2022 Volume 45 Issue 1 Pages 104-113
Individual differences in gut microbiota can affect the pharmacokinetics of drugs. Yokukansan is a traditional Japanese kampo medicine used to treat peripheral symptoms of dementia and delirium. A study examining the pharmacokinetics of the components of yokukansan reported large individual differences in the pharmacokinetics of glycyrrhizic acid (GL). It is known that GL is metabolized by intestinal bacteria to glycyrrhetinic acid (GA), which is absorbed in the gastrointestinal tract. Thus, the gut microbiota may affect GL pharmacokinetics. We aimed to clarify the relationship between the gut microbiota composition and pharmacokinetics of GL in yokukansan. Mice were orally administered yokukansan, following the administration of various antibiotics, and the plasma concentration of GA and composition of gut microbiota were measured. The GA plasma concentration was low in mice treated with amoxicillin and vancomycin. The composition of gut microbiota revealed a different pattern from that of the control group. Mice with low plasma levels of GA had lower levels of the phylum Bacteroides and Firmicutes. Additionally, bacteria, such as those belonging to the genera Parabaceroides, Bacteroides, Ruminococcus and an unknown genus in families Lachnospiraceae and Ruminococcaceae, exerted positive correlations between the gene copies and plasma GA levels. These bacteria may contribute to the absorption of GA in the gastrointestinal tract, and multiple bacteria may be involved in GL pharmacokinetics. The pharmacokinetics of GL may be predicted by evaluating the composition of gut bacteria, rather than by evaluating the amount of a single bacterium.
Individual differences exist among the types, abundance, and health effects of bacteria that make up the gut microbiota.1,2) Intestinal bacteria affect the metabolism and absorption of drugs and nutrients.3,4) Previous studies have reported that gut microbiota modulates drug pharmacokinetics by metabolizing drugs to a form that is easily absorbed in the digestive tract.5,6) Thus, individual differences in the balance of gut microbiota may cause differences in pharmacokinetics.
“Kampo” medicines, which are traditional Japanese medicines, are composed of several natural crude drugs. Interestingly, in vivo pharmacological studies have not detected a considerable number of compounds present in the constituent crude drugs of kampo medicines.7–9) This indicates that these compounds undergo metabolism and exert their actions in the target organs. In many cases, active compounds are present as glycosides in the crude drugs. In general, glycosides are hydrophilic, and they are not easily absorbed in the digestive tract. Glycosides are metabolized by intestinal bacteria to highly lipophilic aglycones, which are further absorbed in the digestive tract.10,11) Therefore, the pharmacokinetics of glycosides may be affected by the gut microbiota.
Yokukansan is a kampo medicine. It is used to treat irritability, children’s night crying, and peripheral symptoms of dementia and delirium.12,13) Thus, it is frequently used in the elderly or pediatric patients. However, some adverse events such as pseudoaldosteronism have been reported.14) Therefore, it is clinically important to control the efficacy and adverse effects of yokukansan, and it is necessary to understand the pharmacokinetics of its components. A pharmacokinetic study on yokukansan reported large individual differences in the plasma concentrations of individual components15); however, the reason for this remains unclear.
Glycyrrhizic acid (GL), one of the main components of yokukansan, is a glycoside that is metabolized by intestinal bacteria to glycyrrhetinic acid (GA) (Fig. 1), which is further absorbed in the digestive tract.10) GL has anti-inflammatory and antioxidant effects and has been reported to be effective against cognitive dysfunction.16–19) Therefore, predicting the pharmacokinetics of GL is important for predicting the efficacy of yokukansan.
Some studies have evaluated the relationship between the pharmacokinetics of GL and gut microbiota. In a previous study, the metabolic activity of GL was reported in stools and the blood GA concentration in rats decreased with the knockdown of intestinal bacteria following antibiotic administration.20) Some other studies have presented the bacteria involved in GL pharmacokinetics. Certain bacteria such as Ruminococcus sp., Clostridium innocuum, and Eubacerium sp. metabolize GL to GA in vitro.21–23) In the actual gastrointestinal tract, multiple bacteria are expected to be associated with the metabolism of GL. Therefore, it is considered necessary to focus on multiple bacteria in the gut to predict the pharmacokinetics of GL.
Therefore, we aimed to clarify the relationship between the gut microbiota composition and pharmacokinetics of GL in yokukansan. Mice were orally administered antibiotics having different spectra to alter the gut microbiota. Following this, the mice were orally administered yokukansan, and GA pharmacokinetics and the abundance of each intestinal bacterium were evaluated.
