Phaeobacter and Ruegeria Species of the Roseobacter Clade Colonize Separate Niches in a Danish Turbot (Scophthalmus maximus)-Rearing Farm and Antagonize Vibrio anguillarum under Different Growth Conditions

Abstract

Members of the Roseobacter clade colonize a Spanish turbot larval unit, and one isolate (Phaeobacter strain 27-4) is capable of disease suppression in in vivo challenge trials. Here, we demonstrate that roseobacters with antagonistic activity against Vibrio anguillarum also colonize a Danish turbot larval farm that relies on a very different water source (the Danish fiord Limfjorden as opposed to the Galician Atlantic Ocean). Phylogenetic analyses based on 16S rRNA and gyrase B gene sequences revealed that different species colonized different niches in the larval unit. Phaeobacter inhibens- and Phaeobacter gallaeciensis-like strains were primarily found in the production sites, whereas strains identified as Ruegeria mobilis or Ruegeria pelagia were found only in the algal cultures. Phaeobacter spp. were more inhibitory against the general microbiota from the Danish turbot larval unit than were the Ruegeria spp. Phaeobacter spp. produced tropodithietic acid (TDA) and brown pigment and antagonized V. anguillarum when grown under shaking (200 rpm) and stagnant (0 rpm) conditions, whereas Ruegeria spp. behaved similarly to Phaeobacter strain 27-4 and expressed these three phenotypes only during stagnant growth. Both genera attached to an inert surface and grew in multicellular rosettes after stagnant growth, whereas shaking conditions led to single cells with low attachment capacity. Bacteria from the Roseobacter clade appear to be universal colonizers of marine larval rearing units, and since the Danish Phaeobacter spp. displayed antibacterial activity under a broader range of growth conditions than did Phaeobacter strain 27-4, these organisms may hold greater promise as fish probiotic organisms.

Bacteria from the Roseobacter clade are widely distributed in marine environments (37, 51), and this group includes 38 different genera (7). The organisms constitute 20 to 30% of prokaryotes in 16S rRNA gene libraries from surface water (11). Even though the clade is detected often, little is known about the species distribution and biogeography in the oceans (7) and in more closed environments like fish farms using oceanic water.

Some members of the Roseobacter clade produce secondary metabolites (23, 31), and the production of, e.g., antimicrobial compounds may contribute to their dominance in several niches. Also, the secondary metabolite production has caused interest in the Roseobacter clade from a biotechnological perspective. For instance, Phaeobacter gallaeciensis (formerly Roseobacter gallaeciensis) (35) and Phaeobacter inhibens are antagonistic against bacteria such as Vibrio anguillarum, Vibrio splendidus, Vibrio cholerae, Bacillus subtilis, Halomonas spp., and Pseudoalteromonas sp. due to production of tropodithietic acid (TDA) (6, 8, 9, 28). The antagonism against Vibrio species and the association of some roseobacters with algae (1, 11, 22) have spurred an interest in P. inhibens as a possible fish probiotic organism, as algae are typically used as live feed in marine larval rearing. Probiotics have been defined by FAO/WHO (17) as “live microorganisms which when administered in adequate amounts, confer a health benefit on the host.” Indeed it has been demonstrated that the survival of scallop, bream, and turbot larvae can be increased by adding cell extracts of roseobacters to the tank water or feeding the larvae with rotifers loaded with the probiont (34, 39, 44).

The Roseobacter clade is one of the dominant groups of colonizers on surfaces in marine environments (13), and in a Spanish turbot larval unit roseobacters were found more often on surfaces (e.g., tank walls) than in the water and a number of specific subtypes appeared as stable colonizers of the rearing unit (29). Specifically, for the Phaeobacter strain 27-4, which was isolated from a Spanish turbot farm, it was demonstrated that production of the anti-Vibrio substance TDA occurred only under growth conditions that also facilitated biofilm formation at the air-liquid interface and on inert surfaces (8, 9).

With the rapidly growing aquaculture industry now supplying more than 40% of the fish used for human consumption (16), there is an intense interest in disease control measures that do not rely on classical antibiotics. The roseobacters, as mentioned, appear to be one such option as fish probiotics. However, it is not known if their colonization of the Spanish turbot larval unit is a unique finding or if the clade in general due to its association with algae and its dominance in the marine environment will be a common component of the microbiota in marine fish-rearing units and, hence, a more universal fish probiotic candidate.

The purpose of the present study was to determine if isolates belonging to the Roseobacter clade also were selected in a turbot larval farm using a water source (water from a Danish fiord) very different from the Atlantic marine waters used in Galician turbot-growing facilities. In the model organism, Phaeobacter strain 27-4, anti-Vibrio activity is tightly coupled with specific (stagnant) growth conditions, and the purpose of the present study was also to determine if this is a general trait in Roseobacter clade strains that produce TDA. From an applied fish-farming perspective, it would be advantageous to select probiotic strains where anti-Vibrio activity occurred under a broad range of growth conditions.

MATERIALS AND METHODS

Bacterial strains.

