COCULTURE OF PROBIOTIC BACTERIA IN ALGAL FEEDSTOCKS FOR DISEASE MANAGEMENT IN BIVALVE HATCHERIES

................................................................................................................. ii ACKNOWLEDGEMENTS ....................................................................................... iii PREFACE .................................................................................................................... iv TABLE OF CONTENTS ............................................................................................. v MANUSCRIPT ............................................................................................................. 1 REFERENCES ............................................................................................................ 34 APPENDICES ............................................................................................................. 40


Abstract
Bivalve hatcheries include microalgae culture operations as a food source for stock, and these algae cultures harbor dynamic bacterial communities. Because algae from these cultures and their commensal microbiota are distributed to stock larval tanks daily upon feeding, the presence of pathogenic Vibrio spp. in hatchery microalgae cultures is a threat to stock health and survival. This study investigates the algal/bacterial and bacterial/bacterial interaction between four popular species of microalgae feedstock, a Vibrio sp. of known pathogenicity to bivalves (V. coralliilyticus RE22 [RE22]), a probiotic marine bacterium with demonstrated effectiveness in reducing larval shellfish mortality in culture operations (Phaeobacter inhibens S4 [S4]), and bacteria strains isolated directly from these algal cultures. Algal growth was unaffected by the addition of probiotics and/or RE22. Results showed that RE22 and S4 have different abilities to grow and persist in coculture, and that the four microalgae species studied have speciesspecific effects on the levels of RE22 and S4. For example, S4 titers were unaffected by coculture with the microalga Pavlova pinguis, yet significantly decreased more than 3-log10 when cocultured with Tisochrysis lutea. An antibiotic knockdown experiment and a zone of inhibition assay with commensal isolates suggest that RE22 and S4 do interact with commensal bacteria. Probiotics cocultured with algae and RE22 did not increase the effect of the algae on RE22 levels, and probiotics did not reach high enough levels in coculture to contribute to stock health via routine algae feeding protocols. For these reasons, benefits of supplementing probiotic to microalgal cultures cannot be claimed without further research and development. This research enhances our understanding of algal/bacterial interaction in shellfish hatcheries and informs methods of probiotic delivery to larvae in bivalve hatcheries.

