Identification of an Antagonistic Probiotic Combination Protecting Ornate Spiny Lobster (Panulirus ornatus) Larvae against Vibrio owensii Infection

Vibrio owensii DY05 is a serious pathogen causing epizootics in the larviculture of ornate spiny lobster Panulirus ornatus. In the present study a multi-tiered probiotic screening strategy was used to identify a probiotic combination capable of protecting P. ornatus larvae (phyllosomas) from experimental V. owensii DY05 infection. From a pool of more than 500 marine bacterial isolates, 91 showed definitive in vitro antagonistic activity towards the pathogen. Antagonistic candidates were shortlisted based on phylogeny, strength of antagonistic activity, and isolate origin. Miniaturized assays used a green fluorescent protein labelled transconjugant of V. owensii DY05 to assess pathogen growth and biofilm formation in the presence of shortlisted candidates. This approach enabled rapid processing and selection of candidates to be tested in a phyllosoma infection model. When used in combination, strains Vibrio sp. PP05 and Pseudoalteromonas sp. PP107 significantly and reproducibly protected P. ornatus phyllosomas during vectored challenge with V. owensii DY05, with survival not differing significantly from unchallenged controls. The present study has shown the value of multispecies probiotic treatment and demonstrated that natural microbial communities associated with wild phyllosomas and zooplankton prey support antagonistic bacteria capable of in vivo suppression of a pathogen causing epizootics in phyllosoma culture systems.


Introduction
The ornate spiny lobster (Panulirus ornatus) is considered a prospective aquaculture species based on encouraging growout potential [1] and lucrative market value [2]. However, closedlife cycle production of P. ornatus is currently commercially unviable due to restricted production of postlarvae resulting from nutritional deficits and bacterial disease during their 4-6 month long larval phase [3][4][5][6][7]. Vibrio owensii is an emerging pathogen, with the type strain DY05 demonstrated as the etiological agent of a disease causing mass mortalities of cultured P. ornatus larvae (phyllosomas) [7,8]. The pathogen can be transmitted through live feed vectors (Artemia) and proliferates in the phyllosoma hepatopancreas (midgut gland), causing extensive tissue necrosis and eventually major systemic infection [7].
In view of the global antibiotic resistance crisis [9] there is considerable interest in developing sustainable biocontrol methods such as probiotics for disease management in aquaculture [10]. The search for probionts is based on screening for beneficial microbial attributes such as antagonism, predation, anti-virulence, competition, attachment to host surfaces, and immunostimulation [10][11][12][13][14]. We have previously shown that the planktonic form is central to vectored transmission of V. owensii DY05 [7], hence it was pertinent to investigate the ability of probiotic candidates to inhibit planktonic growth. Moreover, since biofilms are refuges for pathogens in aquaculture systems [15,16] and pathogen biofilms on natural tissues are inherently tolerant to conventional antimicrobial therapies [17], we also wanted to investigate the ability of probiotic candidates to inhibit biofilm formation under conditions of exclusion, competition and displacement.
Here we present a multi-tiered screening strategy for probiotic candidates, which ultimately led to the identification of a twostrain combination providing efficient protection of phyllosomas against experimental V. owensii DY05 infection. Initially, a shortlist of probiotic candidates was generated from a large pool of bacteria showing in vitro antagonism towards V. owensii DY05. Two additional in vitro screens were developed using a green fluorescent protein (GFP)-transconjugant of the pathogen to assess its planktonic growth and biofilm formation in the presence of shortlisted candidates. Subsequently, promising candidates were assessed for inherent virulence and protective benefit in vivo using a P. ornatus phyllosoma experimental infection model.

