This study was approved by the Institutional Ethics Committee of the Third Xiangya Hospital of Central South University (No. 22148). Written informed consent was obtained from all participants, and all procedures were performed according to the Declaration of Helsinki.

Between January 2021 and September 2022, 23 patients with Cushing’s syndrome (18 women and 5 men, aged 47.24 ± 12.99 years) were consecutively recruited from the Third Xiangya Hospital of Central South University. Cushing’s syndrome was diagnosed according to the Pituitary Society 2021 guideline (10). The inclusion criteria were: 1) typical symptoms of Cushing’s syndrome, such as hyperemia, central obesity, and full moon face; 2) increased cortisol secretion, abnormal circadian rhythm, and twice 24­h urinary free cortisol (UFC) >45 µg/24 h; 3) impaired glucocorticoid feedback with overnight 1 mg dexamethasone suppression test (DST) and serum cortisol concentration ≥18 µg/dL (50 nmol/L); and 4) no exogenous glucocorticoids (oral, injections, inhalers, topical). Of the 23 patients, 18 had adrenocortical adenoma, 2 had Cushing’s disease, and 3 had adrenocorticotropin-independent (ACTH)-independent macronodular hyperplasia. Thirty geographically, age, sex, and BMI-matched healthy controls (18 women and 12 men, 45.03 ± 6.69 years). All healthy controls had normal cortisol and ACTH circadian rhythms and no history of diabetes, hyperlipidemia, or other endocrine diseases. All participants who fulfilled the following exclusion criteria were excluded:1) other organic or infectious diseases, such as immune system diseases, gastrointestinal disease, mental and psychological diseases, tumors, etc.; 2) taking steroids, anti-diabetic drugs, antibiotics, or probiotics three months before sample collection; 3) ectopic ACTH syndrome or adrenocortical carcinoma patients; 4) chronic alcoholics; and 5) any other conditions that may affect the gut microbiota, such as overeating or vegetarian for nearly a month.

The clinical data of the Cushing’s syndrome patients were obtained from their medical records. Fasting blood samples were obtained after fasting for 12 h and were kept at -80°C. Fasting blood glucose (FBG), total cholesterol (TC), triglyceride (TG), high-density lipoprotein cholesterol (HDL-C), and low-density lipoprotein cholesterol (LDL-C) concentrations were quantified using an automatic biochemical analyzer (LST-008; Hitachi, Japan). Glycosylated Hemoglobin Type A1C (HbA1c) was detected using high-performance liquid chromatography (HA8180 automatic glycosylated hemoglobin analyzer). Cortisol and adrenocorticotropic hormone (ACTH) were measured using an electrochemiluminescence immunoassay (Roche Cobas 6000 automatic electrochemiluminescence analyzer), and 24 h UFC was detected using liquid chromatography-tandem mass spectrometry (Changsha Kingmed Center for Clinical Laboratory, China).

Clean stools were collected from individuals before using any medication or DST, transported in ice bags, and stored at -80°C. Genomic DNA was extracted using the CTAB method, and the 16S V3-V4 region of the bacteria was amplified by PCR (Phusion^® High-Fidelity PCR Master Mix with GC Buffer, New England Biolabs). The NEBNext^® Ultra™ II DNA Library Prep Kit was used to construct the library, and NovaSeq6000 was used for sequencing. QIIME2 software was used to calculate Simpson and Pielou_e indices and Unifrac distance, and R software was used to draw PCoA and NMDS dimensionality reduction maps. Adonis and Anosim functions were used to analyze the significance of differences in community structure between groups. LEfSe software was used to complete the significantly different species analysis between groups. Random forest analysis was conducted to screen for biomarkers that play an important role in classification or grouping. PICRUSt2 was used to predict microbial metabolic function.

Acetic acid, propionate acid and butyric acid (the charged modes were all positively charged) concentration of the plasma collected from different participants was performed using trace 1310 gas chromatograph (Thermo Fisher Scientific, USA) and ISQ LT (Thermo Fisher Scientific, USA). GC-MS instrument conditions: sample inlet temperature 250°C; Ion source temperature 230°C; Transmission line temperature 250°C, quadrupole rods temperature 150°C. Electron bombardment ionization (EI) source, full scan and SIM scan mode, electron energy 70eV. The peak area and retention time were extracted using MSD ChemStation software, and a standard curve was drawn to calculate the short-chain fatty acid concentration.

Statistical analysis was performed using SPSS software 25.0. Data are presented as mean ± standard deviation (SD), median and interquartile range M (QL, QU), or proportion (%). The normality of distribution was assessed using the Shapiro-Wilk test. Statistical differences in normally distributed continuous variables were evaluated using an unpaired Student’s t-test. Statistical differences in non-normally distributed continuous variables were examined using the Mann-Whitney U test. Chi-square or Fisher’s exact tests were used to analyze differences in categorical variables. Correlation analysis was performed using Spearman analysis. All tests were two-tailed, and P < 0.05 was considered statistically significant.

