Intermittent Hypoxia and Hypercapnia Alter Diurnal Rhythms of Luminal Gut Microbiome and Metabolome

Abstract

Obstructive sleep apnea (OSA), characterized by intermittent hypoxia and hypercapnia (IHC), affects the composition of the gut microbiome and metabolome. The gut microbiome has diurnal oscillations that play a crucial role in regulating circadian and overall metabolic homeostasis. Thus, we hypothesized that IHC adversely alters the gut luminal dynamics of key microbial families and metabolites. The objective of this study was to determine the diurnal dynamics of the fecal microbiome and metabolome of Apoe^-/- mice after a week of IHC exposure. Individually housed, 10-week-old Apoe^-/- mice on an atherogenic diet were split into two groups. One group was exposed to daily IHC conditions for 10 h (Zeitgeber time 2 [ZT2] to ZT12), while the other was maintained in room air. Six days after the initiation of the IHC conditions, fecal samples were collected every 4 h for 24 h (6 time points). We performed 16S rRNA gene amplicon sequencing and untargeted liquid chromatography-mass spectrometry (LC-MS) to assess changes in the microbiome and metabolome. IHC induced global changes in the cyclical dynamics of the gut microbiome and metabolome. Ruminococcaceae, Lachnospiraceae, S24-7, and Verrucomicrobiaceae had the greatest shifts in their diurnal oscillations. In the metabolome, bile acids, glycerolipids (phosphocholines and phosphoethanolamines), and acylcarnitines were greatly affected. Multi-omic analysis of these results demonstrated that Ruminococcaceae and tauro-β-muricholic acid (TβMCA) cooccur and are associated with IHC conditions and that Coriobacteriaceae and chenodeoxycholic acid (CDCA) cooccur and are associated with control conditions. IHC significantly change the diurnal dynamics of the fecal microbiome and metabolome, increasing members and metabolites that are proinflammatory and proatherogenic while decreasing protective ones. IMPORTANCE People with obstructive sleep apnea are at a higher risk of high blood pressure, type 2 diabetes, cardiac arrhythmias, stroke, and sudden cardiac death. We wanted to understand whether the gut microbiome changes induced by obstructive sleep apnea could potentially explain some of these medical problems. By collecting stool from a mouse model of this disease at multiple time points during the day, we studied how obstructive sleep apnea changed the day-night patterns of microbes and metabolites of the gut. Since the oscillations of the gut microbiome play a crucial role in regulating metabolism, changes in these oscillations can explain why these patients can develop so many metabolic problems. We found changes in microbial families and metabolites that regulate many metabolic pathways contributing to the increased risk for heart disease seen in patients with obstructive sleep apnea.

Keywords: animal models of human disease; atherosclerosis; circadian rhythm; computational biology; metabolome; microbiome.

FIG 1

IHC affects the cyclical dynamics of the gut microbiome. (A) Venn diagram of unique nonzero detected sOTUs in each cohort overall. Purple indicates sOTUs in common. (B) Weighted UniFrac (beta diversity) PCoA of samples. Shading represents different time points as indicated. (C) Weighted UniFrac PCoA of only axis 1 over time. The solid lines indicate the average for the group, and the dotted lines indicate individual mice. (D) Proportional-abundance representation of the top 5 microbial families. Control samples exposed only to normal air conditions are in red (n = 4; 5 to 6 time points per mouse). Experimental samples exposed to IHC conditions are in blue (n = 4; 5 to 6 time points per mouse).

