Microb Ecol (1995) 30:321-334 MICROBIAL ECOLOGY

© 1995 Springer-Verlag New York Inc.

Yeast Co lon iz ing the Intest ine of R a i n b o w Trout (Salmo gairdneri) and Turbot (Scophtalmus maximus)

T. Andlid, R.-V. Jufirez, L. Gustafsson

Department of General and Marine Microbiology, University of G6teborg, Medicinaregatan 9 C, S-413 90 GOteborg, Sweden

Received: 5 October 1994; Revised: 13 January 1995

Abstract. Yeast were isolated from the intestine of farmed rainbow trout (Salmo gairdneri), turbot (Scophtalmus maximus), and free-living fiat-fish (Pleuronectes platessa and P. fiesus). The average number of viable yeasts recovered from farmed rainbow trout was 3.0 × 103 and 0.5 × 10 2 cells per gram homogenized intestine for white and red-pigmented yeasts, respectively. The dominant species were Debaryomyces hansenii, Saccharomyces cerevisiae, Rhodotorula rubra, and R. glutinis. In 5 of 10 free-lying marine fish, >100 viable yeast cells per gram intestinal mucus were recovered. Red-pigmented yeasts dominated and composed >90% of the isolates. Colonization experi- ments were performed by inoculating rainbow trout and turbot with fish- specific, isolated yeast strains and by examining the microbial intestinal coloni- zation at intervals. Inoculation of experimental fish with pure cultures of R. glutinis and D. hansenii HF1 yielded colonization at a level several orders of magnitude higher than before the inoculation. Up to 3.8 × 104, 3.1 × 106, and 2.3 × 10 9 viable yeast cells per gram intestine or feces were recovered in three separate colonization experiments. The high level of colonizing yeasts persisted for several weeks. The concentrations of yeasts in the tank water never exceeded 103 viable cells per milliliter. No traces of fish sickness as a result of high yeast colonization were recorded during any of the colonization experiments. For periods of the experiments, the concentration of aerobic bacteria in the fish intestine was lower than the intestinal yeast concentration. Scanning electron microscopy studies demonstrated a close association of the yeasts with the intestinal mucosa. The mucosal colonization was further demonstrated by separating intestinal content, mucus, and tissue. All compartments were colo- nized by >103 viable yeast cells per gram. No bacteria were detected on the micrographs, indicating that their affinity for the intestinal mucosa was less than that of the yeasts.

Correspondence to: Thomas Andlid

322 T. Andlid et al.

Introduction

Yeasts are a widespread component of the aquatic microbiota of oceans and freshwa- ter systems [12]. Population densities vary with the availability of nutrients and generally decrease with increasing depth and distance from shore [1, 9, 12]. In open oceanic waters, the yeast concentration sometimes falls below 10 cells per liter, while in clean coastal waters, the typical concentration range is between 100 and 500 cells per liter [12]. Plankton blooms and polluted waters are associated with a pronounced increase in yeast concentration, mainly by those species associ- ated with terrestrial substrates, animals, and humans, such as different Candida spp. [9, 26]. In clean, oligotrophic waters on the other hand, the dominating species belong to the genera Rhodotorula, Cryptococcus, and Debaryomyces [2, 9, 23], which unlike the Candida spp., do not respond strongly to pollution [9]. Altogether, the data indicate that some yeasts are well adapted to reproduce in various niches of the marine and freshwater environments and presumably do not depend on terrestrial runoff. These species are aerobic, can use a wide range of compounds as carbon and energy sources, grow well at temperatures below 15°C, and are resistant to osmotic stress [8, 12]. To cope with the commonly low and transient concentration of nutrients and energy [21], the yeasts must survive long periods of starvation and be able to reproduce after such a period. Association with animals would provide a more nutrient-rich niche than the oligotrophic bulk-water.

The microbiology of fish has been extensively studied, both regarding spoilage [30] and natural microbial association [29, 33]. The gastrointestinal tract of fish provides a nutrient-rich habitat for microbial growth, and most data indicate that fish, like mammals, have an indigenous intestinal microbiota [5, 28]. At least fish with a developed gastrointestinal tract have groups of autochthonous microorgan- isms of different composition than the surrounding water [29]. Other authors claim that the intestinal biota mostly reflect the feeding and drinking habits of the animal and are therefore a function of the surroundings [5, 14]. In general, the intestinal bacterial flora of marine fish has been described as different groups of Vibrio and Pseudomonas, while the freshwater fish are colonized by Aeromonas and obligate anaerobes such as Bacteriosides and Eubacterium [27]. Yeasts have occasionally been found in fish intestines at higher concentration than the surrounding water, although little attention has been paid to their significance. The intestinal flora of bluefish, Pomatomus saltarix, was examined, and 2.5% of the colonies were found to be red pigmented yeasts [22]. Tilapia nilotica adapted to seawater contained 105.2 yeast cells per gram of stomach and fore-intestine content [32], and in the feces of cod, Gadus morhua, yeast-like organisms composed 9% of the total flora [10]. Furthermore, Ohwada et al. [24] found up to 94% of randomly selected colonies from the eel Synaphibranchus kaupi to be yeasts, concluding a host specificity of these to the gut of the eel in comparison with other fish studied. Usually, several times more procaryotes than yeasts are present in natural intestinal systems, including fish. However, in spite of the slower growth rate of yeasts, their bigger cell volume and biomass may have ecological significance in the intestinal system. Bearing in mind that the isolation methods (media composition, pH, time, antibiotics, etc.) seldom have been optimized for yeasts, there are reasons to believe that yeasts may inhabit the animal in greater numbers than previously thought and thereby have an impact on the host animal. The colonizing organism needs some

