The experimental protocol was established according to the ethical guidelines of the Declaration of Helsinki and approved by the ethics committee of Qing Hai Provincial People's Hospital, Xining, China.

A cross-sectional study was designed to explore the association between high altitude residency and obesity with respect to lipid levels and gut microbiota diversity by comparing two indigenous communities who live at different altitudes. The study was approved by the Human Rights Committee of the University of Buenos Aires. Each parent provided written informed consent after the study was explained and before it was initiated.

We enrolled 70 subjects in our research study. All the participants were Tibetan children aged 7–12 years with the same eating habits. Using simple random sampling, we selected one primary school as a research site in each altitude area. Based on the China Center for Disease Control and Prevention (CDC) growth charts (27), we divided the subjects into four groups: obese children (N = 15, high-altitude obese children, HOB) and normal-weight children (N = 20, high-altitude normal-weight children, HN) who lived at an altitude of 4,500 m in Ma Duo County (34°52′12″N, 98°15′36″E), Guoluo Tibetan Autonomous Prefecture, Qinghai–Tibet Plateau; and obese children (N = 15, low-altitude obese children, LOB) and normal-weight children (N = 20, low-altitude normal-weight children, LN) who lived at an altitude of 1,500 m in Hai Dong County (36°18′N, 102°48′E), Hai Dong city, Qinghai–Tibet Plateau. A detailed questionnaire that included questions about age, sex, parental education level, history of breastfeeding, and family history of obesity was administered.

The inclusion criteria were as follows: (a) age 7–12 years old (28) and living in the research area for at least 5 years; (b) attendance at the selected boarding school; and (c) a diet containing grains, beef, and fresh vegetables. The exclusion criteria were as follows: (a) use of antibiotics or probiotics within the past 4 weeks; (b) gastrointestinal diseases (e.g., diarrhea) within the past 4 weeks; (c) frequent constipation; or (d) failure to agree to participate in this study. All the enrolled subjects were healthy, with no history of gastrointestinal disease, liver disease, hypertension, or diabetes, as demonstrated by their medical histories and physical examinations.

Height, weight, waist circumference and hip circumference were measured according to the protocol of the International Society for the Advancement of Kinanthropometry (29). Height was measured to the nearest 0.1 cm while the child stood upright against a mounted stadiometer with bare feet. In addition to height, waist and hip circumferences were measured. A digital weighing scale (Huawei3Pro, ShengZhen, China) that was calibrated regularly to the nearest 0.1 kg (after every 10 measurements) was used to measure body mass. We asked children to wear only a t-shirt and pants. Based on the height and body mass measurements, body mass index [BMI (weight/height²)] was calculated using the following formula:

Venous blood samples (5 ml) were collected by venepuncture into Vacutainer tubes, and plasma levels of glucose (GLU), triglycerides (TGs), total cholesterol (TC), high-density lipoprotein cholesterol (HDL-C) and low-density lipoprotein cholesterol (LDL-C) were analyzed with an LW-C400 automatic biochemical analyser (Landwind, Shenzhen, China).

Using sterile fecal boxes, fresh fecal samples (5 g) were collected from obese and control children within 2 h of defecation. Fecal samples were quickly placed in an ultralow temperature freezer (−80°C) for further processing and testing.

