The Roseobacter-Group Bacterium Phaeobacter as a Safe Probiotic Solution for Aquaculture

Phaeobacter inhibens has been assessed as a probiotic bacterium for application in aquaculture. Studies addressing the efficacy and safety indicate that P. inhibens maintains it antagonistic activity against pathogenic vibrios in aquaculture live cultures (live feed and fish egg/larvae), while having no or a positive effect on the host organisms and a minor impact on the host microbiomes.

ABSTRACT

Phaeobacter inhibens has been assessed as a probiotic bacterium for application in aquaculture. Studies addressing the efficacy and safety indicate that P. inhibens maintains its antagonistic activity against pathogenic vibrios in aquaculture live cultures (live feed and fish egg/larvae) while having no or a positive effect on the host organisms and a minor impact on the host microbiomes. While P. inhibens produces antibacterial and algicidal compounds, no study has so far found a virulent phenotype of P. inhibens cells against higher organisms. Additionally, an in silico search for antibiotic resistance genes using published genomes of representative strains did not raise concerns regarding the risk for antimicrobial resistance. P. inhibens occurs naturally in aquaculture systems, supporting its safe usage in this environment. In conclusion, at the current state of knowledge, P. inhibens is a “safe-to-use” organism.

INTRODUCTION

Phaeobacter inhibens is a marine alphaproteobacterium with potential for application as a probiotic bacterium in marine aquaculture systems. P. inhibens is a representative of the Roseobacter group, which is widespread and an environmentally important marine group of bacteria. P. inhibens serves as a heterotrophic marine model organism to understand fundamental processes, such as adaptation to an attached lifestyle and interactions with higher organisms (1). Its probiotic activity against fish pathogenic bacteria is primarily due to the production of the antibacterial compound tropodithietic acid (TDA) (2). TDA is bactericidal to both fish and human bacterial pathogens. The applicability of P. inhibens as a probiotic bacterium in marine aquaculture has been assessed in several studies and has recently been compared with that of other probiotics (3). The bacterium is naturally present in aquaculture systems, and it is able to inhibit fish pathogenic bacteria. Most studies have been performed on its activity against economically important Vibrio spp., such as Vibrio anguillarum and Vibrio vulnificus, but it also inhibits other fish pathogens, such as Aeromonas and Tenacibaculum spp. (J. E. Tesdorpf, A. U. Geers, M. L. Strube, L. Gram, and M. Bentzon-Tilia, unpublished data) (4–6). This activity is seen both in laboratory-based agar assays and in live feed used in marine larval rearing. The ability to inhibit fish pathogenic Vibrio has been demonstrated in both axenic and nonaxenic live feed systems (5, 7–10). Also, in model challenge trials, P. inhibens can decrease the mortality of fish larvae challenged with pathogenic Vibrio spp. (9).

Farming of fish and shellfish is of great importance to supplying protein for food and feed for the growing world population. Catches from wild fish have stagnated (or even declined) since the mid-1980s, and the increase in fish production comes almost exclusively from aquaculture (11). In intense animal rearing systems, infectious agents such as pathogenic bacteria spread rapidly, and one particularly sensitive stage is larval development. Fish larvae do not have a developed immune system, and therefore, vaccination cannot be used as a disease preventive measure (12). During larval rearing, opportunistic and pathogenic bacteria are easily introduced via live feed, and infections can spread rapidly, eradicating the complete larval batch (13). Antibiotics have been used to control these infections; however, due to the risk of bacteria developing and spreading antibiotic resistance, other measures must be found. One strategy for limiting the proliferation of bacterial pathogens in live feed and fish larvae is the use of probiotic bacteria (14). Probiotic microorganisms have been defined by the Food and Agriculture Organization of the United Nations (FAO) and the World Health Organization (WHO) (15) as “live microorganisms which when administered in adequate amounts confer a health benefit on the host.” This approach presents an environmentally sustainable and economically viable solution to counteract the economic loss caused by bacterial pathogens in aquaculture systems.

P. inhibens has been shown to antagonize many fish pathogenic bacteria, such as Vibrio spp. (5, 6). In the laboratory, resistance to the effector molecule TDA could not be selected for in pathogenic bacteria, likely because the molecule as an antiporter destabilizes the bacterial proton motive force (16, 17). Interestingly, despite this broad range effect, addition of the bacterium to live feed causes only minor changes in the microbiome of the live feed (18). A search in the whole-genome sequence of P. inhibens DSM 17395 for the presence of known antimicrobial resistance (AMR) genes relevant to antimicrobial use in humans and animals was performed, as indicated in the European Food Safety Authority (EFSA) guidance on the characterization of microorganisms used as feed additives or as production organisms (EFSA, 2018). For this purpose, a comparison against up-to-date databases (e.g., ARG-ANNOT [antibiotic resistance gene annotation], CARD [comprehensive antibiotic resistance database], and ResFinder) was performed. This article provides an overview of the genetics and physiology of P. inhibens and summarizes the current information on the safety of P. inhibens for its use as a probiotic in aquaculture.

TAXONOMIC CLASSIFICATION, PHENOTYPE, AND GENETIC DIVERSITY OF P. INHIBENS

The obligate marine genus Phaeobacter belongs to the Roseobacter group, the marine subgroup within the family Rhodobacteraceae (19). Due to the ease at which Rhodobacteraceae bacteria can be cultured and the prevalence of this family in marine ecosystems, today, 181 different genera have been reported (according to NCBI taxonomy); however, over the recent years, there have also been several drastic reclassifications at the genus and species levels (20, 21). Currently, the Phaeobacter genus comprises the following six species: Phaeobacter gallaeciencis (20), P. inhibens (20), Phaeobacter marinintestinus (22), Phaeobacter piscinae (23), Phaeobacter porticola (24), and Phaeobacter italicus (25) (Fig. 1). However, the phylogenetic diversity within the genus is low (21), and species-level distinction based on 16S rRNA gene sequences is difficult (26). In particular, differentiating between P. inhibens, P. gallaeciensis, and P. piscinae is challenging (23). When comparing the whole nucleotide data per genome using the average nucleotide identity, P. piscinae (89.7%) and P. gallaeciensis (89.3%) are closely related to P. inhibens (Fig. 1). P. porticola (84.5%) and P. italicus (78.6%) are less similar to P. inhibens.

