Aquaculture

ELSEVIER Aquaculture 193 (2001) 25-37

www.elsevier.nl/locate 1 aqua-online

Search for beneficial bacterial strains for turbot

(Scophthalmus maximus L.) larviculture

L. Huys a,b,*,P. Dhert a, R. Robles a, F. Ollevier c, P. Sorgeloos a,J. Swings c,d

a Laboratory of Aquaculture and Artemia Reference Center, Ghent University, Rozier 44, 9000 Gent, Belgium

b Laboratory of Microbiology, Ghent University, K.L. - Ledeganckstraat 35. 9000 Gent. Belgium

C Laboratory of Aquatic Ecology, Zoological Institute, Catholic University of Lauvain. De Beriotstraat 32.

3000 Louvain, Belgium

d BCCM 1 lMG Culture Collection. Ghent University, K.L. - Ledeganckstraat 35, 9000 Gent, Belgium

Received 5 January 2000; received in revised form 10 July 2000; accepted 21 July 2000

Abstract

The aerobic bacterial flora in the gut of turbot larvae and their influence on larval survival wasexamined. Two turbot experiments were run with six replicates each time. Large variation, from0% up to 44%, was observed in the survival percentage of turbot larvae. There was no correlationbetween the number of bacteria present in the gut of turbot larvae and the larval survival rate.During both experiments, all replicates followed nearly the same rate of bacterial development inthe gut of turbot larvae going from circa 102 CFU larva- I just before first feeding at day 3 posthatchto 105CFUlarva- I at day 9 post hatch.

In total, 127 bacterial isolates from 12 rearing tanks were sampled for further investigation.Based on their fatty acid profile obtained by FAME-analysis, and using principal componentanalysis, the isolates were subdivided in 12 major gaschromatographic-groups or clusters (GC-groups), 11 isolates remained unclustered. Four specific GC-groups (namely cluster A, B, I and J)were selected as potential beneficial bacteria for turbot larviculture as the majority of the isolatesof these clusters derived from rearing tanks with a survival percentage higher than 35%.Representative isolates of these clusters were screened on their ability to enhance the survival rateas well as the poor reproducibility in larval survival in a small-scale turbot confrontation test.Also, a Vibrio mediterranei Q40 strain, isolated from sea bream larvae, was included in thesesmall-scale confrontation tests. Only cluster A and the V. mediterranei Q40 strain had a distinctpositive and reproducible effect on larval survival. In conclusion, cluster A and V. mediterranei

. Corresponding author. Tel.: +32-9-264-3754; fax: +32-9-264-4193.

E-mail address:patrick.sorgeloos@rug.ac.be (L. Huys).

0044-8486/01/$ - see front matter @200l Elsevier Science B.V. All rights reserved.PH: S0044-8486(00)00474-9

Reprinted with kind permission from Elsevier Science BV

26 L. Huys et al. / Aquaculture 193 (200]) 25-37

Q40seemed(0playa roleasfirst coloniserof thegutof turbotlarvaeandcouldpreventthecolonisationof the gut by opportunisticbacteria. @2001 Elsevier Science B.Y. All rightsreserved.

Keywords: Turbot larvae; Bacteria; Microbial management; Probiotic

1. Introduction

Considerable progress has been accomplished in the commercial culture of turbot(Scophthalmus maximus L.) but larval survival remains unpredictable, especially, in thesecond week post hatch when mass mortalities are observed (Planas, 1994; Ring!/!andVadstein, 1998; Ring!/!and Birkbeck, 1999). Experiments have suggested that thesemortalities may be due to inadequate microbial conditions (Minkoff and Broadhurst,1994), however, most of these mortalities could not be associated with primary orspecific pathogens but rather with opportunistic bacteria that attack the host larvae understress conditions (Olafsen, 1993; Vadstein et aI., 1993; Munro et aI., 1995; Verdonckand Swings, 1995).

Since fish larvae establish their bacterial flora partly in a non-selective way (Hansenand Olafsen, 1990; Cahill, 1990), the initial bacterial environment is of utmost impor-tance. In this respect, the early colonisation of the gut by non-opportunistic bacteria mayinitiate a resident microflora which could prevent the proliferation and colonisation ofthe gut of larvae by opportunistic and/or pathogenic bacteria (Westerdahl et aI., 1994;Bergh, 1995; Skjermo et aI., 1997). Colonisation of the digestive tract with beneficial or"probiotic" micro-organisms is a well-recognised practice in veterinary medicine(Vanbelle et aI., 1990) and this concept may be used in larval rearing. Ring!/!andVadstein (1998) were able to colonise the gut of turbot larvae with Vibriopelagius whenthe bacterial species was added 2 days after hatching but lower densities were observedwhen the larvae were exposed to V. pelagius at day 5 or 8 post hatch. Therefore, it wasconcluded that V. pelagius has to compete with the microbiota already present in thelarval gut for suitable attachment sites.

