ARTICLE

Probiotic treatment by indigenous bacteria decreases mortalitywithout disturbing the natural microbiota of Salvelinus fontinalisSébastien Boutin, Céline Audet, and Nicolas Derome

Abstract: Next-generation sequencing is revealing the complex interactive networks of host–bacteria interactions, as it is nowpossible to screen in detail the microbiota harbored by a host. This study investigated the influence of a probiotic treatment onthe survival andmicrobiota of brook charr (Salvelinus fontinalis), focusing on its disturbance of the natural microbiota (dysbiosis).The results indicated that an indigenous probiotic strain (identified as Rhodococcus sp.) colonized neither the fish skin mucusnor the water following the probiotic treatment. Instead, the probiotic strain was detected only in the biofilm of the test tank.Nevertheless, a substantial beneficial effect of the probiotic treatment was observed: the population of the pathogenFlavobacterium psychrophilum decreased in the treated tank water. This study clearly shows that the indigenous strain chosen forthe probiotic treatment did not disturb the natural fish skinmucusmicrobiota but acted directly through the production systemto control the growth of the pathogen and, as a consequence, to enhance fish survival.

Key words: probiotic, Salvelinus fontinalis, Flavobacterium, microbiota.

Résumé : Le séquençage de prochaine génération met en lumière les réseaux interactifs complexes que sont les relationshôte–bactérie en rendant possible la caractérisation détaillée de la microflore qu’abrite l’hôte. La présente étude s’intéresse al’influence d’un traitement de probiotique sur lamicroflore et la survie de l’omble de fontaine (Salvelinus fontinalis), en s’attardantaux perturbations au niveau de la microflore naturelle (dysbiose). Les résultats indiquent qu’une souche probiotique indigène(identifiée comme étant une espèce de Rhodococcus) n’a colonisé ni lemucus de la peau du poisson ni l’eau, a la suite du traitementprobiotique. Au lieu de cela, la souche probiotique n’a été détectée seulement dans le biofilm du bassin d’essai. Or, on a constatéque le traitement probiotique avait un effet bénéfique non négligeable tel qu’attesté par la raréfaction du pathogèneFlavobacterium psychrophilum dans l’eau du bassin d’essai. Cette étude démontre clairement que la souche indigène choisie pourle traitement probiotique ne perturbait pas la microflore naturelle du mucus de la peau du poisson, mais avait un impact directqui passait par le système de production et par ce fait favorisait la survie du poisson. [Traduit par la Rédaction]

Mots-clés : probiotique, Salvelinus fontinalis, Flavobacterium, microflore.

IntroductionIn recent decades, the interest in host–bacteria interactions has

greatly increased. Indeed, this field has significantly advancedboth theoretical and technical aspects due to the development ofnew culture-independent methods, such as metagenomic andmetatranscriptomic analysis (Costello et al. 2009; Roberts et al.2010; Robinson et al. 2010; Gonzalez et al. 2011; Rumbaugh andKaufmann 2012). Thus, host-pathogenic bacteria interaction is nolonger understood as a 2-component system involved in an “evo-lutionary arms race” but rather as an integrative system in whichthe host harbors a consortia of cooperating bacteria, as suggestedin the Black Queen Theory (van Valen 1973; Darch et al. 2012;Morris et al. 2012). Such theoretical advances coincide with theassessment of new methods for the prevention of bacterial infec-tions (Verschuere et al. 2000;Merrifield et al. 2010). Indeed, the useof probiotics is now considered an extremely valuable tool in thestruggle against pathogens in diverse applications in agricultureand aquaculture as well as for improving human health (Fuller1995; Balcázar et al. 2004, 2006; Wang et al. 2008).

The FAO/WHO defines a probiotic agent as “live microorgan-isms which when administered in adequate amounts confer ahealth benefit on the host” (FAO/WHO 2001). The current strategyin developing a probiotic for a given host species is to test aprobiotic agent already proven to be efficient in another host

species (Villamil et al. 2003; Mohamed and Ahmed Refat 2011).However, when transferred into a different environment, the pro-biotic agentmay lose its probiotic properties and possibly becomeharmful for the host (Courvalin 2006). Therefore, to ensure theefficacy of a probiotic in a given host species, the isolation ofendogenous bacteria is favoured for the utilization of probioticthat are effective in another host species. The utilization of anendogenous probiotic would prevent unexpected pathogenicity(Balcazar et al. 2006). However, information about a probiotic’sinfluence on the naturalmicrobiota equilibrium is still scarce. Fortherapeutic use, probiotics are often added in greater concentra-tion relative to the abundance naturally found in hosts (Vine et al.2004). Therefore, there is a pressing need to test whether probiotictreatments have any disturbing effect on endogenous microbiotaand the surrounding environmental bacterial community.

In this study, we focused on the treatment of a cold-waterdisease (CWD) affecting brook charr (Salvelinus fontinalis) by us-ing a mix of probiotics. CWD is caused by the bacterial agentFlavobacterium psychrophilum (Starliper 2011). This bacterium hasbeen recovered from a broad geographic range and is responsiblefor both rainbow trout (Oncorhynchus mykiss) fry syndrome andCWD, 2 important infectious diseases in farmed fish (Wiklundet al. 2000). These diseases especially affect fish in the early lifestages, and it is well documented that stressful conditions and

Received 5 July 2013. Revision received 13 August 2013. Accepted 13 August 2013.

S. Boutin and N. Derome. Institut de Biologie Intégrative et des Systèmes (IBIS), Département de Biologie, Université Laval, 1030 Avenue de la Médecine, Québec, QC G1V 0A6, Canada.C. Audet. Institut des sciences de la mer de Rimouski (ISMER), Université du Québec a Rimouski (UQAR), 310 allée des Ursulines, Rimouski, QC G5L 3A1, Canada.

Corresponding author: Sébastien Boutin (e-mail: sebastien.boutin.1@ulaval.ca).

Pagination not final/Pagination non finale

1

Can. J. Microbiol. 59: 1–9 (2013) dx.doi.org/10.1139/cjm-2013-0443 Published at www.nrcresearchpress.com/cjm on 15 August 2013.

