Bacteriology of the Labrador dog gut: a cultural and genotypic approach


Correspondence to: H.L. Greetham, Food Microbial Sciences Unit, School of FoodBiosciences, The University of Reading, Whiteknights, Reading, RG6 6AP, UK (e-mail:


Aims: To carry out an extensive study of the microflora composition of the Labrador dog gut.
Methods and Results: Faecal specimens from four Labradors were collected and plated onto growth media designed to recover total anaerobes, bacteroides, bifidobacteria, lactobacilli, clostridia, Gram-positive cocci, total aerobes and coliforms. Morphologically different isolates were collected from all agars inoculated with faeces from one canine individual (repeated four times). A total of 157 out of 171 isolates were identified using 16S rRNA gene sequencing. Sequence analysis showed that agar selectivity was poor, especially when bacteroides and Gram-positive cocci were the targets. Bifidobacteria were not detected in any of the samples analysed, indicating their presence at low or negligible levels. The gene sequences of many of the isolates (n=45, representing 29% of the total) did not correlate with known species in the Ribosomal Database Project and EMBL databases, suggesting the presence of novel gut diversity.
Conclusions: Traditional culture methods fail to reflect the bacterial diversity present in Labrador dog faeces.
Significance and Impact of the Study: This study has shown the value of molecular-based methodologies for determining bacterial profiles in the Labrador dog gut microbiota, but has also exposed the limitations of purportedly selective agars.


During the last decade there has been significant interest in the use of dietary procedures targeted towards improved gastrointestinal health. For humans, both probiotics and prebiotics are widely used to fortify bacteria seen as beneficial (Gibson et al. 2000). The approach of human gut flora modulation currently enjoys much scientific and commercial attention; however, historically, probiotics have been more widely used in animals, especially on the farm (Fuller 1992). Here, live microbial feed additions are mainly used to increase weight/yield. More recently, there has been a move towards the use of probiotics in the petfood market, where animal wellbeing is a major concern (Fuller 1989; Martineau 1999). This may be driven by the fact that one of the most useful aspects of probiotic use is improved resistance to transmitted infections. To date, all probiotics for petfood use are strains isolated from humans or farm animals. However, one selection criterion for efficacious forms includes isolation from the same species as the intended use (Huis in't Veld and Shortt 1996).

Interest in probiotic strains has led to recent cultural studies directed towards the isolation of lactobacilli and bifidobacteria (most commonly used genera) from dog faeces (Hartemink and Rombouts 1999; Martineau 1999). Identification of the strains in these studies was based upon Gram reaction and gas liquid chromatography analyses of metabolic end products, and was therefore presumptive. However, both studies concluded that growth media currently used in human gut flora studies may not be the preferred choice for animal work.

To measure ‘beneficial’ effects, quantification of bacterial populations must be accurate. A major current drawback to microbiological analyses of faecal material relates to the methodologies used. Gut bacteria are mostly characterized using phenotypic approaches. These techniques are laborious, because identification of individual colonies is necessary and purportedly selective agars allow growth of bacteria other than target taxa. Microorganisms also exhibit metabolic plasticity, leading to unreliable biochemical-based results. Molecular-based methodologies for high fidelity identification of isolates offer the potential to overcome such drawbacks (Woese 1987; Drancourt et al. 2000).

Quantitative and qualitative knowledge of the structure of the bacterial community in the intestinal tract is essential to understand complex interactive processes and the impact on health status of the host. Moreover, such knowledge can then be applied to the development of new probiotic and prebiotic products. Here, we report the use of a cultural approach to quantify predominant faecal flora components of the canine gut and identification of generated colonies through genotypic procedures. The information is relevant for gut flora diversity studies as well as more directed work on the deriving and testing of probiotics for domestic animals.

Materials and methods


A male Labrador (born 2/8/92) referred to as L195 was used as the main subject. Three other female Labrador subjects were used for culturable comparisons, these being referred to as L198 (born 2/8/92), L325 (born 9/10/93) and L1519 (born 16/11/95). All dogs were maintained on a basic dry petfood diet ‘Formula Advance’, for at least a month previous to sampling, and L195 was maintained on the diet between all sampling dates.

