Short and long-term effects of a synbiotic on clinical signs, the fecal microbiome, and metabolomic profiles in healthy research cats receiving clindamycin: a randomized, controlled trial

Study population

This study was performed at the University of Tennessee’s Veterinary Medical Center and was approved by the Institutional Animal Care and Use Committee at the University of Tennessee, Knoxville (protocol number 2169). Twenty purpose-bred, domestic short hair cats from the University of Tennessee, College of Veterinary Medicine teaching and research colony determined to be healthy based on unremarkable physical examinations at the time of study initiation and lack of pertinent abnormalities on CBC or biochemistry profiles performed within the previous six months were enrolled. Prior antibiotic use was not noted in any of the medical records. All cats ate the same commercial adult dry cat food for the duration of the study (Hill’s Science Diet Light; Hill’s Pet Nutrition, Inc., Topeka, KS, USA). Cats were transitioned to individual housing the week prior to the start of the study and randomized into 1 of 2 treatment groups (placebo vs. synbiotic) at the start of the study using a random number sequence.

Study periods

The study consisted of a baseline period (days 1–7), a treatment period (days 8–28), and a long-term follow-up period (days 631–633) (Fig. 1). During the baseline period, no treatments were administered. During the treatment period, all cats received 150 mg (34 mg/kg/d, range 25.9–44.1 mg/kg/d) of clindamycin (Antirobe Antirobe Aquadrops; Pfizer, Pharmacia & Upjohn Company, New York, NY, USA) PO once daily with food for 21 days. The dose was chosen based on the manufacturer drug insert (Antirobe Antirobe Aquadrops manufacturer insert, Pfizer Pharmacia, March 2010), which noted changes in fecal consistency and emesis in cats receiving higher doses of clindamycin. Cats received either one capsule of placebo or a commercially-available synbiotic PO once daily at the time of antibiotic administration. The synbiotic contains five billion cfu of a proprietary mixture of Bifidobacterium bifidum, Enterococcus faecium and thermophilus, and Lactobacillus acidophilus, bulgaricus, casei, and lantarum per capsule, as well a proprietary blend of fructooligosaccharide and arabinogalactan. The time period for treatment was chosen based on previous veterinary publications (Garcia-Mazcorro et al., 2011; Hart et al., 2012; Koeppel et al., 2006), which showed significant changes in the fecal microbiome or fecal score by 21 days of treatment. After completion of treatment, cats were returned to group housing in the colony. On day 631, cats remaining in the teaching and research pool were reevaluated. Based on review of their medical records, none of the cats had received antibiotics, probiotics, other medications known to affect the microbiome, or dietary change in the intervening time.

Data collection

All cats underwent weekly physical examinations, including determination of body weight, throughout the study. Daily observations, including food intake and vomiting (present vs. absent), were collected throughout the study by an observer (NLL) blinded to the treatment groups. Photographs of defecated feces were taken daily by the same observer. Fecal consistency was scored (Greco, 2011) using still photographs at the completion of data collection by two investigators (JCW and JES), who were blinded to the cat, treatment group, and day for each fecal sample.

Fecal samples

To limit the impact of daily variation in bacterial populations and fecal metabolites, as well as differential distribution of bacterial groups and metabolites within individual fecal samples, first morning naturally-voided fecal samples were collected daily for three days for each cat at each of three timepoints: the conclusion of baseline (days 5–7), the conclusion of treatment (days 26–28), and after long-term recovery (days 631–633). Two grams from the center portion of each fecal sample was subdivided into two aliquots, with each aliquot placed into a 2 mL cryovial. Samples were immediately frozen and remained in storage at −80 °C pending completion of data collection. Samples for each cat from each timepoint were combined directly prior to sample analysis to generate pooled samples for microbiome and metabolomic analysis.

Microbiome analysis

Genomic DNA was extracted from 100 mg of feces from each pooled sample using a commercially available kit according to manufacturer’s protocol (PowerSoil; Mo Bio, Carlsbad, CA, USA).