FIG 1.

Comparison by average nucleotide identity of the genomes of P. gallaeciencis DSM 26640^T, P. inhibens DSM 16374^T and DSM 17395, P. piscinae 27-4^T, P. porticola P97^T, and P. italicus CECT 7645^T using JSpeciesWS (103).

Phaeobacter inhibens strain DSM 16374^T (also named LMG 22475^T or T5^T) was described in 2006 as a reclassification of the species Roseobacter gallaeciensis (20). Additional well-characterized strains include DSM 17395 (initially isolated as type strain of P. gallaeciensis BS107) (27, 28) and 2.10 (also named DSM 24588) (29). The strain DSM 17395 was considered identical to the strain CIP 105210, as both represented culture collection deposits of P. gallaeciensis strain BS107^T; however, further genomic analysis revealed that DSM 17395 is more closely affiliated with P. inhibens T5^T and represents a strain distinct from CIP 105210 (28). CIP 105210^T (= DSM 26640^T = BS107^T) is now the type strain of P. gallaeciensis. In many studies, the descriptions of the source of strain BS107^T are unclear, and accordingly, the biological identity of the strain described in these studies remains unknown.

Extensive physiological data on P. inhibens are available (20). P. inhibens clearly differs morphologically from other marine bacteria due to the formation of brown colonies on nutrient-rich iron-containing medium due to precipitation of a brown TDA-iron complex (30). In liquid medium, cells are motile but tend to form star-shaped aggregates, also called rosettes (31). Based on the 22 currently available genome sequences (Table 1), genomes of P. inhibens strains have an average size of 4.36 Mb (range, 4.02 to 4.84 Mb) and an average GC content of 59.8% (59.5% to 60.0%). The nucleotide information is structured in 1 chromosome and several plasmids (3 to 10 plasmids), including a few very large (up to 262 kb) plasmids, or chromids. The genomes carry on average 4,128 genes (3,771 to 4,609 genes). Chromosomes are largely syntenic, and tetranucleotide frequency distribution is highly homogenous (26). The pangenome of 22 P. inhibens strains comprises 9,683 genes, and the core genome consists of 2,980 genes (Fig. 2). The pangenome did not reach saturation.

TABLE 1.

Phaeobacter inhibens genome features and strain information

Strain	Country	Coordinates	Environment	Habitat	Assembly no.	Size (Mb)	GC% content	No. of scaffolds	No. of genes	No. of proteins
2.10	Australia	NAa	Ulva australis	Algal surface	GCA_003443555.1	4.16	59.8	4	3,913	3,802
DOK1-1	South Korea	37°14′24.0″N 131°52′12.0″E	Seawater	Surface water	GCA_001969345.1	4.29	59.6	5	4,071	3,855
DSM 16374	Germany	53°42′20″N 07°43′11″E	Natural	Water_sediment	GCA_000473105.1	4.13	60.0	6	3,916	3,820
DSM 17395	Spain	43°23′0.128″N 8°24′31.23″W	Aquaculture	Clam_larvae_aquaculture, Pecten maximus	GCA_000154765.2	4.23	59.8	4	3,992	3,880
P10	France	48°21′16″N 04°33′31″W	Natural	Biofilm_boat	GCA_002888685.1	4.24	59.8	4	3,981	3,854
P24	Denmark	56°49′02″N 08°31′47″E	Aquaculture	System_aquaculture	GCA_002891665.1	4.40	59.9	9	4,198	4,084
P30	Denmark	56°49′02″N 08°31′47″E	Aquaculture	System_aquaculture	GCA_002892205.1	4.32	59.9	6	4,089	3,962
P48	Denmark	56°49′02″N 08°31′47″E	Aquaculture	Fish_larvae_aquaculture, Scophthalmus maximus	GCA_002892265.1	4.11	60.0	4	3,898	3,792
P51	Denmark	56°49′02″N 08°31′47″E	Aquaculture	Fish_larvae_aquaculture, Scophthalmus maximus	GCA_002891985.1	4.12	60.0	6	3,926	3,816
P54	Denmark	56°49′02″N 08°31′47″E	Aquaculture	Zooplankton_aquaculture	GCA_002892185.1	4.25	59.8	6	4,010	3,903
P57	Denmark	56°49′02″N 08°31′47″E	Aquaculture	Zooplankton_aquaculture	GCA_002892005.1	4.02	60.0	5	3,771	3,662
P59	Denmark	56°49′02″N 08°31′47″E	Aquaculture	Zooplankton_aquaculture	GCA_002892165.1	4.10	59.9	4	3,845	3,737
P66	Spain	43°23′0.128″N 8°24′31.23″W	Aquaculture	Algae_aquaculture, phytoplankton mixture	GCA_002892065.1	4.78	59.7	10	4,593	4,457
P70	Spain	43°23′0.128″N 8°24′31.23″W	Aquaculture	Algae_aquaculture, phytoplankton mixture	GCA_002892125.1	4.48	59.9	7	4,275	4,153
P72	Spain	43°23′0.128″N 8°24′31.23″W	Aquaculture	Clam_larvae_aquaculture, Ostrea edulis	GCA_002891945.1	4.43	59.8	6	4,209	4,093
P74	Spain	43°23′0.128″N 8°24′31.23″W	Aquaculture	Clam_larvae_aquaculture, Venerupis philippinarum	GCA_002892045.1	4.48	59.9	7	4,276	4,152
P78	Spain	43°23′0.128″N 8°24′31.23″W	Aquaculture	Clam_larvae_aquaculture, Venerupis philippinarum	GCA_002891965.1	4.27	60.0	7	4,045	3,921
P80	Spain	43°23′0.128″N 8°24′31.23″W	Aquaculture	Clam_larvae_aquaculture, Venerupis philippinarum	GCA_002892145.1	4.70	59.6	11	4,478	4,342
P83	Spain	43°23′0.128″N 8°24′31.23″W	Aquaculture	Clam_larvae_aquaculture, Venerupis philippinarum	GCA_002892225.1	4.65	59.6	11	4,419	4,283
P88	Spain	43°23′0.128″N 8°24′31.23″W	Aquaculture	Clam_larvae_aquaculture, Venerupis philippinarum	GCA_002892085.1	4.84	59.5	10	4,609	4,466
P92	Spain	43°23′0.128″N 8°24′31.23″W	Aquaculture	Clam_larvae_aquaculture, Donax trunculus	GCA_002892025.1	4.48	59.9	7	4,275	4,153
S4Sm	USAb	NA	Crassostrea virginica	Inner shell	GCA_001559785.1	4.40	59.8	80	4,221	3,972

