Fecal microbiota in client-owned obese dogs changes after weight loss with a high-fiber-high-protein diet

Study Animals, eligibility criteria and ethical considerations

Client-owned dogs with obesity were referred to the Royal Canin Weight Management Clinic, University of Liverpool, UK. All were presented between June 2009 and August 2017, and completed their weight loss regimens between November 2009 and August 2018. To be included in the study, dogs had to be clinically healthy with no signs of gastrointestinal disease, a BCS of ≥6, and no antimicrobial usage in the past month (Igarashi et al., 2014). None of the dogs had antimicrobials throughout their weight loss program. No fecal analyses were performed on the fecal samples and so occult infection with intestinal parasites could not be excluded. However, no dog showed signs consisted with parasitic infection.

Historical data and fecal samples, before and after participation in the weight loss program, had to be available for the analysis. The final number of dogs with obesity that met the inclusion criteria was 25. At time point zero (T0), fecal samples of all 25 dogs were collected. Twenty dogs completed the weight loss program and reached their target weight, whilst five dogs stopped their program early as request of the owners for undeclared reasons. From dogs that completed the weight loss, a fecal sample from the visit in which they reached their target body weight, was collected to include in the analysis as time point two (T2). Dogs that did not complete the weight loss program but had a fecal sample from the first follow-up visit (time point one (T1)) were included in the analysis to account for the effect of the new diet on the fecal microbiota. The study protocol was reviewed and approved by the University of Liverpool Veterinary Research Ethics Committee (Approval reference: RETH000353 and VREC793), the Royal Canin ethical review committee, and the WALTHAM ethical review committee. Owners of dogs with obesity gave informed consent in writing.

Weight loss regimen

Prior to commencing weight reduction, all dogs were considered to be healthy apart from their obesity. All dogs were screened for overall health by performing complete blood count, serum biochemical analysis, serum free thyroxine concentration (by equilibrium dialysis) and urinalysis. Dogs were weighed at admission, and the body condition score (BCS) was estimated using a 9-integer scale by the attending clinician (AJG). Percentage body fat was measured by dual-energy X-ray absorptiometry (DEXA) as previously described (Raffan et al., 2006). For weight reduction, all dogs were fed a dry therapeutic diet (Canine Satiety^® diet, Royal Canin), with the exception of one dog (OBE16) that was fed a combination of wet and dry food (Canine Satiety^® diet, Royal Canin). Moreover, the formulation of Satiety changed in 2010 (Table 1, Data S1). The endpoint of the study was achievement of ideal body weight which, given differing degrees of adiposity, was individually set for each dog using the results of body composition analysis from before weight loss, as previously reported (German et al., 2012). Briefly, pre-weight-loss body composition data were entered into a computer spreadsheet which contained a bespoke mathematical formula to predict ideal bodyweight. The formula was based upon typical changes in body composition seen from previous weight loss studies at the same clinic (German et al., 2007; German et al., 2010a).

Fecal collection and DNA extraction

Fecal samples from dogs with obesity were collected after spontaneous defecation and stored at −20 °C being shipped to the Gastrointestinal Laboratory at Texas A&M University in February 2019. DNA was extracted from approximately 100 mg of stool using the Mo Bio PowerSoil^® DNA isolation kit (MoBio Laboratories, USA) according to the manufacturer’s instructions.

Quantitative PCR (qPCR) and Dysbiosis Index (DI)

Quantitative PCR was performed using universal bacteria primers and primers for the following bacterial groups: Blautia spp., Clostridium hiranonis (C. hiranonis), Escherichia coli (E. coli), Faecalibacterium spp., Fusobacterium spp., Streptococcus spp., and Turicibacter spp. The analysis was performed using a CFX 96 Touch TM Real-Time PCR Detection system (Biorad Laboratories). Ten µL SYBR-based reaction mixtures: 5 µL of SsoFast™ EvaGreen^® Supermix (Biorad Laboratories), 2.2 µL of water, 0.4 µL of each primer (final concentration: 400 nM), and 2 µL of DNA (1: 10 or 1: 100 dilution) were used for a protocol of 95 °C for 2 min, and 40 cycles at 95 °C 5 s and 10 s at the optimized annealing temperature for each primer set. Afterwards, a melt curve analysis was completed (AlShawaqfeh et al., 2017).

Results from the qPCR analysis for Blautia spp., C. hiranonis, E. coli, Faecalibacterium spp., Fusobacterium spp., Streptococcus spp., and Turicibacter spp. are expressed as the abundance of DNA for each bacterial group (logarithm of starting quantity or logarithm of relative DNA copy number). Relative DNA copy number for the mentioned bacteria were used to calculate a single numerical value known as the Dysbiosis Index (AlShawaqfeh et al., 2017). A value < 0 is indicative of a normal microbiota, numbers between 0 and 2 are considered equivocal, while numbers greater than 2 indicate fecal dysbiosis. The dysbiosis index is a commercially available assay, and the reference intervals have been validated with dogs from various countries, including the UK.