In the present research, the relationship between the intestinal bacterial flora of turbotlarvae and larval survival rates was studied. Possible beneficial bacterial strains wereisolated to investigate their effect on larval survival. The overall goal was to identifybeneficial bacterial strains that may improve the hatchery output in terms of repro-ducibility and larval survival rates.

2. Materials and methods

2.1. Larval rearing conditions

Two turbot experiments (Exp. I and Exp. 11),in whom the bacterial flora of the larvalgut was sampled, were performed under the same larval rearing conditions but with adifference in time. Both experiments were composed of six replicates: Exp. I included

L. Huys et al. / Aquaculture 193 (2001) 25-37 27

tanks 1 to 6 and the following executed Exp. 11 included tanks 7 to 12. The twoexperiments were run until 11 days post hatch during which the water was not renewed.The use of a stagnant system during the early rotifer-feeding stage of larval turbot is ausual practice in most hatchery facilities as newly hatched turbot larvae are very smalland fragile (personal communication, Ringj1let aI., 1996; Skjermo and Vadstein, 1999).

For each experiment, 1-day-old turbot larvae, deriving from a same spawn, wereobtained from a commercial hatchery (France Turbot, France) and acclimated toseawater (35 ppt, 18°C) during a few hours. After acclimatisation, the larvae werestocked in 60-1black conical tanks at a density of 60 larvae I-I and reared following theprocedure of Dhert et al. (1993). Tanks were filled with seawater that was filteredthrough a membrane filter with a pore size of 0.2 j.Lm,disinfected with NaOCl (10ppm), aerated overnight and then neutralised by addition of sodium-thiosulphate.Rotifers (Brachionus plicatilis) were used as a live food source for larval turbotfollowing the method described by Dhert (1996). The survival of each tank wasevaluated at the end of the experiment on day 11.

2.2. Sampling and isolation of the gut microflora of larval turbot

Samples of the aerobic flora of the intestine of 50 turbot larvae were taken at days 3,5, 7 and 9 post hatch. As the small size of turbot larvae renders dissection impractical,the aerobic bacterial flora of the gut was sampled according to the method of Muroga etal. (1987). At each sampling, 50 larvae from each tank were captured with a sterilepipette and placed into a sterile beaker.

To remove the surface bacteria, the larvae were caught on a sterile mesh (10 j.Lm)andsuccessively anaesthetised by immersion in a 0.1% (wIv) benzocaine solution for 10 s,disinfected in a 0.1% (wI v) benzalkoniumchloride solution for 10 s and then rinsedthree times in a sterile Nine Salt Solution (NSS; Olssen et aI., 1992), each rinse for 5 s.The larvae were aseptically transferred to a sterile plastic bag containing 25 ml of NSSand homogenised in a stomacher blender (400SN, Seward Medical, London, UK).Dilutions of the homogenised sample solutions (10- I and 10- 2) were prepared usingsterile NSS and 0.1 ml volumes were plated out on Marine Agar 2216 (MA; DifcoLaboratories, Detroit, MI, USA) and on Thiosulphate-Citrate-Bile salt-Sucrose Agar(TCBS; Oxoid, Basingstoke, Hampshire, England). TCBS-agar was used as a selectivemedium for isolation of Vibrios rather than for quantification. All the plates wereincubated at 18°C for 5 days. After incubation, all different colony types obtained fromthe first dilution, both on MA and TCBS, were isolated and further purified on MA.Pure strains were used for further characterisation.

2.3. Microbiological characterisation techniques

2.3.1. Gas chromatographic analysis of cellular fatty acid methyl esters (FAME)Quantitative analysis of cellular fatty acid compositions was performed using the

gas-liquid chromatographic procedure as described by De Boer and Sasser (1986).Strains were grown for 24 h at 28°C on MA (Difco). Approximately 70 mg of cells wereadded to 1 ml of 3.75 N NaOH in 50% aqueous methanol and heated for 30 min in a

28 L. Huys et al. / Aquaculture 193 (200]) 25-37

boiling waterbathfor saponification.Methylationwas achievedby adding2 ml of 6 Nhydrochloric acid in aqueous methanol and heating for 10 min at 80°C. After cooling toroom temperature, fatty acid methyl esters were extracted with a mixture of hexane andmethyl-iso-butylether. Fatty acid methyl esters were analysed with a Hewlett-Packardmodel 5898A gaschromatograph and identified using MIS software (Microbial ID,Newark, DE, USA). The isolates were compared and grouped into gaschromatographicgroups (GC-groups or clusters) on the basis of the fatty acid fingerprints by principalcomponent analysis (PCA) using the same software package. PCA represents eachoperational unit (OUT) as a point in a multidimensional space and the relationshipsbetween the OUT's are represented by the Euclidean distances between the representa-tive points (Dunn and Everitt, 1982). Clusters were delineated at 90% similarity andisolates being part of the same cluster, were considered belonging to the same species.After grouping into clusters, two or more representatives of each cluster were selectedfor FAME analysis with Tryptic Soy Agar (TSA, Difco Laboratories, Detroit, MI, USA)as culture medium instead of MA. In that way it was possible to compare the FAMEfingerprints of the unknown isolates with the FAME fingerprints of the reference strainsin the commercial databank (TSBA version no. 3.9, Microbial ID, Newark, DE, USA)which were also cultured on TSA. It should be noted that not all the investigated strainsgrew on TSA probably due to the lack of sufficient salt in the culture medium.