Can

. J. M

icro

biol

. Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

Uni

vers

ity o

f L

aval

on

09/2

7/13

For

pers

onal

use

onl

y.

injuries facilitate infections (Madetoja et al. 2000; Bader et al.2003; Starliper 2011). The probiotics used in this study, recentlyisolated from the natural microflora of brook charr, belong to theaccessory microbiome (species with an abundance of <1% of thecommunity) of natural fish skin mucus microbiota and have al-ready proven their efficiency through in vitro and in vivo experi-ments (Boutin et al. 2012). We used Roche 454 pyrosequencing onbacterial ribosomal 16S locus to systematically identify bacteriaandmonitor the naturalmicroflora of the fish skinmucus and thesurrounding water during an in vivo experiment to ensure thatthose probiotics disturbed neither the natural microflora of fishskin mucus nor the bacterial community of the surrounding en-vironment.

Methods

Ethics statementWe raised fish and conducted the experiment according to the

guidelines required by the Comité de Protection des Animaux del’Université Laval (CPAUL, http://www.vrr.ulaval.ca/deontologie/cpa/index.html?accueil).

Experimental conditionsExperimental conditions were already described in detail in

Boutin et al. 2012 but are briefly described here. Four different fishfamilies (i.e., resulting from different parental crosses), all se-lected on the basis of life history traits of interest (growth andreproduction), were used for the in vivo experiment (S1, S5, S9,S10). In this study, we focused on the families S5 and S9 for themicrobiota analysis because they exhibited the greater variabilityin terms of mortality during the infection and were consideredeither resistant (S5) or sensitive (S9) to Flavobacterium disease(Boutin et al. 2012). All of the individuals were juveniles (age: 0+;mean size: 12.2 ± 1.4 cm; mean mass: 16 ± 5.8 g). Fish were dividedamong 4 tanks: 2 control and 2 test tanks (25 fish per family pertank). Although each tank pair shared the same water recircula-tion system, pathogen transmission by skin contact was not pos-sible across separate tanks. Therefore, tanks supplied with thesame water recirculation system were considered as test pseudo-replicates. Test and control tank pairs were separated in 2 differ-ent units (i.e., supplied with independent water recirculationsystems) to avoid migration of probiotic strains into the controltanks. Fish were raised at 12 °C and oxygen concentration wasalways higher than 12 mg/L. Two weeks after transport to LARSA(Laboratoire de Recherche en Sciences Aquatiques), the first symp-toms of CWD appeared due to natural infection in all tanks andwe started the probiotic treatment.

Probiotic applicationThe experiment took place from17 February to 24 March. Seven

bacterial strains isolated from brook charr skin mucus and se-lected for their antagonistic activity in vitro (Boutin et al. 2012)were cocultured in tryptic soy broth. The probiotic solution wasadded to the 2 test tanks twice daily to reach the concentration of105 cell/mL of water (first application at 0800 and the second at1800) until the fish mortality stabilized (18 days; 21 February to10 March). Dead and moribund individuals were checked for clin-ical symptoms of Flavobacterium disease. The moribund fish wereeuthanized with MS-222 (0.075 g/L) as required by the Comité deProtection des Animaux de l’Université Laval.

SamplingTen individuals from each family and from each tank were

sampled at 3 different time periods after the occurrence of infec-tion: (1) before treatment with the probiotic (17 February), (2) thelast day of the probiotic treatment (10 March), and (3) 20 days afterthe probiotic treatment (24March). Mucus samples were collectedfrom the surface of the fish with sterile swabs (Livia et al. 2006).The target sampling area was the skinmucus between the adipose

fin and the caudal fin; we minimized disturbance of this targetarea by handling the fish above the operculum. Samples were putinto sterile microcentrifuge tubes containing lysis buffer (Tris50mmol/L, EDTA 40mmol/L, sucrose 0.75 mmol/L) and stored in a−80 °C freezer until DNA extraction. Skin mucus and gut samplesfrom each fish that died during the probiotic treatment weresampled. At each sampling time, water level was lowered by 5 cmto allow sampling of the biofilm by swabbing the wall of the tank.The original water level was restored after sampling.

For water bacterial community sampling, a 2-step filtration wasperformed using peristaltic filtration equipment (Masterflex L/SPump System with Easy-Load II Pump Head, Cole-Parmer) cleanedup with 5% HCl and rinsed with Milli-Q water before each filtra-tion. The first step was the conditioning of the filtration setupusing 2 L of the rearing water without filters. The last step was tocollect duplicates of water from test and control units, by filteringa total of 4 L of water over two 47-mm-wide nitrocellulose mem-branes (Advantec): prefilter (3.0 �m) and filter (0.22 �m). Immedi-ately after each filtration process, filters were placed in cryotubescontaining 1mL of sterile lysis buffer (40mmol/L EDTA, 50mmol/LTris–HCl, 0.75 mmol/L sucrose) and then stored at −80 °C untilDNA extraction.

DNA extractionDNA was extracted from mucus, biofilm, and water using a

modified salt-extraction protocol (Aljanabi and Martinez 1997).During the first lysis step, 22.6 �L of lysozyme (40 mg/mL) wasadded to the samples, which were then incubated for 45 min at37 °C. Secondly, 22.6 �L of proteinase K (20 mg/mL) and 90 �L of20% SDS were added to the lysate, followed by incubation at 55 °Covernight with agitation. The aqueous phase was transferred intoa clean Eppendorf tube containing 600 �L of 6 mol/L NaCl, mixed,and centrifuged for 20 min at 16 000g. The supernatant was trans-ferred again into a clean Eppendorf tube containing 1 volume ofcold isopropanol,mixed, and stored 30min at −20 °C. Themixturewas centrifuged for 20 min at 16 000g, and the supernatant wasdiscarded. The pellet was washed with cold 70% ethanol, air-dried,and finally resuspended in 25 �L of sterile MilliQ H2O. Subse-quently, DNA integrity and quantity were checked using aNanodrop instrument (ND-1000, Nanodrop).

Detection of Flavobacterium sp. and Rhodococcus sp.Each sample from mucus, biofilm, and water was tested for

the presence of F. psychrophilum, Flavobacterium columnare, andRhodococcus sp. using a diagnostic PCR approach to complete thescreening given by the 454 pyrosequencing (described below). Theadvantage of the PCR diagnostic is that it targets the exact species,whereas assignment to the species level with pyrosequencingis dependent upon the read length. The method used to detectF. psychrophilum from mucus, biofilm, and water samples was devel-oped by Wiklund et al. (2000). The primers developed by Baderet al. (2003) to detect F. columnare and Rhodococcus sp. were used incombination with the PCR method developed to target the catAgene (Táncsics et al. 2008). The specificity of those primers wastested by analyzing 3 strains of Rhodococcus isolated from fish skinmucus. The primers were effective in detecting the specific strainof Rhodococcus contained in the probiotic solution used in thisstudy. The skinmucus and gut of dead fishwere also tested duringthe experiment as were all dead fish which showed symptoms ofCWD.