Introduction
Bivalve aquaculture is a major industry -in 2015, global production reached 16 million metric tons, with a value of over 17 billion USD (Food and Agriculture Organization of the United Nations, 2017). Domestic production of oysters, clams and mussels, the three primary groups of bivalves farmed in the United States, was valued at close to $300 million in 2015 (NOAA Fisheries, 2017). While many different farming methods exist for bivalve culture, the industry relies heavily on hatchery production of seed, as it offers many advantages over the alternative of wild seed collection (Marshall et al., 2010).
These benefits notwithstanding, hatcheries still represent the most volatile stage of bivalve production, as disease outbreaks are common and frequently devastating. Of particular concern are pathogenic bacteria of the genus Vibrio -among bacterial pathogens affecting reared larvae of bivalves, vibrios are considered to be the most significant, causing documented hatchery losses of larval stock of up to 59% (Elston et al., 2008;Rojas et al., 2016). The genus Vibrio contains many species that infect and cause disease in a high diversity of aquatic life (Chatterjee and Haldar, 2012), but certain species, such as V. alginolyticus, V. coralliilyticus, and V. tubiashii, cause the most mortalities in cultured larval bivalves (Beaz-Hidalgo et al., 2010;Dubert et al., 2017).
Currently, bivalve hatcheries rely on water treatment, biosecurity measures and antibiotics to prevent the introduction and/or spread of disease (Dubert et al., 2017;Shumway, 2011). While the first two strategies are largely prophylactic and without negative health, economic, or environmental side effects, the same cannot be said of antibiotics, for several reasons: agricultural usage of antibiotics is widely recognized as a human health concern due to the development and spread of bacterial resistance to the drugs, the detection of antibiotic residues in aquacultural products is alarming to consumers, and antibiotics inherently disrupt the natural bacterial community in rearing tanks and on the animals themselves, leading to a potential increase in disease susceptibility Schmidt et al., 2017). For these reasons, alternative management strategies are in need.
Probiotics represent one such alternative. Defined as "Live, nonpathogenic microorganisms which, when administered in adequate amounts, confer a health benefit on the host" (World Health Organization and Food and Agriculture Organization, 2006), probiotics have already been demonstrated as effective in mitigating disease in aquaculture (de Azevedo and Braga, 2012). Phaeobacter inhibens S4 (S4), a member of the Roseobacter clade of alpha-proteobacteria, is one such microorganism: hatchery-scale experiments have proven this probiotic to be effective at reducing larval oyster mortalities following a challenge with Vibrio coralliilyticus RE22 (RE22), as well as other pathogens Sohn et al., 2016). In order to maintain larval protection, however, S4 must be supplied daily to stock tanks, a common requirement for realizing the benefits of probiotics Kesarcodi-Watson et al., 2008;Verschuere et al., 2000).
In addition to use in bivalve stock tanks, probiotics could be used to prevent the growth of pathogenic organisms in algal cultures. Bivalves are filter feeding organisms at all life stages, with strict nutritional requirements and filtering capabilities at each stage, and the typical bivalve hatchery grows large quantities of multiple species of microalgae as feedstock (Helm et al., 2004). Microalgal monocultures, such as those found in bivalve hatcheries, are supplied with high levels of nutrients and illumination for maximum algal production (Helm et al., 2004). Hatchery microalgal cultures contain an abundance of bacteria, with substantial variation in the bacterial community composition observed from one culturing system to another, between algae species, and over time (Nicolas et al., 2004), and these bacterial communities play a role in disease dynamics in hatcheriespathogenic vibrios are often present in the communities, and bacteria in the microalgae cultures are introduced into bivalve stock tanks upon feeding (Dubert et al., 2017;Elston et al., 2008).
Although the composition of the commensal microbial communities varies considerably from one tank to another, and over time (Nicolas et al., 2004;Wang et al., 2016), certain clades of copiotrophic bacteria, (well-adapted to high nutrient concentrations), and microalgal-symbiotic heterotrophic bacteria have been found to dominate these environments, specifically members of the Roseobacter clade and Cytophaga-Flavobacterium-Bacteroides group (Nicolas et al., 2004). It is well-established that bacteria influence the growth characteristics of microalgae, and in turn microalgae influence the bacterial community (Cooper and Smith, 2015;Fuentes et al., 2016;Kazamia et al., 2012). Studies of the physiology and ecological roles of roseobacters, and P. inhibens S4 specifically, suggest that S4 is an excellent probiotic candidate for coculture with microalgae (Geng and Belas, 2010;Rooney-Varga et al., 2005;Zhao et al., 2016). Closely-related bacteria, i.e. other members of the Roseobacter clade, have been found to consistently associate with microalgae in nature (Amin et al., 2012;Geng and Belas, 2010), as well as in culture (Nicolas et al., 2004;Sandaa et al., 2003). These associations are believed to be due at least in part to the ability of most roseobacters to metabolize an organosulfur byproduct of microalgae, dimethylsulfoniopropionate (DMSP) Cui et al., 2015); preliminary in silico data suggests that S4 does possess this capability (Cui et al., 2015; Genbank accession number NZ_LOHU00000000). Furthermore, S4 produces the antibiotic tropodithietic acid (TDA) and is an excellent biofilm former, characteristics that may provide a sufficient competitive edge for growth and persistence in microalgal coculture (Zhao et al., 2016).
As for the ability of S4 to inhibit vibrios in such a setting, one study has demonstrated that additions of P. inhibens cause a substantial drop in V. anguillarum in a non-axenic culture of the microalgae Tetraselmis suecica (Grotkjaer et al., 2016), a finding consistent with previous research on the closely-related Phaeobacter gallaeciensis and axenic microalgal cultures (D'Alvise et al., 2012).
The objectives of this research were to a) determine the abilities of the probiotic Phaeobacter inhibens S4 and the pathogen Vibrio coralliilyticus RE22 to grow and persist in microalgae cultures, and b) determine the effect of probiotic coculture on microalgal growth, abundance of V. coralliilyticus RE22 added to the cultures, and titers of commensal bacteria. If probiotics can be successfully grown in microalgae cultures, the health benefits of S4 on hatchery stock could be provided with less operational expense -not only would there be no need for daily handing and delivery of probiotic, but bivalve stock are already fed multiple times per day, and the probiotic would be delivered in conjunction with routine feedings. Additionally, cocultured probiotics may succeed in reducing or eliminating the establishment and growth of vibrios in these algae cultures, a potential cause for algal mortality and a recognized entry route of the pathogen into bivalve tanks (Elston et al., 2008).