2.3.2. BIOLOG fingerprintsThe strains that did not grow on TSA and the strains belonging to a Vibrio cluster

were further characterised by BIOLOG metabolic profiles. Strains were grown for 24 hat 25°C on Brain Heart Infusion (BHI, Difco Laboratories, Detroit, MI, USA) supple-mented with 1.5% (wIv) sodium chloride. Inocula were prepared in 1.5% (wIv)sodium chloride and the cell density was standardised between 0.26 and 0.30 O.D. usinga spectrophotometer at 590 nm. Each well in the BIOLOG GN microplate (BIOLOG,Hayward, CA, USA) was inoculated with 150 j.Llof the cell suspension and thesemicroplates were incubated at 25°C for 24 h. Changes in colour were measured using aMultiscan Multisoft filter photometer (Labsystems, Helsinki, Finland) at 590 nm. TheBIOLOG profiles were compared to a database containing BIOLOG fingerprints of 850Vibrio type- and reference strains, by numerical analysis using the Pearson productmoment correlation coefficient. The strains were grouped by unweighted pair-groupmethod of averages (UPGMA).

2.4. Selection of potential probiotic strains for turbot larviculture

The results of the larval rearing cultures were arbitrarily evaluated as successful(survival > 35%), average (survival from 10% to 35%) or a failure (survival < 10%).For each cluster from the PCA analysis, a selectivity index P was calculated followingthe equation P = S/(F + A + S) where S stands for the total number of isolatesderiving from a successful tank, A for the total number of isolates deriving from anaverage culture and F for the total number of isolates from a failure. When the index Pwas higher than 0.5 (i.e., a lot of isolates belonging to a successful culture), the bacteriafrom this cluster were catalogued as potential probionts.

IL. Huys et al. / Aquaculture 193 (2001) 25-37 29

2.5. Small-scale confrontation tests

For each selected potential probiotic gaschromatographic cluster, one representativeisolate was pointed out for further screening on the effect of larval survival. The selectedpotential probiotic strains were tested on first feeding turbot larvae in a I-weekconfrontation test. Therefore, 1 day after hatching, 30 larvae were stocked in a 1-1glassbeaker filled with 500 ml UV-sterilised seawater. Each tested bacterial strain as well asthe control group consisted of eight replicates. No aeration and no feeding weresupplied, the temperature was constantly 18°C and the beakers were continuouslyilluminated. Bacterial suspensions were prepared of each selected potential probioticstrain in a Nine Salt Solution (NSS) and added to the beakers at a concentration of 105bacteria ml- I water. The effect on larval survival was evaluated daily and compared tountreated control groups. In total, five small-scale confrontation tests were carried outfollowing this procedure. Additionally, a Vibrio mediterranei Q40 strain isolated fromsea bream larvae (Grisez et aI., 1997) was included during each trial.

3. Results

3.1. Survival at day 11

Large variation, from 0% up to 44%, was observed in the survival percentage oflarval turbot during both experiments (Table 1). According to the evaluation criteria,tanks 3 and 6 from Exp. I as well as tanks 8 and 12 from Exp. 11were pointed out assuccessful culture tanks since they had a survival percentage above 35%. Tanks 7, 9 and10, all used in Exp. 11,collapsed completely and were therefore evaluated as failures.The remaining tanks 1, 2, 4 and 5 (Exp. I), just as tank 11 (Exp. 11),were considered asaverage cultures.

3.2. Bacterial loading of the gut of turbot larvae

The development of gut-associated bacteria in larval turbot was assessed as changesin the number of colony forming units (CFU) on MA. During both experiments, nosignificant differences between the tanks were obtained in the development of thebacterial flora in the gut of intensively reared turbot larvae. Just before first feeding atday 3 post hatch, the turbot larvae had a few bacteria (circa 102 CFU/larva) associatedwith the gut. This level increased rapidly after feeding commenced until more than 104CFU per larva at day 5 post hatch to approximately 105 C;FUper larva at day 9 posthatch (results not shown).

During both experiments, no Vibrio-counts were observed before addition of the firstfood at day 3 post hatch. Although this number increased rapidly after feeding hadstarted, it rarely exceeded 10% of the value perceived on MA. As the Vibrio-count onTCBS is often found to be of limited value (Bolinches and Egidius, 1987), TCBS-agar isused as a selective medium for Vibrios rather than for quantification (Nicholls et aI.,1976).

