3.2 Diet-induced weight loss changes the gut microbial composition
The microbial community structure detected in fecal samples significantly differed by beta diversity assessed with Bray-Curtis distances between the four study phases (Fib. 2b, PERMANOVA p = 0.011). No significant differences in alpha diversity were identified between study phases using three metrics: observed ASVs, Shannon diversity index, and Inverse Simpson diversity index (Fig. 2c-e).
The relative abundances of ASVs detected in the fecal samples from all cats were assessed at both the phylum and family level (Fig. 3). ASVs belonging to the Firmicutes phylum were predominant in most fecal samples across all phases, though Actinobacteriota and Bacteroidota phyla were also major contributors to community composition (Fig. 3a). This agrees with previous reports of phylum-level bacterial diversity by 16S rRNA gene sequencing in healthy cats, though often members within the phylum Proteobacteria are similarly abundant to Actinobacteriota and Bacteroidota[10]. Bacteroidota was the predominant phylum represented in a fecal sample from cat 4 on study day 5, which was the first sample collected after all cats were transitioned to the calorie restricted diet. While detected in lower relative abundance, ASVs belonging to the phyla Desulfobacterota, Fusobacteriota, Proteobacteria, and Verrucomicrobiota were detected in amounts surpassing > 1% of community structure from at least one sample across the study.
ASVs belonging to 26 families were identified with a relative abundance > 1% in at least one sample. Of these families, five were within Actinobacteriota, two within Bacteroidota, one within Desulfobacterota, 14 within Firmicutes, one within Fusobacteriota, two within Proteobacteria, and one within Verrucomicrobiota (Fig. 3b-c). The relative abundances of ASVs at the family level show that a variety of families contribute to the microbial community structure and vary with both study time course and within individuals. Yet, it is difficult to single one or a few families out as consistently defining that structure.
3.3 Known short-chain fatty acid producing bacteria, including five Blautia genus members, are enriched during and after diet-induced weight loss in cats
LEfSe identified differentially abundant features in any one study phase compared to all other phases (Fig. 4a). Eight ASVs were differentially abundant (LDA score > 2.0). No ASVs were differentially abundant during the Obese MD study phase. During the Obese OM study phase, Prevotella 9 copri (ASV 71) and a Blautia (ASV 216) had significant increases in relative abundance (Supplementary Fig. S2). Specifically, for Prevotella 9 copri, the elevation that occurred during Obese OM phase compared to Obese MD (FDR adj. p = 0.0009) remained significantly increased compared to Obese MD for the additional Lean OM (FDR adj. p = 0.0038) and Lean MD (FDR adj. p = 0.0149) phases.
During the Lean OM phase, Blautia caecimuris (ASV 372) (FDR adj. p = 0.0058) and Solobacterium (ASV 638) (FDR adj. p = 0.0006) had significant increases in relative abundance from Obese MD (Supplementary Fig. S2). The relative abundance was also increased from the Obese OM phase for Blautia caecimuris (ASV 372) (FDR adj. p = 0.0334) and Solobacterium (ASV 638) (FDR adj. p = 0.0479) during the Lean OM phase. Blautia caecimuris (ASV 372) then significantly decreased (FDR adj. p = 0.0117) when the maintenance diet was reintroduced during the Lean MD phase.
During the Lean MD phase, Clostridium sensu stricto 1 (ASV 821), two Blautia ASVs (ASV 358 and ASV 359), and a Lachnospiraceae ASV (ASV 396) had significant increases in relative abundance (Supplementary Fig. S2). One of the Blautia genus ASVs (ASV 359) (FDR adj. p = 0.0061) and Clostridium sensu stricto 1 (ASV 821) (FDR adj. p = 0.0079) were significantly increased from the Obese MD phase. This is notable since the cats were eating the same maintenance diet at both times, leaving the possibility that compositional microbial changes and/or other host factors had shifted during weight loss to allow for this differential response of the gut microbes to the same maintenance diet once the OM weight loss diet was discontinued.
