Faecal microbiota in lean and obese dogs

Abstract

Previous work has shown obesity to be associated with changes in intestinal microbiota. While obesity is common in dogs, limited information is available about the role of the intestinal microbiota. The aim of this study was to investigate whether alterations in the intestinal microbiota may be associated with canine obesity. Using 16S rRNA gene pyrosequencing and quantitative real-time PCR, we evaluated the composition of the faecal microbiota in 22 lean and 21 obese pet dogs, as well as in five research dogs fed ad libitum and four research dogs serving as lean controls. Firmicutes, Fusobacteria and Actinobacteria were the predominant bacterial phyla. The phylum Actinobacteria and the genus Roseburia were significantly more abundant in the obese pet dogs. The order Clostridiales significantly increased under ad libitum feeding in the research dogs. Canine intestinal microbiota is highly diverse and shows considerable interindividual variation. In the pet dogs, influence on the intestinal microbiota besides body condition, like age, breed, diet or lifestyle, might have masked the effect of obesity. The study population of research dogs was small, and further work is required before the role of the intestinal microbiota in canine obesity is clarified.

canine, microbiota, obesity, qPCR, pyrosequencing

Introduction

Obesity constitutes a major health problem in developed countries, in humans as well as in pets. Recent surveys revealed a prevalence of canine obesity of 39% in France (Colliard et al., 2006), about 50% in the UK (Holmes et al., 2007) and 34% in Austria (Handl et al., 2009). It is well documented that obesity is a risk factor for many disorders in dogs, such as orthopaedic disease (Kealy et al., 2000), urinary tract disease (Lekcharoensuk et al., 2000; Lund et al., 2006), cardiovascular disease (Edney & Smith, 1986; Mizelle et al., 1994), respiratory dysfunction (Bach et al., 2007) and hormonal disturbances (Martin et al., 2006). Taken together, obesity reduces longevity (Kealy et al., 2002) and has to be regarded as an animal welfare issue (Houpt et al., 2007). Although the treatment of obesity, reducing energy intake and enhancing physical activity are simple in theory, weight loss programs often fail; therefore, a better knowledge about the underlying pathogenetic mechanisms is urgently needed.

Studies in animal models have highlighted the importance of the intestinal microbiota in energy harvest from the diet. Conventionalization of germ-free mice with normal mice microbiota drastically increased body fat content, despite reduced food intake (Bäckhed et al., 2004). Obesity on the other hand is accompanied by significant changes in the caecal microbiota of mice, characterized by a reduction in the bacterial phylum Bacteroidetes and a significantly greater proportion of the phylum Firmicutes. Furthermore, an increased concentration of acetate and butyrate was found in the caeca of obese mice (Turnbaugh et al., 2006). These changes were reversible through diet-induced weight loss (Ley et al., 2005; Turnbaugh et al., 2006, 2008). Several recent studies also reported differences in the faecal microbiota between lean and obese humans. Brignardello et al. (2010) found a decrease in Clostridium-cluster XIVa in obese adults, while Faecalibacterium prausnitzii was significantly increased in obese children (Balamurugan et al., 2010). Clostridium leptum group (Schwiertz et al., 2009) and Bifidobacteria were significantly decreased in obese subjects, while Lactobacilli (Armougom et al., 2009; Brignardello et al., 2010), Bacteroides (Schwiertz et al., 2009) and Prevotellaceae (Zhang et al., 2009) were enriched in obese men.

All these groups are involved in the fermentation of indigestible carbohydrates. It has been shown, both in genetically manipulated and diet-induced obese mice, that the caecal microbiome of obese animals is enriched for environmental gene tags encoding for enzymes involved in polysaccharide breakdown and for metabolism of simple carbohydrates (Turnbaugh et al., 2006, 2008). It is, therefore, hypothesized that the contribution of the intestinal microbiome to the development of obesity lies in the enhanced fermentation of carbohydrates and that the coexistence of H₂-utilizing Archaea with H₂-producing bacteria facilitates carbohydrate fermentation and causes increased production of short-chain fatty acids, which can be used as energy source by the host (Zhang et al., 2009). Manipulation of the intestinal microbiota via probiotics/prebiotics (Cani et al., 2007; Everard et al., 2011) or even antibiotics (Murphy et al., 2012) to alleviate the negative health affects, like insulin resistance and inflammation, could potentially provide an easily applicable complementary therapy for obesity.

