Multiple circulating alkaloids and saponins from intravenous Kang-Ai injection inhibit human cytochrome P450 and UDP-glucuronosyltransferase isozymes: potential drug–drug interactions

Background Kang-Ai injection is widely used as an adjuvant therapy drug for many cancers, leukopenia, and chronic hepatitis B. Circulating alkaloids and saponins are believed to be responsible for therapeutic effects. However, their pharmacokinetics (PK) and excretion in vivo and the risk of drug–drug interactions (DDI) through inhibiting human cytochrome P450 (CYP) and UDP-glucuronosyltransferase (UGT) enzymes remain unclear. Methods PK and excretion of circulating compounds were investigated in rats using a validated ultra-high-performance liquid chromatography tandem mass spectrometry (UHPLC–MS) method. Further, the inhibitory effects of nine major compounds against eleven CYP and UGT isozymes were assayed using well-accepted specific substrate for each enzyme. Results After dosing, 9 alkaloids were found with Cmax and t1/2 values of 0.17–422.70 μmol/L and 1.78–4.33 h, respectively. Additionally, 28 saponins exhibited considerable systemic exposure with t1/2 values of 0.63–7.22 h, whereas other trace saponins could be negligible or undetected. Besides, over 90% of alkaloids were excreted through hepatobiliary and renal excretion. Likewise, astragalosides and protopanaxatriol (PPT) type ginsenosides also involved in hepatobiliary and/or renal excretion. Protopanaxadiol (PPD) type ginsenosides were mainly excreted to urine. Furthermore, PPD-type ginsenosides were extensively bound (fu-plasma approximately 1%), whereas astragalosides and PPT-type ginsenosides displayed fu-plasma values of 12.35% and 60.23–87.36%, respectively. Moreover, matrine, oxymatrine, astragaloside IV, ginsenoside Rg1, ginsenoside Re, ginsenoside Rd, ginsenoside Rc, and ginsenoside Rb1 exhibited no inhibition or weak inhibition against several common CYP and UGT enzymes IC50 values between 8.81 and 92.21 μM. Through kinetic modeling, their inhibition mechanisms towards those CYP and UGT isozymes were explored with obtained Ki values. In vitro-in vivo extrapolation showed the inhibition of systemic clearance for CYP or UGT substrates seemed impossible due to [I]/Ki no more than 0.1. Conclusions We summarized the PK behaviors, excretion characteristics and protein binding rates of circulating alkaloids, astragalosides and ginsenosides after intravenous Kang-Ai injection. Furthermore, weak inhibition or no inhibition towards these CYP and UGT activities could not trigger harmful DDI when Kang-Ai injection is co-administered with clinical drugs primarily cleared by these CYP or UGT isozymes.