Yokukansan24) extract (Tsumura & Co., Tokyo, Japan, lot 382059900) was used as the investigational product. The three-dimensional HPLC pattern of yokukansan used in this study is demonstrated in Supplementary Fig. 1 (data provided by Tsumura & Co.). We obtained 3.25 g of yokukansan extract via spray drying a hot water extract of a mixture of seven crude drugs, namely, 4 g of Atractylodis Lanceae Rhizoma (Atractylodes lancea Rhizome), 4 g of Poria (Poria Sclerotium), 3 g of Cnidii Rhizoma (Cnidium Rhizome), 3 g of Uncariae Uncis Cum Ramulus (Uncaria Hook), 3 g of Angelicae Acutilobae Radix (Japanese Angelica Root), 2 g of Bupleuri Radix (Bupleurum Root), and 1.5 g of Glycyrrhizae Radix (Glycyrrhiza). The GL content in 1 g of yokukansan extract was 5.8 ± 0.2 mg. The measurement method is presented as follows: Yokukansan hot water extract (1 g of preparation) was extracted by methanol extraction using 20 mL of methanol for 30 min under sonication. The extract (20 µL) was injected into a Shimadzu LC-20 system (Shimadzu Corporation, Kyoto, Japan), consisting of a Shimadzu LC-20 AR HPLC pump, Shimadzu series DGU-20A3R degasser, Shimadzu SIL-20 A autosampler, and TSK gel ODS-80TS (250 × 4.6 mm, Tosoh, Tokyo, Japan) maintained at 40 °C. Mobile solution A comprised 0.2 vol % formic acid/water, and mobile phase B solution was acetonitrile. The mobile phase consisted of solution A and solution B with a gradient of solution B (0 min, 10%; 30 min, 95%; 33 min, 95%) at a flow rate of 0.7 mL/min. A wavelength of 254 nm was measured.
Niflumic acid and diclofenac, which were used as internal standards in this study were purchased from Sigma-Aldrich Co. (St. Louis, MO, U.S.A.) and Wako Co. (Osaka, Japan), respectively. GL and GA were purchased from Wako Co.
Study DesignAll the animal experiments were conducted in accordance with the NIH guidelines and approved by the Institutional Animal Care and Use Committee (IACUC) of Kochi University (Approval No. M-00033). C57BL/6 mice (male, 7-week-old) were purchased from Japan SLC (Shizuoka, Japan), and they were divided into six groups of 28 mice each. Each group was orally administered 100 mg/kg amoxicillin (AMPC), 50 mg/kg metronidazole (MNZ), 100 mg/kg neomycin (NM), or 50 mg/kg vancomycin (VCM). One of the six groups (Mix) received the same dose of all four antibiotics (AMPC, MNZ, NM, and VCM) orally. Distilled water was administered orally to the control group. Antibiotics were administered to mice twice every 24 h. After 1 h of the second antibiotic administration, 2 g/kg yokukansan extract (Tsumura) was orally administered to all the mice, simultaneously. Prior to sacrifice, stool samples were obtained prior to operation from seven mice in each group and were stored at −80 °C until use. They were used for fecal glycoside metabolic activity evaluation and microbiome analysis. Blood was collected from seven mice in each group following 4, 8, 12, and 24 h of a single oral administration of yokukansan extract. It was collected from the descending vena cava in a blood collection tube containing heparin (Mochida, Tokyo, Japan), which was centrifuged at 1200 × g at 4 °C for 30 min. The obtained plasma was stored at −80 °C and used for blood GA concentration measurement. The gut microbiota composition of stool samples from mice treated with each antibiotic (4, 8, 12, 24 h) are presented in Supplementary Fig. 2.
Measurement of Bacterial Metabolic Ability Using GL and HPLCBacterial metabolic activity was measured using GL, as previously described.11,20) Stool samples were weighed and suspended in 100 mg/mL phosphate buffer solution (50 mM, pH 7). Thirty microliters of the suspension were mixed with 270 µL of Gifu anaerobic medium (Nissui, Tokyo, Japan) and 3 µL of 10 mM GL solution. Following this, the mixture was incubated at 37 °C for 12 h under anaerobic conditions. Anaero Pack Kenki (Sugiyamagen, Tokyo, Japan) was used as the anaerobic culture system. Reaction solutions were extracted with 900 µL n-butanol (3 × 300 µL each) to ensure a complete extraction of GL and its metabolites and dried under nitrogen. Dried components were resuspended in 1 mL of 50 vol % methanol/water, and 50 µL of 5 mM diclofenac was added as an internal standard. The solution was applied to a Supel-Select HLB column (Sigma), preconditioned with 1 mL methanol and 1 mL water. Following this, the column was washed with 1 mL of 1 vol % phosphoric acid/water, 1 mL of water, and 1 mL of 40 vol % methanol/water, and elution was performed with 1 mL of 1 vol % acetic acid/methanol. The extracted solution was applied to a Supel-Select SAX column (Sigma), preconditioned with 1 mL of methanol and 1 mL of water. The column was further washed with 1 mL of 50 vol % methanol. Elution was performed with 1 mL of 0.1% formic acid/methanol. The extracted solution was dried under nitrogen. Dried components were resuspended in 200 µL of 50 vol % methanol/water. An aliquot of 20 µL was injected into a Shimadzu LC-20 system (Shimadzu Corporation), consisting of a Shimadzu LC-20 AR HPLC pump, Shimadzu series DGU-20A3R degasser, Shimadzu SIL-20 A autosampler, and Cosmosil 5C18-AR-II column (150 × 4.6 mm, 5 mm, Nacalai Tesque, Kyoto, Japan) maintained at 40 °C. Mobile solution A comprised 0.2 vol % formic acid/water, and mobile phase B solution was acetonitrile. The mobile phase consisted of solution A and solution B with a gradient of solution B (0 min, 43%; 9 min, 57%; 13 min, 95%; 20 min, 95%) at a flow rate of 1 mL/min. The wavelength of 254 nm was measured. The metabolic activity of GL was expressed as the amount of GA produced per stool weight (µmol/mg stool). Each stool sample treated without GL was used as a blank and subtracted from the obtained data to calculate the amount of GA production.