To the best of our knowledge, no studies have been performed evaluating the role of the intestinal microbiota in canine obesity. The aims of the current study were, therefore, to compare the composition of the faecal microbiota of lean and obese pet dogs and further to investigate whether ad libitum feeding causes a shift in the faecal microbial composition in research colony dogs.

Materials and methods

Animals

Faecal samples were collected from 22 lean [median body condition score (BCS), 5/9; range, 4–6] and 21 obese (median BCS, 9/9; range, 7–9) privately owned dogs. Dogs were adult (median age, 5.8 and 6.2 years, respectively), were fed various commercially available diets and had not undergone a diet change for at least 3 weeks prior to sample collection. BCS was estimated by a veterinarian based on a 9-point scale (Laflamme, 1997). Animals were free of any clinical signs indicating gastrointestinal disease and did not receive medications that are expected to alter the gut microbiota (i.e. antibiotics). One faecal sample per dog was collected by the owner after spontaneous defecation and immediately frozen at −20 °C, transported to the laboratory on ice packs and stored at 80 °C until analysis. More detailed information on the pet dogs is provided in Supporting Information, Table S1.

Additionally, nine healthy adult research Beagle dogs (intact females; mean (± SD) age, 4.1 ± 0.6 years; body weight, 8.5 ± 0.4 kg) were enrolled. Dogs were fed a high-fat commercial dry diet (Pro Plan^® Performance Formula; Nestle Purina Pet Care, St. Louis, MO). After an adaption period of 4 weeks, five of the dogs were switched to ad libitum feeding over 6 months to a median BCS of 8/9 (range, 7/9–9/9) as described previously (Grant et al., 2011). The remaining four Beagle dogs were fed the same diet restrictively to maintain baseline body weight and served as lean controls (median BCS, 4/9; range, 4/9–5/9). Faecal samples were collected at baseline (t0), after 1 month (t1), 2 months (t2), 3 months (t3) and after 6 months (t6). At every time point, one sample per dog was collected immediately after spontaneous defecation and frozen without further treatment at −80 °C.

DNA extraction

Genomic DNA was extracted from each faecal sample individually using a bead beating method followed by phenol/chloroform/isoamyl alcohol extraction as described previously (Handl et al., 2011).

Pyrosequencing

Bacterial tag-encoded FLX-Titanium amplicon pyrosequencing (bTEFAP) and data processing were performed as described previously (Handl et al.,2011). The primers used extended from 27F numbered in relation to Escherichia coli 16S ribosome gene (forward 28F: GAGTTTGATCNTGGCTCAG, reverse 519R: GTNTTACNGCGGCKGCTG).

For economical reasons, in the research dogs, pyrosequencing was only performed from the baseline, 3- and 6-month samples.

bTEFAP and fTEFAP sequence processing pipeline

Raw data from bTEFAP were screened and trimmed based on quality scores (nominal Phred20) and were binned into individual sample collections. After sequencing, individual collections of sequences were depleted of chimeras using B2C2 (Gontcharova et al., 2010). The resulting files of short reads (< 350 bp) were depleted, thereby ensuring that all sequences evaluated contained sufficient discriminating data from the V1 to V3 hypervariable regions. Tentative bacterial phylotypes were identified using blastn in comparison with a curated high-quality 16S rRNA gene database derived from the National Center for Biotechnology Information (NCBI) as described previously (Dowd et al., 2008). The compiled data were used to determine the relative percentages of bacteria for each individual sample. Data were also compiled at each individual taxonomic level according to the NCBI taxonomy criteria as described previously (Dowd et al., 2008). Database sequences were characterized as high quality based on similar criteria utilized by RDP ver 9. Using a.NET and C# analysis pipeline, the resulting blastn outputs were compiled, validated using taxonomic distance methods and data reduction analysis performed as described previously. Sequences with identity scores to known or well-characterized 16S rRNA gene sequences > 97% identity (< 3% divergence) were resolved at the species level, between 95% and 97% at the genus level, between 90% and 95% at the family level, between 80% and 90% at the order level, between 80% and 85% at the class and between 77% and 80% at phylum level. After resolving based on these parameters, the percentage of each bacterial ID was individually analysed for each sample providing relative abundance information within and among the individual samples based on relative numbers of reads within each. Evaluations presented at each taxonomic level, including percentage compilations, represent all sequences resolved to their primary identification or their closest relative.

