We recruited participants from the physical examination population of Shandong University of Traditional Chinese Medicine affiliated with Rizhao Hospital of TCM from April 2021 to December 2021 and divided them into 3 groups with 30 participants in each group. PDC and NPDC subjects with prehypertension and BC subjects with ideal blood pressure. All the subjects met the criteria for phlegm dampness or balanced constitution (ZZYXH/T157-2009) (China Association of Chinese Medicine, 2009), diagnostic criteria for prehypertension and ideal blood pressure, according to the Chinese Guidelines for Prevention and Treatment of Hypertension (Joint Committee for Guideline Revision, 2019). Patients with essential hypertension and/or secondary hypertension; individuals with severe diseases of the liver, kidney, lung, brain, endocrine, hematopoietic, or other organ systems; pregnant or pregnant or lactating women; those who received other treatments that may have affected the population observed in this study; and those who also received probiotics and other intestinal microecological preparations were excluded. All participants were not allowed to take any of the following drugs: antibiotics, laxatives, antidepressants, probiotic supplements, PPI preparations, bismuth, hormones, or chemotherapy drugs at least 4 weeks before the study began. The study scheme has passed the ethical review of the China Registered Clinical Trial Ethics Committee (ethics review document number: ChiECRCT20210464), and all participants are required to provide informed consent.

All clinical information was collected according to standard procedures. Peripheral venous blood was drawn after the participants were enrolled in the study, and the participants were provided with a stool sampler and detailed graphic instructions for sample collection. Freshly collected stool samples from each participant were frozen overnight in liquid nitrogen and stored in a -80°C freezer. After collection, all the samples were sent to Shanghai Personalbio Biotechnology Co., Ltd., for high-throughput sequencing.

The fecal sample was removed from the refrigerator and added to a centrifuge tube containing the extracted lysate for grinding at a frequency of 60 Hz (Shanghai Jingxin Multisample Tissue grinder, Model: Tissuelyser-48).

For the pretreated samples, nucleic acid was extracted via the OMEGA Soil DNA Kit (D5635-02) (Omega Bio-Tek, Norcross, GA, USA). The molecular size was determined via 0.8% agarose gel electrophoresis, and the DNA was quantified via a Nanodrop. Extraction was performed via OMEGA’s Mag-Bind Soil DNA Kit (catalog number: M5635-02).

The highly variable V₃V₄ region of the bacterial 16S rRNA gene, with a length of approximately 480 bp, was selected for PCR amplification. The sequences of the primers used were 338F (5’- barcode+ACTCCTACGGGAGGCAGCA-3’) and 806R (5’-GGACTACHVGGGTWTCTAAT-3’). The barcode in the preprimer is a 7–10-base oligonucleotide sequence used to distinguish different samples in the same library. NEB Q5 DNA high-fidelity polymerase was used for PCR. After the components required for the PCR were configured, the template DNA was predenatured at 98°C for 5 minutes on the PCR apparatus to fully denature and then entered the amplification cycle. During each cycle, the template is denatured by holding for 30 s before 98°C, and then the temperature is lowered to 53°C for 30 s so that the primer and the template are fully annealed. At 72°C for 45 s, the primer was extended on the template, the DNA was synthesized, and a cycle was completed. This cycle was repeated 25 times to enrich the amplified DNA fragments. Finally, the product was kept at 72°C for 5 min so that the extension of the product was complete, and the product was preserved at 12°C. The products were amplified via 2% agarose gel electrophoresis, and the target fragments were cut and then recovered with an Axygen gel recovery kit. The PCR products were quantified on a microplate reader (Bio Tek, FLx800) with a Quant-iT Pico Green ds DNA Assay Kit, after which the samples were mixed according to the amount of data required for each sample.

The Illumina’s Tru Seq Nano DNA LT Library Prep Kit was used for library construction, library quality, and quantification. A 1 μL library was collected, and a 2100 quality inspection of the library was conducted with an Agilent High Sensitivity DNA Kit on an Agilent Bioanalyzer machine. The qualified library should have a single peak and no joint peak. A Quant-iT PicoGreen dsDNA Assay Kit was used to quantify the library on a Promega QuantiFluor, and the concentrations of the qualified libraries were above 2 nM after calculation. Qualified libraries were sequenced to 2×250 bp in length via the NovaSeq 6000 SP Reagent Kit (500 cycles) on the Illumina NovaSeq machine.