The target strain used for testing antimicrobial activity of bacteria from the turbot-growing unit was V. anguillarum 90-11-287 (serotype O1) isolated from rainbow trout (46). Roseobacter clade strains 27-4, 8-1, 234-2, 234-9, 234-10, 256-7, 267-1, 270-3, and 632-1 were isolated from two Spanish turbot-rearing units due to their antimicrobial properties (28). Silicibacter sp. strain TM1040 originates from dinoflagellate cultures of Pfiesteria piscicida (36). All strains were stored at −80°C in a freeze medium (30.0 g tryptone soy broth [Oxoid CM129B; Oxoid, Hampshire, England], 5.0 g glucose, 20.0 g skim milk powder, 40.0 g glycerol, 1000 ml H₂O) (21). Strains were cultured on marine agar 2216 (MA; Difco, BD, Sparks, MD) or in marine broth 2216 (MB; Difco) for 2 to 3 days at 25°C.

Sampling and isolation of antimicrobial strains.

Samples were taken for microbiological analysis from a Danish turbot larval farm in October 2006. The turbot unit used water from Limfjorden, which is a shallow fiord transecting the Danish Jutland peninsula, and the salinity is around 2.5%. Samples were collected from walls and water from fish tanks; tanks with rotifers, nauplii, and zooplankton; and bags with algal cultures. Sterile gloves, swabs, and plastic beakers were used for sampling. All samples were 10-fold diluted in sterile seawater and spread onto MA. Liquid samples were diluted directly, whereas swab samples were mixed thoroughly with 1 ml autoclaved seawater (constituting a 10⁰ dilution.). MA plates were incubated for 6 days at 20°C. To isolate colonies with antimicrobial activity, MA plates were replica plated onto seawater agar (1,000 ml seawater, 3.33 g Casamino Acids [Difco 223050], 20 ml 20% glucose, and 10 g agar) plates containing 5 × 10⁵ to 1 × 10⁶ CFU of V. anguillarum 90-11-287/ml agar. The agar was held at 44°C and mixed with V. anguillarum 90-11-287 previously grown in 10 ml MB for 2 days at 25°C. Colonies causing clearing zones after incubation of the seawater agar plates for 24 h at 25°C were isolated from the original MA plate and pure cultured. All strains were tested in a spot inoculation test and a well diffusion assay using colony mass and sterile-filtered supernatants, respectively, to verify the antimicrobial activity. In both assays, the agar was seeded with V. anguillarum 90-11-287 as described above, and Phaeobacter strain 27-4 was used as a positive control. Colonies of strains grown on MA at 25°C for 3 days were inoculated in one spot (approximately 0.5 cm) on seawater agar or Instant Ocean agar (1,000 ml H₂O, 30.0 g Instant Ocean [Aquarium Systems Inc., Sarrebourg, France], 3.33 g Casamino Acids, 20 ml 20% glucose, and 10 g agar). The clearing zone around the spot was read after 24 and 48 h at 25°C. For the well diffusion assay, 70 μl sterile-filtered (pore size, 0.22 μm; Millipore, Bedford, MA) supernatants of each isolate grown in 10 ml MB in 25-ml bottles for 3 days at 25°C under stagnant or shaking (200 rpm) conditions were added to wells (diameter, 6 mm) in solidified seawater agar. Diameters of clearing zones were measured after 24 h at 25°C.

Sampling and isolation of strains representing the general microbiota.

To determine if the Roseobacter clade bacteria would have a pronounced effect on the general microbiota of the turbot larval farm, we investigated the ability of the Roseobacter clade strains to inhibit representative strains of the microbiota. Plate counts on six samples collected from the Danish turbot farm in January and February 2008 from three tanks (water and surfaces) were performed as described above. Colonies were isolated randomly and pure cultured. Spot assays were performed with 17 isolates embedded in Instant Ocean agar, and the Roseobacter clade strains were spotted on top of it as described above. Phaeobacter strain 27-4 was also tested against the 17 isolates. The 17 isolates were identified by biochemical tests and BLASTN (2) search on the National Center for Biotechnology Information (NCBI) database (4) (≥98% identity) using 16S rRNA sequences (for biochemical tests and sequencing procedures, see below).

Identification of bacteria.

Biochemical tests were used to identify the isolates which retained antimicrobial activity as determined by spot assay or that were isolated as representatives of the turbot-rearing microbiota. Gram stain (Bactident aminopeptidase; Merck, Darmstadt, Germany), catalase (3% H₂O₂), and oxidase (BBL Oxidase Dryslide; BD) reactions were tested on cultures grown for 1 day on MA at 25°C. Shape, motility, and ability to form rosettes were examined by phase-contrast microscopy of cultures grown in MB for 3 days at 25°C under stagnant conditions. The ability to ferment or oxidize glucose was tested in OF basal medium (Merck) (30) supplemented with 2% Instant Ocean. Fermentative strains were grown for 1 day at 25°C in 5 ml MB and streaked on plates selective for vibrios (TCBS cholera medium [Oxoid CM333]), and the cultures were also tested for sensitivity to vibriostaticum (2,4-diamino-6,7-di-isopropylteridine; 0/129 DD0014, 10 μg, and DD0015, 150 μg; Oxoid) on MA. All plates were incubated at 25°C for 24 h.

Bioinformatics on Roseobacter clade strains.

