assays
Virulence assays were carried out with turbot (Psetta maxima) fingerlings with an average weight of 5 g. Fish were divided in groups of 30 animals. All fish groups, one per strain tested and a control group, were maintained in 50-L seawater tanks at 18 • C with continuous aeration and water recirculation. The bacterial inoculum was prepared as follows. Several colonies from a 24 h TSA-1 culture were suspended into saline solution (0.85% NaCl) to get a cell suspension with an OD 600 = 0.5. The inoculum used was a 10-fold dilution of this suspension in saline solution. The actual number of injected bacterial cells was determined by plate count on TSA-1. Fish were inoculated intraperitoneally (ip) with 100 µL of bacterial suspensions, ranging the doses used between 2 and 4 × 10 4 CFU per fish. A control group was inoculated with 100 µL of saline solution. Mortalities were recorded daily during 10 days after injection and statistical significance of differences in percentage survival for V. anguillarum strains was determined using the Kaplan-Meier method with Mantel-Cox log-rank test using SPSS (version 20; IBM SPSS Inc., Chicago, IL, United States). P-values were considered significant when P was <0.05, <0.01, and <0.001. All protocols for animal experimentation used in this study have been reviewed and approved by the Animal Ethics Committee of the University of Santiago de Compostela.

V. anguillarum Strain RV22 Harbors a Novel Genomic Island Likely Encoding Piscibactin Synthesis
In a previous work, we described and characterized the gene cluster involved in the synthesis and transport of vanchrobactin system in V. anguillarum RV22 ( Figure 1A) . Inactivation of vanchrobactin synthesis resulted in a drastic decrease in siderophore production by this strain and in its growth ability under iron starvation conditions. However, when the complete genome sequence of RV22 was described (Naka et al., 2011), it was found that chromosome II contains a likely additional siderophore gene cluster with high similarity to the irp genes encoding piscibactin in P. damselae subsp. piscicida (Souto et al., 2012). This irp gene cluster in V. anguillarum RV22 strain (that we named irp ang ) is part of a DNA region of ca. 40 kb located in chromosome II (between locus AEH34691 and AEH34692) ( Figure 1B). This cluster shows identical gene structure and organization as the irp cluster described in P. damselae subsp. piscicida as part of pPHDP70 plasmid ( Figure 1B) (Souto et al., 2012;Osorio et al., 2015). Each CDS present in the irp ang cluster shows an amino acid similarity between 52 and 66% with the irp cluster ortholog from P. damselae subsp. piscicida (Supplementary Table S1). Although irp ang genes include most of the functions for piscibactin synthesis and utilization (Supplementary Table S1), the cluster lacks an entD homolog, which encodes a 4 -phosphopantetheinyl transferase, that is required to activate the peptide synthesis domains of non-ribosomal peptide synthetases (NRPS) (Crosa and Walsh, 2002). However, the vab gene cluster (Figure 1A), encoding the vanchrobactin system, does include a 4 -phosphopantetheinyl transferase gene (vabD) (Balado et al., 2008) that could complement in trans this function (see below).
In addition, there are three orfs (loci: VAR_RS19255, VAR_RS0101310, and VAR_RS19260), adjacent to irp ang , which encode a probable transposon belonging to the Tn7 superfamily ( Figure 1B). The nucleotide sequence (accession No. AEZB01000007.1; region 10925-14484) including these three orfs shows a 86% nucleotide identity to the Tn7 transposon genes of the V. parahaemolyticus pathogenicity island (Vp-PAI) that encodes a type three secretion system (T3SSα) (Okada et al., 2009). Tn7 transposons are widespread in bacteria and are involved in the formation of some genomic islands that have the common feature of not being inserted near of tRNAs genes (Sugiyama et al., 2008). These findings suggest that irp ang could be part of a novel genomic island. Interestingly, a BLASTN search revealed that four of highly virulent V. anguillarum strains (two belonging to serotype O1 and two from serotype O2) (Rønneseth et al., 2017) contain the irp ang genes inserted in the same chromosomal location (Supplementary Figure S1). However, in two of these strains the irp ang genes are not associated with transposon genes (Supplementary Figure S1). The genome sequences of many V. anguillarum strains were recently reported (Busschaert et al., 2015;Castillo et al., 2017;Holm et al., 2018). Thus, we performed an in silico search within the 43 genomes available in GenBank and we could find the presence of irp ang gene cluster in several of these strains, some of which are highly virulent (Table 3). This fact suggests that the piscibactin cluster could have a relevant role in virulence.
