Selection of carbohydrate-active probiotics from the gut of carnivorous fish fed plant-based diets

The gastrointestinal microbiota plays a critical role on host health and metabolism. This is particularly important in teleost nutrition, because fish do not possess some of the necessary enzymes to cope with the dietary challenges of aquaculture production. A main difficulty within fish nutrition is its dependence on fish meal, an unsustainable commodity and a source of organic pollutants. The most obvious sustainable alternatives to fish meal are plant feedstuffs, but their nutritive value is limited by the presence of high levels of non-starch polysaccharides (NSP), which are not metabolized by fish. The composition of fish-gut microbial communities have been demonstrated to adapt when the host is fed different ingredients. Thus, we hypothesized that a selective pressure of plant-based diets on fish gut microbiota, could be a beneficial strategy for an enrichment of bacteria with a secretome able to mobilize dietary NSP. By targeting bacterial sporulating isolates with diverse carbohydrase activities from the gut of European sea bass, we have obtained isolates with high probiotic potential. By inferring the adaptive fitness to the fish gut and the amenability to industrial processing, we identified the best two candidates to become industrially valuable probiotics. This potential was confirmed in vivo, since one of the select isolates lead to a better growth and feed utilization efficiency in fish fed probiotic-supplemented plant-based diets, thus contributing for sustainable and more cost-effective aquaculture practices.

We hypothesize that a selective pressure of diets on fish gut microbiota is a hopeful strategy for an enrichment of bacteria with a secretome able to mobilize dietary components, including the capacity of NSP utilization. Several studies emphasize the presence of carbohydrate-active enzymes in sporeforming species such as Bacillus spp. 39 , but few reports exist on the isolation of carbohydrolytic sporeformers from fish gut with probiotic potential that could be administrated in aquafeeds to help fish in their digestive challenges 11,[40][41][42] . Furthermore, available information on carbohydrases-producing bacteria is mainly restricted to amylolytic, cellulolytic and chitinolytic enzymes [41][42][43] . Thus, screening fish gut microbiota for bacteria capable of producing extracellular digestive enzymes that hydrolyse NSP present in PF (such as mannans, glucans, xylans, arabinans, and galactans), is a promising and unexplored research topic. Gut microbiota isolates showing promising metabolic traits and absence of safety concerns can be used as probiotics in cost-effective and environmental-friendly diets by allowing the host to obtain energy from otherwise indigestible dietary constituents. In fact, native bacteria with probiotic potential will be more apt to become established and persist in the fish gut environment after withdrawal from the diet 44 .
This study describes the isolation, identification and characterization of marine fish gut sporeformers capable of producing carbohydrate-active enzymes that hydrolyse NSP and accesses their potential as probiotics for use in aquafeeds. Sporeformers were isolated from the gut of European sea bass (Dicentrarchus labrax) juveniles challenged with PF diets based on SBM, RSM or SFM, which have different NSP profiles. European sea bass was the model species chosen due to its high commercial importance in European aquaculture and its carnivorous feeding habits, thus being more challenging to cope with PF-based diets.

Results
Isolation and identification of gut sporeformers. More than 200 bacterial fish isolates (FI) were obtained from the heat-treated gut contents of European sea bass fed each dietary situation (CTR, SBM, RSM and SFM; Table 1). Following purification, 160 isolates representing different samples and colony morphologies (illustrated in Fig. 1, Panels A to J) were chosen for analysis. Spore production of each isolate, induced by nutrient exhaustion on Difco Sporulation medium, was confirmed by phase-contrast microscopy ( Fig. 1, Panel K). All isolates were identified by partially sequencing the 16S rRNA gene revealing a predominance (60%) of Bacillus species among European sea bass gut contents ( Fig. 2A). Oceanobacillus were also present, although to a lower extent (~10%), with the remaining isolates distributed between the genera Lysinibacillus and Sporosarcina (with 5% each), Aneurinibacillus and Virgibacillus (with less than 1% of the isolated population, each). Identification to the species level was in most cases inconclusive. Nevertheless, the great majority (>60%) of the isolates belonging to the Bacillus genus fall in the Carbohydrolytic activity of gut sporeformers. The entire collection of 160 isolates was screened for their carbohydrolytic potential by substrate specific culture-based methods, and different profiles of carbohydrate utilization could be assigned to different isolates, as illustrated in Fig. 3. The great majority of isolates grew well on www.nature.com/scientificreports www.nature.com/scientificreports/ glucose-supplemented medium, but not in the other carbohydrates tested. The quantification of each colony density or volume revealed the 43 isolates with higher and/or broader carbohydrolytic capacity ( Supplementary Fig. S1).