Gut Microbiota Analysis and 16S Ribosomal RNA (rRNA) Gene Metagenome Sequencing in Stool SamplesBacterial genomic DNA was isolated using a standard method with slight modifications.25) Briefly, stool samples were freeze-dried and weighed using 10–30 mg Lysing matrix E tubes (MP Biomedicals, LLC., Santa Ana, U.S.A.) and homogenized twice with elution buffer using the FastPrep-24 automated cell disruptor (MP Biomedicals, LLC) at 6 m/s for 40 s. Stool DNA was extracted using the phenol/chloroform/isoamyl alcohol method. The 16S rRNA gene metagenome library for MiSeq (Illumina Inc., San Diego, U.S.A.) was prepared according to the manufacturer’s protocol. The library was applied to the MiSeq Reagent Kit v3 (Illumina Inc.), and it was sequenced according to the manufacturer’s protocol. The sequence data were processed using the 16S rRNA sequence analysis pipeline QIIME 1.9.0.26) First, both sides of the sequences were joined, and sequences with a Phred quality score >20 were removed as poor quality data. Chimella elimination was performed using U-search, and contaminated sequences were removed from the dataset. Open reference operational taxonomic unit (OTU) picking was performed against green gene 97 13_8 as the reference dataset. A summary of the taxonomy in each sample was obtained using script ‘summarize_taxonomy_through_plots.py’ in QIIME 1.9.0. To measure the 16S rRNA gene copy number, quantitative PCR (qPCR) was performed using universal primer for all bacteria.27) The primers were purchased from Thermo Fisher Scientific, Inc. (MA, U.S.A.) as custom primers. To prepare standard curve, equal volume of stool DNA sample was mixed as a template. Target 16S rRNA amplicon PCR was performed using universal primer for all bacteria,27) and the single amplified fragment (292 bp) was purified using Agencourt AMPure XP (Beckman Coulter Inc., Indianapolis, U.S.A.). The accurate concentration of double stranded DNA was measured by Quant-iT PicoGreen dsDNA assay kit (Thermo Fisher Scientific, Inc.) to calculate the copy number and used as a standard curve material. DNA extract from stool samples was used as template. Reactions were performed by a standard method using a SYBR Green PCR Kit (Thermo Fisher Scientific, Inc.) and QuantStudio 7 Flex Real-time PCR system (Thermo Fisher Scientific, Inc.), and the total 16S gene copy number was determined by standard curve. The copy number in each microbe was calculated by multiplying the total 16S gene copy number by relative abundance in each microbe.
Blood GA Concentration Measurement and Liquid Chromatography Tandem Mass Spectrometry (LC/MS/MS) ConditionsPlasma samples (50 µL) were mixed with 5 µL of ethanol, 100 µL of solution (0.1 vol % formic acid/water), and 10 µL of 20 µg/mL niflumic acid as an internal standard. Additionally, they were mixed with 500 µL of tert-butyl-methyl-ether, vortexed, and centrifuged at 10000 × g for 5 min. Subsequently, the supernatant was collected and dried under nitrogen. Dried components were resuspended in 150 µL of 50 vol % acetonitrile/water and used as the measurement sample. A 5 µL of aliquot was injected into a LC/MS/MS system consisting of an LC-20AD system (Shimadzu Corporation) and QTRAP5500 (SCIEX, Tokyo, Japan). The mass spectrometer was operated in positive ionization mode, and data were obtained using Analyst 1.6.2. software (SCIEX). Chromatographic analyses were performed using a CAPCELL PAK C18 MGII column (3 µm, 2.0 × 50 mm; Osaka Soda, Osaka, Japan) maintained at 50 °C, and the autosampler tray temperature was maintained at 4 °C. The mobile phase consisted of solution A (0.5/5/900/100 vol % formic acid/1.0 M ammonium formate aqueous solution/water/acetonitrile) and solution B (0.5/950/50 vol % formic acid/acetonitrile/methanol) with a gradient of solution B (0 min; 50%, 1.1 min; 50%, 1.6 min; 95%, 2.1 min; 95%, 2.2 min; 50%) at a flow rate of 0.5 mL/min.
