Hematological and biochemical assessment
In this study, we examined the hematological and other related determinants between the hypoxic (T group) and control (C group) of rats at the end of the hypoxic exposure, i.e., on day 28. We found that the hemoglobin and the red blood cell value of the T-group of rats housed at 6000 meters, increased significantly as compared to the C-group of rats (Fig. 1). We also evaluated the levels of Glutathione (GSH) and malondialdehyde (MDA) levels in both groups of rats. We observed a significant variation between the levels of GSH and MDA in both groups. We found a lower antioxidant capacity and a higher lipid membrane peroxidation capacity in the T-group of rats as compared to the C-group of rats, which demonstrated the effectiveness of the 28-days hypoxic exposure (Fig. 1). Apart from this, the pathological manifestations of high-altitude polycythemia (HAPC) were surfaced in the T-group of rats.
Metabolomics Results
Metabolic profiling of the rat feces
We performed a non-targeted metabolomics analysis of fecal samples from the T-group of rats. Firstly, we evaluated the stability and reproducibility of the current data by measuring the Quality Control (QC) samples throughout the experiment. The peak area value thus obtained was used to analyze the Pearson correlation coefficient between the QC samples. A higher correlation value of the QC samples indicated better stability of the whole detection process, and the higher the data quality. The results are presented in Supplementary Fig. 1 and demonstrate excellent stability, higher reproducibility, and data quality in our current metabolomics data set.
A quality control check qualified a total of 4,360 peaks in positive ion modes and 4,973 peaks in negative ion modes from the stool samples of rats, which were further annotated with the help of ChemSpider and mzCloud databases comparison (Supplementary Table 1).
Metabolic Findings In Rat Feces
We employed the partial least squares discriminant analysis (PLS-DA)[11] to establish a model of the relationships between the metabolite expression and sample grouping for sample size prediction (Supplementary Fig. 2). Subsequently, we sorted the model to verify if the model was “over-fitting.” The results demonstrated that in the PLS-DA model of each group, R2Y was close to 1, which indicated that each group could explain the grouping well. At the same time, based on the results of the sort verification (Supplementary Fig. 3), we can say that the grouping model was not “over-fitting.” Thus, we can conclude that the experimental results are reliable, and the data set can be subsequently analyzed.
Based on this analysis, we used the Variable Importance in the Projection (VIP) value of the first principal component of the PLS-DA model to represent the contribution rates of metabolite differences in different groups. The difference multiple (Fold Change, FC) was the ratio of the mean of all biological replicate quantitative values for each metabolite in the two groups, and the p-value of the t-test was used to find the differentially expressed metabolites (Table 1).
Compared Samples | Num. of Total Ident. | Num. of Total Sig. | Num. of Sig.Up | Num. of Sig.down |
DAY 28_pos | 4360 | 101 | 41 | 60 |
DAY 28_neg | 4973 | 130 | 25 | 105 |
Table.1 The number of differential metabolites between the experimental group and the control group on day 28. We have set the threshold value to VIP > 1.0, FC > 2.0 or FC < 0.5 and p-value < 0.05, and screened the differential metabolites between the experimental group and the control group.
We performed the hierarchical cluster on each group of differential metabolites. And it was followed by pathway analysis where exact mass and mass spectrometry fragment patterns were searched in the KEGG database. Then a hypergeometric test was applied, as shown in the supplementary Fig. 4, to find pathways enriched in the differential metabolites compared to all identified metabolite backgrounds(With P-value < 0.05 as the threshold). The differential metabolites identified in our study were mapped to 4 metabolic pathways of the KEGG database. They were pyrimidine metabolism pathway, aminoacyl-tRNA biosynthesis pathway, nicotinate and nicotinamide metabolism pathway, and thyroid hormone synthesis pathway. The metabolites in these pathways were found to be significantly different in T group compared with C group (Table 2).