Design of coculture experiments
Four different species of microalgae were investigated, based on their current, widespread use as feedstock in bivalve hatcheries: the diatom Chaetoceros neogracile Chaet B, the chlorophyte Tetraselmis chui PLY-429, and the haptophytes Tisochrysis lutea T-Iso and Pavlova pinguis CCMP-609 (Bendif et al., 2013;Milke et al., 2008;Napolitano et al., 1990;Wikfors et al., 1996). Microalgal culture stocks were obtained from the Aquaculture Breeding Center at the Virginia Institute of Marine Sciences.
The experimental design consisted of 100 mL algal cultures grown at 22°C with constant illumination (see details below). Predetermined quantities of bacteria and algae were introduced into sterile, fertilized seawater for all coculture and control flasks, and samples were drawn at multiple time points to track: a) the levels of algae using cell counts, and b) the levels of each bacterial strain (probiotic, pathogen, and total bacterial counts) in the cultures using selective media. The treatments tested were: algae (all four strains, see Table 1) co-incubated with a) probiotic S4; b) pathogen RE22; and c) both S4 and RE22. Controls included each algae species alone, S4 only, RE22 only, and S4 and RE22. Each treatment was performed in triplicate and each experiment was performed at least twice.

Algal cultures
Culture flasks were soaked for 24 hours in Citranox acid detergent (source) and scrubbed thoroughly prior to experiments to remove biofilms. All materials were autoclaved for sterility, except for those that cannot withstand the autoclave: these were bleachsterilized, rinsed with 0.22 μm filter-sterilized sodium thiosulfate (Na2S2O3), and then rinsed again thoroughly with sterile type-2 deionized water. Instant Ocean® artificial seawater was autoclaved and f/2 fertilizer (Guillard, 1975) and sodium metasilicate (for diatom cultures) were 0.22 μm filter-sterilized. Microalgal cultures were grown up to mid-exponential phase in 1 L flasks containing sterile seawater fertilized with f/2 formula (Fritz Industries, complete algal growth media), with sodium metasilicate (40 mg/L) added to C. neogracile cultures, as diatoms require supplemental silicate for growth (Andersen, 2005). Flasks were lightly aerated with 0.3 μm filtered air. Just prior to the beginning of each experiment, a sample of algal culture (200 μL) was taken, fixed with 10% Lugol's iodine solution, and counted with an improved Neubauer counting chamber on a Nikon Eclipse 50i compound microscope using the 40X objective (400X total magnification) and brightfield condenser setting. This measurement of cell density was used to calculate the volume needed for each flask to achieve the target starting density, typically ~10 5 cells/mL.

Bacterial cultures
The bacterial strains used in this experiment were selected for antibiotic resistance to facilitate enumeration of each strain from coculture samples using selective media (Table   1). Both strains are resistant to the antibiotic streptomycin (Sm) by spontaneous mutation from the parent strains S4 and RE22. S4 has been engineered for chloramphenicol (Cm) resistance as well; the plasmid pRhokHi-2-OFP was transferred from E. coli Sm10 by conjugation, and the resulting strain named WZ02 (Zhao et al., 2016). Bacteria were prepared from glycerol freezer stocks and grown up in broth culture containing the antibiotics to which each strain is resistant. Briefly, bacterial freezer stocks were streaked on mYP30 media plates (30 g Instant Ocean salt, 1 g yeast extract, 5 g peptone, filled to 1000 mL type-2 deionized water, pH of 7.5-7.6, and 16 g Bacto agar)

Coculture of bacterial strains with microalgae
Microalgal cultures were established by adding a volume of algae from stock cultures to achieve a starting concentration of 10 5 cells/ml for C. neogracile, P. pinguis and Ti. lutea, and 10 4 cells/ml for Te. chui, to 100 mL of artificial, autoclaved seawater (28 psu), fertilized with f/2 nutrient formula in 125 mL Erlenmeyer flasks (Fig. 1). Bacteria (S4 and RE22) were added from stock cultures to achieve a starting density of 10 5 CFU/mL.
S4 was added at the beginning of each experiment, while RE22 was added after 24 hours, since prior research has shown that pre-colonization with S4 inhibits RE22 growth more effectively than simultaneous coculture (Zhao et al., 2016) (Fig. 2). Algal-bacteria cocultures were grown at 22°C under constant illumination (two 2600-lumen T12 bulbs at a color temperature of 4100K) and aerated lightly with 0.3 μm filtered air.
Experiments were run for 10-18 days without media exchange, in order to capture any interactions or growth changes that occur over all algal growth stages. One 200 μL sample was taken from each flask at each timepoint (typically day 1, 2, 4, 7, and 10) using sterile technique in a laminar flow hood.

Algal cell counts
One aliquot (90 μL) was taken from each sample obtained from flasks containing algae, fixed with 10 μL Lugol's iodine, and the microalgal cell density was then determined with a Neubauer improved counting chamber on a Nikon Eclipse 50i compound microscope using the 40X objective (400X total magnification) and brightfield condenser setting. Results are expressed in cells per mL of culture.