Fresh fecal samples (2 g) were mixed with 3 ml deionized water and homogenized. The mixture was left for 20 min at room temperature and centrifuged at 14,000 rpm for 15 min. Then, the supernatant was transferred to new EP tubes. Then, 3 ml deionized water was added to the precipitate, and all of the above operations were repeated again. The extraction was repeated twice, and the supernatants were combined. Next, it was filtered through a 0.22 μm filter. A 1 ml sample solution was injected into the GC system. Chromatographic analysis of fecal samples was performed using ISQ Single Quadrupole GC–MSMS (Thermo Fisher Scientific), Trace 1300 Series Gas Chromatograph (Thermo Fisher Scientific) and TG-5MS Separation column (30 m × 0.32 mm, 0.25 μm). The injection port temperature was 160°C. The GC oven was programmed with an increasing starting temperature from 50°C for 1 min to 220°C with 10°C/min for 5 min. A volume of 0.2 μl of the sample was injected. Electron impact ionization (EI) was used as an ionization source for the GC/MS analysis at 70 eV. The injection temperature was 230°C, and the transfer line and ion source were set to 250°C. XcaliburTM software (Thermo Fisher Scientific) was used for the automatisation of the GC–MS system and for data acquisition. Sample quantification was obtained by means of acetic, propionic, and butyric acid standard curves (Sinopharm). Construction of Standard Curve The standard calibration curves were constructed using seven concentrations: 9.325, 18.75, 37.5, 75, 150, 300, and 900 μg/ml. The concentration of each sample was obtained by the standard curve. The standard curves of acetic acid, propionic acid, and butyric acid were Y = 3.796e5X + 9.363e4, Y = 3.936e5X + 4.003e5, and Y = 1.366e6X + 1.151e6, respectively (30, 31).

Total genomic DNA was extracted from the samples using the acetyltrimethylammonium bromide (CTAB) method (28) DNA concentration and purity were analyzed using 1% agarose gels. According to the concentration, the DNA samples were diluted to 1 ng/μl using sterile water. An equal volume of 1 × loading buffer (containing SYBR green) was mixed with the PCR products, and electrophoresis was performed on a 2% agarose gel to visualize the PCR products. PCR products were mixed in equidensity ratios. Then, the PCR products were purified with a Qiagen Gel Extraction Kit (Qiagen, Germany). The DNA purity and concentration were analyzed by measuring the optical density (OD) at wavelengths of 260 and 280 nm with a NanoPhotometer^® spectrophotometer (Implen, Munich, Germany) and then calculating the OD260:OD280 ratio. The DNA concentrations were measured with the Qubit^® dsDNA Assay Kit in a Qubit^® 2.0 Fluorometer (Life Technologies, Camarillo, CA, United States).

Based on many previous studies (32–34), we selected the V3–V4 region to study the microbiome through second-generation sequencing. 16S rRNA gene sequencing was performed by Novogene Bioinformatics Technology Co., Ltd., China. In brief, DNA samples were diluted to a concentration of 1 ng/μl in sterile water and then PCR amplified with the 515F/806R primer set (515F: 5′-GTGCCAGCMGCCGCGGTAA-3′, 806R: 5′-XXXXXXGG ACTACHVGGGTATCTAAT-3′). Sequencing libraries were generated using the TruSeq^® DNA PCR-Free Sample Preparation Kit (Illumina, USA) following the manufacturer's recommendations, and index codes were added. Library quality was assessed with a Qubit@ 2.0 Fluorometer (Thermo Scientific) and Agilent Bioanalyzer 2100 system. Finally, the library was sequenced on the Illumina NovaSeq platform, and 250-bp paired-end reads were generated.

Paired-end reads were assigned to samples based on their unique barcodes and were truncated by removing the barcode and primer sequences. Paired-end reads were merged using FLASH (V1.2.7) (35), which is a very fast and accurate analysis tool that was designed to merge paired-end reads when at least some of the reads overlapped with the read generated at the opposite end of the same DNA fragment. The splicing sequences are called raw tags. Quality filtering of raw tags was performed under specific filtering conditions to obtain high-quality clean tags (36) according to the QIIME (V1.9.1) (37) quality control process. The tags were compared with those in the reference database (Silva138 database) using the UCHIME (28) algorithm to detect chimeric sequences, and the chimeric sequences were removed (38). Then, effective tags were finally obtained.