^aNA, not available.

^bRhode Island.

FIG 2.

Gene saturation curve for the core genome (white boxes) and the pangenome (blue boxes) of 22 Phaeobacter inhibens strains. Nucleotide data annotated with Prokka (104) and pangenome calculated with Roary (105). Visualization in R with ggplot2.

OCCURRENCE OF P. INHIBENS

P. inhibens DSM 16374^T was isolated in 1999 from the German Wadden Sea (53°42′20″N, 07°43′11″E) in marine broth (MB 2216; Difco) due to its strong antibacterial activity (2). Further strains were isolated globally (Europe, Asia, North America, and Australia) from coastal waters (26, 32) and marine aquaculture facilities (algae, zooplankton, clam, and fish) (6, 33–35). At the genus level, Phaeobacter represents approximately 0.03% of all sequences of bacteria in marine surface waters, as found in the global marine metagenomic sequence data (36). As experimentally assessed by CARD-fluorescence in situ hybridization (FISH), the Roseobacter group is of cosmopolitan distribution and represents 2% to 7% of the bacterioplankton in surface waters (37). A recent genomic analysis of the Phaeobacter genus (being represented by nearly 50% of genomes of P. inhibens) demonstrated that the genomic clustering of the strains was independent of their geographic distribution (26).

Interestingly, Phaeobacter species have not been isolated from open ocean waters (38) but instead occur in harbor areas or aquaculture tanks (6, 24, 32, 39, 40). Roseobacter bacteria could be isolated from water samples of aquaculture larval rearing tanks (40), but most Phaeobacter isolates in culture today were obtained from biofilm samples, e.g., from walls of fish tanks or macrobiological (algal, bryozoan, and barnacle) or inert surfaces (wood, metal, and plastic) submerged in seawater (Table 1) (26, 27, 29, 33, 41). Overall, the knowledge pertaining to the natural environmental habitat of the genus remains limited. However, laboratory experiments and genomic analyses indicate an adaptation of P. inhibens to a biofilm lifestyle (1). In laboratory cultures, P. inhibens forms rosettes that assemble into biofilms on the air-liquid interface and the walls of the culture container when growing under agitation and even more so when growing under stagnant conditions (42, 43). Also, P. inhibens was demonstrated to attach to microalgae and their exudates in laboratory cultures (9, 44–46), indicating not only a preference to thrive in a biofilm but also an adaptation to an eukaryote-associated lifestyle.

EFFECT OF P. INHIBENS CELLS ON AQUACULTURE ORGANISMS

Antagonistic activity of P. inhibens against Vibrio spp.

Pathogenic outbreaks in hatcheries can cause great economic damage, such as the loss of 40% of juvenile turbot in a Norwegian hatchery in 1988 due to vibriosis (47, 48). Vibriosis affecting marine fish and shrimps still represents a major issue for the wealth of the aquaculture sector (49, 50). While vaccines can be used in adult fish, treatment of larvae and juveniles with an undeveloped or underdeveloped immune system must rely on other procedures. Traditionally, these procedures are antibiotic treatments, which also pose the risk of antibiotic resistance development (51, 52). For more than 2 decades, scientists have evaluated probiotics for aquaculture (48, 53). In the 1990s, Phaeobacter strains were first isolated and proposed as probiotics for aquaculture due to their inhibitory activities against fish pathogenic bacteria, such as Vibrio anguillarum (27, 53). Subsequently, many more Phaeobacter isolates were obtained and their anti-Vibrio activity confirmed (6, 33–35, 40). Over a 7-day cocultivation experiment, Phaeobacter inhibens DSM 17395 (inoculum, ∼10⁶ CFU/ml) was capable of preventing growth of V. vulnificus (inoculum, ∼10³ CFU/ml), while monocultures of V. vulnificus would reach a cell concentration of ∼10⁶ CFU/ml (5). More recently, strong in vitro antagonistic effects of P. inhibens DSM 17395 toward various Vibrio species of aquaculture interest, such as Vibrio aestuarianus, Vibrio alginolyticus, Vibrio harveyi, Vibrio parahaemolyticus, and Vibrio splendidus, were also confirmed (Lallemand, unpublished).

