Introduction

Essential hypertension is the most prevalent chronic disease worldwide and is a major risk factor for cardiovascular morbidity and mortality1, 2, 3. Although previous studies have shown that essential hypertension is a multifactorial disease associated with metabolic disorders, its etiology and pathogenesis remain poorly understood4, 5. The “omics” technique offers a high-throughput screening strategy to explore the root causes of hypertension. The genome-wide map** of human loci has linked many candidate genes to essential hypertension6, 7, although none of them has been widely confirmed in different population groups. Recently, the metabolic profiling of hypertension has drawn the attention of researchers8, 9, 10, 11, 12, 13, 22, 23.

In China, it has long been recognized that total ginsenosides (TG) can adjust blood pressure by balancing hyper- and hypotensive states24, 25, 26. Although dozens of investigations have reported the regulatory effects of ginseng extract or ginseng products on lipids and sugars, these studies have offered little information on the specific metabolites involved in these effects (with the exception of glucose and cholesterol). The reduction of systolic blood pressure (SP) and the regulation of metabolism in vivo mediated by TG and frontline clinical agents have never been thoroughly investigated and correlated. To study their regulatory effects on both blood pressure and the perturbed metabolism associated with hypertension, we used SHRs as the animal model in this study and explored their metabolic responses to TG using a metabonomic platform based on a gas chromatography/time-of-flight mass spectroscopy (GC/TOFMS) technique coupled to chemometric analysis27, 28, 29. Four commercially available antihypertensive agents, which act via diverse mechanisms, were included as positive antihypertensive control drugs.

Materials and methods

Chemicals

Authentic reference compounds were purchased from Sigma Aldrich (St Louis, MO, USA) or Sigma-Aldrich (Steinhein, Germany), Fluka (Buchs, Switzerland), Merck (Darmstadt, Germany), or Serva (Heidelberg, Germany); all were of analytical or chromatography grade. Total ginsenosides were composed of the following: ginsenoside Rb1 (19.1%, w/w), Rb2 (13.8%), Rc (12.3%), Rd (9.7%), Re (11.8%), Rf (3.5%), and Rg1 (7.1%) (Jilin University, China). Amlodipine maleate (amlodipine), terazosin hydrochloride (terazosin), captopril, and hydrochlorothiazide were obtained from accredited pharmaceutical companies in China. The following were purchased from Cambridge Isotope Laboratories (Andover, MA, USA), Fluka (Buchs, Switzerland), and Pierce Chemical Company, (IL, USA), respectively: stable-isotope-labeled internal standard (IS), [13C2]-myristic acid; alkane series (C8–C40), pyridine (silylation grade); and N-methyl-N-trimethylsilyl-trifluoroacetamide (MSTFA)+1% trimethylchlorosilane. Methanol, heptane, and sodium ethylenediaminetetraacetate (EDTA-Na) were of chromatography or analytical grades.

Animals and dosing

The animal studies were carried out after approval by the Animal Ethics Committee of China Pharmaceutical University. Male SHRs and normotensive control Wistar Kyoto rats (WKYs), all aged 8 weeks, were purchased from Wei Tong Li Hua Animal Center (Bei**g, China). All rats were housed individually in metabolic cages in a standard animal laboratory with a 12-h light/dark cycle. Water and standard rat chow were available ad libitum. After 2 weeks of acclimatization, the rats were randomly divided into groups (n=6) based on body weight to receive treatment for 8 weeks (10–18 weeks of age). The following were administered to the rats based on the group to which they were assigned: vehicle (0.5% carboxymethycellulose sodium aqueous solution) or agents suspended in 1 mL of the vehicle, including total ginsenosides at 30 or 3 mg/kg (ip per day), captopril at 30 mg/kg (ig per day), amlodipine at 5 mg/kg (ig per day), terazosin at 5 mg/kg (ig per day), or hydrochlorothiazide at 10 mg/kg (ig per day). All of the doses were withdrawn after 8 weeks.

