Diet Mediate the Impact of Host Habitat on Gut Microbiome and Influence Clinical Indexes by Modulating Gut Microbes and Serum Metabolites

Abstract

The impact of external factors on the human gut microbiota and how gut microbes contribute to human health is an intriguing question. Here, the gut microbiome of 3,224 individuals (496 with serum metabolome) with 109 variables is studied. Multiple analyses reveal that geographic factors explain the greatest variance of the gut microbiome and the similarity of individuals' gut microbiome is negatively correlated with their geographic distance. Main food components are the most important factors that mediate the impact of host habitats on the gut microbiome. Diet and gut microbes collaboratively contribute to the variation of serum metabolites, and correlate to the increase or decrease of certain clinical indexes. Specifically, systolic blood pressure is lowered by vegetable oil through increasing the abundance of Blautia and reducing the serum level of 1-palmitoyl-2-palmitoleoyl-GPC (16:0/16:1), but it is reduced by fruit intake through increasing the serum level of Blautia improved threonate. Besides, aging-related clinical indexes are also closely correlated with the variation of gut microbes and serum metabolites. In this study, the linkages of geographic locations, diet, the gut microbiome, serum metabolites, and physiological indexes in a Chinese population are characterized. It is proved again that gut microbes and their metabolites are important media for external factors to affect human health.

Keywords: China population; diet; geography location; gut microbiome; physiological indexes; serum metabolome.

Figure 1

Overview of gut microbiota composition and associated host factors. A) Graphical summary of the cohort and overview of variables (N = number of variables collected, n = sample size). Background in the map of China is colored according to region classification. The red curve indicates the boundary of two area in this study. The red straight line indicates Hu Huanyong line of China's population boundary. B) The relative abundances of the top five phylum‐level microbiota among individuals (n = 3,224). C) The relative abundances of the top ten genera. D) The effect sizes of host factors significantly associated with gut microbial variations were evaluated by PERMANOVA (Adonis) filtered based on FDR < 0.05. The bars are colored according to metadata categories. HLJ, Heilongjiang (n = 235); LN, Liaoning (n = 141); BJ, Beijing (n = 115); SD, Shandong (n = 131); JS, Jiangsu (n = 146); SH, Shanghai (n = 140); ZJ, Zhejinag (n = 142); SX, Shaanxi (n = 140); HeN, Henan (n = 390); HB, Hubei (n = 134); HN, Hunan (n = 457); CQ (n = 148), Chongqing; GZ, Guizhou (n = 342); YN, Yunnan (n = 130); GX, Guangxi (n = 433).

Figure 2

Characterization of geographic‐specific gut microbiota signatures. A) The effect sizes of city/county factor and community factor for gut microbiota variations using Adonis. On the left side of the x‐axis, the effect sizes of city/county factor in each province. On the right side of the x‐axis, the effect sizes of community factor in each city/county (FDR < 0.05). B) The microbiota dissimilarity assessed by Bray‐Curtis distance and microbiota similarity indicated by Pearson coefficient in different geographic ranges. C,D) The linear regression of the relationship between Bray‐Curtis distance or Pearson values and the actual geographic distances from the northernmost resident to the other 3,223 individuals (C) or from the southernmost residents to the other 3223 individuals (D). E) The microbiota α diversity indicated by Observed OTUs in 15 provinces. Background colors in the map of China show the value of local Observed OTUs. F) Linear Discriminant Analysis (LDA) and principal component analysis (PCA) visualizing the beta‐diversity. The values on the PC1 axis in 15 provinces showing downside. The dots were colored according to different province. G) MaAsLin analysis on the microbiome composition and specific genera in 15 provinces (FDR < 0.01). H) Random forest model to determine a person's province location based on province‐specific genera. The number represents the prediction accuracy.

Figure 3

Mediation linkages among geographic location, diet, and the gut microbiome. A,B) Correlations between foods, nutrient, demographic, physiological parameters, and different geographic ranges using Cramer's V (A) and Adonis (B) (p < 0.05 and Cramer's V > 0.2). C) Correlation among 10 geography‐associated food analyzed by Cramer's V based on all samples (n = 3,181). D) The relationship between geography‐associated food and gut microbiota identified by Boruta (on the left) and the relative abundances of food‐associated microbiota in 15 provinces. The peak plots are colored according to phylum. E) Causal linkages among latitude, food, and the gut microbiota by mediation analysis (p < 0.05). F,H) Examples of causal relationships between latitude, food, and the gut microbiota. The gray lines indicate the associations. The red and blue arrowed lines indicate the latitude effects on microbiota mediated by specific food. The beta coefficient and p values are labeled at each edge. The proportions of indirect effect (mediation effect) and mediation p values are labeled at the center of the ring charts.