Replica Plate Assay
Wild Panulirus spp. phyllosomas and putative zooplankton prey items (not endangered or protected) were collected at Osprey Reef (Coral Sea, Australia; 13u 569S, to 14u 039 S and 144u 269 E to 146u 489 E) between 24 May-9 June 2008. No specific permits were required as the sites were located in the Australian Exclusive Economic Zone outside the Great Barrier Reef Marine Park and were not protected. Capture was achieved using a modified Isaac-Kidd mid-water trawl net according to Smith et al. [3]. Briefly, animals were washed 36 in 0.22 mm filtered artificial sea water (ASW; Instant OceanH) to remove debris and loosely attached epibionts, homogenised in ASW, and spread plated on minimal marine agar (MMA; 0.3% casamino acids; 0.4% glucose; 1% bacteriological agar in 1 L ASW), modified from Hjelm et al. [18]. After incubation at ambient temperature (24uC) for 24-48 h, MMA plates with ,300 colonies were replica plated [18] onto MMA seeded with 10 mL mL 21 of V. owensii DY05 grown overnight (24uC, 170 rpm) in marine broth 2216 (MB; BD). Replica plates were incubated for 72 h (24uC) and inspected for inhibition zones signifying antagonistic activity against V. owensii DY05. Antagonistic colonies were picked and cultured to purity on MMA, re-cultured in MB overnight (28uC, 170 rpm), and cryopreserved in 30% (v/v) glycerol (280uC).

Well-diffusion Agar Assay (WDAA)
Antagonistic isolates recovered from replica plates and the Australian Institute of Marine Science (AIMS) culture collection were tested for growth-inhibitory activity against V. owensii DY05 in a well diffusion agar assay (WDAA). In brief, the pathogen was seeded into molten MMA as outlined above. Following solidification, wells (diameter 5 mm) were cut into the agar and loaded with 40 mL of dense cultures (1-3 day old) of test isolates grown in MB (28uC, 170 rpm). Plates were incubated (28uC) and observed every 24 h for 72 h for inhibition zones. Phaeobacter (formerly Roseobacter) strain 27-4 was used as a positive antagonistic control on each plate because of its broad spectrum inhibitory activity against Vibrio pathogens [18][19][20]. Antagonism was classified according to the size of the inhibition zones as low (5-10 mm), moderate (11-20 mm) or strong ($21 mm).

Phylogenetic Identification
Colony PCR was performed on antagonistic isolates with universal primers 27F and 1492R [21] under standard conditions. PCR products were purified and sequenced using 27F (all strains) and 1492R (16 shortlisted candidates) as sequencing primers by Macrogen (Seoul, Korea). Sequences were edited with Sequencher 5.0 software (GeneCodes Corporation) and analysed using the BLAST algorithm (http://www.ncbi.nlm.nih.gov/BLAST/) to determine nucleotide-nucleotide similarity with sequences in the nr/nt database. Isolates were grouped according to phylogenetic relatedness by partial 16S rRNA gene sequence alignment using MEGA4 [22]. For the 16 shortlisted candidates sequences were submitted to GenBank under accession numbers JX075050-JX075065.

Inoculum Preparation
A GFP-labelled transconjugant of V. owensii DY05 (DY05[GFP]) was used as a proxy for pathogen growth and attachment in microgrowth co-culture and multispecies biofilm assays, respectively (described below). DY05[GFP] stably expresses the GFP, and does not differ in growth profile or virulence towards stage 1 P. ornatus phyllosomas compared to wild type V. owensii DY05 [7]. DY05[GFP] was cultured on LB20 agar plates (5 g L 21 yeast extract; 10 g L 21 neutralised peptone; 20 g L 21 NaCl; 15 g L 21 agar) supplemented with 40 mg mL 21 kanamycin and 50 mg mL 21 colistin. Probiotic candidates and wild type V. owensii DY05 were cultured on marine agar 2216 (MA) at 28uC for 24 h. For each strain, colony material was suspended in 2 mL phosphate buffered saline (PBS: 8 g L 21 NaCl; 0.2 g L 21 KCl; 1.44 g L 21 Na 2 HPO 4 ; 0.24 g L 21 KH 2 PO 4 ; pH 7.2) and absorbance adjusted to OD 600 nm 0.1 (Nanodrop ND1000). The corresponding total viable counts (expressed as CFU mL 21 ) were determined for each strain in triplicate initial experiments by spiral plating (Eddy Jet; IUL) on MA and enumeration by an automatic colony counter (Flash and Grow v1.2; IUL). This information was used to calculate the volume of each OD 600 nm 0.1 suspension needed to achieve desired starting concentrations for the assays described below.