Amplification and sequencing of the V4 variable region (primers 515F/806R) of the 16S rRNA gene was performed on a MiSeq (Illumina; MR DNA (Molecular Research LP), Shallowater, TX, USA) as previously described (Bell et al., 2014). The software Quantitative Insights Into Microbial Ecology (QIIME, v. 1.8; http://www.qiime.org) was used for processing and analyzing the sequences. The raw sequence data were de-multiplexed, and low quality reads were filtered using default parameters for QIIME. Chimeric sequences were filtered from the reads using USEARCH against the 97% clustered representative sequences from the Greengenes database (v. 13.8) and removed. The remaining sequences were assigned to operational taxonomic units (OTUs) using an open-reference OTU picking protocol using the default QIIME parameters, uclust consensus taxonomy assigner, and Greengenes database. The OTU table was rarefied to 35,000 sequences per sample. Alpha rarefaction plots, alpha diversity metrics (Chao1, Shannon, Goods Coverage, and Observed Species), and beta diversity metrics (weighted and unweighted UniFrac distance matrices) were created using QIIME scripts.

Quantitative PCR was performed for selected bacterial groups (total bacterial, Faecalibacterium spp., Turicibacter spp., Streptococcus spp., Escherichia coli, Blautia spp., Fusobacterium spp., Clostridium hiranonis) using extracted DNA as has been previously described (AlShawaqfeh et al., 2017). Briefly, 2 µl of normalized DNA (final concentration: 5 ng/ µl) was combined with 5 µl of a DNA-binding dye (SsoFast EvaGreen supermix; Bio-Rad Laboratories, CA, USA), 0.4 µl each of a forward and reverse primer (final concentration: 400 nM), and 2.6 µl of PCR water to achieve a total reaction volume of 10 µl. Oligonucleotide primers and probes, as well as respective annealing temperatures, are summarized in Table 1. Data were expressed as log amount of DNA (fg) for each particular bacterial group per 10 ng of isolated total DNA.

Metabolomics analysis

10 mg of lyophilized feces from each pooled sample were analyzed at a metabolomics facility using gas chromatography time-of-flight mass spectrometry (GC-TOF-MS) and standardized protocols as described in detail elsewhere (Fiehn et al., 2008; Minamoto et al., 2015). Raw data files were processed using ChromaTOF v. 2.32. BinBase algorithm was applied to match spectra to database compounds (West Coast Metabolomics Core, University of California, Davis, CA, USA) or to characterize as an unknown compound as reported previously (Honneffer et al., 2017). Quantification was reported by peak height of an ion at the specific retention index characteristic of the compound across all samples.

Statistical analyses

Descriptive statistics were generated for each parameter. Samples were analyzed for normality using the Shapiro–Wilk test and, for the presence of outliers, using box-and-whisker plots. Age and weight for the two treatment groups were compared using an independent two-sample Student’s t-test. Mean percent food intake, percent days of vomiting, and mean fecal scores were determined for each week within each study period (baseline and treatment weeks 1, 2, and 3). Mean food intake for each week in each study period was calculated as a percentage of food intake of each cat during the week prior to the start of the study. Inter-rater correlation coefficients were calculated for fecal scores. The mean of fecal scores assigned by the two investigators were used for all further statistical analyses. Mean food intake, percent days vomiting per week, and mean fecal score were compared between treatment groups using a multivariate repeated measures ANOVA. Treatment group (placebo or synbiotic), week, and cat were included as categorical variables. Treatment, week, and the treatment-by-week interaction were included as fixed effects. Week was included as a repeated measure with subject as cat. Age, sex, and weight were initially included as covariates but were not retained in the final models. Model assumptions regarding normally distributed residuals were verified with the Shapiro–Wilk test for normality. Model assumptions regarding equality of variances were verified with Levene’s Test for Equality of Variances.