The 16S rRNA genes of presumed Roseobacter clade strains (gram-negative, nonfermentative/nonoxidative, motile rods with positive catalase and oxidase reactions and the ability to form rosettes and brownish pigment) were sequenced not only to identify the isolates but also to determine similarities of the strains isolated from the Danish and Spanish turbot-rearing units. The gyrase B (gyrB) genes were sequenced to determine if the very homogenous clusters found by comparing 16S rRNA genes were also reflected in this housekeeping gene.

DNA was purified from cultures grown for 3 days at 25°C in 5 ml MB using the Dynal Dynabeads DNA Direct System (Dynal Biotech ASA, Oslo, Norway). Two microliters DNA was mixed with 13 μl 2× Brilliant IIQPCR Master Mix (Stratagene, La Jolla, CA), 8 μl sterile MilliQ water, and 1 μl 12.5 M of each primer. The primers were synthesized by DNA Technology A/S (Aarhus, Denmark), and we used 27F (5′ AGAGTTTGATCMTGGCTCAG 3′) and 1492R (5′ TACGGYTACCTTGTTACGACTT 3′) for 16S rRNA genes and UP-1 (5′ GAAGTCATCATGACCGTTCTGCAYGCNGGNGGNAARTTYGA 3′) and UP-2 (5′ AGCAGGGTACGGATGTGCGAGCCRTCNACRTCNGCRTCNGTCAT 3′) for gyrB genes (52). The PCRs for 16S rRNA genes were run (9800 Fast Thermal Cycler; Applied Biosystems, Foster City, CA) for 10 min at 95°C before 35 cycles of 95°C for 30 s, 51°C for 1 min, and 72°C for 1.5 min and, after the last cycle, 7 min at 72°C. For the gyrB genes, the reactions were run for 10 min at 95°C followed by 40 cycles of 1 min at 95°C, 1 min at 60°C, and 2 min at 72°C. The program ended with 7 min at 72°C. All PCR products were analyzed by 1% agarose gel electrophoresis, bands were cut out, and DNA was purified (GFX PCR DNA and gel band purification kit; GE Healthcare, Buckinghamshire, Great Britain). Sequencing was done by DNA Technology A/S using the primer set 518F (sequence, 5′ CCAGCAGCCGCGGTAATACG 3′) and 800R (sequence, 5′ TACCAGGGTATCTAATCC 3′) for 16S rRNA genes and UP1S (5′ GAAGTCATCATGACCGTTCTGCA 3′) and UP2Sr (5′ AGCAGGGTACGGATGTGCGAGCC 3′) for gyrB genes (52).

Sequences were assembled using Vector NTI (Invitrogen, Carlsbad, CA). BLASTN searches using 16S rRNA sequences were performed on the NCBI database to find sequences with ≥98% identity. Relevant type strains for species identified were found online in the List of Prokaryotic Names with Standing in Nomenclature (15) and included in the analysis. Multiple alignments were done using the program ClustalX (48), and the alignments were edited in BioEdit (25). Distance matrix JC was calculated using ClustalX and neighbor joining (45), and bootstrap (number of trials, 100) (18) trees were drawn using the program MEGA4 (47). Type strains Rhodobacter capsulatus ATCC 11166 plus Rhodobacter sphaeroides ATCC 17023 and Roseobacter denitrificans Och114 plus Roseobacter litoralis Och149 served as outgroups in the 16S rRNA and gyrB trees, respectively.

Subtyping of Roseobacter clade strains by RAPD.

Roseobacter clade isolates were random amplified polymorphic DNA (RAPD) typed as described earlier (50) to determine subspecies homology. In brief, 2 μl purified DNA plus one Ready-To-Go RAPD Analysis Bead (Amersham Pharmacia Biotech Inc., Piscataway, NJ) was dissolved in 23 μl of one of the following 1 μM primers (DNA Technology A/S): UBC 104 (sequence 5′ GGGCAATGAT 3′) and UBC 106 (sequence 5′ CGTCTGCCCG 3′) (29). The PCRs were run at 95°C for 2 min followed by 10 cycles of 1 min at 94°C, annealing at 45°C for 1 min, and extension at 72°C for 2 min. The annealing temperature was decreased by 1°C per cycle. Then 30 cycles followed with denaturing at 94°C for 1 min, annealing at 35°C for 1 min, and extension at 72°C for 2 min. The program was completed with 10 min of final extension at 72°C. The bands were visualized after electrophoresis in 2% agarose gels by staining with ethidium bromide. Phaeobacter strains (27-4, 8-1, and 632-1) and a 100-bp ladder (Amersham) standard were included in all gels.

Influence of growth conditions on attachment, pigment formation, and antibacterial activity of Roseobacter clade strains.