Pathways | Metabolites | VIP | FC | P.value | Up.Down |
Pyrimidine metabolism | Uridine | 2.78 | 0.35 | 5.79E-03 | down |
Pseudouridine | 2.75 | 0.36 | 7.67E-03 | down |
Thymidine* | 3.31 | 0.20 | 2.30E-02 | down |
4.22 | 0.11 | 4.15E-02 | down |
Cytidine | 1.88 | 0.47 | 6.93E-03 | down |
Aminoacyl-tRNA biosynthesis | L-Aspartic acid | 2.07 | 2.28 | 5.95E-04 | up |
L-Tyrosine | 2.04 | 0.39 | 4.72E-02 | down |
Nicotinate and nicotinamide metabolism | L-Aspartic acid | 2.07 | 2.28 | 5.95E-04 | up |
Maleic acid | 1.97 | 2.02 | 3.39E-02 | up |
Thyroid hormone synthesis | Glutathione disulfide | 1.84 | 2.00 | 7.86E-03 | up |
Table.2 Major differential metabolic markers and the enriched pathways on day 28. *The expression of thymidine was different in both positive and negative ion modes
Microbial Community Analysis
We used Illumina HiSeq sequencing platform for sequencing, and obtained 112,695.48 Mbp of Raw Data. After single sample assembly and mixed assembly, a total of 2,231,469,534 BP scaftigs were obtained. We got 2,746,053 open reading frames (ORFs) by using MetaGeneMark. Then, the blastp algorithm was used to compare with the MicroNR Library, and annotated species with LCA algorithm. The proportions of Genera and Phyla were 52.90% and 78.65%, respectively. The core-pan genetic analysis (Supplementary Fig. 5) denoted that samples tended to saturate the platform, which indicates that the sequence coverage was sufficient to capture the diversity of bacterial communities in the sample.
Beta diversity of gut microbiota between the two groups with multivariate statistics analysis
We used Non-Metric Multi-Dimensional Scaling (NMDS) as a simple method of visual interpretations to compare the overall structure of fecal microbiota between two samples (Fig. 2A). NMDS was performed using the Bray-Curtis similarity index based on the ORFs. Moreover, we also described the use of analysis of similarity (Anosim) to statistically test the significant difference between groups (Fig. 2B). Analysis ofsimilarity (ANOSIM) revealed that significant different were observed between T group and C group in microbiota community structure (R = 0.344; P = 0.006).
Fig 2. Non-Metric Multi-Dimensional Scaling (NMDS) plot (A) and ANOSIM analysis (B). NMDS is a simple method for visual interpretations to compare the overall structure of fecal microbiota between two samples while ANOSIM is used to statistically test the significant difference between groups.
Fecal Microbiota Composition Between Two Groups
Bacterial communities with inter-group differences were analyzed at the genus level. Metastat analysis showed that the relative abundance of genus Lactobacillus, Alistipes in the experimental group were significantly increased than that in the control group, while genus Flavonifractor, Faecalibacterium and Dorea were decreased.
Table 3
Metastat analysis between two groups at genus levels. The mean of group indicates the relative abundance of this genus, the screening threshold are P value less than 0.05 and relative abundance greater than 0.1%.
Taxa | Mean(T group) | Mean(C group) | P value |
Genus Lactobacillus | 2.24% | 0.97% | 1.07E−02 |
Genus Alistipes | 1.20% | 0.58% | 8.52E−04 |
Genus Flavonifractor | 0.23% | 0.55% | 1.68E−02 |
Genus Faecalibacterium | 0.14% | 0.46% | 4.06E−03 |
Genus Dorea | 0.11% | 0.30% | 3.47E−02 |
Association of the intestinal microbial species with the metabolites.
In order to explore the correlations between differential genera and differential metabolites. We used pearson statistical method to calculate the correlation coefficients rho and P values (p༜0.05) between the relative abundance of each genus and metabolites. The results are shown in Fig. 3. Genus Alistipes has significant positive correlations with glutathione disulfide, l-aspartic acid, and maleic, and negative correlations with uridine, pseudouridine, cytidine and l-tyrosine; Genus Dorea has significant positive correlations with metabolites thymidine and l-tyrosine; Genus Faecalibacterium has a significant positive correlation with pseudouridine, and negative with glutathione disulfide; Genus Flavonifractor has a significant negative correlation with the metabolite l-aspartic acid; Genus Lactobacillus has no significant correlation with these differential metabolites.