Global changes in microbiota communities (beta diversity) between individuals were determined using unweighted Unifrac distance metrics; principal coordinates analysis (PCoA) plots and rarefaction curves were plotted using QIIME software. The ANOSIM function in PRIMER 6 (PRIMER-E Ltd, Ivybridge, UK) was used to compare beta diversity metrics across time and between treatment groups (Suchodolski et al., 2012). A dysbiosis index was calculated through methods previously described (AlShawaqfeh et al., 2017). The dysbiosis index is a quantitative PCR-based assay that summarizes the quantitative abundances of 8 major bacterial groups in feces in 1 number. The higher the dysbiosis index, the more deviation of the microbiota from normobiosis. Global changes in untargeted metabolomic profiles were determined using principal component analysis (PCA) plots developed using an online metabolomics software analysis suite (MetaboAnalyst 3.0: http://www.metaboanalyst.ca) as per standard protocols (Xia et al., 2015; Xia & Wishart, 2011). Pathway analysis was performed in the same software suite using the Homo sapiens pathway library, interquantile range data filtering, log transformation, and Pareto scaling.

Shannon indices, goods coverage, the Chao 1 metric, the dysbiosis index, proportions of bacteria groups, and fecal metabolite profiles were compared between treatment groups using a mixed model, split-plot repeated measures ANOVA that included fixed effects of treatment (placebo or synbiotic), time period, and treatment-by-time period interaction. Cat nested within group was included as a random effect. Age and sex were included as covariates in the initial analysis, but they were not retained in the final model due to lack of significant effects. The repeated measure of time period was accounted for in a repeated statement. A compound symmetry variance/covariance structure was incorporated into each model to account for the potential inclusion of constant covariates over time. The Shapiro–Wilk test of normality of the residuals was evaluated for each marker to confirm the assumption of normally distributed residuals had been met. Model assumptions regarding equality of variances were verified with Levene’s Test for Equality of Variances. Differences in least squares means were determined for markers with significant main effect or interaction terms. Due to a marked lack of variability in microbiome results for cats on days 26–28 (see Results below), rank transformation had to be employed to allow convergence of the model for analysis of the proportions of bacteria groups and ensure the statistical assumptions regarding normally distributed residuals and equality of variances were met. Only bacteria taxa that were present in at least 50% of cats in ≥1 group at ≥1 time point were included in statistical analyses.

P < 0.05 was considered statistically significant. P-values were corrected for multiple comparisons on each phylogenetic level and for untargeted metabolomics using the Benjamini & Hochberg’s False Discovery Rate (fdr). Publicly-accessible software packages (http://www.qiime.org; http://www.metaboanalyst.ca; SAS 9.4 release TS1M3; MedCalc 15.8: MedCalc, Ostend, Belgium; SAS 9.4 release TS1M3: SAS Institute Inc., Cary, NC, USA) were used for all analyses.

Study population

There were four female spayed and 6 male castrated cats initially enrolled in the placebo group and five female spayed cats and five male castrated cats in the synbiotic group. Median age was six years (range 3–8 years) for cats in the placebo group and five years (range 3–8 years) for cats in the synbiotic group. Median weight was 4.2 kg (range 3.4–5.4 kg) for cats in the placebo group and 4.3 kg (3.4–5.8 kg) for cats in the synbiotic group. No cat vomited any of the administered capsules during the study. Four cats were removed from the study for marked hyporexia (two placebo, one synbiotic) or high baseline fecal scoring (1 synbiotic).

After completion of treatment, cats were returned to the colony. On day 631, 15/16 of the cats that completed the trial were available for re-examination. One cat (from the synbiotic group) had been retired from the colony with subsequent private adoption. None of the cats had received antibiotics, probiotics, or other medications known to affect the microbiome in the intervening time, and none had undergone diet change. None of the cats were experiencing vomiting, abnormal fecal scores, or inadequate body condition scores.

Antibiotic-associated gastrointestinal signs

Food intake, percent days vomiting, and fecal scores over time by treatment group are summarized in Table 2. Food intake was decreased during treatment in 3/8 cats in the placebo group and 4/8 cats treated with the synbiotic. Food intake was significantly associated with week of treatment (F-value 5.2, P < 0.01) but did not differ between treatment groups (P = 0.5). Percent days vomiting per week was significantly associated with week of treatment (F-value 8.6, P < 0.01) but not treatment group (P = 0.4). It was highest in the first week of treatment, during which 14/16 cats had at least 1 episode of vomiting, seven in each group. Fecal scores increased significantly during treatment (F-value 9.4, P < 0.01) but did not differ between treatment groups (P = 0.2). All cats in each treatment group had fecal scores ≥5 at some point during treatment, although high daily variability was noted. There was no significant effect of age, weight, or sex on any of the clinical parameters evaluated.

