Biochemical and molecular characterization of three serologically different Vibrio harveyi strains isolated from farmed Dicentrarchus labrax from the Adriatic Sea

Vibrio harveyi is recognized as one of the major causes of vibriosis, a disease that threatens the long-term sustainability of aquaculture. Current research shows that the Mediterranean strains of V. harveyi are serologically heterogeneous, though research comparing the traits of different strains is scarce. This study aims to describe the biochemical, physiological and genetic characteristics of three serologically different strains of V. harveyi isolated from farmed European Sea bass (Dicentrarchus labrax) from the Adriatic Sea. A total of 32 morphological and biochemical markers were examined and, the susceptibility to 13 antimicrobials tested, and then compared the results of high-throughput sequencing and in silico analyses. This study also presents the first whole genome sequences of V. harveyi isolated from European sea bass. A large number of nonsynonymous variations were detected among sequences of the three strains. The prediction analysis of resistance genes did not correspond with the in vitro antimicrobial susceptibility tests. Six virulence genes previously unrelated to virulence of vibrios were detected in all three studied strains. The results show that differences were detected at every level of comparison among the three studied strains isolated from the same fish species originating from a small geographic area.


Results
Biochemical and serological characteristics of V. harveyi. Strains 72/16, 150/16 and 94/17 were confirmed as V. harveyi by using Nucleotide BLAST against NCBI Nucleotide database with the obtained 16S rRNA and toxR sequences. Strains 72/16 and 150/16 were part of the intestine microbiome collected from farmed ESB during winter months, while strain 94/17 was isolated from ESB with typical symptoms during August. Serological assays showed that strain 94/17 belongs to serotype A and strain 150/16 belongs to the new serotype recently reported in Croatian isolates (tentatively, serotype B), while strain 72/16 is serologically different from those belonging to the known serotypes. The biochemical and physiological characteristics of the studied strains compared with reference characteristics published by Austin & Austin 39 and Pretto 40 are presented in Table 1. Of the 32 tested characteristics, all three strains were identical for 25 and differed in seven traits. All three strains are gram-negative, oxidase and catalase positive, capable of fermenting saccharose, motile and grow as yellow colonies on TCBS agar. They are all positive for lysine decarboxylase and can utilise citrate, but are negative for β-galactosidase, arginine dihydrolase, urease, acetoin and production of hydrogen sulphide. Furthermore, all three strains can ferment glucose, mannose, sucrose and amygdalin while none can ferment inositol, rhamnose, melibiose or arabinose. Regarding the salt tolerance, the strains can all grow in peptone water with up to 6% added salt but are unable to grow in peptone water with 10% salt. Regarding their differences, strains 94/17 and 150/16 are α-haemolytic while strain 72/16 is β-haemolytic. Likewise, strains 94/17 and 150/16 are positive for ornithine decarboxylase and tryptophanase and can ferment sorbitol, while strain 72/16 does not possess those traits. For tryptophan deaminase, strains 94/17 and 72/16 are positive while strain 150/16 is negative. The reverse is true for gelatinase, with only strain 150/16 positive. Finally, strain 94/17 is the only one able to reduce nitrates to nitrites same as reference strains 39,40 . Compared to previously published characteristic, most of the properties are the same excluding the fact that none of Croatian studied strains fermented arabinose.
Susceptibility of V, harveyi to different antimicrobials. The results of the antimicrobial susceptibility test are presented in Table 2 as the measured diameters of inhibition zones obtained by the disc diffusion method since there is no harmonised interpretive criteria for V. harveyi 41 Obtained results showed that none of the tested strains inhibited growth of ampicillin, O129 10 μg and Novobiocin where only strain 72/16 showed slight inhibition. Lower inhibition was obtained for Gentamicin for three strains while there are smaller zones of inhibition for all substances of the strain 72/16 compared to other two tested strains.
Sequencing results. Sequencing 43 we identified the same 374 genes linked to antimicrobial resistance (AMR) in each of the reference mapped consensus genomes of the studied strains. Of those, there were zero perfect hits, four strict hits and 370 loose hits. The four strict hits were the same in every strain, and are presented in Table 4. All hits are shown in Supplementary Table S1.