All Roseobacter clade strains were grown in 5 ml MB for 3 days at 25°C under stagnant conditions, and 200 μl was reinoculated in 20 ml MB in 250-ml bottles and grown for 3 days at 25°C under stagnant or shaking (200 rpm) conditions. Sterile-filtered supernatants were used for testing antimicrobial activity in a well diffusion assay in Instant Ocean agar as described above and for measuring pigment by spectroscopy (Novaspec II; Pharmacia Biotech, Cambridge, England) at 398 nm (9). Cell numbers were determined on cultures by 10-fold dilutions in sterile 0.85% saline and plated on MA, and plates were incubated at 25°C for 3 days. A subset of seven Roseobacter clade strains from the Danish turbot larval farm, four from the Spanish turbot larval farm, the strain from a dinoflagellate culture, and the V. anguillarum strain not able to form rosettes were examined for their ability to attach to surfaces as described by Bruhn et al. (8). Briefly, glass coverslips (Knittel Glässer, Braunschweig, Germany) were dipped for 5 s in cultures grown in 20 ml MB in 250-ml bottles for 3 days at 25°C under stagnant or shaking (200 rpm) conditions as described above. Loosely attached cells were removed by placing the coverslip on absorbent paper, after which the remaining cells were fixed at 60°C for 30 min. Attached cells were stained for 15 min in 0.1% crystal violet, and unbound dye was washed off using phosphate-buffered saline (BR0014G; Oxoid). Dye bound to attached cells was dissolved in 2 ml 33% acetic acid, and the optical density at 590 nm (OD₅₉₀) was measured. The attachment was done in duplicate. The crystal violet OD resulting from dipping glass coverslips in pure MB were subtracted from all measurements. Phase-contrast microscopy was performed on cells from the air-medium interface of the cultures. Pigment was measured as described earlier (9). The antibacterial compound TDA was measured in sterile-filtered supernatant as described below.

HPLC-tandem mass spectrometry analysis of TDA.

Extracts of 5 μl were analyzed on an Agilent (Torrance, CA) 1100 high-pressure liquid chromatography (HPLC) system controlled by MassLynx V4.1. Samples were separated on a Gemini C₆-phenyl 3-μm, 2-mm-inside-diameter × 50-mm column (Phenomenex, Torrance, CA), using a flow rate of 0.300 ml/min at 25°C. A linear water-acetonitrile (ACN) gradient was used, starting at 10% ACN, going to 45% ACN in 8 min and then 100% ACN in 0.5 min, holding this for 2 min before reverting to 10% ACN in 1 min, and maintaining this for 8 min. Both solvents contained 120 mM formic acid. The HPLC was a coupled Quattro Ultima triple mass spectrometer (Waters-Micromass, Manchester, United Kingdom) with a Z-spray electrospray ionization source using a flow rate of 700 liters/h nitrogen at 350°C; hexapole 1 was held at 30 V, and the cone was held at 25 V. Nitrogen was used as collision gas, and the mass spectrometer operated in positive multiple-reaction monitoring mode (dwell time, 100 ms), monitoring m/z 213 to 151 (25-V collision energy) and m/z 213 to 167 (20-V collision energy) as quantifier and qualifier ions, respectively.

Nucleotide sequence accession numbers.

The 16S rRNA and gyrB gene sequences have been deposited in GenBank under the accession numbers FJ014969 to FJ01503 and FJ014947 to FJ014968, respectively.

RESULTS

Sampling, isolation, and identification of antimicrobial strains.

Forty-three samples from the Danish turbot larval farm (Table 1) all had aerobic plate counts of 10⁵ to 10⁶ CFU/ml on MA. Thirty-one samples contained colonies that in the replica assay indicated inhibition of V. anguillarum. We isolated cells from 117 of these colonies, and upon retesting, the inhibitory activity was retained for 100 isolates. Fifty-four of the 100 isolates also inhibited V. anguillarum 90-11-287 when we used sterile-filtered supernatant from stagnant cultures in the well diffusion assay. These isolates originated from water (23 isolates), tank surfaces (22 isolates), and algal cultures (nine isolates). Instant Ocean agar was used for all subsequent experiments, as this is easier to standardize than an agar based on natural seawater, and gave very similar inhibitory profiles. Fifty-one of the 54 isolates were gram-negative, nonfermentative/nonoxidative, motile rods with positive catalase and oxidase reactions and rosette formation and therefore were likely to belong to the Roseobacter clade. It was not possible, based on the biochemical tests used, to identify the last three isolates. The 46 isolates that were antagonistic only in the spot assay and not in the well diffusion assay were primarily identified as Vibrio spp. (38 out of 46), being fermentative, gram-negative rods able to grow on TCBS agar and in the presence of vibriostaticum.

TABLE 1.

Numbers of samples taken at a Danish turbot larval farm and numbers of isolates with antagonistic activity against Vibrio anguillarum strain 90-11-287 in seawater agar

Sample site	No. of samples		No. of antagonistic isolates
	Total	No. with antagonistic bacteria determined by replica plating	Total	Determined by spot assay	Determined by well diffusion assay
Inside production
Fish tanks
Water	4	3	12	11	5
Surface	9	3	7	7	6
Water cleaning system
Water	3	3	13	12	5
Surface	2	2	10	9	9
Tank with copepod cultures
Suspension	4	3	9	2	0
Surface	4	4	11	8	0
Bag with algal culture, suspension	4	3	11	11	9
Subtotal	30	21 (70%)	73	60 (82%)	34 (47%)
Outside production
Tanks with larvae
Water	2	2	11	11	6
Surface	2	1	1	1	1
Tanks with zooplankton
Water	2	2	11	9	5
Surface	2	1	7	6	6
Water cleaning system, surface	1	1	2	2	0
Tank with water from the inlet
Water	1	1	5	5	0
Surface	1	1	3	2	0
Tank with nauplii
Water	1	1	4	4	2
Surface	1	0	0	0	0
Subtotal	13	31 (76%)	44	40 (90%)	20 (45%)
Grand total	43	31 (76%)	117	100 (85%)	54 (46%)

BLASTN searches supported the Roseobacter clade affiliation of the 51 isolates, as all 16S rRNA gene sequences showed ≥98% similarity to species such as Roseobacter spp., P. inhibens, P. gallaeciensis, and Phaeobacter daeponensis, whereas isolates from algal cultures were ≥98% similar to Ruegeria mobilis, Ruegeria pelagia, and Silicibacter spp. (Table 2).

