This cross-sectional study was conducted between October 31, 2022, and October 30, 2023, at the Hospital of the Chengdu Office of the People’s Government of the Tibetan Autonomous Region (Hospital C.T.). The protocol was approved by the Ethics Committee of Hospital C.T. (Approval No.: [(2024) No.41 Scientific Research] postpone). All participants provided written informed consent prior to enrollment.

Fasting venous blood (10 mL) and urine samples were collected in the morning. Blood samples were processed within 2 hours of collection via standardized centrifugation (3,000g, 10 min, 4°C). Hemolyzed specimens (free hemoglobin >0.2 g/L, detected spectrophotometrically at 414 nm) were excluded. Aliquots of plasma and urine were stored at -80°C in the Hospital C.T. Biobank for metabolomics analysis. Residual samples were archived for future studies on diabetes complications (e.g., metabolite, protein, or gene sequencing assays), with de-identification via unique study IDs to protect privacy.

Clinical data: Demographics, medical history, and medication use were collected via EHR review and standardized questionnaires.

Follow-up: Endocrinologists and diabetes nurses conducted regular phone follow-ups to monitor blood glucose, blood pressure, and medication adherence, providing online interpretation of test reports and management guidance.

Confidentiality: Personal identifiers were replaced with codes; data access was restricted to the research team, with only de-identified data used for analysis and publication.

IBM SPSS for Windows (version 22.0, Armonk, NY, USA) was used for statistical analysis and receiver operating characteristic (ROC) curve generation. All collected data are expressed as the mean ± standard deviation (SD). Differences between H_T2DM patients and L_T2DM patients were evaluated with Student’s t test followed by the Kolmogorov–Smirnov test. A P value less than 0.05 was considered to indicate statistical significance.

In this study, we compared the clinical data of 100 H_T2DM patients and 100 L_T2DM patients. There were no significant differences between H_T2DM patients and L_T2DM patients in terms of sex distribution, body mass index (BMI), height, sex. However, the body weight of the H_T2DM group was significantly higher than that of the L_T2DM group (P = 0.040). This might be related to the high intake of animal fat, red meat and refined carbohydrates in the high-altitude area, combined with the lack of sufficient oxygen in the high-altitude region, which makes it impossible to engage in extensive physical activities. The average age in the H_T2DM group was younger than that in the L_T2DM group (P < 0.001), which reflects a significant change in the diabetes epidemic trend in the plateau region. This may be related to the dietary structure shift: the traditional high-fiber, low-fat diet (such as foods made from barley) is being replaced by modern diets rich in saturated fats, refined carbohydrates, and sugary beverages, which accelerates insulin resistance in the younger population. It may also be due to the differences in the spread of health education, where in low-altitude urban centers, young people pay more attention to health preservation, exercise, and can more conveniently access regular health check-ups, enabling early intervention and treatment. In contrast, the health education coverage rate is low in high-altitude areas, compounded by the altitude factor, limited aerobic exercise, and limited medical resources, making it impossible to receive timely diagnosis and treatment. This results in a lower average age of patients in the H_T2DM cohort. The results of the biochemical tests revealed that alanine aminotransferase (ALT) (P < 0.001), aspartate aminotransferase (AST) (P = 0.001) and hemoglobin A1 (HbA1c) (P = 0.014) levels were significantly greater in the H_T2DM group than in the L_T2DM group. High-density lipoprotein cholesterol (HDL-C) levels were significantly lower in the H_T2DM group than in the L_T2DM group (P = 0.010). To exclude potential confounding effects, we performed subgroup analyses based on smoking status using one-way ANOVA. The results revealed no significant differences in HDL-C levels across different smoking categories within either population (P = 0.088), confirming that smoking habits do not contribute to the observed HDL-C discrepancy between the two groups. Instead, this difference may be primarily linked to the distinct dietary patterns of high-altitude populations, who predominantly consume animal oil and meat. Epidemiological evidence indicates that such a diet, rich in saturated fat, can impair reverse cholesterol transport and downregulate hepatic HDL synthesis, thereby reducing HDL-C levels and altering lipid metabolism in plateau residents. Later, we will further study the differences in lipid metabolism between the plateau population and the plain population. In addition, there was no significant difference in serum total cholesterol (TC, p = 0.205), triglycerides (TG, p = 0.13) and blood glucose AUC (p = 0.071) between high- and low-altitude groups. Although statistically non-significant, the observed trend suggests that high-altitude T2DM patients may experience a higher overall 24-hour glycemic load. Suggesting that dietary factors related to lipid and antioxidant intake may not substantially contribute to the observed metabolite variations (Table 1). Thus, in our study population, the high-altitude environment did not have an effect on lowering blood glucose.