Carbohydrate-active gut sporeformers as putative probiotics for aquaculture. The selected 43 isolates were checked for minimal biosafety requirements to be considered as putative probiotics, following the guidelines from the European Food Safety Authority (EFSA) an the World Health Organization (WHO) 27,45 . The majority (33) of the isolates exhibited some degree of hemolytic activity when cultivated on 5% sheep blood agar plates, with 14 isolates showing strong or β hemolysis ( Table 2). Half of the isolates revealed to be resistant to at least 1 antimicrobial, and 10 isolates were resistant to 2 or more antimicrobials, defined as MR in Table 2 and detailed in  Supplementary Table S2. These tests allowed selecting a strict group of 11 isolates as good candidates to become a probiotic for European sea bass ( Table 2, highlighted in bold lettering), as isolates showing strong hemolytic activity or any antimicrobial resistance to the different classes of antibiotics tested were not further studied.      www.nature.com/scientificreports www.nature.com/scientificreports/ The selected 11 isolates were then simultaneously cultured in M9 liquid medium to quantify bacteria growth after 24 h in liquid M9 supplemented with the different carbohydrates (Fig. 4A). The results from 3 independent experiments (Fig. 4A) allowed to eliminate fish isolates FI87 and FI89 from the follow-up tests, after revealing the lowest capacity to metabolize the carbohydrates tested.
Sporeforming isolates FI92, FI99, FI123, FI142, FI57, FI162, FI164, FI187, and FI226, that simultaneously met the minimal safety requirements to be eligible as probiotic and were the most efficient isolates in metabolizing the carbohydrates tested, were further characterized to determine their sporulation efficiency, an important characteristic for future industrial production and feed incorporation. By comparison with the well-studied standard strain B. subtilis 168 46 , isolates FI164, FI187, and FI226 did not reach a minimum titer of 107 ml −1 heat-resistant cells, after 24 h sporulation induction by nutrient exhaustion in DSM liquid medium (Fig. 5) and were discarded from the subsequent tests. Furthermore, FI187 and DI226 did not even reach that minimum level of total (viable) cells, revealing to be inadequate for future industrial applications. With the exception of FI123, the remaining six isolates presented an efficiency of sporulation higher than 70%, which anticipates a high suitability for cost-effective spores production (Fig. 5).
Next, the potential to survive passage through the gastrointestinal tract, important for in vivo efficacy, was determined by exposure to sequential simulated stomach and intestinal conditions. Purified spores of isolates FI92, FI99, FI123, FI142, FI57 and FI162 were first subjected during 4 h to acidified NaCl containing pepsin, to mimic stomach conditions, followed by 24 h exposure to alkalinized LB medium containing pancreatin and bile salts. While 4 h in simulated stomach conditions had nearly no effect on the isolates survival, the subsequent 24 h exposure to simulated intestinal conditions lead to a reduction in each bacterial population (Fig. 6). In particular, cell survival was dramatically decreased in isolates FI92 and FI157, similarly to what was observed to the standard strain B. subtilis 168 (Fig. 6). Isolates FI99 and FI162, which showed higher sporulation efficiency, and consequently higher cell number at time 0, seem to be the best fit to survive in the gut.
The remaining four isolates (FI99, FI123, FI142, FI162) were characterized for their antimicrobial activity against several fish pathogenic strains, namely P. damselae, V. harveyi, T. maritimum, A. bivalvium, and S. aureus. As illustrated in Fig. 7A, all isolates showed some extent of antimicrobial activity. Strain FI99 was successful in inhibiting the growth of S. aureus, T. maritimum and to a lower extent V. harveyi. FI123 was only active against Ph. damselae. FI142 inhibited the growth of S. aureus and of Ph. damselae while FI162 was active against Ph. damselae, V. harveyi and T. maritimum. The control B. subtilis 168 could also effectively inhibit the growth of S. aureus, Ph. damselae and T. maritimum, but this last inhibitory activity was lost when using its cell-free supernatant (Fig. 7B) as opposing to the killing activity observed with the cell-free supernatant of FI162, clearly indicating that this strain produces an extracellular inhibition molecule(s) capable of inhibiting T. maritimum growth (Fig. 7B).