TABLE 2.

Identification of antagonistic bacterial strains isolated from a Danish turbot larval farm

Sampling site	No. of isolates antagonistic in:
	Both spot and well diffusion assaysa			Spot but not well diffusion assayb
	Phaeobacter spp.	Ruegeria spp.	Others	Roseobacter-like	Vibrio spp.	Others
Indoor production
Fish tanks
Water	5	0	0	0	6	0
Surface	6	0	0	0	1	0
Water cleaning system
Water	5	0	0	0	7	0
Surface	9	0	0	0	0	0
Tank with copepod cultures
Suspension	0	0	0	0	1	1
Surface	0	0	0	0	4	4
Bag with algal culture, suspension	0	9	0	1	0	1
Subtotal	25	9	0	1	19	6
Outdoor production
Tanks with larvae
Water	6	0	0	0	5	0
Surface	1	0	0	0	2	0
Tanks with zooplankton
Water	4	0	1	0	4	0
Surface	6	0	0	0	0	0
Water cleaning system, surface	0	0	0	0	2	0
Tank with water from the inlet
Water	0	0	0	0	5	0
Surface	0	0	0	0	0	0
Tank with nauplii
Water	0	0	2	0	1	1
Surface	0	0	0	0	0	0
Subtotal	17	0	3	0	19	1
Grand total	42	9	3	1	38	7

^aIdentified by BLASTN search with 16S rRNA gene sequences on the NCBI database.

^bIdentified by biochemical tests (Gram stain, catalase and oxidase tests, ability to metabolize glucose, reaction on TCBS plates, and sensitivity/resistance to the vibriostatic agent 0129) and microscopy (motility, shape, and ability to form rosettes).

RAPD subtyping of Roseobacter clade strains.

RAPD typing divided the 51 Roseobacter clade strains into 14 different subtypes when two primers were used (data not shown). None of these types were similar to three major RAPD types that had been found in the Spanish turbot-rearing farms (29). Three major subgroups found in the Danish turbot-rearing unit were designated AA, BB, and CC and harbored 23, 4, and 9 strains, respectively. All the 23 subtype AA strains originated from parts of the production site that were indoors. Only two indoor isolates belonged to other subtypes (BB and DD). The RAPD type CC encompassed nine strains, and all were isolated from the algal cultures. All of these nine strains were subsequently identified by 16S rRNA gene sequencing as Ruegeria spp. Subtype BB harbored four strains isolated at different sites. The remaining 17 isolates were isolated from the outdoor sites of the production unit and made up a total of 11 RAPD types.

16S rRNA and gyrB gene comparison of Roseobacter clade strains.

Three major clusters appeared when comparing 989 bases of the 16S rRNA gene from 48 Danish and nine Spanish Roseobacter clade isolates (Fig. 1a). Three Danish isolates could not be amplified. One cluster was identical to type strains R. mobilis MBIC01146 and R. pelagia HTCC2662 and contained nine isolates representing two samples from Danish algal cultures and one Spanish isolate. Twenty-one Danish isolates from seven indoor samples (all RAPD type AA) and eight isolates from the Spanish turbot farm clustered together with the type strain P. gallaeciensis BS107. The last cluster consisted of the type strain P. inhibens T5 and Danish isolates of different RAPD types. Fifteen isolates were from five outdoor samples, and two isolates were taken from two different indoor samples. The gyrB gene (998 bases) was sequenced for 19 Danish and three Spanish strains representing the different clusters obtained in the 16S rRNA tree. It resulted in the same three major clusters when this housekeeping gene was used as did the 16S rRNA gene sequences; however, a greater separation between the Danish and the Spanish strains was achieved using the gyrB sequences (Fig. 1b). Furthermore, the Spanish isolate 8-1 was affiliated with the P. gallaeciensis BS107 cluster in the 16S rRNA tree whereas in the gyrB tree it grouped with the strains from the P. inhibens T5 cluster.

FIG. 1.

Phylogenetic trees constructed using the 16S rRNA gene (a) and gyrase B (gyrB) gene (b) sequences. Numbers at the nodes are bootstrap values from 100 replicates. Type strains Rhodobacter capsulatus ATCC 11166 and Rhodobacter sphaeroides ATCC 17023 served as outgroups in the 16S rRNA tree, and Roseobacter denitrificans Och114 and Roseobacter litoralis Och149 served as outgroups in the gyrB gene trees (not shown). and ▪, Danish and Spanish turbot-rearing farm strains, respectively. ^T, type strains.

