Meta-analysis identifying gut microbial biomarkers of Qinghai-Tibet Plateau populations and the functionality of microbiota-derived butyrate in high-altitude adaptation

Figures & data

Figure 1. Sample integration and analysis process.

(a) The geographic distribution of samples from six studies and information on the number of samples from each region is shown. (b) The proportion of sample sequencing regions, proportion of groups, proportion of different independent institutes, and proportion of samples in each region, respectively. (c) Workflow: ① Gather 668 16S rRNA samples from databases, categorized into LH, HH, LT, and HT groups. ② Process the raw data using software such as Qiime and vesearch, followed by conducting differential analysis using R packages. ③ Identify gut microbiota biomarkers in high and low-altitude populations using p-value, caret, and Random Forest methods. ④ Validate the gut microbiota through LODO, study-to-study transfer validation, and new cohorts. ⑤ Utilize Mendelian randomization to infer causal relationships between gut microbiota and high-altitude adaptation. ⑥ Predict and quantify the butyrate production capacity of gut microbiota in high and low-altitude populations. ⑦ Validate the impact of butyrate using a high-altitude rat model. ⑧ Investigate the mechanism of action of butyrate through cell experiments

Figure 2. Differences in gut microbiota between high- and low-altitude people.

(a) Richness index and evenness index scatter plot after filtering some OTUs and ConQuR processing (LH, n = 243; HH, n = 105; LT, n = 59; HT, n = 240). The box plot on the top indicates the observed index results of different groups, while the box plot on the right indicates the evenness index results of samples of other groups. The table on the top right shows the difference between groups in observed and evenness index compared with LH. p 1 was from the Wilcoxon Rank sum test. p 2 was from the blocked Wilcoxon Rank sum test. (b) PCoA analysis of samples after filtering some OTUs and ConQuR processing (LH, n = 243; HH, n = 105; LT, n = 59; HT, n = 240). Beta diversity was based on Bray-Curtis Dissimilarity. p value of the PCoA was from PERMANOVA (999 permutations). The p value of the boxplot was from the Kruskal test. (c) The relative abundance and variation trend of phyla-level species in 6 study cohorts. Draw the picture using unprocessed data. The image is drawn through the microbiome database (http://egcloud.cib.cn/). (d, e) The relative abundance in genera level of HH, HT, and LT were significantly increased (d) or decreased (e) compared to LH. FDR-corrected p value was from the Wilcoxon Rank sum test.

Figure 3. Identification of gut microbiota biomarkers in plateau population.

(a – c) The optimal number of gut microbiota biomarkers for LH_vs_HH (a), LH_vs_HT (b), and LH_vs_LT (c) were determined by random forest. (d – f) The performance of RF, GLM, and SVM classification models constructed by LH_vs_HH (d), LH_vs_HT (e), and LH_vs_LT (f) using the gut microbiota biomarkers, respectively. Blue, red, orange, yellow, purple, and green represent AUC, Accuracy, Sensitivity, Specificity, Precision, and F1 Score, respectively.

Figure 4. Validation of predictive performance of gut microbiota biomarkers in plateau population.

(a,b) Cross-validation matrices of LH_vs_HH (a) and LH_vs_HT (b), respectively. The model was constructed using the RF method with 10-fold cross-validation. The value represents the AUC, and the color depth was positively correlated with the AUC. The top squares represent the result of study-to-study transfer validation, and the bottom squares represent the result of LODO validation. The training and test samples are marked on the left and top, respectively. All samples except for test samples were used as training samples in LODO validation. (c,d) AUC value of LH vs HH (c) and LH vs HT (d) LODO validation classifiers with different characteristics. Circles, triangles, squares and diamonds represent top-ranking biomarkers, all biomarkers, all significantly different OTUs and all OTUs, respectively. The color of the line represents the test set sample. The Wilcoxon Rank sum test calculated the significant difference in OTUs (FDR-corrected p value < 0.05). (e) The gut microbiota co-biomarkers of high-altitude population. Fold change is equal to the ratio of the average relative abundance of OTU in HH or HT to the average relative abundance of OTU corresponding to LH. The length of the bar graph is log2 (fold change). The blue bar indicates a significant difference between LH and LT (FDR-corrected p value < 0.05), and the red indicates a non-significant difference between LH and LT. (f) ROC curves of LH_vs_HA(High Altitude[HH and HT]). 20% of the total samples were randomly selected as the test set, and the remaining samples were used as the training set, repeated ten times. Only the optimal ROC curve and average AUC values are shown here. (g) A set of newly collected samples (LH, n = 26; HA, n = 40) was used as a test set to verify the model constructed with LH and HA samples as the training set.

Figure 5. The causal relationship between gut microbiome and altitude adaptation.

Mendelian randomization results of causal effects between the gut microbiome and polycythemia-associated phenotypes

Figure 6. Changes in the abundance of gut microbiome SCFA-producing bacteria in high- and low-altitude population.