Fig. 2.
Introductions of algae and S4 to experimental flasks took place at the beginning of each experiment, while RE22 was added after 24 hours. Sampling was typically conducted on day 1, 2, 4, 7 and 11.

Bacterial cell counts
Planktonic bacterial density (CFU/mL) was determined via serial dilutions and platecounts of samples, using the spot-plating method as described by Miles & Misra, 1938; since 10 μL was selected for the spotting volume, the lower detection limit for all bacteria in these experiments is 10 2 CFU/mL. Total culturable bacteria were enumerated by growth on mYP30 agar plates, while the two bacterial strains of interest, S4 and RE22, were each enumerated via selective agar media plates: mYP30 + Sm (200 μg/mL) + Cm (5 μg/mL) was used to enumerate S4 colonies, and thiosulfate-citrate-bile salts-sucrose agar (TCBS; Difco) + Sm (200 μg/mL) was used to enumerate RE22. Cycloheximide (40 μg/mL) was also added to the S4-selective media plates to prevent the growth of microalgae on the agar plates.

Effect of knockdown of the commensal bacteria in algal cultures on S4 cell counts
A coculture experiment was designed to knock down the commensal bacteria population, in order to determine whether commensal bacteria may affect the growth of S4 when coincubated with algae. To this end, antibiotics (streptomycin, 200 μg/mL, and chloramphenicol, 5 μg/mL) were added directly to treatment flasks at the beginning of the experiment, and probiotic (S4) growth curves were compared with control cocultures to which no antibiotics were added.

Zone of inhibition assay
In an effort to determine if commensal bacteria associated with C. neogracile, P. pinguis, and Te. chui could be affecting levels of S4 or RE22 in microalgal coculture via direct inhibition (Ti. lutea was not available at the time of the assay), culturable bacteria were isolated from algae cultures and a zone of inhibition assay was conducted to identify isolates with inhibitory effects on S4 and RE22. The assay was conducted as described by Karim et al. (2013) with minor changes. Briefly, samples were collected from each of the three algae cultures, serially-diluted, and 10 μL volumes of each dilution were then spotted on mYP30 media and incubated for 72 h. All morphologically-distinct colonies from each algae sample were picked and grown up overnight in mYP30 broth at 150 rpm and 28°C; 5 μL of each of these overnight cultures was then spotted onto an RE22 lawn prepared on mYP30 media, and allowed to incubate at 28°C for 48h. S4 was selected as a positive control because it has known inhibitory effects against RE22 . Following incubation, zones of inhibition were identified and measured: the values reported are the average of two measurements (perpendicular to each other) of the difference in diameter between the zone of inhibited growth of RE22 and the colony size of each isolate. Results are expressed in mm.

Microscopy
Since another P. inhibens strain has been found to attach to microalgae cells (Bramucci et al., 2018;Mayers et al., 2016), this study sought to determine if any such interactions occur between P. inhibens S4 and the microalgae species investigated. To this end, P.
inhibens strain WZ02 (Table 1) was selected for its constitutive expression of orange fluorescent protein (OFP), which can be visualized against a large background bacterial community via epifluorescent microscopy (Zhao et al., 2016). Samples were taken on days 5 and 9 of each coculture experiment except for Ti. lutea, and visualized without fixation within one hour from sample collection on a Zeiss Axioimager M2 Imaging System, utilizing ZEN 2011 software; micrographs were captured with a Zeiss Axiocam HRc high resolution camera. Observations and micrographs were taken with a 10X, 40X, and 100X oil immersion objective (total magnification with 100X objective = 1000X) and the following settings: transmitted light brightfield, phase contrast, and epifluorescent observation with a Rhodamine filter set for visualization of OFP-expressing S4.
Nonlinear regression and the extra sum-of-squares F-test were used to determine whether growth curves differed significantly between treatments in all experiments. All data were log10 transformed prior to curve-fitting. Two different models were used: (a) algae and total bacteria, which showed traditional logistic growth curves, were modeled with a logistic function: Where: N0 is the starting population NM is the maximum population and K is the rate constant.
(b) all other bacterial growth curves were modeled with a quadratic (second-degree polynomial) function: Where: B0 is a constant and B1 and B2 are coefficients to the first and second order terms, respectively.
An initial F-test was applied to all treatments in each experiment to test whether one curve adequately fits all data sets; a p-value of <0.05 was chosen to identify data sets that are better represented by more than one curve. Following the initial F-test of all treatments (except algae growth curves, none of which differed significantly), biologically-relevant pairwise comparisons were tested as well: (1)  to identify data sets better represented by two curves than one, to account for multiple comparisons.