Background
Herbal medicines are gaining popularity for the treatment of chronic and complex diseases as it works by hitting multiple molecular targets to modulate different signaling pathways [1][2][3]. Major advances have been made regarding their activities related their clinical indications, and more importantly in understanding their pharmacokinetics (PK) and excretion [3]. The PK profiles, just like the pharmacological activities, can be used as a "sieve" to assess the effectiveness of the individual compounds, and identify the chemical basis in an herbal injection [3][4][5]. Therefore, PK behaviors and excretion of circulating compounds should be first implemented in laboratory animals to reveal their systemic exposure. The corresponding PK parameters could show the ability of herbal compounds to be absorbed and transferred from the site of administration to the action targets through multiple biological barriers, and metabolic stability in vivo.
Kang-Ai injection is an herbal medicine approved in 2002 by the China Food and Drug Administration (China FDA). It consists of Astragali Radix (Huangqi), Ginseng Radix et Rhizoma (Renshen) and kushenin (Kushensu), and is indicated for many cancers, leukopenia and chronic hepatitis B [6][7][8]. Previously, we have finished the chemical identification and content determination of the main alkaloids (oxymatrine, matrine), astragalosides (astragaloside IV) and ginseng saponins (ginsenosides Rg1, Re) in Kang-Ai injection [9]. Despite the increasing understanding of Kang-Ai injection in chemical and pharmacological studies, the systemic exposure and excretion of its bioactive compounds after administration remain to be elucidated.
Previous PK researches of individual herb or its characteristic compounds have been carried out, which greatly facilitated the holistic PK properties of Kang-Ai injection. Oxymatrine, the most abundant herbal compound in Kang-Ai injection, occupies over 95.37% of total contents. When intravenously treated with oxymatrine at 30 mg/kg, the AUC 0-12 h , C max , t 1/2 and CL values in rat plasma were 16234.27 ng h/mL, 32847.57 ng/mL, 0.57 h and 3.10 L/h, respectively [10]. Besides, only about 19.9% of oxymatrine was reduced to matrine following intravenous infusion of 600 mg oxymatrine in healthy male volunteers [11]. In addition, systemic exposure of astragaloside IV appeared to be dose proportional over the dose ranged from 200 to 500 mL astragaloside IV (9 mg/100 mL) in healthy volunteers [12]. Furthermore, metabolism of astragaloside IV by intestinal microbiome was a crucial step in its excretion, rather than by hepatic and intestinal enzymes [13][14][15]. Moreover, ginsenosides Rg1, Re, and Rf were not extensively bound in plasma, and also were with substantially shorter t 1/2 values and significantly larger systemic clearance (CL tot,p ) than other-type ginsenosides [16]. Although several bioanalytical assays have been applied to the PK study of several circulating alkaloids and saponins, PK and excretion information about Kang-Ai injection is still limited. This also brings a huge obstacle in our understanding of the chemical basis responsible for the therapeutic effects of Kang-Ai injection.
In addition, drug-drug interaction (DDI) is significant safety concerns in clinical medication [17][18][19]. Inhibition or induction of human CYP or UGT isozymes could potentially adverse clinical DDI and result in metabolic disorders for endogenous components [20][21][22]. It would be more serious for the clinical drugs with narrow therapeutic indices, such as digoxin, warfarin, SN-38 and so on [17]. Similarly, UGT1A1 is the only one UGT isozyme in the conjugative detoxification of bilirubin [21]. Once UGT1A1 function is inhibited, there is a high risk of liver injury [21]. Considering the intravenous administration of Kang-Ai injection and abundant exposure of its circulating herbal compounds, it should be of particularly clinical significance to investigate the potential risks of DDI.
To this goal, this current report first identified the major circulating alkaloids, astragalosides, ginsenosides in rat samples after an intravenous 30 min-infusion of Kang-Ai injection at 6 mL/kg. In addition, the detailed PK behaviors, and exposure of these herbal substances in plasma were systemically carried out. Meanwhile, we also investigated the plasma binding rate assays of these circulating alkaloids and saponins. Furthermore, the excretion of these compounds in urine and bile were determined by dividing the cumulative amount excreted into urine (Cum.A e-U ) and bile (Cum.A e-B ) by their respect doses. Moreover, we explored the inhibitory effects and mechanism of nine major circulation compounds against several important CYP and UGT enzymes based on their widely recognized probe substrates. Taken together, the current PK and excretion information would facilitate the identification of active herbal compounds responsible for the therapeutic actions of Kang-Ai injection, and would also be helpful to inform its potential DDI in rational clinical use and achieve the optimal herbal therapy.
Keywords: Kang-Ai injection, Pharmacokinetics, Excretion, Plasma protein binding rate, Cytochrome P450, UDPglucuronosyltransferase, Drug-drug interactions Animals Specific pathogen free (SPF) grade male Sprague-Dawley rats (250 ± 20) g were provided by the Experimental Animal Center of Zhengzhou University (Zhengzhou, China). These rats were kept in a designated animal room at constant temperature (22 ± 2) °C and humidity (50 ± 20) % with 12 h of light/dark per day and free access to water and food. All protocols of animal experiments were approved in accordance with the Regulations of Experimental Animal Administration issued by the Ethics Review Committee for Animal Experimentation of Jinan University (Ethical Review NO. 20190301010). Prior to the experiments, the rats were fasted for 12 h but with free access to water. After study use, the rats were euthanized with CO 2 gas.

Sample preparation
Each plasma sample (50 μL) was transferred to a 1.5-mL polypropylene tube containing 50 μL of internal standard (IS) solution (including 20 nM olanzapine-d3 and 1000 nM mycophenolic acid-d3) and 50 μL of methanol. The mixture was vortex mixed for 30 s, and then 150 μL methanol was precipitated. The tubes were vortex mixed vigorously for 1 min and centrifuged at 13,800g for 10 min at 4 °C. Finally, 4 μL aliquots of the supernatant was injected into the UHPLC/TQD-MS system. Similarly, rat urine (50 μL) and bile (50 μL) samples were precipitated with methanol at a volumetric sample-tomethanol ratio of 1:3; after centrifugation, the resulting supernatants were applied for analysis. All the biological samples were stored at − 80 °C.