Measurement of SP and collection of plasma sample

The systolic blood pressures of conscious rats were measured using the indirect plethysmographic tail cuff (Shanghai Alcott Biotech Co, Ltd, China) every week as described previously12. After SP measurement, 0.5 mL of blood was collected in EDTA-Na anti-coagulated tubes, and the plasma was obtained by centrifuging at 1600xg for 10 min at 4 °C and stored at −70 °C until use.

Sample preparation and GC/TOFMS analysis

Plasma samples were thawed and the endogenous compounds were extracted, mathoximated, and trimethylsylilated as described previously27. To minimize systematic variations, the plasma samples were analyzed in random order. Using an Agilent 7683 autosampler, 1 μL of sample was injected into an Agilent 6890N system equipped with a fused silica capillary column chemically bonded with 0.18 μm DB5-MS stationary phase (10 m×0.18 mm ID). The inlet temperature was set to 250 °C, and helium was used as the carrier gas at a constant flow rate of 1.0 mL/min through the column. To achieve good separation, the column temperature was initially maintained at 70 °C for 2 min and then increased at a rate of 35 °C/min from 70 °C to 305 °C, where it was held for 2 min. The column effluent was introduced into the ion source of a Pegasus III MS (Leco Corp, St Joseph, MI, USA). The transferline temperature was set at 250 °C and the ion source temperature at 200 °C. Ions were generated by a 70-eV electron beam at a current of 3.2 mA. Masses were acquired with m/z 50–680 at a rate of 30 spectra/s.

Automatic peak detection and mass spectrum deconvolution were performed as described previously27, 28. To minimize interference from drug metabolites, only the peaks (of the same mass spectra and retention time/index) found in the blank control (ie, SHR control and WKY control without drug) were included in the data matrix for further data processing. The retention index of each peak was calculated by comparing the retention time of the peak with those of the alkane series C8-C40. The compounds were identified by comparison of the mass spectra and retention indices of all the detected compounds with the authentic reference standards and those available in the National Institute of Standards and Technology (NIST) library 2.0 (2005).

Principal component analysis and statistics

After normalization against the stable isotope internal standard, the data matrix was constructed with the observation/samples in columns and the responses/peaks as variables in rows. To improve the validity of the mathematical model, response variables of higher average standard deviation (>60%) within each group were excluded from the data matrix. Data processing was carried out using SIMCA-P 11 software (Umetrics, Umeå, Sweden). Principal component analysis (PCA) and partial least squares projection to latent structures & discriminant analysis (PLS-DA) were employed to process the acquired GC/TOFMS data, following established methodology28, 29, 30. PCA involves a mathematical procedure that transforms a number of correlated or uncorrelated variables into a smaller number of uncorrelated, new variables called “principal components” (PCs); in other words, it projects a K-dimensional space (K, the number of variables/peaks) and reduces it to a few PCs that describe the maximum variation in different groups or samples while retaining as much information as possible. Thus, the comparative analysis of data is greatly facilitated and can be visualized; for example, in two- or three-dimensional space. PCA is an unsupervised way to show the original scatter of the plots, whereas the supervised methods, partial least squares projection to latent structures (PLS) and discriminant analysis (DA), allow differentiation of the groups. The results of PCA or PLS-DA are displayed as score plots that represent the scatter of the samples. When tightly clustered, these indicate similar metabolic phenotypes. The relative positions of the samples suggest their similarity or dissimilarity; when loosely clustered, they indicate compositionally different metabolic phenotypes. The purpose of PLS-DA was to develop models that differentiate groups or classes. Samples from the same groups were classified into a single group for PLS_DA modeling. Cross-validation with seven cross-validation groups was used throughout to determine the number of principal components31.

The following statistics for the regression models are discussed throughout this paper. R2X is the percentage of all GC/TOFMS response variables explained by the model, R2Y is the percentage of all observation/sample variables explained by the model, and Q2Y is the percentage of all observation/sample variables predicted by the model. The range of these parameters is between 0 and 1, with values approaching 1 indicating better explanation or prediction. Statistical analysis between groups was validated for biochemical parameters or metabolite concentrations using one-way ANOVA with a significance level of 0.05 or 0.01.