After having established the anti-Vibrio activity of Phaeobacter inhibens in laboratory dual-cultivation setups, its activity against fish pathogenic Vibrio strains was also assessed in aquaculture-related multispecies experiments (see Table S1 in the supplemental material). Since pathogenic bacteria may enter the larval tanks via live feed (54, 55), challenge trials were conducted in cultures of rotifers, brine shrimps, copepods, and microalgae, as well as in marine fish larvae. In algal cultures, the reduction of Vibrio bacteria is typically 2- to 4-log₁₀-fold (5, 7–10, 56). The extent of the effect may sometimes lead to the pathogen being undetectable by cultivation (9), while no adverse effect of the probiotic bacterium on the algae was observed. The same level of activity was observed in small zooplankton cultures (copepod, brine shrimp, rotifer, and cod larvae) (7, 9, 10, 56), while no reduction of Vibrio spp. by P. inhibens was observed in adult oysters (5, 41, 57, 58). Phaeobacter inhibens was able to level off the mortality of cod larvae challenged with Vibrio anguillarum (accumulated control mortality reaching 100%) while substantially increasing the survival of unchallenged larvae administered with the probiotic (9).

Effect of P. inhibens on the host.

In general, the probiotic bacterium P. inhibens appears to have no or a positive effect on higher organisms (Table S1). In Vibrio-challenge trials as well as in dual cocultivation with microalgae, P. inhibens had no effect on the host (7–9, 56). In cocultivation with the microalga Emiliania huxleyi, P. inhibens initially promoted algal growth but subsequently reduced algal cell numbers during senescence (46, 59).

Phaeobacter inhibens has a growth-promoting effect on rotifers (9). It caused improved survival of various host organisms (brine shrimp, cod larvae, and fish cell line) (7, 9), while other studies found no significant effect of P. inhibens on survival of the brine shrimp Artemia nauplii or copepods (56, 60). P. inhibens had no adverse effect on the survival of the nematode Caenorhabditis elegans or turbot larvae (34, 60). The strain LSS9 of the related species P. piscinae induced symptoms of bleaching disease in defense-deficient red alga Delisea pulchra, but it did not invade algal cells (61). To our knowledge, there is only one report of a potential beneficial effect of a Phaeobacter sp. strain on shrimp survival (62), even though the family Rhodobacteraceae and Phaeobacter spp. are largely represented within shrimp aquaculture systems. The Rhodobacteraceae family was found to comprise 10% to 30% of the microbial community (relative abundance) in water of shrimp hatchery and nursery systems in Vietnam, while its level can reach 70% to 80% within the shrimp-associated microbiome. A Phaeobacter daeponensis strain was detected at a relative abundance of 3% in shrimp larvae (Lallemand, unpublished). Interestingly, Zhao et al. (63) recently performed a feeding trial where the improving effect of a P. daeponensis strain on abalone health, V. harveyi resistance, performance parameters, and modulation of the gut microbiota was determined.

Effect of P. inhibens on the host microbiome.

The microbiome associated with an organism significantly contributes to its health, and accordingly, maintaining a balanced microbiome on all trophic levels is key to sensitive and intensive fish-rearing systems. Phaeobacter spp. are endemic to aquaculture systems (6, 27, 33); however, artificial addition of a high load of a given microorganism might cause imbalance in aquaculture-related microbiomes. So far, high-dose addition of P. inhibens to the microbiomes of microalgae (Tetraselmis suecica, Thalassiosira rotula, and Emiliania huxleyi), copepods (Acartia tonsa), fish larvae (Scophthalmus maximus), and oysters (Ostrea edulis) has been investigated using cultivation-dependent techniques and amplicon sequencing approaches (5, 18, 44, 64). In natural biofilm samples, the relative abundance of P. inhibens is estimated to be 0.02% to 0.68% of the bacterial community (32). In contrast, the dosages of P. inhibens administered in laboratory experiments were higher, resulting in relative abundances of 5% of a natural seawater community added to a microalga, 0.9% to 90% of a native microalgal microbiome, or 23% of a native oyster microbiome (44, 64). The results of the microbiome studies indicate that the effect of P. inhibens was dependent on the overall complexity of the native microbiome; i.e., low-complexity microbiomes, such as those of algae and copepods, were structurally changed due to addition of P. inhibens, while the more complex microbiomes of oysters and turbot fish larvae were less affected. Addition of P. inhibens reduced or maintained the abundance of individual operational taxonomic units (OTUs) of putative pathogenic groups, such as vibrios and Pseudoalteromonas sp.

To test if the antibacterial activity of P. inhibens would affect the microbiome assembly of a microalga, Thalassiosira rotula was incubated with microbial communities from natural seawater and either P. inhibens 2.10 or a variant strain lacking the antibacterial activity (65). Only minor differences between the microbiomes in the two setups were identified (44), suggesting that the antibacterial activity of P. inhibens did not have a major effect on bacterial community assembly. Overall, the available data suggest that P. inhibens has a minor impact on the microbiomes of marine eukaryotes. The observed changes are highly specific, i.e., partially decreasing the abundance of putative pathogenic or opportunistic bacteria, such as vibrios. Available studies were conducted over a maximum time period of 8 days, and long-term exposure of microbiomes to TDA-producing bacteria is required.

BIOACTIVE MOLECULES PRODUCED BY P. INHIBENS

Apart from TDA (2), several other small molecules of biological interest have been isolated from P. inhibens, as follows: methyl troposulfenin (66), roseobacticides A to K (67), roseochelin A and B (68), acylated homoserine lactones (AHLs), and siderophores (1) (Table 2). Small molecules are key to the interaction between bacteria as well as between bacteria and their host and are thus likely involved in the putative probiotic effect of P. inhibens.

TABLE 2.

