This research was approved by the Ethics Committee of Guangxi University (approval number: gxdxyxll05). For the experiment design, 46 subjects from Donglan County, Hechi City, Guangxi, were recruited, and all subjects signed written informed consent. The general conditions of the research subjects were recorded, including their age, gender, height, weight, and blood pressure of the subjects. The age information was verified by comparing the ID card or household registration. Clinical and medical history were collected based on the self-report and physical examination reports (Supplementary Table 1). Blood pressure (BP) was measured using a calibrated arm electronic sphygmomanometer (Omron). When measuring blood pressure, subjects were asked to sit still for 5 min with their feet flat on the floor without crossing them, and their upper arms were relaxed and placed on a table at the same level as the heart. Two measurements were taken 1 min apart, and if systolic BP was >15 mm Hg apart or diastolic BP > 10 mm Hg apart, a third measurement was performed and averaged. Inclusion criteria: (i) age ≥ 60 years, and either sex, (ii) body mass index (BMI) 17–25 kg/m², (iii) systolic blood pressure (SBP) ≥ 140 mmHg, and not using BP-lowering medication, (iv) from autochthonous families, none was coming from other regions originally. Exclusion criteria: (i) history of chronic medical conditions (diabetes, gastroenteritis, chronic kidney disease, morbid obesity, chronic pancreatitis, or other malabsorption disorder), (ii) use of probiotics, prebiotics, or antimicrobial medication (antibiotic or antifungal treatments) 1 year before sampling. Four participants were excluded due to high BMI, four were excluded due to the SBP lower than 140 mmHg, and five were excluded due to chronic medical conditions. One participant dropped out due to unwillingness to continue (Supplementary Figure 1). Following criteria, a total of 32 participants remained: 19 participants aged 60–90 were divided into Hypertension-Elderly (HTE, average 70.5) group, and 13 participants aged 92+ were divided into Hypertension-Longevity (HTL, average 100). The characteristics of the participants are shown in Table 1.

Participants’ dietary and nutritional status was assessed using a semi-quantitative food frequency questionnaire (FFQ), which was recorded for consecutive 7-days at home (23, 24). All participants were asked to adopt the mode of individual dining and were urged to maintain their usual eating patterns at home during the dietary survey. Trained investigators used electronic food scales, measuring cups, and spoons to weigh and record each food and drink the participants consumed. For unmeasurable food, the standard portion size was used. Qualified dietitians evaluated and verified the dietary data at a dietary survey location for quality control purposes. Dietary data was converted to nutritional consumption levels according to the 2022 Dietary Guidelines for Chinese Residents. The average daily nutritional intake was determined by multiplying the food consumed (in grams) or the portion size consumed by the nutrient content per 100 g of the food listed in the guideline (25).

Our study followed guidelines for gut microbiota studies in hypertension (26) and the STORMS (Strengthening the Organization and Reporting of Microbiome Studies) reporting (27) (Supplementary Table 2). Fecal samples of the participants were collected by themselves or their family members with an icebox at home in sterilized empty tubes (for SCFA determination) or tubes containing RNAlater (Thermo Scientific) for microbial DNA extraction. Morning urine samples were collected in sterilized empty tubes with an icebox. Tubes were immediately stored in a small car refrigerator (Alpicool, A30) at −20°C for <24 h. Then, the samples were transported to the laboratory by professionals and stored at −80°C after dispensing into 2 mL sterilized EP tubes until analysis. DNA was extracted using the Stool Genomic DNA Extraction Kit (Solarbio) according to the manufacturer’s instructions. After extraction, PCR amplification of the full-length region of bacterial 16S rRNA was performed using universal primers 27F (5′-AGRGTTTGATYNTGGCTCAG-3′), 1492R (5′-TASGGHTACCTTGTTASGACTT-3′), and purified with magicpure size selection DNA beads (TANGEN Biotech Corporation Ltd., Beijing, China).