QIIME2 2019.4 was used to analyze microbiome biology information (https://docs.qiime2.org/2019.4/tutorials/).

To evaluate the diagnostic effect on intestinal microorganisms, we determined the number of optimal genera through the analysis of the tenfold cross-validation error rate curve, constructed a random forest model (RFM), combined it with the screening results of LEfSe analysis (LDA>2), and calculated the AUC value of the bacteria that jointly met the above conditions. ROC curves were generated to evaluate the efficacy of the diagnostic model for gut microbes. The confidence intervals (CIs) of the area under the curve (AUC) and receiver operating characteristic (ROC) curves were calculated with GraphPad Prism (v 10.0, San Diego, California, CA, USA).

One-way analysis of variance (ANOVA) was applied for continuous variables with a normal distribution and homogeneity of variance; otherwise, a nonparametric test was used. Multiple comparisons were performed via the least significant difference (LSD) t test. Continuous variables are expressed as the mean ± standard deviation (SD) or median and quartile spacing (P25, P75). Categorical variables were compared via the χ² test and are expressed as numbers (percentages). Statistical analysis was performed via GraphPad Prism (10.0, San Diego, California, CA, USA) and SPSS 26.0 (IBM, Armonk, New York), and P < 0.05 was considered statistically significant. Relative abundance differences in flora were analyzed for correlations between clinical indicators and the Personalbio gene cloud platform (https://www.genescloud.cn/).

In addition to blood pressure, individuals in the PDC and NPDC groups presented greater body weight, BMI, body fat mass, and waist circumference (WC) than those in the BC group did. Among them, the differences in WC and BMI were statistically significant in both the NPDC and PDC groups compared with the BC group (P < 0.0001). In addition, there were no statistically significant differences in sex or age distribution among the three groups. Although the weight, WC, BMI, body fat, and TC in the PDC group were greater than those in the NPDC group were, the differences were not statistically significant. In a sense, distinguishing PDC and NPDC individuals with prehypertension on the basis of only obesity, glucose, and lipid metabolism indicators was difficult ( Table 1 ).

Analysis of the α diversity of the intestinal flora in 90 samples from the 3 groups revealed that the indices (Chao1, observed species, Shannon, Simpson, Good’s coverage, and Pielou’s evenness) were not significantly different among the 3 groups ( Figure 1A ). The species rarefaction curve revealed that the sequencing depth was sufficient, as the species abundances of the NPDC and PDC groups were similar but lower than that of the BC group ( Figure 1B ). PCoA (based on the Bray−Curtis distance) was used to evaluate the changes in the overall community composition. Figure 1B shows that there are many overlapping parts among the three groups, and it is difficult to distinguish them clearly ( Figure 1C ). In addition, analysis of similarities (Anosim) based on the Jaccard distance was used to determine whether there were significant differences among the groups. The results revealed that the R values of PDC vs. BC (R= 0.047, P = 0.017) or PDC vs. NPDC (R=0.025, P = 0.123) were greater than 0, which indicates that the differences between groups were greater than those within groups ( Table 2 ). The difference was statistically significant between the BC and PDC groups, whereas the β diversity in the NPDC and BC groups (R=-0.002, P=0.494) was not significantly different.

A comparison of the compositions of the intestinal microorganisms revealed that Bacteroidetes, Firmicutes, Proteobacteria, Fusobacteria, and Actinobacteria were the main dominant bacteria in each group at the phylum level. It accounts for more than 98% of the relative abundance of all bacteria ( Figure 2A ; Table 3 ). Bacteroidetes were the main bacteria (BC group: 54.63%, NPDC group: 51.62%, PDC group: 53.93%). The second dominant phylum was Firmicutes (BC group: 39.72%, NPDC group: 40.26%, PDC group 40.63%), and the ratio of Firmicutes to Bacteroides (F/B) in the PDC and NPDC groups was greater than that in the BC group. However, these differences were not statistically significant at the phylum level. Next, we analyzed the contributions of the top 20 bacterial genera to each group. As shown in Figure 2B ; Table 4 , Bacteroides was the most important genus (BC group: 35.29%, NPDC group: 31.31%, PDC group: 30.07%). A Venn diagram can directly reflect the similarity and specificity of species number composition among samples of different groups. The results revealed 5279 ASVs in the three groups, 30,470 ASVs in the BC group, 29,524 ASVs in the NPDC group, and 31,846 ASVs in the PDC group, which were not found in the other two groups ( Figure 2C ). Species composition heatmap analysis at the genus level revealed that, compared with PDC or NPDC subjects with prehypertension, Bacteroides, Blautia, and [Ruminococcus] were enriched in the BC group, and the relative abundance of Blautia was significantly different (P < 0.05) both in NPDC vs. BC and in PDC vs. BC. Streptococcus, Prevotella, Alistipes, Coprococcus, Oscillospira, Parabacteroides, and Dialister were enriched in PDC subjects relative to NPDC subjects. The difference in the abundance of Alistipes between PDC and NPDC was statistically significant (P < 0.05). These results suggest that the intestinal flora structure of the prehypertensive population is different from that of individuals with ideal blood pressure and that the intestinal flora structures of prehypertensive individuals with PDC and NPDC are also different.