PyrosequencingEach DNA sample was PCR amplified using Takara Ex Taq pre-

mix (Fisher). All PCR reactions were performed in a final volumeof 50 �L containing 25 �L of Premix Taq, 1 �mol/L of each primer,500 ng of template, and sterile MilliQ H2O was added to bring thefinal volume to 50�L. To achieve the PCR amplifications, a generalreverse primer (R519) combinedwith B primer (Roche) was used in

Pagination not final/Pagination non finale

2 Can. J. Microbiol. Vol. 59, 2013

Published by NRC Research Press

Can

. J. M

icro

biol

. Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

Uni

vers

ity o

f L

aval

on

09/2

7/13

For

pers

onal

use

onl

y.

combination with a unique tagged forward primer (F63-targeted)combined with A primer (Roche) (Marchesi et al. 1998; Turneret al. 1999). Amplification primers were chosen to target the hy-pervariable region 1 to 3 of the 16S RNA gene (Chakravorty et al.2007). PCR conditions were as follows: after a denaturing step of30 s at 98 °C, samples were processed through 30 cycles consistingof 10 s at 98 °C, 30 s at 55 °C, and 30 s at 72 °C. The final extensionstep was done at 72 °C for 4 min 30 s. Following amplification,samples were purified using AMPure Beads (Beckman CoulterGenomics). Samples were adjusted to 100 �L with EB (Qiagen),63 �L of beads was added. Samples were mixed and incubated for5 min at RT. Using a Magnetic Particle Concentrator (MPC), thebeads were pelleted against the wall of the tube and the superna-tant was removed. The beads were washed twice with 500 �L of70% ethanol and incubated for 30 s each time. The supernatantwas removed and beads were allowed to air-dry for 5 min. Tubeswere removed from the MPC, and 24 �L of EB was added. Sampleswere vortexed to resuspend the beads. Finally, using the MPC, thebeads were pelleted against the wall once more, and the superna-tant was transferred to a new clean tube. Samples were quantifiedwith the Nanodrop instrument and mixed equally before beingsent to the Plateforme d’Analyses Biomoléculaires (Institut deBiologie Intégrative et des Systèmes, Université Laval) for sequenc-ing on a 454 GS FLX DNA Sequencer with Titanium Chemistry(Roche), according to the procedure described by the manufac-turer.

Sequence analysisThe data were analyzed in 2 steps. In the first step, CLC Genom-

ics Workbench 3.1 (CLC Bio, Aarhus, Denmark CLC work benchBIO) was used to trim amplicon sequences for quality and to re-cover the primers sequences and tags (minimum average qualityscore: 35 for a window of 50 bp, number of differences to theprimer sequence = 0, maximum number of differences to thebarcode sequence = 0, number of ambiguous base calls = 0, max-imum homopolymer length = 8). In the second step, preprocess-ing and analysis were performed using the MOTHUR software(Schloss et al. 2009). Chimeras were excluded from the data set,and amplicons were screened in all the data sets to extract thesequences reaching 300 bp to ensure an accurate taxonomic as-signment at the genus level. We used the Operational TaxonomicUnit (OTU) based method described by Costello et al. (2009). Theindex retained to assess the depth of pyrosequencing was thesequence coverage index (Good’s coverage estimator). This index

is a nonparametric index used to estimate the quality of a data set(Esty 1986). All sequences were clustered into OTUs using a 97%identity threshold, and OTUs were classified from phylum to ge-nus using the programMOTHURwith the default setting (Blast onRibosomal Database Project). To visualize similarities betweenmucus communities andwater communities, the distances betweencommunities, which were computed using Unifrac Weighted Dis-tances, were then represented using dendrograms based on ThetaYC (weighted) index because Theta YC takes into account the nor-malized count of each OTU (Yue and Clayton 2005). To visualizethe distances between communities, we used Principal Coordi-nates Analysis (PCoA) using an eigenvector-based approach (Joliffeand Morgan 1992) to represent multidimensional data in as fewdimensions as possible. Thismethod is preferred over the classicalPCA, because PCoA works with any dissimilarity measure. Fur-thermore, this methodology better handles the problem of thepresence of many double zeros in data sets.

ResultsThe probiotic addition significantly reduced the mortality in

test tanks (p < 0.001). The family S9 exhibited a very high sensitiv-ity to infection (24% mortality in control tanks), whereas the S5family was more resistant (4.4% mortality in control tanks), asdetailed previously (Boutin et al. 2012). Fish that died during theexperiment exhibited symptoms imputable either to Columnarisdisease or CWD (i.e., fin erosion, necrosis, etc.). The diagnosticPCR approach was used to accurately identify the causative agent.Only F. psychrophilum was successfully detected in 87% of the skinmucus samples from dead fishes and in all water samples, exceptfor control tanks at the end of the experiment (Table 1). In the gutfrom dead fish, the pathogen was detected in 43.4% of our sam-ples. Gut samples from control and test tanks were not equallyinfected: 60% of the dead fish from control tanks were positive forF. psychrophilum by PCR, whereas 30% of the dead fish from testtanks were positive, although this difference was not significant(p = 0.159). Flavobacterium psychrophilumwas also found at the initialstep of the experiment in mucus samples from one duplicate offamilies S5 and S9 in test tanks. After 2 weeks of treatment (controlor test), F. psychrophilum was found in all the samples from mucus.Finally, 2 weeks after the probiotic treatment, F. psychrophilum wasstill present in all control samples frommucus but only in 50% ofthe test samples from mucus. Using pyrosequencing, we notedthat the genus Flavobacterium was weakly represented in the

Table 1. Detection of Flavobacterium psychrophilum, Flavobacterium columnare, and Rhodococcus probiotic strain on samples fromskin mucus and gut of brook charr, water, biofilms, and biofilters.