Quantitative real-time PCR

Quantitative real-time PCR (qPCR) was performed on a C1000 Thermal Cycler with the CFX96™ Real Time System, associated with the CFX Manager™ Software version 1.6 (all from BIO-RAD Laboratories, Hercules, CA). All qPCRs were performed in a volume of 10 μL using low profile clear 96-well Multiplate^® PCR Plates™ with Microseal^® ‘B’-Film PCR Sealers™ (both from BIO-RAD). Primers were purchased at the Gene Technologies Laboratory, Institute of Developmental and Molecular Biology, Texas A&M University. Each reaction mixture consisted of 5 μL of IQ™ SYBR Green Supermix (BIO-RAD), giving a final concentration of 50 mM KCl, 20 mM Tris–HCl, pH 8.4, 0.3 mM of each dNTP, iTaq hot-start polymerase (50 units mL⁻¹), 3 mM MgCl₂ and 10 nM fluorescein and further 0.2 μM of each primer and 2 μL of DNA template or water (no template control). 2.8 μL of PCR water was added to adjust to the final volume of 10 μL. The reaction conditions were 95 °C for 3 min, 40 cycles of 95 °C for 10 s, annealing temperature and time given in Table 1 and 72 °C for 10 s. A melting curve analysis was performed after amplification. All samples were analysed in duplicate.

Sequences of primers and protocols used for qPCR

Assay	Primers 5′–3′	Annealing T and time	References
Universal bacteria
F341	CCTACGGGAGGCAGCAG	57 °C for 10 s	Cunliff & Kertesz (2006)
518R	ATTACCGCGGCTGCTGG
Phylum Firmicutes
Lgc353	GCAGTAGGGAATCTTCCG	58 °C for 10 s	Fierer et al. (2005)
Eub518	ATTACCGCGGCTGCTGG
Phylum Bacteroidetes
Cfb319	GTACTGAGACACGGACCA	60 °C for 30 s	Fierer et al. (2005)
Eub518	ATTACCGCGGCTGCTGG
Prevotella group
g-Prevo	FCACRGTAAACGATGGATGCC	55 °C for 25 s	Matsuki et al. (2002)
g-Prevo	RGGTCGGGTTGCAGACC
Lactobacillus group
LacRT F	AGCAGTAGGGAATCTTCCA	58 °C for 15 s	Walter et al. (2001)
LacRT R	CACCGCTACACATGGAG
Bifidobacterium group
BifRT F	TCGCGTC(C/T)GGTGTGAAAG	60 °C for 15 s	Rintillä et al. (2004)
Bifrt-R	CCACATCCAGC(A/G)TCCAC
Clostridium-cluster I
CperRT F	ATGCAAGTCGAGCGA(G/T)G	55 °C for 15 s	Malinen et al. (2005)
CperRT R	TATGCGGTATTAATCT(C/T)CCTTT
Clostridium-cluster IV
c lept F	GCACAAGCAGTGGAGT	58 °C for 20 s	Matsuki et al. (2004)
c lept R	CTTCCTCCGTTTTGTCAA
Clostridium-cluster XI
C Clust XI F	ACGCTACTTGAGGAGGA	55 °C for 30 s	Song et al. (2004)
C Clust XI R	GAGCCG AGCCTTTCACT
Clostridium-cluster XIVa
CcocRT F	CGGTACCTGACTAAGAAG	53 °C for 15 s	Malinen et al. (2005)
CCocRT R	AGTTT(C/T)ATTCTTGCGAAC
Genus Catenibacterium
Cat F	AACGCCGCGTGAGCGAAGAA	58 °C for 10 s	This study
Cat R	TTCGGTACCTTTGGCCCGCC

Statistical analysis

To account for unequal sequencing depth, a random subset of 2400 sequences per sample was analysed. To assess the diversity of the faecal microbiota in the pet dogs, the Shannon–Weaver diversity index and evenness were calculated from the number of OTUs assigned at the genus and the species level as described in Steele et al. (2005). Percentages of bacterial sequences and numbers of bTEFAP reads are given as mean ± standard deviation (SD). Results of qPCR are presented as the ratio of the copy number of the bacterial group to the copy number obtained in the universal bacterial assay. The relative abundances of the bacterial groups on various phylogenetic levels and the diversity indices were compared between pet dog samples using various statistical procedures (GraphPad Prism 5.0, San Diego, CA). Data were tested for normal distribution using the D'Agostino and Pearson omnibus normality test. An unpaired t-test was used to compare normally distributed data; a Mann–Whitney U-test was used for nonparametric data. The resulting P-values were corrected for multiple comparisons using Benjamini and Hochberg's false discovery rate (Benjamini & Hochberg, 1995). A Fisher's exact test was used to compare the number of animals harbouring a specific bacterial group. For the research dogs, a general linear model for repeated measures was performed using the feeding regimen (restricted or ad libitum) as within-subjects factor and the individual dog as covariate (paws Statistics 17.0; IBM Software). To identify clustering of samples along the first three axes of maximal variance, principal component analysis plots (PCoA) were constructed based on the unweighted UniFrac distance metric.