Preparation of standard solutions and quality control samples
An appropriate amount of each authentic standard was separately dissolved with 60% methanol-water in a 10 mL volumetric flask to prepare the standard solution. Furthermore, an appropriate volume of each standard solution was transferred into a 10 mL volumetric flask to the desired concentration as the stock solution. The mixed IS solution contained olanzapine-d3 and mycophenolic acid-d3 at final concentration of 20 nM and 1000 nM, respectively. All the solutions were stored at 4 °C.

Chromatographic and mass spectrometric conditions
Ultra-high-performance liquid chromatography (UHPLC) analysis was performed on a Waters Acquity UHPLC I-Class system (Manchester, UK) for the quantitative analysis of Kang-Ai injection. Chromatographic separation was achieved on an BEH C18 column (2.1 × 50 mm, 1.7 μm) with water (A) and acetonitrile (B) (both including 0.1% formic acid) at a flow rate of 0.45 mL/min with column temperature of 35 °C. The gradient flow profile was optimized as follows: 10% B from 0 to 0.5 min, 10-20% B from 0.5 to 1.5 min, maintaining 20% B from 1.5 to 4.5 min, 20-30% B from 4.5 to 5.5 min, 30-35% B from 5.5 to 7.5 min, 35-90% B from 7.5 to 9.5 min, keeping 90% B from 9.5 to 10.5 min, 90-10% B from 10.5 to 11.0 min, and maintaining 10% B from 11 to 12 min.
UHPLC system was coupled to a triple quadrupole mass spectrometer (Waters Xevo TQD, Waters, Manchester, UK). The detailed mass spectrometers were adjusted as follows: capillary voltage, 3.5 kV (ESI+) or 1.5 kV (ESI−); cone voltage, 50 V (ESI+) or 50 V (ESI−); source temperature, 350 °C; desolvation gas flow, 650 L/h; The mass spectrometer was performed in the multiple reaction monitoring mode (MRM) using both positive and negative ionization. The quantitative parameters in MRM modes were same as our previous study. All experimental data were collected and processed using a Quanlynx software in Masslynx 4.1 platform.

Method validation
The developed UHPLC-MS/MS method was validated according to the Guidance for Industry: bioanalytical method validation from the US FDA for specificity, linearity, LLOQ, extraction recovery, matrix effects, precision, accuracy, stability and dilution integrity [23]. The assay validation of each analyte in rat biological samples including plasma, urine and bile were all performed based on the standard specification above.
Specificity was determined by comparing the chromatograms of blank samples (from six individual rats), with dosed samples after an intravenous bolus of Kang-Ai injection, and blank samples spiked with each analyte at LLOQ and IS. The calibration curves were constructed by the peak area ratios of the analytes to IS (Y), versus respective sample concentrations (X) applying a weighted (1/x 2 ) least squares linear regression analysis. LLOQ was determined based on the lowest concentrations for which acceptable linearity, accuracy and precision were demonstrated. The accuracies (the relative error, RE) and inter/ intra-day precisions (the relative standard deviation, RSD) of the assay were evaluated by determining six replicates of QC samples (at low, middle, and high concentrations) on three consecutive days.
Extraction recoveries (ER) and matrix effects (ME) were evaluated by a published experimental protocol [24]. The response of extracted QC sample was defined as A1. In addition, A2 referred to the corresponding standard solutions added into the post-extracted supernatant from biological matrix, while A3 corresponded to the responses of each analyte at three QC levels. The ER and ME values were calculated as follows: ER % = A1/ A2 × 100%, and ME % = A2/A3 × 100%.
Stability tests were investigated at three QC levels under different conditions, including short-term stability (8 h exposure at room temperature), autosampler storage stability of the methanol-treated samples (at 8 °C for 18 h), freeze/thaw stability (three − 80 °C ↔ 23 °C cycles), long-term stability (14 days storage at − 80 °C). The dilution integrity was investigated by analyzing six replicate samples with each analyte, and each analyte were diluted 10-fold with blank rat matrix. Furthermore, the diluted samples were analyzed using freshly prepared calibration curve to calculate RE and RSD, which should be no more than ± 15% as criterion.

