Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
Lone Gram, Danish Institute for Fisheries Research, Department of Seafood Research, Søltofts Plads, c/o Technical University of Denmark, bldg. 221, DK-2800 Kgs. Lyngby, Denmark (e-mail: firstname.lastname@example.org).
Aims: To identify the dominant culturable and nonculturable microbiota of rainbow trout intestine.
Methods and Results: Microbial density of rainbow trout intestine was estimated by direct microscopic counts (4′,6-diamidino-2-phenylindole, DAPI) and by culturing on tryptone soya agar (TSA). Differential gradient gel electrophoresis analysis of bacterial DNA from intestinal samples, re-amplification of bands and sequence analysis was used to identify the bacteria that dominated samples where aerobic counts were ≤2% of the DAPI counts. 16S rDNA gene sequences of 146 bacterial isolates and three sequences of uncultured bacteria were identified. A set of oligonucleotide probes was constructed and used to detect and enumerate the bacterial community structure of the gastrointestinal tract of rainbow trout by fluorescence in situ hybridization (FISH). Members of the gamma subclass of Proteobacteria (mainly Aeromonas and Enterobacteriaceae) dominated the bacterial population structure. Acinetobacter, Pseudomonas, Shewanella, Plesiomonas and Proteus were also identified together with isolates belonging to the beta subclass of Proteobacteria and Gram-positive bacteria with high and low DNA G + C content. In most samples, the aerobic count (on TSA) was 50–90% of the direct (DAPI) count. A bacterium representing a previously unknown phylogenetic lineage with only 89% 16S rRNA gene sequence similarity to Anaerofilum pentosovorans was detected in intestinal samples where aerobic counts were ≤2% of direct (DAPI) counts. Ten to 75% of the microbial population in samples with low aerobic counts hybridized (FISH) with a probe constructed against this not-yet cultured bacterium.
Conclusions: Proteobacteria belonging to the gamma subclass dominated the intestinal microbiota of rainbow trout. However, in some samples the microflora was dominated by uncultivated, presumed anaerobic, micro-organisms. The bacterial population structure of rainbow trout intestine, as well as total bacterial counts, varied from fish to fish.
Significance and Impact of the Study: Good correlation was seen between cultivation results and in situ analysis, however, a molecular approach was crucial for the identification of organisms uncultivated on TSA.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Within other ecological niches it has been shown that large subpopulations of micro-organisms exist that are not easily isolated and cultivated on traditional agar substrates (Van Elsas and van Overbeek 1993). The culturability of bacteria in water environments (seawater, freshwater and mesotrophic lake) is below 1% (Amann et al. 1995) and one study reported a culturability of micro-organisms from fish skin of <0·01% (Bernadsky and Rosenberg 1992).
Molecular methods based on the PCR, rDNA sequence analysis or fluorescence in situ hybridization (FISH) have reached a high level of acceptance in microbial ecology as techniques for identification and specific enumeration of bacteria unbiased by the limitations of culturability. Group-specific rRNA-targeted oligonucleotide probes offer means to rapidly assess those bacterial phylogenetic groups that are most abundant in a given sample. So far they have been successfully applied to various environmental samples such as activated sludge, marine picoplankton, biofilms, soil and the human intestine (Amann et al. 1995; Franks et al. 1998). Using 16S rRNA-targeted FISH, Amann et al. (1990) detected Fibrobacter intestinalis in the intestinal ecosystem of mice. Langendijk et al. (1995) applied quantitative FISH analysis to the human intestine for the detection of Bifidobacterium spp. with genus-specific oligonucleotide probes. With 16S rRNA probe hybridization using oligonucleotide probes specifically constructed for the environmental sample of interest, a high resolution in situ identification can be achieved. Denaturing gradient gel electrophoresis (DGGE) of PCR-amplified 16S rRNA genes of total bacterial DNA from an environmental sample offers a visualization of the bacterial diversity as a DNA fingerprint. It gives the opportunity for subsequent identification and enumeration of community members by sequence analysis and in situ hybridization with probes specifically developed for the bacteria identified from the environment (Muyzer et al. 1993).