To test if irp ang genes are actually expressed, a series of reverse-transcriptase PCR reactions were conducted. The results showed that this gene cluster is transcribed in a polycistronic mRNA comprising araC1, araC2, frpA, irp1-5, irp8 and irp9 genes (Figures 1B,C). Therefore, the 10 genes putatively encoding piscibactin biosynthetic, regulatory, and uptake functions could be co-transcribed from the promoter upstream of araC1. A similar result was found for the piscibactin gene cluster harbored by P. damselae subsp. piscicida plasmid pPHDP70, which is transcribed in a single polycistronic mRNA from the promoter upstream of dahP . Interestingly, the irp ang cluster does not include a dahP homolog and many other differences were found at the nucleotide level in intergenic regions. Thereby, even though irp ang genes can be transcribed from a promoter upstream of araC1, the existence of additional promoters, particularly upstream of frpA, cannot be ruled out (see below).

V. anguillarum RV22 Produces Vanchrobactin and Piscibactin Simultaneously
To demonstrate that RV22 synthesizes piscibactin in addition to vanchrobactin (Soengas et al., 2006(Soengas et al., , 2007, we used the previously described methodology, based on the use of HLB cartridges and liquid chromatography-mass spectrometry (LC-MS), that we developed for isolation of piscibactin (Souto et al., 2012). Vanchrobactin was first detected with a retention time (t R ) + denotes the presence of a complete siderophore gene cluster. (+) denotes the insertion of an IS element into an essential biosynthetic gene. Gray shadows indicate the active siderophore systems in each strain.
of 2.74 min (Figures 2A,B), which showed the [M+H] + and [M+Na] + ions at m/z 398.1667 and 420.1487, respectively ( Figure 2C). Furthermore, from the cell-free supernatants of RV22 cultures under iron restriction and after addition of GaBr 3 , in order to stabilize the siderophore as its corresponding complex, it was detected the characteristic piscibactin-Ga(III) complex with (t R ) of 9.56 min ( Figure 2D). The chemical identification was done on the basis of its t R value, UV spectral data and HRESIMS analysis. Thus, piscibactin-Ga(III) complexes obtained from V. anguillarum RV22 and from P. damselae subsp. piscicida (Souto et al., 2012) showed similar retention times and absorbance maxima (223, 267, and 354 nm) in its UV spectrum ( Figure 2F)   ( Figure 2E), indicating the same molecular mass to that reported for piscibactin-Ga (III) complex isolated from P. damselae subsp. piscicida (Souto et al., 2012). As expected, chemical analysis of cell-free supernatants of the V. anguillarum RV22 vabF mutant (unable to produce vanchrobactin), following the same methodology, showed the absence of vanchrobactin (Figures 3A-C) and the presence of a peak with a t R of 9.71 min (Figure 3D), displaying a m/z value of 519.9945/521.9933 ( Figure 3E) and a UV spectrum ( Figure 3F) which correspond to piscibactin-Ga(III) complex. The same methodology was used to test piscibactin synthesis by V. anguillarum RV22 vabD mutant (Balado et al., 2008), which lacks an active 4 -phosphopantetheinyl transferase. In this mutant, neither vanchrobactin or piscibactin were detected (Supplementary Figure S2), confirming the vabD gene is essential for the synthesis of both siderophores. In addition, cross-feeding assays were used to ascertain piscibactin production and utilization by V. anguillarum RV22 (Figure 4). RV22 wild type and RV22 vabD mutant could be cross-fed by a piscibactin-producing strain (P. damselae subsp. piscicida DI21). The V. anguillarum mutant RV22 vabF (impaired for vanchrobactin production) still could cross-fed both indicator strains, indicating the production of a second siderophore. However, no growth halo was obtained when using RV22 vabD mutant as tested strain (Figure 4). All these results prove that V. anguillarum RV22 is able to produce and use piscibactin, which denotes that irp ang genes are functional. Moreover, the results clearly suggest that irp ang gene cluster needs a functional vabD gene (encoding the 4 -phosphopantetheinyl transferase) to complete the biosynthesis of piscibactin.