Results

Pet dogs

To reveal whether there were differences in the composition of the intestinal microbiota between lean pet dogs and dogs, which had been obese for a longer period of time, we analysed the microbial composition of the faeces using high-throughput pyrosequencing and qPCR. A total of 250 982 bTEFAP reads were analysed across all faecal samples with a mean ± SD of 5837 ± 2122 reads per dog. A strong dominance of the phylum Firmicutes was noted in all dogs (mean abundance > 90%). Other phyla frequently identified were Actinobacteria, Fusobacteria, Proteobacteria and Bacteroidetes. The mean abundance of Actinobacteria was, however, significantly greater (P = 0.044) in the obese dogs (3.1 ± 3.8%) compared with the lean dogs (1.7 ± 2.9%). OTUs belonging to the phyla Chloroflexi, Thermotogae, Verrucomicrobia and Candidate division OP10 were detected only in individual lean dogs, but in none of the obese animals.

The composition of the faecal microbial population of the pet dogs on the bacterial order level is depicted in Fig. 1. The orders Clostridiales, Lactobacillales and Erysipelotrichales (all within the phylum Firmicutes) and Fusobacteriales (phylum Fusobacteria) were the predominant groups. While the abundance of Clostridiales was less in the lean animals (67.8 ± 35.0% in lean dogs vs. 87.6 ± 10.0% in obese dogs), and the abundance of Bacillales (15.4 ± 27.5% vs. 3.2 ± 6.1%) and Fusobacteriales (6.4 ± 19.1% vs. 0.3 ± 1.2%) was less in the obese group; these differences did not reach statistical significance.

Composition of the fecal microbial population of lean (N = 22; a) and obese (N = 21; b) pet dogs on order level evaluated with 454 pyrosequencing.

At genus level (5% dissimilarity, McKenna et al., 2008), 147 OTUs were identified in the study population. Bacterial diversity indices, the number of observed OTUs per animal (mean ± standard deviation) and the number of total observed OTUs at the genus level in lean and obese pet dog, respectively, are listed in Table 2. No significant differences in the Shannon diversity index, the evenness and the number of OTUs were observed between lean and obese pet dogs.

Diversity indices, number of observed OTUs per animal (mean ± standard deviation) and number of total observed OTUs in lean and obese pet dogs using massive parallel pyrosequencing

	Lean pet dogs	Obese pet dogs	P
Genus level (5% dissimilarity)
Shannon diversity index	1.14 ± 0.38	1.26 ± 0.40	0.328
Evenness	0.36 ± 0.11	0.39 ± 0.10	0.300
Mean observed OTUs per animal	24 ± 7.2	25 ± 10.0	0.733
Total observed OTUs	123	98	–

All bacterial genera identified in the pet dogs, the percentages of pyrosequencing tags affiliated with each genus and the numbers of lean and obese dogs harbouring each genus are listed in Table S2.

The most abundant genera were Clostridia, Ruminococcus and Blautia from the order Clostridia, and Streptococcus and Enterococcus from the order Lactobacillales. The mean abundance of many genera showed considerable interindividual differences in leans as well as obese dogs. After correction for multiple comparisons, a significant difference was only detected in Roseburia (order Clostridiales, P = 0.012), which was more abundant in obese dogs (0.66 ± 0.78% vs. 0.21 ± 0.24%). PCoA plot based on the unweighted UniFrac distance metric of the faecal microbiota of the lean and obese pet dogs is depicted in Fig. 2. No clear separation was seen.

PCoA plot based on the unweighted UniFrac distance metric of the fecal microbiota of lean (red squares) and obese (blue squares) pet dogs using 16S rRNA gene pyrosequencing.

The results of the qPCR performed on the pet dog samples are presented in Table 3. The proportions of the investigated bacterial groups were not significantly different between lean and obese dogs.