Sample Condition Family F. psychrophilum F. columnareRhodococcus sp.(probiotic strain)

Dead fishes skin mucus (n = 10) Treatment S9 80% 0% 0%Dead fishes skin mucus (n = 3) Treatment S5 100% 0% 0%Dead fishes skin mucus (n = 8) Control S9 87.50% 0% 0%Dead fishes skin mucus (n = 2) Control S5 100% 0% 0%Dead fishes gut (n = 10) Treatment S9 20% 0% 0%Dead fishes gut (n = 3) Treatment S5 66.70% 0% 0%Dead fishes gut (n = 8) Control S9 62.50% 0% 0%Dead fishes gut (n = 2) Control S5 50% 0% 0%Fish skin mucus (n = 60) Treatment S9 66.70% 0% 0%Fish skin mucus (n = 60) Treatment S5 66.70% 0% 0%Fish skin mucus (n = 60) Control S9 66.70% 0% 0%Fish skin mucus (n = 60) Control S5 66.70% 0% 0%Water (n = 3) Treatment 66.70% 0% 0%Water (n = 3) Control 100% 0% 0%Biofilter (n = 3) Treatment 0% 0% 0%Biofilter (n = 3) Control 0% 0% 0%Biofilm (n = 3) Treatment 0% 0% 33%Biofilm (n = 3) Control 0% 0% 0%

Note: Presence of the bacteria was assessed by PCR methods and represented as percentage of positive samples.

Pagination not final/Pagination non finale

Boutin et al. 3

Published by NRC Research Press

Can

. J. M

icro

biol

. Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

Uni

vers

ity o

f L

aval

on

09/2

7/13

For

pers

onal

use

onl

y.

bacterial community (0.7% in the all data set). This pathogen wasnot found inmucus samples, except in one sample from family S9in a test tank at the end of the probiotic treatment. In the water,Flavobacterium was also poorly represented. It was never found intest tanks. In control tanks, it was absent at the first sampling butreached 3.9% in subsequent sampling, to finally decrease to 0.3%at the end of the experiment.

A total of 117 260 reads (mean length: 134 bp) were obtainedfrom the 31 samples through 454 pyrosequencing analysis. Afterthe process of filtering out short (<300 bp) and bad quality se-quences, 61 881 reads were kept (52% of the data set; mean length:430 bp). The relatively low rate of retrieved sequences is due to thestrong threshold of quality. Usually, the recommended lengththreshold for taxonomic assignment is 150 bp (Costello et al. 2009).According to this threshold (i.e., filtering out short (<150 bp) se-quences), 96 931 sequences (82.6% of the data set) were retrieved.Although all analyses done with this data set gave consistent re-sults for spatial (PCoA) and phylogenetic (Unifrac) relationshipsbetween samples, this low threshold of filtering greatly increasedthe uncertainty of the taxonomic assignment to the genus leveldue to shorter sequences (<150 bp) (Okubo et al. 2012). This phe-nomenon has been discussed elsewhere (Liu et al. 2007). To com-bine accurate genus assignment with community analysis, wechose a threshold of 300 bp, as recommended previously by Liuet al. (2008). To focus on themore biologically relevant taxa, OTUsrepresented by <10 reads overall were discarded for the analysiswhere taxonomic assignment was important (relative abundancein Fig. 1) but were kept for the other analyses (PCoA and Unifrac.The resulting data setwas distributed among 38 genera and 5 groupsor phyla. The Good’s coverage estimations ranked between 69%and 100%.

The results from pyrosequencing indicated that only one genuswas detected in the mix of probiotics, namely Rhodococcus (100%of the reads from the probiotic sample). The genus Rhodococcuswas also detected by pyrosequencing in one mucus samplefrom the family S9 in a test tank at the end of the treatment.There were 2 strains of Rhodococcus in the probiotic mix. Amongthe 3 Rhodococcus strains isolated from fish skin mucus, only onestrain from the 2 used as probiotic in our experimentwas detectedby PCR. This Rhodococcus strain was detected with the PCRmethodonly in the biofilm of test tanks at the end of the addition ofprobiotic but not in any fish or water samples (Table 1). Further-more, we validated this taxonomic assignment by culturing thisbacterium from the mix of probiotic and resequencing it. Thestrain was assigned to Rhodococcus genus based on a 1500 bp frag-ment of 16S rDNA.

The bacterial community of fish skin mucus was quite stableexcept for the family S9 sample at the first sampling in thecontrol unit (CAS91) (Fig. 1). Samples from mucus were domi-nated by the occurrence of the genus Sphingomonas, whichreached a mean of 91.6% of the community. The other majorgenera were Pseudomonas, Pseudoxanthomonas, Kerstersia, Dichelobacter,and Propionibacterium (2.7%, 1.4%, 0.9%, 0.5%, and 0.4%). Bacterialcommunities from the water were more dynamic during the ex-periment even in control tanks in which no probiotic were added.Bacterial communities from the water samples also showed morediversity than fish skin mucus microbiome. In general, the mostabundant genus was again Sphingomonas (47.1%), followed byLegionnella (16.5%), Smithella (15.6%), Microvirga (4%), Dinoroseobacter(2.2%), Curvibacter (1.8%), Hydrocarboniphaga (1.8%), Kestersia (1.6%),Pseudomonas (1.4%), and Kordiimonas (1.2%).

PCoA shows that our data are highly dimensional (30 dimen-sions), but the first and second axes represent 44.6% and 16.2% ofthe variation (60.8% total) for the theta YC distances. Furthermore,the R2 value (0.978) between the original distance matrix and thedistance between the points in 2D PCoA space indicates that theycorrelate closely. As visualized in the PCoA plot, all the samplesfrom mucus clustered in the same groups except for, once again,

the sample coming from the family S9 at the first sampling in thecontrol unit (CAS91) (Fig. 2). At the first sampling interval, watersamples are very similar tomucus samples (TEAUA1 and CEAUA1).The bacterial communities from water samples in control tanksfor the second and third sampling times (CEAUA2 and CEAUA3)grouped into a second cluster with the water sample in test tanksfor second sampling time (TEAUA2). The dendrogram represent-ing the phylogenetic distances between the bacterial communi-ties shows exactly the same results as the PCoA (Fig. 3). All thesamples fromfishmucus are phylogenetically close and related tothe water samples from the first sampling time (CEAU1 andTEAU2). The only exception was the CAS91 sample and the watersamples of the second and third sampling times. The probioticmix, as it was composed of only one genus that was not recoveredin the other samples, forms an outgroup.

DiscussionThe aim of this study was to monitor the effect of a probiotic

treatment on both the natural microbiota of brook charr skinmucus and the surrounding water microflora. Using 454 pyro-sequencing, the variation in abundance of the different bacterialstrains was screened to test whether the treatment using a probi-otic candidate (Rhodococcus sp.) would disturb the fish skin mucusmicrobiota and (or) the surrounding natural microflora.