The LC system was coupled with a QTRAP5500 mass spectrometer equipped with an electrospray ionization source. Quantification was performed using the multiple reaction monitoring method with transitions of m/z 471.4→149.1 and m/z 283→245 for GA and niflumic acid (internal control), respectively. Nitrogen was used as the nebulizer, heater, curtain gas, and collision activation dissociation gas. The mass spectrometric parameters were set as follows: Curtain gas, 30 psi; collision gas, 6 psi; ionspray, 4000 V; temperature, 600 °C; nebulizing gas (gas1), 60 psi; and heater gas (gas2), 60 psi. The method was validated by peak selectivity, calibration curve linearity, and intra- and inter-day reproducibility under mouse plasma matrix conditions.
For peak selectivity, chromatograms of plasma from 6 mice confirmed that there were no contamination peaks near the retention time of GA and internal standard. The calibration curve of GA was prepared at concentrations of 2, 5, 10, 20, 50, 200, and 500 ng/mL by spiking blank plasma with GA standard solution. The calibration curve revealed linearity in the concentration range of 2–500 ng/mL with correlation coefficients r2 > 0.997. The precision and accuracy were determined at four different concentrations (i.e., 2, 5, 50, 400 ng/mL) on three separate days. The intra-assay precision ranged from 5.1 to 10.7% with the accuracy ranging from 100.8 to 106%. The inter-assay precision ranged from 5.5 to 10.7% with the accuracy ranging from 98.5 to 100.6%.
Statistical AnalysisAll statistical analyses were performed using the Stat Flex program (View Flex, Tokyo, Japan). Descriptive statistical analysis was used to calculate means, standard deviations, medians, and interquartile ranges. To determine the significant differences in plasma GA concentrations in each group, two-way ANOVA was used for variance analysis. The Tukey test was used as a post-hoc test. To evaluate glycoside metabolic activity and each bacterial 16S rRNA copy, the Kruskal–Wallis test was used for variance analysis. Subsequently, the Steel–Dwass test was used as a post-hoc test.
Correlation between the plasma GA concentrations and each bacterial 16S rRNA copy was evaluated using the Spearman's test. Statistical significance was set at p < 0.05.
Data on bacterial 16S rRNA copies, plasma GA concentrations, and fecal glycoside metabolic activities were converted to Z-scores, and they are presented in the heat map. The bacterial 16S rRNA copies were converted to logarithmic values prior to calculation. The Z-score was calculated in each row based on the data of all mice used in this study (zero was the average value of all mice).
Z-score was calculated as follows: Z-score(i) = (value in valuable(i)−mean value in valuable(i))/standard deviation of valuable(i)), where i = 16S rRNA gene copy number in each bacterium, plasma GA concentration, or fecal glycoside metabolic activity.
β-Diversity analysis was performed by non-metric multidimensional scaling (NMDS) using the Bray−Curtis dissimilarity “metaMDS” in package “vegan”28) in R 4.0.2 (The R Foundation Conference Committee). The hierarchical cluster analysis was performed using “hclust” in package “stats” in R 4.0.2. The distance between each variable used the Bray−Curtis dissimilarity indices, and the distances between each cluster were obtained by Ward’s method.
There was no apparent abnormality in each group. Yokukansan extract was orally administered to antibiotic-treated mice, and the plasma GA concentration was examined over time (Fig. 2). In the control group, GA was detected in the plasma following 4, 8, and 12 h of yokukansan extract administration. The concentrations were 39.6 ± 11.8, 14.9 ± 6.8, and 17.5 ± 8.6 ng/mL at 4, 8, and 12 h, respectively. In the control group, the highest plasma GA concentration was revealed at 4 h following administration of yokukansan extract, and the profile revealed double peak phenomenon, which reflected the enterohepatic circulation of GA. The double peak phenomenon was also revealed in NM group while MNZ group presented large individual diversity of plasma GA concentration. No significant change was observed between the GA plasma concentration at each time point in MNZ and NM groups and control group. However, the GA plasma concentrations at 4, 8, and 12 h were significantly lower in AMPC, VCM, and Mix groups than in the control group. The GA concentrations of AMPC group were 0.7 ± 1.8 ng/mL (p = 0.019) and 1.1 ± 2.9 ng/mL (p = 0.020) at 8 and 12 h, respectively, and they were not detected at 4 h. GA concentrations in the VCM group were not detected at all times. The GA concentrations of the Mix group were 0.4 ± 1.0 ng/mL (p = 0.002) and 1.7 ± 11.8 ng/mL (p = 0.047) at 4 and 8 h, respectively, and they were not detected at 12 h. The two-way repeated measures ANOVA results indicated that these effects were significant with time (F4,166 = 6.72, p < 0.001), group (F5,166 = 4.96, p = 0.001), and interactions between time and group (F20,166 = 2.34, p = 0.002).