Figure 4

Influence of the microbiome and diet on inter‐individual variation of serum metabolome. A) Contributions of indicated factors to inter‐individual variation in the serum metabolome estimated by the Adonis method (FDR < 0.05). B) Venn diagram indicating the number of metabolites significantly associated with specific foods and gut microbiota genera, as estimated using Spearman's correlation (FDR < 0.05). C) Association of serum metabolites and foods or microbiota genera in HN and GZ province (n = 496, FDR < 0.05). D) Mediation links between food, gut microbiota, and serum metabolites showed by parallel coordinates chart that are significant at FDR < 0.05. Shown are foods (left), gut microbiota (middle), and serum metabolites (right). The curved lines connecting the panels indicate the mediation effects. E) Analysis of the effect of vegetable oil intake on the levels of 1‐palmitoyl‐2‐palmitoleoyl‐GPC (16:0/16:1) as mediated by Blautia. F) Analysis of the effect of fruit intake on the levels of threonate as mediated by Blautia. G) Analysis of the effect of wine intake on the levels of 2‐hydroxy‐3‐methylvalerate through Clostridium XVIII. In (E–G), the gray lines indicate the associations. The blue arrowed lines indicate the food effects on serum metabolites mediated by specific genera. The beta coefficient and p values are labeled at each edge. The proportions of indirect effect (mediation effect) and mediation p values are labeled at the center of the ring charts.

Figure 5

Associations between food, microbiome, metabolites, and host physiological parameters. A) Clustered heatmaps indicating the associations between food, gut microbiome, metabolites, and physiological parameters that are significantly correlated with each other by Spearman's correlation analysis (FDR < 0.25 and p < 0.01). Three schematic examples of identifying multi‐factor biological links along the food‐microbiome‐metabolite‐host phenotype axis showing in the box embedded in between the heatmaps. B) Parallel coordinates chart showing the mediation links between gut microbiota, serum metabolites and physiological parameters that were significant at FDR < 0.05. Shown are gut microbiota (left), serum metabolites (middle), and physiological parameters (right). The curved lines connecting the panels indicate the mediation effects. C) Analysis of the causal relationships among vegetable oil intake, Blautia, 1‐palmitoyl‐2‐palmitoleoyl‐GPC (16:0/16:1), and systolic blood pressure combined with Figure 4E. D) Analysis of the causal relationships among fruit intake, Blautia, threonate and systolic blood pressure combined with Figure 4F. E) Analysis of the causal relationships among wine intake, Clostridium XVIII, 2‐hydroxy‐3‐methylvalerate and aspartate aminotransferase combined with Figure 4G. In (C–E), the gray lines indicate the associations. The blue arrowed lines indicate the food effects on serum metabolites mediated by specific genera. The green arrowed lines indicate the specific genera effects on physiological parameters mediated by serum metabolites. The beta coefficient and p values are labeled at each edge. The proportions of indirect effect (mediation effect) and mediation p values are labeled at the center of the ring charts.

Figure 6

Associations between aging‐related clinical indexes and age‐related gut microbiota, serum metabolites. A) The gut microbiota genera that significantly different in the youth (18–44 years), middle‐aged (45–59 years), and old (60–80 years) age people (FDR < 0.01) using the Envfit method with p values < 0.05. The checkers are colored according to phylum. B) The linear regression of the relationship between relative abundance of Bifidobacterium or Escherichia/Shigella of each individual and age (n = 3,224). C) The linear regression actual age and predicted age performed by RF analysis based on microbiota. D) The microbial interaction networks across the entire population at six‐year intervals constructed with FastSpar, FDR < 0.05, cor < mean‐ sd and cor > mean + sd. E) Pearson correlation between age and physiological parameters. The colors of dots indicate the value of cor. The size of dots indicates the value of q. F) The linear regression between the value of systolic blood pressure or low‐density lipoprotein cholesterol of each individual and age. G) The linear regression between the levels of serum 1‐palmitoyl‐GPG (16:0) or 1‐stearoyl‐2‐arachidonoyl‐GPC (18:0/20:4) of each individual and age. H) Spearman's correlation among age‐related gut microbiota, serum metabolites and physiological parameters (FDR < 0.05).