In an attempt to infer the germination capacity of these strains inside the animal gut, spores of the same four isolates (FI99, FI123, FI142 and FI162) were subject to different germinants, namely L-alanine and a mixture of KCl, glucose, fructose and L-asparagine (AGFK). For the conditions tested, isolates FI123 and FI142 were unable to germinate (Fig. S3), leading to the selection of isolates FI99 and FI162 as the most promising probiotic strains.
A second fish-growth trial assay using challenging plant-based diets (CTR−), revealed that supplementation with 1 × 10 9 spores g feed −1 of FI99 and FI99 + FI162 (Mix) has a positive effect on the final body weight, the weight gain, the feed efficiency and the protein efficiency ratio of European sea bass juveniles, with a tendency to get closer to a FM-based diet (CTR+) ( Table 3).
Finally, both strains have been deposited in the Spanish culture collection (CECT-Colección Espanola de Cultivos Tipo), under the Budapest Treaty on the International Recognition of the Deposit of Microorganisms for the Purposes of Patent Procedure, and, were submitted for protection with a Provisional Patent Application (PT No. 115101).

Discussion
The role of gut microbiota in shaping human and animal health is well established, and the potential health benefit of manipulating the gut ecosystem using probiotics is increasingly being accepted 5 . In carnivorous fish such as European sea bass, an ideal probiotic should not only enhance resistance to pathogens (i.e. by competitive exclusion, the most common criteria for selection of probiotic strains) but also help fish in their current dietary challenges, including the utilization of plant-feedstuffs (PF). In this study, the application of a PF-based dietary pressure to modulate European sea bass gut microbiota composition and corresponding metabolic functions revealed to be a successful strategy to find carbohydrate-active bacteria with probiotic potential. In particular, we targeted and isolated spore-forming Bacilli, commonly used in probiotic preparations, mainly due to their extreme resistance characteristics and indefinitely survival, advantageous for industrial applications 29,32,34,38 .
Long viewed as soil inhabitants, Bacillus spores are nowadays believed to alternate between the soil/water environment and the gastrointestinal tract as part of their natural life-cycle, being able to germinate, grow and sporulate again inside the animal gut 30,[47][48][49] . Thus it was not surprising that a great variety of sporulating organisms (most rod-shaped but also including the cocci Sporosarcina spp.) could be found associated with European sea bass gut samples. By sequencing the 16S rRNA gene, all isolates could be assigned to a genus, being Bacillus the most prevalent (>60%). Affiliation to a species based on a single molecular marker (16S rRNA) was limited, as expected 50  β-hemolysis, the bacterial hemolytic enzymes completely break down the blood cells; α-hemolysis, the bacterial hemolytic enzymes only partially break down the blood cells; γ-hemolysis corresponds to essentially no hemolytic activity detected. f AbR-Antimicrobial resistance determined by the E-test method against several antibiotics (Table S2). R-resistance to one antimicrobial; MR-resistance to 2 or more antimicrobials; -no resistance phenotype detected. g Closest known species found using RDP based on partial sequences (600 to 800 nt) of the 16S rRNA gene.  51,52 .