A single antibiotic (MNZ, NM, AMPC, or VCM), all four antibiotics (Mix), or distilled water was administered to each mouse twice every 24 h. One hour following the second antibiotic administration, yokukansan extract was orally administered to all the mice. Blood was collected from mice in each group following 0, 4, 8, 12, and 24 h of a single oral administration of yokukansan extract. The GA concentration in plasma was measured using LC/MS/MS. The data are expressed as mean ± standard deviation (S.D.) (n = 6–7), and they were evaluated using two-way repeated ANOVA, followed by the Tukey’s test. ** p < 0.01 vs. control group. When GA was not detected, it was described as ND.
Stool obtained from mice 4 h following yokukansan administration was incubated with GL, and GA production ability was analyzed (Fig. 3). No significant change was observed in NM (p = 0.16) and MNZ groups (p = 0.70) than in the control group. However, the GA production ability was significantly lower in the AMPC (p = 0.032), VCM (p = 0.031), and Mix groups (p = 0.032) than in the control group. The Kruskal–Wallis test results indicated the significant effect of antibiotics (df = 5, X2 = 32.2, p < 0.001).
A single antibiotic (MNZ, NM, AMPC, or VCM), all four antibiotics (Mix), or distilled water was administered to each mouse twice every 24 h. One hour following the second antibiotic administration, yokukansan extract was orally administered to all the mice. Stools were collected from mice following 4 h of a single oral administration of yokukansan extract. Stool suspensions were incubated with GL, GA production levels were measured using HPLC, and GA production ability (nmol/mg stool) was calculated. Data are expressed as medians and interquartile ranges, and they were evaluated using the the Kruskal–Wallis test, followed by the Steel-Dwass test. ** p < 0.01 vs. control group.
Stool microbiota samples from antibiotic-administered mice were analyzed using 16S rRNA metagenome sequencing. The number of trimmed, qualified reads in each sample were 31105 ± 7895. The total OTU detected in all samples were 40730. The changes in the relative abundance of microbiota at the phylum and genus levels are presented in Figs. 4A and B, respectively. The shape of microbiota in the MNZ and NM groups revealed a tendency to increase in the phylum Bacteroidetes. In the AMPC and VCM groups, there was a tendency for a decrease in the phylum Bacteroidetes and increase in the phylum Proteobacteria.
A single antibiotic (MNZ, NM, AMPC, or VCM), all four antibiotics (Mix), or distilled water was administered to each mouse twice every 24 h. One hour following the second antibiotic administration, yokukansan extract was orally administered to all the mice. Stools were collected from mice following 4 h of a single oral administration of yokukansan extract. The relative abundance of gut microbiota was determined using 16S metagenome sequence analysis. The average relative abundance is presented as bar charts of (A) phylum and (B) genus levels (n = 6–7). (C) β-diversity analysis of each mouse gut microbiota composition analyzed for 16S rRNA gene copies of each bacterium is presented. The distance between each variable was calculated using the Bray − Curtis dissimilarity index, and the distance between each cluster was calculated using Ward's method. (D) The 16S rRNA gene copies of each bacterial species in the plots are color-coded from less (blue) to more abundant (red) in the heat map. The heatmaps were color-coded based on the row Z-score calculated from the logarithmic value of 16S rRNA gene copies in stools.
Since antibiotic treatment was thought to affect not only the proportion of gut microbiota, but also the number of microbes, we measured the copy number of 16S rRNA using universal primer. These data are presented in Supplementary Fig. 3. Except for the Mix group, the copy number of microbes in each group was not significantly different compared with that in the control group throughout the experiment. In the Mix group, the copy number of microbes was significantly decreased at 8 and 24 h following administration of antibiotics compared with that in control group.
The distribution of microbes in each group was evaluated using non-metric multidimensional scaling and cluster analysis using the Bray–Curtis dissimilarity distances (Fig. 4C). The AMPC, VCM and Mix groups formed a cluster placed farther from the control group than from the MNZ and NM groups.