Piscibactin Synthesis Contributes to the Growth Ability Under Iron Restriction Depending of Environmental Conditions
To test whether irp1 is required for piscibactin production and to study the influence of piscibactin on V. anguillarum cell fitness, in-frame deletion mutants of irp1 were obtained in RV22wt and RV22 vabF (deficient in vanchrobactin synthesis) backgrounds. The growth ability of resultant RV22 irp1 and RV22 vabF irp1 mutant strains was measured at 25 • C under iron excess and under iron-deprivation conditions using the strong iron-chelating agent ethylenediamine-di-(o-hydroxyphenyl-acetic acid) (EDDA), or a weaker one, 2,2 -dipyridyl. RV22 wild type strain (that produces both vanchrobactin and piscibactin) was used as positive growth control showing a minimal inhibitory concentration (MIC) of EDDA and 2,2 -dipyridyl of 10 µM and 150 µM, respectively. Conversely, RV22 vabD strain (Balado et al., 2008), impaired to produce any of these siderophores, showed MICs of 5 µM for EDDA and 100 µM for 2,2 -dipyridyl. When the mutants were cultured under iron-excess conditions (CM9 plus 10 µM ferric chloride) no significant differences in growth levels were observed with respect to the wild type strain ( Figure 5A). However, significant differences were observed between RV22wt and some of the siderophore mutants when cultivated under iron deficiency. When EDDA (CM9 plus 5 µM EDDA) was used as chelating agent, the RV22 irp1 mutant (producing only vanchrobactin) and the parental strain showed identical growth levels. However, the mutant producing only piscibactin (RV22 vabF) was unable to grow ( Figure 5A). These results corroborate those previously published  and confirm that inactivation of vanchrobactin synthesis strongly reduce the growth ability of V. anguillarum under in vitro iron starvation. Interestingly, when growth was measured in CM9 containing 100 µM of the chelator 2,2 -dipyridyl the mutant RV22 vabF reached around 50% of the growth levels showed by the parental strain ( Figure 5A). Siderophore-like activity present in cell free supernatants, measured by the CAS assay after growing each V. anguillarum strain at 50 µM 2,2 -dipyridyl, correlated with the growth levels observed.
FIGURE 4 | Cross feeding assay to determine the production and use of piscibactin by V. anguillarum RV22 wild type and RV22 vabD mutant. Indicator strains are those inoculated within the CM9 plates containing 120 µM 2,2 -dipyridyl. Tested strains are those tested by placing a loopful of bacterial biomass on the agar surface. A growth halo indicates that indicator strains can use the siderophores produced by tested strains to overcome the iron limitation conditions. Photobacterium damselae subsp. piscicida DI21 wild type strain (irp + , piscibactin producer) and DI21 cured from plasmid pPHDP70 (irp − , unable to produce piscibactin) were used as controls. RV22 vabF does not produce vanchrobactin; RV22 vabD does not produce vanchrobactin nor piscibactin. Siderophore production (indicated as A 630 values), determined by the CAS test performed on culture supernatants, is also displayed for all strains. In each panel, bars labeled with the same letter (a, b, c, d) are not significantly different from each other (t-test). Error bars denote standard deviations.