Microgrowth Co-culture Assay
The relationship between fluorescence and CFU mL 21 of V. owensii DY05[GFP] monocultures was tested using the Pearson correlation coefficient. Triplicate samples were withdrawn from the microgrowth assay (described below) at 4 h intervals and there was a strong positive correlation (p,0.0001) between fluorescence and pathogen growth for the first 24 h ( Figure S1). This showed that the fluorescence signal generated by V. owensii DY05[GFP] could be used to indirectly quantify pathogen growth during 24 h co-culture with probiotic candidates.
To assess the activity of planktonic candidates, MB was inoculated with PBS suspensions of V. owensii DY05[GFP] (initial concentration 1610 3 CFU mL 21 ) separately or in combination with PBS suspensions of candidate probiotics (final concentrations 1610 3 , 1610 5 , or 1610 7 CFU mL 21 ) in Nunc TM (NUN137101) black microwell plates (final volume 200 mL). Separate plates were used for each candidate-pathogen combination and all treatments were performed in hextuplicate well sets. Sterile milli-Q water (200 mL) was added to perimeter rows and columns to minimise evaporative loss and plate covers were treated with 0.1% Triton X-100 in 20% ethanol to smooth condensation [23]. Plates were sealed with parafilm and incubated for 24 h (28uC, 170 rpm). Growth of the GFP-tagged pathogen was monitored indirectly by measuring fluorescence (excitation/emission 485/520 nm) with a Wallac Victor 2 1420 multilabel counter. Fluorescence values were adjusted by subtracting the average autofluorescence generated from corresponding controls (wild type V. owensii DY05 monocultures or co-cultures).
The antagonistic activity of each strain was classified based on their ability to reduce the pathogen fluorescence signal after 24 h relative to the maximum signal reduction recorded for the respective assay. In this way, strain antagonistic activity was classified as low (,50% of max), moderate (50-75% of max), or strong (.75% of max).

Biofilm Assays
Monostrain biofilm production by the pathogen and candidate probionts was quantified using a microwell crystal violet (CV) staining assay [24], modified so culture conditions were consistent with other microwell assays used in the present study (Protocol S1). Pathogen biofilm formation in the presence of probiotic candidates was tested under conditions of exclusion, competition, and displacement using a modified attachment assay [25]. The sensitivity of the assay was increased by extending the incubation time from 1 h to the time of maximum biofilm formation by the reporter strain (DY05[GFP]) as determined by the CV assay (t = 48 h and t = 72 h).
For each test, MB was inoculated with PBS suspensions of bacterial strains (each at initial concentration 1610 7 CFU mL 21 ) in Optical bottom Nunc TM (NUN165305) microwell plates (final well volume 200 mL). In all cases the time point for addition of DY05[GFP] was set as t = 0 and plates were incubated statically at 28uC until t = 48 h or t = 72 h. Specifically, for the exclusion assay, half volume MB was inoculated with probiotic candidates and incubated for 24 h prior to inoculation with DY05[GFP] (t = 0 h) and fresh MB. For the competition assay, MB was inoculated simultaneously with DY05[GFP] and probiotic candidates (t = 0 h). For the displacement assay, half volume MB was inoculated with DY05[GFP] (t = 0 h) and incubated for 24 h before inoculation with candidate probionts and fresh MB.
Separate plates were used for each interaction and incubation period and treatments were carried out in hextuplicate well sets. Plates were sealed and treated as described for microgrowth coculture assay. After incubation, wells were washed 36 in 200 mL PBS to remove planktonic and nonadherent cells. To prevent dehydration, washed wells were loaded with PBS (200 mL). Biofilm attachment was measured as a function of fluorescence and fluorescence values adjusted as described above. Candidates which caused significant signal increase (Student t-test, p,0.05) at any time point were removed from the candidate pool as these were considered pathogen biofilm facilitators. From the remaining observations, the average signal decrease was calculated from the two time points and used to classify the antagonistic activity of each strain as described above.