Yeast Colonizing the Intestine of Fish 323

s t ra tegy to p reva i l in the in tes t ine , e i ther by a t t ach ing to the m u c o s a or by g r o w i n g

fas ter than the d i lu t ion rate. Af f in i ty for the m u c o s a has b e e n s h o w n to be an

impor t an t co lon i za t i on fac to r for m a n y c o m m e n s a l and p a t h o g e n i c bac te r i a [4, 6,

11 ] and for Candida a lbicans c o l o n i z i n g the an ima l and h u m a n in tes t ine [13, 15]. T h e p resen t i nves t i ga t i on was unde r t aken in an a t t empt to r evea l w h e t h e r cer ta in

yeas ts strains i so la ted f r o m the in tes t ine o f f ish h a v e the b i o l o g i c a l p roper t ies

neces sa ry for u n e q u i v o c a l co lon i za t i on o f the intest ine. T h e co lon i za t i on capac i ty was e x a m i n e d by m e a n s o f in v i v o e x p e r i m e n t s us ing turbot and r a i n b o w trout.

S c a n n i n g e l ec t ron m i c r o s c o p y was used to obse rve the m o r p h o l o g y , adhes ion , and spat ial d i s t r ibu t ion o f c o l o n i z i n g m i c r o o r g a n i s m s .

Materials and Methods

Fish

Four colonization experiments (I to IV) were performed on Salmo gairdneri and Scophtalamus maximus obtained from three fish farms in Sweden:

I. Juvenile S. gairdneri with an approximate average weight of 12 g were obtained from the commercial fish farm Amen laxodling AB (Sweden). The fish were held in 200 liters of recirculating fresh water passing through a biofilter (Fluval 303) with a pumping capacity of 840 liters& at a temperature of 8°C.

IX. A second collection of S. gairdneri of the approximate weight of 15 g (Anten laxodling) was maintained at 15°C in three tanks each of 100 liters of fresh water, using the same filter and pumping system as described in I (above).

III. Salmo gairdneri of the average weight of 10 g were obtained from Laxforsen fish farm (Ewos AB, Kungsbacka, Sweden). The fish were held at 15°C in three tanks each of 100 liters of recirculating fresh water passing through the same biofilter as described in I (above).

IV. Seophtalamus maximus of an average weight of 10 g obtained from Havsfiskelaboratoriet, Swedish Fishery Board (Lysekil, Sweden) were maintained in a 100-liter tank of recirculating natural deep seawater. The same pumping and filtering system as described in I (above) was used.

Isolation o f Yeast Strains and Culturing Conditions

Yeasts were isolated from S. gairdneri belonging to fish in group I. The intestine of newly killed S. gairdneri were aseptically removed and homogenized in 0.9% NaC1. The homogenate was serially diluted and plated for viable yeast on YPG-plates (yeast extract 10 g, peptone 20 g, glucose 20 g, and agar 20 g in 1 liter of water), containing 250 mg chloramphenicol per liter. Eight different yeast strains were distinguished by colony morphology and light microscopy. The strains were sent to CBS Yeast Collection (Delft, the Netherlands) for identification. Yeast strains used in the colonization experiments were cultivated in 100 ml yeast nitrogen base (YNB) Difco, USA, containing 1 g glucose and 0.5 g (NH4)2SO4, at 30°C for 30 h on a rotary shaker, harvested, and washed in N-2- hydroxyethylpiperazine-N-2-ethanesulfonic acid (HEPES)-Hanks buffer, pH 7.2.

Colonization Experiments and Microbial Examination

Colonization Experiment L Before introduction of yeasts, 25 rainbow trout were kept for at least 2 weeks in the 200-liter tank for adaptation. During the experiment, the fish were fed every day with commercial pellets. At time zero, 20 ml of a washed stationary phase culture of Rhodotorula glutinis

324 T. Andlid et al.

CCUG 22164 (obtained from Department of Clinical Bacteriology S-413 46 Gtiteborg, Sweden) was added directly to the water of the fish tank. The intestinal flora was then examined at intervals by the following procedure: A fish was sacrificed, the ventral surface was scrubbed with 70% ethanol, and the intestine was aseptically removed. The intestine was cut free from fat and separated into two sections: (a) esophagus and stomach with pylorus and (b) the mid and posterior part of the intestine, including the rectum. The parts were transferred to sterile dishes and weighed. A 10-fold dilution was made in sterile 0.9% NaCI; the (intestinal) sections were homogenized for 3 rain and serially diluted in sterile 0.9% NaC1. Samples were spread on tryptic soy agar (TSA, Difco), which was found to give the highest number of bacteria, and on YPG-plates (yeast extract 10 g, peptone 20 g, glucose 20 g, and agar 20 g in 1 liter of water), containing 250 mg chloramphenicot per liter, for viable yeast enumeration. The plates were incubated at 15°C for 5-7 days. Water samples from the fish tanks were continuously withdrawn and plated for viable microorganisms.