Subsequently, the quality inspection was performed on the formed sequencing library, and processing, including barcode recognition, was performed on the high-quality circular consensus sequencing (CCS) sequence obtained. The generated optimization CCS was clustered at 97% similarity (USEARCH, version 10.0), and its species classification was obtained based on the operational taxonomic unit (OTU) sequence composition. The platforms of 16S: Silva database and RDP Classifier were used to analyze species annotation and taxonomy and gut microbiota diversity. Alpha diversity analysis was performed to examine the species richness and diversity within a single sample, and each sample’s ACE, Chao, Shannon, and Simpson indices were calculated. The differences in the community composition and structure of different samples were compared through Beta diversity analysis. Finally, Metastats analysis was carried out to compare the significant differences between the groups at the genus level, and linear discriminant analysis effect size (LEfSe) analysis was used to screen biomarkers that were statistically different between the groups (criteria: LDA score > 3.5). The raw sequencing data generated in this study have been deposited to the NCBI SRA database (BioProject: PRJNA888351).

Serum SCFAs were measured in 400 μL, and fecal SCFAs were measured from 0.2 g of fecal sample, all in triplicates. For the fecal sample extraction, 0.2 g of −80°C frozen feces were mixed with 2 mL ultra-pure water in a centrifuge tube. Samples were shaken for 2 min and centrifuged at 12,000 rpm for 20 min (4°C). After centrifugation, 1 mL supernatant was mixed with 0.25 mL of 25% metaphosphoric acid, kept in ice bath at −20°C for 2 h. After thawed, the samples were centrifuged at 12,000 rpm for 10 min (4°C). The supernatant (1 mL) was mixed with 1 mL ethyl acetate, vortexed and centrifuged for 12,000 rpm for 5 min (4°C) to obtain the clear organic layer, which was aspirated and filtered through an organic green membrane into the cold GC glass vials for analysis. For the serum sample extraction, 400 μL serum was mixed with 50 μL of 25% metaphosphoric acid and kept in ice bath at −20°C for 2 h. Follow steps were similar to fecal extraction and took care to add proportion. The analysis of acetate, propionate, isobutyrate, butyrate, isovalerate, and valerate was performed by gas chromatography-mass spectrometry (Agilent 7890A Series) using a capillary DB-Wax column (30 m, 0.25 μm, 0.25 μm; SGE, Cromlab SL, Barcelona, Spain), coupled with a flame ionization detector (GC-FID). The column temperature was programmed at 80°C, reached 110°C at a rate of 10°C/min, maintained for 1 min, then increased to 160°C at 4°C/min, and maintained 1 min (total run time 17.5 min). Helium was the carrier gas (0.5 mL/min). The injection was carried out with a split injector (1:100) at 220°C, the detector temperature was 250°C, and 1 μL of the solution was injected into the GC-FID system. The SCFAs was identified according to standard compounds’ retention time of standard compounds (acetate, propionate, isobutyrate, butyrate, isovalerate, and valerate; Sigma-Aldrich).

Aliquots of 500 μL of urine were transferred into a 2.0 mL centrifuge tube and added with 500 μL of 100 mM phosphate buffer containing 0.05% TSP-labeled heavy water (10% heavy water, K₂HPO₄/NaH₂PO₄, PH = 7.4). The mixtures were vortexed and centrifuged (15 min, 12,000 × rpm, 4°C) to obtain the urine supernatant.

The three main SCFA-sensing receptors GPR41 (FFAR3), GPR43 (FFAR2), and GPR109A (HCAR2) have a role in cardiovascular dysfunction (28). As these receptors are highly expressed in immune cells (29), we quantified the mRNA expression of the 3 receptors in circulating immune cells in 30 participants. Since no blood samples were collected from the participants, we selected 30 samples(15 in HTE and 15 in HTL) from long-lived populations with hypertension characteristics collected earlier (23). The basic information of these samples is presented in Supplementary Table 3. Whole blood was treated with Red Blood Cell Lysis Buffer (ThermoFisher Scientific), and RNA was extracted using the Total RNA Extraction Kit (Solarbio). RNA was quantified in a Nanodrop Spectrophotometer, and the first-strand complementary synthesis reaction (cDNA) was made using the BeyoRT™ III First Strand cDNA Synthesis Kit (Beyotime). SYBR Green I was used in a Roche LIGHTCYCLER 96 Real-time quantitative PCR (qPCR) system (Roche Diagnostics Co., Ltd., Basel, Switzerland), with GAPDH as housekeeping genes (Supplementary Table 4). All expression experiments were run in duplicates, and significance was assessed by the 2–ΔΔCT method.