In vivo rat studies
The rats in this study all received a single 30-min intravenous infusion of Kang-Ai injection via the tail veins using TYD01-02 infusion pumps (Lead Fluid, Baoding, China). The dose was derived from the label daily dose of Kang-Ai injection (60 mL/person, once daily) according to dose normalization by body surface area [25]. For plasma pharmacokinetic studies, the rats were randomly received a single intravenous 30 min-bolus dose at 6 mL/ kg. Blood samples (100 μL) were collected from the rats' external jugular veins into heparinized polypropylene tubes at 5, 15, 30, 45 min, and 1, 1.5, 2, 3, 4, 6, 8, 12 h. After gently shaking for 10 s, the blood samples were centrifuged at 13000 g for 10 min to yield plasma samples, which were kept frozen at − 80 °C until analysis.
For urine sampling, the rats were housed singly in metabolic cages. And then, the urine samples were collected at 0-4, 4-8 and 8-12 h after starting infusion and were weighed. Likewise, bile samples were also collected at 0-3, 3-6, 6-9 and 9-12 h after starting infusion and were weighed. The urine and bile collection tubes were all frozen at − 80 °C.

Protein binding assay
Binding of circulating alkaloids and saponins to rat plasma was measured by equilibrium dialysis using Spectra/Por 2 dialysis membranes (molecular weight cutoff, 12-14 kDa; Rancho Dominguez, CA, USA) as described [26]. After equilibrating for 24 h at 55 rpm and 37 °C, the dialysate and the plasma were sampled for analysis. In brief, the concentrations of each analyte in dialysate and rat plasma samples were determined for analysis after equilibrating at 55 rpm and 37 °C for 24 h. The concentrations of tested matrine, oxysophocarpine and oxymatrine were 2 μmol/L, while those of astragaloside IV, III, and ginsenosides Rh1, Rg2, Rg1, Rf, Re, Rd, Rc, Rb1, Rb2, notoginsenosides R1, R2, were 4 μmol/L. The unbound fraction (f u-plasma ) in rat plasma was calculated as follow: , where the C d and C p values were the concentrations of each analyte in the dialysate sample and in the post-dialysis plasma sample, respectively.

Inhibition kinetic analysis
The inhibition constant (K i ) values were obtained by multiple concentrations of substrates in the absence or presence of multiple concentrations of herbal compounds as described previously [18,27,28]. Competitive inhibition, noncompetitive inhibition, and mixed-type inhibition models were used to determine the K i values by nonlinear regression analysis using the Eqs. (1)-(3), respectively. The model with the smallest Akaike information criterion (AIC) and Schwartz information criterion (SC) values was considered as the best model, following the obtained appropriate K i values. Model fitting and parameter estimation were performed using Graphpad Prism V5 software (SanDiego, CA).
The detailed parameters for three equations were as follow. V is the velocity of the reaction. The [S] and [I] are the concentrations of substrate and herbal compounds, respectively. K i is the constant describing the affinity between herbal compounds and the enzyme. K m is the substrate concentration at half of the maximum velocity (V max ) of the reaction. The αKi describes the affinity of herbal compounds to the complex of enzyme and substrate; When α is very large (α ≫ 1), the binding of inhibitor would prevent the binding of substrate, and the mixed inhibition model becomes identical to competitive inhibition.

Data processing
Pharmacokinetic parameters were estimated by noncompartmental analysis using WinNonlin 6.