Colonization Experiment II. Three tanks were used: A, seven fish inoculated with Debaryomyces hansenii HFI ; B and C, each containing seven control fish. To enable examination of the same animals throughout the whole experiment and to quantify an average of colonizing yeasts from several fish, the bacteria and yeasts in fish feces were examined. The number of yeasts and bacteria in the intestine before the introduction of yeasts was examined by sacrificing two fish (as described earlier) from each tank, where seven fish had been kept for 1 week. The remaining five fish in tank A were anesthetized with 3,3-diaminobenzidine tetrahydrochloride (0.1 g/liter, Sigma) and carefully fed orally with 1 mI of washed stationary phase culture of D. hansenii HE1, using a soft tube and a syringe. The fish were then replaced in the initial tank (A), which indicated time zero of the experiment. The two other tanks with equal numbers of fish were used as controls. The fish were fed in the morning with a commercial salmon pellet, and 3 hours later the feces from each fish tank were collected with a siphon tube. The excess water was poured back to the fish tank, and the feces were filtered (Munktell No. 8) under vacuum to remove most of the water and weighed. A 50 times dilution in 0.9% NaC1 containing 0.1% Tween 80 was made to facilitate the following homogenization. Samples from serial dilutions were finally spread on TSA (Difco) and YPG plates.

Colonization Experiment IlL Feed pellets containing viable yeasts were prepared as follows: 30-h precultures of D. hansenii HFI and Saccharomyces cemvisiae CBS 7764 were washed once in phosphate-buffered saline (PBS) buffer, pH 7.0, and the optical density at 610 nm was adjusted to 5. A 20-ml suspension of each strain was mixed with 40-g feed pellets by vortexing for 10 s. The suspension containing the pellets was filtered through filter paper (Whatman No. 8), spread on sterile dishes, dried for 20 h at 30°C and stored at 8°C. One week after the preparation, a pellet contained 1.9 × 106 viable yeast cells per gram. At intervals of approximately 14 days, new batches of yeast- containing pellets were prepared. Three tanks containing 15 rainbow trout each were used: A, fish fed every third day with the yeast pellets and the other days with normal feed; B and C, control fish, fed with normal feed once a day. Feces and water were examined for microorganisms as described.

Colonization Experiment lV. Five turbots were anesthetized with 3,3-diaminobenzidine tetrahydrochlo- ride (0.1 g/liter Sigma) and orally inoculated with D. hansenii HFI (according to the procedure described for colonization experiment II). Eleven days after inoculation, the yeasts present in the intestinal tract and water were enumerated by plating water samples and homogenized intestine from two fish according to the procedure described in colonization experiment I. The intestine was divided into the three regions: (1) stomach, (2) upper intestine, and (3) lower intestine.

For determination of the intestinal yeast flora in untreated farmed rainbow trout, the same procedure as described in colonization experiment I was used except for the addition of yeasts. Free-living Pleuronectes platessa and P. flesus were examined for yeast colonization by diluting the intestinal mucus scraped from intestines in synthetic seawater (NSS; nine salt solution: NaCI, 17.6 g; Na~SO4, 1.47 g; NaHCO3; 0.08 g; KC1, 0.25 g; KBr, 0.04 g; MgCI~ • 6H20, 1.87 g; CaC12 • 2H20, 0.41 g; SrC12 • 6 H20, 0.008 g; H3BO3 0.008 g in 1 liter of double distilled water) and plating on YPG media containing 250 mg chloramphenicol/per liter.

Yeast Colonizing the Intestine of Fish 325

Scanning Electron Microscopy

Pieces of tissue from different regions of the intestine of freshly sacrificed farmed rainbow trout and turbot were fixed in 2.5% glutaraldehyde in 0.07 M phosphate buffer, pH 7.2, and washed in phosphate buffer. Samples were dehydrated in a graded series of ethanol followed by acetone, critical point dried, sputter coated with gold-palladium, and finally examined in a scanning electron microscope (DSM 940; Zeiss, Germany).

Examination of Colonization Site

After colonization experiment II was completed, two fish from tank A were killed and the posterior part of the intestine was aseptically removed. Intestinal content and mucus were separately transferred to sterile tubes and weighed. A part of the remaining actual intestine (the tissue without content and mucus) was transferred to a petri dish and weighed. The three parts were diluted 100 times and homogenized in HEPES Hanks buffer, and microorganisms were determined by plating of viable cells on YPG (with 250 mg chloramphenicol per liter) and TSA.

Results

Yeasts in Farmed and Free-Living Fish

More than 95% of the white yeast colonies isolated from the intestine of farmed rainbow trout had the morphology of the strains identified as S. cerevisiae and D. hansenii. Such isolates were repeatedly sent to CBS Yeast Collection for identifica- tion. Cryptococcus and Leucosporidium were morphological ly easy to distinguish from Debaryomyces and Saccharomyces and present in such low numbers that their influence probably was of little importance. The group of red-pigmented yeasts was next in abundance after D. hansenii and S. cerevisiae. Among the distinguishable morphologies, Rhodotorula rubra was the most common followed by R. glutinis. In the colonization experiments (next section), the yeasts are referred to as red and white yeasts, which principally equal the Rhodotorula strains (red) and S. cerevisiae plus D. hansenii (white).