Effect of bacterial additions on microalgal growth
This study sought to determine the effect of S4 and/or RE22 additions on microalgal growth. Growth curves of all four microalgae species were unaffected by bacterial (P. inhibens S4 and/or V. coralliilyticus RE22) additions ( Fig. 3; global F-test p-values for all four species ranged from 0.828 to 0.9981; Supplementary Table S1).

Dynamics of pathogen RE22 in coculture with microalgal strains
This study sought to determine the dynamics of the pathogen RE22 in coculture with microalgae. The levels of RE22 in fertilized seawater increased from 10 5 to 10 6 CFU/mL during the first day of incubation, and then levels were maintained at 10 6 CFU/mL for the length of the experiment (Fig. 4). As previously reported (Zhao et al., 2016), the    Table S2). No significant differences were found between growth curves of RE22 co-incubated with algae alone vs. RE22 coincubated with algae and S4 (p-values ranged from 0.2507 to 0.8983) (Fig. 4, Appendix Table S2).

Dynamics of probiotic S4 in coculture with microalgal strains
This study sought to determine the dynamics of the probiotic S4 in coculture with microalgae. S4 showed a decline of 0.5-1 log10 of planktonic CFU/mL in the first 24h of incubation in fertilized seawater (control), followed by a more gradual, steady decline to 2x10 4 -2x10 5 CFU/mL after 12 days (Fig. 5). The addition of RE22 (final concentration  Table S3). The effect of coculture with algae on S4 CFU/mL was found to be species-specific: no significant difference was observed for C. neogracile and P. pinguis, while coculture with Te. chui and Ti. lutea caused a significant decline in S4 CFU/ml beginning in the first 1-2 days of each experiment (p<0.0001) (Fig. 5C, 5D, Appendix Table S3). The extent of this decline differed between the two species: coculture with Te.
chui caused an average 1 log10 decline in S4 levels over the course of the experiment, decreasing to a final level of 2x10 4 CFU/mL after 12 days (Fig. 5C), while coculture with Ti. lutea caused a much more substantial decline in S4 levels, dropping to 3x10 2 CFU/mL ( Fig. 5D). Additionally, Te. chui was the only algae species for which RE22 addition significantly affected S4 levels in coculture, causing an average 0.5 log10 increase in S4 CFU/mL in the final two sampling timepoints (days 8 and 12; p=0.0154; Fig. 5C).

Effect of probiotic and pathogen additions on commensal bacterial density
This study sought to determine if S4 and/or RE22 additions affected total culturable bacterial density in microalgae cultures. Total culturable bacterial density in cultures of all four microalgae species was unaffected by additions of S4 and RE22 ( Fig. 6; global Ftest p-values for all four species ranged from 0.2726 to 0.8508).

Dynamics of S4 in coculture with antibiotic-treated microalgae
This study sought to determine if knockdown of the commensal bacteria in algae cultures affects the dynamics of S4 in coculture with algae. Addition of streptomycin (200 μg/mL) and chloramphenicol (5 μg/mL) to C. neogracile cultures did not eliminate the commensal bacterial population, but did reduce total culturable bacteria by 2.5 log10 on day 4 and 1.5 log10 on day 8 (compared to control cultures to which no antibiotics were added) (Appendix, Fig. S2). Antibiotic additions had minimal effect on S4 density on day 4, but on day 8 S4 was significantly higher in coculture with algae to which antibiotics were added, compared to both controls (S4 only with antibiotics, and S4 and algae without antibiotics; p=0.0057 and 0.0178, respectively [Appendix, table S4]) (Fig.   7). An increase in planktonic S4 CFU density of ~0.5 log10 was observed compared to the S4-only with antibiotics control, and an increase of ~1 log10 was observed compared   to the S4 and algae without antibiotics control. The antibiotics had no effect on algae cell density, based on a preliminary trial (Appendix , Table S6).

