The botanical drugs used to prepare the decoctions were purchased from the Third People’s Hospital of Fujian University of Traditional Chinese Medicine (FJUTCM). The botanical drugs were authenticated by the staff in the Herb Identification, Teaching and Research Division of the College of Pharmacy of FJUTCM. The voucher specimens were kept in the College of Pharmacy of FJUTCM, Fujian, China. The decoctions, including those of Chaihu-Shugan-San (CSS), Yougui, and Zuogui, were prepared according to the Pharmacopoeia of the People’s Republic of China (2015 Edition). The combined decoction of Zuogui and CSS (Zuogui-CSS) was also prepared. The ingredients are listed in Tables 1–4.

For preparing the decoctions, botanical drugs for each formula were mixed and macerated in distilled water at room temperature for 1 h. The mixture was then decocted twice with distilled water for half of an hour, at a ratio of 1:10 w/v and at a ratio of 1:5 w/v, respectively. The two resultant decoctions were combined, centrifuged, and filtrated. The decoctions of CSS, Yougui, Zuogui, and Zuogui-CSS were concentrated to 0.2777, 0.476, 0.304, and 0.5817 g/ml of the crude drugs, respectively. All decoctions were stored at −80°C.

SPF-grade (specific pathogen-free grade) female Sprague-Dawley (SD) rats (5 months old) were purchased from Shanghai Xipuer-Bikai Laboratory Animal Co., Ltd. (Shanghai, China). Animals were housed in individually ventilated cages and kept in an environmentally controlled room with a 12 h light/dark cycle at constant temperature (23°C) and humidity (55%). The rats had free access to the standard laboratory water and diet.

Perimenopausal rats were generated based on previous studies (Cai et al., 2016; Chen et al., 2020). These rats were with irregular menstrual cycles at the age of about 11–13 months. Forty of these perimenopausal rats were selected and randomly divided into five groups with eight rats each. The rat model with KD and LQS was generated by the method of chronic restraint stress (CRS) (Liang et al., 2015). The food intake, water intake, body weight, appearance, and activity of the rats were monitored weekly throughout the experiment. Plasma concentrations of estrogen (E2), adrenocorticotropic hormone (ACTH), cortisol (CORT), corticotropin-releasing hormone (CRH), 5-hydroxytryptamine (5-HT), dopamine (DA), and beta-endorphin (β-EP) were determined by ELISA (with kits from CUSABIO, China).

The human equivalent dose (HED) for rat has a multiplication factor of 6:37 (Reagan-Shaw et al., 2008). So the final concentration of crude drugs for Yougui, Zuogui, CSS, and Zuogui-CSS was adjusted to 0.476, 0.304, 0.2777, and 0.5817 g/ml, respectively. The decoctions were administered by gavage with a feeding tube once daily for 4 weeks at a dose of 1 ml for every 100 g of body weight. After overnight fasting, blood samples were collected and centrifuged. All plasma samples were stored at −80°C.

The NMR spectra were acquired at 298 K on a 600 MHz Bruker AVANCE II NMR spectrometer (Bruker BioSpin, Rheinstetten, Germany) operating at 600.13 MHz for ¹H signals. The ¹H NMR spectrum was acquired using a Carr-Purcell-Meiboom-Gill (CPMG, RD-90-(τcp-180-τcp)-acquisition) spin-echo pulse sequence to suppress the water with a total spin-spin relaxation delay (RD) of 320 ms to attenuate broad signals from proteins and lipoproteins due to their short transverse relaxation time. Each ¹H NMR spectrum was obtained with 80 scans with a spectral width of 12335.5 Hz, spectral size of 65,536 points, pulse width (PW) of 30° (12.7 μs), and RD of 2.0 s. The FIDs were Fourier transformed with LB = 0.3 Hz.

The ¹H NMR spectra were manually corrected for phase and baseline with the Topspin 3.2 software. Integrations of water resonance (4.70–5.15 ppm) in the spectra of aqueous samples were excluded. The metabolites were normalized to the total integrated spectral area (−0.55–8.55 ppm) for aqueous samples. The data sets were log-transformed and Pareto-scaled (mean-centered and divided by the square root of the standard deviation of each variable) prior to statistical analysis.

