The microbial ecology of the canine gastro-intestinal (GI) tract is a rapidly expanding research area in veterinary medicine. The intestinal tract harbours a large number of prokaryotes, mainly bacteria, which exceed the number of host cells. Complex interactions exist between the eukaryotic and prokaryotic components; the latter are important in maintaining the health of the former by playing a vital role in the normal nutritional, physiological, immunological and protective functions of the host [1]. The amount and form of food, feeding frequency and diet composition are known to have important effects on GI function. Both nutrients and non-nutritional dietary components influence gut health in terms of intestinal microbiota [2]. Alterations in the intestinal microbiota or aberrations in immune responses to its components are hypothesized to play a crucial role in the pathogenesis of enteropathies (e.g., inflammatory bowel disease, dietary intolerance, sensitivity and allergy) [1]. An important focus of canine research has been the effect of different diets on satiety, faecal consistency and quantity of Clostridium perfringens in faeces [3, 4]. Recently, two studies have been published about the fluctuations in canine faecal bacterial populations caused by dietary changes [5, 6]. Human studies using conventional culturing techniques have indicated that the protein and fat content of the diet as well as the nature of the carbohydrates (simple sugars vs. complex carbohydrates) does affect microbiota composition and activity [7]. Studies in chickens, rats and mice support the hypothesis that the intestinal microbiota can be modified by diet [8–10].

To date, seven bacterial groups (Bacteroides, Clostridium, Lactobacillus, Bifidobacterium, Fusobacterium, Enterobacteriaceae and Coriobacterium) in five predominant phyla (Firmicutes, Fusobacteria, Bacteroidetes, Proteobacteria and Actinobacteria) have been identified from different parts of the canine intestine using culture techniques and/or various molecular methods [6, 11–13].

Previously, the main technique for studying the canine GI tract microbial community structure has been the cultivation of bacteria from intestinal contents [12, 14]. However, the cultivable bacteria in the animal intestinal tract represent only a fraction of the microbes actually present in the gut, obviating the need for high-throughput molecular approaches, many of which rely on the sequence of the bacterial 16 S rRNA genes that serve as phylogenetic markers [15]. Although these molecular techniques have apparent benefits over the traditional bacterial culturing, they possess several potential pitfalls that must be taken into account. The main issues for quantification of nucleic acids in complex communities relate to the repeatability of DNA extraction from different bacteria or sample types with constant efficiency and sufficient purity [16]. The profiling of guanine-cytosine content (%G + C) is a technique for initial investigation of bacterial populations of previously unknown structure. The great benefit of this method over other DNA-based methods is its lack of dependence on any a priori information about the bacteria being analysed. Moreover, no PCR amplification is required, which is known to introduce artefacts with an increasing number of cycles [1, 17]. This technique relies on the separation of chromosomal DNA of various bacterial species by density gradient centrifugation and yields a profile based on their characteristic guanine-cytosine content [18]. The individual G + C fractions from the pool of bacterial chromosomal DNA with any G-C content can be collected for subsequent detailed analyses, including cloning and sequencing of the 16 S rRNA genes [18, 19]. This approach been successfully employed to study microbial community structures in a variety of environments, such as soils, or the GI tracts of humans or different animals [18–22].

As mentioned above, specific interactions between diets differing in macronutrient content and microbiota composition have rarely been investigated with 16 S rDNA-based molecular tools in dogs. Therefore, the aim of the current study was to investigate the alterations in canine intestinal microbiota due to dietary changes by applying %G + C profiling for total community analysis, followed by sequencing of relevant fractions of nearly complete 16 S rRNA gene fragments.

Graeco-Latin square design was used to evaluate the influence of high-carbohydrate (HC), high-protein (HP) and dry commercial (DC) diets on the colonic microbiota of five Beagle dogs. Isolated bacterial DNA from canine faecal samples obtained during the feeding of one of the three specialized diets was used for %G + C profiling and sequencing of valid fractions (referred to herein as fractions 5, 10 and 14) from the %G + C profile. Fractions 5, 10 and 14 corresponded to %G + C ranges of 27–32, 46.5-51.5 and 62–67, respectively.

%G + C profiling of DNA samples - The DC diet faecal samples displayed a significantly higher abundance of microbes with %G + C between 33 and 41 than the HP diet samples (p = 0.03) (Figure 1). Moreover, samples from the HP diet contained a peak at %G + C between 46 and 50, which was completely lacking from the DC diet samples (p = 0.02, Figure 1). A low %G + C peak was present in the HP diet and lacking in DC diet samples in the %G + C range of 25–29 (p = 0.05, Figure 1).

No major differences between the DC and HC diet samples were observed. However, the profiles were significantly different at %G + C 39–40, 57–58 and 65–66 (p < 0.05, Figure 1).