The intimate relationship between brook charr and the genusSphingomonas seems to be a basis for the homeostasis of the skinmucus microbiome and protection against pathogens. The analy-sis of bacterial communities present in fish mucus showed thatbacterial communities were stable through the 3 sampling timesand across samples from different families. Thismay be explainedby the high abundance of the genus Sphingomonas, which repre-sents 91.6% of the bacterial communities. This genus is widelydistributed in freshwater environments and has already beenisolated from environments with high concentrations of organicnutrients, such as blood and sputum (Hsueh et al. 1998). It istherefore not surprising to detect this bacterial genus in fish skinmucus. Sphingomonas is also known for consuming a wide range oforganic compounds and to exhibit remarkable biodegradativeand biosynthetic capabilities (Balkwill et al. 2006). Suchmetabolicabilities are correlated with the domination of Sphingomonasalaskensis in marine picoplankton (Eguchi et al. 2001; Vancanneytet al. 2001). Therefore, a high competitiveness for nutrients couldexplain how Sphingomonas present in skin mucus would become aprotective barrier against other environmental bacteria. The ben-eficial effect of Sphingomonas for the host could also be related tointerference competition. Indeed, another species of the samegenus, Sphingomonas paucimobilis, is known to inhibit the growth offungi in plants (White et al. 1996). We can hypothesize that brookcharr coevolved with this bacterial genus to ensure protection bythe bacteria in exchange of nutrients offered by the host (Sachset al. 2004).

The bacterial communities from the mucus and water sharedmany species in common. The proximity between bacterial com-munity present in mucus and water samples is notable. Prior toprobiotic treatment, PCoA analysis and Unifrac dendrogramshowed considerable similarity between the 2. This is in accor-dance with the hypothesis that skin mucus microbiota resultsfrom the colonization by strains inhabiting the surrounding wa-ter (Cahill 1990). Here, the results showed that this proximity wasdisturbed by the appearance of the disease and the increase ofF. psychrophilum abundance in control tanks.

The decrease of the disease was not mediated by the coloniza-tion of the skin mucus by the probiotic as expected. Surprisingly,the probiotic strain was neither detected in fish skinmucus nor inwater samples but rather only in the biofilm of test tanks, at theend of the treatment. The absence of the probiotic from fishskin mucus can be explained by the inability of our probiotic to

Pagination not final/Pagination non finale

4 Can. J. Microbiol. Vol. 59, 2013

Published by NRC Research Press

Can

. J. M

icro

biol

. Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

Uni

vers

ity o

f L

aval

on

09/2

7/13

For

pers

onal

use

onl

y.

Fig. 1. Bacterial composition and structure at the genus level of fish skin mucus, water samples, and probiotic solution (Prob).

Pagination not final/Pagination non finale

Boutin et al. 5

Published by NRC Research Press

Can

. J. M

icro

biol

. Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

Uni

vers

ity o

f L

aval

on

09/2

7/13

For

pers

onal

use

onl

y.

out-compete the bacteria already present enough to firmly estab-lish a presence in these environments. This strain was isolatedfrom skin mucus of brook charr, thus demonstrating its ability tolive on mucus and its competence for adhesion. However, theprobiotic strain was isolated from another brook charr familywhere Sphingomonas was not the most abundant genus present.Hypothetically, this probiotic strain may not be coadapted withSphingomonas and, in turn, is unable to colonize the mucus. Asidefrom this, the absence of the probiotic in water suggests either alow survival of the probiotic strain in open water or quick sedi-mentation or a biofilm formation. Furthermore, the probioticstrain was not detected in the gut from dead fish in test tankseither. This suggests that this Rhodococcus strain was not able tocolonize the gut. The survival differences recorded between testsand the control tanks were significant (mean reduction of 47%death, p< 0.001) and correlated (p< 0.001) with the addition of thisprobiotic strain in test tanks (Boutin et al. 2012). Surprisingly, theresults indicate that this beneficial effect was not correlated witha colonization of skin mucus by probiotic strains that were iso-lated from fish skinmucus. In this respect, no community changewas observed in the skin mucus microbiota between the controland test tank samples. As the study is dealing with a naturallyoccurring infection episode, we acknowledge that we had nocontrol over pathogen load. However, the quantification of

F. psychrophilum in the water tank at the initial step of the experi-ment showed the same amount of pathogen in the water. Thisresult allows us to confirm that difference observed in mortalitybetween the test and the control is not due to the highestprevalence of the pathogen in the tank. Interestingly, in thesurrounding water of control tanks, we observed an increase ofF. psychrophilum abundance during the second sampling time thatwas significantly correlated with fish mortality (Pearson correla-tion, p < 0.001). Such a finding makes sense because the diseasesymptoms encountered in this study were clearly associated withCWD.

The action of the probiotic is localized in the surrounding en-vironment. The localization of the probiotic in the biofilm of thetank associated with the decrease of F. psychrophilum in the watergives us some clues about the modus operandi. It is likely that thecolonization of the biofilmby the probiotic prevents the pathogenusing this niche as a second reservoir. The probiotic interruptsthis life cycle and indirectly prevents colonization of themucus. Itis already known that F. psychrophilum cannot survive long in wa-ter, and shedding of F. psychrophilum by infected fish has previ-ously been shown to be correlated with mortality (Madetoja et al.2000). In test tanks, no such increase was observed after the addi-tion of the probiotic strain, so we can conclude that the probioticcontrolled the growth of F. psychrophilum in the surroundingwater

Fig. 2. PCoA analysis of bacterial communities of fish skin mucus and water samples.

Pagination not final/Pagination non finale

6 Can. J. Microbiol. Vol. 59, 2013

Published by NRC Research Press

Can

. J. M

icro

biol

. Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

Uni

vers

ity o

f L

aval

on

09/2

7/13

For

pers

onal

use

onl

y.

but not on skinmucus, aswe first hypothesized. This conclusion isalso supported by the fact that the third water sample from testtanks (TEAUA3) tightly clustered with the mucus samples and theinitial water samples in the dendrogram and the PCoA, whereasthe water sample from the control tank was differentiated fromthe initial sample. This result suggests that the probioticmitigatesmucus colonization by the pathogens by controlling the growth ofF. psychrophilum in the water. Furthermore, it was recently discov-ered that pathogens from the genus Flavobacterium can increasetheir survival in the system by colonizing the biofilm of the tank,therefore, staying available for infection longer (Sundell andWiklund 2011; Cai et al. 2013). As hypothesized in the in vitrostudy, the mechanism of action is probably an exclusion via acompetition for the nutrients and (or) a synthesis of antimicrobial

compounds (Verschuere et al. 2000). Furthermore, the proportion ofgeneralized infection (i.e., infection in the gut by F. psychrophilum)tends to be lower with addition of probiotic (60% of infection inthe control samples versus 30.8% with the treatment). Such antag-onistic action would in turn favor the resilience of the skinmucusmicrobiota.