The present study on the microbial diversity of the gastrointestinal tract of rainbow trout is part of a larger study in which we evaluated the potential disease preventive effect of bacterial strains isolated from fish (Gram et al. 1999; Spanggaard et al. 2000, 2001). The work has focused on the dominant part of the microflora assuming that these organisms may play an important role in vivo. It is therefore crucial to our strategy that the dominant micro-organisms are isolated and characterized. In a previous study (Spanggaard et al. 2000) we found that, in general, the dominant bacterial microflora of rainbow trout intestine was culturable. In some fish, however, the bacteria dominating the intestinal microflora as visualized by 4′,6-diamidino-2-phenylindole (DAPI) staining could not be cultured by standard methods (Spanggaard et al. 2000).
The present study reports identification of the dominant culturable as well as nonculturable bacterial microflora of rainbow trout intestine detected by 16S rDNA sequencing and comparative sequence analysis. The retrieved 16SrDNA sequences were used for the construction of a set of group-specific oligonucleotide probes. These probes were subsequently applied by FISH analysis in conjunction with other group-specific probes to localize and enumerate the dominating bacterial population structure of rainbow trout intestine.
Materials and methods
Sampling of rainbow trout
Rainbow trout were collected from three freshwater fish farms in Denmark (Table 4) and have been described elsewhere (Spanggaard et al. 2000). All fish were killed by a blow to the head and placed on ice and examined in the laboratory within 3 h. The intestinal contents were extracted by dissecting the fish, removing the intestine and squeezing out the contents for microbial analysis. All samples were weighed and homogenized for 1 min on Ultra Turrax T25, (Janke & Kunkel, IKA®-Labortechnik, Staufen, Germany) after addition of 2 ml phosphate-buffered saline (PBS: 130 mmol l−1 NaCl, 10 mmol l−1 NaH2PO4, pH 7·2).
*Total number of fish examined per sampling point.
†Total number of isolates purified and characterized by physiological criteria.
‡Total number of 16S rDNA sequences retrieved from rainbow trout intestine.
§Total bacterial cell count for all fish.
¶Bacterial bacterial plate counts on TSA agar.
**Numbers of PFA-fixed fish intestines analysed in detail for microflora composition (DAPI, CFU, FISH).
1 (22 August 1997)
1 × 105–2 × 108
2 × 104–9 × 107
1·8 × 108
8·9 × 107
2 (11 November 1997)
9 × 103–2 × 107
8 × 102–2 × 106
1·8 × 107
2·0 × 106
3 (10 September 1997)
5 × 103–1 × 107
3 × 102–5 × 106
1·0 × 107
5·0 × 106
3 (9 December 1997)
2 × 105–7 × 108
6 × 103–6 × 108
1·6 × 108
3·2 × 106
Isolation and characterization by traditional techniques
The samples were serially diluted in physiological saline and appropriate dilutions were spread on tryptone soya agar (TSA, Oxoid CM131). Bacterial colonies were isolated after incubation at 15°C for 7 days. A total of 504 isolates were pure cultured and identified (Spanggaard et al. 2000). A rapid tentative grouping of the isolates was performed by random amplified polymorphic DNA (RAPD) analyses of the isolates with the aim of reducing the number of bacterial strains for sequence analysis of the 16S rRNA gene and phylogenetic identification (Spanggaard et al. 2000).
Identification by sequence data analysis
Bacterial pure cultures (146) representing different physiological groups and different RAPD-groups were selected for sequence analyses of the most variable region of the 16S rRNA gene. Almost complete 16S rRNA gene fragments corresponding to nucleotides 8-1541 of Escherichia coli 16S rRNA molecule were amplified in vitro (Springer et al. 1992) and excess primers and deoxynucleotide triphosphates were removed using MicroSpin S-400 HR Columns (Pharmacia Biotech, Inc., Uppasala, Sweden). Direct sequencing of the purified PCR products was performed using rDNA-specific primers (Springer et al. 1992). Cycle sequencing was carried out on both strands of the amplified 16S rDNA using the Thermo Sequenase fluorescent-labelled primer cycle sequencing kit (Amersham, Life Science, Hillerød, Denmark) according to the instructions of the manufacturer and were analysed on an automated DNA sequencer (A.L.F. express, Pharmacia Biotech).