The same growth experiments were conducted also at 18 • C since it is a temperature closer to the conditions encountered in vivo during infection. At this temperature of incubation, and in presence of 100 µM 2,2 -dipyridyl, RV22 vabF (producing only piscibactin) behaved similarly to the wild type strain (Figure 5B), which suggests that piscibactin production by V. anguillarum could be favored at low temperatures (see below). Reactivation of piscibactin synthesis in RV22 vabF irp1 double mutant by introduction of pPHDP70 plasmid (Osorio et al., 2015), resulting in strain RV22 vabF irp1(pPHDP70), showed a phenotype almost identical to RV22 vabF strain in both temperatures assayed (Figures 5A,B).
Overall, these results suggest that (i) V. anguillarum may use vanchrobactin as primary siderophore and that (ii) piscibactin efficacy to provide iron to support growth depends on the environmental conditions, in particular on the iron source and the temperature. Additionally, the results also suggest that the role of vanchrobactin in the biology of V. anguillarum could be previously overestimated  probably owing to the fact that vanchrobactin defective mutants have been studied only under stringent conditions of iron limitation by the use of EDDA as iron chelator.

Piscibactin Contributes to a Greater Extent Than Vanchrobactin to Virulence for Turbot
In order to study the contribution of each siderophore (vanchrobactin and piscibactin) to V. anguillarum pathogenesis, the virulence of the RV22 wild type strain, as well as of the mutants deficient in one of the two siderophores or in both of them, was evaluated by experimental infections in turbot. The fish were kept at 18-20 • C and were inoculated intraperitoneally with 2-4 × 10 4 CFU of the wild type or with one of the single or double siderophore mutants. The wild type strain RV22 killed 80% of the fish after 5 days of challenge (Figure 6). When the challenge was done with the mutant that produces only vanchrobactin (RV22 irp1) the mortality reached 56% after 6 days of infection. However, when the fish were inoculated with the mutant that produces only piscibactin (RV22 vabF) the mortality reached 88% after 5 days, at the same level of the wild type strain. The double mutant RV22 vabF irp1 and the RV22 vabD (both unable to produce any of the two siderophores) give mortalities of 12% and 8%, respectively, after 7 days. Interestingly, when piscibactin production was reactivated by the introduction of pPHDP70 plasmid (that contains the piscibactin synthesis gene cluster) (Osorio et al., 2015) into RV22 vabF irp1 double mutant, this complemented strain caused a mortality of 78% after 6 days (Figure 6). These results clearly demonstrate that the ability to synthesize a siderophore has a great impact in the virulence of V. anguillarum. However, each siderophore system contributes in different grade to virulence being piscibactin production, more than vanchrobactin, the main virulence factor in the tested conditions.

Transcriptional Analysis of the irp ang Gene Cluster: The Expression of Piscibactin Genes Are Favored at Low Temperatures
The synthesis of siderophores are tightly regulated at transcriptional level which ensures that siderophore production occurs only when it is necessary (Cornelis et al., 2011). In order to analyze the expression levels of the irp ang putative FIGURE 6 | Percentage survival after 7 days of turbot fingerlings challenged with 2-4 × 10 4 CFU/fish of V. anguillarum RV22 wild type strain or with the siderophore mutants analyzed. Asterisks denote statistically significant differences between strains: * P < 0.05; * * P < 0.01; * * * P < 0.001; ns, no statistically significant differences. promoters, DNA fragments of ca. 700 nucleotides upstream frpA ang , araC1 ang and araC2 ang genes, that were named PfrpA ParaC1 and ParaC2, respectively (Figure 1B), were cloned into the plasmid pHRP309 upstream of a promoterless lacZ gene. frpA encodes the presumptive ferri-piscibactin outer membrane receptor while araC1 and araC2 encode two putative AraC-type transcriptional regulators (Supplementary Table S1). Resulting plasmids were mobilized into V. anguillarum RV22wt and the transcription levels were measured by determining β-galactosidase activities under different iron availability conditions at 25 • C (Figure 7A). The expression levels were compared with the promoter of the vanchrobactin system located upstream of vabH (Balado et al., 2008). Interestingly, the use of the PfrpA, ParaC1, and ParaC2 presumptive promoters produced significant β-galactosidase activity (Figure 7) which demonstrates that the three sequences serve as transcriptional starts. In addition, under the iron-restricted conditions tested, the expression from promoter PfrpA was three-times higher than from ParaC1 or ParaC2 ( Figure 7A). Thus, although the transcription of piscibactin genes could start from any of the three promoters, it seems to occur to a greater extent from the promoter immediately upstream of frpA (PfrpA), which encodes the probable piscibactin receptor.