Reports have shown that the prevalence of diabetes at high altitudes is lower than that at low altitudes. However, for patients who already have diabetes, H_T2DM patients tended to be younger, have higher HbA1c levels, and even have more severe symptoms of diabetes than L_T2DM patients.

Therefore, we believe that a high-altitude environment affects the metabolic processes of diabetic patients resulting in higher HbA1c levels than in low-altitude diabetic patients. We further analyzed the metabolism of patients with high-altitude diabetes and low-altitude diabetes.

We used LC–MS/MS to profile serum samples from 100 H_T2DM, 100 H_HC, 100 L_T2DM, and 100 L_HC individuals using the criteria of FC > 1.2 or <0.833, P < 0.05 and VIP > 1. By comparing and analyzing the metabolomes of H_T2DM and H-HC (H_T2DM.vs.H_HC) we obtained differentially abundant metabolites (DMs) related to high-altitude diabetes. The results showed that 412 DMs (32 upregulated and 380 downregulated) related to H_T2DM (Figure 2A). This stringent approach ensured that only metabolites showing both statistical significance and biological relevance were selected for downstream analysis. To obtain the DMs related to low-altitude diabetes we subsequently conducted a comparative analysis of the metabolomics between L_T2DM and L_HC (L_T2DM.vs. L_HC, L_HC as a comparison). There were 308 DMs related to L_T2DM, among which 51 DMs were upregulated and 257 DMs were downregulated (Figure 2B). To further investigate the metabolic differences between diabetes at high altitudes and low altitudes, we conducted a comparative analysis of 308 DMs associated with L_T2DM and 412 DMs associated with H_T2DM. Finally, 26 DMs were identified, of which 8 DMs were upregulated and 18 DMs were downregulated in L_T2DM (Table 2). In other words, there were 26 DMs between diabetic patients at high altitudes and at low altitudes, with 8 playing a downregulated role in diabetic patients at high altitudes and 18 playing an upregulated role (Figure 2C).

To comprehensively investigate the levels and regulatory processes of these 26 DMs we applied Z score transformation (a value based on the relative level of the metabolite) and conducted cluster analysis. The absolute value of the Z score distribution in the L_T2DM group was smaller than in the H_T2DM group, indicating that the levels of these metabolites were generally lower in the L_T2DM group than in the H_T2DM group (Figure 2D). The clustering results showed that the levels of 4-(2-thienyl) benzoic acid, N’1-(2-chlorobenzoyl)-2-[(2-methyl-1H-indol-3-yl)thio]ethanohydrazide, 2-hydroxy-6-aminopurine, and guanosine were significantly different between the H_T2DM and L_T2DM groups (Figure 2E). In addition, correlation analysis of the levels of the top 20 metabolites revealed that N’1-(2-chlorobenzoyl)-2-[(2-methyl-1H-indol-3-yl)thio]ethanohydrazide and guanosine levels were strongly positively correlated with 2-hydroxy-6-aminopurine levels (Figure 2F). These findings suggest that N’1-(2-chlorobenzoyl)-2-[(2-methyl-1H-indol-3-yl)thio]-ethanohydrazide, guanosine and 2-hydroxy-6-aminopurine were closely related to H_T2DM.