To further identify species with significant differences between groups, we performed linear discriminant analysis (LDA) and effect size (LEfSe) analysis (LDA =2). While performing differential analysis directly for all taxonomic levels simultaneously, more emphasis is placed on finding robust differential species between groups. Three classes, six orders, seven families, and five genera met the conditions that the LDA was greater than 2, and no phyla met the conditions ( Table 5 , Figure 3A ). The relative abundances of o_RF39, f_Porphyromonadaceae, f_Christensenellaceae, g_parabacteroides, and g_nitrobacteria were significantly greater in the population of the PDC group than in either the NPDC or BC group. ( Table 5 ; Figure 3B ).

Accordingly, we analyzed the differences in the fecal microbiota between PDC vs.NPDC, NPDC vs. BC, and PDC vs. BC at the genus level and selected bacteria for which P < 0.05 and LDA > 2 ( Table 6 ). Compared with those in the BC group, the abundances of four species of bacteria in the PDC group significantly changed, and the abundances of Actinomyces, Allobaculum, Corynebacterium, and Sutterella decreased in the PDC group ( Table 6 ). Compared with those in the BC group, the abundance of Flavobacterium in the NPDC group was significantly greater, and the abundances of Allobaculum, Blautia, Coprococcus, Parabacteroides, and Shigella were significantly lower ( Figure 3C ), suggesting that these bacterial genera are specific to prehypertension. The differences between the NPDC and BC groups reflect the distinction between prehypertension and ideal blood pressure. To further clarify the effects of WC and BMI on the differentiation of the NPDC and BC groups, we performed univariate and multivariate logistic regression analyses, respectively. Univariate logistic regression analyses revealed that the ORs of WC and BMI were 1.114 (95% CI: 1.041–1.193, P =0.002) and 1.273 (95% CI: 1.047–1.548, P=0.015), respectively, suggesting that both BMI and WC are risk factors for prehypertension. Through multifactorial logistic regression analysis, BC and BMI were corrected, and WC was included in the regression equation (OR=1.128, 95% CI: 1.016–1.253, P=0.024), suggesting that WC was an independent risk factor for prehypertension. The taxa with significant differences between the PDC group and the NPDC group were related to the phlegm dampness constitution of prehypertension. As shown in Figure 3D , a total of six bacterial genera presented significantly different abundances between PDC and NPDC. Among them, Alistipes, Butyricimonas, Odoribacter, Oscillospira, and Parabacteroides were enriched in the PDC group, and Corynebacterium was enriched in the NPDC group.

To further screen the specific flora of PDC in the prehypertensive population, random forest analysis was performed between the PDC group and NPDC group via LEfSe analysis. The results revealed that Alistipes, Butyricimonas, Odoribacter, Parabacteroides, and Corynebacterium were the bacterial genera whose abundances significantly differed between the NPDC and PDC and strongly contributed to group differentiation ( Figure 4A ; Table 7 ). We noted that these five bacteria are also genera with significant intergroup differences in NPDC vs. PDC according to LEfse analysis ( Table 6 ). ROC analysis was used to further explore the potential of these 5 strains in the identification of phlegm dampness in the prehypertensive population. AUC, sensitivity, specificity, and threshold values were used to assess diagnostic and predictive accuracy ( Table 8 ). The AUC of the PDC compared with the NPDC ranged from 0.653 (95% CI: 0.511–0.794) to 0.684 (95% CI 0.547–0.822) ( Table 8 ; Figure 4B ). We further constructed a combined diagnostic model of these 5 bacterial genera via logistic regression, and the results revealed that the model could better distinguish PDC from NPDC, with an AUC of 0.706 (95% CI: 0.573–0.838) ( Table 8 ; Figure 4B ).