Multivariate analyses of the NMR spectra were carried out with the algorithm of PLS-DA (Partial Least Square Discriminant Analysis) implemented in MetaboAnalyst 4.0 (Chong et al., 2018). The model quality was validated based on two parameters, R² (goodness-of-fit parameter) and Q² (predictive ability). A model with a large R ² (close to 1) and Q² (Q² ≥ 0.5) was considered an excellent model. The PLS-DA model was also validated by the permutation test in which the class membership was randomly shuffled by 100 times for calculating the response values. The new R ² and Q² values were lower than the original ones indicating that the model was not overfitting (Chang et al., 2007).

The important metabolites were identified based on their respective variable influence on projection (VIP) score in the PLS-DA analysis. Significant differences of the selected signals of the main metabolites, which were responsible for class discrimination, were analyzed using a t-test in GraphPad Prism 5 software. The data are presented as p < 0.05 (*), p < 0.01 (**), and p < 0.001 (***).

The human ligand-binding domain of estrogen receptor α (ERα LBD) containing residues 297–554 in the ERα sequence was expressed in E.coli with BL21 (DE) cells and a pET-22b vector. It was purified following an established protocol (Bruning et al., 2010). STD NMR (Mayer and Meyer, 2001; Meyer and Peters, 2003) was employed to detect the ERα LBD interaction with small molecules in the TCM decoctions at 298 K on an 850 MHz Bruker ADVANCE III spectrometer (Bruker BioSpin, Rheinstetten, Germany) equipped with a TCI CryoProbe. The RD was set to 2 s. Selective on-resonance irradiation frequency was set to 0.47 ppm with a saturation time of 2 s. The selective saturation was achieved by a train of 50 ms Gauss-shaped pulses separated by a 2 ms delay. The duration of the presaturation of 2 s was adjusted using n = 128 cycles. Off-resonance irradiation frequency for the reference spectrum was applied at 50 ppm. The decoction of Zuogui-CSS was mixed with the recombinant ERα LBD (20 μM) in phosphate buffer (50 mM sodium phosphate, pH 7.4, 150 mM NaCl, and 5% glycerol) in the presence of 10% D₂O (Cambridge Isotope Laboratories, United States). The NMR data were analyzed with the Topspin 3.2 software.

Simulated gastric fluid (SGF) and simulated intestinal fluid (SIF) were prepared following the protocols in Pharmacopoeia of the People’s Republic of China (2015 Edition). SGF was prepared by mixing 16.40 ml of diluted hydrochloric acid, 800 ml of Milli-Q water, and 10 g of pepsin (Sigma-Aldrich, P7012) to a final volume of 1,000 ml. SGF thus prepared had a pH value of about 1.4. Solution A for SIF was prepared by dissolving 6.80 g of potassium dihydrogen phosphate in Milli-Q water to a final volume of 500 ml. The pH value of solution A was adjusted to 6.8 with a NaOH solution (0.1 mol/L). Solution B for SIF was prepared by dissolving 10 g of pancreatin (Sigma-Aldrich, P7545) in Milli-Q water. Solutions A and B were mixed and the final volume was adjusted to 1,000 ml with Milli-Q water to make SIF.

The SGF-treated Zuogui-CSS was prepared by incubating Zuogui-CSS with SGF at a ratio of 1:10 at 37°C with shaking at 100 rpm for 2 h. The sample was then dried using a rotary evaporator at 70°C. The SGF-treated samples were rehydrated with Milli-Q water. Similarly, Zuogui-CSS was treated with SIF by mixing Zuogui-CSS and SIF in a ratio of 1:4 for a 4 h incubation. It was similarly dried and then rehydrated as for the SGF-treated samples. Zuogui-CSS was also sequentially treated with SGF for 2 h and SIF for 4 h. The SGF- and SIF-treated samples were analyzed by ¹H NMR spectroscopy.

The perimenopausal rats (PM rats) were with irregular menstrual cycles starting at the age of about 12 months old and these rats were deemed to be with KD (Miao et al., 2018; Chen et al., 2020). The weight, sugar preference rate, and behavior indices of rats were of no significant difference between 6-month-old rats and perimenopausal rats, based on sugar preference test (SPT) and open-field test (OFT) (Strekalova et al., 2004; Kraeuter et al., 2019) (Supplementary Figure S1). While the levels of CRH, CORT, β-EP, and 5-HT showed no significant difference, the levels of E2 and DA were lower in perimenopausal rats, whereas the level of ACTH was upregulated (Supplementary Figure S2). These data were compatible with the physiological state of the perimenopausal rats (Cai et al., 2007; Fu et al., 2017).