Although several Bacillus spp. are quite common in the gut of different animals, including the ones with high-fiber feeding habits, such as soil invertebrates or the giant panda 53,54 , few studies have focused on their carbohydrolytic potential 39 . For example, predominant B. subtilis strains from the intestinal microbial community of the giant panda, seem to have the capacity to growth in a higher fiber environment 55 , opening the possibility that also in fish, Bacillus spp. may have a decisive role in shaping their host digestive capacity towards the efficient utilization of PF-diets. In fact, two recent studies, although limited to cellulase and xylanase activities, reported the isolation of carbohydrate-active Bacillus spp. from the gut of different fish species 41,42 . The 160 isolates tested in our study showed different, and in some cases potent, hydrolytic capacities when using as sole carbon source selected carbohydrates including xylose, galactose, arabinose, or mannose. This observation was further sustained by the presence of genes coding for specific extracellular CAZymes that can help fish in obtaining the otherwise unavailable energy trapped in PF. The absence of amplification for these specific genes in some isolates, despite showing broad carbohydrolytic activities, is not surprising considering that some studies suggests that new or substantially different CAZymes involved the metabolization pathways are yet to be found in the Bacilli group of organisms 39,56 . Furthermore, the lack of PCR amplification of these genes, observed with some isolates, may also be caused by mismatches of the primer pairs, due to the difficulty to design gene-specific primers regarding genomic regions poorly conserved.
Probiotic approval within European Union for incorporation into animal feed, including aquafeeds, is subject to strict and exhaustive exigencies following EFSA guidelines on quality, safety, and efficacy of the candidate(s) bacterial strain(s) [57][58][59] . Besides the obligation of strain deposition in an internationally recognised culture collection, candidate probiotic isolates must be tested for the presence of any acquired antibiotic resistance genes 31,45,60 . www.nature.com/scientificreports www.nature.com/scientificreports/ Probiotics, which are given to animals in massive amounts, should not contribute to the escalation of antimicrobial resistance by acting as vehicles of transferable genetic determinants. Unfortunately, these rules do not apply worldwide and very recently antimicrobial resistant strains were found in probiotic products used in Vietnamese shrimp culture 61 or in Chinese human commercial products 62 , with all the risks those findings pose to the aquaculture production sector and to public health 63 . Although EFSA guidelines only require the absence of acquired (transmittable) resistance genes, allowing the use of bacterial strains whose antibiotic resistance is chromosomally encoded, we opted to eliminate from our study all the strains showing any antimicrobial resistance to the different classes of antibiotics tested. Adding to that criterion, strains showing strong hemolytic activity, indicative of virulence potential in several pathogenic bacterial species, including sporeforming ones [64][65][66] , were also not further tested. These tests allowed to select a group of 11 probiotic candidates that qualify with the minimal biosafety issues to be approved by EFSA.
To demonstrate efficacy, EFSA requires three in vivo studies showing statistically-significant effects on each target animal species 59 . To conduct such follow-up studies in European sea bass growth and digestibility trials, it was necessary, for practical reasons, to narrow the group of interesting and potential probiotic candidates. These were subjected to a series of consecutive tests to analyze some desired characteristics on a future probiotic product. First the sporulation yield, an important parameter in industrial and economical terms, was determined by comparison with the well-studied standard laboratory strain B. subtilis 168 46 . Six isolates demonstrated high yield spore formation, which anticipates a good suitability for cost-effective spores´ production in industrial scale. Additionally, higher sporulation levels might also act as a form of propagation inside the animal gut, maximizing these strains beneficial effect 30,67 . Second, exposure of purified spores to sequential simulated gastric and intestinal conditions, revealed the four isolates best equipped to survive passage through the gastrointestinal tract, important to guarantee their in vivo efficacy. In particular, isolates FI99 and FI162, which also showed higher  . Viability of spores from each sporeformer isolate (codes in x axis) when exposed for 4 h (T4, blue) to simulated stomach conditions (0.85% NaCl, pH 2, containing 3 mg ml −1 pepsin) followed by 24 h (T24, yellow) exposition to simulated intestinal condition (LB, pH 8 containing 1 mg ml −1 pancreatin and 0.3% bile salts). The initial viable counts (time 0 or T0) are depicted in grey Bacillus subtilis 168 was used as control and the results are the average of three independent experiments with error bars representing the standard deviation.
www.nature.com/scientificreports www.nature.com/scientificreports/ sporulation efficiency, seem to be the best suited to reach, at higher numbers, the intestine where their probiotic action can take place.