(a) Prediction of the abundance of butyrate-producing bacteria by acetyl-CoA pathway. (b) Prediction of the bacterial abundance of butyryl-CoA: acetate CoA transferase. (c) Prediction of the bacterial abundance of Butyrate kinase. (d) Prediction of the abundance of propionate-producing bacteria by succinate pathway. (e) Prediction of the abundance of propionate-producing bacteria by propanediol pathway. (f) Prediction of the abundance of total propionate-producing bacteria. (c) – (f) FDR-corrected p value was from the Wilcoxon Rank sum test. (g,h) qPCR results. The difference between the abundance of BCoAT in LH (n = 5) and HA (n = 12) is shown in (g) BCoAT primers designed according to R. intestinalis. (h) BCoAT primers designed according to F.prausnizii. The 2^−ΔΔCT denotes the abundance. p values were obtained from a one-sided t-test.

Figure 7. SCFA relieved intestinal damage caused by the plateau environment.

(a) HE staining of duodenum, jejunum, and ileum of the CON, HPBS, NaPr, and NaBu groups (n = 5, bar = 50 μm). (b–d) The villi length of the duodenum, jejunum, and ileum, respectively. (e) Immunohistochemical staining of occludin in duodenal tissue sections of CON, HPBS, NaPr, and NaBu groups (bar = 50 μm). (f) Mean IOD of occludin in the duodenum. (g,i) Immunohistochemical staining of ZO-1(g) and occludin(i) in colon tissue sections of CON, HPBS, NaPr, and NaBu groups (bar = 20 μm). (h,j) Mean IOD of ZO-1(h), occluding(j) in the colon. Con/C: control group; HPBS/HP: high-altitude PBS group; NaBu/Bu: high-altitude butyrate group; NaPr/Pr: high-altitude propionate group. The results are expressed as Mean±SE (n = 5). “*” stands for significant difference. *p < 0.05. **p < 0.01. ***p < 0.001. ****p < 0.0001. The p value is calculated by t-test.

Figure 8. Butyrate alleviates intestinal damage caused by hypoxia by down-regulating HIF-1α in NCM460.

(a) Western blot of ZO1, HIF-1α and occludin. Alpha-tubulin is the internal reference protein. (b) Densitometry quantification of occludin level by Western blot (Con: n = 4; CoCl₂: n = 4; NaB: n = 4). (c) Densitometry quantification of ZO-1 level by Western blot (Con: n = 4; CoCl₂: n = 4; NaB: n = 4). (d) Densitometry quantification of HIF-1α level by Western blot (Con: n = 4; CoCl₂: n = 4; NaB: n = 4). (e) mRNA expression of VEGF (Con: n = 3; CoCl₂: n = 3; NaB: n = 3). (f) mRNA expression of BNIP3 (Con: n = 3; CoCl₂: n = 3; NaB: n = 3). (g) mRNA expression of HK2 (Con: n = 3; CoCl₂: n = 3; NaB: n = 3). (h) mRNA expression of PDK1 (Con: n = 3; CoCl₂: n = 3; NaB: n = 3). (i) mRNA expression of LDHA (Con: n = 3; CoCl₂: n = 3; NaB: n = 3). (j) LDHA relative activity (Con: n = 4; CoCl₂: n = 4; NaB: n = 4). (k) Western blot of HIF-1α. Alpha-tubulin is the internal reference protein. (l) Densitometry quantification of HIF-1α level by Western blot (Con: n = 4; CoCl₂: n = 4; lactate: n = 4). “*” stands for significant difference. *p < 0.05. **p < 0.01. ***p < 0.001. ****p < 0.0001. The p value is calculated by t-test.

Figure 9. The potential mechanism of gut microbiota biomarkers to help maintain human health at the plateau.

The high-altitude hypoxic environment can be detrimental to the body. On the one hand, reduced oxygen levels lead to high-altitude polycythemia, but specific gut microbiota biomarkers can dull the erythrocyte production response, aiding adaptation to the high-altitude environment. On the other hand, the high-altitude hypoxic environment can lead to intestinal and intestinal barrier damage and gut microbiota imbalance. Gut microbiota biomarkers in people living in plateau regions protect against intestinal damage caused by the harsh environment. Butyrate produced by gut microbiota biomarkers may reduce lactic acid build up in the intestinal tract and decrease the overexpression of HIF-1α, which is caused by hypoxia at high altitudes, by inhibiting LDHA activity. Inhibiting the downstream target gene of HIF-1 can help the host better adapt to the plateau environment, maintain the stability of the intestinal barrier, and reduce intestinal damage. RBC: Erythrocyte count; Hb: Hemoglobin concentration; HCT: Hematocrit. The figure is drawn using Figdraw. (https://www.figdraw.com/).