Isolation of commensal bacteria able to inhibit the growth of RE22 in vitro
In an effort to determine if commensal bacteria could be causing the observed decline of RE22 (and to a lesser extent, S4) in microalgal coculture, isolates cultured from algal cultures were tested for inhibitory activity against RE22 and S4 in a zone of inhibition assay. A total of 40 bacterial isolates were tested: 14 from C. neogracile, 15 from P.
pinguis, and 11 from Te. chui (Ti. lutea was not available at the time of the assay). None of the 40 bacterial isolates inhibited S4 in a zone of inhibition assay. Seven of these isolates inhibited RE22 to varying degrees (

Association of S4 with algal cells
Fluorescent imaging conducted on samples of each treatment throughout all experiments (except for Ti. lutea, which was not investigated) revealed no physical associations between S4 and microalgae cells of any of the three species (C. neogracile, P. pinguis and Te. chui). Therefore, the methods used here for measuring bacterial levels in media mainly reflect planktonic cells. S4 was observed to exist planktonically as both single cells and in rosette structures (multicellular aggregations) (Bruhn et al., 2007) (Fig. 8) on both sampling timepoints (d 5 and 9). Qualitative observations did not identify any variation in the proportion of S4 existing as single cells vs. rosettes across timepoints, treatments, or co-incubated algae species.