To take advantage of these isolates as probiotics, upon passage through the stomach and anterior intestine, spores must germinate to originate new vegetative cells that can produce the enzymes/molecules thought to benefit their host. In nature, spore germination is believed to occur in response to specific nutrients 68,69 . For example B. subtilis spores are known to germinate in response to L-alanine, L-valine and L-asparagine but not in response to their D-enantiomers 68,69 . Taken this, and although the mechanisms of germination of spores of different Bacilli (independently of their specific species) are thought to be essentially the same, one cannot rule out that some of our isolates might respond efficiently to other germination molecules that might be abundant in vivo (inside the animal intestine) 68,69 , explaining the germination failure of isolates FI123 and FI142 under the conditions assayed.
Probiotic candidates FI99 and FI162 were further studied, namely by addressing their in vivo efficacy in improving PF diets utilization by European sea bass. This evaluation revealed the potential of isolate FI99 in   www.nature.com/scientificreports www.nature.com/scientificreports/ improving fish performance and feed utilization efficiency, when used as a supplement in plant-feedstuffs-based diets. Fish fed PF-based diet supplemented with FI99 showed better zootechnical results than fish fed the un-supplemented PF-based diet (CTR−), tending to the results of the FM-based diet (CTR+). Future analyses including digestive enzymes activity and gut microbiota modulation, might help explain the results obtained. Also, new growth trials, with increasing levels of both isolates incorporation in the diets, will clarify if fish zootechnical results can get similar to the ones obtained with FM-diet. Furthermore, a comprehensive screening of FI99 and FI162 genomes will potentially allow the identification of new carbohydrases.
Besides their carbohydrolytic potential, these probiotics might also benefit the fish host by minimizing colonization with pathogenic species, known to be especially problematic in marine aquacultures. This is the case of T. maritimum whose growth was efficiently inhibited in vitro, when exposed to both cells and cell-free culture medium of isolate FI162. Probiotic candidates FI99 and FI162 are thus being further studied, namely by addressing their in vivo efficacy in improving disease resistance in a T. maritimum infection model 70 . From the preliminary identification based on the 16S rRNA sequence, these strains are B. subtilis or B. licheniformis, both species recognised by EFSA as qualifying for the QPS status 31 and thus excluding them from the need to perform additional toxicological studies. Nevertheless, sequencing both strains genomes is currently being performed allowing, in a short period, a comprehensive screening of their genomic potential to better meet the EFSA criteria for probiotics. Dissecting these genomes will permit to detail and further document their promising biotechnological value as probiotics and/or as sources of carbohydrases or antimicrobial molecules. For instance, determining the number and type of CAZymes present in each genome will provide deeper understanding on their carbohydrolytic potential but might also allow identification of genomic features responsible for adaptation to life within the gut that may support the role of Bacillus spp. as probiotics 39,[71][72][73] . The growing applications of spores in biomedicine and biotechnology (as oral vaccines, probiotics or display systems) [32][33][34][35]38,74 , and the fact that there are approximately 30 probiotic strains approved as feed additives in European Union, but only one for aquaculture (Bactocell ® , which is not a sporeformer formulation), underscore the importance of this study.

Materials and Methods
All methods were carried out in accordance with relevant guidelines and regulations, namely in the construction of figures and their compliance with the digital image and integrity policies. All animal experiments were approved by the Animal Welfare Committee of the Interdisciplinary Centre of Marine and Environmental Research (CIIMAR) and carried out in a registered installation (N16091.UDER) and were performed by trained scientists (following FELASA category C recommendations) in full compliance with national rules and following the European Directive 2010/63/EU of the European Parliament and the European Union Council on the protection of animals used for scientific purposes.

Diets composition.
In the first trial, three experimental diets were formulated to be isonitrogenous (47% crude protein), isolipidic (17% crude lipid) and to contain 30% of soy bean meal (SBM diet), 30% of rapeseed meal (RSM diet) or 30% of sunflower meal (SFM diet). A fish meal (FM) based diet was used as the control diet (CTR diet). Fish oil and pregelatinized maize starch were the main lipid and carbohydrate sources, respectively. For the second trial, five experimental diets were formulated to be isoproteic (46%) and isolipidic (18%). A negative control diet (diet CTR−), using fish meal (FM) and plant feedstuffs (PF) (soybean meal, rapeseed meal, corn gluten, wheat gluten, and pea protein concentrate) as protein sources at a ratio of 20:80 of protein from FM:PF, respectively. Three other diets were formulated identical to diet CTR−, with the incorporation separately of 2 lyophilised pure spores preparations (diet FI99 and diet FI162, respectively) at a dose commonly used in fish diets (1 × 10 9 spores g feed −1 ), or mixed in equal parts (diet MIX). The fifth diet was a FM-based diet and was used as positive control (diet CTR+). In all diets of the second trial, fish oil was used as the main lipid source.