16S rRNA gene sequencing

The V4 variable region of the 16S rRNA gene was sequenced at the MR DNA laboratory (http://www.mrdnalab.com, Shallowater, TX, USA). Primers 515F (5′-GTGYCAGCMGCCGCGGTAA) (Parada, Needham & Fuhrman, 2016) to 806RB (5′-GGACTACNVGGGTWTCTAAT) (Apprill et al., 2015) and HotStarTaq Plus Master Mix (Qiagen, USA) were used to amplify samples and perform the Illumina MiSeq protocol following the manufacturer’s guidelines. Raw sequences were uploaded into Sequence Read Archive of the NCBI GenBank database under the accession number PRJNA580258.

Analysis of sequences

Quantitative Insights into Microbial Ecology 2 (QIIME 2.0) was used for analysis of the 16S rRNA amplicon sequences (Bolyen et al., 2019). Sequences were demultiplexed and the OTU table was created using DADA2 (Callahan et al., 2016). Operational taxonomic units (OTUs) were defined as sequences with at least 97% similarity within the Greengenes v 13.8 database (DeSantis et al., 2006). Prior to downstream analysis, sequences assigned as chloroplast, mitochondria, and low abundance OTUs, containing less than 0.01% of the total reads in the dataset were removed (McDonald et al., 2012). All samples were rarefied to an even depth of 36,775 sequence reads, based on the lowest read depth of samples.

Alpha diversity was evaluated with Chao1, Shannon diversity, and observed OTUs. Beta-diversity metric was estimated by unweighted and weighted phylogeny-based UniFrac distances and visualized using PCoA (Principal Coordinate Analysis) plots.

Statistical analysis

Normality was tested for all continuous variables using the Shapiro–Wilk test. Results were reported as mean (standard deviation [SD]) or median (range), when data were normally- or not normally-distributed, respectively. Differences in dog characteristics (i.e., age) between groups at baseline were compared using either the t-test. The ANOSIM (Analysis of Similarity) test within PRIMER 6 software package (PRIMER-E Ltd., Luton, UK) was used to analyze significant differences in microbial communities between groups.

Alpha diversity indices (Shannon, Chao1, and Observed OTUs), Dysbiosis Index, and quantitative PCR results were compared between groups using Wilcoxon test. A statistical software package (Prism version 8.0; GraphPad Software, San Diego, CA, USA), was used for the described analyses. To minimize false discoveries in univariate statistics, OTUs not present (0%) in at least 50% of the samples of at least one of the compared groups were considered rare and excluded from analysis. Filtered taxa were tested with Wilcoxon test for paired analysis using statistical software (R Studio Software version 1.2.1335 ©, R Studio, Boston, MA, USA; and JMP Pro 14; SAS, Durham, NC, USA). P-values were adjusted using the Benjamini–Hochberg Step-up method with a false discovery rate (FDR) of 0.05. For all statistical analyses, significance was set at p < 0.05.

Animal population characteristics

After a mean of 330.9 (SD 203.4) days on weight loss diet, BCS changed significatively in dogs with obesity when compared before (BCS 8; range 6–9) and after weight loss (median BCS 5; range 4–7; p < 0.001). However, BCS was not significantly different at baseline between the group of dogs that lost weight and those that did not complete the weight loss program ( p = 0.551). At baseline, the mean age in dogs with obesity that lost weight was 69.4 (SD 32.3) months, and 75.0 (SD 23.6) months for those that did not lose weight (p = 0.730), (Table 2, Data S1). The breeds included in the obese group were: Labrador Retriever (n = 7), Golden Retriever (n = 2), Cavalier King Charles spaniel (n = 2), Border Collie (n = 2), Pug (n = 2), Lhasa Apso (n = 1), American Bulldog (n = 1), Dachshund (n = 1), Rottweiler (n = 1), Newfoundland (n = 1), Bichon Frise (n = 1), Rough Collie (n = 1), and mixed breed (n = 3).

Body composition measurements were available for 19 out of the 20 dogs with obesity that lost weight (n = 20) and mean body fat mass was 44.8% (SD 5.0%) before weight loss (T0) and 30.4% (SD 6.5%) after weight loss (T2, p < 0.001). Mean rate of weight loss of starting body weight was 0.68% (SD 0.29%) per week, while the energy intake during the weight loss period was 60.8 (SD 5.5) kcal per kg^0.75 of ideal body weight per day (Table 2, Data S1).

