*2.6. Method Validations*

Method validations for the pharmacokinetic and cocktail assays, in terms of specificity, linearity, and sensitivity; precision and accuracy; recovery and matrix effect; and stability, were individually carried out by following the U.S. Food and Drug Administration (FDA) Guidance on Bioanalytical Method Validation and Drug Interaction Studies [12].

#### *2.7. Data Processing*

Calibration data were fitted to linear calibration curves using 1/x <sup>2</sup> weighting. For the PK study, the half-time (*T*1/2), maximum plasma concentration (*C*max), time to reach the maximum concentration (*T*max), and area under concentration-time curve (AUC) were determined by non-compartmental method using Drug and Statistics 3.0 (DAS 3.0, Mathematical Pharmacology Professional Committee of China, Shanghai, China). For the cocktail assay, the activity is expressed as the percentage of activity remaining comparing with that of a control sample containing no inhibitor. Substrate inhibition data were analyzed using GraphPad Prism 6 (version 6.01; San Diego, CA, USA) in logistic regression. All the data were described as mean ± standard deviation (SD). Normality assumptions were tested by the Kolmogorov–Smirnov statistic setting *p* = 0.10 as the limit for rejection of the null hypothesis of normality. If the distribution of the data was normal with equal variances, a two-sided test was performed at the 5% level of significance. The Welch's correction was then applied when the underlying variances were not equal. When the assumption of normality must be rejected, the Mann-Whitney test, a non-parametric equivalent of the independent-measures *t*-test, was used.

#### **3. Results**

#### *3.1. Injection Solvent Optimization*

It is well known that the sample solvent and volume have large effects on the peak asymmetry and column efficiency [13]. Herein, we established a LVDI-UHPLC-MS/MS method, assisted by injection solvent optimization, for sensitive bioassays of the pharmacokinetic interactions and cocktail analysis of CTE and NGTS. Considering the solubility and the sample pretreatment methods (Supplementary Materials) of the target analytes, the mixed standard solution of the six reference compounds was diluted with methanol (MeOH)/water (H2O) in a ratio of 20% increment ranging from 100% H2O to 100% MeOH for the PK study, and all the six standards of the PK study showed the highest responses when using 40% or 60% aqueous MeOH as the injection solvent. The CYP450 probe substrates and their corresponding metabolites for the in vitro cocktail assay were chosen according to the FDA guide [14]. Likewise, the six metabolites for the cocktail assay showed the best chromatograms when the 25% aqueous MeOH was served as the sample solvent, with exception of 1′ -hydroxymidazolam, for which 50% aqueous MeOH was selected as the injection solvent (Figure S2).

### *3.2. Optimization of the Loading Phase for the LVDI-UHPLC-MS*/*MS-Based Method*

The instrument stability of the LVDI-UHPLC-MS/MS setup was first investigated, and the results indicated the LVDI-UHPLC-MS/MS could meet the demands of quantitation (Table S3). An analytical run of the LVDI-UHPLC-MS/MS-based method was fragmented into a loading phase and an elution phase. The gradient condition for the elution phase turned out to be the same as that of the regular UHPLC-MS/MS analysis. Thus, the optimization works were concentrated on the loading phase, including the flow rate, mobile phase, and dilution time. According to the optimized gradient programs of the elution phases, both PK and cocktail studies employed water as the mobile phase for their loading phase after evaluating the solvents ramped from 0% to 20% aqueous acetonitrile. The flow rate of the loading phase for the PK study was finally optimized as 3.0 mL·min−<sup>1</sup> after assays of 0.4, 1.0, 2.0, and 3.0 mL·min−<sup>1</sup> flow rates. Because the higher the flow rate, the lower the signal responses of paracetamol and 6-hydroxychlorzoxazone, the cocktail assay finally chose 0.4 mL·min−<sup>1</sup> as the flow rate of the loading phase. It turned out that the dilution time did not have a profound effect on the

total chromatogram by comparing 0.5, 1, and 2 min. Thus, the shortest time, 0.5 min, was chosen to promote the analytical efficiency. The corresponding bioanalytical method validations were carried out by following the FDA guidance [12], and the results demonstrated that the newly developed LVDI-UHPLC-MS/MS-based methods enabled reliable detection and precise determination of the multiple-component in PK and cocktail studies. The lower limits of quantitation (LLOQ) of most analytes (except Rb<sup>1</sup> and Rd) were lower than 60 pg/mL.