The data were analyzed and compared using SPSS version 27.0 (Chicago, Illinois, USA) and R software (v 2.15.3). Data are summarized as the mean (standard deviation, SD)/median (interquartile range, IQR) for continuous variables depending on normality. Student's t test was used to analyse significant differences in age, sex, height, weight, waist circumference, hip circumference, BMI and lipid levels among the schoolchildren who lived at different altitudes. Analysis of variance (ANOVA) was used to compare the content of short-chain fatty acids (SCFAs) in feces among the four groups. Significant differences in categorical variables are expressed as numbers and percentages. The chi-square test was used to analyse count data. The Kruskal–Wallis test was used to investigate significant differences in operational taxonomic units (OTUs) and the abundance-based coverage estimator (ACE), Chao 1, Simpson's, and Shannon's indexes among the four groups. To correct for type I errors, we applied the Bonferroni method for multiple comparisons between two groups. Differences in microbial community abundances between the obesity group and the control group were analyzed using the Wilcoxon rank sum test, and the significance of these differences was assessed using the false discovery rate (FDR). Principal coordinate analysis (PCoA) was performed using the WGCNA package, stat packages and ggplot2 package in R software (v 2.15.3) to compare the similarity of community structures. Multiresponse permutation procedures (MRPPs) (39) were used to analyse differences in the microbial community structure between groups. Linear discriminant analysis effect size (LEfSe) was performed to identify particular taxa with significantly different abundances between the two groups. An unweighted pair-group method with arithmetic means (UPGMA) clustering was performed as a type of hierarchical clustering method to interpret the distance matrix using average linkage by QIIME (2 v2022.2) software. Correlations between blood lipid levels and intestinal flora diversity were analyzed by Spearman's correlation analysis. p ≤ 0.05 indicated statistical significance.

A total of 70 subjects were included in the high-altitude groups (n = 35, normal/obese = 20/15) and low-altitude groups (n = 35, normal/obese = 20/15; Table 1). There was no significant difference in age, sex or height (p > 0.05; Table 1). Body weight, BMI, waist circumference and hip circumference were significantly different between the obesity group and the control group from the same altitude (p < 0.05; Table 1). There was also a significant difference between the obese groups from different altitudes (p < 0.05, Table 1). TG levels were significantly different between the obesity group and the control group from different altitudes (p < 0.05, Table 1), and LDL-C levels were significantly different between the obese groups from different altitudes (p < 0.05, Table 1).

A total of 6,130,974 high-quality and classifiable raw reads were obtained from 70 samples. We retained 5,991,598 clean reads after removing the low-quality sequences. A total of 3,109,972 and 2,881,626 clean reads were obtained for the high-altitude groups and the low-altitude groups, respectively. Clustering analysis assigned the microbial sequences to the same OTU if the samples had at least 97% similarity. We obtained 3,843 OTUs by high-throughput sequencing. These reads were classified into 43 phyla, 101 classes, 226 orders, 331 families, 576 genera, and 340 species. The rarefaction curves showed that our sequencing depth was sufficient (Supplementary Figure S1).

α-diversity analysis was performed using four indexes, namely, OTUs and the Shannon index, Simpson index, Chao 1 index, and ACE index. There were highly significant differences in all four indexes among the four groups according to the Kruskal–Wallis test (p < 0.05; Table 2). The Shannon and Simpson indexes were significantly different in the different BMI groups from the same altitude according to the Bonferroni multiple comparisons test (p < 0.05; Table 2). The OTUs and Chao 1 and ACE indexes were significantly different between groups with the same BMI who were from different altitudes according to the Bonferroni multiple comparisons test (p < 0.05; Table 2).

β-diversity analysis was used to evaluate differences in species composition among the samples from different groups. Fecal microbial OTU data were examined by principal coordinate analysis (PCoA) to evaluate similarities among the four groups. Multiresponse permutation procedures (MRPPs) were used to analyse differences in microbial community structures between groups. The results showed that the groups from different altitudes had different microbiota compositions. The LN and LOB groups had similar microbiota compositions, and the sequences could be grouped into the same clusters, whereas the HN and HOB groups did not have similar microbiota compositions, and the sequences could not be grouped into the same clusters. In addition, the results revealed obvious differences in the community compositions and structures in the groups from different altitudes (Table 3 and Figure 1).