Bacterial groups in faecal samples from lean (N = 22) and obese (N = 21) pet dogs measured by qPCR. Results are presented as proportion of total bacteria copies

	Phylum Firmicutes	Phylum Bacteroidetes	Lactic acid–producing bacteria	Bifidobacterium group	Prevotella group	Clostridium-cluster I	Clostridium-cluster IV	Clostridium-cluster XI	Clostridium-cluster XIVa
Lean	0.40 ± 0.25	0.69 ± 0.12	0.32 ± 0.23	0.09 ± 0.24	0.15 ± 0.18	0.41 ± 0.25	0.36 ± 0.21	0.51 ± 0.19	0.57 ± 0.26
Obese	0.40 ± 0.25	0.70 ± 0.08	0.34 ± 0.19	0.15 ± 0.26	0.20 ± 0.24	0.28 ± 0.25	0.37 ± 0.27	0.52 ± 0.13	0.46 ± 0.32
P	0.927	0.942	0.643	0.303	0.629	0.078	0.574	0.865	0.366

Research dogs

To evaluate whether ad libitum feeding caused a change in faecal bacterial population in dogs, we sampled five Beagle dogs fed ad libitum and four animals fed a restricted amount of the same diet to maintain optimal body condition over a period of 6 months. A total of 103 116 bTEFAP reads were analysed across all faecal samples from the nine Beagle dogs with a mean of 3819 ± 1319 reads per sample.

Eight different phyla were identified, with Firmicutes being predominant in all dogs (> 94%). Further phyla found at low abundance in the majority of dogs were Actinobacteria, Proteobacteria and Bacteroidetes. The phyla Fusobacteria, Tenericutes and Spirochaetes were present only in a subset of dogs. The bacterial orders showing the greatest abundance are listed in Table 4. Considerable fluctuations in the mean abundance of bacterial orders were seen between individual dogs at the same sampling time point as well as between sampling time points in the same dog. A statistically significant influence of the feeding regimen (restricted vs. ad libitum) was detected on the relative abundance of the phylum Firmicutes (P = 0.024), which increased during the course of the study in the dogs fed restrictively, and the phylum Bacteroidetes (P = 0.034), which decreased in the dogs fed restrictively. A statistically significant influence of the feeding regimen was further detected on the relative abundance of the classes Gammaproteobacteria (P = 0.016) and Alphaproteobacteria (P = 0.028; both decreased in the dogs fed restrictively). The relative abundance of Clostridia was higher at t3 as at t0 and was lower again at t6; but these changes were more pronounced in the dogs fed ad libitum (P = 0.025). The order Coriobacteriales decreased continuously in the dogs fed restrictively, while it decreased first, but rose again at t6 in the dogs fed ad libitum (P = 0.043). The order Pseudomonadales decreased in the animals fed restrictively, while it increased in the dogs fed ad libitum from t0 until t3 (P = 0.016) and was undetectable in both groups at t6. The individual dog as covariate had a significant influence on the abundance of Firmicutes (P = 0.004), Bacteroidetes (P = 0.016), Gammaproteobacteria (P = 0.01), Alphaproteobacteria (P = 0.004), Coriobacteriales (P = 0.009) and Pseudomonadales (P = 0.01). Only for the order Clostridia, the influence of feeding regimen was significant, while the influence of the individual dog was not.

Major bacterial orders (mean abundance ± SD) identified in a colony of healthy, adult research beagles using 454 pyrosequencing

Bacterial order	Group	t0	t3	t6	P
Clostridiales	A	56.5 ± 26.6	89.7 ± 7.4	78.8 ± 10.6	0.025
	B	64.3 ± 22.7	90.3 ± 7.6	73.3 ± 21.2
Erysipelotrichales	A	29.3 ± 32.6	8.4 ± 5.7	20.5 ± 10.5	0.239
	B	22.8 ± 18.1	5.9 ± 5.7	14.1 ± 8.6
Lactobacillales	A	13.2 ± 19.1	0.8 ± 0.6	0.1 ± 0.1	0.490
	B	6.9 ± 5.9	1.3 ± 1.7	11.9 ± 22.5
Coriobacteriales	A	0.5 ± 0.3	0.8 ± 0.6	0.5 ± 0.1	0.043
	B	2.3 ± 3.3	1.6 ± 1.3	0.3 ± 0.3
Pseudomonadales	A	0.3 ± 0.5	0.0 ± 0.0	0.0 ± 0.0	0.016
	B	2.2 ± 4.7	0.7 ± 1.5	0.0 ± 0.0

A = dogs fed restrictively (N = 4) and B = dogs fed ad libitum (N = 5). One faecal sample per dogs was taken at baseline (t0) and after 3 (t3) and 6 months (t6).