The host genotype also has a role in the relationship among thehost, the commensal microbiome, and the pathogen. It remainsunclear why we observed that the beneficial effect of this probi-otic strain had different impacts on fish mortality depending onthe family (Boutin et al. 2012). These 4 families are more or lesssensitive to the treatment, and consequently, this result does notfit with the direct effect on surrounding water. This nonadditiveprobiotic effect may be imputable to the host genotype. It is

Fig. 3. Dendrogram analysis based on Theta YC index of fish skin mucus, water samples, and probiotic solution. Symbols in parenthesesindicate the type of samples and the condition: &, mucus samples (control); #, water samples (control); +, mucus samples (treatment); *, watersamples (treatment). For mucus samples, labels indicate the family (S5–S9) and the time of sampling (1 (17 February), 2 (10 March),3 (24 March)). Water samples are labeled only with time sampling (1, 2, 3) and probiotic solution is labeled Prob1.

Pagination not final/Pagination non finale

Boutin et al. 7

Published by NRC Research Press

Can

. J. M

icro

biol

. Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

Uni

vers

ity o

f L

aval

on

09/2

7/13

For

pers

onal

use

onl

y.

already known that S. fontinalis families differ in their response toa given stress (Crespel et al. 2011). Then, a recent study revealedthat the genetic background in rainbow trout (Oncorhynchus mykiss)is an important factor in the susceptibility to F. psychrophilum,particularly differences in the host transcriptome response(Langevin et al. 2012). Ultimately, the host genotype was observedto control endogenous bacterial strains in mice (McKnite et al.2012) and brook charr (Boutin et al. in revision). Therefore, themost likely factors explaining mortality differences betweenthose families are inherent to the genetics of the host, whichunderlies both immune response andmicrobiota functional prop-erties.

In conclusion, the probiotic strain Rhodococcus sp. used in thisexperiment did not disturb the natural microflora of fish skinmucus. Our results indicate that the beneficial effect on fish sur-vival (mean diminution of 47% death) was most likely associatedwith an improvement of the surrounding water quality. Theseresults indicate that the probiotic strain used is harmless to thefish and is a promising candidate for aquaculture. The likelymechanism is the probiotic outcompeting the pathogenic bac-teria in the biofilm, which acts as a secondary habitat (like asecondary host) for later colonization of the mucus. However,this conclusion has to be taken with a caution because it remainsunclear how the host can respond differently if there is no colo-nization. To provide a better understanding of this question, weneed to understand the biology of the probiotic strain Rhodococcussp. by analyzing further its antimicrobial functions againstF. psychrophilum both in vitro and further characterizing the hostgenetic basis of resistance to this bacterium.

AcknowledgementsWe thank S. Pavey and M. Llewellyn for critical reading of the

manuscript. This work was financially supported by a Partnershipgrant (Strategic program) from theNatural Sciences and Engineer-ing Research Council of Canada (NSERC) to Louis Bernatchez,Nicolas Derome, and Céline Audet, and from the CollaborativeResearch and Training Experience Program (CREATE). This is acontribution to the research program of Réseau AquacultureQuébec (RAQ). The authors are very grateful to the staff of theLaboratoire Regional des Sciences Aquatiques for their invaluablehelp in developing the biofilters and culturing bacterial strainsused in this experiment. Thanks are extended to Louis Bernatchez(Laval University, Québec, Canada) for critical reading of the man-uscript.

ReferencesAljanabi, S.M., and Martinez, I. 1997. Universal and rapid salt-extraction of high

quality genomic DNA for PCR-based techniques. Nucleic Acids Res. 25(22):4692–4693. doi:10.1093/nar/25.22.4692. PMID:9358185.

Bader, J.A., Shoemaker, C.A., and Klesius, P.H. 2003. Rapid detection of colum-naris disease in channel catfish (Ictalurus punctatus) with a new species-specific 16-S rRNA gene-based PCR primer for Flavobacterium columnare.J. Microbiol. Methods, 52(2): 209–220. doi:10.1016/S0167-7012(02)00208-7.PMID:12459241.

Balcázar, J., de Blas, I., Ruiz-Zarzuela, I., Vendrell, D., and Muzquiz, J. 2004.Probiotics: a tool for the future of fish and shellfish health management.J. Aquac. Trop. 19: 239–242.

Balcázar, J., Blas, I., Ruizzarzuela, I., Cunningham, D., Vendrell, D., andMuzquiz, J. 2006. The role of probiotics in aquaculture. Vet. Microbiol.114(3–4): 173–186. doi:10.1016/j.vetmic.2006.01.009. PMID:16490324.

Balkwill, D., Fredrickson, J., and Romine, M. 2006. Sphingomonas and relatedgenera. In The prokaryotes. Edited by M. Dworkin, S. Falkow, E. Rosenberg,K.-H. Schleifer, and E. Stackebrandt. Springer, New York, USA. pp. 605–629.

Boutin, S., Bernatchez, L., Audet, C., and Derôme, N. 2012. Antagonistic effect ofindigenous skin bacteria of brook charr (Salvelinus fontinalis) against Flavobac-terium columnare and F. psychrophilum. Vet. Microbiol. 155(2–4): 355–361. doi:10.1016/j.vetmic.2011.09.002. PMID:21958747.

Cahill, M.M. 1990. Bacterial flora of fishes: a review. Microb. Ecol. 19(1): 21–41.doi:10.1007/BF02015051.

Cai, W., De La Fuente, L., and Arias, C.R. 2013. Biofilm formation by the fishpathogen Flavobacterium columnare: development and parameters affect-ing surface attachment. Appl. Environ. Microbiol. 79(18): 5633–5642. doi:10.1128/AEM.01192-13.

Chakravorty, S., Helb, D., Burday, M., Connell, N., and Alland, D. 2007. A detailedanalysis of 16S ribosomal RNA gene segments for the diagnosis of pathogenicbacteria. J. Microbiol. Methods, 69(2): 330–339. doi:10.1016/j.mimet.2007.02.005. PMID:17391789.