Usually, expression of siderophore systems promoters are down-regulated under iron availability (Cornelis et al., 2011). As expected, the addition of an iron chelator to the medium produced a significant increase of PfrpA, ParaC1, and ParaC2 activities ( Figure 7A). However, the maximal expression levels achieved by piscibactin promoters under high iron deprivation (by addition of 2,2 -dipyridyl 100 µM) at 25 • C were fourfold lower with respect to the vanchrobactin promoter ( Figure 7A). These findings strongly contrast with the high relevance of piscibactin in virulence, as demonstrated above (Figure 6). This low expression at 25 • C could be due to: (1) irp ang genes could remain down-regulated because enough iron is entering the cell through the vanchrobactin system; or (2) piscibactin promoters need additional environmental signals, not only low iron levels, to achieve their maximal activity. These two hypothesis need to be further studied.
Since, as shown above, temperature could have a relevant role in piscibactin production, we wanted to ascertain if the expression levels of piscibactin genes could be favored at low temperatures. For this purpose, transcriptional levels of the piscibactin promoters were measured in RV22wt strain incubated at 18 • C and compared to those observed at 25 • C. While ParaC2 showed almost the same activity at both temperatures, we found an increment of ca. twofold in the PfrpA and ParaC1 activity at 18 • C with respect to that measured at 25 • C ( Figure 7A). In contrast, the activity of the vanchrobactin promoter decreased twofold. When constitutive proC-promoter was tested, no significant differences in lacZ activity were observed between conditions assayed either temperature or iron availability (Figures 7A,B). These findings show that expression of piscibactin genes are favored when V. anguillarum grows at low temperatures. In addition, the increase of piscibactin promoters expression, together with the decrease of vanchrobactin promoters at low temperatures, strongly FIGURE 7 | Transcriptional activity (β-galactosidase units) of the lacZ fusions to the P1, P2 and vabH promoters in V. anguillarum RV22wt (A) or RV22 vabD mutant impaired to produce siderophores (B) grown at 25 • C or 18 • C in CM9 medium under different iron availability conditions: iron excess and CM9 containing 50 µM 2,2 -dipyridyl. lacZ fusion with the constitutive promoter of proC (PproC) was used as control. Asterisks indicate statistically significant differences between bars (P < 0.05); ns, no statistically significant differences. Error bars denote standard deviations.
suggests that V. anguillarum is able to balance the synthesis of each siderophore responding to environmental signals like iron levels and temperature (Figure 7). To ascertain if the decrease in vancrobactin promoter activity at 18 • C was due to the temperature or if it was caused by the intake of iron mediated by piscibactin, the expression levels of the three promoters were measured in the RV22 vabD strain, which is unable to produce any siderophore. While PfrpA and ParaC1 piscibactin promoters showed a strong temperature-dependent expression pattern again, the vanchrobactin promoter (PvabH) showed almost identical β-galactosidase activity at 25 or 18 • C ( Figure 7B). These results suggest that the lower activity of vanchrobactin promoter at 18 • C in the RV22wt strain may be due to the increased intake of iron through the piscibactin system more than to a temperature-dependent regulation. Overall, these results strongly suggest that the expression of irp ang genes in V. anguillarum are favored at low growth temperatures.