In the preceding study, the metabolic profiles of rats entering the perimenopausal state and that of younger age were clearly different with metabolites for energy metabolism, such as lipid, glucose, trimethylamine-n-oxide, glutamine, pyruvate, acetoacetate, citrate, betaine, and acetone, influencing the shift of metabolomes (Chen et al., 2020).

To generate a rat model for PMS, the perimenopausal rats were treated with CRS. These rats were considered to be with both KD and LQS (Li et al., 2015; Miao et al., 2018). The PMS rats were with a weight loss (Figure 1A), decreased sugar preference rate (Figure 1B), and reduced level of activity compared to either the 6-month-old or the PM rats without the stress (Supplementary Figure S3). The analyses also showed that the concentrations of E2, DA, β-EP, and 5-HT in the peripheral blood of the PMS rats decreased rapidly (Figures 1C–E), while the levels of ACTH and CORT in direct correlation with the HPA axis reactivity were elevated. These data indicated that PMS rats were under stress and depression.

The metabolomic states between PM rats and PMS rats were distinctly different (Figure 2A) and the metabolites that dictated the metabolomic change in the PMS rats were glucose and lipid (Supplementary Figure S4 and Figure 2B). The levels of creatine, glycerol, trimethylamine-n-oxide, betaine, glutamine, pyruvate, acetoacetate, methionine, and glycoproteins increased in the plasma of PMS rats. In addition, the concentrations for unsaturated lipid, glycine, acetone, acetate, and isoleucine decreased (Supplementary Figure S5 and Supplementary Table S1).

The metabolomic profile of PMS rats was also distinct from that of the 6-month-old rats (Figure 2C) with elevated levels of glucose, creatine, trimethylamine N-oxide, betaine, choline, creatinine, citrate, glutamine, succinate, pyruvate, acetoacetate, methionine, glycoprotein, acetate, and valine, but a downregulated level of glycine (Figure 2D, Supplementary Figures S6, S7, and Supplementary Table S2).

The above data indicated that the glucose metabolism and lipid metabolism of perimenopausal rats were upregulated from those of the 6-month-old rat. Entering the LQS state, the glucose metabolism was upregulated but the lipid metabolism was suppressed in PMS rats comparing to the perimenopausal rats. This aberrant state of glucose metabolism was in line with the fact that the rats were under stress and LQS was associated with downregulated lipid metabolism.

After entering the PMS state, the metabolomic state changed rapidly without any treatment (Figure 3A). Since PMS would be with both KD and LQS by TCM diagnosis, it would be reasonable to treat the model with the corresponding prescriptions in TCM. KD was subdivided into kidney yin deficiency, kidney yang deficiency, and the combination of the two. Traditionally, Yougui, a decoction made from a mix of Chinese botanical drugs (Table 2), is prescribed for kidney yang deficiency and Zuogui (Table 3) is prescribed for kidney yin deficiency (Zhao et al., 2011; He et al., 2014; Fu et al., 2017; Chen et al., 2019). Previously, we discovered that both decoctions of Zuogui and Yougui could adjust the metabolomic states of PM rats (Chen et al., 2020). Metabolomic regulations of PMS rats by these two decoctions were also investigated.

PMS rats were treated with the decoction of Yougui (Yougui) for comparison with rats of 6-months old and of perimenopause. The treatment of Yougui showed little influence on the weight (Figure 3B) and the level of E2 of PMS rats (Figure 3C). However, the levels of ACTH and CORT were reduced, while the level of DA increased (Figures 3D,E). Overall, following the treatment with Yougui, the weight and the abovementioned endocrine indexes of PMS rats were not significantly changed. The E2 level was not significantly different either (Figure 3).

Yougui treatment did alter the metabolomic state of PMS rats by bringing the metabolomic state of PMS rats closer to that of perimenopausal rats (Figures 4A,B). The metabolomic difference between PMS rats treated with Yougui and rats of 6 months old though was still significant (Figure 4C). Overall, the data indicated that Yougui treatment improved the metabolomic state of PMS rats and brought its metabolomic profile closer to those of perimenopausal rats without LQS (Figure 4D). The dominant metabolites altering the metabolomic profile in the Yougui treatment were lipid, glucose, amino acids, and proteins (Supplementary Figures S8–S10).

After Yougui treatment, the levels of acetate and isoleucine increased significantly in the PMS rats, but the level of glucose decreased. In comparison to the 6-month-old rats, the levels of creatine, trimethylamine-n-oxide, choline, creatinine, glutamine, acetoacetate, glycoprotein, acetate, valine, isoleucine, and leucine were significantly higher (Supplementary Figure S11 and Supplementary Table S3).