Fifteen full and 131 partial 16S rDNA sequences of the isolates were added to an alignment of about 16 000 homologous full and partial primary structures using the respective automated tools of the ARB software package (Ludwig and Strunk 1997). The partial sequences (5′-end of the 16S rRNA molecule) were used for maximum parsimony analysis and a phylogenetic classification of the isolates. After adding the 16S rRNA gene sequences to the sequence database of the Technical University Munich, the ARB_EDIT tool of the ARB software package (Ludwig and Strunk 1997) was used for sequence alignment. The alignment was corrected manually. The 16S rRNA-based phylogenetic trees were based on the results of maximum parsimony analysis of all available 16S rRNA primary structures for bacteria. The topologies of the different trees were evaluated by performing maximum parsimony analysis of the full data set and subsets, respectively. The full sequences retrieved were also analysed for distance matrix criteria using the neighbour-joining treeing method. The phylogenetic positions of all organisms represented by partial sequences were roughly reconstructed by applying the parsimony criteria without changing the overall tree topology.
Oligonucleotide probes complementary to 16S rRNA target sites of phylogenetic groups of bacterial isolates from the rainbow trout intestine were designed using the PROBE_DESIGN Tool of the ARB program package. The selected oligonucleotide target sites were tested for specificity against the 16S rRNA sequences available in the ARB program package using the PROBE_MATCH tool of the program. Fluorescein-, TAMRA- and CY3-labelled oligonucleotide probes were synthesized and purified by Hobolth DNA Syntese (Hillerød, Denmark). All probe sequences, their hybridization conditions and references are shown in Table 2. Hybridization conditions for the new oligonucleotide probes were optimized as described previously (Manz et al. 1992). Paraformaldehyde (PFA)-fixed bacterial pure cultures with one, two and three mismatches in the probe target region served as reference strains for testing the probes (Table 3). Probe stringency was adjusted by adding formamide to the hybridization buffer in concentrations ranging from 0 to 45% (v/v).
Table 2. Oligonucleotide probes used for FISH analysis of rainbow trout intestinal microbiota
E. coli isolates of Enterobacteriaceae (A28, T25, T116, T120)
Ps. fluorescens AH2
(rRNA group 1)
isolates of Pseudomonads¶
isolates of Acinetobacter spp.¶
(A17, A20, A86, A87)
Ps. fluorescens AH2
Ps. fluorescens AH2
Gill isolate of Pseudomonads¶ (B128)
V. ordali, V. mimicus
Isolate of Curtobacterium¶ (B85)
Isolates of Carnobacterium¶ (D35, D73)
Total cell counts by DAPI staining
The intestinal samples were fixed in ethanol/PBS (1 : 1) and stained with DAPI (Sigma D 9542; 1 μg ml−1) for 20 min in the dark, filtered on polycarbonate filters (diameter 25 mm, pore size 0·2 μm, type GTTP; Millipore, Glostrup, Denmark) and placed on nitrocellulose support filters (25 mm, 0·45 μm; Sartorius, Roskilde, Denmark). Excess DAPI solution was removed, the filters air-dried and immersed in Citifluor (Citifluor Ltd, London, UK). Cell numbers were determined by counting with a grid ocular. Epifluorescence microscopy at ×1000 magnification was performed with a Zeiss Axioskop 20 (Brock & Michelsen, Birkerød, Denmark) using a Chroma filterset (F41-000).
Bacterial strains and whole-cell hybridization
Specificity and stringency of the designed oligonucleotide probes were tested on pure bacterial cultures by whole-cell hybridization. All strains were grown overnight in TSB at room temperature and fixed in 3% PFA (Amann et al. 1990). The whole-cell hybridization with PFA-fixed bacterial strains as positive and negative controls (Table 3) was performed as described by Manz et al. (1992). For probe testing the virulent strains of the fish pathogens Vibrio anguillarum [strain 90-11-287, serotype 01 (Skov et al. 1995)], Yersinia ruckeri (7/97; Biomar A/S, Brande, Denmark) and Aeromonas salmonicida (970618-1/1, Dr I. Dalsgaard, RVAU, Copenhagen, Denmark) as well as the Pseudomonas fluorescens strain AH2 with inhibitory effect towards fish pathogens (Gram et al. 1990) were used. All other bacterial reference strains were obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA) and the German Type Culture Collection (DSMZ, Deutsche Sammlung von Mikroorganismen und Zellkulturen, Braunschweig). The strains Acinetobacter junii (ATCC 17908T), Aeromonas eucrenophila (ATCC 7966T), Aeromonas hydrophila (ATCC 7966T), Aeromonas trota (ATCC 49657T), Cellulomonas flavigena (DSM 20109), Citrobacter freundii (ATCC 29935), Erwinia carotovora (ATCC 15713T), E. coli K-12 (DSM 30083T), Ps. aeruginosa (DSM 20109), Ps. aureofaciens (DSM 6698T) and Rahnella aquatilis (ATCC 33989) were used.