A total of 89 OTUs were identified at genus level (5% dissimilarity, McKenna et al., 2008) across all samples. After the study period of 6 months, 41 OTUs were detected in the lean dogs fed restrictively and 49 OTUs in the obese ad libitum fed dogs. Due to the small number of dogs in each group, diversity indices were not calculated.

For all bacterial genera identified in the research dogs, the percentages of pyrosequencing tags affiliated with each genus and the numbers of lean and obese dogs harbouring each genus are listed in Table S3.

The genera with the greatest abundance were Clostridium, Ruminococcus and Blautia (all in the order Clostridiales), and Turicibacter (order Erysipelotrichales). Again, high variations in the relative abundance of bacterial genera were noted between the dogs at each time point as well as between the three sampling time points in each individual dog. The results of the statistical evaluation of the influence of the feeding regimen using a general linear model for repeated measures are also given in Table S3. The feeding regimen had a significant influence on the relative abundance of Clostridium (P = 0.025), which increased during the course of the study in all dogs, but more pronounced with ad libitum feeding. The relative abundance of Bacteroides (order Bacteroidales), Serratia (order Enterobacteriales), Enterobacter (also order Enterobacteriales), Afipia (order Rhizobiales) and Pseudomonas (order Pseudomonadales) was also significantly influenced by the feeding regimen (Table S3); however, in those genera, the influence of the individual dog as covariate was also significant (results not shown). PCoA plot based on the unweighted UniFrac distance metric of the faecal microbiota of the research colony is shown in Fig. 3. No obvious separation between time points was found.

PCoA plot based on the unweighted UniFrac distance metric of the fecal microbiota of research colony Beagles (N = 5) analysed using 16S rRNA gene pyrosequencing at baseline (red squares), after 3 months (green squares) and 6 months (blue squares) of ad libitum feeding with a commercial dry diet.

Table 5 displays the results of the qPCR analysis from the research dogs. Corresponding to the pyrosequencing results, fluctuations were observed between time points. No significant influence of the feeding regimen (restricted vs. ad libitum) was detected.

Bacterial groups in faecal samples from research colony dogs fed restrictively (A, N = 4) and ad libitum (B, N = 5) measured by qPCR. One faecal sample per dogs was taken at baseline and after 1, 2, 3 and 6 months. Results are presented as proportion of total bacteria copies

Bacterial group	Group	t0	t1	t2	t3	t6	P
Phylum Firmicutes	A	0.65 ± 0.12	0.48 ± 0.21	0.49 ± 0.15	0.55 ± 0.10	0.56 ± 0.10
	B	0.35 ± 0.21	0.31 ± 0.22	0.37 ± 0.11	0.43 ± 0.07	0.46 ± 0.18	0.796
Phylum Bacteroidetes	A	0.70 ± 0.05	0.69 ± 0.06	0.64 ± 0.05	0.72 ± 0.02	0.75 ± 0.02
	B	0.65 ± 0.07	0.65 ± 0.08	0.66 ± 0.05	0.72 ± 0.03	0.63 ± 0.09	0.092
Lactic acid–producing bacteria	A	0.87 ± 0.11	0.46 ± 0.12	0.49 ± 0.16	0.59 ± 0.06	0.56 ± 0.06
	B	0.73 ± 0.18	0.53 ± 0.09	0.58 ± 0.05	0.70 ± 0.10	0.70 ± 0.16	0.752
Bifidobacterium group	A	0.24 ± 0.30	0.07 ± 0.13	0	0.19 ± 0.23	0.17 ± 0.23
	B	0	0	0.04 ± 0.08	0	0.08 ± 0.17	0.719
Prevotella group	A	0.25 ± 0.20	0	0	0.17 ± 0.26	0.22 ± 0.26
	B	0	0	0	0	0	0.161
Clostridium-cluster I	A	0.43 ± 0.30	0.51 ± 0.13	0.31 ± 0.21	0.30 ± 0.06	0.56 ± 0.06
	B	0.30 ± 0.21	0.52 ± 0.13	0.36 ± 0.21	0.56 ± 0.13	0.52 ± 0.06	0.615
Clostridium-cluster IV	A	0.51 ± 0.17	0.38 ± 0.18	0.36 ± 0.18	0.51 ± 0.20	0.47 ± 0.20
	B	0.26 ± 0.25	0.42 ± 0.16	0.25 ± 0.21	0.33 ± 0.20	0.34 ± 0.22	0.121
Clostridium-cluster XI	A	0.55 ± 0.10	0.64 ± 0.04	0.58 ± 0.05	0.58 ± 0.08	0.54 ± 0.08
	B	0.53 ± 0.14	0.59 ± 0.04	0.59 ± 0.06	0.67 ± 0.04	0.58 ± 0.08	0.978
Clostridium-cluster XIVa	A	0.50 ± 0.24	0.51 ± 0.20	0.44 ± 0.21	0.57 ± 0.29	0.60 ± 0.29
	B	0.34 ± 0.32	0.57 ± 0.21	0.57 ± 0.15	0.54 ± 0.25	0.60 ± 0.26	0.372