While Yougui is generally prescribed for kidney yang deficiency, the decoction of Zuogui (Zuogui) (Table 3) is for kidney yin deficiency in centuries of TCM practice (He et al., 2014). Treatment of PMS rats with Zuogui showed minimal influence on the body weight (Supplementary Figure S12). Similar to the treatment with Yougui, treatment with Zuogui did not lead to a significant change in the level of E2 in PMS rats (Figure 5A), but the levels of CORT and ACTH were reduced, and the level of DA was increased. Unlike Yougui, Zuogui treatment increased the level of β-EP (Figures 5B,C).

Zuogui modulated the metabolomic state of the PMS rats (Figure 5D). Lipid, glucose, proteins, and trimethylamine-n-oxide were important metabolites influencing the metabolomic profile (Supplementary Figure S13). Although the metabolomic profile moved in the direction of the perimenopausal rats, the adjustment was not sufficient to bring the metabolomic state of PMS rats to that of 6-months old (Figure 5E).

Zuogui upregulated the levels of unsaturated lipid, glutamine, acetate, valine, and lipid while downregulated the levels of creatine, trimethylamine-n-oxide, creatinine, glutamine, pyruvate, acetoacetate, and glycoproteins in the PMS rats. Notably, after the Zuogui treatment, the levels of creatine, creatinine, glutamine, pyruvate, glycoproteins, and citrate were brought to the metabolic levels of the 6-month-old rats (Supplementary Figure S14 and Supplementary Table S4). It seemed that Zuogui attenuated the metabolism of PMS rats effectively, but it might also indicate that the PMS rats used here were more of kidney yin deficiency than kidney yang deficiency.

CSS (Table 1) is a classical prescription of TCM to treat LQS-associated illness. CSS was documented as early as 1624 in a book of TCM, Jing Yue Quan Shu. The decoction is an effective prescription for ameliorating depression and anxiety (Su et al., 2011; Su et al., 2014).

As shown in Supplementary Figure S15, not much change was observed in the weight of PMS rats treated with CSS as compared to that of the 6-month-old or perimenopausal rats. The changes in the levels of E2 and CORT were not significant either (Figures 6A,B). There was a significant reduction in the level of ACTH however and a considerable increase in the levels of β-EP, 5-HT, and DA in PMS rats treated with CSS, suggesting that CSS altered the endocrine parameters of PMS rats in the favorable direction (Figure 6C).

CSS treatment effectively modulated the metabolomic profile of PMS rats (Figure 6D). It brought the metabolomic state of the PMS rats closer to that of the PM rats but still not to that of the 6 months old (Figure 6E). It upregulated the levels of the unsaturated lipids, isoleucine, and lipid while downregulating the levels of glucose, creatine, betaine, creatinine, glutamine, pyruvate, and acetoacetate. The dominant metabolites for changing the metabolomic state by CSS were glucose and lipid (Supplementary Figure S16). In comparison with the 6-month-old rats, the levels of most of the metabolites were higher, including trimethylamine-n-oxide, betaine, choline, citrate, glutamine, acetoacetate, acetone, methionine, glycoprotein, acetate, valine, isoleucine, leucine, and lipid. Moreover, the levels of glucose, creatine, pyruvate, glycerol, creatinine, and succinate were adjusted to that of the 6 months old Supplementary Figure S17, Supplementary Table S5). Overall, CSS upregulated the metabolism of lipid and downregulated the metabolism of glucose. Considering that LQS rats were with upregulated glucose metabolism and downregulated lipid metabolism, the data is consistent with the fact that CSS is with a tranquilizing potency.

The aforementioned data indicated that both glucose and lipid metabolisms were upregulated in perimenopausal rats. In LQS rats, however, the glucose metabolism was further upregulated, while the lipid metabolism was downregulated. It was previously shown that the glucose and lipid metabolisms were both upregulated for the perimenopausal rats by the treatment with either Yougui or Zuogui (Chen et al., 2020). For the PMS rats though the glucose metabolism was downregulated, the lipid metabolism was upregulated for PMS rats treated with either Yougui, Zuogui, or CSS. It seemed that each of these decoctions was of various degrees of benefits in treating PMS rats based on the metabolomic indicators. However, none of these decoctions was effective enough to bring the metabolomic state of PMS rats to that of rats before entering the perimenopause.