Results

SP of SHR treated with TG, amlodipine, captopril, terazosin, or hydrochlorothiazide

We clearly observed that the SP of the SHR control group increased gradually from week 10 to week 20, whereas the SP of the normotensive control group remained stable (Figure 1). As the model SHRs age ranged from 10 to 20 weeks, their SP remained steady at a high level, 30%–50% higher than that of the normotensive WKY. The SP of the animals was measured every week after treatment with TG, amlodipine, captopril, terazosin, or hydrochlorothiazide. SP decreased rapidly and significantly in the amlodipine-, captopril-, and terazosin-treated groups, whereas TG and hydrochlorothiazide reduced SP only slightly (Figure 1). Interestingly, SP was elevated during the first week of treatment in the TG group. As the treatment continued, SP continued to decrease slowly in all groups, although the captopril-treated group showed a sharp decline in SP after six weeks of treatment. Eight weeks later, we stopped all treatments and SP increased quickly in all groups, with the exception of the TG group. TG displayed its greatest antihypertensive effect on SP in week 8, when SP was 15.6% lower than that of the SHR controls. Two weeks after the cessation of treatment, the average increase in SP was only 8.0% of the TG group, which was still 8.5% lower than the SP of the SHR controls (P<0.01, one-way ANOVA). Although the SP of the captopril group reached the same level as that of the TG group, the percentage increase was much greater, 29.6% higher than TG at week 8. Clearly, TG yielded a persistent and sustained long-term reduction in SP, even two weeks after the cessation of treatment. This result suggests that TG acts on an unknown target to lower SP.

Figure 1
figure 1

Systolic blood pressure (SP) measured from zero to eight weeks of treatment with total ginsenosides (TG), terazosin, captopril, amlodipine, or hydrochlorothiazide, and for two weeks after withdrawal of the drugs. Treatment with TG, amlodipine, captopril, terazosin, and hydrochlorothiazide led to a rapid reduction in SP. This effect was significant in the amlodipine-, captopril-, and terazosin-treated groups, whereas TG and hydrochlorothiazide exhibited a trend towards reduced SP. Two weeks after the withdrawal of the drugs, the SP of the animals in all groups increased rapidly, with the exception of those in the TG-treated groups. bP<0.05, cP<0.01 significantly different from SHR (one-way ANOVA).

Full size image

Metabolic phenotype of SHR and the regulatory effects of TG

According to the PCA algorithm, every dot represents a sample and contains information regarding all the metabolites measured in the sample as well as their concentrations. The metabolites and their concentrations determine the sample's position in the scores plot. Therefore, the relative positions of the plots suggest similarities or differences among the samples; the closer the dots, the more similar the metabolite compositions and concentrations within the samples. Conversely, the further apart the dots, the greater the differences among the samples. To evaluate the effects of the therapeutic agents on the metabolic phenotype, a PLS-DA model was calculated between the SHRs and the WKY normotensive controls between weeks 10 and 18 (ie, zero, two, four, six, and eight weeks after treatment with the agents). This produced a good PLS-DA model of four principal components (PC1: R2X=0.302, R2Y=0.140, Q2Y=0.123; PC2: R2X=0.510, R2Y=0.304, Q2Y=0.214; PC3: R2X=0.536, R2Y=0.387, Q2Y=0.291; and PC4: R2X=0.576, R2Y=0.423, Q2Y=0.298; Figure 2A), revealing a clear separation of SHRs and WKYs and a continual movement in the age-related variation from week 10 to week 18.

Figure 2
figure 2

A) The scores plot shows the metabolomic profiles of SHR, WKY, and SHR treated with total ginsenosides (TG) for two, four, six, and eight weeks. Treatment with TG led to a gradual repositioning of the plots from SHR and toward WKY, suggesting a clear rectification of the metabolome. B) The scores plot shows the metabolomic patterns for the total ginsenosides (TG)-, terazosin-, captopril-, amlodipine-, and hydrochlorothiazide-treated groups at week 8. Significantly different metabolic phenotypes were observed after eight weeks of treatment. The TG group clustered close to WKY, whereas the other groups overlapped with or were close to SHR, which suggests the rectification of endogenous metabolites after TG treatment and a lesser regulatory effect after treatment with the other drugs.