The 16S rRNA gene copies of bacteria in each group at 4 h following administration of yokukansan were compared at the phylum level, whose results are presented in Fig. 5. In the phylum Bacteroidetes, the 16S rRNA gene copy number was significantly elevated in the NM group (p = 0.046) and significantly decreased in the AMPC (p = 0.048), VCM (p = 0.032), and Mix groups (p = 0.032) than in the control group. No significant change was observed in the MNZ group (p = 0.39). In the phylum Firmicutes, it was significantly more decreased in the Mix group (p = 0.032) than in the control group. There was a tendency for decrease in the VCM group (p = 0.072). No significant change was observed in the MNZ (p = 0.99), NM (p = 0.99), or AMPC groups (p = 0.99). In the phylum Proteobacteria, the 16S rRNA gene copy number was significantly more increased in the MNZ (p = 0.046), AMPC (p = 0.032), and VCM groups (p = 0.032) than in the control group. There was a tendency for an increase in the NM group (p = 0.072). No significant change was observed in the Mix group (p = 0.99). The Kruskal–Wallis test results indicated the significant effects of antibiotics (Bacteroidetes: df = 5, X2 = 30.5, p < 0.001, Firmicutes: df = 5, X2 = 24.8, p < 0.001, Proteobacteria: df = 5, X2 = 29.2, p < 0.001).
A single antibiotic (MNZ, NM, AMPC, or VCM), all four antibiotics (Mix), or distilled water was administered to each mouse twice every 24 h. One hour following the second antibiotic administration, yokukansan extract was orally administered to all the mice. Stools were collected from mice following 4 h of a single oral administration of yokukansan extract. 16S rRNA gene copies in each phylum were determined using 16S metagenome sequence analysis and qPCR. The 16S rRNA gene copies of (A) phylum Bacteroidetes, (B) phylum Firmicutes, and (C) phylum Proteobacteria were compared in each group (n = 6–7). These data are expressed as medians and interquartile ranges, and they were evaluated using the Kruskal–Wallis test, followed by the SteelDwass test. * p < 0.05, vs. control group.
The relationship between gut microbes and plasma GA levels after 4 h of antibiotic administration was speculated (Fig. 6). In the phylum Bacteroidetes, the genera Prevotella in the family Paraprevotellaceae (r = 0.68, p < 0.001), Bacteroides (r = 0.63, p < 0.001), Parabacteroides (r = 0.63, p < 0.001), Prevotella (r = 0.82, p < 0.001), an unknown genus in the order Bacteroidales (r = 0.65, p < 0.001), an unknown genus in the family Rikenellaceae (r = 0.79, p < 0.001), and an unknown genus in the family S24-7 (r = 0.78, p < 0.001) exerted a positive correlation between the gene copies and plasma GA levels. In phylum Firmicutes, the genera Lactobacillus (r = 0.38, p = 0.016), Ruminococcus in the family Lachnospiraceae (r = 0.81, p < 0.001), Coprococcus (r = 0.71, p < 0.001), an unknown genus in the family Lachnospiraceae (r = 0.81, p < 0.001), Oscillospira (r = 0.84, p < 0.001), an unknown genus in the family Ruminococcaceae (r = 0.81, p < 0.001), Ruminococcus (r = 0.80, p < 0.001), Coprobacillus (r = 0.63, p < 0.001), and an unknown genus in the family Erysipelotrichaceae (r = 0.77, p < 0.001) exerted a positive correlation, and an unknown genus in the family Enterococcaceae (r=−0.35, p = 0.026) exerted a negative correlation between the gene copies and plasma GA levels. In the phylum Proteobacteria, the genus Sutterella (r = 0.37, p = 0.019) exerted a positive correlation, and an unknown genera in the family Enterobacteriaceae (r=−0.55, p < 0.001), Klebsiella (r=−0.56, p < 0.001), and Morganella (r=−0.50, p < 0.001) exerted negative correlations between the gene copies and plasma GA levels. In other phyla, the genus Akkermansia in the phylum Verrucomicrobia (r = 0.54, p < 0.001) revealed a positive correlation between the gene copies and plasma GA levels.
Stool samples were collected from mice following 4 h of a single oral administration of yokukansan extract. The 16S rRNA gene copies of bacteria were determined using 16S metagenome sequence analysis and qPCR. The correlation between 16S rRNA gene copies of each bacteria in stools and GA concentration in plasma following 4 h of yokukansan extract administration was evaluated using the Spearman’s rank correlation coefficient.
This is the first study that investigated the gut microbiota composition and pharmacokinetics of GL in yokukansan using experimental animals. This study suggested that oral antibiotics affect the plasma levels of GA and blood level of GA changes depending on the gut microbiota composition. Furthermore, the GA plasma levels were decreased in the mice decreased in the phyla Bacteroides and Firmicutes, suggesting that these bacteria may affect GL pharmacokinetics.