Discussion

The current study is the first to investigate possible differences in the composition of the faecal microbiota between lean and obese dogs. We included pet dogs that had been obese for an indefinite period of time, as well as dogs from a research colony, which were fed ad libitum for 6 months. Using both qPCR and pyrosequencing techniques, we did not observe clear differences in microbial populations of lean and obese dogs as previously reported in animal models (Turnbaugh et al., 2008, 2009) or human patients (Ley et al., 2006; Turnbaugh et al., 2009).

However, also the studies on the role of the intestinal microbiota in humans and model animals are far from conclusive. Ley et al. (2005) discovered a marked change in the microbial composition in the caecum of genetically obese mice compared with wild-type mice, characterized by a 50% reduction in the abundance of the phylum Bacteroidetes and a proportional increase in Firmicutes in the obese mice. The same research group reported similar results in humans (Ley et al., 2006; Turnbaugh et al., 2009). While some studies have obtained similar results as mentioned above (Armougom et al., 2009), others have either noted the opposite (Schwiertz et al., 2009) or have found no differences in the proportions of Bacteroidetes and Firmicutes between lean and obese subjects (Duncan et al., 2008; Zhang et al., 2009; Balamurugan et al., 2010). Considering the growing knowledge about the complexity of the intestinal microbiome, it becomes obvious that the analysis of the composition on phylum level cannot reveal compositional changes that affect only certain bacterial subgroups, but which might still have serious consequences for the function of the microbial ecosystem.

Differences in certain bacterial subgroups between lean and obese humans have thus been investigated. The level of F. prausnitzii was significantly greater in obese Indian children (Balamurugan et al., 2010), while Brignardello et al. (2010) detected a decrease in Clostridium-cluster XIVa in obese adults. Clostridium leptum group was significantly decreased in morbidly obese (BMI > 30) subjects, as were Bifidobacteria (Schwiertz et al., 2009), while Lactobacilli (Armougom et al., 2009; Brignardello et al., 2010), Bacteroides (Schwiertz et al., 2009) and Prevotellaceae (Zhang et al., 2009) were enriched in obese men. In the current study, the genus Roseburia within the order Clostridiales was more abundant (P = 0.012) in obese pets dogs. In the research dogs, the order Clostridiales was significantly influenced by the feeding regimen (restricted vs. ad libitum). However, all these studies used different molecular methods (pyrosequencing, qPCR or fluorescence in situ hybridization) making the results difficult to compare. It is, nonetheless, hypothesized that an ‘obese microbiome’ has a considerable influence on energy metabolism and fat storage in the host. Members of the Clostridiaceae and Prevotellaceae possess enzymes that degrade complex indigestible carbohydrates and produce short-chain fatty acids, which can be used as energy source by the host. Lactobacilli and Bifidobacteria are commonly used as probiotics, also for their growth-promoting effects (Khan et al., 2007; Angelakis & Raoult, 2010).