Small molecules produced by Phaeobacter inhibens^a

Compound(s)	Bioactivity	Concentration	Biosafety	Reference(s)
Tropodithietic acid	Antibacterial, iron chelating, anticancer	MIC of Mycoplasma gallisepticum, 0.39 μg/ml; MIC of Vibrio anguillarum, 40 μg/ml; LC50 of U251 CNS renal cancer, 1.2 μg/ml	Potential effect on aquaculture microbiome	2, 17, 70
Methyl troposulfenin	Antibacterial	IC50 of Vibrio anguillarum, 0.14 mM	Unknown	66
Roseobacticides A–K	Algicidal	IC50 of Emiliania huxleyi, 0.2 μM	Potential effect on feed algae	45
Roseochelin B	Algicidal, antibacterial	IC50 of Emiliania huxleyi, 64 μM; IC50 of Vibrio orientalis, 75 μM	Potential effect on feed algae or aquaculture microbiome	68
Acylated homoserine lactones	Quorum sensing	NDb	Potential effect on aquaculture microbiome	1, 89, 106, 107
Siderophore	Iron chelating	ND	Potential effect on aquaculture microbiome	1

^aSee Table S2 for extended data.

^bND, not defined.

Tropodithietic acid.

TDA is a disulfide-containing tropone derivative (2). The current knowledge on the molecule has been recently reviewed, and we therefore focused on the safety-related aspects (69). TDA was first discovered from a Pseudomonas sp. from soil as its tautomer thiotropocin (70, 71). Besides its biosynthesis by Phaeobacter spp. (2, 59), TDA is also produced by strains of the genera Epibacterium (formerly Ruegeria or Silicibacter) (59, 72, 73) and Pseudovibrio (74, 75). An S-methylated analogue of TDA, namely, methyl troposulfenin, was recently characterized (66). Its inhibitory effect against Vibrio anguillarum was 4-fold to 100-fold lower than that of TDA, indicating that methylation of TDA turns the potent compound inactive.

TDA is produced under iron-rich conditions and has a weak iron-chelating activity (30). Its production is regulated by acyl homoserine lactone quorum sensing and autoinduction (76), but cell aggregation and biofilm formation are not physiological prerequisites for TDA production (42, 58). Also, it was proposed that TDA biosynthesis is posttranscriptionally regulated and that it requires a cross-regulation of carbon, nitrogen, and sulfur metabolism (77, 78). It is bactericidal against Gram-positive and Gram-negative bacteria, including human pathogens, such as Proteus mirabilis, Mycoplasma spp., Staphylococcus aureus, and Klebsiella pneumoniae (MIC, ≤3.13 μg/ml) (70), and fish pathogens, such as Vibrio anguillarum (MIC of 90-11287, 40 to 80 μg/ml; MIC of NB10Sm, 1.25 μg/ml), V. parahaemolyticus (MIC, 3.31 μg/ml), and V. vulnificus (MIC, ≤0.83 μg/ml) (5, 16, 58) (see Table S2 for all details). Antibacterial activity is dependent on pH and was found to be stronger in the acidic range (test range, pH 5 to 9); e.g., the MIC against Escherichia coli was 50 μg/ml at pH 9, decreasing to 0.1 μg/ml at pH 5 (70, 79). The stability of anti-Vibrio activity of (thus presumably TDA-containing) P. piscinae 27-4 supernatant was assessed across ranges of temperature (−80 to 37°C) and pH (1 to 9) (80). The antibacterial activity of the supernatant was stable after 210 days of storage at −80°C but decreased after approximately 30 days at −20°C. Over a period of 3 days, the activity was stable after storage at 5°C but decreased with increasing temperature and time of storage (∼50% loss at 37°C after 1 day of storage and complete loss of activity after 2 days in comparison to 5°C). In this assessment of Phaeobacter supernatant, exposure to different pH values for 1 to 1.5 h did not affect bioactivity. The stability of anti-Vibrio activity of P. inhibens will have to be studied and potentially improved in order to develop a robust product for use in aquaculture systems, which are exposed to a broad range of temperatures and where storage conditions are not always optimal for biological products.

TDA is also active against pathogenic fungi, including Rhizoctonia solani and Pyricularia oryzae (MIC, 3.13 μg/ml) (Table S2) or Candida albicans (inhibition zone of 43 mm in an agar diffusion assay with 50 µl TDA on a 9-mm-diameter filter paper disk; concentration, 700 μg/ml) (81). An agar diffusion assay against three strains of microalgae (of Chlorella and Scenedesmus) resulted in inhibition zones of 17 to 24 (50 µl TDA on a 9-mm-diameter filter paper disk; concentration, 700 μg/ml) (81). The survival of the aquaculture feed alga Tetraselmis suecica was affected by TDA at a concentration of 10.5 μg/ml, but no effect was observed at 0.2 μg/ml (9). A total of 42.4 μg/ml of TDA caused a rapid loss of motility in the amoeba Dictyostelium discoideum followed by cell death (17). It had no significant effect on survival of Artemia nauplii at a concentration of 3.2 μg/ml (60), but 100% mortality was observed at 1 mg/ml (81). The 50% inhibitory concentration (IC₅₀) of TDA against the nematode Caenorhabditis elegans was determined to be 25 µg/ml (81).

TDA was cytotoxic to mammalian neuronal and glial cell lines at concentrations of >0.1 μg/ml (82) and to cancer cell lines (MCF7 breast carcinoma, HM02 gastric carcinoma, and HEPG2 hepatocellular carcinoma) (50% of maximal inhibition of cell proliferation [GI₅₀], 5.0 to 6.7 µg/ml) (81). A screening against the NCI-60 cell collection demonstrated the broad-spectrum lethal and growth-inhibitory activities of TDA against cancer cells, particularly against certain renal cancer cell lines, with 50% lethal concentrations (LC₅₀s) of around 1.3 µg/ml (17). In contrast, the IC₅₀ of TDA against noncancerous MCF10A epithelial cells was 4.1 µg/ml. In neuronal cells, TDA may cause disruption of the mitochondrial membrane potential and activation of the oxidative stress response (82). Thus, the mechanism of TDA might be dependent on the cell type and target organism. However, as mentioned above, no negative effect of the live P. inhibens cells producing TDA was so far observed on eukaryotic host organisms, indicating that TDA is produced at concentrations unlikely to cause concerns for the host.