Dogs were classified as having discontinued prematurely (n = 5) when their body weight at the time that they dropped out of the study (T2) were still significantly above of their ideal weight. The time between enrollment (T0) and the first follow-up (T1) was 15 (14–37) days, during which they had lost a median of 2.7% (0.0%–4.8%) of their starting body weight. At the time they dropped out of the program, after 414 (414–781) days on weight loss, they lost 9.9% (−3.6%–21.3%) of body weight. Although one dog did eventually lose 21.3% of its body weight, it was still significantly (∼13%) above its ideal weight at the time it discontinued and had also not lost any weight when the follow-up fecal sample was taken. All other dogs in this group lost <11% of their starting weight and were also above their ideal weights at the point that their weight loss was ended. The median rate of weight loss of starting body weight was 0.10% (−0.06%–0.39%) per week and the energy intake was 54.0 (51.8%–60.9%) kcal per kg^0.75 of ideal body weight per day. Body composition measurements were available in 4 out of 5 of the dogs at baseline, and median body fat was 44.7% (44.0%–47.5%). Body condition score did not change significatively during the period of weight loss (Data S1). No significant differences were identified for alpha and beta diversity when sex, age and neutered status were investigated for association with fecal microbiota in all dogs at baseline (Data not shown).

qPCR and dysbiosis index

On analysis by qPCR, there was a decreased abundance of E. coli (T0: 4.8 vs. T2: 3.5; p = 0.030), and an increased abundance of Fusobacterium spp. (T0: 7.4 vs. T2: 8.0; p = 0.017) after weight loss (Fig. 1). The values for the abundances of the evaluated bacteria by qPCR and for the Dysbiosis Index showed that the changes observed in the fecal microbiota before and after weight loss were mostly within the established reference intervals for clinically healthy dogs.

Changes in the fecal microbiota with weight loss in dogs with obesity

Weighted UniFrac analysis of similarities revealed significant clustering of the microbial communities in dogs with obesity before and after weight loss (weighted ANOSIM; p = 0.016, R = 0.073; Fig. 2A). Alpha diversity evenness and richness, as indicated by Shannon (p = 0.007), Chao1 (p = 0.007), and Observed OTUs (p = 0.007) indices were significantly increased after weight loss (Fig. 2B).

When the bacterial abundance was investigated at the taxa level, significant differences were found for the phyla Bacteroidetes, Firmicutes and Fusobacteria (Fig. 3A). The median Firmicutes/Bacteroidetes ratio decreased from 0.123 to 0.014 (p = 0.004) in dogs with obesity after weight loss (Fig. 3B). The relative abundance of Bacteroidetes (T0: 1.4% vs. T2: 10.1%; q = 0.002; Fig. 4A) and Fusobacteria (T0: 1.6% vs. T2: 6.2%; q = 0.040; Fig. 5A) increased significantly after weight loss, whilst the abundance of the phylum Firmicutes, instead decreased (T0: 92.3% vs. T2: 78.2%; q = 0.001; Fig. 6A).

The increase in Bacteroidetes was driven by the genera Bacteroides (T0: 0.7% vs. T2: 7.9%; q = 0.017; Fig. 4C) and Paraprevotella (T0: 0% vs. T2: 0.1%; q = 0.033; Fig. 4F).

From the phylum Fusobacteria, the genus Fusobacterium (T0: 1.6% vs. T2: 6.2%; q = 0.099; Fig. 5C) increased after weight loss. However, the corrected q-value did not reach significance.

Belonging to the phylum Firmicutes, the family Clostridiaceae decreased in abundance after weight loss (T0: 37.3% vs. T2: 24.6%; q = 0.068), but this difference did not reach significance after Benjamini correction (Fig. 6B). The same was noticed for the genus Clostridium (T0: 0.7% vs. T2: 0.6%; q = 0.119; Fig. 6C). The genus Megamonas (T0: 0.2% vs. T2: 0.0%; q = 0.027; Fig. 6E) and the genus Catenibacterium (T0: 2.3% vs. T2: 0.5%; q = 0.017) decreased significantly after weight loss (Fig. 6F), and the genus Coprobacillus increased in abundance (T0: 0% vs. T2: 0.5%, q = 0.033; Fig. 6G).

A short-term change in diet does not alter fecal microbiota

The fecal microbiota beta diversity of the dogs with obesity that stopped the weight loss program before reaching the endpoint (n = 5) was analyzed before and during the weight loss program. No significant differences were evident between the two time points (Weighted ANOSIM; p = 0.778 R =  − 0.080; Fig. 7A). Alpha diversity evenness and richness, as indicated by Shannon (p = 0.313), Chao1 (p = 0.438), and Observed OTUs (p = 0.438) indices did not show significant differences after a median period of 15 days on weight loss diet.