Several investigations indicate that dietary composition might be the factor with the most profound influence on microbial composition. Switching mice to a high-fat diet caused large alterations in microbial composition, in wild-type mice as well as in resistin-like molecule β-knockout mice, which are resistant to dietary obesity (Hildebrandt et al., 2009). The influence of dietary fibre on the structure of the canine microbiota was recently shown using 454 pyrosequencing. Research beagle dogs fed a diet containing 7.5% beet pulp had higher proportions of Firmicutes and Bacteroides/Chlorobi group and lower proportions of Fusobacteria (Middelbos et al.,2010; Swanson et al., 2011). The diet is of course a major contributor to the development of obesity in humans as well as in dogs and, therefore, these two factors are impossible to separate in field studies. In the present study, information on the diet was available in 18 of the 22 lean dogs and all obese dogs (Table S1). A wide variety of diets was used (commercial dry diets, prescription diets, raw diets and human food items). Unfortunately, information on composition and nutrient contents of the diets was not collected, nor was the actual nutrient intake compared to the requirement of the dogs. Besides energy intake, other factors, like management, exercise, concurrent diseases or medications, are relevant for the development of obesity. In the present study, only clinically healthy pet dogs were included, but lifestyle factors were not evaluated. All these variations might explain the strong interindividual differences in the faecal microbiota of the pet dogs, which may overlay effects caused by obesity.

Besides body fat content and diet, several other factors have been discussed that might determine the composition of the intestinal microbiota in mammals, including gender (McKenna et al., 2008), age (Benno et al., 1992; Simpson et al., 2002), breed (Simpson et al., 2002) and kinship (Turnbaugh et al., 2009). While these biases can be easily controlled in experiments on laboratory rodents, they are difficult to exclude in field studies. In contrast to the pet dogs, the research dogs included in the present study were entirely of the same breed, were all intact females, were kept under the same environmental conditions and were fed the same diet, varying only in the amount of food given. We, therefore, expected that environmental influences would be less prominent in the research dogs, and that alterations in the microbial composition caused by obesity would be more distinct. However, although possible influences were excluded as far as possible, and all dogs were allowed to adapt to the diet for 4 weeks, faecal microbial communities showed considerable interindividual variations already at baseline. Therefore, we included ‘individual dog’ as a covariate in the statistical analysis, revealing that the influence of the individual was indeed stronger in many cases than the influence of the feeding regimen.

So far, only limited information is available on the microbial composition of canine faeces. In a previous study by our group investigating the faecal bacterial communities in lean healthy pet dogs also using pyrosequencing (Handl et al., 2011), we found similar proportions of the major bacterial phyla (> 90% Firmicutes, about 2% of Bacteroidetes and Actinobacteria, respectively) as reported in the present investigation. Similar proportions were also identified in a large group of healthy dogs in a recent study (Suchodolski et al., 2012). Middelbos et al. (2010), who also used the pyrosequencing technique and 16S rRNA gene–based data, had reported a lesser abundance of Firmicutes (25%) and a greater prevalence of Bacteroidetes (35%) and Fusobacteria (40%). In addition to potential animal differences, sample preparation may also contribute to these discrepancies. In the present study, as well as in Handl et al. (2011), bead beating was used for DNA extraction, while Middelbos et al. (2010) avoided a bead beating step. If a certain methodology leads to better extraction, identification or quantification results for some bacterial groups, changes in other, less well-represented groups might be overlooked. Further investigations are, therefore, needed to clarify how discrepancies due to variances in the methodology may be minimized, including DNA extraction methods and primer design.

In the present study, we collected only one faecal sample per dog and time point, respectively. However, it seems that the microbial composition of faeces can show considerable day-to-day variations (Garcia-Mazcorro et al., 2012). Therefore, in future studies, it may be useful to pool samples over a collection period of several days when investigating faecal microbial composition and possible influences thereon.

In conclusion, we did not observe major shifts in the faecal microbial composition between lean and obese dogs. As this is the first study on this topic and was carried out under field conditions with various confounding factors, future studies with better controlled environmental factors are recommended to study the role of the intestinal microbiota in obesity development in dogs. Our results indicate changes in the order Clostridiales, which is in agreement with findings in obese humans and rodents. The number of research dogs used in our study was small (five vs. four animals), and experiments with a large number of animals kept under controlled conditions are needed to reveal the influence of obesity on the microbial community more clearly.

Financial disclosure

Both A.J.G.'s and S.L.H.'s posts at the University of Liverpool are financially supported by Royal Canin. The funders had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript.

Supporting Information

Table S1. Pet dogs taking part in the study.

Table S2. Bacterial genera in fecal samples from lean and obese pet dogs identified using 16S rRNA gene pyrosequencing.

Table S3. Bacterial genera in fecal samples from research colony Beagles fed restrictively (N = 4) or ad libitum (N = 5) at baseline (t0), after 3 months (t3), and 6 months (t6) analyzed using 16S rRNA gene pyrosequencing including the results of the statistical evaluation of the influence of the feeding regimen using a general linear model for repeated measures.