The usage in aquaculture (and elsewhere) of probiotic bacteria that produce antibacterial molecules raises the concern of resistance development against these bioactive compounds—in this case, TDA. P. inhibens carries genes conferring its self-resistance to TDA (17), and TDA-resistant or -tolerant bacterial strains were isolated from eukaryote-associated microbiomes (74); however, no tolerance or resistance of pathogenic vibrios to TDA could be induced in long-term exposure experiments (16, 83). The genes conferring self-resistance to P. inhibens are colocated to the TDA gene cluster on the largest plasmid (262 kbp), which has so far not been demonstrated to be transmissible or to carry the necessary genes for transmission (84).

In E. coli, TDA was proposed to act as an electroneutral proton antiporter, creating an acidic cytosol by the import of hydrogen ions while exporting metal ions (17). Subsequently, TDA would cause the disruption of the proton motive force. At subinhibitory concentrations, TDA caused cell regeneration, such as the biogenesis of the cell envelope and upregulation of defense mechanisms (oxidative stress defense and iron uptake), proposing a more complex mode of action in the fish pathogen Vibrio vulnificus (85). Also, TDA was shown to act as a regulator of quorum sensing, similar to homoserine lactones, and can affect ∼10% of the producer’s gene expression (86).

Roseobacticides.

Phaeobacter inhibens can produce the algicidal compounds roseobacticides A to K upon induction by the algal breakdown products ferulic acid, sinapic acid, or p-coumaric acid (67). The biosynthetic pathway and chemical structure of roseobacticides are similar to TDA (87). In contrast to TDA production, the ability to produce roseobacticides is, however, limited to a narrow phylogenetic subgroup within the Phaeobacter genus (59). The antibacterial and algicidal activities of roseobacticides A and B, or roseobacticide-containing extract, have been assessed against the bacteria P. gallaeciensis, Ruegeria sp., Bacillus subtilis, V. anguillarum, and Pseudoalteromonas tunicata and against the algae Emiliania huxleyi, Rhodomonas salina, Chaetoceros muelleri, Thalassiosira pseudonana, Tetraselmis suecica, and Isochrysis sp. (45, 59). Roseobacticides did not inhibit any bacterial strains at the tested concentrations (≤160 μM). The lowest IC₅₀s (0.1 to 0.2 μM) were detected against the microalgae Emiliania huxleyi and Rhodomonas salina. The aquaculture strains Tetraselmis suecica and Isochrysis sp. were not affected by roseobacticides, and also, R. salina was able to recover from addition of roseobacticide-containing extract. In the event that production of roseobacticides might cause concern in usage of Phaeobacter as a probiotic in aquaculture systems, a TDA-positive Phaeobacter strain has already been isolated that does not produce roseobacticides (59).

Siderophores and acylated homoserine lactones.

The addition of sinapic acid also induces production of the small molecules roseochelin A and B in P. inhibens (68). The molecules were chemically characterized, but the biosynthetic genes are unknown. Roseochelin B has iron-chelating activity and is algicidal against the microalga Emiliana huxleyi, but no growth inhibition was detected for Chaetoceros muelleri and Dunaliella tertiolecta. Antibacterial activity of roseochelin B was detected against Vibrio strains (IC₅₀ of 75 to 891 μM), while other bacterial pathogens (Pseudomonas aeruginosa, Bacillus subtilis, Staphylococcus aureus, and Enterococcus faecalis) and Saccharomyces cerevisiae were not affected.

The iron-chelating activity of P. inhibens DSM 17395 was also described in a phenotypic chrome azurol S assay (without induction by sinapic acid), and a putative biosynthetic gene cluster (BGC) with homology to the petrobactin BGC of Bacillus cereus and Marinobacter aquaeolei was identified (1, 88) (Table 3). However, the chemical structure of this siderophore and its broader activity range remain to be determined. Siderophores are iron-chelating compounds, but since marine waters are already iron limited and the natural microbiome produces siderophores, the addition of a single organism with siderophoric activity is unlikely to have any specific effects or raise any concerns.

TABLE 3.

Bioinformatic analysis for biosynthetic gene clusters of Phaeobacter inhibens DSM 16374^Ta

BGC no.	BGC type	NCBI accession no. of contig	Nucleotide position		On contig edge	Most similar known cluster	MIBiGb BGC ID
			From	To
1	Bacteriocin	NZ_KI421498.1	368979	379828	−	None	None
2	HSLc	NZ_KI421498.1	490450	511158	−	None	None
3	HSL	NZ_KI421498.1	1725917	1746526	−	None	None
4	HSL	NZ_KI421498.1	1856118	1876691	−	None	None
5	Type I PKS	NZ_KI421498.1	2000683	2052870	−	None	None
6	HSL	NZ_KI421498.1	2179961	2200603	−	None	None
7	Siderophore	NZ_AXBB01000007.0	53610	67178	−	None	None
8	NRPS	NZ_AXBB01000008.0	35116	68567	+	Polysaccharide B; 6% similarity	BGC0001411_c1

^aExtracted from the antiSMASH database (in fast mode) (108). See https://antismash-db.secondarymetabolites.org/output/GCF_000473105.1/.

^bMIBiG, Minimum Information about a Biosynthetic Gene cluster (https://mibig.secondarymetabolites.org/).

^cHSL, homoserine lactone.