Discussion
This study explores the incorporation of probiotics into algae cultures in bivalve hatcheries for improved disease management in bivalve hatcheries. Results revealed no effect of probiotic or pathogen coculture on the growth of microalgae, reaffirming a previous report on RE22 -algal interactions (Elston et al., 2008). Coculture of bacteria with microalgae revealed species-specific effects on both RE22 and S4 dynamics in algal coculture, and these dynamics differed between the two bacterial strains, particularly when cocultured with P. pinguis and Te. chui. Results from the antibiotic knockdown experiment and zone of inhibition assay with commensal isolates suggest that RE22 and S4 interact with commensal bacteria in algal coculture, but direct inhibition of RE22 and S4 by commensal bacteria is unlikely to be the only driver of their observed dynamics.
While S4 inhibits RE22 in vitro , Zhao et al. 2016, the same effect was not observed when these two species were cocultured with algae, probably due to the effect of the algal species on RE22. This research sheds new light on microbial-algal interactions and provides practical applications for bivalve hatcheries.
RE22 maintained a titer in media alone at a density of 10 6 CFU/ml. Previous work has identified Vibrio spp. titers of 10 4 -10 5 CFU/mL in surface seawater (Elston et al., 2008) and titers of up to 2-3x10 9 CFU/ml in nutrient-rich growth media under controlled laboratory conditions (Zhao et al., 2016). When added to media pre-colonized by S4, RE22 titers dropped approximately 1 log10, exhibiting the probiotic effect in vitro and corroborating prior observations (Zhao et al., 2016).
S4 also maintained a high titer in media alone, and declined gradually. Research has found roseobacters can be abundant in natural seawater, especially among algal blooms: the relative abundance of roseobacters among the marine bacterial community can reach 30% (Buchan et al., 2005). S4 dynamics in media alone were similar to RE22, except titers of S4 were slightly lower, and decreased slightly over time; this may be due to nutrient depletion or settlement of planktonic cells onto surfaces (Bruhn et al., 2007;Kolter et al., 1993). When cocultured with RE22, S4 titers increased significantly in 2 out of 4 experiments. Previous work did not detect such an effect (Zhao et al., 2016), although experimental conditions (e.g. media used) were different. This observed increase in some of the experiments may be due to increased nutrient availability following killing of RE22 cells by S4, resulting in increased growth; a study of nutrient limitation of S4 in this media could provide greater insight. A second explanation could be an effect of RE22 on the "swim or stick" phenotype of S4: studies have found that environmental and chemical cues control the tendency of some roseobacters to synthesize flagella and enter media as motile, single cells, or to shed their flagella and aggregate in biofilms and rosettes (Belas et al., 2009;Sule and Belas, 2013). Since the methods here only account for colony-forming units of planktonic or particle (algae)-associated cells (algae in cultures were also plated), a change in S4 phenotype towards the motile life stage would be detected as an increase in S4 titer.
In most cases, dynamics of RE22 and S4 differed between bacteria-only controls and coculture with algae; for RE22, titers declined in coculture with all four algal species, compared to the RE22-only control. C. neogracile supported and maintained the highest levels of RE22, while coculture with the other three algae species caused a steady decline in RE22 levels as time progressed. The results reported here corroborate previous research which found that Te. chui and a close relative of Ti. lutea, Isochrysis galbana, inhibit the growth of vibrios in coculture (Giménez Papiol et al., 2018;Molina-Cárdenas et al., 2014). The species-specific nature of these interactions has great relevance to bivalve aquaculture, since abundance of pathogenic vibrios in algal feedstocks is directly linked to stock exposure to these pathogens (Elston et al., 2008). Possible mechanisms of this species-specific inhibition of RE22 include algal production of antimicrobials (Austin et al., 1992) and interference in bacterial quorum sensing (Natrah et al., 2011), as well as inhibition by commensal bacteria specific to each algae culture; the antibiotic knockdown experiment and zone of inhibition assay (discussed below) were designed to probe this possibility. Furthermore, relationships between microalgae and bacteria can be dynamic: in one case, vibrios were found to proliferate in dense cultures of the microalgae Nannochloropsis oculata, but drop to undetectable levels in lower density As for S4, coculture with two of the four algal species caused a decline in S4 compared to the S4-only control, while the other two species had no effect. Furthermore, the two species that inhibited S4 did so to different degrees -Ti. lutea caused nearly a 2 log10 greater decline than Te. chui. In addition to the potential mechanisms driving bacterial dynamics in coculture as discussed above for RE22, variable production of DMSP between algae species may also contribute to the species-specific effect on S4 growth.
The common natural association between roseobacters and microalgae is believed to be due at least in part to the relatively unique capability of most roseobacters to metabolize DMSP, an organosulfur byproduct of microalgal metabolism which is abundant among dense algae growth Cui et al., 2015); preliminary in silico data of the draft genome (Genbank accession number NZ_LOHU00000000) by our team suggests that S4 does possess this capability.
Interestingly, S4 did not cause a decline in RE22 when cocultured with algae, even though the effect was observed in bacteria-only controls (as expected) (Zhao et al., 2016).
Possible explanations for this observation include: a) RE22 is already declining on its own in coculture with three out of four algal species, obscuring the effect of S4 (C. neogracile being the exception, which cannot be explained by RE22-algal dynamics); b) titers of S4 are lower in algal coculture than the bacteria-only controls, so a lesser inhibitory effect (if any) is to be expected ; and/or c) the environmental changes (both physical and chemical) caused by the presence and growth of microalgae alters the phenotype of S4 in such a way that it does not inhibit RE22, such as decreased production of TDA (Bruhn et al., 2007). Although no effect was observed here, it has been demonstrated that additions of P. inhibens DSM17395 cause a substantial drop in V. anguillarum in a non-axenic culture of the microalgae Tetraselmis suecica (Grotkjaer et al., 2016), a finding that was further supported by research on the closely-related P. gallaeciensis and axenic microalgal cultures (D'Alvise et al., 2012).
Interrelated species effects have been observed as well -coculture with Nannochloropsis was found to enhance the antibiotic effect of a roseobacter against V. anguillarum (Sharifah and Eguchi, 2011).
Even though S4 did not affect RE22 in coculture with any of the four algal species tested, RE22 additions did result in significantly higher S4 levels when cocultured with Te. chui (not observed with the other three algae species). This increase due to RE22 addition was also observed in the bacteria-only coculture controls, and may be due to the same factors hypothesized above: increased nutrient availability following killing of RE22 cells by S4, and an effect of RE22 on the "swim or stick" phenotype of S4.