30 L. Huys et al. / Aquaculture 193 (200]) 25-37

Table I

Survival percentages at day 11 of experiment I and 11with notification of the isolate numbers per tank

Experiment Tank Survival % Evaluation: Number of Isolate no.no. at day 11 F, A or sa isolates

per tank

3.3. Characterisation of the isolates by FAME and BIOLOG fingerprints

During these two turbot experiments, 149 isolates were obtained but when culturingthem on MA, 15% did not grow and therefore only 127 isolates were considered for thefatty acid analysis (FAME). Table 1 includes a schematic overview of the number ofisolates originating in each tank.

The isolates were compared and grouped on the basis of the fatty acid fingerprintsusing principal component analysis (PCA). This operation resulted in the definition of12 major gaschromatographic FAME-groups or clusters while 11 isolates remainedunclustered (namely isolate no. la, 55, 58, 66, 67, 87, 92, 95, 109, 113 and 118). Thedelineation of these clusters and single strains was also found in a numerical analysis ofthe fatty acid of the isolates (dendrogram not shown), using the Euclidean distancecoefficient and clustering by the unweighted pairgroup average method (Sneath andSokal, 1973). For further identification of the clusters, 3 or more representatives of eachcluster were selected for the FAME analysis with TSA as a culture medium and theywere also subjected to the BIOLOG technique. Table 2 shows that five clusters (A, D, F,G and K) did not grow on TSA or BRI and remained unidentified. Cluster B was readily

30 A 10 la, Ib, 2, 15,31,32,41,42,51,52

2 27 A 7 3a, 3b, 4, 5, 19, 54, 553 44 S 13 6,7, 8, 21, 22, 33, 34,

35, 43, 44a, 44b,45, 534 30 A 8 9, lOa, lOb, 23, 24,

25,57,585 17 A 5 11,12,39,59,606 36 S 12 Ba, Bb, 14,28,30,36,

37, 47, 48, 49, 61a, 61b11 7 0 F 11 63,65,66,67,102,103,

104, 105, 106, 107, 1088 42 S 13 69, 70, 71, 72, 73, 74, 109,

110, 111, 112, 113, 114, 1159 0 F 14 75, 76, 77, 79, 80, 81, 116,

117, 118, 119, 120, 122,

123, 12410 0 F 12 82,83,84,85,86,87,88,

125, 126, 128, 129, 13011 12 A 11 89,90,91,92,93,94,96,

131, 132, 133, 13512 38 S 11 95,97,98,99, 100, 101,

137, 139, 140, 141, 149

as = successful culture, A = average culture, F = failure.

L. Buys et al. / Aquaculture 193 (200]) 25-37 31

Table 2

Survey of the 12 FAME-clusters obtained by FAME analysis with MA as growth medium. Representativestrains of each FAME-cluster were identified by comparison with databases of reference strains using i)FAME analysis with TSA as growth medium and iD BIOLOG metabolic fingerprinting

FAME cluster Isolate no. FAME identification BIOLOG Identification

(# iso1ates/cluster)

A (8) no growth on BHI

BOO

C (5)D (2)E (22)

F (3)G (2)H (9)

1(3)

](4)

K (2)L (35)

Ba, 14, 53, 86, 93,

98, 112, 1391b, lOb, 11, 13b, 22,

24, 28, 61a, 61b,

69, 892,9, lOa, 12,52

85,994,6,7,32,33,34,37,

39,54,57,60,63,71,75,82,91,96,120,

125, 129, 131, 137

70, 77, 1003a,3b51,59,65,80,90,97,102, 116, 12821, 42,44a, 44b,45,

48,49,104,111,126,132, 140, 149

19,35,36,47

79,83

5, 8, 15, 23, 25, 30, 31,

41,43,72,73,74,76,81,84,88,94, 101,

103, 105, 106, 107, 108,

110,114,115,117,119,122, 123, 124, 130, 133,

135, 141

no growth on TSA

Pseudomonas sp.

no growth on TSA

no growth on TSAPseudoalteromonas

haloplanktis(4,7,32,34,39,54)

no growth on TSA

no growth on TSAno growth on TSA

Vibrio sp.

Vibrio sp.

no growth on TSAno growth on TSA

no match found

no match found

no growth on BHIno match found

no growth on BHIno growth on BHIno match found

Vibrio mediterranei

(21,48,49, 132)V. nereis or V. costicola

(45, 104, 111, 126)V. nereis or V. costicola

(35, 36, 47)

no growth on BHI

Vibrio campbellii(8, 25, 30, 43, 76,

94, 105, 110, 133)

recognisable as Pseudomonas with the FAME technique but identification down tostrain level was not possible. Likewise, no match was found with the BIOLOGtechnique as the databank only contains reference strains of Vibrio species. Clusters Cand H, which did not grow on TSA probably due to the lack of sufficient salt, showedwith BIOLOG less than 40% similarity with the Vibrio genera. Cluster E was recognisedas Pseudoalteromonas haloplanktis with FAME analysis while there was no matchfound with BIOLOG. According to BIOLOG, four isolates of cluster I were identified asbeing V. mediterranei while four other isolates were recognised as Vibrio nereis orVibrio costicola. According to the FAME analysis, all the isolates of cluster J wereassigned to the genus Vibrio but without species specification. The results of BIOLOGindicated that cluster J was V. nereis or V. costicola. The FAME-method was unable to

32 L. Huys et al. / Aquaculture 193 (200]) 25-37

identify cluster L as no growth was detected on TSA, while BIOLOG identified clusterL as Vibrio campbellii with a similarity level of 75%.

