Integrative Longitudinal Analysis of Metabolic Phenotype and Microbiota Changes During the Development of Obesity

Abstract

Obesity has increased at an alarming rate over the past two decades in the United States. In addition to increased body mass, obesity is often accompanied by comorbidities such as Type II Diabetes Mellitus and metabolic dysfunction-associated fatty liver disease, with serious impacts on public health. Our understanding of the role the intestinal microbiota in obesity has rapidly advanced in recent years, especially with respect to the bacterial constituents. However, we know little of when changes in these microbial populations occur as obesity develops. Further, we know little about how other domains of the microbiota, namely bacteriophage populations, are affected during the progression of obesity. Our goal in this study was to monitor changes in the intestinal microbiome and metabolic phenotype following western diet feeding. We accomplished this by collecting metabolic data and fecal samples for shotgun metagenomic sequencing in a mouse model of diet-induced obesity. We found that after two weeks of consuming a western diet (WD), the animals weighed significantly more and were less metabolically stable than their chow fed counterparts. The western diet induced rapid changes in the intestinal microbiome with the most pronounced dissimilarity at 12 weeks. Our study highlights the dynamic nature of microbiota composition following WD feeding and puts these events in the context of the metabolic status of the mammalian host.

Keywords: bacteria-phage dynamics; bacteriophage; metabolic phenotype; microbiota; obesity.

Figure 1

Percent body weight change. Weekly percent weight change over time is shown for the Chow and WD groups. Group differences over the course of dietary treatment were analyzed by ANOVA. All data points are shown as group mean ± SE. (**p < 0.01, ***p < 0.001 compared to Chow).

Figure 2

Energy expenditure at 2-, 4- and 12-weeks. (A) Mean circadian analysis of energy expenditure at each hour in the 24-hour cycle in the Chow and WD groups after 2-weeks of dietary exposure. (B) Mean circadian analysis of energy expenditure at each hour in the 24-hour cycle in the Chow and WD groups after 4-weeks of dietary exposure. (C) Mean circadian analysis of energy expenditure at each hour in the 24-hour cycle in the Chow and WD groups after 12-weeks of dietary exposure. All data points are shown as group mean ± SE. (*p < 0.05, **p < 0.01, ***p < 0.001 compared to Chow).

Figure 3

Respiratory exchange ratio at 2-, 4- and 12-weeks. (A) Mean circadian analysis of respiratory equivalent ratio at each hour in the 24-hour cycle in the Chow and WD groups after 2-weeks of dietary exposure. (B) Mean circadian analysis of respiratory equivalent ratio at each hour in the 24-hour cycle in the Chow and WD groups after 4-weeks of dietary exposure. (C) Mean circadian analysis of respiratory equivalent ratio at each hour in the 24-hour cycle in the Chow and WD groups after 12-weeks of dietary exposure. All data points are shown as group mean ± SE. (*p < 0.05, **p < 0.01, ***p < 0.001 compared to Chow).

Figure 4

An nMDS ordination of the microbiota samples (A) by diet and (B) over time. The taxonomic profiles of the samples were used to compute the sample dissimilarity matrix using Bray-Curtis dissimilarity index. The matrix was used to compute an ordination of the samples in two dimensions (MDS1 and MDS2). The stress associated with this ordination is 0.156. The shapes in plot B denote the diet (Chow and WD), where the color denotes the time point (0 days, 2 days and 2, 8 and 12 weeks on the diets).

Figure 5

Changes in relative abundance of microbial composition after administration of the Western diet. (A) Phylum, (B) Class and (C) Family level bacterial composition in mice fed Chow or WD after 2 days, 2, 8 and 12 weeks of dietary exposure. The mean relative abundance (%) of bacterial phyla are shown. Statistical results outlined in Tables S1–S3.

Figure 6

Changes in relative abundance of bacteriophage composition after administration of the Western diet. (A) Family and (B) Genus level viral composition derived from the order Caudovirales in mice fed Chow or WD at 2 days, 2, 8 and 12 weeks of dietary exposure. Data shown are based on those families belonging to the order Caudovirales. The mean relative abundances (%) of viral families are shown. Statistical results outlined in Tables S4, S5.

Figure 7

Diet Induced Bacteria-Bacteriophage Patterns in Obesity. Pearson’s correlation plot of the top 55 most abundant Bacterial Families and all Bacteriophage Genera (Viral OTUs within the order Caudovirales) for mice fed Chow or WD at 12 weeks. Positive values (red circles) indicate positive correlation coefficients above 0.6, and negative values (blue circles) indicate inverse correlation coefficients below -0.6. The size and shading of the circles indicate the magnitude of the correlation, where darker shades indicate a stronger correlation than lighter shades. Organization of the plot is based on the correlation coefficient values of Brachyspiraceae. Correlation coefficient values outside of the threshold of 0.6 are not included in this plot, but are outlined in Table S9. Putative host information and correlation coefficient values between bacteriophage genera and putative bacterial host family are listed in Table S8.

Figure 8

Diet Induced Bacteria/Bacteriophage – Metabolic Patterns in Obesity. Pearson’s correlation plot of (A) Bacterial Families (top 55 most abundant) or (B) Bacteriophage Genera (Viral OTUs within the order Caudovirales) and metabolic parameters for mice fed Chow or WD for data after 12 weeks of dietary exposure. Statistical significance was determined for all pairwise comparisons. Positive values (red circles) indicate positive correlation coefficients above 0.6, and negative values (blue circles) indicate inverse correlation coefficients below -0.6. The size and shading of the circles indicate the magnitude of the correlation, where darker shades indicate a stronger correlation than lighter shades. Correlation coefficient values outside of the threshold of 0.6 are not included in this plot (Tables S10, S11).