An antibiotic knockdown experiment and a zone of inhibition assay with commensal isolates were carried out in an effort to determine if (and to what degree) commensal bacteria affect the growth dynamics of RE22 and S4 in coculture, and if the observed declines of RE22 and S4 in coculture with some algal species could be attributed to direct antagonism by the commensal bacteria. The antibiotic experiment was conducted with C.
neogracile only, and revealed minor inhibition of S4 by commensal bacteria; commensal bacteria of other algal strains may affect S4 differently. By day eight, S4-only controls increased 0.5 log10 with antibiotic additions, and S4 co-incubated with algae and antibiotics increased 1 log10, compared to coculture with algae and no antibiotics, which leaves a 0.5 log10 increase that may be explained by the knocked-down bacterial community. A possible explanation for the surprising increase in planktonic S4 CFU/mL following antibiotic additions (to which it is resistant) may be due to a phenotypic change triggered by the antibiotics associated with less aggregation into rosettes or settlement onto surfaces. Although the 0.5 log10 CFU/mL increase due to the knocked-down bacterial community is slight (but statistically significant), the antibiotic treatment most probably merely disrupted the commensal bacterial community, and did not eliminate it completely. Future experiments with axenic algae cultures will shed more light on this interaction; previous work has found that competition for limited space and nutrients can be intense in microbial communities, such as those associated with microalgae cultures, and these communities can prevent the establishment of strains that would otherwise grow alone in media or with algae Wietz et al., 2013).
A zone of inhibition assay was conducted to further explore the nature of this observed interaction between S4 and commensal bacteria, and to shed light on the species-specific effect of RE22 in algal coculture. It was expected that if the assay discovered inhibitory bacterial isolates with inhibitory activity against RE22 and/or S4, they would be from the algal cultures in which RE22 and S4 showed the greatest decline. While no bacterial isolates from algal cultures inhibited S4, we were surprised to find that six out of the seven isolates that are inhibitory towards RE22 in vitro came from C. neogracile -the algae culture that inhibits RE22 growth the least (of the four studied). Furthermore, none of the bacteria isolated from P. pinguis inhibited RE22, even though RE22 dropped over 3 log10 in coculture with this species. As seen here and previously, bacterial inhibition in vitro does not always correspond to inhibition in vivo (Gram et al., 2001). Possible explanations include a) phenotypic differences between isolates grown on agar media vs.
algae culture, where differences in nutrient availability and cell density may play a role in expression of inhibitory cell products and/or mechanisms (Gram et al., 2001;Kinnula et al., 2017); b) complex interactions between inhibitory isolates and algae and/or commensal bacteria, such as competition and quorum sensing interference (Natrah et al., 2014;Wietz et al., 2013); and c) the possibility that inhibition still occurs in coculture, but the signal is lost among other factors of higher impact affecting RE22 growth or the inhibitory activity is due to commensal bacteria that do not grow on the media used for isolations. These findings taken together suggest that direct inhibition of RE22 by culturable commensal bacteria may not be the primary factor contributing to the observed decline in RE22 when cocultured with the algae species investigated here. Furthermore, the slight inhibition of S4 growth that may be attributable to interaction with commensal bacteria appears not to be due to direct (antagonistic) inhibition by culturable strains, since no S4-inhibitory isolates were discovered. Fig. 9. Summary diagram of observed effects of algal coculture on planktonic abundance of probiotic S4 and pathogen RE22. Coculture with algae inhibited growth of S4 and RE22 in most cases (red arrows); arrow thickness signifies the observed magnitude of inhibition. Results of the antibiotic knockdown experiment suggest that commensal bacteria do play a role in probiotic dynamics in algal coculture. Experiments with axenic algal cultures can shed more light on the mechanisms underlying these interactions (Fig. S4).
This research informs the practical application of probiotic usage in bivalve hatcheries.
While S4 did persist in coculture with all four algae species, albeit to varying degrees, the levels never reached the high density required for supplying this probiotic to stock tanks at an effective dose. Previous research found that a minimum S4 density of 10 4 CFU/mL is required to confer a health benefit on stock , and the dilution factor of feeding has been estimated at 1000X (Elston et al., 2008); these values taken together suggest that for microalgae cultures to serve as an effective incubator and delivery vehicle of S4 to stock tanks, the probiotic would have to grow to a density of ~10 7 CFU/mL. The highest densities attained after early timepoints were in coculture with P. pinguis and Te. chui, where S4 levels remained in the range of 10 4 -10 5 CFU/mL, 2-3 log10 below the required density. Importantly, this study has revealed that for maximum delivery of S4 to stock tanks, S4 must be mixed with algal feedstocks just prior to feeding, versus allowing a period of co-incubation, during which time S4 titers are likely to drop (differences in environmental conditions and media composition from those used here may affect S4 dynamics (Zech et al., 2013)).
Further implications of this research for bivalve hatcheries include the observed resilience of feedstock microalgae to bacterial antagonism, and the interspecific variation in their effect on the pathogen RE22. Since many bacteria have been found to parasitize or inhibit the growth of microalgae (Ramanan et al., 2016), the lack of antagonism of algae by S4 is encouraging for future applications of this or related probiotics in microalgae coculture. Regarding the species-specific effect of algae on RE22, hatchery managers seeking to minimize vibrio introductions to stock tanks are advised to closely monitor for contamination algal feedstocks that support higher levels of pathogenic vibrios, such as C. neogracile, which this study has shown to support relatively high levels of RE22.
The results presented here emphasize the complex algal/bacterial interactions that dictate the growth and/or decline of aquaculturally-important microbes in bivalve feedstock cultures. More research is needed to shed light on the drivers of these bacterial dynamics in order to develop more effective probiotic-based disease management strategies.
Ethics statement: Not applicable.

Conflict of interest:
The authors declare that they have no conflicts of interest.
Appendix A: all F-test results.

Results/Discussion
RI titers declined rapidly in both the fertilized seawater control treatment as well as in coculture with C. neogracile (Fig. S3). No significant differences in growth curves were observed between treatments (Global F-test: p=0.1105, Table S7); in all treatments, RI titers stabilized at ~10 3 CFU/mL by day 2, and declined marginally (<0.2 log10) for the remaining 6 days of the experiment. Neither RE22 titers nor total culturable bacterial density were affected by RI additions in coculture with C. neogracile (Fig. S3). Since RI declined rapidly in media alone, its effect on other algae species was not investigated.