All diets were supplemented with bicalcium phosphate to avoid phosphorus imbalance. All diet ingredients were finely ground, thoroughly mixed and dry-pelleted in a laboratory pellet mill (California Pellet Mill, CPM Crawfordsville, IN, USA), through a 3.0 mm die for the first trial and a 2.0 mm die in the second trial. Pellets were dried in an oven at 50 °C for 24 h, and then stored at −20 °C in airtight bags until used. Ingredients and proximate composition of all experimental diets are presented in Table 1.

Animals and experimental conditions. Both animal experiments were performed at CIIMAR, Porto
University, Portugal, with European sea bass juveniles obtained from a commercial fish farm (Maresa S.A., Ayamonte, Huelva, Spain). After transportation to the experimental facilities fish were first submitted to a quarantine period of 30 days before transfer to the experimental system where they were allowed to adapt during 15 days. Before the experimental period, fish were fed a commercial diet (Aquasoja Sustainable Feed; Sorgal, Ovar, Portugal) containing 18% lipids and 44% protein. Both trials were conducted in a recirculating aquaculture system (RAS) equipped with 15 fiberglass tanks of 100 L water capacity, thermo-regulated to 22.7 ± 0.8 °C and supplied with a continuous flow of seawater (36.0 ± 0.5 g L −1 salinity, circa 7 mg L −1 oxygen). The photoperiod was set to 12:12 h light:dark which is provided using artificial illumination. In the first trial, 12 groups of 20 European sea bass with an initial mean body weight of 34.4 g were distributed to each tank and the experimental diets randomly assigned to triplicate groups. The trial lasted 45 days. In the second trial, that lasted 9 weeks, 15 groups of 18 fish with an initial mean body weight of 29.0 g were established and the experimental diets randomly assigned to triplicate tanks. In both trials, fish were fed by hand, twice daily, 6 days a week, until apparent visual satiation. Utmost care was taken to avoid feed losses.
sampling. Fish in each tank were bulk-weighed at the beginning and at the end of the trial, after 1 day of feed deprivation. For that purpose, fish were slightly anaesthetized with 0.3 ml l −1 ethylene glycol monophenyl www.nature.com/scientificreports www.nature.com/scientificreports/ ether (Sigma-Aldrich, Steinheim, Germany). On the sampling days (at day 15 after the beginning of the trial and at the end of the trial or day 45), fish were fed several times over the day to guarantee that intestines were full at sampling time. At 4 h after the first meal, 3 fish per tank were randomly sacrificed with an overdose of ethylene glycol monophenyl ether, for collection of biological samples under aseptic conditions. To overcome inter-fish variation the resulting material was pooled into one sample per tank to assess differences between dietary groups. Whole-intestines (without pyloric caeca) were aseptically excised and squeezed to collect the digesta contents.
Isolation of sporeforming bacteria. Each sample of digesta (1 g) obtained from fish fed the different dietary treatments was homogenized in 9 ml of buffered saline solution (0.9%). Serial dilutions were prepared in Bott & Wilson (B&W) salts and 100 μl aliquots spread on the surface of Luria Bertani (LB) agar medium, after 20 min heat treatment at 65 °C, for sporeformers selection 75 . Plates were incubated at 30 °C in aerobic conditions for up to 5 days. Following selection, sporeformers were isolated and characterized for morphology in Difco Sporulation Medium (DSM), to confirm spore production by phase-contrast microscopy 37,75 . Colonies representing different morphologies were picked at random and purified by restreaking on agar plates of the same media, before storage at −80 °C in LB broth with 30% glycerol. Sporeformers isolates were routinely grown aerobically at 37 °C in LB or DSM. The laboratory strain B. subtilis 168 (BGSC1A1) was used as a control in most of the experiments described in this study.