To illustrate the difference between the HP and HC diet, we also carried out a direct comparison between these two diets. As expected, considerable differences were observed in the average %G + C profiles (Figure 1). The HC diet resulted in a significant higher abundance of microbes with %G + C between 33 and 40 than the HP diet (p < 0.01, Figure 1). The peak present in the HP diet samples at %G + C between 46 and 50 was completely absent in the HC diet samples (p = 0.02, Figure 1). The HC diet samples favoured bacteria with %G + C higher than 60 (p < 0.05, Figure 1).

Phylogenetic analysis of fraction 5 sequences

Ninety-six sequences per sample were obtained from %G + C fraction 5. Clostridiales was the most representative bacterial order in DNA obtained from faecal samples of dogs fed DC and HC diets (78 % and 85 % of clones, respectively). Overall, the proportion of Clostridiales sequences in the HP diet sample was much lower (37 %, Figure 2). This difference was statistically significant with order-level library comparison analysis between the HP and DC diet samples as well as between the HC and DC diet samples (p < 0.01).

At family-level classification, the Clostridiales clones from DC diet samples were distributed into two main bacterial families, namely Lachnospiraceae (72 % of Clostridiales clones) and Peptostreptococcaceae (24 % of Clostridiales clones), whereas in HC diet samples, the vast majority of clones (99 % of Clostridiales clones) were affiliated with Peptostreptococcaceae. A distribution similar to that of the HC diet sample was observed in the HP diet sample, where 94 % of the Clostridiales clones were also classified into the family Peptostreptococcaceae (Figure 2).

In the HP diet samples, Fusobacteriales was the most prevalent order (62 % of clones, Figure 2). Interestingly, only one representative of Fusobacteriales was found in HC diet samples, whilst no members of this order were discovered in DC diet samples. The sequences belonging to the order Fusobacteriales showed the closest similarity with Fusobacterium varium and Fusobacterium mortiferum.

In addition to the members of Clostridiales, 14 % of the sequenced clones in the HC diet sample were affiliated with the order Erysipelotrichales (Figure 2), which is classified into the same phylum as Clostridiales. More specifically, these sequences belong to Clostridial Cluster XVIII. By contrast, only one sequence that classified into the order Erysipelotrichales was discovered in HP diet samples (Figure 2).

In the DC diet, the sequence diversity was generally higher than in the HC and HP diets, as the sequences were classified into five different orders, two of which, Lactobacillales and Bacteroidales, were completely absent from the samples from the HC and HP diets (Figure 2). At the family level, all Lactobacillales sequences associated the Streptococcaceae and Bacteroidales sequences with Prevotellaceae.

Phylogenetic analysis of fraction 10 sequences

Ninety-six sequences per sample were obtained from %G + C fraction 10. Sequences that affiliated with Clostridiales dominated in the DNA obtained from canine faecal samples during HP and HC diet phases (93 % and 90 % of all clones, respectively). At family-level classification, the Clostridiales sequences in the HP diet sample were allocated mainly to Lachnospiraceae (57 % of Clostridiales clones), Peptostreptococcaceae (35 % of Clostridiales clones) and Ruminococcaceae (6 % of Clostridiales clones). On the other hand, the majority of Clostridiales sequences in the HC diet sample (96 % of Clostridiales clones) were classified into the family Peptostreptococcaceae (Figure 3). A total of 85.5 % of these sequences were affiliated with Clostridium hiranonis, a member of Clostridial Cluster XI.

The abundance of Clostridiales sequences (44 %; mainly members of the family Lachnospiraceae) in the DC diet sample was smaller than in the HP and HC diet samples (Figure 3, p < 0.01). The most dominant group of bacteria in canine faecal samples with the DC diet in fraction 10 was Lactobacillales, more specifically Streptococcaceae, which comprised 54 % of the clones sequenced. Only two Streptococcaceae clones were found in HP diet samples, and no lactic acid bacteria were detected in HC diet samples (Figure 3).

Phylogenetic analysis of fraction 14 sequences

Forty-eight sequences per sample were obtained from %G + C fraction 14. All 141 clones from the high %G + C fraction 14 that yielded a sequence of adequate quality were affiliated with the order Coriobacteriales (Figure 4). At the genus level, the majority of clones (n = 138) appeared to belong to Collinsella spp., with the remaining clones representing Slackia spp. and Eggerthella spp. Surprisingly, members of the order Bifidobacteriales were not discovered in any of the three samples (Figure 4).

The influence of dietary animal-derived proteins and carbohydrates on canine intestinal microbiota was investigated. The %G + C profiles, as well as order-level sequence distribution in fraction 5, between the DC and HC diet samples did not differ considerably, most likely indicating that the modulatory influence of the HC diet on canine fecal microbiota is smaller than that of the HP diet when compared with the DC diet phase of the trial. This result is not surprising since both DC and HC diets consisted mainly of carbohydrate-rich components.