#### *3.3. Comparative Multiple-Component PK Studies*

Considering the bioavailability improvement of active ingredients is a key point of traditional Chinese medicine compatibility, we executed a multiple-component PK study. HSYA, GRb1, GRd, GRe, GRg1, and NGR<sup>1</sup> were the primary circulating compounds and the main cardio-protective components in CNP [15,16], and thus were chosen as the PK markers. Meanwhile, the optimized injection solvent and LVDI-UHPLC-MS/MS method were integrated to achieve a much more sensitive method for reliable quantification of these components in vivo. After validation (Figure 3, Tables S4–S7), the developed method was first applied to characterize the pharmacokinetic characters of HSYA, GRb1, GRd, GRe, GRg1, and NGR<sup>1</sup> in rat plasma. Their plasma concentrations versus time profiles are displayed in Figure 4 and Table S8. The combination group showed greater *C*max and AUC0-t values (Table 1) of HSYA, GRg1, GRb1, NGR1, and GRe over the individual extract groups. Exceptionally, GRd exhibited considerable AUC0-t and *C*max between NGTS and CNP dosing groups, whereas significantly different *T*1/<sup>2</sup> values, which indicated the combination use of CTE and NGTS, may have accelerated the elimination processes of GRd. The reason may have been due to the hydrolysis of GRb<sup>1</sup> to GRd in vivo [17], resulting in a more complicated PK behavior of GRd than other compounds in CNP.

**Figure 3.** Representative multiple reaction monitoring (MRM) chromatograms of target analytes in rat plasma in a positive mode: (**A**) blank plasma; (**B**) blank plasma spiked with six chemical standards and internal standards (IS); (**C**) plasma sample collected at 2 h following oral administration of extract CNP (*Carthamus tinctorius* extract (CTE) 50 mg/kg + notoginseng total saponins (NGTS) 60 mg/kg) to rats.

**Figure 4.** Mean plasma concentration-time profiles of the six analytes in rats after oral administration of CTE, NGTS, and CNP. Each point represents the mean ± SD (*n* = 6).

**Table 1.** Pharmacokinetic parameters of hydroxysafflor yellow A (HSYA), ginsenoside Re (GRe), ginsenoside Rb1 (GRb<sup>1</sup> ), ginsenoside Rd (GRd), ginsenoside Rg<sup>1</sup> (GRg<sup>1</sup> ), and notoginsenoside R<sup>1</sup> (NGR<sup>1</sup> ) after oral administration of CTE, NGTS, and CNP to rats. Each point represents the mean ± SD (*n* = 6).


\*: *p* < 0.05, versus the combination group. *T*max: the time of peak concentration; t1/2: half-life; *C*max: the peak or maximum concentration; AUC: area under concentration-time curve. For abbreviations of analytes A6–A18, please refer to the Supplementary Materials section.

### *3.4. CYP450-Mediated Herb–Herb Interactions*

To investigate the possible HHIs between NGTS and CTE, an in vitro cocktail assay involving seven probe substrates was conducted, with the assistance of LVDI-UHPLC-MS/MS method to avoid CYP450 crossovers (Figure 5, Tables S9–S11). The incubation system was optimized in the aspects of substrate choice, enzyme concentration, incubation time, and substrate concentration (Figures S3–S4, Table S12), following the guidance of the FDA [14]. According to the half inhibiting concentration (IC50) values of CTE, NGTS, and CNP (Table 2), CTE showed weak inhibition on CYP1A2, CYP2D6, and CYP2C9 and moderate inhibition on CYP2B6 and CYP2E1, whereas NGTS presented much more potent inhibitions on all these detected CYP450s than CTE (Table 2). After combination, CNP showed

more potent inhibition on CYP1A2 than CTE and NGTS, and more potent inhibition on CYP2C9, CYP2C19, CYP3A4, and CYP2D6 than CTE (Table 2).

**Figure 5.** Representative MRM chromatograms of all probe metabolites and two ISs in the incubated rat microsomal sample with no substrate cocktails: (**A**) blank microsomal sample, (**B**) blank microsomal sample spiked with seven chemical standards and two ISs monitored in a polarity switching mode.


**Table 2.**IC50values of CTE, NGTS, CNP, and the 18 representative compounds for inhibiting cytochrome p450 (CYP450) isozymes.

#: Values were obtained from triplicate tests, and presented as mean ± SD. The units of IC50 values for CTE, NGTS, and CNP are <sup>µ</sup>g·mL−1, whereas for the 18 single components are µM. Potent inhibitors (IC50<15µM for 1A2, 2C9, 2C19, 2D6, 2B6, and 3A4) are presented in italic. For abbreviations of analytes A6–A18, please refer to the Supplementary Materials section.