Production of the quorum sensing signals acylated homoserine lactones (AHLs) by P. inhibens was identified in an Agrobacterium tumefaciens bioassay (89, 90). This assay typically detects regular, 3-hydroxy-, and oxo-HLs and is sensitive to longer-chain AHLs (>C₄) (91). The AHLs N-(3-hydroxydecanoyl)-l-HL and N-(octadecanoyl)-l-HL were confirmed chemically to be produced by T5 (89) and N-(3-hydroxydecanoyl)-l-HL, N-(dodecanoyl-2,5-diene)-l-HL, and N-(3-hydroxytetradecanoyl-7-ene)-l-HL by S45m. The S45m AHLs inhibited the protease activity of Vibrio coralliilyticus RE22 with IC₅₀s of 0.26 µM, 3.7 µM, and 2.9 µM, respectively. N-(3-Hydroxydecanoyl)-l-HL is produced by the AHL synthase PgaI (BGC no. 6) (Table 3) and partly regulates the production of TDA (76). The supernatant of a 24-h culture of an AHL-negative mutant was not inhibitory against V. anguillarum as was the wild type; however, after 3 days of incubation, the inhibitory activity of the mutant was at the level of the wild type (8).

OTHER FACTORS POTENTIALLY MEDIATING INTERACTIONS OF P. INHIBENS

According to the current state of knowledge, P. inhibens has shown no or a positive effect on aquaculture host organisms. To identify potential functional traits of concern, the genome annotation was screened for putative virulence factors. However, all genes identified can also be associated with traits of basic cellular functions or communication. The genome of DSM 17395 encodes a complete type IV secretion system (1, 84), a versatile system found in both Gram-negative and Gram-positive bacteria and that secretes a wide range of substrates and compounds (92). Furthermore, the genome annotations of P. inhibens DSM 16374^T and DSM 17395 contain genes associated with type I, II, and IV secretion proteins; toxin-antitoxin systems; hemolysin; and up to 19 proteases (see Table S3 in the supplemental material). Genomic analysis of DSM 17395 using the KEGG database (93) indicates the presence of several complete ABC transporters, such as ZnuABC (Zinc) or YejABEF (microcin C), and two-component systems, such as ChvGI. Phaeobacter inhibens carries the necessary genes for the secretory pathways twin arginine translocation system (TatABC) and the Sec-signal recognition particle (SRP) pathway (FtsY, Ffh, SecABDEFGY, YajC, and YidC). Many of these secretion pathways were also experimentally confirmed to be produced (94). Furthermore, the genome indicates the ability of P. inhibens to perform chemotaxis, and as mentioned above, P. inhibens is capable of quorum sensing, biofilm formation, and motility. No genes encoding other virulence factors, such as adhesin and invasin, were found. Although genetically encoded, the herein identified proteins and pathways have so far not been tested for their association with possible virulence in P. inhibens.

Antimicrobial resistance.

Antimicrobial susceptibility can be determined by phenotypic testing and searching for AMR genes through whole-genome sequence analysis. In agar disk diffusion assays, P. inhibens DSM 16374^T was found to be susceptible to many antibiotics and strongly susceptible (inhibition zones, >40 mm) to ampicillin (10 µg), cephalosporins (30 µg), chloramphenicol (30 µg), imipenem (10 µg), linezolid (30 µg), penicillin G (10 IU), piperacillin-tazobactam (40 µg), and ticarcillin (75 µg) (23). These susceptibilities were also observed for DSM 17395, P. gallaeciensis DSM 26640^T, and two P. piscinae strains. Other studies described the susceptibility of DSM 16374^T to streptomycin sulfate (6 µg), kanamycin (30 µg), gentamicin (30 µg), and neomycin (30 µg) using disk assays, which also applied to the close relatives P. gallaeciensis, Leisingera daeponensis, and Leisingera methylohalidivorans (20, 95).

P. inhibens DSM 16374^T and DSM 17395 were resistant (disk assay; inhibition zone, 0 mm) to nystatin (100 IU), clindamycin (10 µg), lincomycin (15 µg), teicoplanin (30 µg), and fosfomycin (50 µg); these results were also observed for P. gallaeciensis DSM 26640^T and four P. piscinae strains (23) and thus appear inherent to this phylogenetic group. Furthermore, resistance of the strain (and DSM 26640^T and the aforementioned Leisingera strains) to novobiocin (5 µg) was reported by Yoon et al. (95). Otherwise, P. inhibens strains were found to be sensitive to copper (≥0.5 mM CuCl₂) (23).

In the genome of P. inhibens DSM 17395 (chromosome and three plasmids), no acquired antimicrobial resistance genes conferring resistance to the available antibiotics (i.e., aminoglycoside; beta-lactam; colistin; fluoroquinolone; fosfomycin; fusidic acid; glycopeptide; macrolide, lincosamide, and streptogramin B (MLS); nitroimidazole; oxazolidinone; phenicol; rifampin; sulfonamide; tetracycline; and trimethoprim) were detected using ResFinder 4.0 (database version 2020-07-28) (with default settings of 90% identity threshold and 60% minimum length, and with other, less-stringent settings of 70% threshold for identification [ID] and 60% coverage of the query sequence) (96).

The blast results for the DSM 17395 genome against the latest ARG-ANNOT (97) database (ARG-ANNOT_V6_NT_July2019) using the two available online tools (e.g., NCBI BLAST tool version +2.10.1 [98] with low stringency settings of 70% threshold for ID and 60% coverage of the query sequence and the BLAST ARG-ANNOT NT tool) did not show the presence of any known antimicrobial resistance gene conferring resistance to the antimicrobials listed in the EFSA guidance.