Using the VFanalyzer, we identified 155 genes associated with virulence. They can be divided, according to function, into 55 genes related to mobility, 48 genes related to secretion, 19 genes related to adherence, 16 genes related to antiphagocytosis, 11 genes related to iron uptake, two genes related to quorum sensing, two genes related to toxins, one gene related to cell surface components and one gene related to fimbrial adherence. The entire result of the analysis is presented in Supplementary Table S2. Table 5 singles out the genes found in this analysis, but not in all three studied strains, and also shows the genes that were not associated with virulence of  www.nature.com/scientificreports/ other Vibrio species in the VFDB. In total, seven virulence genes are identified only in one or two of the studied strains, and there are eight genes previously not linked with virulence in other Vibrio species.

Discussion
V. harveyi is one of the major species responsible for vibriosis, a disease globally causing severe economic losses in aquaculture. Still, the exact mechanism and interplay of different factors underlying this disease are largely unclear, and different strains of V. harveyi show large phenotypic differences in regards to their virulence 18,28,30 . This study describes the first approach to comparing the biochemical, physiological and genetic characteristics of serologically different V. harveyi strains originating from the Mediterranean Sea. The comparison includes two strains belonging to identified V. harveyi serotypes (A and tentatively "B") and a strain serologically different from those.
Comparing the biochemical characteristics of the studied strains obtained by the API 20E test, we noted that all three strains are similar with 15 of 21 reactions giving the same result. Still, certain unexpected differences were obtained, with strain 72/16 negative for ornithine decarboxylase and tryptophanase, strains 72/16 and 150/16 negative for the reduction of nitrates and strains 94/17 and 72/16 negative for gelatinase. Positive results were expected for these traits, as reported by Austin and Austin 39 . With two more reactions showing dissimilar results, tryptophan deaminase and sorbitol tests, there is clearly some ambiguity between V. harveyi strains, and therefore we advise caution in solely relying on this analysis for identification. Further, strain 72/16, which is serologically different from previously identified serotypes, deviates even further from expectations and is more diverse from the strains of the known serotypes than they are in relation to each other.
To date, the zone diameter clinical breakpoints of the disc diffusion test are published for only one aquatic bacteria, A. salmonicida 41 , and have yet to be defined based on standardised testing protocol and harmonised interpretive criteria in susceptibility studies for V. harveyi 44 . For this reason, the results of antimicrobial susceptibility are presented as measured diameters of inhibition zones without interpretation as to whether the bacteria are sensitive or resistant to a given antibiotic. Still, it appears that the studied bacterial strains are inhibited by all tested antimicrobials except ampicillin, novobiocin and the 10 μg disc of vibriostat O129. Comparing the results of the rest of the antimicrobials, strain 72/16 shows smaller inhibition zones than strains 94/17 and 150/16 in each test, as seen by the average size of inhibition zones for each strain (strain 94/17: 43.7 ± 9.6; strain 150/16: 41.9 ± 8.3; and strain 72/16: 33.2 ± 6.3) across all tested antimicrobials demonstrating the inhibition of growth. The strains belonging to serotypes A and "B" show highly similar zones of inhibition.
In terms of the consensus genome sequences, it appears that some genomic data are missing in all three strains. Compared to the reference genome, which comes from one of the most studied strains (ATCC 33,843) 45 , the consensus sequences of all three of our studied strains are shorter by 0.3 Mb, with over 400 fewer gene annotations. This is most easily observed using the whole genome alignment (Fig. 1), indicating that a part of the reference genome is not present in the studied strains. Meanwhile, the number of unmapped reads is relatively high for all three strains (strain 94/17: 19.34%, strain 72/16: 16.22%; strain 150/16: 19.83%). This indicates that a large amount of sequenced DNA found in the strains from the Adriatic Sea is not present in the reference strain. While the existence of large plasmids in V. harveyi has been demonstrated 46,47 , considering the percentage of unmapped reads it is unlikely that they all belong to plasmid DNA. Comparing the whole genome alignment of the reference mapped consensus sequences of the studied strains, several areas can be observed with a higher concentration of variation, and there are some smaller differences in size between the collinear blocks. There is a large number of nonsynonymous variations between the three studied strains and it would be interesting to explore whether the same mutations are found in other strains belonging to the same serotypes, linking exact variations to functional diversity.