To further screen the key differentially abundant metabolites between high-altitude and low-altitude diabetic patients, KEGG enrichment analysis(The version is 2/20/17(c)Kanehisa Laboratories) (9)was performed on the 26 DMs. The results showed that the DMs were enriched mainly in the KEGG pathways of lysine degradation, fatty acid biosynthesis, purine metabolism and metabolic pathways, and four metabolites, namely, pipecolic acid, lauric acid, guanosine and kaempferol, were the most prominent (Figure 3A). We subsequently conducted a random forest analysis and the results showed that pipecolic acid, lauric acid and guanosine were among the top 20 biomarkers (Figure 3B). Moreover, the area under the ROC curve (AUC) was 0.818, suggesting that the random forest model had high predictive accuracy (Figure 3C). Combined with the results shown in Figure 2F, N’1-(2-chlorobenzoyl)-2-[(2-methyl-1H-indol-3-yl)thio]-ethanohydrazide, guanosine, 2-hydroxy-6-aminopurine, pipecolic acid and lauric acid, these five metabolites are highly correlated with known mechanisms of altitude adaptation, strongly suggesting that altitude environment (rather than diet) is the primary driver of differences between patients with altitude diabetes and those with plains diabetes (10).

We subsequently analyzed the levels of N’1-(2-chlorobenzoyl)-2-[(2-methyl-1H-indol-3-yl)thio]-ethanohydrazide, guano-sine, 2-hydroxy-6-aminopurine, pipecolic acid and lauric acid in each group. The results showed that there were no significant differences in the levels of these five metabolites between healthy individuals at high altitudes and at low altitudes (Figures 3D-H). In addition, 2-hydroxy-6-aminopurine and lauric acid did not significantly differ between H_T2DM and L_T2DM. However, the level of N’1-(2-chlorobenzoyl)-2-[(2-methyl-1H-indol-3-yl)thio]-ethanohydrazide in H_T2DM was lower than in L_T2DM. However, the levels of pipecolic acid and guanosine in H_T2DM were both greater than those in L_T2DM.

To further verify the reliability of these biomarkers, we matched the top 20 biomarkers to the mzCloud (https://www.mzcloud.org/) and mzVault databases (https://mytracefinder.com/tag/mzvault/) (the match between the actual detected mass–charge ratio and the mass–charge ratio reported in the database; the highest score was 100 points) and found that only guanosine and pipecolic acid had high matches in the three databases: both match scores were 88 points except for N’1-(2-chlorobenzoyl)-2-[(2-methyl-1H–indol–3-yl)thio]-ethanohydrazide (Table 3). This result illustrates that guanosine and pipecolic acid had high predictive accuracy and are the greatest potential determinants of diabetic metabolic differences between high and low altitudes.

To verify whether guanosine and pipecolic acid are determinants of diabetes at high and low altitudes we conducted a correlation analysis between metabolites and disease indicators in diabetic patients to identify metabolites that are only affected by altitude differences. The results showed that guanosine levels were not correlated with biochemical indicators (Figure 3I). These findings suggest that changes in guanosine levels are not affected by diabetes and are only associated with changes in altitude.

To confirm the validity of guanosine levels as a potential determinant of metabolic differences in diabetes between high-altitude and low-altitude individuals, we further recruited 30 diabetic patients living at high altitudes (HDMs) and 30 diabetic patients living at low altitudes (LDMs). After application of the inclusion and exclusion criteria, their serum was collected for LC–MS/MS analysis which detected 800 and 792 metabolites in the samples from the HDM and LDM groups, respectively, with 790 metabolites in common (Figure 4A). The results of PLS-DA revealed that R2Y was greater than Q2Y, indicating that the model was well established (Figure 4B). KEGG enrichment analysis was subsequently performed and showed that the DMs were involved mainly in lipid metabolism (Figure 4C). The network structure analysis results revealed that guanosine was still among the top 20 DMs (Figure 4D).

Thus, we believe that guanosine is the greatest potential determinant of metabolic differences in diabetic patients between high-altitude and low-altitude individuals.