Costello, E.K., Lauber, C.L., Hamady, M., Fierer, N., Gordon, J.I., and Knight, R.2009. Bacterial community variation in human body habitats across spaceand time. Science, 326(5960): 1694–1697. doi:10.1126/science.1177486. PMID:19892944.

Courvalin, P. 2006. Antibiotic resistance: the pros and cons of probiotics. Digest.Liver Dis. 38(Suppl. 2): S261–265. doi:10.1016/S1590-8658(07)60006-1. PMID:17259087.

Crespel, A., Bernatchez, L., Garant, D., and Audet, C. 2011. Quantitative geneticanalysis of the physiological stress response in three strains of brook charrSalvelinus fontinalis and their hybrids. J. Fish Biol. 79(7): 2019–2033. doi:10.1111/j.1095-8649.2011.03149.x. PMID:22141902.

Darch, S.E., West, S.A., Winzer, K., and Diggle, S.P. 2012. Density-dependentfitness benefits in quorum-sensing bacterial populations. Proc. Natl. Acad.Sci. U.S.A. 109(21): 8259–8263. doi:10.1073/pnas.1118131109. PMID:22566647.

Eguchi, M., Ostrowski, M., Fegatella, F., Bowman, J., Nichols, D., Nishino, T., andCavicchioli, R. 2001. Sphingomonas alaskensis strain AFO1, an abundant oligotro-phic ultramicrobacterium from the North Pacific. Appl. Environ. Microbiol.67(11): 4945–4954. doi:10.1128/AEM.67.11.4945-4954.2001. PMID:11679312.

Esty, W.W. 1986. The efficiency of Good’s nonparametric coverage estimator.Ann. Stat. 14(3): 1257–1260. doi:10.1214/aos/1176350066.

FAO/WHO. 2001. Health and nutritional properties of probiotics in food includ-ing powder milk with live lactic acid bacteria. Report of a joint FAO/WHOexpert consultation on evaluation of health and nutritional properties ofprobiotics in food including powder milk with live lactic acid bacteria. FAO/WHO, Cordoba, Argentina.

Fuller, R. 1995. Probiotics: their development and use. InOld Herborn UniversitySeminar Monograph 8. Probiotics: Prospects of Use in Opportunistic Infec-tions. Edited by R. Fuller, P.J. Heidt, V. Rusch, and D. Van Der Waaij. Institutefor Microbiology and Biochemistry, Herborn-Dill, Germany. pp. 1–8.

Gonzalez, A., Clemente, J.C., Shade, A., Metcalf, J.L., Song, S., Prithiviraj, B.,Palmer, B.E., and Knight, R. 2011. Our microbial selves: what ecology canteach us. EMBO Rep. 12(8): 775–784. doi:10.1038/embor.2011.137. PMID:21720391.

Hsueh, P.-R., Teng, L.-J., Yang, P.-C., Chen, Y.-C., Pan, H.-J., Ho, S.-W., and Luh, K.-T.1998. Nosocomial infections caused by Sphingomonas paucimobilis: clinical fea-tures and microbiological characteristics. Clin. Infect. Dis. 26(3): 676–681.doi:10.1086/514595. PMID:9524843.

Joliffe, I.T., and Morgan, B.J. 1992. Principal component analysis and explor-atory factor analysis. Stat. Methods Med. Res. 1(1): 69–95. doi:10.1177/096228029200100105. PMID:1341653.

Langevin, C., Blanco, M., Martin, S.A.M., Jouneau, L., Bernardet, J.-F., Houel, A.,Lunazzi, A., Duchaud, E., Michel, C., Quillet, E., and Boudinot, P. 2012. Tran-scriptional responses of resistant and susceptible fish clones to the bacterialpathogen Flavobacterium psychrophilum. PLoS One, 7(6): e39126. doi:10.1371/journal.pone.0039126. PMID:22720048.

Liu, Z., Lozupone, C., Hamady, M., Bushman, F.D., and Knight, R. 2007. Shortpyrosequencing reads suffice for accurate microbial community analysis.Nucleic Acids Res. 35(18): e120–e120. doi:10.1093/nar/gkm541. PMID:17881377.

Liu, Z., DeSantis, T.Z., Andersen, G.L., and Knight, R. 2008. Accurate taxonomyassignments from 16S rRNA sequences produced by highly parallel pyrose-quencers. Nucleic Acids Res. 36(18): e120. doi:10.1093/nar/gkn491. PMID:18723574.

Livia, L., Antonella, P., Hovirag, L., Mauro, N., and Panara, F. 2006. A nondestruc-tive, rapid, reliable and inexpensive method to sample, store and extracthigh-quality DNA from fish body mucus and buccal cells. Mol. Ecol. Notes,6(1): 257–260. doi:10.1111/j.1471-8286.2005.01142.x.

Madetoja, J., Nyman, P., and Wiklund, T. 2000. Flavobacterium psychrophilum, in-vasion into and shedding by rainbow trout Oncorhynchus mykiss. Dis. Aquat.Organ. 43(1): 27–38. doi:10.3354/dao043027. PMID:11129378.

Marchesi, J.R., Sato, T., Weightman, A.J., Martin, T.A., Fry, J.C., Hiom, S.J., andWade, W.G. 1998. Design and evaluation of useful bacterium-specific PCRprimers that amplify genes coding for bacterial 16S rRNA. Appl. Environ.Microbiol. 64(2): 795–799. PMID:9464425.

McKnite, A.M., Perez-Munoz, M.E., Lu, L., Williams, E.G., Brewer, S.,Andreux, P.A., Bastiaansen, J.W.M., Wang, X., Kachman, S.D., Auwerx, J.,Williams, R.W., Benson, A.K., Peterson, D.A., and Ciobanu, D.C. 2012. Murinegut microbiota is defined by host genetics and modulates variation of meta-bolic traits. PLoS One, 7(6): e39191. doi:10.1371/journal.pone.0039191. PMID:22723961.

Merrifield, D.L., Dimitroglou, A., Foey, A., Davies, S.J., Baker, R.T.M., Bøgwald, J.,Castex, M., and Ringø, E. 2010. The current status and future focus of probi-otic and prebiotic applications for salmonids. Aquaculture, 302(1–2): 1–18.doi:10.1016/j.aquaculture.2010.02.007.

Mohamed, M.H., and Ahmed Refat, N.A.G. 2011. Pathological evaluation of pro-biotic, Bacillus subtilis, against Flavobacterium columnare in tilapia nilotica(Oreochromis niloticus) fish in Sharkia Governorate, Egypt. Am. J. Sci. 7(2): 244–256.