In this study, we found 484 common OTUs among the four groups, and there were 357, 164, 266, and 313 unique OTUs in the LN, LOB, HN, and HOB groups, respectively. The relative abundance of the microbiota was lowest in the LOB group (Figure 2). Further analysis of the microbiota at the phylum, class, genus and species levels in the four groups showed that five of the top 10 most dominant bacteria at any level were the same among the four groups. The microbiota compositions in all four groups were dominated by the following phyla (order: LN, LOB, HN, HOB): Firmicutes (61.2, 60.1, 60.6, and 53.2%, respectively); Bacteroidota (22.2, 15.4, 28.0, and 30.4%, respectively); Actinobacteriota (11.3, 15.7, 8.5, and 9.9%, respectively); Proteobacteria (3.0, 7.2, 0.9, and 4.2%, respectively); and Verrucomicrobiota (0.3, 0.05, 0.7, and 0.2%, respectively; Figure 3). At the phylum level, the relative abundance of Firmicutes (F) was higher than that of Bacteroides (B) in the obese groups. Bacteroides abundances and F/B ratios were significantly different among the four groups (Supplementary Table S1). At the phylum level, the abundances of Bacteroidota, Proteobacteria, and Verrucomicrobiota were significantly different among the four groups (Table 4). At the class level, the abundances of Clostridia, Negativicutes, Gammaproteobacteria, Bacteroidia, Actinobacteria, and Verrucomicrobiae were significantly different among the four groups (Table 4). At the genus level, the abundances of Megamonas, Prevotella, Akkermansia, and Bifidobacterium were significantly different among the four groups (Table 4). At the species level, the abundances of Romboutsia, Bacteroides caccae, and Escherichia coli were significantly different among the four groups (Table 4).

We further analyzed differences in the microbiota composition between the two different groups at the genus level using the Metastat analysis method. First, between groups from the same altitude, there were significant differences in the abundances of Akkermansia, Streptococcus, Escherichia-Shigella, Megamonas and Sarcina between the LN and LOB groups (p < 0.05; Figure 4A), and there were significant differences in the abundances of Akkermansia, Ruminococcus torques, Holdemanella, Barnesiella, Fusicatenibacter, Megasphaera, Subdoligranulum, Blautia, Eubacterium hallii, Alistiples, Parabacteroides, Ruminococcus, Escherichia-Shigella and Megamonas between the HN and HOB groups (p < 0.05; Figure 4B). Second, between groups with the same BMI, the abundances of Eubacterium hallii, Clostridium sensu stricto, Akkermansia, Ruminococcus torques, Streptococcus, Escherichia-Shigella, Holdemanella, Klebsiella, Romboutsia, Lachnoclostridium, Parabacteroides, Citrobacter, Megamonas, Sarcina, Catenibacterium, Alloprevotella, Succinlvibrlo and Megasphaera were significantly different between the HN and LN groups (p < 0.05; Figure 4C), and the abundances of Blautia, Eubacterium hallii, Holdemanella, Klebsiella, Romboutsia, Fusicatenibacter, Lachnoclostridium, Ruminococcus, Streptococcus, Escherichia-Shigella, Sarcina, Methanobrevibacter, Catenibacterium, Prevotella, Succinlvibrlo, Megasphaera, Akkermansia, Paraprevotella and Alloprevotella (p < 0.01) were significantly different between the HOB and LOB groups (p < 0.05; Figure 4D). In conclusion, we found that Akkermansia, Holdemanella, Streptococcus, Prevotella, Escherichia-Shigella, and Megamonas were strongly associated with BMI, and these six members of the microbiota plus Sarcina and Catenibacterium were associated with altitude.

LEfSe was performed to identify particular taxa with significantly different abundances between groups. The results of the UPGMA analysis showed that there were differences in the abundances of the genera Bifidobacterium, Escherichia-Shigella, Blautia, Romboutsia, Streptococcaceae, Megasphaera, and Prevotella in obese children living at different altitudes. Bifidobacterium, Escherichia-Shigella, Blautia, Romboutsia, and Streptococcaceae were highly abundant in children living at low altitudes, while Megasphaera and Prevotella were abundant in children living at high altitudes. These results indicated that these microbes were closely related to altitude when obesity conditions were the same. We further analyzed intestinal differences in normal-weight children living at different altitudes. Prevotella and Holdemanella were highly abundant at high altitudes. It was interesting to further confirm that Prevotella was positively correlated with high altitude (Figure 5).