The amount and type of fermentable carbohydrates reaching the colon are primary factors influencing the abundance and variety of the resident bacterial population. The bacteria that can most rapidly degrade and use the digesta will proliferate beyond the others [23]. Corn starch, which was also included in our diets, has a high small intestinal digestibility and is therefore not expected to reach the large intestine in high amounts [24]. However, the DGGE band patterns obtained in a previous study indicated that this carbohydrate affected the composition of faecal bacteria in rats [9]. Therefore, the passage of corn starch into the large intestine might have been one of the reasons for sequence differences between DC and HP diet samples as well as between HC and HP diet samples.

The sequence diversity in the DC diet sample was generally higher than in the HC and HP diet samples, as the sequences were classified into five different orders, two of which, Lactobacillales and Bacteroidales, were completely absent in faecal samples from the HC and HP diets. The higher diversity was most likely due to the different ingredients of the DC diet sample, and the combination of corn and processing conditions might have resulted in a more versatile spectrum of fermentable substrates for various bacterial types [24].

In fraction 5 of the HP diet sample, the most abundant sequences belonged to the order Fusobacteriales and showed close similarity with the species F. varium and F. mortiferum. F. varium F. necrophorum F. nucleatum and F. equinum have been found to play roles in the pathogenesis of colonic, oropharyngeal, gingival, periodontal and other inflammatory processes, such as abscesses, pneumonia and sinusitis [25–27]. Given that the HP diet led to diarrhoea for all dogs in our study (data not shown), it could be hypothesized that species from the order Fusobacteriales could have caused the loss in faecal consistency, together with the high collagen concentration in the HP diet. The Greaves-meal diet, having a high digestibility, is known to soften the faeces, increase Clostridium perfringens levels and decrease bifidobacteria in dogs [28, 29]. To our knowledge, F. varium and F. mortiferum have not been previously detected in canine faecal samples. Further characterization of isolates of these species should clarify whether they are commensals or opportunistic pathogens, or both, which is the case in the human intestine [27].

The results obtained from fraction 10 again indicate that the increased sequence diversity with the DC diet relative to the HP and HC diets was most likely due to the more versatile nutrient composition. In the HC diet sample, the most abundant sequences belonged to the order Clostridiales, showing the closest similarity with Clostridium hiranonis, which has been discovered previously in the canine GI tract and is considered to belong to the normal canine intestinal microbiota [12].

In fraction 14, all sequences in faecal samples of all dietary groups belonged to the order Coriobacteriales, suggesting that members of Coriobacteriaceae may be indicators of a healthy GI microbiota. For instance, in humans a high abundance of Collinsella aerofaciens has been associated with a lowered risk of colon cancer and inflammatory bowel disease [30, 31]. To our knowledge, the presence of bacteria belonging to the order Coriobacteriales in canine faecal samples has been reported only in a recent 16 S rRNA gene sequencing study [6]. The order Coriobacteriales within the phylum Actinobacteria was found to be more abundant than previously estimated with conventional sequencing studies also in human faecal samples [22]. This is most likely due to the sequencing studies having been carried out without %G + C fractioning. It is evident that fractionating the total faecal DNA preparations minimizes PCR and cloning-derived bias, which is common in multi-template sequencing studies. In other words, fractionating facilitates the amplification and subsequent cloning of species with high G + C contents from diverse microbial communities [19, 20].

No bifidobacteria were found, consistent with an earlier study [11]. However, contradictory data also exist, as many studies have found bifidobacteria in dogs [12, 32]. Possibly, bifidobacteria were not part of the predominant intestinal microbiota of the Beagle dogs participating in our study. Another potential explanation for this unanticipated result may be that the universal 16 S rRNA gene-targeted primer pair contained mismatches to many bifidobacterial species, which could have led to significant underestimation of bacteria belonging to this genus.

Clostridiales and Coriobacteriales were the most prevalent bacterial orders in the faecal samples of all dietary groups. Suchodolski and coworkers [13], by contrast, reported that Fusobacteriales and Bacteroidales were the most representative orders in the canine colon. It is noteworthy, however, that we analysed only three %G + C fractions, which showed the most pronounced alterations between the dietary groups. Our aim was not to obtain an overall picture of the canine faecal microbiota, but to elucidate the diet-derived effects on the microbial community structure in the lower intestine.

Significant dietary effects on canine intestinal microbiota were detected. The DC diet sample showed a high abundance of representatives of the orders Clostridiales, Lactobacillales and Coriobacteriales and the presence of representatives of the order Bacteroidales. Sequence diversity was higher with the DC diet sample, as representatives of the orders Lactobacillales and Bacteroidales were not detected in the HP and HC diet samples. During feeding of the HC and HP diets the representatives of Clostridial Cluster XIVa were suppressed in canine faecal samples. The HC diet favoured representatives of Clostridial Cluster XVIII. The HP diet favoured representatives of the order Fusobacteriales, which could play a role in induction of diarrhoea together with the lower carbohydrate concentration entering the large intestine.

IH, WMV, JH, JZ and TS conceived the study design; IH performed sample collection; TR and JA performed bacterial DNA extraction, %G + C profiling and sequencing analysis; AK and SA performed statistical analysis; all authors contributed to the writing of the manuscript.