The mean number of yeasts in the intestine of nine untreated farmed rainbow trout f rom three different batches of fish (obtained from Anten laxodling AB) was 3 × 103 (ranging from 1 × 102 to 2 × 10 4) and 0.5 × 103 (0.5 × 102 to 2 × 10 3) cells per gram homogenized intestine for white and red yeasts, respectively. The number of yeast colony-forming units (CFUs) from the water of the fish tanks was always lower. Yeasts seemed to be a component of the indigenous flora in the farmed trout. Free-living marine fish, P. platessa and P. flesus caught at the Swedish west coast, were also examined. In 5 of 10 fish, yeasts were isolated at > 100 cells per gram intestinal mucus, and in 4 of these, red-pigmented yeasts presumed to be RhodotoruIa dominated and accounted for > 9 0 % of the recovered intestinal yeasts.

Colonization of Salmo gairdneri

Group L A higher concentration of both yeasts and bacteria was present f rom the beginning in the water of the fish tank compared to the fish intestine (Fig. 1, day

326

10 v

c9

105

103 !

,0 i 10 0 I 0 20 30 40 50 60

Time (days)

T. Andlid et al.

70

Fig. 1. Colonization experiment I. Concentration (CFUs) of yeasts and aerobic bacteria in the posterior intestine of rainbow trout. Fifteen fish were maintained at 8°C in recirculating fresh water. At time zero, R. glutinis CCUG 22164 was incoulated directly into the water, and the intestinal flora was examined by homogenizing the intestine of sacrificed fish. Filled symbols represent intestinal samples, and open symbols represent water samples. Total yeasts in the intestine (.), total yeasts in the tank water (o), red-pigmented yeasts in the intestine (t), total aerobic bacteria in the intestine (m), and total aerobic bacteria in the tank water (a).

-2). Following introduction of R. glutinis to the tank water at day zero, the yeast concentration was significantly higher in the fish intestine compared to the sur- rounding water (Fig. 1). The bacteria on the other hand remained at a higher concentration in the water than in the fish intestine (Fig. 1). At day 17, the number of colonized red-pigmented yeasts had increased to 45 times the initial level and thereafter gradually decreased in concentration. The major increase after feeding the fish with R. glutinis was in fact not due to this strain, but to three white strains, identified at CBS in Delft as D. hansenii (strain HFI) , S. cerevisiae CBS 7764, and S. cerevisiae CBS 7765. At day 17, the red yeasts accounted for 31% of the total yeast colonies recovered. This fraction decreased to 6% at day 65. The maximum number of yeasts was found in the fish sacrificed 38 days after time zero, where the total amount of yeast was 2.8 X 10 4 cells per gram of section A (upper) and 3.8 X 104 cells per gram of section B (posterior), 380 times more than before inoculation of yeast (Fig. 1). This occurred without an increased number of yeasts in the water. The increased level of yeast colonization seemed to have an effect on the aerobic bacteria, which were reduced in number at the sampling times when the yeast concentration was at maximum (Fig. 1). Up to 19 times more viable yeasts than aerobic bacteria were isolated from the lower region of the intestine and 15 times more in the upper region. The data in Fig. 1 originate from samples from the lower intestine. The upper intestine contained fewer microorgan- isms. The number of yeasts per gram of upper intestine was 77% of the number in the lower intestine (i.e., in the same fish) on day 17 and decreased gradually to 11% at day 65.

Group II. Debaryomyces hansenii HF1 was inoculated, and the number of yeasts in feces (except for Sample 1) was used as an indicator of the intestinal microbial population. The starting situation in the three tanks was (A) 900, (B) 3000, and (C) 245 yeast cells per gram homogenized intestine (Fig. 2). After time zero, which was the time of feeding all the fish in tank A with the HF1 strain, the average yeast concentration of feces in A increased to a high level. The peak (considering the discrete sampling) was reached after 13 days, when the total yeast concentration was 3.1 x 10 6 cells per gram feces (Fig. 2), corresponding to an increase of more than 3000 times, and should be compared with the two control tanks where the corresponding value was 1.2 X 104 cells per gram feces in each. Also, the yeast

Yeast Colonizing the Intestine of Fish 327

z v

107

I ()('

105

104

1 0 3

102 -10 0 I0 20 30 40 50

Time (days) 60

Fig. 2. Colonization experiment II. Concentration (CFUs) of yeasts and aerobic bacteria in the feces or rainbow trout and in the tank water. Three tanks, each with seven rainbow trout, were kept at 15°C with recirculating fresh water. At time zero, the experimental fish (tank A) were orally inoculated with D. hansenii HF1. The fish in tank B and C were used as controls. The microbial examination was made by homogenizing collected feces. Total yeasts in feces of tank A (e); white yeasts in feces of tank A (t); total yeasts in feces, average of tank B and C (~,); total yeasts in the water of tank A (o); and total yeast in the water, average of tank B and C (zx).