Screening sporeforming bacteria for carbohydrates metabolization. Each sporeformer isolate was cultured on solid M9 minimal medium 76  Growth after 24 h at 37 °C was recorded by photographing colonies in a Gel Doc XR System (Bio-Rad) using the Image Lab software v.4.0.1 (Bio-Rad). Growth quantification was assessed by measuring the colony volume on fixed areas with local background subtraction (adjusted volume = [CNT*mm 2 ] data counts/mm 2 ) using the Quantity One software v.4.6.9 (Bio-Rad). Quantification of carbohydrates utilization in liquid M9 was performed after an overnight enrichment in liquid LB at 37 °C with agitation. Each isolate was diluted to an initial optical density (OD600; absorbance measured at 600 nm) of 0.1 in liquid M9 minimal medium alone or supplemented with 0.2% of the different carbohydrates previously tested. Bacterial growth was followed during 48 h and quantified by measuring the OD600. In both solid and liquid medium assays, results presented were corrected by subtracting the colony volume/OD600 measured in M9 alone.
Taxonomic identification of putative probiotic isolates. Identification was carried out for all the isolates with promising extracellular carbohydrolytic activities. Total genomic DNA extraction was performed from overnight LB cultures, using the EZNA bacterial DNA purification kit (Omega Bio-Tek, USA), according to the manufacturer's instructions. PCR amplification of the small-subunit rRNA (16S rRNA) was carried at an annealing temperature of 55 °C using primers 27 F and 1492 R 77 . Each 20 µL reaction contained 1 x DreamTaq Buffer (Thermo Scientific, Vilnius, Lithuania), 0.2 mM of each dNTP (Thermo Scientific, Vilnius, Lithuania), 0.2 μM of each primer (STAB Vida, Lisboa, Portugal), 1 U of DreamTaq DNA Polymerase (Thermo Scientific, Vilnius, Lithuania) and 25 ng of DNA template. The Bioinformatics resources Sequence Match package of the Ribosomal Database Project 11 (http://rdp.cme.msu.edu) and BLAST of the GenBank nonredundant (nr) nucleotide database (http://www.ncbi.nlm.nih.gov) were used to analyse the sequencing data. screening putative probiotic isolates for Nspases. To tentatively obtain a set of primers specific for the genes encoding Non-Starch Polysaccharides degrading enzymes (NSPases), an initial search was conducted at the Protein Knowledgebase -UniProtKB with terms "family:hydrolase AND annotation:(type:location AND secreted) AND taxonomy:"Bacteria [2]". A file containing bacterial secreted glycosyl hydrolases (GH) was then created and the ones involved in the utilization of NSP of interest were chosen for further analysis. Enzymes chosen included mannanases, mannosidases, arabinofuranosidases, arabinanases, glucosidases, glucanases, fructosidases (fructanases), fructafuranosidases, galactorunases, xylosidases, and xylanases. The protein sequence of each individual enzyme was used to search for similar proteins in the translated nucleotide database (tblastn) (http:// www.ncbi.nlm.nih.gov) and to make nucleotide alignments between the sequences obtained with ClustalW algorithm using Geneious R7 v7.1.7 (Biomatters, Auckland, New Zealand). Regions of sequence conservation were chosen to design primer pairs (Supplementary Table S1) with the Vector NTI 10 software (Invitrogen, Carlsbad, CA), with a calculated annealing temperature of approximately 55 °C and an amplicon size of 200 to 250 base pairs (bp). PCR amplification was done essentially as described for the 16S rRNA (previous section), adjusting the annealing temperature to 55 °C and the extension time to 30 s. www.nature.com/scientificreports www.nature.com/scientificreports/ Antimicrobial activity screening assays. The antimicrobial activity of selected sporeforming isolates was assessed by a colony overlay assay, essentially as described in 37 , using as targets different fish pathogens. Zones of growth inhibition around the producer strains spots after 24 h incubation at 25 °C (for Photobacterium damselae, Vibrio harveyi, Tenacibaculum maritimum and Aeromonas bivalvium) or 37 °C (for Staphylococcus aureus) were considered as positives and the corresponding growth-inhibition halos diameter measured (mm). A cell-free supernatant screening assay was performed by inoculating BHI or Marine Agar (for T. maritimum) plates with overnight cultures of indicator strains, assuring a uniform and complete coverage of the agar plate. After 15 min rest to allow plates to dry, 1 cm holes where done in the agar and consequently filled with 200 μl of cell-free supernatant of each producer strain, previously centrifuged and filtered through a 0.2 μm cellulose filter, from stationary phase LB cultures (grown overnight at 37 °C). Zones of growth inhibition around the producer strains supernatant holes obtained after 24 h incubation at 25 °C or 37 °C (as before) were considered as positive. All observations were recorded by photographing in a Gel Doc XR System (Bio-Rad) using the Image Lab Software (Bio-Rad).