Log2 fold changes (log2FC) of abundance compared to Obese MD were also performed (Fig. 4b-d). Like the LEfSe analysis, Prevotella 9 copri (ASV 71) was significantly enriched during the Obese OM phase (log2FC = 2.01, FDR adj. p = 0.0254). Three additional ASVs were also enriched during the Obese OM phase: Turicibacter sanguinis (ASV 818) (log2FC = 19.5, FDR adj. p = 0.0006), Clostridium sensu stricto 1 paraputrificum (ASV 738) (log2FC = 17.8, FDR adj. p = 0.0009), and Capnocytophaga (ASV 112) (log2FC = 18.0, FDR adj. p = 0.0012). The three ASVs with the greatest log2FC decrease during the Obese OM phase were Butyricicoccus pullicaecorum (ASV 241) (log2FC = -26.1, FDR adj. p < 0.0001), Clostridium sensu stricto 1 (ASV 840) (log2FC = -25.0, FDR adj. p < 0.0001), and Enterococcus (ASV 773) (log2FC = -7.3, FDR adj. p < 0.0001). This highlights that members within the same genus can fluctuate with changing microbial community dynamics, as ASVs belonging to Clostridium sensu stricto 1 were both significantly enriched (ASV 738) and significantly decreased (ASV 840).
When the Lean OM phase was compared to Obese MD, the same Solobacterium (ASV 638) identified in the LEfSe analysis was significantly enriched (log2FC = 8.8, FDR adj. p < 0.0001). The three ASVs with the greatest log2FC increase during the Lean OM phase compared to Obese MD were Blautia (ASV 427) (log2FC = 20.9, FDR adj. p < 0.0001), Lachnospiraceae UCG-008 (ASV 400) (log2FC = 17.9, FDR adj. p = 0.0010), and Turicibacter sanguinis (ASV 818) (log2FC = 16.8, FDR adj. p = 0.0025). The three ASVs with the greatest log2FC decrease during the Lean OM phase were Blautia (ASV 422) (log2FC = -26.5, FDR adj. p < 0.0001), Clostridium sensu stricto 1 (ASV 840) (log2FC = -24.7, FDR adj. p < 0.0001), and Alistipes massiliensis (ASV 141) (log2FC = -19.7, FDR adj. p = 0.0002). This demonstrates another situation where ASVs within the same genus had opposing responses to fluctuating microbial community dynamics, as Blautia (ASV 427) had the largest log2FC increase and Blautia (ASV 422) had the largest log2FC decrease.
Comparing the Lean MD phase to Obese MD, the same Clostridium sensu stricto 1 (ASV 821) identified in the LEfSe analysis was significantly enriched (log2FC = 6.1, FDR adj. p = 0.0201). The three ASVs with the largest log2FC increase were Clostridium sensu stricto 1 paraputrificum (ASV 738) (log2FC = 19.9, FDR adj. p < 0.0001), Blautia (ASV 427) (log2FC = 19.7, FDR adj. p = 0.0001), and Turicibacter sanguinis (ASV 818) (log2FC = 16.4, FDR adj. p = 0.0051). The three ASVs with the greatest log2FC decrease during the Lean MD phase were Clostridium sensu stricto 1 (ASV 840) (log2FC = -25.2, FDR adj. p < 0.0001), Alistipes massiliensis (ASV 141) (log2FC = -20.1, FDR adj. p = 0.0002), and Streptococcus (ASV 649) (log2FC = -19.0, FDR adj. p < 0.0001). Interestingly, Blautia (ASV 427) and Turicibacter sanguinis (ASV 818) were among the top three enriched and Clostridium sensu stricto 1 (ASV 840) and Alistipes massiliensis (ASV 141) were also among the top three decreased ASVs during the Lean OM phase. This may suggest that these ASVs represent persistent gut microbial community changes following weight loss even once the OM weight loss diet was discontinued.