FISH and probe specific cell counts
Intestinal samples were fixed with 3% PFA and immobilized on glass slides. FISH analysis of the intestinal microflora of rainbow trout was performed by applying general group-specific oligonucleotide probes and oligonucleotide probes specifically constructed for this study. The hybridization conditions are described by Manz et al. (1992). Micrographs were taken with a Zeiss Axioskop 20 epifluorescence microscope using Chroma HQ-FITC (F41-001) and HQ-CY3 (F41-007) high quality filtersets as well as using a LSM 510 scanning confocal microscope (Carl Zeiss, Jena, Germany) equipped with an Argon laser (450–514 nm) and a He-Ne laser (543 nm).
DNA-extraction from the fish intestine
DNA from rainbow trout intestine of all fish from fish farm 1 was isolated following the instructions of the FastDNA® Spin Kit for Soil (Bio 101, La Jolla, CA, USA). The DNA was extracted from 200 μl homogenized intestinal content and eluted in 50 μl H2O. The amount and integrity of the nucleic acids were determined visually after electrophoresis on a 1% agarose gel containing ethidium bromide.
The DNA extracted from fish intestine was used for PCR amplification by applying the primer combination GM5F (E. coli position 341 of the 16SrDNA, with GC-clamp) and 907R (E. coli position 907 of the 16SrDNA). A 550-bp fragment of the 16S rDNA was amplified, suitable for subsequent DGGE analysis. PCR amplifications were performed as described earlier (Casamayor et al. 2000) using a hot-start and touchdown annealing reaction to increase the specificity of the amplification. DGGE was performed with a Bio-Rad DCodeTM system (Bio-Rad, Hercules, CA, USA). About 400 ng of PCR product were deposited in each well of the DGGE gel. The urea/formamide gradient of the gel ranged from 20 to 80%. Electrophoresis was performed for 5.5 h at a constant voltage of 200 V at 60°C. After electrophoresis the gels were stained in aqueous SYBR-Green I solution (10 000× diluted) and photographed on a u.v. (302 nm) transillumination table with a polaroid camera. Selected DGGE bands were excised from the DGGE gel, eluted, and reamplified with the same primers but without GC-clamp. Purification and sequencing of the purified PCR products were carried out as described above.
Nucleotide sequence accession numbers
All full 16S rDNA sequences of the bacterial isolates from rainbow trout obtained in this study have been submitted to GenBank under the accession numbers AY374104 to AY374118. Sequences from DGGE analysis have been submitted to GenBank under the accession numbers AY374101 to AY374103.
Identification of bacterial isolates from fish gut by 16S rDNA sequence analysis and phylogenetic analysis
Fifteen bacterial isolates representing the major RAPD groups were fully sequenced and 131 isolates were partially sequenced at the most variable 5′-end of the 16S rRNA gene representing the dominant RAPD groups (Spanggaard et al. 2000). Most of the isolates could be identified to genus level (95% sequence similarity, Stackebrandt and Goebel 1994), but some only to family or subclass level (Table 1). The bacteria isolated from rainbow trout intestine were mainly Proteobacteria of the gamma subclass, but Proteobacteria of the beta subclass, Flavobacteria and Gram-positive bacteria with low and high DNA G + C-content were also identified.