101o

10 8

z

r..) 10 6

10 4

-10

~ , i , i ,

10 20 30 40 50

Time (days)

60

Fig. 3. Colonization experiment III. Concentration (CFUs) of yeast and aerobic bacteria in feces and in the tank water. Debaryomyces hansenii HF1 and S. cerevisiae CBS 7764 were inoculated in combination into the experimental fish (tank A) continuously every third day, by feeding fish with pellets containing approximately l06 viable yeast cells per gram. Tank B and C were used as controls. Total yeasts in feces, tank A (e); total yeasts in faeces, average of tank B and C ( ,) ; total aerobic bacteria in feces, tank A (m); and total aerobic bacteria in feces, average of tanks B and C (#).

cell densi ty of the water in tank A, as well as in B plus C, is shown in Fig. 2. The yeas t cell densi ty in the water did not exceed 103 yeast cells per mil l i l i ter in any of the tanks. The redis t r ibut ion be tween red and white strains bas ica l ly means the dis tr ibut ion be tween R. rubra and R. glutinis on the one hand and D. hansenii and S. cerevisiae on the other. There was a prominent difference be tween the yeast content o f fish in A, where HF1 had been introduced, and the two controls, B and C. The dominat ing fract ion of the total yeasts f rom fish in tank A was white yeasts. The average concentra t ion ratio of whi te / red yeasts for the whole exper iment was 7.3 (_+0.9) in tank A. The corresponding values of the controls, B and C, were 0.22 (_+0.14) and 0.37 (+0 .26) , respect ively. In conclusion, the white strains domina ted in the inocula ted fish, and the red strains domina ted in the controls. Al though D. hansenii was the yeast species fed to the fish, there was a concomitant increased of Rhodotorula spp.

Group IlL A third method of int roduct ion of a specif ic yeas t strain in fish was practiced. Debaryomyces hansenii HF1 and S. cerevisiae CBS 7764 were dis t r ibuted to the exper imenta l fish by feeding them with a pel le t conta ining approx imate ly 106 viable yeas t cells per gram. Af ter 23 days, the feces contained 3.2 X 108 v iable yeast cells per gram and increased to 2.4 × 109 cells per g ram at 53 days (Fig. 3). The number of bacter ia increased to at h ighest 6.0 × 106 viable bacter ia per g ram

328 T. Andlid et al.

L

4.0 10 4 ~

3 .0 10 4 z

t,.) 2 .0 10 4

I.O 10 4

Con ten t M u c u s Tissue

Fig. 4. Distribution of viable yeast (CFUs) between intestinal content, mucus, and tissue of the lower intestine of rainbow trout. The fish belonged to the group of tank A in colonization experiment II and were sacrificed 80 days after the single inoculation with D. hansenii HFI (see text for details).

feces. Further, the control fish showed a successive, but much lower, increase in number of viable yeast concomitantly with a tendency toward decreasing number of bacteria. The number of yeasts in the tank water ranged from 3.5 × 102 to 1.8 X 104 cells per milliliter in tank A and in the controls from 1.5 X 101 to 1.8 x 103 cells per milliliter (data not shown). At least 95% of the recovered colonies exhibited a morphology in accordance with D. hansenii HF1.

Colonization Site

Viable yeast cells were recovered from the intestinal content, mucus, and tissue (including the epithelial surface). From the intestinal content to the mucosa, the number of yeasts progressively declined (Fig. 4). Exact comparison is difficult because of the tissue weight. What is important, however, is that the data indicate that the yeast strains have the capacity to penetrate and establish in the rainbow trout intestinal mucosa.

Colonization of Scopthalmus maximus With D. hansenii HF1

Eleven days after inoculation with D. hansenii HF1, one of the two examined fish contained a significant number of yeast in all regions of the intestine. The colony morphology and microscopic examination indicated that the re-isolated yeasts were D. hansenii. In contrast to rainbow trout, in turbot the stomach was the site of largest yeast colonization, with 8.4 _+ 3.3 x 103 viable yeast cells per gram. The corresponding values for the upper and lower intestine were 5.3 x 1.6 X 103 and 0.68 _+ 0.5 x 103, respectively. The second fish contained many fewer microorgan- isms (data not shown). The water of the fish tank was free of viable yeasts. Furthermore, in these fish, the number of isolated aerobic bacteria growing on TSA medium was zero.

Scanning Electron Microscopy

It was easy to detect colonizing microorganisms in the lower region of the intestine (Fig. 5), whereas in the anterior portion, only a few microorganisms were visible. Although bacteria were present in the homogenized intestine (including the content)

Yeast Colonizing the Intestine of Fish 329

Fig. 5. Scanning electron micrographs of the lower intestine of rainbow trout inoculated with D. hansenii HF1 (a-d) and turbot (e). The white bar represents 10 Ixm. The trout were taken from tank A in colonization experiment II (see text for details) 85 days after a single yeast inoculation, a, Typical view of the dominating spherical yeast cells. Note the size distribution and the adhesive cells, b, Different yeast morphology of cells associated with each other and mucus. The oval cells are presumed to be Rhodotorula or Saccharomyces spp., and the spherical cells Debaryomyces (sp. see text for explanation). c, Early bud on an oval mother cell, proving that the cells are yeasts, d, Close association of yeast cells and intestinal mucus, which after dehydration appears as fibrous strands of material. On the left, an area of naked epithelia is seen. These areas were only sparsely colonized, e, A lawn of mucus in lower intestine of turbot. The micrograph shows spherical yeast-like cells, tightly adhered to strands of mucus. Bacteria were not detected on any micrograph.