Faecal specimens

Within a two-week time period, two samples were collected from L195, one week apart, and a single sample was collected from each of the other three subjects. The samples were collected immediately after defaecation and used to prepare a 10% (w/v) slurry using prereduced 0·1 mol l−1 phosphate-buffered saline (pH 7·0). This was then transferred into an anaerobic cabinet (10% H2; 10% CO2; 80% N2) and homogenized for 10 min. Serial tenfold dilutions of the faecal material were prepared (up to 10−9) using half strength peptone water and cysteine-HCL (0·5 g l−1). Samples were then plated out onto agars designed to select for a range of predominant gut microorganisms, i.e. total aerobes, total anaerobes, clostridia, bifidobacteria, bacteroides, coliforms, Gram-positive cocci, and lactobacilli. Details of the growth media are given in Table 1. For each dilution, triplicate samples were inoculated onto triplicate plates.

Table 1.  Media used and their target bacterial groups from faeces
MediumTarget populationReference
  1. *Brucella agar with kanamycin (75 mg l−1), vancomycin (7·5 mg l−1), hemin (5 mg l−1), vitamin K (10 mg l−1) and laked horse blood (50 ml l−1).

  2. †Columbia agar with agar (5 g l−1), glucose (5 g l−1), cysteine HCL (0·5 g l−1) and propionic acid (5 ml l−1) to pH 5.

  3. ‡Reinforced Clostridial agar with novobiocin (8 mg l−1) and colistin (8 mg l−1).

Nutrient agar (NA)total aerobesLapage et al. 1970
MacConkey agar (MAC)coliformsWilson and Miles 1964
Wilkins–Chalgren agar (WC)total anaerobesWilkins and Chalgren 1976
Beerens agar (BEE)bifidobacteriaBeerens 1990
Rogosa agar (ROG)lactobacilliRasic 1984
Azide agar (AA)Gram-positive cocciWood 1965
Reinforced Clostridial agar (CLOS)clostridiaFujisawa et al. 1995
Bacteroides* agar (BAC)bacteroidesHoldeman and Moore 1973

Pre-reduced agar plates were incubated in the anaerobic cabinet at 37°C for up to 5 d, whilst those selective for facultative anaerobes/aerobes were placed in a 37°C incubator for up to 48 h. All plates were examined daily for new colony types and enumerated. All colonies recovered from L195 were isolated on to their original growth medium, until pure (as determined by Gram staining and plate examination) and subsequently stored on cryogenic beads (in Microbank tubes) at − 70°C. A further two samples from L195 were processed as described before, six months later.

Genotypic identification of isolates

Pure colonies for genotypic identification were vortexed, centrifuged at 13 000 g for 3 min and the supernatant aspirated. The resultant bacterial pellet was stored at − 20°C until required for DNA extraction. Total DNA was extracted using an InstaGene Matrix (Biorad, Hercules, California, USA) according to the manufacturers instructions. The 16S rRNA genes were then amplified by PCR using conserved primers proximal to the ends of the gene (primer pA, sequence 5′−3′, AGAGTTTGATCCTGGCTCAG, positions 8–27 E. coli numbering; primer pH, sequence 5′−3′, AAGGAGGTGATCCAGCCGCA, positions 1541–1522 E. coli numbering; primer pD*, sequence 5′−3′, GTATTACCGCGGCTGCTG, positions 539–522 E. coli numbering; primer γ, sequence 5′−3′, ACTGCTGCCTCCCGTAGGAG, positions 358–339 E. coli numbering; primer ANTI γ, sequence 5′−3′, CTCCTACGGGAGGCAGCAGT, positions 339–358 E. coli numbering; primer pD, sequence 5′−3′, CAGCAGCCGCGGTAATAC, positions 522–539 E. coli numbering; primer 3, sequence 5′−3′, GTTGCGCTCGTTGCGGGACT, positions 1109–1090 E. coli numbering; primer ANTI 3, sequence 5′−3′, AGTCCCGCAACGAGCGCAAC, positions 1090–1109 E. coli numbering; primer A, sequence 5′−3′, CGGTGTGTACAAGGCCC, positions 1398–1382 E. coli numbering). In a final volume of 50 μl, the reaction mixture contained amplification primers, premixed deoxynucleoside triphosphates, MgCl2, template, and Taq DNA polymerase, which was added after a precycling stage whereby the reaction mixture was heated to 96°C for 4 min and then held at 0°C. The PCR amplification conditions were 1 min at 95°C, 1 min at 55°C, and 1·5 min at 72°C for 30 cycles. Following the final cycle, the reaction was extended at 72°C for 10 min and then held at 10°C. Amplification products were visualized by electrophoresis using a 1% (w/v) agarose gel in 1 × TAE (40 mmol Tris-acetate, 1 mmol EDTA) containingethidium bromide (0·5 μg ml−1). Reaction products which displayed bands corresponding to the correctly sized products (1500 base pairs for primers pA-pH) were purified using a QIA-quick PCR purification kit (QIAgen, West Sussex, UK) according to the manufacturer's instructions.