Method validation
The typical MRM chromatograms obtained from blank plasma, blank plasma spiked with the analyte (LLOQ), and dosed rat plasma after intravenous Kang-Ai injection αK i (30 min) were presented in Additional file 1: Fig. S4. As seen from MRM chromatograms, there were no endogenous interference for each analyte and two IS compounds, which suggested that the specificity of the quantitative method was satisfactory. Additional file 1: Table S2 summarized the mean values of linear regression equation of the sixteen compounds. The correlation coefficients (r 2 ) were ranged from 0.9900 to 0.9994, which exhibited good linearity. The LLOQs of each circulation compound were between 5 and 40 nM, with the accuracies and precisions less than 20%, which were enough for the pharmacokinetics study of sixteen analytes following a 30-min infusion of Kang-Ai injection to rats.
The results of the intra-and inter-day precision and accuracy of all analytes and three QC samples were displayed in Additional file 1: Table S3. The intraday and inter-day precision (RSD) of sixteen analytes were within the range from 2.35 to 12.64%, whereas the accuracy (RE) derived from three QC samples was between − 14.30 and 13.66%, all meeting the requirements. These results proved that the developed method was accurate and reproducible.
Additional file 1: Table S4 exhibited the matrix effect and extraction recovery results of all analytes and two IS. The mean matrix effects of 16 circulation compounds were in the range from 92.06 to 110.45% with RSD values less than 14.14%, and the average absolute matrix effects of olanzapine-d3 (IS-1) and mycophenolic acid-d3 (IS-2) were 100.29% and 109.12% with RSD values of 11.42% and 3.13%, respectively. Similarly, the mean extraction recoveries ranged from 82.85% to 108.98% with RSD values no more than 14.73% at three QC levels, whereas the mean absolute recoveries of IS-1 and IS-2 were 80.54% and 83.60% with RSD values of 2.11% and 5.19%, respectively. These results demonstrated that the recoveries obtained were consistent and reproducible, and the ionization competition (enhancement or suppression effects) between the analytes and the endogenous co-elution was negligible.
The mean stability results (Additional file 1: Table S5) indicated the RE and RSD values of sixteen circulating alkaloids and saponins at three QC levels were in the range from − 14.53% to 14.40% and 13.87%, respectively, indicating that all the analytes were stable under different storage conditions.
In addition, a total of twenty circulating compounds were detected in rat urine. The regression equations, linear ranges and LLOQs for the determination of the analytes in rat urine was shown in Additional file 1: Table S6. Over the considered calibration range, the regression coefficient (r 2 ) was > 0.9906 for the curves of each analyte. Furthermore, twenty-one circulating compounds were obtained in rat bile, and the calibration curves of twenty-one analytes exhibited the good linearity with r 2 values over 0.9901 (Additional file 1: Table S7) at a certain concentration range in rat bile. From the data, we can see the developed method showed good linearity and sensibility in rat urine and bile. Table 1 summarized the PK parameters of circulating alkaloids (A1-A3), astragalosides (B2, B3) and ginsenosides (C1, C3, C5-C9, D2-D5) from intravenously dosed Kang-Ai injection. Figure 1a displayed the total plasma concentrations of the alkaloids over time, during and after a 30-min intravenous infusion of Kang-Ai injection in rats. Clearly, the plasma C max value of matrine (A1) were determined after stopping the infusion for 30 min. After the 30-min infusion, its plasma concentration increased with mean t 1/2 values of 4.33 h. In addition, the C max values of oxysophocarpine (A2) and oxymatrine (A3) were both measured just before stopping the infusion, following the declined plasma concentration with shorter t 1/2 values of 1.78 and 2.73 h, respectively. The mean V SS values of A1, A2 and A3 were 9.85, 25.20 and 220.11 mL/kg, respectively, suggesting these three alkaloids all had small V d values. Besides, the CL tot, p values of A3 was 111.34 mL/h/kg and those of A1 and A2 were 1.58 and 12.31, respectively, indicating that A3 and A1, A2 had moderate and low total clearance, respectively. Furthermore, these circulating alkaloids (A1-A3) were not extensively bound with f u-plasma values over 90%.