sporulation, germination and resistance to gut environment. The kinetics of spore formation and germination was quantified using adaptations of well-established methods 37,75,76 . Sporulation occurred in DSM for 24 h at 37 °C in an orbital shaker at 200 rpm, and its efficiency was determined by plating serial dilutions made in B&W isotonic buffer (Bott and Wilson salts: 1.24% K 2 HPO 4 , 0.76% H 2 PO 4 , 0.1% trisodium citrate, 0.6% [NH 4 ] 2 SO 4 , pH 6.7) on LB agar, before and after a 20 min heat treatment at 80 °C to eliminate vegetative cells. Following 24 h incubation at 37 °C, visible colonies were counted and sporulation efficiency calculated as the titre of colony forming units (CFU mL −1 ) before and after the heat treatment.
Preparation of highly purified spores was done according to a new purification method recently described 78 . In brief, 48 h spores preparations (in liquid DSM) of each isolate were centrifuged for 10 min at 10000 g and 4 °C. Cell pellets were suspended in 50 mM Tris-HCl (pH 7.2) containing 50 µg ml −1 of lysozyme, and incubated for 1 h at 37 °C. After a single wash with 1 volume of distilled water (10 min at 10000 g, 4 °C), cell pellets were suspended in 0.05% SDS, followed by three washes with distilled water and finally suspended in 1 volume of distilled water. Spores purity and recovery yields were determined by plating serial dilutions on LB agar, before and after a 20 min heat treatment at 80 °C.
Spore germination in response to the addition of 100 mM L-alanine or to a mixture of 100 mM KCl, 56 mM glucose, 56 mM fructose and 33 mM L-asparagine (AGFK), was performed at 37 °C in 50 mM Tris-HCl, pH 7.5, essentially as previously described 75 .
Potential resistance to gut transit was evaluated by determining the acid and bile tolerance of each selected isolate. For that purpose, 48 h DSM spores preparations were heat-treated for 20 min at 80 °C to eliminate vegetative cells and harvested by centrifugation. After a double wash with Phosphate-buffered saline (PBS), serial dilutions made in B&W salts were plated onto LB agar plates to determine the initial bacterial counts. Spores were then diluted in 1 volume of 0.85% NaCl, pH 2, containing 3 mg ml −1 pepsin (Sigma-Aldrich, Steinheim, Germany), to mimic stomach conditions. Following 4 h incubation at 37 °C with agitation, serial dilutions made in B&W were again plated onto LB agar plates to determine bacterial counts, and, after a single wash with PBS, spores were ressuspended in LB, pH 8 containing 1 mg ml −1 pancreatin (Sigma-Aldrich, Steinheim, Germany) and 0.3% bile salts (Sigma-Aldrich, Steinheim, Germany). Bacterial incubation continued for 24 h at 37 °C with agitation to mimic passage through the intestine. Finally, serial dilutions made in B&W were again plated onto LB agar plates to determine the final bacterial counts. All plates were incubated at 37 °C during 24 h prior to colonies count. statistical analysis. Data related to growth performance and feed utilization efficiency of European sea bass are presented as mean ± standard deviation. Statistical analysis was conducted by one-way ANOVA using the SPSS 21 software package for Windows (IBM ® SPSS ® Statistics, New York, USA). Data were tested for normality and homogeneity of variances by the Shapiro-Wilk and Levene's test, respectively. When normality was not verified, data were transformed prior to ANOVA. Significant differences among groups were determined by the Tukey's multiple range test. The probability level of 0.05 was used for rejection of the null hypothesis.