For sequencing of PCR products, approximately 500 nucleotides proximal to the 5′ end of the rRNA were targeted using a dRhodamine terminator cycle sequencing kit (PE Applied Biosystems, Inc., Foster City California, USA) and a model 377 automatic DNA sequencer (PE Applied Biosystems). After initial screening of isolates was carried out, if organisms displayed ≤98% sequence similarity to described species, then sequence analysis was extended over the full PCR fragment. Generated sequences were compared with 16S rRNA gene sequences available in the GenBank/EMBL database using the FASTA program and the Ribosomal Database Project (RDP) to ascertain percentage similarity scores (Pearson and Lipman 1988; Maidak et al. 1999).

All chemical reagents were obtained from Sigma (Poole, UK) and growth media were purchased from Oxoid (Basingstoke, UK) unless otherwise stated.


Bacterial enumerations

Total bacterial counts (colony-forming units log10/gram [wet weight]) were recorded from eight different agar types, designed to select for predominant groups of gut bacteria, in stool samples from four Labradors (Fig. 1). The agars used are conventionally employed in studies on the human gut microflora. Generally, populations of total anaerobes outnumbered aerobes by 100-fold, with counts being similar to those previously reported by Matsumoto and Baba (1972) and Buddington and Sunvold (1998). Bacterial counts between the same individual showed little deviation. Moreover, microbial counts of L195 were similar to those of L198 and these dogs came from the same litter. The youngest dog, L1519, exhibited highest counts on all agars tested apart from MacConkey's. Counts of ‘presumed’ bacteroidesspecies were highest in all samples, followed by Gram-positive cocci, clostridia, aerobes, and lactic acid bacteria, respectively. All counts were tentative identities at this stage.

Figure 1.

Canine total nondiscriminative counts for cultured faecal sample microflora, on a range of eight agar types. Four different samples were enumerated from L195 (sample collection dates 03/11/98, 10/11/98, 20/05/99 and 27/05/99 respectively), and single samples from three other individuals. Standard deviation error bars are displayed. Triplicate enumerations of each agar were carried out from a triplicate dilution series. A, Nutrient Agar; B, MacConkey agar; C, Wilkins–Chalgren agar; D, Rogosa agar; E, Beerens agar; F, Bacteroides agar; G, Reinforced Clostridial agar; H, Azide agar

Genotypic culture identities on the agars used for dog L195

Four separate faecal samples were processed from dog L195. All colonies that grew on these agars were stored andsubsequently analysed using the genotypic procedures described earlier. This involved the sequencing of 157 bacterial colonies, with a further 14 not recovering from the storage procedure used. Results of the 16S rRNA sequencing of these isolates are shown in Table 2. The species identifications are based on sequence similarity values of 98% or higher, to existing nucleotide sequences contained in the databases. Using this approach, it has been possible to assess the reliability of selective agars. In some cases, the sequences showed less than 98% similarity, indicative of potentially new microbial diversity. In these cases, more detailed genotypic and phenotypic analyses are ongoing. However, based on the phylogenetic identifications shown, it is still feasible to ascertain the reliability of the agars used here.

Table 2.  Percentage coverage and counts of L195 faecal sample isolates on agar; identities determined by 16S rRNA gene sequencing
 Sample 1Sample 2Sample 3Sample 4
AgarPercentage coverageCountsPercentage coverageCountsPercentage coverageCountsPercentage coverageCounts
  1. N, negligible counts; %, percentage coverage of isolates on agar; bold text signifies an organism which the agar is selective for; figures displayed are bacterial counts (log10 cfu g−1 of wet weight faeces).