Plasma pharmacokinetics of circulating alkaloids and saponins in rats
The plasma concentration-time curves and PK parameters of astragaloside IV (B2) and III (B3) were shown in Fig. 1b and Table 1, respectively. The plasma C max (2363.99 nmol/L) and AUC 0-∞ (4523.65 nmol/L*h) values of B2 were far higher than those of B3 with C max and AUC 0-∞ values of (54.75 nmol/L) and AUC 0-∞ (34.29 nmol/L*h), due to the differences in the compounds' doses from the injection. Besides, the t 1/2 values of these two astragalosides were both small and less than 2 h. Similarly, B2 and B3 both exhibited the small V d values (< 100 mL/kg), whereas B2 (28.39 mL/h/kg) and B3 (206.96 mL/h/kg) had low and moderate total clearance, respectively. Protein binding assay demonstrated that B2 and B3 were extensively bound in rat plasma (nearly 90%).
These circulating ginsenosides in plasma could be classified into two categories, protopanaxatriol (PPT) type ginsenosides (C1, C3, C5-C9) and protopanaxadiol (PPD) type ginsenosides (D2-D5). Fig. 1c, d summarized the plasma concentrations of the PPT-type ginsenosides and PPD-type ginsenosides over time, respectively. It was notably shown that the T max values of PPT-type ginsenosides were 30 min, whereas those of PPD-type ginsenosides were measured after stopping infusion for 15 min. In addition, the C 4h values of the PPT-type ginsenosides were very low, while those of the PPD-type ginsenosides still maintained a moderate plasma concentration. These suggested that the PPT-type ginsenosides possessed the longer MRT, and were eliminated more quickly than PPD-type ginsenosides. Another strong evidence was that CL tot, p values of PPT-type ginsenosides (72.70-255.80 mL/h/kg) were approximately 10 times of those of PPT-type ginsenosides (5.60-7.80 mL/h/kg). Besides, except for ginsenoside Rg2 (C5) about the t 1/2 value of 9.77 h, the PPT-type ginsenosides exhibited the shorter elimination half-lives values of i.e., 0.29-1.20 h for ginsenoside Rh1 (C1), Rg1 (C6), Rf (C7), Re (C9) and notoginsenoside R1 (C8), R2 (C3) than those of 4.12-7.22 h for ginsenosides Rd (D2), Rc (D3), Rb2 (D4) and Rb1 (D5). The V SS values of PPT-type ginsenosides and PPD-type ginsenosides were 50.05-791.61 and 35.69-46.96 mL/kg, respectively, indicating that the PPT-type ginsenosides were more easily distributed in various body fluids and tissues than the PPD-type ginsenosides. Additionally, the PPT-type ginsenosides (C1, C3, C5-C9) were not significantly bound, whereas the PPD-type ginsenosides (D2-D5) were obviously bound (around 99%) in rat plasma.
Moreover, due to lack of the available authentic standards, virtual quantification of other main circulating herbal compounds in rat plasma, including baptifoline (A5), 9α-hydroxy-sophocarpine (A6), leontalbinine N-oxide (A7), mamanine (A8), sophoranol (A9), were achieved using the calibration curve of an available reference standard of an analog that bore close structural similarity to the analyte. Figure 1e, Table 1 showed their mean plasma concentration-time curves and calculated relative PK parameters, respectively. The C max values ranged from 202.37 to 421.27 nmol/L, and the T max values were all determined just before stopping the infusion. Besides, A7 exhibited the longest MRT (7.94 h) and t 1/2 (5.73 h) values among these five alkaloids. The small V SS (16.89-150.70 mL/kg) and CL tot, p (7.93-56.05 mL/h/kg) values indicated that these alkaloids were slowly distributed in various body fluids and tissues, and slowly cleared from the systemic circulation.

Urinary and biliary excretion of circulating alkaloids and saponins in rats
Astragaloside III (B3) was eliminated more extensively via hepatobiliary excretion (f e-B , 67.06%) than renal excretion (f e-U , 5.25%) ( Table 1). Unlike for B3, renal excretion was a minor elimination pathway for astragaloside IV (B2) (f e-U , 4.84%), and the same fact was that the fraction of dosed B2 was excreted into bile (f e-B ) about 2.49%.
The PPT-type ginsenosides (C1-C9), except ginsenoside Rg2 (C5), all exhibited the large f e-U values (9.59-72.17%) ( Table 1), suggesting that the renal excretion was the predominant elimination route responsible for their systemic clearance. Similarly, the PPD-type ginsenosides (D2-D5), exhibited moderate f e-U values (6.41-27.8%), which was around a half of the values of PPT-type ginsenosides. In addition, in rat bile, the PPT-type ginsenosides were the most abundant herbal compounds with f e-B values of 9.71-67.93%, whereas only two PPD-type ginsenosides, ginsenoside F2 (D1) and Rd (D2) could be detected and displayed the small f e-B values of 7.46% and 5.55%, respectively. Consistent with the inter-compound differences in CL tot, p values, the CL R and CL B values for these PPT-type ginsenosides (C1-C9) were significantly greater than those respect values for the PPD-type ginsenosides (D1-D5).

Analysis of inhibition mechanism
The AIC and SC values ( Table 2) were obtained after the inhibition data were modeled by three conventional inhibition equation (i.e., competitive, noncompetitive and mixed-type). Based on the smallest AIC and SC values principle, C6 displayed competitive inhibition against UGT1A1, while C9 exhibited mixed-type inhibition against CYP2C8. Noncompetitive inhibition kinetics were observed for B2 against CYP2C9, D2 towards CYP2B6 and 3A4, D3 against UGT1A9 and D5 towards CYP2C9.