In vivo Protection against V. owensii DY05 Infection
Selected candidates were tested for inherent pathogenicity towards cultured stage 1 P. ornatus phyllosomas using an infection model described previously [7]. Briefly, phyllosomas were exposed to bacterial strains using live Artemia stage II (nauplii) as vectors. Formalin-disinfected nauplii were enriched with probiotic candidates through filter-feeding in tissue culture flasks (Sarstedt) for 2 h (initial concentration 1610 7 CFU mL 21 ). Positive pathogen control nauplii and negative control nauplii were treated similarly, except 1610 6 CFU mL 21 V. owensii DY05 or no bacteria were added, respectively. Apparently healthy P. ornatus phyllosomas, as assessed by photopositive response and active motility, were sourced from the AIMS larviculture facility [3], distributed to 12-well cell culture plates (1 larva well 21 ) and fed live enriched or non-enriched nauplii (t = 0 h). All treatments were performed in quintuplicate (n = 60) and survival was assessed every 24 h for 5 days.
The protective benefit of selected candidates was evaluated in two separate initial experiments using the same probiotic administration strategy (Strategy 1). In brief, stage 1 P. ornatus phyllosomas were fed nauplii enriched with probiotic candidates separately, or in combination (t = 0 h; each candidate at 1610 7 CFU mL 21 ). After 24 h, phyllosomas were challenged with nauplii enriched with 1610 6 CFU mL 21 V. owensii DY05 for 6 h, after which phyllosomas were transferred to new cell culture plates and again fed nauplii enriched with probiotic candidates (t = 30 h). Subsequently, in a third experiment the most promising candidates were tested using an alternative administration strategy (Strategy 2) which differed from strategy 1 by enriching nauplii with the pathogen in combination with probiotic candidates at t = 24 h (pathogen at 1610 6 CFU mL 21 ; each probiont at 1610 7 CFU mL 21 ). The most promising treatment from the latter experiment was replicated twice to validate observations. All experiments included a negative control (non-enriched nauplii) and a positive pathogen control (nauplii enriched in V. owensii DY05 at t = 24 h; non-enriched nauplii at t = 0 and t = 30 h). All treatments were performed in quintuplicate (n = 60) and survival was assessed every 24 h for 5 days.

Statistical Analysis
Differences between survival curves in the experimental infection models were determined using the product-limit (Kaplan-Meier) estimator and confirmed with an ANOVA. A post hoc Dunnett's test was used to compare treatments to the defined control groups. All statistical analyses were performed using the statistical software package JMPH7 (SAS) standardised at significance level a = 0.05.
To select antagonistic strains likely capable of interaction with phyllosoma hosts, representatives were shortlisted based on their phylogenetic identity, the strength of their in vitro antagonism in WDAA, and their environmental origin, with preferences given towards natural prey-derived isolates and bacteria associated with P. ornatus phyllosomas or their environments (natural or artificial). Strains that were closely related to known human pathogens or were difficult to keep in pure culture were excluded from further analysis. In this way, the pool of probiotic candidates was reduced to 16 isolates for further in vitro screening (Table S1).

Co-culture Assay
Probiotic candidates were co-cultured with V. owensii DY05[GFP] at three different starting concentrations (Figure 1). At equal starting concentrations (1610 3 CFU mL 21 ), the probiotic candidates had little or no inhibitory effect on pathogen growth, except for the three pigmented Pseudoalteromonas strains (EPP07, K25 and PP107), which demonstrated moderate to strong activity. At higher initial concentration (1610 5 CFU mL 21 ), the candidates generally had increased inhibitory effect, with strong inhibition of pathogen growth observed for the pigmented Pseudoalteromonas strains (EPP07, K25 and PP107) and three of the four Vibrio strains (C013, Ma31, and PP05). At the highest initial concentration (1610 7 CFU mL 21 ), all candidates belonging to the Pseudoalteromonas and Vibrio genera, and Ruegeria strain K2 caused strong growth inhibition, in some instances resulting in total elimination of fluorescent signals. Irrespective of initial concentration, the Bacteroidetes strains (AH26 and PPM04) had minor impact on pathogen growth during co-culture.