Morris, J.J., Lenski, R.E., and Zinser, E.R. 2012. The black queen hypothesis:

Pagination not final/Pagination non finale

8 Can. J. Microbiol. Vol. 59, 2013

Published by NRC Research Press

Can

. J. M

icro

biol

. Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

Uni

vers

ity o

f L

aval

on

09/2

7/13

For

pers

onal

use

onl

y.

evolution of dependencies through adaptive gene loss. mBio 3(2): e00036–12.doi:10.1128/mBio.00036-12. PMID:22448042.

Okubo, T., Ikeda, S., Yamashita, A., Terasawa, K., and Minamisawa, K. 2012.Pyrosequence read length of 16S rRNA gene affects phylogenetic assignmentof plant-associated bacteria. Microbes Environ. 27(2): 204–208. doi:10.1264/jsme2.ME11258. PMID:22791055.

Roberts, C., Mao-Jones, J., Ritchie, K.B., Jones, L.E., and Ellner, S.P. 2010. Howmicrobial community composition regulates coral disease development.PLoS Biol. 8(3): e1000345. doi:10.1371/journal.pbio.1000345. PMID:20361023.

Robinson, C.J., Bohannan, B.J.M., and Young, V.B. 2010. From structure to func-tion: the ecology of host-associated microbial communities. Microbiol. Mol.Biol. Rev. 74(3): 453–476. doi:10.1128/MMBR.00014-10. PMID:20805407.

Rumbaugh, K.P., and Kaufmann, G.F. 2012. Exploitation of host signaling path-ways by microbial quorum sensing signals. Curr. Opin. Microbiol. 15(2): 162–168. doi:10.1016/j.mib.2011.12.003. PMID:22204809.

Sachs, J.L., Mueller, U.G., Wilcox, T.P., and Bull, J.J. 2004. The evolution of coop-eration. Q. Rev. Biol. 79(2): 135–160. doi:10.1086/383541. PMID:15232949.

Schloss, P., Westcott, S., Ryabin, T., Hall, J., Hartmann, M., Hollister, E.,Lesniewski, R., Oakley, B., Parks, D., Robinson, C., Sahl, J., Stres, B.,Thallinger, G., Van Horn, D., andWeber, C. 2009. Introducing mothur: open-source, platform-independent, community-supported software for describ-ing and comparingmicrobial communities. Appl. Environ. Microbiol. 75(23):7537–7541. doi:10.1128/AEM.01541-09. PMID:19801464.

Starliper, C.E. 2011. Bacterial coldwater disease of fishes caused by Flavobacteriumpsychrophilum. J. Adv. Res. 2(2): 97–108. doi:10.1016/j.jare.2010.04.001.

Sundell, K., and Wiklund, T. 2011. Effect of biofilm formation on antimicrobialtolerance of Flavobacterium psychrophilum. J. Fish Dis. 34(5): 373–383. doi:10.1111/j.1365-2761.2011.01250.x. PMID:21488905.

Táncsics, A., Szoboszlay, S., Kriszt, B., Kukolya, J., Baka, E., Márialigeti, K., andRévész, S. 2008. Applicability of the functional gene catechol 1,2-dioxygenaseas a biomarker in the detection of BTEX-degrading Rhodococcus species. J. Appl.Microbiol. 105(4): 1026–1033. doi:10.1111/j.1365-2672.2008.03832.x. PMID:18479346.

Turner, S., Pryer, K.M., Miao, V.P.W., and Palmer, J.D. 1999. Investigating deep

phylogenetic relationships among Cyanobacteria and plastids by small sub-unit rRNA sequence analysis. J. Eukaryot. Microbiol. 46(4): 327–338. doi:10.1111/j.1550-7408.1999.tb04612.x. PMID:10461381.

van Valen, L. 1973. A new evolutionary law. Evolutionary Theory, 1: 1–30.Vancanneyt, M., Schut, F., Snauwaert, C., Goris, J., Swings, J., and Gottschal, J.C.

2001. Sphingomonas alaskensis sp. nov., a dominant bacterium from a marineoligotrophic environment. Int. J. Syst. Evol. Microbiol. 51(1): 73–79. doi:10.1099/00207713-51-1-73. PMID:11211276.

Verschuere, L., Rombaut, G., Sorgeloos, P., and Verstraete, W. 2000. Probioticbacteria as biological control agents in aquaculture. Microbiol. Mol. Biol. Rev.64(4): 655–671. doi:10.1128/MMBR.64.4.655-671.2000. PMID:11104813.

Villamil, L., Figueras, A., Planas, M., and Novoa, B. 2003. Control of Vibrio algino-lyticus in Artemia culture by treatment with bacterial probiotics. Aquaculture,219(1–4): 43–56. doi:10.1016/S0044-8486(02)00515-X.

Vine, N.G., Leukes, W.D., and Kaiser, H. 2004. In vitro growth characteristics offive candidate aquaculture probiotics and two fish pathogens grown in fishintestinal mucus. FEMS Microbiol. Lett. 231(1): 145–152. doi:10.1016/S0378-1097(03)00954-6. PMID:14769479.

Wang, Y., Li, J., and Lin, J. 2008. Probiotics in aquaculture: challenges and out-look. Aquaculture, 281(1–4): 1–4. doi:10.1016/j.aquaculture.2008.06.002.

White, D.C., Sutton, S.D., and Ringelberg, D.B. 1996. The genus Sphingomonas:physiology and ecology. Curr. Opin. Biotechnol. 7(3): 301–306. doi:10.1016/S0958-1669(96)80034-6. PMID:8785434.

Wiklund, T., Madsen, L., Bruun, M.S., and Dalsgaard, I. 2000. Detection ofFlavobacterium psychrophilum from fish tissue and water samples by PCR am-plification. J. Appl. Microbiol. 88(2): 299–307. doi:10.1046/j.1365-2672.2000.00959.x. PMID:10735999.

Yue, J.C., and Clayton, M.K. 2005. A similarity measure based on species propor-tions. Commun, Stat. Theory Methods, 34(11): 2123–2131. doi:10.1080/STA-200066418.

Pagination not final/Pagination non finale

Boutin et al. 9

Published by NRC Research Press

Can

. J. M

icro

biol

. Dow

nloa

ded

from

ww

w.n

rcre

sear

chpr

ess.

com

by

Uni

vers

ity o

f L

aval

on

09/2

7/13

For

pers

onal

use

onl

y.