Potential DDI risks
Based on the [I]/K i values in vitro, potential DDI risks in vivo could be predicted [27,28]. Traditionally, [I]/ Ki < 0.1, low possibility DDI; 1 > [I]/Ki > 0.1, medium possibility DDI; [I]/Ki > 1, high possibility DDI [18]. In this study, [I]/K i values for these herbal compounds were all no more than 0.1 (Table 2). Therefore, it seemed impossible to trigger DDI when Kang-Ai injection were co-administrated with clinical drugs which were metabolized by these enzymes.

Discussion
Bioactive compounds with adequate abundance in herbal injection and favorable PK properties are mostly believed to be responsible for the pharmacological effects and therapeutic efficacy [4]. In the current study, alkaloids and saponins derived from Kang-Ai injection are two Fig. 3 The effects of several circulating compounds towards eight expressed CYP isozymes and three recombinant UGT enzymes. a Kang-Ai injection, b oxymatrine, c astragaloside IV, d ginsenosides Rg1, and e ginsenoside Rc; The probe substrates were incubated at 37 °C in the absence (control, 0 μM) and presence of tested mixture and compounds (Kang-Ai injection and samples that Kang-Ai injection was diluted 10 times; 1, 10, and 100 μM for oxymatrine; 0.1, 1, 10 and 100 μM for astragalosides and ginsenosides). Data represent the mean ± standard deviation of triplicate, (* compared with those of control, * p < 0.05) classes of bioactive compounds which are likely responsible for the adjuvant therapy of many cancers (Additional file 1: Table S1) [6,8]. After dosing, oxymatrine (A3), matrine (A1), astragaloside IV (B2), ginsenosides Rc (D3), Rd (D2), Rb1 (D5), Rg1 (C6) and Re (C9), all exhibited considerably high levels of systemic exposure (over 1000 nmol/L*h) (Fig. 2a), indicating that these circulating compounds are worth special consideration in the pharmacological research. Besides, the abundant exposure of A1 may be attributed the metabolism of A3 by CYP3As [29]. In addition, the considerable systemic exposure for A3, A1, D3, D2 and D5, seemed to be correlated with the significantly longer t 1/2 values of 2.73-7.22 h than those of B2, C6 and C9 (1.07-1.64 h) ( Table 1). This phenomenon that long elimination half-life could significantly counterbalance influence on systemic exposure to herbal compounds with poor bioavailability has been proved previously [30].
Except the systemic exposure of major circulating compounds, we also investigate their elimination in the current study (Table 1). After dosing Kang-Ai injection, oxymatrine (A3) and matrine (A1) mainly involved in the renal and hepatobiliary excretion of the parent compounds (over 90%), of which parts of circulated A1 was originated from the metabolism of A3 by CYP3A4 [11]. Only about 25% of dosed astragaloside IV (B2) were detected as parent compound in urine and bile, and this was probably attributed to the de-glycosylation of B2 to its sapogenin [15]. Additionally, ginsenosides Rg1 (C6) and Re (C9) were eliminated mainly to urine and bile Fig. 4 The IC50 values of astragaloside IV against CYP2C9 (a), ginsenoside Rg1 against UGT1A1 (b), ginsenoside Re towards CYP2C8 (c), ginsenoside Rd towards CYP2B6 (d) and CYP3A4 (e), ginsenoside Rc against UGT1A9 (f), and ginsenoside Rb1 against CYP2C9 (g). The data were fit to log (concentration) and normalized response equations. Each data point represented the mean value ± the S.D. of triplicate determination (nearly 60%), whereas it proved that approximately 25% of ginsenoside Rd (D2) was excreted via renal and hepatobiliary elimination. Ginsenosides Rc (D3) and Rb1 (D5) were not detected in bile, and mainly were eliminated through renal excretion. Except the in vivo metabolism, i.e., oxidation and de-glycosylation [16], the delivery to the rat tissues of circulating herbal compounds through blood flow were also an important factor to determine the systemic exposure and excretion of circulating compounds [31]. The V SS values of dosed A3, C6 and C9 were 100.55-220.11 mL/kg, which were near the rat extracellular volume (300 mL/kg) [32], indicating that these herbal compounds could be distributed evenly in various body tissues, except several barriered tissues. However, this elimination route warrant future investigation.