The gyrB sequences of some isolates revealed putative stop codons in five isolates (Phaeobacter strains M3-1.2, M4-3.1, M6-4.1, M9-4.2, and M9-4.3) or frameshift mutations in 12 isolates (Phaeobacter strains 8-1, 27-4, M2-3.1, M2-4.2, M3-1.3, M4-3.1, M6-4.1, M9-4.1, M9-4.2, and M9-4.3 and Ruegeria strains M41-3.1 and M43-1); however, the reason for this or the functional implications have not been further pursued.

Inhibition of the general microbiota by Roseobacter clade strains.

Seventeen isolates randomly collected from water and surface samples from the Danish turbot larval farm were inhibited by the 51 Roseobacter clade strains, although some of the strains were inhibited to a lesser degree than was V. anguillarum 90-11-287, as indicated by the size of the inhibition zone (Table 3). The inhibition zones surrounding the 42 Phaeobacter sp. strains in the spot assay were larger than the zones produced by the nine Ruegeria sp. strains. This pattern was especially pronounced for target strains isolated from water samples. The 17 isolates were identified as eight different genera or species, and one isolate was impossible to identify based on the methods used. The Phaeobacter spp. strongly inhibited all but one isolate (Rhodococcus sp.), whereas the Ruegeria spp. inhibited all but six isolates (Halomonas sp./Cobetia marina, Pseudomonas sp., Pseudoalteromonas sp., and Rhodococcus sp.). There was no pattern in the sensitivity of the target strains to Roseobacter clade strains depending on genus or species of the target organism.

TABLE 3.

Abilities of Phaeobacter spp. and Ruegeria spp. to inhibit 17 isolates from a Danish turbot larval farm^e

Identification of isolates from samplea	Grading of the size of inhibition zone in spot assay forb:
	Phaeobacter spp. (42 strains)	Ruegeria spp. (9 strains)
Water, indoor fish tank
Halomonas sp./Cobetia marinac	3.0 ± 0.0	1.0 ± 0.0
Halomonas sp./Cobetia marina	3.0 ± 0.0	1.6 ± 0.5
Micrococcus sp.	3.0 ± 0.2	2.9 ± 0.3
Halomonas sp./Cobetia marina	3.0 ± 0.0	0.7 ± 0.5
Water, outdoor tank with phytoplankton and zooplankton
Pseudomonas sp.	3.0 ± 0.0	0.9 ± 0.8
Pseudomonas sp.	3.0 ± 0.2	0.4 ± 0.9
Pseudoalteromonas sp.	2.0 ± 0.5	0.0 ± 0.0
Water, indoor fish tank
Pseudoalteromonas sp.	3.0 ± 0.0	2.0 ± 0.9
Micrococcus sp.	2.9 ± 0.3	2.4 ± 0.5
Marinomonas sp.	3.0 ± 0.0	2.3 ± 1.0
Surface, indoor fish tank
Rhodococcus sp.	1.0 ± 0.0	0.0 ± 0.0
Surface, outdoor tank with phytoplankton and zooplankton
Micrococcus sp.	3.0 ± 0.0	3.0 ± 0.0
Surface, indoor fish tank
Olleya marilimosa	2.9 ± 0.3	2.1 ± 0.9
Unidentifiedd	2.6 ± 0.5	3.0 ± 0.0
Kordia algicida	3.0 ± 0.2	2.0 ± 0.5
Pseudoalteromonas sp.	2.8 ± 0.7	2.4 ± 0.7
Micrococcus sp.	2.7 ± 0.7	2.1 ± 0.9
Diseased rainbow trout (Oncorhyncus mykiss)
Vibrio anguillarum 90-11-287	3.0 ± 0.3	2.7 ± 0.7

^aIsolates were identified by biochemical tests (Gram stain, catalase and oxidase tests, ability to metabolize glucose, reaction on TCBS plates, sensitivity/resistance to vibriostatic agent 0129), microscopy (motility and shape), and BLASTN searching of 16S rRNA gene sequences in the NCBI database.

^bGrades (0 to 3) indicate the size of the clearing zone around the spot of colony mass: 0, no zone; 1, <0.5-mm zone; 2, 0.5- to 2-mm zone; 3, >2-mm zone. Values are averages and standard deviations of 42 Phaeobacter sp. and nine Ruegeria sp. strains, respectively.

^cCobetia marina (reclassified from Halomonas marina [3]) and Halomonas sp. both belong to the Halomonadaceae family.

^dNot possible to identify based on the biochemical tests, and BLASTN search gave low identities to other known species (≤96%).

^eTarget bacteria were seeded in agar, and Phaeobacter spp. or Ruegeria spp. were added on top in a spot assay.

Influence of growth conditions on attachment, pigment formation, and antibacterial activity of Roseobacter clade strains.