3.4. Distribution of the FAME clusters

There was a great variation between the twelve different tanks with regard to thepresence of the FAME-clusters among the bacterial isolates from the intestine of turbotlarvae of each tank (Table 3). Isolates of clusters C, G and J were detected in at mostthree tanks of the first experiment while isolates of clusters D, F and K appeared inmaximally three tanks of the second experiment. On the other hand, isolates from clusterE and L were rather ubiquitously present in all tanks, i.e., that, with regard to theexperimental design used in this work, P. haloplanktis and V. campbellii were bothpredominant bacteria in the intestinal tract of larval turbot. The remainder clusters A, B,H and I occurred in at least six tanks and appeared in both experiments.

Table 3

Distribution of the number of isolates per FAME cluster according to (a) the tank number, (b) the samplingday (3, 5 and 9) and (c) the success rate of the larval culture (S, A or F)

According to Number of isolates per FAME cluster (total number of isolates per cluster)

N (8) Ba (J 1) C (5) D (2) E (22) F (3) G (2) H (9) I" (13) J" (4) K (2) L (35)

(a) Tank no.1 1 2 1 1 1 32 2 2 I I3 1 1 4 4 1 24 2 2 1 25 I I 26 2 4 1 2 2 I7 I 2 1 58 I 1 1 1 1 69 2 1 2 1 710 1 1 3 1 1 1 311 1 1 3 1 1 312 2 1 1 1 1 2 2

(b) Sampling day3 4 4 1 2 9 3 0 6 0 0 2 95 4 7 4 0 8 0 2 3 6 1 0 239 0 0 0 0 5 0 0 0 7 3 0 3

(cl Success rate of larval culturesF = failure 1 0 0 1 6 1 0 5 2 0 2 15A = average 1 5 5 0 9 0 2 3 2 1 0 9S = success 6 6 0 1 7 2 0 1 9 3 0 11Index pb 0.75 0.55 0.00 0.50 0.32 0.67 0.00 0.11 0.69 0.75 0.00 0.31

a Selected potential probionts.bp=S/(F+ A+S).

L. Huys et al. / Aquaculture 193 (200]) 25-37 33

Table 4

Small-scale confrontation tests: survival percentages at day 5 post hatch (48 h after administration of thebacterial suspension)

Bacterial strain Survival %:t SO (day 5 post hatch)

Trial 1 Trial 2 Trial 3 Trial 4 Trial 5

NO = not determined.

.Vibrio mediterranei Q40: strain isolated from sea bream larvae (Grisez et aI., 1997).

Table 3 shows that the clusters D, F and K only appeared in the first sampling at day3. Later in the experiment, no more isolates of these clusters were detected, suggestingthat the respective clusters could not be established in the intestinal tract of the larvae.The same conclusion can be made for cluster G that only appeared at sampling day 5.Clusters A, B, C and H were only isolated at sampling day 3 and 5 while clusters I and Jwere detected from day 5 onwards.

3.5. Small-scale confrontation tests

For the identification of a potential probiont, clusters with a high P-index wereselected, meaning that the majority of the isolates of this cluster originated from asuccessful turbot culture, which was the case for clusters A, B, F, I and J (Table 3).Apart from cluster F that only contained three isolates and one of them belonging to afailure, representative isolates of these clusters were tested in vivo for their effect on thesurvival of turbot larvae in a small-scale confrontation test (Table 4). The influence onlarval survival was most visible at day 5 post hatch, Le., 48 h after addition of thebacterial strain, while around day 7 or 8 post hatch, the majority of the larvae were deadthrough lack of feed (personal observation).

The strains representing the unidentified cluster A and V. mediterranei Q40 both hada distinctive positive and reproducible effect on the survival of turbot larvae comparedto the untreated control groups. The isolates representing cluster I and J were onlyevaluated in the third trial and showed both a positive effect on larval survival comparedto the control although they seemed to be less effective than cluster A and V.mediterannei Q40. The reproducibility of these results needs to be verified. Noreproducible results were obtained with the strain representative of the Pseudomonascluster B.