Monostrain Biofilm Production
A microwell crystal violet assay was used to study monostrain biofilm production ( Figure 2). Attachment of V. owensii DY05 and DY05[GFP] was detected after 24 h but maximum biofilm density was achieved between 48 and 72 h prior to dispersal. Pigmented Pseudoalteromonas strains (EPP07, K25 and PP107) rapidly formed dense biofilms which dispersed after 48 h. Non-pigmented Pseudoalteromonas strains (EPP11, PP86, and PP87) were strong biofilm formers reaching maximum density between 36 and 48 h. Strain PP81 differed from other non-pigmented pseudoalteromonads by rapidly forming and maintaining biofilm density for 48 h until sloughing. Vibrio strains generally produced low density biofilms, with exception of Ma31 which formed a dense and stable biofilm after 12 h. The Ruegeria isolates (AH10, K2, and EPP04) were strong and stable biofilm formers. Bacteroidetes strain PPM04 rapidly formed and maintained a dense biofilm between 12 and 36 h before dispersal, while Bacteroidetes strain AH26 was a weak biofilm producer.

Multistrain Biofilm Assays
A multistrain biofilm assay was used to investigate pathogen biofilm formation in the presence of probiotic candidates under conditions of exclusion, competition and displacement ( Figure 3).
All non-pigmented Pseudoalteromonas strains (EPP11, PP81, PP86, and PP87), the pigmented Pseudoalteromonas strain EPP07, and Vibrio strain C013 caused significant increases (Student's t test p,0.05) in fluorescence signals in at least one multistrain interaction, and hence were regarded as biofilm facilitators and eliminated from the candidate pool. In terms of percentage signal reduction, the remaining 10 candidates were more successful at inhibiting pathogen biofilm by exclusion (52-96%; Figure 3a), followed by competition (26-90%; Figure 3b) and displacement (0-37%; Figure 3c). Under conditions of exclusion, the pigmented Pseudoalteromonas strains K25 and PP107, Vibrio strains Ma31 and PP05, and all Ruegeria isolates (AH10, K2 and EPP04) demonstrated strong inhibition. In the competition assay, strong inhibitory activity was shown by two pigmented Pseudoalteromonas strains (K25 and PP107), Vibrio strain PP05 and the Bacteroidetes strains (AH26 and PPM04), and weakest activity exhibited by the Ruegeria candidates AH10, K2 and EPP04. During conditions of displacement, the strongest inhibitory activity was exhibited by two Ruegeria candidates (AH10 and K2) and pigmented Pseudoalteromonas strain K25.

Pathogenicity Testing and in vivo Protective Benefit
The selection of candidates showing strong antagonistic activity under all tested conditions was regarded as the most promising strategy for successful inhibition of the pathogen during in vivo conditions. Based on in vitro screening results the strains Vibrio sp. PP05, Pseudoalteromonas sp. PP107, and Pseudoalteromonas sp. K25 were selected based on their overall superior performance in the well diffusion, microgrowth and biofilm assays. In addition, the best performing Roseobacter clade isolate (Ruegeria sp. K2) was included since many strains belonging to this group of bacteria have elicited promising probiotic effects for other aquaculture species.  An initial experiment tested if the probiotic candidates themselves were pathogenic to phyllosomas. Vectored challenge with PP05, PP107, K25 or K2 (Dunnett's test p.0.05) did not alter survival of P. ornatus phyllosomas (stage 1) relative to the unchallenged control (Figure 4a) and no anomalous phototactic responses or swimming behaviours were noted, indicating the candidates were non-pathogenic. In comparison, vectored challenge with V. owensii DY05 (pathogen control) caused a significant increase (Dunnett's test p,0.0001) in phyllosoma mortality (Figure 4a).
The in vivo protective potential of candidates was tested in an experimental setup where probiotic candidates were delivered to phyllosomas via Artemia nauplii before (t = 0 h) and after (t = 30 h) vectored challenge with the pathogen V. owensii DY05 for 6 h (t = 24-30 h) (Strategy 1). Phyllosoma survival was significantly enhanced for candidates PP05 and PP107 (Dunnett's test p,0.05) compared to the pathogen control (Figure 4b). In contrast, treatment with K25, K2 or a mixture of the four candidates (Dunnett's test p.0.05) did not significantly enhance survival relative to the pathogen control (Figure 4b). Based on these results, PP05 and PP107 were selected as the most promising candidates and their protective benefit singularly or in combination was further investigated. In the second experiment, survival of PP05 or PP107-treated phyllosomas did not significantly differ from the pathogen control (Dunnett's test p.0.05) (Figure 4c). However, used in combination the probiotic candidates resulted in a significant (Dunnett's test p,0.01) benefit, enhancing phyllosoma survival by 30% (Figure 4c).
When the administration strategy was altered so that probionts were present also during the pathogen challenge (t = 24-30 h) (Strategy 2), phyllosoma survival was enhanced by 23% for PP107 (Dunnett's test p,0.01), by 42% for PP05 (Dunnett's test p,0.0001), and by 53% for PP05/PP107 in combination (Dunnett's test p,0.0001) relative to the pathogen control (Figure 4d). The experiment was repeated twice for PP05/ PP107 in combination to validate observations (Figure 4e-f), and phyllosoma survival was significantly enhanced by 80% (Dunnett's test p,0.0001) and 75% (Dunnett's test p,0.0001) respectively, compared to the pathogen control (Figure 4e-f). It should be noted that survival of PP05/PP107-treated phyllosomas did not differ significantly (Dunnett's test p.0.05) from non-challenged control phyllosomas in each of the three replicated experiments when nauplii were enriched simultaneously with pathogen and probionts (Figure 4d-f).