330 T. Andlid et al.

and in the feces, as determined by plate counting, they were, with few exceptions, absent in scanning electron microscopy (SEM) micrographs. Colonizing yeasts were, on the other hand easy, to detect. Figure 5a shows a typical view of the rainbow trout intestine, with spherical yeast cells, adhesive to each other and with a pronounced size distribution. The characteristics are all typical for the D. hansenii HF1 strain, with which these fish were inoculated at the beginning of colonization experiment II. Figure 5b shows different cell morphologies in close association with each other and to the mucus. The oval cells are probably RhodotoruIa or Saccharomyces. Figure 5c indicates that the cells are yeast by the presence of an early bud on an oval mother cell. Figure 5d shows a close association between the spherical yeasts and the mucus gel, which after complete dehydration appears as fibrous strands of material. The left side of Figure 5d illustrates an area of naked epithelial cells without coating mucus. Areas like this were found to be only sparsely colonized. The dominating cell type in SEM micrographs from rainbow trout was the spherical cell with a pronounced variation in size, reaching up to 35-txm diameter. This variation in size distribution was also seen with light micros- copy for the D. hansenii HF1 strain after growth in mucus or synthetic media. Figure 5e illustrates a section of the lower intestine of turbot. This fish was killed at the farm and not previously fed with live yeast; consequently, fewer detectable colonizing microorganisms were present. However, spherical and oval (not shown) yeast-like cells adhered to and penetrated the mucosa.

Discussion

The results of this study show that the gastrointestinal tract of rainbow trout and turbot may be colonized by dense populations of yeasts. No traces of fish sickness as a result of high yeast colonization were recorded during any of the colonization experiments. Rhodotorula glutinis, R. rubra, and D. hansenii are all species that constitute a large portion of the aquatic yeasts. These species were frequently isolated from the intestine of fish. If reintroduced in a higher number, some of these strains colonized the fish intestine at a level several orders of magnitude higher than usually found in free-living fish. Maximum values recovered were 3.8 X 10 4, 3.1 × 106, and 2.3 × 109 viable yeast cells per gram intestine or feces for experiments I, II, and III, respectively. At these levels, the yeasts most likely have an impact on the intestinal ecology and nutrition of the animal, which, however, remains to be clarified.

From our data and the literature, it seems feasible to suggest that the fish intestine may be an important reproduction site for these yeasts in natural aquatic environments. There is no doubt that fish continuously ingest live yeast cells by eating and drinking. A marine teleost drinks seawater at a rate of 0.5% of its body weight per hour [20]. We have previously shown that the fish-isolated yeasts change considerably in cell surface properties and therefore also the degree of adhesion to various substrata [34]. During exponential growth, these yeasts have a high cell surface hydrophobicity, while starving cells are much more hydrophilic. Further- more, we have shown that the yeasts can use the intestinal mucus as the sole source of nutrients [34]. By combining these findings, we suggest an ecological model in which the normally starving yeast cells in the bulk water enter the fish by its

Yeast Colonizing the Intestine of Fish 331

drinking, become exposed to an increasing nutrient concentration, and start to grow, accompanied by a progressive change in cell surface to a high degree of hydrophobicity, leading to adhesion and colonization. Cells that are lost through the rectum go back to the hydrophilic nongrowing stage, and the cycle is complete.

Austin et al. [2] used scanning electron microscopy to determine the distribution of gastrointestinal microorganisms in rainbow trout. Little evidence for intimate microbial colonization of the gastrointestinal tract was found. Organisms if coccoid morphology were instead found associated with food particles in the gastrointestinal tract. Most of these seem too large to be bacteria, and we suggest that they were yeasts, exhibiting similarity with the D. hansenii HF1 strain used in this study. It is even more striking that in no investigations of the gastrointestinal flora of fish, including our own, has any evidence for bacteria intimately colonizing the actual intestinal mucosa been presented [5]. If a microorganism has little or no capacity to penetrate the mucus, it may be lost during the dehydration procedure (which also must be a selective extraction of material since ethanol and acetone are used) for SEM preparation. The absence of bacteria in the micrographs may indicate less adherence or that bacteria principally grow in the intestinal content. The yeasts, on the contrary, were easy to detect on the micrographs and look very similar to C. albicans in association with the mucosa of mice in the SEM investigation by Kennedy and collaborators [16, 17]. For both rainbow trout and turbot not fed with live yeast, as well as fish used in the colonization experiments, yeasts in this investigation adhered firmly to the mucosa (Fig. 5a-e). Furthermore, the gastrointestinal tract of rainbow trout has a much thicker mucus layer in the posterior part than elsewhere, where we found the yeasts in the SEM study, and the highest number of yeast CFUs (Fig. 1). This study provides conclusive evidence that yeasts can colonize the actual intestinal mucosa of fish (Figs. 4, 5).