4. Discussion

During both turbot experiments, large variation in viability was observed between thelarval culture tanks although these were submitted to identical rearing conditions. As aresult, one must assume that that the difference in survival is caused by factors other

Control 14.29:t 8.54 60.83:t 7.39 50.00:t3.00 75.00:t 15.84 15.84:t 10.55

V. med. Q40' 55.24:t 12.00 76.67:t 12.17 73.40:t 11.30 81.67:t 12.22 70.63:t 10.47

Cluster A NO 85.83:t 13.44 60.60:t 11.24 NO 79.30:t 10.81Cluster B NO 44.17:t 14.24 57.40:t 13.35 NO NO

Cluster I NO NO 66.60:t 10.36 NO NO

Cluster J NO NO 66.80:t 15.38 NO NO

34 L. Huys et al. / Aquaculture 193 (2001) 25-37

than egg quality, nutrition and culture techniques. Many studies have revealed micro-organisms to be involved in the problems during the early larval stages of marine fishlarvae (Verdonck and Swings, 1995; Skjermo and Vadstein, 1999; Hansen and Olafsen,1999).

The results of this study, in line with Munro et al. (1994), clearly demonstrated thatthere was no correlation between the number of bacteria present in the larval gut andlarval survival rates as all replicates followed nearly the same bacterial development inthe larval gut throughout the whole experiment. Just before first feeding (day 3 posthatch), turbot larvae contained few bacteria (circa 102 CFU/larva) associated with thegut. After feeding had started, this level increased to more than 104 CFU/larva at day 4post hatch and grew further up to 105 CFU/larva at day 9 post hatch. These results arein accordance with the findings of Munro et al. (1993), Ring!1let al. (1996) and Ring!1land Vadstein (1998).

The observation that bacteria can enter the tract before first feeding has initiated, canbe explained by the osmoregulation during which the marine fish larvae start drinkingbefore the yolk is completely absorbed (Reitan et aI., 1998). Moreover, Hansen andOlafsen (1999) suggest that ingestion of bacteria at the yolk sac stage might result in theestablishment of a primary intestinal microflora, which seemed to persist beyond firstfeeding.

Munro et al. (1995) postulated that control on the bacterial diversity present, ratherthan the total bacterial density in the gut might be important in ensuring high larvalsurvival rates. If the gut bacteria influence the survival rate, there should be a differencein the composition of the gut flora between turbot larvae derived from a failure culture,an average culture or a successful culture during the two turbot experiments presented inthis study. Although there was a great variation in the presence of the 12 FAME-clustersover the 12 rearing tanks, isolates from cluster E and L were ubiquitously present in alltanks at each sampling day. This implies that, with respect to the experimental designused in this research, P. haloplanktis and V. campbellii are predominant species in thegut of larval turbot. The dissimilarity in bacterial composition between the differenttanks, observed in both Exp. I and Exp. 11,highlight the inter-tank variability that wasalso reported by Grisez et al. (1997). Yet, no explanation could be found for thistank-effect although it is remarkable that the presence of both the non-Vibrio cluster Aand the Vibrio cluster I was peculiar to all successful rearing cultures (tanks 3, 6, 8 and12), while only one isolate belonging to cluster A was detected in solely one failureculture (tank 10). In conclusion, these observations indicate that the composition of thegut microflora of larval turbot may differ from one tank to another, even when the samerearing conditions are applied.

Importantly, a series of studies investigating the taxonomic composition of the larvalgut of various marine fish species (review Ring!1land Birkbeck, 1999) demonstrated thatthe intestinal microflora is very diverse as the composition may change with age,nutritional status and environmental conditions (Hansen and Olafsen, 1999). Besides, itshould be noted that differences in methodology make interlaboratory comparisonsdifficult, explaining conflicting statements in literature.

In addition, there seemed to be a transition in the composition of the gut microfloraduring the first week of feeding the larvae with rotifers. Apart from cluster E and L,

L. Huys et al. / Aquaculture 193 (2001) 25-37 35

whom were attendant at all sampling days, most clusters were only temporarily present.It is interesting to note that the unidentified non-Vibrio cluster A was isolated beforefirst feeding had commenced and disappeared later on in the experiment. The outcomethat some bacterial strains may become established while others may be digested orexpelled, can be attributed to the complex interactions between the host and theintestinal microbiota (Hansen and Olafsen, 1999), which reinforces the notion thatfurther research on this matter is warranted.

As fish larvae establish their gut microflora partly in a non-selective way (Hansenand Olafsen, 1989; Cahill, 1990) and as there is proof for the existence of an indigenousmicroflora in marine and fresh water fish (Horsley, 1977; Sakata, 1990; Ring!1SandGatesoupe, 1998), the initial transient bacteria are of utmost importance as they maybecome established and evolve into a more persistent microflora. Moreover, the abilityof specific bacteria to occupy attachment sites in the larval gut preventing opportunisticbacteria proliferating and colonising the larval gut, is assumed to be an importantdefence mechanism, especially during very early larval stages when the immune systemis not fully developed (Vanbelle et aI., 1990).