3.4 Propionic acid production is promoted during diet-induced weight loss in cats and correlates with abundance of two bacterial species
The concentrations of SCFAs detected in individual feline fecal samples were assessed. Acetic acid concentration predominated in most instances, followed by propionic acid and butyric acid in similar concentrations, and then valeric acid, isovaleric acid, and isobutyric acid being in lower concentrations (Supplementary Figs. S3 and S4). When grouped by study phase, concentrations of fecal SCFAs clustered by study phase using principal component analysis (Fig. 5a). Propionic acid had the highest mean decrease in accuracy when permuted in a Random Forest machine learning algorithm (Fig. 5b). This demonstrated the importance of propionic acid concentration as a discriminating feature when attempting to predict which study phase a fecal sample belonged to based solely upon the detected SCFA concentrations. No difference in total SCFA concentration was identified between study phases (Fig. 5c). When individual SCFA concentrations were assessed between study phases (Fig. 5d-i), the median concentration of propionic acid was increased during the Obese OM study phase (median 76.25 µmol/mL) compared to Obese MD (median 47.9 µmol/mL) but did not reach statistical significance (FDR adj. p = 0.2191). During the Lean MD phase, the concentration of propionic acid was significantly reduced compared to Obese MD (FDR adj. p = 0.0404), Obese OM (FDR adj. p = 0.0024), and Lean OM (FDR adj. p = 0.0098) phases. The significant reduction in propionic acid concentration during the Lean MD phase compared Obese MD is notable since cats were eating the same maintenance diet during both phases, suggesting that diet alone was not the only factor determining fecal propionic acid concentration.
Separately from propionic acid, concentrations of isobutyric acid (FDR adj. p = 0.0324) and isovaleric acid (FDR adj. p = 0.0111) were significantly reduced from Obese MD after diet-induced weight loss had occurred in the Lean OM study phase (Fig. 5e-f). Isobutyric acid and isovaleric acid represent the two branched-chain fatty acids (BCFAs) that were measured. No statistically significant changes in concentration occurred between study phases for acetic acid, butyric acid, or valeric acid (Fig. 5g-i).
The relative composition that each of the six measured SCFAs contributed to the total concentration of SCFAs within each sample was also explored (Supplementary Fig. S5). Propionic acid contributed a significantly greater percentage of the total composition of SCFAs within feces during diet-induced weight loss in both the Obese OM (FDR adj. p < 0.0001) and Lean OM (FDR adj. p < 0.0001) phases (Fig. 6a-b). During the Lean MD phase, the composition of propionic acid was reduced to a degree not significantly different from the Obese MD study phase (FDR adj. p = 0.2535). Compositional changes also occurred with a reduction in the BCFA isobutyric acid during the Obese OM (FDR adj. p = 0.0068) and Lean OM (FDR adj. p = 0.0068) phases (Fig. 6c). Significantly reduced composition occurred for the other measured BCFA, isovaleric acid, during Lean OM phase (FDR adj. p = 0.0464) (Fig. 6d). These reductions were then followed by a significant increase of both isobutyric acid (FDR adj. p = 0.0011) and isovaleric acid (FDR adj. p = 0.0010) composition in the Lean MD phase. Though no differences in butyric acid concentration were identified between study phases (Fig. 5h), compositionally butyric acid was significantly increased in the Lean MD phase compared to Obese MD (FDR adj. p = 0.0426), Obese OM (FDR adj. p = 0.0481), and Lean OM (FDR adj. p = 0.0295) (Fig. 6f). No differences in acetic acid composition were identified between phases (Fig. 6e).
Given that propionic acid was the microbial metabolite identified as enriched during diet-induced weight loss in cats, it was investigated whether changes in microbial relative abundance were correlated with the observed increase in fecal propionic acid. To answer this question in the context of diet-induced weight loss, fecal samples from the Lean OM phase were used to capture the period where weight loss had been achieved and cats had remained on the calorie-restricted diet for at least 11 weeks. Fecal samples from the Obese MD phase, when propionic acid composition was originally lower with a distinct microbiota community structure, were also included in this analysis. Both Prevotella 9 copri (Spearman’s ρ = 0.6385, p = 0.0006) and Blautia caecimuris (Spearman’s ρ = 0.5269, p = 0.0068) relative abundance were significantly positively correlated with the composition of propionic acid (Fig. 7a-b). This identifies two bacterial species positively correlated with propionic acid production in obese cats undergoing weight loss and are relevant to microbial community structure in the context of diet-induced weight loss.