Discussion
The present study has demonstrated that antagonistic bacteria recovered from natural prey items of P. ornatus phyllosomas were capable of protecting cultured phyllosomas from the serious hatchery pathogen V. owensii DY05. The used probiotic screening strategy targeted antagonistic activity by candidate strains in both planktonic and attached forms, resulting in the selection of a two strain combination (Vibrio sp. PP05 and Pseudoalteromonas sp. PP107) that conferred a substantial additive survival benefit to pathogenchallenged phyllosomas.
Antagonism is a widespread trait implicated in the competiveness and ecological success of many marine bacteria [26][27][28][29] and is thus considered an important attribute of aquaculture probionts. In the present study, the antagonistic bacteria most readily culturable from wild phyllosomas and zooplankton belonged to Pseudoalteromonas and Vibrio. Both genera are frequently recovered from the marine environment and aquaculture systems [18,[27][28][29][30] and are commonly associated with eukaryotic hosts [31,32]. It is not known which antagonistic mechanisms were used by strains tested in the current study, however broad-spectrum anionic proteins and non-proteinaceous antibiotics produced by Pseudoalteromonas spp. [33,34], and aliphatic hydroxyl ethers and andrimid antibiotics [30,35] synthesised by Vibrio spp. are implicated in inhibition of aquatic vibrios.  Inhibition of planktonic V. owensii DY05 depended on both initial concentration and taxonomic grouping of the candidates. The results support previous studies showing that antagonists are generally required at higher concentrations than the pathogen for elimination [36][37][38]. Planktonic forms of all Vibrio and Pseudoalteromonas strains and the Ruegeria strain K2 strongly inhibited pathogen growth at the highest inoculum concentration. In contrast, the other strains (eg. Bacteroidetes and the other Ruegeria candidates) showed only moderate or low inhibition of planktonic pathogen growth despite showing antagonistic activity in well diffusion assays. This observation is consistent with several studies suggesting that free-living forms of marine bacteria may be less prone to producing antibacterials [26,29]. Moreover, some compounds may only be bioactive during certain interactions. For example, Dheilly et al. [39] found anti-biofilm exoproducts of Pseudoalteromonas sp. 3J6 had no antibacterial properties against free-living Paracoccus and Vibrio strains.
Overall, the strongest biofilm inhibitory activity was seen in the exclusion assay, followed by the competition and displacement assays, respectively. This was probably due to the ability of bacteria such as Pseudoalteromonas and Ruegeria to rapidly form biofilms on the microwell surface and synthesise compounds with antifouling and antibacterial activity [20,40,41] that resist incoming pathogen propagules. The pigmented Pseudoalteromonas strains were the strongest inhibitors of pathogen attachment and pigmentation is known to be linked to production of bioactive molecules in this genus [40,42]. Some Vibrio candidates (PP05 and PP25) were poor biofilm formers on the microwell surface yet were among the strongest inhibitors of pathogen attachment, suggesting inhibition was probably more related to the potency of the secreted compound rather than the biofilm biomass. Attached Bacteroidetes strains (AH26 and PPM04) were more successful than their planktonic conspecifics in outcompeting the pathogen, inferring an ecological preference for surface attachment and supporting a growing body of evidence that attached forms are more likely to exhibit antibacterial activity [29]. Reduced ability of probiotic candidates to displace pathogen biofilms could partly be related to the biofilm exopolymeric matrix trapping or slowing diffusion of antimicrobial compounds, leading to increased resistance [43,44]. Biofilms may also tolerate antimicrobials through changes in genotypic pathways, including upregulation of genes encoding efflux pumps which facilitate the efflux of antimicrobials [45].