DISCUSSION
In the present work, we could identify in V. anguillarum a siderophore gene cluster, named irp ang , that is contained within a pathogenicity island and that shows a strong homology to its counterpart described in the pPHDP70 plasmid from P. damselae subsp. piscicida (Osorio et al., 2015). This PAI is located in the second chromosome of some highly pathogenic strains of V. anguillarum, including strain RV22, from which vanchrobactin genes were previously identified (Balado et al., , 2008. The irp ang cluster shows the same genetic organization than the irp genes of P. damselae subsp. piscicida and both confer the ability to produce the siderophore piscibactin. The PAI island harboring the piscibactin genes in P. damselae subsp. piscicida encodes all functions required to produce and use piscibactin (Osorio et al., 2015). However, piscibactin synthesis in V. anguillarum depends on the presence in the recipient genome of a functional 4 -phosphopantetheinyl transferase gene (entD). A similar case was reported for Aeromonas salmonicida, where acinetobactin and amonabactin biosynthetic pathways share the same set of genes encoding the biosynthesis of 2,3-DHBA, a common moiety of both siderophores, and also share an EntD homolog .
Chemical analysis demonstrated that piscibactin is indeed being synthesized by the vanchrobactin-producing V. anguillarum strains. Although the production of more than one siderophore by a bacterial cell could be deleterious (Cordero et al., 2012), in some cases the ability to produce two siderophores can enhance niche flexibility (Sandy et al., 2010;Dumas et al., 2013) and pathogenesis (Garcia et al., 2011). Some pathogens carry multiple iron acquisition systems that seem redundant in laboratory culture conditions. For instance, the fish pathogen Aeromonas salmonicida subsp. salmonicida produces acinetobactin and amonabactin, two catechol siderophores and, although there are strains that produce both siderophores simultaneously, most strains have the amonabactin cluster inactivated . By contrast, the co-expression of siderophores with different chemical properties can play specialized roles at the host-pathogen interface (Fetherston et al., 2010;Koh and Henderson, 2015). Piscibactin, along with yersiniabactin and pyochelin, constitute a group of siderophores that share similar structural characteristics (Crosa and Walsh, 2002;Souto et al., 2012). It has been shown that the co-expression of yersiniabactin with other catecholate siderophores enhances the virulence properties of uropathogenic Escherichia coli (Garcia et al., 2011). Yersiniabactin and pyochelin seem to play several roles, other than iron uptake, during infection since they can efficiently bind other metals besides iron. For instance, yersiniabactin participates in zinc (Zn +2 ) acquisition and also confers protection against copper toxicity, which promotes bacterial colonization, dissemination and resistance against phagocytosis (Fetherston et al., 2010;Chaturvedi et al., 2012;Bobrov et al., 2014;Koh and Henderson, 2015;Holden et al., 2016). Pyochelin has been also associated with some traits such as oxidative or inflammatory activities in Pseudomonas aeruginosa infections (Britigan et al., 1997), decomposition of organotin compounds (like TPT, triphenyltin chloride or DPT, diphenyltin dichloride) (Sun et al., 2006) or plants defense against fungal plagues (Audenaert et al., 2002). Some of these or other related roles could also be attributed to piscibactin in V. anguillarum.
Interestingly, the inactivation of piscibactin synthesis in V. anguillarum RV22 produced just a small decrease in the growth ability under iron restriction. This suggests that these mutants would use vanchrobactin to internalize iron. Thus, vanchrobactin would serve in vitro as an efficient iron carrier (Iglesias et al., 2011) for V. anguillarum while the piscibactin efficacy to supply iron to the cell would depend on which is the iron source available (Figure 5). However, this observation highly contrasts with our results of the virulence assays where piscibactin production alone was sufficient to obtain the maximal virulence. The inactivation of piscibactin system resulted in a severe loss of the virulence degree even if vanchrobactin system was functional (Figure 6). Hence, although the ability to synthesize one siderophore, either vanchrobactin or piscibactin, has a great impact in V. anguillarum virulence, the ability to produce piscibactin seems to be sufficient to confer maximal virulence to V. anguillarum, at least in the tested conditions.