Glycosides are often included as active ingredients in natural medicines. Many glycosides are highly water-soluble and metabolized by intestinal bacteria prior to being absorbed in the digestive tract.5) For example, ginsenoside Rb1 in Panax ginseng is a glycoside. A continuous administration of Daikenchuto, a kampo medicine containing Panax ginseng, affects the gut microbiota and increases the glycoside metabolic activity of gut microbiota and its blood concentration.11) Thus, the pharmacokinetics of glycosides are influenced by intestinal bacteria.
Yokukansan, one of the most frequently used kampo medicines, contains several glycosides that contribute to its pharmacological effects.29) In a previous pharmacokinetic study of yokukansan, a large difference in the blood concentration of GL was observed,15) albeit the reason behind this difference was not clarified. GL, a glycoside present in yokukansan, contains two glucuronic acids.30) GL is metabolized to GA by the intestinal bacterium β-glucuronidase (Fig. 1) and subsequently, absorbed in the digestive tract.10,31) Therefore, it is presumed that the pharmacokinetics of GL is related to its metabolism to GA by gut microbiota. In this study, as a part of the purpose of clarifying the differences in pharmacokinetics of the components of yokukansan, we examined the relationship between pharmacokinetics of GL and intestinal bacteria. The results revealed lower GA plasma levels and lower metabolizing activities of GL in the stools of AMPC, VCM, and Mix groups than those in the control group (Figs. 2, 3). This suggests that the metabolic activity of gut microbiota may affect the bioavailability of GL. Previous studies have examined the reduction of GA absorption by knockdown of intestinal bacteria20) and examination of bacteria that metabolize GL to GA in vitro21–23); however, no studies have examined the intestinal bacterial composition and pharmacokinetics of GL. In this study, the GA plasma levels were lower in the AMPC, VCM, and Mix groups than those in the control group, whereas the levels in the MNZ and NM groups did not change significantly as compared to those in the control group (Fig. 2). This suggests that GL pharmacokinetics differ depending on the antibiotic spectrum. In addition, the composition of intestinal bacteria in each group was confirmed as follows: NM and MNZ presented a composition similar to that of the control group, while the AMPC, VCM, and Mix groups presented a composition different from that of the control group (Fig. 4C). Antibiotics may alter the pharmacokinetics of GL by changing the composition of intestinal bacteria. Additionally, oral administration of AMPC and VCM caused low blood levels of GA. These antibiotics are also used orally in clinical practice. They may alter the pharmacokinetics of GL and reduce the efficacy of medicines containing GL when used in combination.
Metabolites of GL such as 3-monoglucuronyl-glycyrrhetinic acid and 22α-hydroxy-18β-glycyrrhetyl-3-O-sulfate-30-glucuronide inhibit 11β-hydroxysteroid dehydrogenase type 2 (11β-HSD2), leading to pseudoaldosteronism.32–34) Therefore, predicting the pharmacokinetics of GL is important for predicting the adverse effects of yokukansan. Recent studies have suggested that 18β-glycyrrhetol-3-O-sulfate, which is formed by sulfate conjugation of GA, plays a major role in the development of pseudoaldosteronism, and its high serum concentration is related to lower renin, aldosterone, and potassium levels.32,35) In contrast, GA may not be a direct cause of pseudoaldosteronism.35,36) However, it has been reported that higher doses of licorice cause a higher risk of developing pseudoaldosteronism, which may be related to the dose of GL.37) It has been reported that GA can be metabolized to 18β-glycyrrhetol-3-O-sulfate by SULT2A1 in the liver, and its serum levels positively correlate with serum levels of 18β-glycyrrhetol-3-O-sulfate.35) Thus, the greater the amount of GA absorbed, the greater the serum concentration of 18β-glycyrrhetol-3-O-sulfate will be. To discuss how the altered pharmacokinetics of GL by gut microbiota affects the risk of developing pseudoaldosteronism, it is necessary to examine GL metabolites such as 18β-glycyrrhetyl-3-O-sulfate in the future.
The intestinal bacteria in the group with low blood levels of GA was compared to that in the control group. In the AMPC group, the 16S rRNA gene copies of the phylum Bacteroidetes decreased (Fig. 5). In the VCM group, the 16S rRNA gene copies of the phylum Firmicutes and Bacteroidetes decreased (Fig. 5). It has been reported that Ruminococcus sp., Clostridium innocuum, and Eubacterium sp. in the phylum Firmicutes21–23) and Bacteroides J-37 in the phylum Bacteroidetes metabolize GL to GA.38,39) Although the processes leading to the metabolism of GL have not yet been studied, it has been reported that in the families Lachnospiraceae and Ruminococcaceae in the phyrum Firmicutes40,41) and some bacteria in Bacteroides sp. and Parabacteroides sp. in the phylum Bacteroidetes have strong β-glucuronidase activity.39,40) In this study, some bacteria in the phylum Firmicutes and Bacteroidetes, such as the genera Ruminococcus, Bacteroides, Parabacteroides, and an unknown genus in families Lachnospiraceae and Ruminococcaceae were positively correlated with GA plasma levels (Fig. 6). A number of these bacteria may contribute to the differences in the gut bacterial composition compared with those in the control group, and changes in the number of these bacteria may affect the pharmacokinetics of GL.