Full size image

From week 10 to week 18, the plots moved gradually from the bottom to the top of the figures, for both SHRs and WKYs, indicating continual metabolic modifications with aging. However, the development of hypertension resulted in the repositioning of the SHR plots further from the WKY plots, suggesting that the metabonomic variations became increasingly significant.

Although the SHR data points were clearly separated from those of the normotensive controls and moved further away with time, the profiles of the SHRs treated with TG showed a gradual movement away from the starting point at the lower right of the positive control and SHR plots toward the upper left, closer to that of the normotensive WKY controls (Figure 2A). The close clustering of the TG group and the WKY group eight weeks after the treatment strongly suggests a similarity between the two groups in terms of their metabolic phenotypes (metabolite compositions and concentrations). This movement pathway suggests a gradual adjustment of the endogenous compounds in the SHRs toward those in the normotensive control. Because endogenous metabolites are intrinsically involved in biochemical processes in vivo, this finding suggests that TG regulates the metabolic disorders that correlate with hypertension.

Comparative regulatory effects of TG, amlodipine, captopril, terazosin, and hydrochlorothiazide on the metabolic phenotype

The regulatory effects on the metabolism of the antihypertensive agents were also compared. Treatment with TG, amlodipine, captopril, terazosin, or hydrochlorothiazide for four weeks resulted in a shift from the SHR control toward the WKY phenotype in the TG-treated group and a slight shift in the amlodipine-treated group. However, the effects observed in the other groups were minor (data not shown). The continuation of the treatments for eight weeks resulted in more pronounced trends.

A snapshot of the clustering of the treated groups revealed a clear differentiation between the treated SHRs and the SHR and WKY controls (Figure 2B). The PLS-DA model of the three principal components (R2X, 0.621; R2Y, 0.439; Q2Y, 0.286) shows that the TG group clustered close to the normotensive control; this result suggests a metabonomic similarity. The amlodipine-treated group was located far from both the SHR and the WKY controls, suggesting a partial regulatory effect on the metabonomic profile. In contrast, the groups treated with captopril, terazosin, or hydrochlorothiazide clustered close to the SHRs control, but far from the WKY control, suggesting only a minor regulation of metabolism. To investigate the dose-dependent metabolic response, we also examined the effects of a low dose of TG. Our data demonstrate that in terms of the observed metabolic adjustment, a low dose of TG was less effective than a high dose of TG.

Regulation of endogenous molecules and the metabolic network

To investigate the metabolic pathways and the specific compounds regulated by the TG intervention, we identified compounds that exhibited differential variation in SHRs and WKYs. In total, we identified 27 compounds (potential markers) that differed significantly between SHRs and WKYs (Table 1). These compounds included free fatty acids (FFAs), amino acids, lipids, and low-molecular-weight organic acids. The variations in their levels suggested the abnormal metabolism of lipids and the tricarboxylic acid cycle (TCA) and abnormalities in glucose and amino acid turnover.

Table 1 The relative abundance of the identified compounds differentiating SHR from WKY and the regulation by TG. bP<0.05, cP<0.01 statistically different from that in SHR (one-way ANOVA); eP<0.05, fP<0.01 statistically different from that in WKY (one-way ANOVA).
Full size table

Treatment with TG led to the restoration of 14 (51.8%) of the 27 discriminatory metabolites (Table 1) to normal levels, that is, significantly different from those of the SHR control (P<0.05, one-way ANOVA) but not different from those of the normotensive WKY control (P>0.05, one-way ANOVA). Amlodipine, captopril, terazosin, and hydrochlorothiazide adjusted the levels of eight (29.6%), four (14.8%), four (14.8%), and two (7.4%) metabolites toward normal values, respectively. In contrast, an analysis of the regulatory effects of the specific agents revealed that captopril significantly regulated α-ketoglutarate, tryptophan, cystine, and cysteine to the normal levels observed in WKY. Amlodipine rectified all the metabolites regulated by captopril, and citrate, pyruvate, and creatinine. Terazosin adjusted tryptophan, cystine, cysteine, and creatinine, whereas hydrochlorothiazide adjusted 3-hydroxybutyrate and cystine.