Nutrient98%Escherichia coli7·5100%Escherichia coli6·380%Escherichia coli5·797% Escherichia coli6·2
 2%Streptococcus bovis5·8  19%Streptococcus bovis5·13%Staphylococcus hominis4·7
     1%mixed culture ruminal  bact.3·5  
MacConkey98%Escherichia coli6·9100%Escherichia coli6·085%Lactobacillus murinus5·898%Escherichia coli6·1
 2%mixed culture ruminal bact.5·2  15%Escherichia coli5·12%Lactobacillus murinus4·5
     N mixed culture ruminal  bact.2·7  
Wilkins– Chalgren60%Collinsella intestinalis7·553%Collinsella intestinalis7·143%Collinsella-like sp. 28·740%Bacteroides-like sp. 18·1
 27%Pectinatus-like sp. 17·133%Pectinatus-like sp. 16·923%Clostridium hiranonis8·432%Staphylococcus pasteuri8·0
 11%Streptococcus bovis6·76%Clostridium perfringens6·19%Collinsella-like sp. 18·011%Collinsella intestinalis7·5
 1%Escherichia coli5·84%Fusobacterium-like sp. 16·08%Pectinatus-like sp. 18·05%Escherichia coli7·2
 1%Rothia-like sp. 15·64%unknown5·95%Collinsella intestinalis7·74%Clostridium hiranonis7·1
 N unknown4·7N Fusobacterium-like sp. 24·95%Coriobacterium-like sp. 17·74%Sutterella-like sp. 17·1
 N Staphylococcus sp.4·7N Clostridium hiranonis4·93%Fusobacterium varium7·4 4%Fusobacterium varium7·1
     2%Clostridium-like sp. 37·4  
     1%Staphylococcus epidermidis7·1  
     N Clostridium-like sp. 46·4  
Rogosa65%Streptococcus bovis6·388%Lactobacillus ruminus5·9100%Lactobacillus murinus5·898%Lactobacillus murinus5·8
 24%Lactobacillus animalis5·912%Lactobacillus reuteri5·0  2%Lactobacillus reuteri4·1
 11%Lactobacillus ruminus5·5    N Lactobacillus acidophilus2·7
Beerens67%Streptococcus bovis6·771%Streptococcus bovis6·072%Lactobacillus murinus5·896%Lactobacillus murinus5·7
 18%Lactobacillus ruminus6·127%Staphylococcus sp.5·628%Clostridium-like sp. 25·44%Streptococcus bovis4·3
 15%Staphylococcus epidermidis6·02%Lactobacillus ruminus4·2    
Azide93% unknown7·267% unknown6·578%Propionibacterium acnes7·773%Clostridium perfringens7·6
 7%Streptococcus bovis6·124%Pectinatus-like sp. 16·113%Eubacterium-like sp.16·910%Catenibacterium mitsuokai6·7
 N Fusobacterium varium4·76%Clostridium perfringens5·56%Clostridium perfringens6·69%Collinsella intestinalis6·7
 N Clostridium perfringens4·73%Fusobacterium-like sp. 15·22%Pectinatus-like sp. 16·15%Fusobacterium varium6·4
     1%Lactobacillus murinus5·82%Pectinatus-like sp. 16·1
       1%Lactobacillus murinus5·4
Clostridial80%Staphylococcus sp.6·899%Eubacterium-like sp.16·8100% unknown7·647%Citrobacterium freundii6·7
 20% unknown6·21%Streptococcus hansenii4·8N Clostridium-like sp. 14·740%Clostridium-like sp. 16·6
     N Clostridium -like sp. 44·713%Staphylococcus epidermidis6·1
     N Clostridium perfringens4·7  
     N Corynebacterium group4·7  
Bacteroides56% unknown6·856%Fusobacterium-like sp. 16·699% unknown7·791%Escherichia coli8·4
 33%Escherichia coli6·617%Staphylococcus pasteuri6·11%Staphylococcus pasteuri5·78% unknown7·3
 6%Staphylococcus haemolticus5·915%Fusobacterium varium6·0N Lactobacillus murinus5·21%Fusobacterium varium6·2
 5%Lactobacillus animalis5·812%Lactobacillus ruminus5·9N Fusobacterium-like sp. 15·2N Lactobacillus murinus5·7
 N Staphylococcus hominis4·7  N Staphylococcus epidermidis4·7N Micrococcus luteus5·7
 N Lactobacillus ruminus4·7  N Staphylococcus capitis4·7  