FIG 2

IHC affects the cyclical dynamics of the fecal metabolome. See Table S1 in the supplemental material for a list of the full annotations and abbreviations of the metabolites displayed. (A) Canberra PCoA of metabolomics samples. Shading represents different time points. Significance was determined by PERMANOVA. (B) Canberra PCoA of axis 2 over time. Solid lines indicate means for the group, and dotted lines indicate individual mice. The yellow box indicates the time under IHC exposure for the treatment group. (C) Pie charts of key groups of metabolites, separated by condition. (D) Heat map of level 1 bile acids, organized using hierarchical clustering based on controls. Yellow indicates the time under IHC exposure for the treatment group. For other level 3 bile acids, see Fig. S3C. (E) Heat map of selected phosphocholines, organized using hierarchical clustering based on controls. The value of each square of the heat map represents the mean relative abundance value for all mice under that condition for that time point. The heat maps are also row normalized across both conditions and placed on a standard scale, referenced in the center, to allow easier comparison. # indicates a metabolite that is also shown in Fig. 4 and Fig. S4. Air is in red (n = 4; 5 to 6 time points per mouse); IHC is in blue (n = 4; 5 to 6 time points per mouse).

FIG 3

Microbes and metabolites with linked expression levels as determined by mmvec analysis. (A) mmvec (58) cooccurrence analysis (y axis) based on songbird (56) multinomial regression differential ranking analysis (x axis). Bile acids generally have level 1 identifications, except for cholic acid (CA), CDCA, and murideoxycholic acid (MDCA), which are level 3 annotations. (B) Log conditional probability heat map, organized using hierarchical clustering, with the top 4 differentially abundant microbial families and the top differentially abundant bile acids. Pink and green boxes highlight the top 2 points with the highest correlation values. (C) Log ratios of the top correlated microbes (x axis) and metabolites (y axis) identified in panel B. Microbial log ratios were determined as the number of all reads from sOTUs that belong to the family Ruminococcaceae divided by the number of all reads from sOTUs that belong to the family Coriobacteriaceae. Metabolite log ratios were determined as the raw values from CDCA divided by the raw values of TβMCA. (D) Linear regression plot using the same log ratios as the ones in panel C, with best-fit lines and shaded areas representing 95% confidence intervals. Log ratios are based on natural log. Control samples with exposure only to normal air conditions are in red (n = 4; 5 to 6 time points per mouse). Experimental samples exposed to IHC conditions are in blue (n = 4; 5 to 6 time points per mouse). Complete metadata can be found in Table S3 at https://doi.org/10.6084/m9.figshare.14614434.

FIG 4

Cyclical dynamics of log ratios of key microbes and metabolites. Additional selected log ratios (natural log), their cyclical dynamics over time (double-line plots) (top), and their relative abundances grouped by cycle phase (box plots) (bottom) are shown. (A) Log ratios of all reads from sOTUs that belong to the family Ruminococcaceae divided by all reads from sOTUs that belong to the family Verrucomicrobiaceae. (B) Log ratios of all reads from sOTUs that belong to the family Coriobacteriaceae divided by all reads from sOTUs that belong to the family Verrucomicrobiaceae. (C) Log ratios of all reads from sOTUs that belong to the family Ruminococcaceae divided by all reads from sOTUs that belong to the family Coriobacteriaceae. (D) Log ratios of raw values of CDCA divided by raw values of TβMCA, the two most differentially abundant bile acids identified in Fig. S5B in the supplemental material. (E) Log ratios of the raw values of UDCA divided by the raw values of TβMCA. Solid lines represent the means, and error bars indicate standard errors of the means. Individual mice are indicated by dashed lines. Shading indicates when room lights are off (i.e., active/feeding time for the mice). Yellow squares indicate the 10 h of the day where mice under the IHC conditions would be exposed to experimental conditions (ZT2 [after collection] until ZT12). MetaCycle with the JTK method was used to determine cyclicity. *, P < 0.05; **, P < 0.01; ***, P < 0.001; ****, P < 0.0001 (by a Mann-Whitney-Wilcoxon test). @ indicates diurnal oscillations as determined by MetaCycle (JTK) with a P value of <0.05. Control samples with exposure to only normal air conditions are in red (n = 4; 5 to 6 time points per mouse). Experimental samples exposed to IHC conditions are in blue (n = 4; 5 to 6 time points per mouse). Error bars were not placed for time points where there were fewer than 3 log ratios available.