Table 1. Sequence analysis and whole-cell hybridization results of bacterial strains isolated from TSA plates from fish farms 1, 2 and 3
No. of isolates
Length of sequence (bp)
Whole-cell hybridization to specific probes tested
16S rDNA sequencing of 42 isolates from fish farm 1 showed that Enterobacteriaceae dominated the cultured flora (86%). Two full and 29 partial sequences of Enterobacteriaceae were determined. Other bacterial isolates were identified as Plesiomonas, Proteus, Acinetobacter and Aeromonas (Table 2). From fish farm 2 (11 November 1997) 13 bacterial isolates were identified by 16S rDNA sequencing and phylogenetic analysis. By molecular identification (RAPD-analyses and whole-cell hybridization with the probe BET42a) 61% of the cultured bacteria were characterized as members of the beta subclass of Proteobacteria. These isolates were grouped into two phylogenetic groups, one closely related to Duganella zoogloeides, with 98% sequence similarity, and the other to Iodobacter fluviatile as the closest related bacterial pure culture with 93% sequence similarity (Fig. 1). The gamma subclass of Proteobacteria isolates were identified as heterogeneous aeromonads and the Gram-positive bacteria as members of the Staphylococcus and Curtobacterium. Flavobacterium was also identified by 16S rDNA sequencing (Table 1). From fish farm 3 (10 September 1997), 76 bacterial strains were sequenced and the population was dominated by heterogeneous aeromonads (71%) (Table 4). Enterobacteriaceae, Shewanella, Acinetobacter, Plesiomonas, Pseudomonas, Arthrobacter and Rhodococcus were also identified. At a later sampling at fish farm 3 on 9 December 1997, the cultured microflora was dominated by Enterobacteriaceae (54%) as well as by Carnobacterium (33%). Four Carnobacterium, five isolates of Enterobacteriaceae and more representatives of the gamma subclass of Proteobacteria (isolates of the genera Shewanella, Proteus, Aeromonas and Pseudomonas) were identified (Table 1). However, the microscopic analysis of the fish intestines, and also the low culturability rate of 0·1–2%, indicated that the dominant part of the microflora from this sampling was not cultured.
Specific probes constructed for the microflora of rainbow trout
An alignment of 16S rRNA sequences using the ARB ribosomal RNA database containing all 146 sequences of identified isolates was used for the construction of nine specific oligonucleotide probes for the microflora of rainbow trout intestine (Table 2). Because of the close phylogenetic relationship between the fish intestine isolates, most of the probes for the 16S rDNA sequences were designed to be group-specific and tested against species/strains from culture collections and against isolates from rainbow trout intestine (Table 3). In addition, three probes (YER445, ASA446 and VAN71) were developed for the important fish pathogens Yersinia ruckeri, Aeromonas salmonicida and Vibrio anguillarum.
The probes designed for Acinetobacter (ACA432), Aeromonas sobria (ASO642), Yersinia ruckerii (YER445), Yersinia spp. (YER1251 and Curtobacterium (CUR145) did not detect representatives of the bacterial species with only one base pair mismatch (using 20–45% formamide) (Table 3). The probes ACA432 and ASO642 hybridized with fish intestinal strains identified by 16S rDNA sequencing as Acinetobacter and A. sobria, respectively.
The probe ECO180 detected Enterobacteriaceae, including rainbow trout isolates identified as Citrobacter and Enterobacter. The PROBE_MATCH tool of the ARB program showed single strains existing within the family Enterobacteriaceae that could not be detected by the probe, indicating that this probe cannot detect the whole family of the Enterobacteriaceae. The probe AER642 detected the whole genus Aeromonas and hybridized (by whole-cell hybridization) with all 56 isolates identified by 16S rDNA sequencing as Aeromonas (Tables 3 and 4). Representatives of Rahnella aquatilis and Yersinia ruckeri with two mismatches in the probe target region of AER642 were not detected when 35% formamide was used. No bacterial strain with only one mismatch in the probe target region is known. All 10 isolates identified as A. salmonicida by 16S rRNA gene sequencing were detected by the probe ASA446 in whole-cell hybridization experiments, whereas strains with two mismatches were not detected. Twenty-four Aeromonas isolates of rainbow trout intestine could be identified as A. sobria with the probe ASO642.
The probe PSE224 designed for rRNA group 1 pseudomonads hybridized with fish isolates and did not detect bacterial isolates (Acinetobacter) with two mismatches in the probe target region. Under stringent conditions the probe PAH64 developed for the detection of the probiotic strain Ps. fluorescens AH2, also detected P. viridiflava and Ps. aureofaciens. It was not possible to construct a 16S rRNA-targeted probe specific for the strain P. fluorescens AH2 alone because of the close relationship within the bacterial genus Pseudomonas. The probe VAN71, which was constructed specifically for Vibrio anguillarum, did not detect bacterial strains with two mismatches in the probe target region, but did, even under stringent hybridization conditions, also detect the species V. ordalii, V. mimicus and V. cholerae.