Selectivity of the agars used

Table 1 lists the agars used in this study and the microbial groups they were designed to recover. Table 2 shows the percentage recovery, counts and genotypic identities of different isolates from these agars. As shown in Table 2, E.coli was the predominant microorganism detected on Nutrient agar. This was also the case for MacConkey's medium, which was reasonably selective for coliforms. Varying diversity was isolated from the Wilkins–Chalgren plates, with Gram-positive rods (mainly Collinsella and Clostridium) being predominant in all samples. Several novelisolates (previously not genotypically described) were also found using this agar. Rogosa appeared to be areasonably selective growth medium for lactobacilli, although streptococci were also recovered in sample 1. Species of Lactobacillus varied, with Lactobacillus ruminus, L. animalis and L. reuteri being recovered in runs 1 and 2, but L.murinus predominating in samples 3 and 4. Beerens agar was designed for the isolation of bifidobacteria. However, a mixture of streptococci, lactobacilli, staphylococci and a Clostridium-like organism were isolated with no bifidobacteria being detected. Streptococcus bovis was the mostfrequently found species in samples 1 and 2, and Lactobacillus murinus in samples 3 and 4. On the Azide agar, clostridia were frequently detected but Gram-positive cocci (the target organisms) were absent. On Reinforced Clostridial agar a high proportion of the microflora detected did not correspond to clostridia. Using the Bacteroides agar, strains of lactobacilli, fusobacteria, E.coli and staphylococci were found to be present in all samples tested, with bacteroides not being recovered on this selective agar. Populations were seen to vary over time (different samples). An overall similarity can be seen between sample sets 1 + 2 and 3 + 4.


Through genotypic identification of bacterial isolates obtained in the culture work, agar selectivity was assessed. MacConkey's agar performed best, in that growth of coliforms alone was observed, as was found previously by Jones and Martineau (1999). Escherichia coli, streptococci, and occasionally staphylococci, were found to populate the Nutrient agar. Rogosa agar was found to be fairly selective for lactobacilli, although Streptococcus bovis was also isolated and even predominated in one of the samples tested. Selectivity of the other agars used was poor, e.g. Reinforced Clostridial and Azide agars. Beerens and Bacteroides agars never recovered their target organisms. It may be argued that these genera were not present in the Labrador dog gut. In particular, bifidobacteria were never recovered on any of the agars used; however, this was not the case for Gram-positive cocci. Nevertheless, the reliability of selective agars seems dubious, thereby necessitating the use of high fidelity procedures to confirm culture identities.

Populations were seen to vary between samples. Lactobacillus murinus appeared to be the predominant Lactobacillus species in samples 3 and 4, isolated from both Rogosa and Beerens agars. However, L. murinus was not isolated from samples 1 or 2. In agreement with Benno et al. (1992) who concludes microbial populations change with time, this observation suggests that bacterial populations may vary independently from the diet. When L. murinus was present, a lower variety of other lactobacilli was observed, possibly indicating some form of competitive exclusion.

Bifidobacteria were not isolated in this study. Martineau (1999) and Hartemink and Rombouts (1999) reported a similar finding, although without the support of genotypic evidence. This may be due to bifidobacteria being absent or present in very low numbers and being out competed by other taxa such as the lactobacilli. Willard et al. (2000) was aware of an inability to constantly find bifidobacteria in faecal samples from dogs. This may be due to an absence of appropriate growth substrates in the canine diet for bifidobacteria. As such, Beerens agar may be the selective choice for human studies but cannot be relied upon for canine work.

Some clearly novel bacterial species were isolated in this study. These showed large variations from any of the sequence entries in the databases searched (Ribosomal Database Project and EMBL). Because of the laborious nature of bacterial identifications, previous studies have only attempted identification of 5–6 predominant colony types from each agar type used (Clapper 1970; Martineau 1999). The 14 isolates, which remained unknown due to difficulties with their regrowth, again demonstrate the limitations of culture-based work.

The main conclusion of the present study is that the cultural approach fails to reflect much of the diversity present in faecal samples. Moreover, a large amount of the microbiota diversity remains undetermined, and selective agars particularly designed for strict anaerobes show poor recovery efficiencies. As interest in the use of probiotics for domestic animals increases, it is important that robust recovery and identification procedures are used.


HLG was in receipt of a BBSRC Case studentship (with the WALTHAM Centre for Pet Nutrition).