A total of 43 V. anguillarum genomes are currently available in the genomic databases and the virulence properties of 32 of these strains were recently reported (Busschaert et al., 2015;Rønneseth et al., 2017). An in silico analysis showed that many of these virulent V. anguillarum strains harbor both, vanchrobactin and piscibactin, gene clusters (Table 3). Thus, piscibactin genes are widespread in V. anguillarum strains and they are associated with a highly virulent phenotype for several fish species, such as turbot (Psetta maxima), cod (Gadus morhua), halibut (Hippoglossus hippoglossus), rainbow trout (Oncorhynchus mykiss) or sea bass (Dicentrarchus labrax) (Busschaert et al., 2015;Rønneseth et al., 2017). These results support the previously suggested hypothesis that vanchrobactin is the ancient siderophore of V. anguillarum (Lemos et al., 2010). However, according to our results and to previously published works (Sandy et al., 2010), vanchrobactin production could be more related to the persistence of V. anguillarum in the marine environment than to pathogenesis. Hence, piscibactin would be the most important siderophore during the vibriosis outbreaks in fish. Interestingly, it was recently found that piscibactin is one of the most extended siderophore systems in Vibrionaceae family. Phylogenetic analysis of three siderophore systems (piscibactin, vibrioferrin, and aerobactin) suggested that their current distribution could be explained by an old insertion that was followed by the action of diverse evolutionary forces (Thode et al., 2018). An interesting observation is that the piscibactin gene cluster is never present in anguibactin-producing strains ( Table 3). It has been previously reported that the presence of pJM1-like plasmids encoding the anguibactin system inactivates the vanchrobactin cluster through a transposition event (Naka et al., 2008). A similar genetic incompatibility could explain the mutual exclusion between piscibactin and anguibactin gene clusters.
Many of the genes encoding virulence factors, including iron uptake systems, are preferentially expressed within the host in response to particular environmental signals (Lee and Camilli, 2000). The analysis of promoters activity showed that irp ang genes in V. anguillarum are not only strongly iron-regulated, but also, that piscibactin production has a requirement for low temperatures. These findings suggest that V. anguillarum is able to modulate the synthesis of vanchrobactin and piscibactin according to the surrounding temperature, which adjust the energy costs of siderophore production to the environmental circumstances. Some bacterial species that are able to produce more than one siderophore can modulate the expression as a response to environmental factors, e.g., pH, carbon source, amino sugars or temperature (Valdebenito et al., 2006;Craig et al., 2012;Cornelis and Dingemans, 2013;Dumas et al., 2013;Tanabe et al., 2014). The temperature-dependent regulation is characteristic of pathogenic bacteria affecting ectothermic hosts such as fish, where the infection processes generally occur at temperatures lower than those required for bacterial optimal growth (Guijarro et al., 2015). However, these regulation mechanisms are yet poorly understood and few examples were reported of siderophore synthesis regulated by temperature. One of these is ruckerbactin, a catecholate siderophore produced by Yersinia ruckeri (Fernández et al., 2004); other one is the hydroxamate bisucaberin produced by Aliivibrio salmonicida, the causative agent of cold-water vibriosis (Colquhoun and Sørum, 2001;Winkelmann et al., 2002). In both bacteria, the maximal siderophore production take place around the temperatures at which the diseases outbreaks occur. Our results also suggest that piscibactin synthesis in V. anguillarum is higher at temperatures (16-19 • C) at which the vibriosis outbreaks usually occur in turbot (Toranzo et al., 2005(Toranzo et al., , 2017.

CONCLUSION
Pathogenic V. anguillarum strains lacking the anguibactin system produce two chromosomally encoded siderophores, vanchrobactin and piscibactin. While vanchrobactin synthesis seems to be regulated only by the iron availability, piscibactin is preferentially produced at low temperatures at which fish vibriosis outbreaks usually occur in natural conditions. Piscibactin seems to contribute to a greater extent than vanchrobactin to the virulence for turbot. Further studies are currently under way to decipher the regulation pathways that control piscibactin synthesis in V. anguillarum.