The probe CAR193 was specific for the genus Carnobacterium including the species Lactobacillus maltaromicus. All isolates of rainbow trout intestine that were identified as Carnobacterium were detected by the probe. The probe was not tested against bacterial negative controls, but no known bacterium (corresponding to the ARB database of ca 16 000 16S rRNA sequences) showed fewer than three mismatches in the probe target region. The probe was used with 25% formamide in the hybridization buffer and the signal disappeared if higher formamide concentrations were applied in whole-cell hybridization experiments.
As described in Spanggaard et al. (2000), the counts of the intestinal microflora varied by 3–5 log units between fish within the same sampling point (Table 4) and, in general, high culturability (11–50%) was found. At fish farm 3, several samples contained high counts of noncultured bacteria (Spanggaard et al. 2000). One fish intestine contained 98% of not culturable bacteria (clusters of small coccoid bacteria) and was selected for further analysis. Subsequently, these coccoid bacteria were also identified in other fish at this sampling point (9 December 1997).
Identification of the uncultured bacterial microflora of the rainbow trout intestine by DGGE fingerprinting and phylogenetic affiliation
Three DGGE bands could be detected in all intestinal samples of the sampling point 9 December 1997 from fish farm 3. These were excised, re-amplified and sequenced (Fig. 2). By phylogenetic analysis, the first band sequence was identified as belonging to the Carnobacteria with 98% sequence similarity to Carnobacterium piscicola (DFI 3), the second band belonged to the Clostridia with 95% sequence similarity to Clostridium botulinum (DFI 2). The third sequence did not belong to any well-known phylogenetic lineage (DFI 1). The closest related bacterial pure culture was Anaerofilum pentosovorans with 89% sequence similarity of the 16S rRNA (Fig. 3).
Oligonucleotide probes for the identification of the yet uncultured intestinal microflora
Three specific oligonucleotide probes were constructed for the recovered sequences from the DGGE fingerprints of rainbow trout intestine from fish farm 3 (9 December 1997). The probes labelled with CY3 were applied to PFA-fixed intestinal samples of the same sampling point. The probe UNK730 constructed for the sequence DFI 1 could specifically detect the clusters of small coccoid bacteria (Fig. 4). By in situ hybridization of rainbow trout intestines, no bacteria could be detected with the probes UNK712, constructed for the sequence DFI 3 with 98% sequence similarity to Carnobacterium piscicola, and UNK835, constructed for the sequence DFI 2 with 95% similarity to Clostridium botulinum. The probes were applied without adding formamide to the hybridization buffer. They were not tested with bacterial pure cultures as positive controls in whole-cell hybridization experiments as the probes detected uncultured bacteria. The probe UNK730 was subsequently applied to nine fish samples from fish farm 3 where DAPI counts were 2–3 log units higher than the culturable counts (Spanggaard et al. 2000). In five of nine samples, cells hybridizing to UNK730 were found in high concentrations whereas such cells were found on lower amounts in four samples. The UNK730 hybridized to 10–100% of the cells in a microcopic view field.
In situ analysis of rainbow trout intestine with oligonucleotide probes
One fish of each fish farm and sampling point was assessed for a high resolution FISH analysis of the microbial composition of the intestine. This approach is limited by the number of bacteria present per microscopic field and a detailed in situ analysis was only carried out on the fish with highest bacterial counts in the intestine, varying from 1 × 107 to 7 × 108. Bacterial numbers were analysed using general group-specific probes and all probes constructed in this study (Table 2). The intestinal microflora was quantified by microscopic counting using individual probes compared with total bacterial counts with the probe EUB338-mix detecting all bacteria in the sample (Fig. 5). For the fish analysed by in situ hybridization in detail the DAPI : EUB338-mix ratio was also determined. The specific probes revealed that each fish carried its own predominant bacterial population (Fig. 5).