The RGI analysis using strict criteria predicted resistance to four drug classes in all three strains. All four genes can be found in CARD associated to antimicrobial resistance in different Vibrio species, but to our knowledge, only tet35 has previously been associated with V. harveyi 48 . Comparing the results of the RGI analysis with the disk diffusion test of antimicrobial sensitivity, several observations can be made. The only in silico predicted resistance that manifested in vitro is for ampicillin (penam). While the RGI analysis predicted resistance to tetracycline antibiotics and fluoroquinolone antibiotics, this was not confirmed with disc diffusion. Quite the opposite, the inhibition zones were from 31 to 40 mm for oxytetracycline disks, and 34 to 53 mm for enrofloxacin disks. Finally, the in vitro analysis showed that the strains were resistant to novobiocin, an aminocoumarin antibiotic, and while the RGI analysis identified 28 genes linked to aminocoumarin antibiotic resistance, all were loose hits. Considering these results, we can conclude that the results of the RGI analysis should be considered with caution and they clearly cannot replace the in vitro assay. Unfortunately, analysing RGI in CARD resistance genes associated to the reference strain of V. harveyi ATTCC 33,843 showed single adeF gene what additionaly emphasize the need of more comprehensive research with inclusion of more isolates of this important species of the genus Vibrio.
Currently, there are 242 genes in the VFDB associated with virulence of at least one of the following Vibrio species: V. cholerae, V. fischeri, V. parahaemolyticus and V. vulnificus. Of those 242 genes, 142 are found in all three of the studied V. harveyi strains, while five other genes were found in at least one of the studied strains. Additionally, eight virulence genes were detected that were previously unrelated to the virulence of vibrios, and six of those eight genes were detected in all three of our studied strains. Five genes were associated with only one strain of V. harveyi (strain 94/17: wecA; strain 72/16: pilA; and strain 150/16: mshB, wbfY and stbA). Future research should examine whether these genes are always found in specific serotypes.
In conclusion, we found differences at every level of comparison among the three studied V. harveyi strains isolated from the same fish species originating from a small geographic area. We should keep in mind that studied strains were isolated in different environmental condition what could influence on the expression of the virulence www.nature.com/scientificreports/ and resistance genes. However, neither biochemical nor genetic studies are not supporting this hypothesis. The best way to expand on this research would be to isolate and serotype more strains, possibly from other parts of the Mediterranean Basin, to see which characteristics are shared by certain serotypes. Furthermore, using long range high throughput sequencing in combination with Illumina sequencing would allow for the assembly of high quality genomes with high contiguity without the use of reference mapping, allowing for comparisons of most of the data, and leading to detection of more structural differences in the genome. The future directions of this research are certainly promising, as there are interesting questions still left to explore.

Methods
Examined strains. The bacteria compared in this study were isolated from ESB originating from the Eastern Adriatic Sea. Two strains, designated 72/16 and 150/16, were isolated during a study of the gut microbiome of the European sea bass during winter months 49  Positive reactions were submitted for sequencing to Macrogen Europe (Amsterdam, Netherlands). Sequence alignments and BLAST were performed using Geneious Prime 2019.2. Once the genus was identified as Vibrio, we performed a second PCR using the same DNA samples, to test whether the bacteria belong to the species V. harveyi. For this purpose, we specifically amplified the toxR gene using primers toxRF1 (5'-GAA GCA GCA CTC ACC GAT -3') and toxRR1 (5'-GGT GAA GAC TCA TCA GCA -3') 53 . Conditions used for the second PCR were the same as those used in the first PCR, except the annealing temperature was 55 °C, and the duration of the elongation step in each cycle was 1 min. Electrophoresis was performed as above, and all positive samples were submitted to Macrogen Europe for sequencing.

Serology.
Rabbit polyclonal antisera against formalin-killed cells of strains belonging to dominant serotypes were prepared as previously described 35 . Then, slide agglutination with whole cell suspensions containing 10 8 colony forming units (cfu)/mL and O-antigens in PBS were performed according to the procedure of Fouz and Amaro 36 .