The examination of the yeast distribution between intestinal content, mucus, and epithelia indicated that the yeasts were able to penetrate into the mucus and maybe even to attach to the epithelium (cf. Fig. 5a). It is believed that the mucus gel has an important role as a barrier against invading pathogens [3]. The mucus has also been suggested to be a colonization site for commensal microorganisms [31] and may serve as a source of nutrients and adhesion receptors [4, 6, 11]. The yeasts isolated from the intestine of fish in this study, in in vitro experiments, have been shown to be adhesive to and to grow readily in isolated intestinal mucus (unpublished data).

The number of microorganisms detected in the first colonization experiment was significantly lower than in the other experiments. It is reasonable to presume that this is partly due to the lower temperature of 8°C compared with 15°C in the other experiments. Also, the method was homogenization of the intestine, including the whole tissue and the content, compared to homogenization of feces. The number of microorganisms associated with a poikilothermic animal is very dependent on the temperature of the surroundings [18]. This may be one of the reasons why there are such large variations in the groups and numbers of microorganisms found in fish from different investigations.

In colonization experiments I and II, the establishment of one strain of yeast in the intestine was followed by colonization of another. The reason for this apparent synergistic effect is unclear. There was, furthermore in colonization experiment I, a pronounced decrease in the number of bacteria after the level of colonized yeast

332 T. Andlid et al.

increased. Also in colonization experiment III, the number of bacteria in the feces from the control fish decreased as the number of yeasts increased. These results may indicate in vivo microbial antagonism. In vitro experiments have shown antagonistic effects of D. hansenii HF1 and S. cerevisiae 7764 against the fish pathogens Aeromonas salmonicidae and Vibrio angiullarum (unpublished data). Naturally, the concentration of live yeasts was high in the experiment III, in which the fish had a continuous input of approximately 106 viable yeast cells per gram. However, the feces concentration of > 10 9 viable yeast cells per gram feces must be explained in either one or a combination of three ways. The yeasts are adhesive and retained on surfaces and therefore concentrated in the intestine, the yeasts grow in the intestine, or they grow in the feces. Two more observations from experiment III were that approximately 50% of the yeast in the control fish (which increased during the experiment) were red-pigmented strains and that D. hansenii HF1 strain was a better colonizer than S. cerevisiae CBS 7764, when introduced in combination.

Many reports exist on yeasts improving the health of different animals and humans, when introduced as live cells [19]. Additionally, [3-1,3 glucans of the yeast cell wall are known to act as nonspecific immunomodulators in animals (fish, mice, crustaceans [7]) and thereby to mediate increased resistance against pathogens. In one investigation, rainbow trout injected with purified S. cereviciae cell wall glucans were significantly more resistant against Yersina ruckeri, V. anguillarum, and V. salmonicida than control fish [25]. Whether this effect or antagonism against pathogenic bacteria may be achieved by an increased level of colonized yeasts is at present under investigation. If the aim is to try yeast as an alternative health-improving agent in fish farms, it is important to chose an efficient method to feed the yeast to the fish. In this study, three different methods were performed. The oral inoculation causes a stress for the animals, which may lead to a disturbed intestinal microbiota, and this method would not be practical in larger scale studies. Water inoculation is imprecise and may be less efficient if the water is not recirculated. The preparation of dried but still viable yeasts in the feed pellets appears, on the other hand, controllable and practical, even for a larger scale study.

Acknowledgments. We thank Christer Olsson and Allan Westerdahl (Department of General and Marine Microbiology, G{3teborg University) for contributing to the preparation of the SEM micrograph of turbot intestine. We also express sincere thanks to Professor Paul Cohen for valuable discussions and to Dr. Ruth Kennan for linguistic help. The work was supported by grants from Gullmarsfon- den, Sweden.

References

I. Ahearn DG, Roth FJ, Meyers SP (1968) Ecology and characterization of yeasts from aquatic regions of South Florida. Marine Biol 1:291-308

2. Austin B, A1-Zahrani AMJ (1987) The effect of antimicrobial compounds on the gastrointestinal microflora of rainbow trout, Salmo gairdneri Richardson. J Fish Biol 1-34

3. Berg RD (1992) Translocation and the indigenous gut flora. In: Fuller R (ed) Probiotics. The scientific basis. Chapman & Hall, Cambridge, UK, pp 55-85

4. Blomberg L, Cohen PS, Conway PL (1992) A study of adhesive capacity of Escherichia coli strain Bd 1107/7508 (K88ac) in relation to growth phase. Microb Pathol 14:67-74

Yeast Colonizing the Intestine of Fish 333

5. Cahill MM (1990) Bacterial flora of fishes: a review. Microb Ecol 19:21-41 6. Cohen PS, Arruda JC, Williams TJ, Laux DC (1985) Adhesion of human fecal Escherichia coli

strain to mouse colonic mucus. Infect Immun 48:139-145 7. Di Luzio NRD (1985) Update on the immunomodulating activities of glucans. Springer Semin