Enterobacteriaceae (ECO180) dominated the intestinal microflora of the fish from fish farm 1 (22 August 1997) with total bacterial counts of 1.8 × 108 (Figs 5 and 6a,b). Bacteria belonging to the genera Aeromonas (AER642) and Acinetobacter (ACA432) and low G + C Gram-positive bacteria could be detected in lower numbers. About 85% of all bacteria were detected by FISH analysis and 89% of the total cell count (DAPI) was detected by the bacterial probe EUB338-mix. In the fish investigated from fish farm 2 (11 November 1997), with total bacterial counts of 1·8 × 107, Proteobacteria belonging to the beta and gamma subclass dominated (Figs 5 and 6c,d). The gamma-Proteobacteria were dominated by Aeromonas (AER642). Acinetobacter (ACA432) and Pseudomonas (PSE224) were detected in situ but had not been isolated on TSA. Twelve per cent of the flora were Gram-positive bacteria and were detected by LGC354a/b/c (10%) and HGC69a (2%). About 84% of all bacteria (EUB338-mix) were detected by FISH analysis and 83% of the total cell count (DAPI) were detected by the bacterial probe EUB338-mix. From fish farm 3 the intestinal contents of two fish of two different sampling points were analysed. One fish with a total of 1 × 107 bacteria in the intestine, from 10 September 1997, was dominated by representatives of the genus Aeromonas (AER642, Fig. 6e,f), while the fish with total bacterial counts of 1·6 × 108 from 9 December 1997 was dominated by clusters of very small coccoid bacteria (UNK730, Fig. 4a,b), which could not be cultured (72% of the bacterial flora). Twelve per cent of the microflora was identified as belonging to the Gram-positive bacteria with low DNA G + C content. About 80 and 84% of the bacterial flora were detected by the oligonucleotide probes applied, respectively. The DAPI : EUB338-mix ratio was 86% on 10 September 1997 on 97% on 9 December 1997.
The use of molecular methods, and in particular the 16S rRNA approach for microbial communities, has previously provided insight into the diversity and population dynamics of bacterial communities in different habitats including seawater (Giovannoni et al. 1990), activated sludge (Wagner et al. 1993), the rumen (Stahl et al. 1988), termite intestines (Ohkuma and Kudo 1996) and soil (Liesack and Stackebrandt 1992). In the present study, we performed a molecular analysis for cultivation-independent identification of the dominant microflora. Although complete 16S rDNA sequences should be used for phylogenetic reconstruction (Ludwig and Strunk 1997), partial sequences of the most variable part of the 16S rRNA gene were, in our study, sufficient to determine the closest relatives for unknown sequences and to assign them to well-established phylogenetic groups (Muyzer et al. 1995; Stackebrandt and Rainey 1995).
In Arctic charr some micro-organisms appear to be associated with the gut wall (Ringøet al. 2001a), but in rainbow trout electron microscopy studies (Austin and Al-Zahrani 1988) and direct FISH analysis (I. Huber, unpublished data) point to a general lack of colonization of the gut wall, with the vast majority of the organisms residing in the lumen. We therefore believe that our data (based on the squeezed intestinal contents) reflects the microbial population of the gastrointestinal tract of trout.
For many ecosystems it has been shown that only a fraction of the total number of bacteria present in the sample can be cultured (Amann et al. 1995; Berg 1996) or that cultivation-based identification can be biased because of media selectivity (Wagner et al. 1994; Amann et al. 1995). However, we typically found high culturability rates and, in general, good agreement was seen between direct in situ identification and flora composition as based on the culturable flora.
Although culturability rates were relatively high (11–50%), our data also indicate that some organisms in the trout intestine have not yet been cultured. Thus FISH analysis demonstrated that 20% of the total cell counts (DAPI) were not detected by the bacteria-specific probe EUB338-mix. These cells could either be nonbacterial or be cells that have not been permeabilized by the applied conditions for FISH.