Based on this hypothesis, representative isolates of cluster A, B, I and J were selectedas potential beneficial strains and they were screened for their effect on larval survival ina small-scale confrontation test, in which also a V. mediterranei strain Q40 wasincluded. Only the non-Vibrio cluster A and V. mediterranei Q40, administered at themoment of mouth opening, showed a distinct positive and reproducible effect on larvalsurvival. It was concluded that both these strains could play a role as first colonisers ofthe gut of turbot larvae and as a consequence protect the gut from colonisation bypossible harmful bacteria.

These observations indicate that the concept of introducing bacterial species to therearing water at very early larval stages, may favour the growth of a protective normalflora and display a continued effect on the further bacterial development in the larvalintestinal tract. This finding is supported by the earlier observations of Str!1Smand Ring!1S(1993), Ring!1Set al. (1996) and Ring!1Sand Vadstein (1998).

Acknowledgements

LH acknowledges a grant of the Flemish Institute for the Promotion of ScientificTechnological Research in the Industry (IWT). This study was also supported by theFund for Scientific Research - Flanders (Belgium, FWO grant no. G0063.96).

References

Bergh, 0., 1995. Bacteria associated with early life stages of halibut, Hippoglossus hippoglossus L., inhibit

growth of a pathogenic Vibrio species. Journal of Fish Diseases 18, 31-40.Bolinches, J., Egidius, E., 1987. Heterotrophic bacterial communities with the rearing of larval halibut,

Hippoglossus hippoglossus, with special reference to Vibrio species. Journal of Applied Ichthyology 3,165-173.

36 L. Huys et al. / Aquaculture 193 (200]) 25-37

Cahill, M.M., 1990. Bacteria] flora of fishes: a review. Journal of Microbial Ecology 19, 21-41.

De Boer, S.H., Sasser, M., 1986. Differentiation of Erwinia carotovora ssp. carotovora and E. carotovora

ssp. atroseptica on the basis of cellular fatty acid composition. Canadian Journal of Microbiology 32,796-800.

Dhert, P., 1996. Rotifers. Manual on the Production and Use of Live Food for Aquaculture. In: Lavens, P.,Sorgeloos, P. (Eds.), FAO Fisheries Technical Paper 361, pp. 61-100.

Dhert, P., Lavens, P., Dehasque, M., Sorgeloos, P., 1993. Improvements in the larviculture of turbot

Scophthalmus maximus: zootechnicals and nutritional aspects, possibility for disease control. Turbot

Culture: Problems and Prospects. In: Lavens, P., Remmerswaal, R. (Eds.), Gent, European AquacultureSociety, Special Publication 22, pp. 32-46.

Dunn, G., Everitt, B.S., 1982. Multidimensional scaling. In: Cannings, C., Hoppensteadt, F. (Eds.), AnIntroduction to Mathematical Taxonomy. Cambridge Univ. Press, New York, USA

Grisez, L., Reyniers, J., Verdonck, L., Swings, l, Ollevier, F., 1997. Dominant intestinal microflora of sea

bream and sea bass larvae, from two hatcheries, during larval development. Aquaculture 155, 387-399.

Hansen, G.H., Olafsen, J.A., 1989. Bacterial colonization of cod (Gadus morhua L.) and halibut (Hippoglos-sus hippoglossus) eggs in marine aquaculture. Applied and Environmental Microbiology 55, 1435-1446.

Hansen, G.H., Olafsen, J.A, 1990. Endocytosis of bacteria in yolk sac larvae of cod (Gadus morhua L.). In:

Lesel, R. (Ed.), Microbiology of Poecilotherms. Elsevier, Amsterdam, pp. 187-191.

Hansen, G.H., Olafsen, J.A, 1999. Review article: bacterial interactions in early life stages of marine coldwater fish. Microbial Ecology 38, 1-26.

Horsley, R.W., 1977. A review of the bacterial flora of teleosts and elasomobranchs, including methods for itsanalysis. Journal of Fish Biology 10, 529-553.

Minkoff, G., Broadhurst, A.P., 1994. Intensive production of turbot, Scophthalmus maximus, fry. TurbotCulture: Problems and Prospects. In: Lavens, P., Remmerswaal, R. (Eds.), Gent, European AquacultureSociety, Special Publication 22, pp. 14-29.

Munro, P.D., Barbour, A., Birkbeck, T.H., 1994. Comparison of the gut bacterial flora of start feeding larvalturbot reared under different conditions. Journal of Applied Bacteriology 77, 560-566.

Munro, P.D., Barbour, A., Birkbeck, T.H., 1995. Comparison of the growth and survival of larval turbot in the

absence of culturable bacteria with those in the presence of Vibrio anguillarum, Vibrio alginolyticus, or amarine Aeromonas species. Applied and Environmental Microbiology 61 (12), 4425-4428.