Microbiome data were analyzed as explained above. Graphs were generated using GraphPad Prism version 9.0. Adobe Illustrator version 2020 was used to figure combinations and graphical abstract drawings. All the analyses were performed using SPSS version 26.0 (SPSS Inc., United States) and a 5% significance level. The data were presented as mean ± SD values. The Mann–Whitney U test was used to determine the statistical difference between the two groups gut microbiota, SCFAs, and urine metabolites. A spearman correlation test was conducted to explore correlations between the differential genus and significant microbial pathways and nutrient intake. Pearson correlation analysis was performed to examine correlations between blood pressure and levels of SCFAs, and urine metabolites, after adjustment for age and BMI. Further analyses were conducted using simple linear regression models for BP, acetate, butyrate, and propionate.

The nutrient intake levels in the two groups are presented in Table 2. The two groups had significantly different nutrient intake levels. Notably, the level of energy, fat, saturated fatty acid (SFA), monounsaturated fatty acids (MUFA), polyunsaturated fatty acids (PUFA), carbohydrate, thiamine, riboflavin, vitamin E, folic acid, phosphorus, potassium, calcium, zinc, and manganese were significantly lower in subjects in the HTL compared with the group HTE (p < 0.05). Furthermore, the intake of protein, dietary fiber, cholesterol, vitamin C, vitamin K, nicotinic acid, phosphorus, sodium, magnesium, iron, and selenium were at the same level between the groups. In addition, there were no significant differences in the number of food items consumed by participants, probably due to the similar dietary habits of the participants in the longevity areas. The spearman correlation was used to calculate the relationship between genus and nutrient intake (Supplementary Figure 2). The genus of Faecalibacterium was positively correlated with sodium, whereas the genus of Subdoligranulum was positively correlated with energy, fat, SFA, PUFA, vitamin A, thiamine, riboflavin, vitamin E, phosphorus, magnesium, calcium, iron, and zinc. Clostridium was negatively correlated with vitamin A, iron, and zinc. Christensenellaceae_R7_group had a negative correlation with vitamin C.

Through high-throughput sequencing and sequence screening, 400,724 CCS sequences were obtained through the Barcode identification of 32 samples after sequencing. Each sample generated at least 10,254 CCS sequences, with an average of 12,523 CCS sequences. Operational taxonomic units (OTUs) were divided according to 97% sequence similarity, and a total of 360 OTUs were obtained for subsequent analysis. Through the Shannon curve and the rank abund curve, it was found that the number of OTUs in all samples increased with the increase in sequencing amount and eventually plateaued, indicating that the sequencing volume is sufficient to achieve the ideal sequencing depth (Supplementary Figure 3). The shared and unique OTUs within and between groups were analyzed and plotted as Venn diagrams. HTL had six shared OTU features, HTE had three shared OTU features, and 306 shared OTU features between the two groups (Supplementary Figure 4). Alpha diversity was assessed by five parameters and showed no significant difference between HTE and HTL (Supplementary Table 5).

A total of 11 phylum were dominated by Firmicutes, Proteobacteria, Bacteroidetes, and Verrucomicrobiota, which accounted for 97.32% (HTE) and 99.01% (HTL), respectively. The ratio of Firmicutes to Bacteroidetes(F/B) of HTL (2.14) was significantly lower than HTE (3.71). Figure 1 shows species with relative abundances greater than 1% at the family and genus levels. Both groups were dominated by five families, Enterobacteriaceae, Ruminococcaceae, Akkermansiaceae, Bacteroidaceae, and Lachnospiraceae, with cumulative abundances of 72.61% (HTE) and 75.90% (HTL), respectively. Compared with HTL, the HTE group had much higher levels of the hypertension-related bacteria genera Klebsiella and Streptococcus (+655% and +242%), while it had lower levels of SCFA-producing bacteria Bacteroides (–60%), Roseburia (–80%), Faecalibacterium (–50%), and Alistipes (–79%), respectively. Table 3 further presents the significant difference between the groups in taxa at the genus and species level. The above results showed that the HTE experienced gut dysbiosis, while HTL were abundant in SCFA-producing bacteria. These results are consistent with the commonly reported hypertensive gut genera in addition to identifying two new health-associated bacterial genera (Supplementary Table 6).