In samples where cultivation rates were only 0·1–2%, a direct amplification followed by DGGE analysis and sequencing allowed us to identify the uncultured clusters of cocci bacteria as a new bacterial phylogenetic lineage with 89% sequence similarity to Anaerofilum pentosovorans. This new microbial population detected in the sample reflected the strongest band in the DGGE gel. FISH analysis confirmed the dominance of this microbial population in rainbow trout intestine in several samples with low culturability rates. The other two sequences identified were members of the genera Clostridium and Carnobacterium both of which have been identified from fish intestine (Sugita et al. 1997; Ringøet al. 2001b). The specific probes constructed for these two bacterial populations did not detect bacterial cells in situ which could either be due to malfunctioning probes or to a significantly lower proportion than that of the dominant cocci clusters. Working with mixtures of pure cultures, Muyzer et al. (1993) found that DGGE analysis enabled detection of populations with an abundance of <1% of the total cell count. As it is assumed that the primers used for PCR amplification are general for bacteria, DGGE fingerprints should not be biased towards known bacterial groups, as they rely on direct rRNA gene retrieval from environmental DNA. However, biases in this molecular approach may occur in the efficiency of cell lysis (Picard et al. 1992), DNA extraction and purification (Liesack et al. 1991) as well as during PCR amplification (Reysenbach et al. 1992). Therefore it cannot be expected that the full species richness of total microbial diversity in a complex habitat can be estimated with this method. Strictly anaerobic bacteria were not investigated in our study, but have previously been isolated from fish gut (Sakata et al. 1980b; Austin and Al-Zahrani 1988). Ringøet al. (1995) have previously suggested that the predominant bacteria isolated from the salmonid gut are aerobes or facultative anaerobes.
It has been suggested that the microbial flora of fish intestine consists of bacteria that are also present in the surrounding water, but which are able to persist and multiply in the environment provided by the intestinal tract (Sugita et al. 1988; Cahill 1990). In agreement with this, both the cultivation approach and the direct molecular FISH approach demonstrated that Proteobacteria of the gamma subclass dominated the microflora. Fermentative bacteria of the Enterobacteriaceae, Vibrionaceae and Aeromonodaceae are fast growing organisms that are likely to thrive in the fish gastrointestinal tract with its conditions of low pH, lack of oxygen and ample nutrients. It has previously been described that faecal streptococci and coliforms can enter fish farms via incoming water and feed, and have the potential to multiply in the fish intestine (Cahill 1990). One group of isolates within the beta-Proteobacteria could not be identified to genus level. This group showed 93% 16S rRNA sequence similarity to Iodobacter fluviatile. Closely related bacterial isolates (more than 99% 16S rDNA sequence similarity) were identified on the skin of fish from the same fish farm at different sampling points (results not shown), indicating that this group of isolates represents bacteria from the surrounding water that were established on the outer surfaces of the fish.
Rearing conditions of the fish and seasonal changes (e.g. temperature) may influence fish microflora (Cahill 1990). Our results support these findings for the rainbow trout intestine as the fish of every sampling point had a specific predominant intestinal microflora. Large animal-to-animal variations in the population sizes of the dominant bacteria have been found in the gastrointestinal tract of pigs by the use of culturing methods (Robinson et al. 1984). High variations in bacterial counts between individuals have also been reported for fish (Trust and Sparrow 1974; Yoshimizu and Kimura 1976; Sakata et al. 1980a) and were confirmed by our results.
In conclusion, the combination of bacterial isolation, DGGE analysis, 16S rDNA sequencing and subsequent FISH analysis with oligonucleotide probes specifically developed for the set of rRNA sequences retrieved proved to be a powerful tool for detailed insight into the microbial diversity of rainbow trout intestine. A high bacterial diversity was found in the gut, although the fish intestine seemed to have a predominant bacterial flora at each sampling point. Molecular tools proved essential to answer the many questions concerning the composition, structure and stability of this microbial ecosystem. We showed that molecular methods can be applied for a detailed analysis of the microbial community structure of fish intestine. Future studies should use the molecular approach described to determine the stability of the microflora of individual fish over time as dependent on feeding, season and antibiotic administration. Such studies will be crucial in determining whether the microflora is established in the fish gut or whether it is indeed transient, representing the flora of the surrounding water.
Knud Fischer, Ib Knudsen, Binderup Mølle fish farm and Biomar A/S (Esben B. Sick) in Hirtshals kindly provided the rainbow trout. The authors thank Michael Schmid for generous assistence in image processing, Thomas H. Roberts for critical reading of the manuscript and Randi Brundstedt and Flemming Jørgensen for kind help with preparation of the manuscript. The study was financed by the Danish Ministry of Food, Agriculture and Fisheries.