An experimental phyllosoma infection model was used to investigate if treatment with probiotic candidates could prevent or interrupt the infection cycle of V. owensii DY05 in P. ornatus phyllosoma. A pathogen exposure time of 6 h was selected, as previous studies using the phyllosoma infection model showed that V. owensii DY05 cells have entered the hepatopancreas at this time point [7]. However, as opportunistic carnivores [46,47] some phyllosomas had not consumed all Artemia nauplii after 6 h, likely contributing to increased standard deviations relative to the robust and reproducible survival data produced in our previous study [7]. This can in part explain the discrepancies between the first two protection experiments (Figure 4b-c), where a significant protective benefit was recorded for separate treatments with Vibrio sp. PP05 or Pseudoalteromonas sp. PP107 in the first but not the second experiment.
An altered delivery strategy where the pathogen was added in combination with probiotic candidates during vectored transmission dramatically increased the protective benefit towards phyllosomas. Importantly, survival of phyllosomas receiving Vibrio sp. PP05 and Pseudoalteromonas sp. PP107 in combination did not differ significantly from non-treated controls across three replicated experiments (Figure 4d-f), and a reproducible survival enhancement (53-80%) was seen relative to pathogen controls. Pseudoalteromonas and Vibrio species have previously shown good potential as probiotics by enhancing survival of cultured invertebrates [33,48,49] and fish [50,51] following challenge with pathogenic vibrios. However, horizontal gene transfer has contributed significantly to the evolution and dissemination of virulence genes in Vibrio genomes [52], so a certain amount of risk is involved in the selection of Vibrio probiotic strains. We consider the risk to be acceptable given the lack of evidence of probiotic vibrios acquiring virulence traits and the dramatically increased protective effect on phyllosomas when Vibrio sp. PP05 was included in the probiotic mixture. The possibility of transfer of virulence traits to PP05 does however exist, and this will have to be considered if disease outbreaks persist or re-emerge.
Multispecies probiotic applications have shown clear advantages over monospecies formulations in improving pathogen resistance also in previous studies [53]. At this stage it is not clear which mechanisms are responsible for the additive probiotic effects of PP05 and PP107 and this warrants further investigation.

Conclusions
Vibrio sp. PP05 and Pseudoalteromonas sp. PP107 were prospected from a large pool of antagonistic candidates based on their ability to inhibit planktonic and attached forms of pathogenic V. owensii DY05. Used in combination, these bacteria significantly and reproducibly protected P. ornatus phyllosomas from experimental infection with V. owensii DY05. Thus, the use of miniaturised coculture and biofilm assays enabled rapid processing of numerous candidates and selection of probiotic bacteria capable of promoting survival. The study showed that natural microbial communities of wild phyllosomas support antagonistic bacteria capable of suppressing pathogens originating from the larviculture ecosystem and affirmed natural prey items as reservoirs of beneficial microorganisms. Figure S1 Correlation between fluorescence and concentration (CFU mL 21 ) of V. owensii DY05[GFP] during 24 h monoculture growth. (DOCX) Protocol S1 Monostrain biofilm production assay. (DOCX)