Munro, P.D., Birkbeck, T.H., Barbour, A, 1993. Influence of rate of bacterial colonisation of the gut of turbotlarvae on larval survival. In: Reinertsen, H., Dahle, L.A., J~rgensen, L., Tvinnereim, K. (Eds.), Proceed-

ings from International Conference on Fish Farming Technology, Trondheim, Norway, August 1993, pp.85-92.

Muroga, K., Higashi, M., Keitoku, H., 1987. The isolation of intestinal microflora of farmed red seabrearn

(Pagrus major) and black seabream (Acanthopagrus schlegeli) at larval and juvenile stages. Aquaculture65, 79-87.

Nicholls, K.M., Lee, lV., Donovan, TJ., 1976. An evaluation of commercial thiosulphate citrate bile saltsucrose agar (TCBS). Journal of Applied Bacteriology 41, 265-269.

Olafsen, lA., 1993. The microbial ecology of fish aquaculture. In: Heen, K., Monahan, R.L., Utter, F. (Eds.),

Salmon Aquaculture. Fishing New Books, Oxford, pp. 166-175.

Olssen, J.C., Westerdahl, A., Conway, P.L., Kjelleberg, S., 1992. Intestinal colonization potential of turbot

(Scophthalmus maximus)-and dab (Limanda limanda)-associated bacteria with inhibitory effects againstVibrio anguillarum. Applied and Environmental Microbiology 58, 551-556.

Planas, M., 1994. R&D on production systems. Turbot Culture: Problems and Prospects. In: Lavens, P.,

Remmerswaal, R. (Eds.), Gent, European Aquaculture Society, Special Publication 22, pp. 57-73.

Reitan, K.!., Natvik, C.M., Vadstein, 0., 1998. Drinking rate, uptake of bacteria and microalgae in turbotlarvae. Journal of Fish Biology 53, 1145-1154.

Ring~, E., Birkbeck, T.H., 1999. Review: intestinal microflora in fish larvae. Aquaculture Research 30, 73-93.

Ring~, E., Gatesoupe, FJ., 1998. Lactic acid bacteria in fish: a review. Aquaculture 160, 177-203.

Ring~, E., Vadstein, 0., 1998. Colonization of Vibrio pelagius and Aeromononas ca viae in early developingturbot (Scophthalmus maximus L.) larvae. Journal of Applied Bacteriology 84, 227-233.

Ring~, E., Birkbeck, T.H., Munro, P.D., 1996. The effect of early exposure to Vibrio pelagius on the aerobical

flora of turbot, Scophthalmus maximus (L) larvae. Journal of Applied Bacteriology 81, 207-211.

L. Huys et al. / Aquaculture 193 (2001) 25-37 37

Sakata, T., 1990. Microflora in the digestive tract of fish and shell fish. In: Usel, R. (Ed.), Microbiology of

Poecilotherms. Elsevier, Amsterdam, pp. 171-176.

Skjermo, J., Vadstein, 0., 1999. Techniques for microbial control in the intensive rearing of marine fish

larvae. Aquaculture 177 (1-4), 333-343.Skjermo, J., Salvesen, I., 0ie, G., Olsen, Y., Vadstein, 0., 1997. Microbially matured water: a technique for

selection of a non-opportunistic bacterial flora in water that may improve performance of marine larvae.

Aquaculture International 5, 13-28.Sneath, P.H.A., Sokal, R.R., 1973. Numerical Taxonomy. The Principles and Practice of Numerical Classifica-

tion. W.H. Freeman, San Francisco, CA, USA.

Strjijm, E., Ringjij, E., 1993. Changes in the bacterial composition of early developing cod, Gadus morhua L.,

larvae following inoculation of Lactobacillus plantarum into the water. In: Walther, B., Fyhn, HJ. (Eds.),Physical and Biochemical Aspects of Fish Larval Development. Grafisk Husk, Bergen, pp. 226-228.

Vadstein, 0., 0ie, G., Olsen, Y., Salvesen, I., Skjermo, J., Skjak-Brrek, G., 1993. A strategy to obtain

microbial control during larval development of marine fish. In: Reinertsen, D., Jjijrgensen, T. (Eds.),Proceedings from International Conference on Fish Farming Technology, Trondheim, Norway, August

1993, pp. 69-75.Vanbelle, M., Teller, E., Focant, M., 1990. Probiotics in animal nutrition: a review. Archives of Animal

Nutrition Berlin 40, 543-567.

Verdonck, L., Swings, J., 1995. Microbial problems related to marine fish hatcheries in southern Europe: a

case study. Memorias 11 Congreso Ecuatoriano de Aquacultura 1993, 197-199.Westerdahl, A., Olsson, J.e., Conway, P.L., Kjelleberg, S., 1994. Characterization of turbot (Scophthalmus

maxim us ) associated bacteria with inhibitory effects against the fish pathogen Vibrio anguillarum. Acta

Microbiologica et Immunologica Hungarica 41, 403-409.