Furthermore, the DSM 17395 genome sequence was screened with CARD’s Resistance Gene Identifier (RGI) version 5.1.1 (99) and the CARD database version 3.1.0. “Perfect,” “strict,” and “loose” hits were evaluated for complete and partial prodigal gene predictions with the default filter parameters (e.g., threshold percentages for identity and minimum sequence length) applied by the RGI tool (RGI tool does not provide information on such default filter parameters nor does it offer the user the possibility of selecting filter parameters). The CARD analysis (RGI criteria) resulted in 0 perfect hits, 0 strict hits, and 251 loose hits in the P. inhibens DSM 17395 whole-genome sequence. After filtering out the 251 loose hits (with less than 70% identity against their matched CARD references and hits matching less than 60% of the length of their CARD reference sequences), only 2 AMR gene loose hits remained, both corresponding to the antibiotic elfamycin, which is not included in the EFSA list of antimicrobials of human and veterinary importance in the EFSA guidance. However, with the CARD database, in addition to detecting an AMR gene or a variant of an AMR gene (100% identical sequence), very distant homologs of AMR genes can also be detected—so-called loose hits. Thus, homologous sequences and partial hits not playing a role in antibiotic resistance are always detected.

In conclusion, under the new EFSA guidance about characterization of microorganisms used as feed additives (100), even though phenotypic resistance was detected for clindamycin and fosfomycin, no known AMR genes were identified that can be linked to these phenotypes; therefore, no further studies would be required. Interestingly, resistance to clindamycin was also detected for closely related P. gallaeciensis and P. piscinae strains (23), hence suggesting an intrinsic resistance mechanism.

CONCLUSIONS AND PERSPECTIVE

In conclusion, P. inhibens is a promising probiotic candidate for marine aquaculture systems due to efficient inhibition of fish pathogenic vibrios. P. inhibens maintains its antagonistic effect in the presence of aquaculture-relevant live feed organisms (algae, rotifers, and crustaceans) and fish eggs and larvae, while there is no effect or a positive effect on the eukaryotic host. Also, the effect on the host-associated microbiomes is minor. Importantly, Phaeobacter spp. have repeatedly been isolated from aquaculture systems and thus are intrinsic to these environments. P. inhibens produces few natural products, and its bioactive genomic potential is low compared with other marine bacteria, such as Pseudoalteromonadaceae and Vibrionaceae (88, 101, 102); however, among those molecules is the potent antibacterial agent TDA. TDA is the major driver of Vibrio inhibition. Algicidal molecules produced by P. inhibens did not show an inhibitory effect on aquaculture-relevant algae. P. inhibens demonstrated resistance to several antimicrobial compounds; however, this phenotype was observed across several strains and Phaeobacter species, and no dedicated resistance genes were identified in P. inhibens genomes.

While the results of the herein comprised studies are promising for the probiotic application of P. inibens, a number of tests remain to be conducted to assess its safe use. The probiotic has been tested in fish eggs and larvae (turbot), which are the trophic levels in need of treatment against vibrios. However, carryover of the probiotic to juvenile and adult fish cannot be excluded, and thus, its effect on the higher trophic levels and their microbiomes needs to be evaluated. The ubiquitousness and natural high abundance of the family Rhodobacteraceae, which includes Phaeobacter sp., in water-associated microbiota, including those of healthy eukaryotes, further support the safe use of such organisms as potential probiotics. Current tests have been limited to feed cultures and turbot larvae on a laboratory scale. However, aquaculture systems as well as the fish species produced are very diverse. Also, the production units comprise different ecological niches (such as tank water, tank walls, and biofilters) with their own unique microbiota. Accordingly, the effect of the probiotic in these environments would require further assessment.

Addressing these challenges will not only confirm the applicability and safe usage of P. inhibens in aquaculture systems but also drive the sustainable development of the industry for the production of high-quality protein according to the United Nations (UN) sustainable development goals.

Biographies

Eva C. Sonnenschein studied at Kiel University, Germany, and received her PhD in marine microbiology from Jacobs University Bremen and the Max Planck Institute for Marine Microbiology in 2011. After exploring the secondary metabolism in microalgae for two years as a postdoctoral researcher at UC San Diego, she worked first as a postdoctoral researcher and now is a senior scientist at the Technical University of Denmark. Here, she investigates the molecular interactions of bacteria and microalgae and exploits them for biotech applications.

Guillermo Jimenez studied at the Faculty of Veterinary Medicine, Autonomous University of Barcelona (UAB), and received his degree in Veterinary Medicine (DVM) in 1989 and his MSc in animal production in 2002. In 1990, he started his professional career at Rubinum, S.A. (former Asahi Vet, S.A.) in the research, development, and registration of microorganisms intended for use in animal nutrition, thus gaining experience in the fields of taxonomy, genomics, bioinformatics, and safety and efficacy of probiotics. He has participated in some scientific papers either as author or coauthor. As from 2016 he is R&D Project Manager at Lallemand Animal Nutrition.

Mathieu Castex studied at Paris Institute of Technology for Life, Food and Environmental Sciences (AgroParisTech) and obtained his PhD from the same institute in 2009. He has since then occupied various roles, from research to business development, at Lallemand, a global leader in the development, production, and marketing of yeast, bacteria, and specialty ingredients. Since 2013 he has acted as the RnD director of Lallemand Animal Nutrition, managing a team of microbiologists, physiologists, nutritionists, and animal husbandry experts dedicated to the development of microorganisms intended for use in animal nutrition and health.

Lone Gram received her MSc in 1985 and her PhD in 1989 from the Royal Veterinary and Agricultural University in Copenhagen. She has since 2000 been a professor in bacteriology at the Technical University of Denmark. Lone studies bacterial ecophysiology and biotechnology and is especially interested in marine bacteria. One key interest, as featured in the manuscript, is the use of marine bacteria as probiotics in aquaculture. She spent research visits at University of New South Wales and at Harvard Medical School. She received the Villum Annual Award in 2016 and has since January 2018 been leading a Centre of Excellence on Microbial Secondary Metabolites. She is decorated with the Order of Dannebrog and is a member of the Royal Danish Academy of Sciences and Letters. Lone has published more than 240 scientific papers that are cited more than 12,000 times.