All the 51 Danish and nine Spanish Roseobacter clade strains produced brown pigment (as determined by OD₃₉₈) when grown under stagnant conditions in MB. However, 42 Danish and seven Spanish isolates also produced pigment when grown under aerated (shaking) conditions, although the amount of pigment was less than that in stagnant cultures (data not shown), and all these strains were identified as Phaeobacter spp. The brown pigment correlated with the antibacterial activity of the supernatant in the well diffusion assay, as inhibition was observed only when pigment was produced. The remaining nine Danish and two Spanish isolates expressed these phenotypes only during stagnant growth. Of these, all Danish strains and one Spanish strain were identified as Ruegeria spp., and the last Spanish strain behaving like the Ruegeria spp. was Phaeobacter strain 27-4. All stagnant and shaken cultures grew to approximately 5 × 10⁸ CFU/ml and 1 × 10⁹ CFU/ml, respectively. It was examined for a subset of samples (eight Phaeobacter sp. strains, three Ruegeria sp. strains, Silicibacter strain TM1040, and V. anguillarum 90-11-287) if rosette formation and attachment ability co-occurred with pigment production and antibacterial activity as seen for Phaeobacter strain 27-4 (8, 9). All Roseobacter clade isolates attached better to the glass surface as determined by crystal violet staining when grown under stagnant conditions (Fig. 2), and more cells appeared in a rosette morphology under stagnant conditions than under shaking conditions (Fig. 3) (data are shown only for the Danish Phaeobacter strain M23-3.1, the Spanish Phaeobacter strain 27-4, and the Danish Ruegeria strain M43-2.3 but are typical for the other Roseobacter clade strains). This was also the case for Phaeobacter spp. and Silicibacter strain TM1040, which produced brown pigment and had antibacterial activity when grown under shaking conditions. For all tested strains, production of TDA correlated with pigment formation and antibacterial activity (Fig. 2).

FIG. 2.

Attachment (a), production of pigment (b), production of TDA (c), and ability to inhibit Vibrio anguillarum 90-11-287 in well diffusion assay (d) of Roseobacter clade strains and Vibrio anguillarum 90-11-287 grown under shaking (200 rpm) (black bars) or stagnant (0 rpm) (gray bars) conditions. The attachment experiment was conducted in duplicate, and error bars represent 1 standard deviation. When measuring TDA, the HPLC-tandem mass spectrometry peak area from the m/z 213 to 151 transition was used. In the well diffusion assay, the diameter of the well itself has been subtracted from the diameter of the inhibition zone.

FIG. 3.

Cell morphology of and pigment formation by Roseobacter clade strains grown in MB under static (0 rpm) (a, c, and e) or shaking (200 rpm) (b, d, and f) conditions. The Danish Phaeobacter sp. strain M23-3.1 (a and b), the Spanish Phaeobacter sp. strain 27-4 (c and d), and the Danish Ruegeria sp. strain M43-2.3 (e and f) are shown. Microscopy pictures are from phase-contrast microscopy at ×1,000 magnification. Bars, 10 μm.

DISCUSSION

Roseobacter clade strains capable of in vitro inhibition of the fish-pathogenic bacterium V. anguillarum were found in many sites of a Danish turbot larval farm. In fact, these bacteria strongly dominated the antagonistic culturable microbiota (51 out of 54 isolates). Roseobacter spp. have been isolated from a Spanish turbot larval unit (29) and from scallop larval cultures (43), and Brunvold et al. (10) detected a band using denaturing gradient gel electrophoresis with sequence homology to a Roseobacter sp. strain from a cod hatchery. This, collectively, suggests that Roseobacter clade members are common in this type of environment. We did, like Hjelm et al. (29), also identify antagonistic Vibrio spp. from the turbot larval rearing unit; however, these were not further characterized, as they are likely pathogenic for the turbot (28).

Roseobacter clade strains are often associated with surface colonization in the marine environment (12, 13) and were isolated predominantly from surfaces in a Spanish turbot farm (29). However, in the present study of the Danish turbot farm, they were equally prevalent in the tank water, and they have indeed also been isolated as pelagic bacteria in marine environments (11, 40). Roseobacter clade strains are often associated with algal blooms in marine environments and can account for as much as 50% of the bacterial rRNA genes (1, 11, 22). In concordance with this niche preference, we isolated several strains from the algal cultures. This particular association may render the algae an interesting vector for supplying the probiotic culture to the fish larval rearing process.

Comparing 989 bases of the 16S rRNA gene revealed three clusters among the Danish and Spanish turbot-rearing farm isolates. Strains similar to P. gallaeciensis BS107 (type strain) dominated among the indoor isolates, whereas outdoor strains were similar to P. inhibens T5 (type strain), and all isolates from algal cultures were similar to type strains R. mobilis MBIC01146 and R. pelagia HTCC2662. The two latter type strains are identical on the 16S rRNA level. This indicates that different species of the Roseobacter clade colonize specific niches in the Danish rearing unit. All four species from the Roseobacter clade are very common in marine environments (11), and as they are easily culturable (6, 32, 38, 43), it is not surprising to find them in a nutrient-rich environment favoring fast-growing heterotrophic organisms. The 16S rRNA gene is very useful as a phylogenetic comparative gene for a number of genera and species; however, it may not be the optimal differentiating molecule in some genera (14, 49). We considered that the Roseobacter clade species could, in principle, be difficult to differentiate based on 16S rRNA genes and therefore chose to compare the strains using the gyrB gene also, which has been used successfully to discriminate between genera and species with closely related 16S rRNA gene sequences (19, 26, 42, 53). In this study, the gyrB gene supported the finding of the clusters obtained with 16S rRNA gene sequences, but at the same time it gave a greater evolutionary distance between the Danish and Spanish strains. To our knowledge, this is the first time that the gyrB gene has been used for phylogenetic analysis of Roseobacter clade members.