The LEfSe method was used to analyze and screen out the biomarker with an LDA score > 3.5. A total of 23 differential species were found between the two groups, Klebsiella pneumoniae, Lactobacillus gasseri, Streptococcus salivarius, Ruminococcus, Actinomyces, Rikenellaceae, f_Saccharimonadaceae, Clostridium perfringens in the HTE and Bacteroids, Faecalibacterium prausnitzii, Parabacteroides, Alistipes in the HTL group, respectively (Figure 2).

PICRUSTs2 was performed to predict the functional prediction of KEGG microbial pathways in gut microbiota, and a total of 248 Class3 microbial pathways were obtained. The biosynthesis of secondary metabolites, biosynthesis of antibiotics, microbial metabolism in diverse environments, and biosynthesis of amino acids were significantly enriched in both HTE and HTL groups. There were 114 Class3 microbial pathways with relative abundance greater than 0.1% and 16 microbial pathways were significantly different between HTE and HTL groups, including ABC transporters, alanine, aspartate and glutamate metabolism, arginine biosynthesis, benzoate degradation, degradation of aromatic compounds, GABAergic synapse, glycosaminoglycan metabolism, inositol phosphate metabolism, lysosome, other glycan degradation, phosphotransferase system (PTS), propanoate degradation, quorum sensing, selenocompound metabolism, sphingolipid metabolism, and tyrosine metabolism (Supplementary Table 7). The correlations between the top 20 species and 16 significant microbial pathways were further analyzed by Spearman correlation (Figure 3). Interestingly, the association of the hypertensive marker species Klebsiella pneumoniae with microbial pathways was consistent with Escherichia coli, but the correlation was reversed for Akkermansia muciniphila.

Six short-chain fatty acids (SCFAs) in fecal samples, namely acetate, propionate, isobutyrate, butyrate, isovalerate, and valerate, were qualitatively detected by GC-MS. Acetate (p = 0.008) and propionate (p = 0.033) were significantly lower in HTL feces than in HTE (Figures 4A, C). Correlation analysis showed that there was no correlation between SBP and SCFAs, but DBP was negatively correlated with acetate (r = −0.368, p = 0.045) and propionate (r = −0.387, p = 0.035) (Figures 4B, D). There were significant correlations between SCFAs, and acetate was significantly positively correlated with propionate (r = 0.708, p = 0.000), butyrate (r = 0.545, p = 0.002), and valerate (r = 0.453, p = 0.012) (correlation results after adjusted for age and BMI).

Through urine ¹H-NMR metabolomic analysis, 36 metabolites were identified (Supplementary Figure 5). Compared with HTE, urine acetate (p < 0.001), pyruvate (p = 0.022), alanine (p = 0.014), and acetoacetate (p = 0.010) were significantly higher in HTL, while creatinine was significantly lower (p = 0.041) (Figure 5). Supplementary Table 8 shows the urine metabolites significantly associated with SBP and DBP. In particular, N-acetyl glycoprotein was negatively correlated with both SBP and DBP (correlation results were adjusted by age and BMI).

To further explore the relationship between SCFAs and hypertension, we then directly measured the levels of serum SCFAs and plasma SCFAs receptors. Since no blood samples were collected from the participants, 30 samples (15 in HTE and 15 in HTL) from long-lived populations with hypertension characteristics collected earlier were selected (23). Hypertension long-lived elderly had higher serum acetate levels than hypertension elderly (p = 0.011) (Figure 6A). There were no significant differences between the groups’ levels of serum propionate, butyrate (Figures 6B, C), isobutyrate, valerate, and isovalerate. We then quantified the mRNA levels of SCFA-sensing receptors, GPR41, GPR43, and GPR109A, in white blood cells from these samples. There were no changes in the expression of GPR41, GPR43, and GPR019A (Figures 6D–F).