A Spearman's correlation matrix was generated to analyse correlations between BMI, TC, TG, HDL-C, and LDL-C levels and the dominant genera in the gut microbiota in all children. Significant relationships were observed between the gut microbiota composition and BMI, TC, TG, and LDL-C levels. Correlation analysis revealed that the abundances of the genera Methanobrevibac, Sarcina, Succinivibrio, Catenibacterium, Alloprevotella, Lactobacillus, Megasphaera, Holdemanella and Prevotella were positively correlated with BMI (p < 0.05; Figure 6), and the abundances of Citrobacter, Klebsiella, Lachnoclostridum, Clostridium-sens, X. Eubacterium.-h, Fusicatenibacter, Blautia, Streptococcus, Escherichia. Shige, Romboutsia, Bifidobacterium and Megamonas were negatively correlated with BMI (p < 0.05; Figure 6). The abundance of Megasphaera was positively correlated with the TC level (p < 0.05; Figure 6), and the abundances of Clostridium and Paraprevotella were negatively correlated with the TC level (p < 0.05; Figure 6). The abundances of Sarcina, Megasphaera and Holdemanella were positively correlated with the TG level (p < 0.05; Figure 6), and the abundances of Fusicatenibacter and Romboutsia were negatively correlated with the TG level (p < 0.05; Figure 6). The abundances of Citrobacter, Klebsiella, Subdoligranulum and Romboutsia were positively correlated with the HDL-C level (p < 0.05; Figure 6), and the abundances of Sarcina, Scuccinivibrio, Catenibacterium, Holdemanella and Prevotella were negatively correlated with the HDL-C level (p < 0.05; Figure 6). The abundances of Succinivibrio and Megasphaera were positively correlated with the LDL-C level (p < 0.05; Figure 6), and the abundances of Clostridium_sens, X. eubacterium._h, Akkermansia, Paraprevotella, Alistipes, Blautia, and Romboutsia were negatively correlated with the LDL-C level (p < 0.05; Figure 6). In addition, the difference in the abundance of Prevotella was positively correlated with the difference in the LDL level between the obese groups from different altitudes (Figure 6).

To further explore the function of gut microbes in children at different altitudes, Tax4Fun was used to analyse the functional prediction of the fecal microbiota in different groups. Figure 7 shows the 35 most abundant pathways at the third level of KEGG pathways. Of these 35 pathways, 10 pathways (amino sugar and nucleotide sugar metabolism; quorum sensing; transcription factors; ABC transporters; transporters; two component system; starch and sucrose metabolism; galactose metabolism; secretion system, bacterial motility proteins.) were enriched in the fecal microbial community of the LOB groups. Eighteen pathways (ribosome biogenesis; prokaryotic defense system; aminoacyl tRNA biosynthesis; transfer RNA biogenesis; DNA replication proteins; purine metabolism; glycine, serine and threonine metabolism; carbon fixation pathways in prokaryotes; pyrimidine metabolism; amino acid-related enzymes; ribosome; chromosome and associated proteins; DNA repair and recombination proteins; mitochondrial biogenesis; exosome; chaperones and folding catalysts; alanine, aspartate and glutamate metabolism; peptidases.) were enriched in the HOB groups, and three pathways (glycolysis/gluconeogenesis, butanoate metabolism and pyruvate metabolism) were enriched in the HN groups. The KEGG pathways for predicting the function of gut microbiota were mainly expressed in four aspects: metabolism, genetic information processing, environmental information processing and cellular processes. The differences in these pathways may be related to the influence of altitude on the abundance of intestinal flora. In addition, the prominence of the butyric acid metabolic pathway in normal weight children at high altitude has attracted our attention.

Through KEGG pathway annotation and enrichment analysis, the results showed that the butyric metabolism pathway was significantly different in the four groups, with the highest abundance in the HN group, followed by the LN group, and the lowest abundance in the LOB group (Figure 7). Combined with this feature, we measured the short-chain fatty acids in the feces of different groups, and the results showed that acetic, propionic and butyric acid levels were significantly different among the four groups (Table 5). We further carried out a correlation analysis between blood lipids and the content of short-chain fatty acids, and the results showed that acetic acid was correlated with LDL, while butyric acid was correlated with TC, TG, and LDL (Table 6).