The regulatory effects of these antihypertensive agents were also reflected in the diverse regulation of specific compounds. For example, high-dose TG adjusted the level of 9-(Z)-hexadecenoic acid to normal (ie, to a similar level observed in WKYs), which was significantly different from its level in SHRs. In contrast, the other agents did not significantly regulate 9-(Z)-hexadecenoic acid (Supplementary Information, Figure S1A). Interestingly, TG regulated most FFAs to normal or near normal levels. High-dose TG, amlodipine, and captopril reduced the levels of α-ketoglutarate to normal, whereas the other drugs had a smaller effect (Figure S1B). Only TG and hydrochlorothiamide markedly reduced the levels of 3-hydroxybutyrate, whereas the other drugs had minor effects (Figure S1C). The regulation of arachidonic acid, fumarate, and ornithine by TG was also observed (Figure S1D–F). In general, TG, terazosin, amlodipine, captopril, and hydrochlorothiazide had different effects on the regulation of 9-(Z)-hexadecenoic acid, arachidonic acid, α-ketoglutarate, 3-hydroxybutyrate, fumarate, and ornithine.

Correlation between SP and potential hypertension markers

The results discussed above show that many of the potential hypertension markers are regulated by TG. It is important to determine whether there is a quantitative relationship between SP and these markers. Linear regression between the SP values and the normalized peak areas of the metabolites revealed that many lipids (eg, oleic acid, palmitic acid, and 9-(Z)-hexadecenoic acid) correlated with SP, whereas several amino acids (eg, serine, threonine, and ornithine) correlated inversely with SP (Supplementary file, Figure S2). This suggests that these compounds are not only potential markers of hypertension, but also quantitative markers of the therapeutic response to these antihypertensive agents.

Discussion

Although metabolic dysfunction is closely related to hypertension, the cause of hypertension remains unknown and the cause-effect relationships between hypertension and metabolic disorders have not been established. In this study, we not only provided evidence of the marked perturbation of lipid and glucose metabolism in SHR but also found that TG normalizes this metabolic perturbation and shows mild, sustained antihypertensive effects; in contrast, the conventional antihypertensive agents efficiently reduced SP but had only minor regulatory effects on the perturbed metabolism. Captopril, amlodipine, and terazosin act on different targets to reduce the peripheral resistance of blood vessels and hence reduce blood pressure. This is illustrated by the following example: captopril acts on the angiotensin-converting enzyme; amlodipine selectively inhibits calcium ion influx across cell membranes, thus acting directly on vascular smooth muscles; terazosin produces a hypotensive effect mainly by blocking the α1 adrenoceptors; and hydrochlorothiazide is a diuretic agent that interferes with the renal tubular mechanism to reduce electrolyte reabsorption. Unlike TG, these agents significantly reduced SP, but had little regulatory effect on the perturbed metabolism of SHR, even after the hypertension had been reduced for eight weeks. The significant regulation of the perturbed metabolism of SHRs resulted primarily from the normalization of glucose metabolism and lipid metabolism, whereas the effects of the other agents on the perturbed metabolism of SHRs were probably associated with their efficient reduction of SP. The withdrawal of the drugs led to a rapid rebound of SP. These results confirm that these agents act on hypertension at a superficial level. Because the efficient reduction of hypertension had only a slight effect on the perturbed metabolism, hypertension may not be the cause of this perturbation of the metabolism.