Determination of biochemical characteristics. Several tests were performed to compare the biochemical and physiological characteristics of the studied strains. Strains were grown on MA to examine colony morphology and in the selective medium TCBS agar to test their ability to ferment saccharose. Gram-stained smears of bacterial colonies were examined under a reverse microscope to differentiate Gram-positive and Gram-negative bacteria. Motility was determined using an API M Medium (BioMerieux, France), while the glucose oxidation-fermentation test was performed using an OF Medium (BioMerieux, France), both according to the manufacturer's instructions. To test for the presence of indophenol oxidase, bacterial colonies were rubbed onto filter paper treated with Oxidase Reagent Droppers (Becton, Dickinson and Company, USA), and the colour change was observed. The production of catalase was tested by observing the production of oxygen bubbles after the addition of 3% solution of hydrogen peroxide to the bacterial colony on the glass slide. To determine the type of haemolysis, bacterial colonies were grown on blood agar supplemented with 1.5% NaCl for 24 h at 23 ± 2 °C. Using the API 20E commercial test kit (BioMerieux, France) according to the manufacturer's instructions, strains were examined for 21 biochemical properties: the presence of β-galactosidase, arginine dihydrolase, lysine decarboxylase, ornithine decarboxylase, urease, tryptophan deaminase, tryptophanase and gelatinase, utilization of citrate, production of hydrogen sulphide, fermentation of glucose, mannose, inositol, sorbitol, rhamnose, sucrose, melibiose, amygdalin and arabinose, reduction of nitrates to nitrites and Voges-Proskauer test (detection of acetoin). Salt tolerance of the strains was determined in peptone water with the addition of 0.5%, 3%, 6% and 10% NaCl. The identification was performed based on the reference characteristics described by Austin & Austin 39 and Pretto 40 .
Antimicrobial susceptibility testing. Antimicrobial susceptibility of the studied strains was determined by disc diffusion method 50  www.nature.com/scientificreports/ UK), oxolinic acid 2 µg (Biolab, Hungary), oxytetracycline 30 µg (BD, USA), sulfamethoxazole 300 µg (Biolab, Hungary) and trimethoprim/sulfadiazine 1.25/23.75 µg (BD, USA). Three criteria were used in selection of antimicrobial substances 44 ; (i) agents used in therapy of aquatic vibriosis namely amoxicillin, oxolinic acid, flumequine, enrofloxacin, oxytetraciycline, florfenicol and trimethoprim/sulfametoxazole; (ii)agents used in therapy of humans infected with Vibrio sp. (quinolones, the third generation cephalosporins and the tetracyclines) 54 ; (iii) miscelaneous criteria which includes chloramphenicol, prohibited substance that could indicate illegal use, Ceftazidime as an effective prescreen for isolates containing extended-spectrum β-lactamase-/ AmpC β-lactamase resistance mechanisms as major concern for human pathogens, meropenem is included to screen for isolates producing carbapenems 55 and sulfamethoxazole was included to facilitate the detection of isolates possessing sul genes reported as occurring with a high frequency in some aquatic environments.
Whole genome sequencing. Library preparation for high throughput sequencing was performed using NexteraXT Library Prep Kit (Illumina, USA) with two modifications to the manufacturer's protocol: for input DNA we used two nanograms instead of one, and PCR elongation time was increased to one minute. DNA quantification and library preparation were carried out on a Hamilton Microlab STAR automated liquid handling system. Pooled libraries were quantified using the Kapa Biosystems Library Quantification Kit for Illumina on a Roche light cycler 96 qPCR machine. Libraries were sequenced on an Illumina MiSeq System using a 250 bp paired end protocol.
Sequence analysis. Reads obtained by sequencing were adapter trimmed using Trimmomatic v0.30 56 with a sliding window quality cut-off of Q15. After trimming, reads were mapped to the genome sequence of the V. harveyi strain ATCC 33,843 42 in Geneious Prime 2020.1.2 using Geneious mapper. For mapping, we used medium-low sensitivity setting and fine tuning with 10 iterations. Annotations were transferred from the reference genome to consensus sequences using the following settings: 90% similarity cut-off, cost matrix 65% similarity (5.0/− 4.0) and only transfer best match. After mapping, the variants were called and annotated using the Find Variations/SNPs analysis in Geneious Prime 2020.1.2 with the following settings: minimum coverage 10, minimum variant frequency 0.5, maximum variant P-value 10 -6 , minimum strand-bias P-value 10 -5 when exceeding 65% bias, find only nonsynonymous variants, analyse the effect of variants on translations using bacterial genetic code, merge adjacent variations and use separate annotations for each variant at a position.
To align the consensus genomes we used the progressive Mauve algorithm with default settings 57 . Consensus genome sequences were exported in a FASTA format file which was used as an input file for the VFanalyzer in the VFDB 34 and for RGI analysis in CARD 39 .