As different decoctions for KD and LQS seem to adjust the metabolomics in different directions and attenuate different sets of metabolites, one would naturally wonder what would be the effect if both KD and LQS were ameliorated simultaneously. To this end, a decoction containing ingredients from Zuogui-CSS was prepared for the gavage of PMS rats.

The weight of PMS rats was not dramatically changed by the treatment with this decoction containing ingredients from Zuogui-CSS (Supplementary Figure S18A). The level of E2 was not significantly altered with the treatment either (Supplementary Figure S18B). The ACTH and CORT levels of peripheral blood dropped after the treatment, but the levels of DA, β-EP, and 5-HT increased (Supplementary Figures S18C,D).

The metabolomic analysis indicated that the combined treatment of Zuogui-CSS was effective for restoring a metabolic state of PMS (Figure 7). Treatment with Zuogui-CSS not only adjusted the metabolomic state of the PMS rats to that of the perimenopausal rats but also brought the metabolomic profile of PMS rats to the proximity of the metabolomic profile of the 6-month-old rats (Figure 7B). Further analysis indicated that Zuogui-CSS modulation of the metabolome of PMS rats was with lipid and glucose as the dominant factors (Supplementary Figure S19). The Zuogui-CSS treatment significantly lowered the levels of glucose, glycerol, trimethylamine-n-oxide, betaine, glutamine, acetoacetate, methionine, glycoprotein, valine, and leucine, but the levels were still higher than those of the 6-month-old rats. In contrast, the pyruvate concentration in PMS rats was even lower than that in the 6-month-old rats after the treatment. Other metabolites, such as creatine, citrate, succinate, acetoacetate, acetone, glycine, and isoleucine, were of similar levels between PMS rats treated with Zuogui-CSS and the 6-month-old rats (Supplementary Figure S20; Supplementary Table S6).

It was possible that the effective attenuation of the metabolomic state of PMS rats without significant change in the level of estrogen was through modulating the activity of estrogen receptors with interacting compounds in Zuogui-CSS. To address this possibility, the ligand-binding domain of estrogen receptor α (ERα LBD) was expressed and purified (Supplementary Figure S21). STD, a ligand-based NMR technique (Mayer and Meyer, 2001), was employed to detect the possible interacting molecules in the decoction. The STD experiment involves subtracting a spectrum, in which the protein was selectively saturated with magnetic radiation, from another spectrum without the protein saturation. In the difference spectrum, only the signals of those compounds that received the magnetic radiation transferred from the protein will remain. Other compounds that do not bind to the protein will not receive the saturation transfer so their characteristic signal will not appear in the difference spectrum (Meyer and Peters, 2003). The techniques can detect K_d values ranging between 10^–3 and 10^–8 M for the interacting compounds and is especially useful for detecting weak binders (Mayer and Meyer, 2001). Many phytoestrogens are with affinities in this range to estrogen receptors. For example, the K_d value for daidzein binding ERα is 5.9 × 10⁻⁶ M (van Lipzig et al., 2004) and that for coumestrol is 1 × 10⁻⁴ M (Qiu et al., 2020). So the technique is applicable to detecting the like compounds in the decoctions if they are of reasonable concentration. The spectrum without the saturation transfer from the protein was obtained for Zuogui-CSS (Figure 8A). In the STD difference spectrum (the upper spectrum in Figure 8B), the only observed peaks were those from glycerol (3.54, 3.63, and 3.77 ppm) and H₂O (4.70 ppm) in the solution (Lu et al., 2018). These were nonspecific peaks due to their high concentrations as they also were present in the spectrum without the protein (lower spectrum in Figure 8B). So, there was no detectable compound by STD in the decoction to interact with ERα.

As these decoctions were through the stomach and intestinal tracts after the gavage, the gastrointestinal digestion could yield interacting compounds. To test this possibility, Zuogui-CSS was treated with SGF and SIF for the STD NMR experiments. The treatment with either SGF or SIF indeed produced new compounds (Figure 9A). The decoction was also first treated with SGF and then with SIF simulating its passage through the gastrointestinal system. The resulting spectrum for the sequential treatment of Zuogui-CSS though was the superposition of spectra from Zuogui-CSS treated with SGF and SIF separately (Figure 9A). These newly generated compounds by the gastrointestinal digestion were with no detectable interaction with ERα LBD either based on the STD NMR analyses (Figures 9B–D).