Immunopathol 8:401412 8. Fell JW, Ahearn DG, Meyers SR Roth FJ, Jr (1960) Isolation of yeasts from Biscany Bay, Florida

and adjacent benthic areas. Limnol Oceanogr 5:366-371 9. Fell JW, van Uden N (1963) Yeasts in marine environment. In: Oppenheim CH (ed) Symposium

on marine microbiology. Charles C. Thomas, Springfield, Illinois, pp 329-340 10. Frances DE (1947) Microorganisms from Atlantic cod. J Fish Res Can 7:128-136 11. Freter R (1981) Mechanisms of association of bacteria with mucosal surfaces. Ciba Foundation

Symposium 80:36-55 12. Hagler AH, Ahearn DG (t987) Ecology of aquatic yeasts. In: Rose AH, Harrison JS (ed) The

yeasts. Harcourt Brace Jovanovich, Orlando, Florida pp 181-205 13. Hazen KC (1989) Participation of yeast cell surface hydrophobicity in adherence of Candida

aIbicans to human epithelial cells. Infect Immun 57:1894-1900 14. Horsely RW (1976) A review of bacterial flora of teleosts and elasmobranchs, including methods

for its analysis. J Fish Biol 10:529-553 15. Kennedy MJ (1990) Models for studying the role of fungal attachment in colonization and

pathogenesis. Mycopathologia 109:123-137 16. Kennedy M J, Volz PA (1985) Ecology of Candida albicans gut colonization: inhibition of Candida

adhesion, colonization and dissemination from the gastroinestinal tract by bacterial antagonism. Infect Immun 49:654-663

17. Kennedy MJ, Volz PA, Edwards CA, Yancey RJ (1987) Mechanisms of association of Candida albicans with intestinal mucosa. J Med Microbiol 24:333-341

18. L6sel R (1989) Thermal effect on bacterial flora in the gut of rainbow trout and African catfish. In: L6sel R (ed) Microbiology in poeciolotherms. Elsevier, Paris, pp 33-38

19. Lyons TR Jacques KA, Dawson KA (1993) Miscellaneous products from yeast. In: Rose AH, Harrison JS (eds) The yeasts. Academic Press, London, pp 293-324

20. McWilliams P (1992) Ioneregulering. In: D6ving K, Reimers E (eds) Fiskens fysiologi. John Grieg Forlag AS, Stavanger, Norway, pp 198-215

21. Morita RY (1988) Bioavailability of energy and its relationship to growth and starvation survival in nature. Can J Microbiol 34:436441

22. Newman JT, Cosenza B J, Buck DJ (1972) Aerobic microflora of the bluefish (Pomatomus saltatrix) intestine. J Fish Res Board Can 29:333-336

23. Norkrans B (1966) On the occurrence of yeasts in an estuary of the Swedish west coast. Svensk Botanisk Tidskrift 60:443482

24. Ohwada K, Tabor PS, Colwell RR (1980) Species composition and barotolerance of gut microflora of deep-sea benthic macrofauna collected at various depths in the Atlantic ocean. Appl Environ Microbiol 40:746-755

25. Robertsen B, R~Srstad G, Engstad R, RaaJ (1990) Enhancement of nonspecific disease resistance in Atlantic salmon, Salmo saIar L., by glucan from Saccharomyces cerevisiae cell walls. J Fish Dis 13:391-400

26. Roth FJ, Jr, Abeam DJ, Fell JW, Meyers SP, Meyer SA (1962) Ecology and taxonomy of yeasts isolated from various marine substrates. Limnol Oceanogr 7:178-185

27. Sakata T (1989) Microflora in the digestive tract of fish and shell-fish. In: L6sel R (ed) International symposium on microbiology in poecilotherms. ISBN, Paris, pp 171-176

28. Sakata T, Okabayashi J, Katimoto D (1979) Variations in the intestinal microflora of Tilapia reared in fresh and seawater. Bull Jpn Soc Scie Fish 46:313-317

29. Sera H, Ishida Y (1972) Bacterial flora in the digestive tracts of marine fish. III. Classification of isolated bacteria. Bull Jpn Soc Scie Fish 38:853-858

30. Simudu U, Kaneko E, Aiso K (1969) Microflora of fresh and stored flatfish, Kareius bocoloratus. Bull Jpn Soc Scie Fish 35:77-82

31. Stanley RA, Ram SE Wilkinson RK, Robertson AM (1986) Degradation of pig gastric and colonic mucins by bacteria isolated from the pig colon. Appl Environ Microbiol 51:1104-1109

334 T. Andlid et al.

32. Sugita H, Oshima K, Tamura M, Deguchi Y (1982) Bacterial flora in the gastrointestine of freshwater fishes in the river. Bull Jpn Soc Scie Fish 49:1387-1395

33. Trust TJ, Sparrow RAH (1974) The bacterial flora in the alimentary tract of salmonid fishes. Can J Microbiol 20:1219-1227

34. VS_zquez-Ju~rez R, Andlid T, Gustafsson L (1994) Cell surface hydrophobicity and its relation to adhesion of yeasts isolated from fish gut. Colloids Surfaces B: Biointerfaces 2:199-208