In addition, the obtained PK parameters ( Table 1) also could enable us to compare the plasma concentrations of the alkaloids and saponins with the reported effective concentrations (IC 50 or EC 50 values) for pharmacological activities. Oxymatrine (A3) exhibited the most abundant concentration (422.70 μmol/L) after dosing Kang-Ai injection, which is still obviously below the reported effective concentrations (473. .84 μmol/L) for several tumor cell lines, including HepG2, MCF-7, HeLa, A549 cells [33]. However, for SW-620, BGC823, A375 cells, A3 is sufficiently effective to inhibit the tumor cell viability (0.10-40.00 μmol/L) [33]. Similar results were obtained for matrine (A1) and its derivatives [34], astragaloside IV (B2) [35], and several main ginsenosides [36]. Although the low plasma concentrations do not result in the ineffectiveness of Kang-Ai injection for the therapeutic effects, they may limit the contribution of bioactive compounds to the current adjuvant therapy. How to achieve the higher plasma concentrations is an important idea to develop new and superior adjuvant agents derived from the current Kang-Ai injection.
Identification of the induction or inhibition of the abundant circulating alkaloids and saponins towards uptake and efflux transporters was another significant risk factor for the DDI. In clinics, oxymatrine (A3) could interfere with hOCT1-mediated hepatic uptake and renal elimination, and intestinal absorption by hOCT3, whereas matrine (A1) only had the potential to decrease The Dixon plots for the inhibition effects of circulating saponins against recombinant CYP and UGT isozymes. a Inhibition effect of astragaloside IV against tolbutamide-4-hydroxylation for CYP2C9; b Inhibition effect of ginsenoside Rg1 towards β-estradiol-3-O-glucuronidation for UGT1A1; c Inhibition effect of ginsenoside Re against paclitaxel-6-hydroxylation for CYP2C8; d Inhibition effect of ginsenoside Rd towards bupropion-hydroxylation for CYP2B6; e Inhibition effect of ginsenoside Rd against nifedipine-oxidation for CYP3A4; f Inhibition effect of ginsenoside Rc towards propofol-O-glucuronidation for UGT1A9; g Inhibition effect of ginsenoside Rb1 against tolbutamide-4-hydroxylation for CYP2C9; All data represent the mean ± SD of triplicate determinations hOCT3 expressed in enterocytes [42]. Besides, A1 and A3 did not demonstrate the inhibition of P-gp in vitro and in vivo [43]. Additionally, it was reported that astragaloside IV (B2) could induce the protein levels of P-gp, MRP1, MRP3, but produced the inhibitory effects on BCRP [44]. In contrast, in MDR1-MDCKII cells, B2 only exhibited weak inhibition activity of 8.22% when treated with 100 μM [43]. Recently, several scholars got the fact that B2 could induce the upregulation of P-gp and BCRP, and might be a potential substrate of P-gp [45]. For ginsenosides, ginsenoside Rg1 (C6) was potential inhibitor of NTCP (IC 50 = 50.49 μM), whereas ginsenoside Re (C9) was potential substrates of NTCP [46]. Further, DDI between ginsenosides Rb1 (D5), Rc (D3), Rd (D2) and methotrexate (a probe substrate for Mrp2) might occur owing to the decrease in the mRNA and protein levels of Mrp2 [47]. In addition, D5, D3, D2 and D9 were predicted to possess the high potential for human OATP1B3-mediated drug interactions [16]. Moreover, PPT and PPD, the aglycones of PPT-type ginsenosides and PPD-type ginsenosides, respectively, were potent P-gp inhibitors, suggesting the unpredictable interactions when ginsenosides were co-administered with the P-gp substrate drugs [48]. The interactions of complex circulating compounds and clinical drugs based on transporters may be one of the important compatibility mechanisms for Kang-Ai injection.
In clinical consumption, the adverse drug reactions (ADR) derived from herbal injections, should be gaining more attention. It was essential to evaluate the objectivity of ADR, develop a reasonable application strategy, and ensure the security and validity of Kang-Ai injection. Data and information obtained from PK studies on the circulating alkaloids and saponins could help understand the complex action of herbal injections, and predict a variety of events related to the efficacy and toxicity of herbal injections. In Kang-Ai injection, the most abundant compounds, oxymatrine (A3) and matrine (A1) both could cause the hepatic toxicities, of which the toxicity of A1 was greater than that of A3 [49]. The elevated metabolism of A3 by CYP3A4 may result in more severe liver damage. Therefore, the clarification of the PK studies could be helpful in deciphering the ADR and hepatorenal toxicity of herbal injections.