Neomycin has no antibacterial activity against Bacteroidetes.42) In this study, 16S rRNA gene copies of phylum Bacteroidetes were higher in the NM group than those in the control group (Fig. 5). However, the GA plasma levels did not increase in the NM group as compared to those in the control group. This suggests that GL pharmacokinetics may not be related to only the phylum Bacteroidetes but also to other bacteria, such as those belonging to the phylum Firmicutes. Mice with high GA plasma levels possessed high 16S rRNA gene copies of the genus Ruminococcus, Bacteroides, or Parabacteroides (Fig. 4D). Therefore, the pharmacokinetics of GL may depend on multiple bacteria and not just on one of these bacteria.
Gut bacteria have been reported to be associated with the metabolism of nutrients and drugs.3,4) Therefore, it is expected that the study of intestinal bacteria is applied to the prediction of disease incidence. For example, it has been reported that the ratio of phylum Bacteroidetes and Firmicutes is associated with a high probability of obesity and thus, expected to be a biomarker for obesity.43,44) However, in GL pharmacokinetics, it is difficult to infer from a single bacterium and may be important to observe multiple enterobacteria as the composition of gut microbiota. Thus, in this study, we classified the antibiotic-treated mice into clusters based on the copy number of 16S rRNA of all detected bacteria. All the mice with high absorption of GA were classified into close groups. Consequently, it may be possible to predict the pharmacokinetics of GL by comprehensive analysis of gut microbiota before administering yokukansan to the patients. This may also be possible for other glycosides.
This study had some limitations, which are as follows: 1) We used experimental animals in this study and it was necessary to verify the results in humans in clinical trials. The intestinal flora of mice and humans differ. Although the abundance of each bacterium differs between mice and human gut microbiota, the types of bacteria present in the gut are qualitatively similar. Parabaceroides sp., Bacteroides sp., and Ruminococcus sp., which were analyzed in this study, are also present in the human intestine.45) 2) Regarding the pharmacokinetics of GL, blood sampling was not performed frequently. Thus, the maximum blood concentration, arrival time, and area under the curve of plasma GA levels could not be evaluated in this study. These questions need to be examined in the future studies. 3) We believe that it is necessary to clarify whether the same phenomenon occurs in other GL-containing medicines, since the experiment was conducted using only yokukansan in this study. Although there have been no reports on yokukansan, daikenchuto, one of the kampo medicines, has been reported to change the composition of gut bacteria.11) Therefore, yokukansan possibly affects the composition of the intestinal bacteria. In contrast, since no change in the composition of gut bacteria was observed following a single administration of yokukansan (Supplementary Fig. 2), we believe that the effect of yokukansan on gut bacteria maybe small. 4) The doses of antibiotics and yokukansan used in this study may be higher than those used for humans in clinical practice. These doses were determined by referring to previous pharmacokinetic studies on kampo medicine in experimental animals.20,46) The doses of antibiotics were set at approximately 12 times the human daily dose per body weight by considering the conversion based on the body surface area from human to mice. For yokukansan, the daily clinical dose for humans as a base powder was 3.25 g, 12 times of which was about 0.65 g/kg. The dosage of 2 g/kg was increased by considering the conversion based on the body surface area from human to mice to detect sufficient GA in the plasma of yokukansan-treated mice. It needs to be verified in future whether the same results can be obtained with normal doses of the drug in clinical practice.
This study is the first to verify the relationship between the pharmacokinetics of GL in yokukansan and composition of gut microbiota in experimental animals. The results suggested that there are multiple bacteria that may affect the metabolism of GL. Although this study was an animal experiment, it will need to be re-evaluated for clinical trials. It was suggested that the pharmacokinetics of GL may be predicted by evaluating the composition of gut bacteria, rather than by evaluating the amount of a single bacterium. For other glycosides, multiple gut bacteria are also expected to be associated with their metabolism in the gastrointestinal tract. During evaluating the relationship between the pharmacokinetics of glycosides and gut bacteria, it may be necessary to investigate the composition of gut bacteria.
Participated in research design: Ishida, Jobu, Ogawa, Hanazaki, Miyamura.
Conducted experiments: Ishida, Jobu, Kawada, Morisawa, Kawazoe, Shiraishi, Fujita, Nishimura, Kanno, Nishiyama, Morita.
Performed data analysis: Ishida, Jobu, Kanno, Nishiyama.
Wrote or contributed to the writing of the manuscript: Ishida, Jobu, Kawada, Hanazaki, Miyamura.
The authors declare no conflict of interest.
The online version of this article contains supplementary materials.