Fecal microbiome

Cats in both treatment groups had marked changes in their fecal microbiome over time (Figs. 2– 3). Alpha diversity was significantly lower (P < 0.01) on days 26–28, based on the Shannon index, goods coverage, and Chao1 metric (Table 3). There was no significant association between treatment group and changes in Shannon index, goods coverage, or the Chao1 metric. Similarly, beta diversity was significantly altered on days 26–28 compared to baseline and days 631–633 in each treatment group (P < 0.01; R = 0.73–0.76) with no significant differences between treatment groups (Fig. 2). The dysbiosis index was significantly higher on days 26–28 than at baseline or on days 631–633 in each treatment group (Table 3), and a significant association was found between treatment group and dysbiosis index (Table 3, F-value 13.6, P < 0.01).

Five phyla were identified based on sequencing analysis (mean baseline prevalences): Actinobacteria (67.95%), Firmicutes (30.39%), Proteobacteria (1.03%), Bacteroidetes (0.73%), and Fusobacteria (0.02%), with marked and significant differences over time in relative abundances of all phyla except Fusobacteria (Fig. 3, Table 4). There was no significant effect of treatment group on changes in relative bacterial abundances at any timepoint, and there was no significant association between age or sex and changes in any of the analyzed fecal microbiome parameters.

Metabolome

Untargeted fecal metabolomics profiles differed markedly over time for cats in both treatment groups (Figs. 4– 5), with a variety of patterns of changes found. Fecal profiles for 178 of 252 assayed compounds (71%) differed significantly over time (see Additional file 1). Of those that differed over time, 47 of 178 (26.4%) metabolites had equivalent profiles at baseline and days 631–633. Partial return to baseline profiles were identified for an additional 14 metabolites (7.9%), and 21 metabolites (11.8%) had altered profiles on days 26–28 compared to baseline with no significant changes between days 26–28 and days 631–633. Changes in profiles for the remaining 96 metabolites (53.9%) were more complex. Profiles for 33 compounds (18.5%) differed significantly on days 631–633 from both baseline and days 26–28 with no difference between baseline and days 26–28, and profiles for 14 metabolites (7.9%) were significantly higher or lower at days 26–28 compared to baseline with further significant changes in the same direction by days 631–633. In contrast, profiles for 30 metabolites (16.9%) were significantly higher at days 26–28 than baseline but also significantly lower at days 631–633 than baseline; profiles for another 14 metabolites (7.9%) were decreased on days 26–28 with much higher profiles on days 631–633 than baseline. Finally, 5 metabolites had individual patterns of change.

Significant differences (fdr adjusted P < 0.05) were noted between treatment groups for seven metabolite profiles (Fig. 6). Significant group by time interactions were found for fecal profiles of putrescine, isopentadecanoic acid, cellobiose, ethanolamine, lactose, and D-erythro-sphingosine, while fecal profiles for N-acetylglutamate differed significantly by treatment group alone.

With regard to metabolites of known biological importance, there were significant changes in profiles of some short-chain fatty acid (SCFA) synthesis metabolites, bile acids, tryptophan degradation pathway metabolites, sphingolipids, polyamines, and benzoic and cinnaminic acids over time (Table 5). Pathway analysis revealed significant changes (fdr adjusted P < 0.05) over time for 55 metabolic pathways; 47 of which had impact factors >0 (Table 6). The majority of pathways affected were related to SCFA, amino acid, sugar, and nucleotide synthesis and degradation. Amino acid pathways that were affected included those related to tryptophan degradation. Other pathways significantly affected included those related to linoleic acid, sphingolipid, and glycerolipid metabolism.