The clinical characteristics of Cushing’s syndrome patients and controls are summarized in Table 1 . No significant differences were observed in body mass index (BMI), age, or sex between the control and CS groups ( Table 1 ). Compared to the controls, Cushing’s syndrome patients had higher HbA1c and TGs levels, lower HDL-C levels, and higher systolic and diastolic blood pressure. In addition, Cushing’s syndrome patients had a higher cortisol concentration and an abnormal circadian rhythm.

Compared to controls, the Simpson and Pielou_e of α diversity index were significantly decreased in the Cushing’s syndrome group, which indicated that the species richness and evenness of the gut microbiota were decreased ( Figures 1A, B , P < 0.05). The bacterial community composition was distinct from that of the control group according to principal coordinate analysis (PCoA) based on weighted UniFrac ( Figure 1C ). The gut microbiota community structure in the Cushing’s syndrome group was also different from that in the control group by non-metric multidimensional Scaling (NMDS) analysis ( Figure 1D ). There was a significant difference in the gut microbiota community structure of Cushing’s syndrome patients and controls by Adonis analysis (R2 = 0.081, P= 0.003) and Anosim analysis (R value= 0.151, P= 0.005).

To further investigate differences in specific taxa between Cushing’s syndrome patients and controls, we analyzed the relative abundance of 16S rRNA reads at the phylum and genus levels. Compared to controls, the relative abundance of Proteobacteria was significantly increased, and the relative abundance of Firmicutes was significantly decreased at the phylum level in Cushing’s syndrome patients ( Figure 2A , P<0.05). The Firmicutes/Bacteroidetes ratio was significantly decreased in the Cushing’s syndrome group compared to controls (P=0.025). Taxonomic analysis at the genus level demonstrated that the relative abundance of Escherichia-Shigella from Proteobacteria was significantly higher in Cushing’s syndrome patients (P<0.05), while the relative abundance of Agathobacter, Blautia, Anaerostipes, Eubacterium_eligens_group, Eubacterium_hallii_group, Ruminococcus, Lachnospiraceae_ND3007 and Lachnospira from Firmicutes were significantly lower in Cushing’s syndrome patients ( Figure 2B , P<0.05). As displayed in Figure 2C , Escherichia-shigella from Proteobacteria was the dominant genus in the gut microbiota of Cushing’s syndrome patients, while the dominant bacteria genera in the control group were Blauia, Faecalibacterium, and Agathobacter from Firmicutes. Random forest analysis revealed that Anaerostipes, Lachnospira, Lachnospiraceae_ND3007_group, and Escherichia-Shigella were the four most important biomarkers ( Figure 2D ).

Spearman correlation analysis was conducted to evaluate the relationship between gut microbial dysbiosis and clinical indicators in Cushing’s syndrome populations, showing that the HbA1c, cortisol level, SBP, and DBP were significantly positively correlated with Proteobacteria, Escherichia-Shigella, and significantly negatively correlated with the relative abundance of Agathobacter, Blautia, Anaerostipes, Erysipelotrichaceae_UCG.003, Eubacterium_eligens_group, Eubacterium_hallii_group, Lachnospira, Ruminococcus, etc. ( Figures 3A, B ).

Functional prediction analysis using PICRUST2 was performed to predict the functional gene changes in the gut microbiota. PCA analysis revealed an obvious separation in the functional genes of the gut microbiota in the PC2 axis, which indicated that the bacterial genes might differ between groups ( Figure 4A ). The Venn graph depicted that the number of genes shared by the two groups was 7,253, with 176 genes unique to the Cushing’s syndrome group and 323 genes unique to the control group ( Figure 4B ). The volcano map demonstrated differential genes ( Figure 4C ).

Because the short-chain fatty acid-producing microbiota, such as Blautia, Anaerostipes, Eubacterium_eligens_group, Eubacterium_hallii_group, and Lachnospira were significantly decreased in Cushing’s syndrome patients compared to controls, the concentration of short-chain fatty acids was quantified in Cushing’s syndrome patients and controls by GC-MS. Compared to controls, propionic acid concentration was significantly decreased in the Cushing’s syndrome group (0.151±0.054 vs. 0.205±0.032 µg/mL, P=0.039) ( Figure S1 ) and significantly negatively correlated with systolic pressure and cortisol levels ( Table 2 , P<0.05). There was no difference in butyric acid and acetic acid in those two groups (CS vs Control: butyric acid 0.041±0.049 vs 0.036±0.035 ug/ml, P=0.831; acetic acid 1.382±0.612 vs 1.517±0.512 ug/ml, P=0.664) ( Figure S1 ).