SHRs are genetically defective in fatty acid translocase/CD3632. Studies have confirmed that defective CD36 contributes to insulin resistance and increased serum levels of fatty acids in SHR and humans33, 34. The identified metabolites that differentiate SHRs from WKYs (Table 1) suggest perturbations in the metabolism of lipids, glucose, and the TCA cycle as well as amino acid turnover (Figure S3). The elevation of FFAs in SHRs confirms the genetic defect in the fatty acid translocase/CD36 gene. Theoretically, the downregulated expression of the fatty acid translocase/CD36 gene will reduce the β-oxidation of FFAs and therefore the animal's energy supply. To our surprise, we detected a marked elevation of 3-hydroxybutyrate, one of the ketone body oxidation products of FFA, possibly because SHRs tend to utilize energy sources other than ketobodies. As supporting evidence, we observed the elevation of TCA cycle intermediates in SHRs relative to those in WKYs. This finding indicates that SHRs rely more on the TCA cycle for energy than on lipid metabolism. High levels of glucose always favor the synthesis of lipids35, and we observed higher levels of FFA and glucose in SHR. However, SHRs are smaller and thinner than WKYs. This obvious contradiction suggests that both the catabolism and the anabolism of lipids are retarded in SHRs.

In general, the TG treatment not only led to a prolonged reduction in SP, even two weeks after its withdrawal, but also showed a distinct regulation of the metabolic pattern (Figure 2A, 2B) and restored the levels of most FFAs in SHRs toward normal (Table 1). These results suggest that TG acts on fatty acid translocase and indicates a relationship between the SP-lowering effects of TG and its regulation of the perturbed metabolism of SHRs. Brindle and colleagues were able to distinguish the serum samples from patients with low/normal SP from those with borderline or high SP. However, the borderline- and high-SP samples were indistinguishable from each other9. This result indicates that the metabolic phenotypes of borderline and high-SP patients are similar and that metabolic disorders have developed in borderline patients. In other words, metabolic disorders develop before high SP can be determined, and the occurrence of metabolic disorders does not immediately result in high SP. These lines of evidence may explain why TG treatment clearly regulated the metabolism of SHR but did not immediately lower SP. These results also provide indirect support for the hypothesis that TG exerts its antihypertensive effect by normalizing the perturbed metabolism involved in hypertension and that metabolic perturbation may play an important role in the development of hypertension. If this hypothesis is correct, the regulation of the perturbed metabolism will be a novel target for the treatment and cure of hypertension.

An infusion of FFAs results in elevated blood pressure in both animals and humans36, 37, 38, 39, 40, 41, 42, 43. A large-scale population study concluded that the elevation of FFAs in human individuals is a highly significant risk factor for the subsequent development of hypertension44. In the present study, we not only determined the presence of higher levels of FFAs in SHRs but also found a good correlation between SP and FFAs (eg, oleic acid, 9-(Z)-hexadecenoic acid, and palmitic acid). Together with evidence from other animal studies and large-scale studies in human subjects, these results suggest that the levels of FFA or FFA-related metabolism may play an important role in the development and remission of hypertension and that FFA may be a good biomarker or indicator of hypertension. Based on the fact that TG significantly downregulated most FFA toward normal levels and produced a mild and sustained reduction in SP, even two weeks after its withdrawal, we propose that TG exerts its antihypertensive effects via the regulation of FFA-mediated biochemical processes.

In summary, the five agents that were investigated produced different hypotensive effects and distinct regulation patterns of the perturbed metabolism that occurs in SHRs. Captopril, amlodipine, and terazosin efficiently reduced SP but had only a minor regulatory effect on the perturbed metabolism of SHRs, indicating that the efficient reduction of hypertension did not result in the correction of metabolic perturbation and that metabolic perturbation is not the consequence of hypertension. Total ginsenosides regulated the perturbed metabolism and showed prolonged hypotensive effects, even two weeks after its withdrawal. These results suggest that metabolic perturbation plays an important role in the development of hypertension and that TG is a unique antihypertensive agent that can regulate perturbed metabolism.

Author contribution

Ji-ye AA, Guang-ji WANG, and Hai-** HAO designed the research; Qing HUANG, Yi-hong LU, Bei YAN, Wei-bin ZHA, Lin-sheng LIU, and An KANG performed the research; Ji-ye AA, Qing HUANG, Yi-hong LU, Bei YAN, and Wei-bin ZHA analyzed the data; Ji-ye AA and Qing HUANG wrote the paper.