Subtyping the Danish Roseobacter clade strains using RAPD with two primers revealed types other than those found in the Spanish turbot larval farm (29). The two turbot farms have different water sources (Limfjorden, a fiord in Denmark, versus the Atlantic Ocean off the Galician coast). Also, specific subtypes may have preferences for special niches or may have been introduced at random and remained. This is indicated by the fact that even though the same water source was used in the Danish turbot larval farm for inside and outside production, only one out of 14 RAPD types was found at both sites. The diversity in terms of subtypes was much lower in the indoor production than in the outside production sites. A similar phenomenon has been reported for the human-pathogenic bacterium Listeria monocytogenes (27), which is an environmental bacterium capable of colonizing fish-processing units. This could be because the outdoor production is an open production form and thereby more easily influenced by the surrounding environment than is the indoor production.

The Phaeobacter spp. and Ruegeria spp. isolated due to their antimicrobial activity against V. anguillarum also inhibited bacterial strains randomly isolated from surfaces and water from fish tanks in the turbot farm. This is in agreement with the other studies, as alphaproteobacteria can inhibit different marine bacteria (24, 33), and Phaeobacter spp. inhibit organisms as diverse as flavobacteria, Acinetobacter, vibrios, Pseodoalteromonas, Alteromonas, Bacillus, and Halomonas (5, 6, 8, 9, 28, 41). Hence, one should be aware that if such bacteria are deliberately added as probiotic organisms, they may not just inhibit the pathogenic agents but also alter the general microbiota. Therefore, careful supervision of the changes in microbiota upon probiotic additions is required.

Phaeobacter spp. appeared more inhibitory than Ruegeria spp.; however, the two species produced TDA in similar amounts when grown in stagnant broth cultures (Fig. 2c). We therefore speculate that Phaeobacter spp. may produce inhibitory compounds other than TDA and that these act as antimicrobials themselves or act synergistically with TDA. The antagonistic activity of Phaeobacter spp. and Ruegeria spp. is likely to be an important factor explaining their dominance in several niches, and indeed they outcompete other marine organisms such as Pseudoalteromonas tunicata in competition experiments (41).

The ability of all the Roseobacter clade strains tested to inhibit V. anguillarum co-occurred with their production of brown pigment and was independent of the origin of the isolates. A similar coupling has been reported for P. inhibens T5, Phaeobacter strain 27-4, and Silicibacter strain TM1040 (6, 8, 9). While the inhibitory compound TDA is not the pigment itself, the coupling between these two phenotypes has also been demonstrated by transposon mutagenesis creating noninhibitory mutants in a Phaeobacter strain and a Silicibacter strain (20). The production of pigment and antibacterial activity was influenced by growth conditions; however, this varied with species, as the Ruegeria spp. behaved like Phaeobacter strain 27-4 (9), producing pigment and showing antimicrobial activity only when grown under stagnant conditions. In contrast, Phaeobacter spp. from the Danish turbot larval farm expressed these phenotypes after both stagnant and aerated growth. This is interesting from an application point of view, as only a few sites in a fish tank will be stagnant, and hence, the Phaeobacter spp. isolated in the present study may hold greater promise as probiotic organisms than Phaeobacter strain 27-4. It must be determined if the Danish Phaeobacter strains are capable of disease suppression in vivo as has been documented for strain 27-4 (39). As seen in this study, strains with 100% similarity in the 16S rRNA gene sequences do not necessarily express phenotypes in the same pattern. The same phenomenon was observed by Grossart et al. (24), as isolates showed different inhibitory activities against marine bacteria. Also, Silicibacter strains (Silicibacter strain TM1040 and Silicibacter pomeroyi strain DSS-3) vary in how the culture conditions influence pigment production and antimicrobial activity (8). In the present study, the production of TDA, pigment, and antimicrobial activity co-occurred for a subset of 12 Roseobacter clade strains (eight of Phaeobacter spp., three of Ruegeria spp., and one of Silicibacter sp.). TDA has been detected from Phaeobacter spp. and Silicibacter strain TM1040 (6, 9, 20), and to the best of our knowledge the present study is the first report of TDA production by R. mobilis/R. pelagia strains.

A prominent characteristic of Phaeobacter and some other Roseobacter clade strains is their ability to grow as rosettes (8, 9), and this mode of growth appears to enhance the surface attachment ability of organisms. Also, the strains isolated in the present study grew as rosettes, and this did correlate, for Ruegeria, with their attachment capability.

In conclusion, Phaeobacter spp. and Ruegeria spp. with antibacterial activity colonized different units in the Danish turbot larval farm. We suggest that members of the Roseobacter clade are common colonizers of marine larval rearing units. This makes the clade a suitable candidate as a universal marine fish larval probiotic bacterium. In particular the Danish strains of Phaeobacter spp. are of applied interest as their antibacterial activity and TDA production occurred under several types of growth conditions. However, the true probiotic potential will have to be further evaluated in in vitro and challenge trials.