To identify the major CNP components responsible for the inhibition, 18 representative components from CNP were evaluated by the cocktail assays. The results (Table 2) showed that HSYA and ginsenosides showed high or moderate inhibition activities on the seven CYP450 isozymes. Although flavonoid glycosides showed moderate or weak inhibition activities, or even no inhibition activities (IC<sup>50</sup> values of >200 µM, Table 2), the flavonoid aglycones quercetin (A16), kaempferol (A17), and 6-hydoxykaempferol (A18), in contrast, exhibited potent inhibitory activities against the seven CYP450 isozymes. In particular, 6-hydroxykaempferol (A18) showed remarkably stronger inhibitory activities on the seven CYP450 isozymes (IC<sup>50</sup> < 5 µM, Table 2). In conclusion, the ginsenosides GRg<sup>1</sup> (A1), GRb<sup>1</sup> (A2), and GRd (A3), and the flavonoids 6-hydoxykaempferol-3-*O*-glucoside (A6), kaempferol-3-*O*-glucoside (A7), anhydroxysafflor yellow B (AHSYB, A8), hydroxysafflor yellow A (HSYA, A9), 6-hydroxykaempferol-3,6-di-*O*-glucoside (A13), quercetin (A16), kaempferol (A17), and 6-hydroxykaempferol (A18) were presented as being the intensive inhibitors to different CYP450s (IC<sup>50</sup> ≤ 15 µM). Among them, GRg<sup>1</sup> (A1), GRb<sup>1</sup> (A2), GRd (A3), quercetin (A16), kaempferol (A17), and 6-hydroxykaempferol (A18) were the main active components of CNP for the inhibition of CYP1A2.

#### *3.5. Discussion*

Herbal pair, the most fundamental and simplest form of Chinese herbal medicine formula, has been favored for centuries because of its better therapeutic outcomes and fewer side effects [18]. Herein, we primarily aimed to clarify the compatibility mechanisms between NGTS and CTE from PK interactions. Given the low dosage, more efforts were paid onto the detection and quantification of trace CNP-derived components in vivo. We found the injection solvent extensively affected the chromatographic performances. Increasing the injection volume could advance the sensitivity owing to the subjection of larger amounts of analytes [19]. However, the solvent effect might be initiated by directly injecting large volume of solution onto the chromatographic column without any additional treatment. Increasing the flow rate of the mobile phase could guarantee the dilution of the injection solvent and retention of the target analytes, which should be a practical choice to minimize, or even avoid the solvent effect. However, a rapid flow rate gave rise to a higher back pressure. Therefore, the evaporation–reconstitution step often involved loading the sufficient quantity of sample onto the column for LC–MS analysis. Fortunately, the electronic six-port/two-channel valve mounted on the QTRAP system can be applied as a viable solution to split the back pressure. Therefore, LVDI-UHPLC-MS/MS method was proposed. With the LVDI-UHPLC-MS/MS method, the sample can be directly injected into LC-MS analysis without any evaporation–reconstitution step, which is time-consuming and risks crucial chemical degradation during the evaporation procedure. Under the assistance of the injection solvent optimization, the validated LVDI-UHPLC-MS/MS-based method turned out to be extremely sensitive, accurate, and qualified for the bioassay measurement.

The pharmacokinetic results of HSYA, GRg1, GRe, GRb1, and NGR<sup>1</sup> indicated that after combination, the absorption of these active components was increased inferred from their higher *C*max and AUC0-t values over that of the individual extract groups (Table 1). The increment of *C*max and AUC0-t values suggested that CYP450-mediated HHIs between CTE and NGTS may primarily account for the compatibility mechanisms of CTE and NGTS. An in vitro cocktail assay was then carried out to find the clues being responsible for HHIs between CTE and NGTS. The results showed CNP exhibited more potent inhibition on CYP1A2 (Table 2), the key enzyme involved in the oxidation reactions of most xenobiotics [20], compared with CTE and NGTS. In order to search for the single components contained in CNP responsible for the inhibition of CYP1A2 and other CYP450s, 18 main components from CNP were evaluated for their inhibition on CYP450s.

The results showed that GRg<sup>1</sup> (A1), GRb<sup>1</sup> (A2), GRd (A3), 6-hydoxykaempferol-3-*O*-glucoside (A6), kaempferol-3-*O*-glucoside (A7), anhydroxysafflor yellow B (AHSYB, A8), hydroxysafflor yellow A (HSYA, A9), 6-hydroxykaempferol-3,6-di-*O*-glucoside (A13), quercetin (A16), kaempferol (A17), and 6-hydroxykaempferol (A18) were the main active components for the CYP450 inhibition, and GRg<sup>1</sup>

(A1), GRb<sup>1</sup> (A2), GRd (A3), quercetin (A16), kaempferol (A17), and 6-hydroxykaempferol (A18) were the main active components for CYP1A2 inhibition.

To further discuss the possibility of in vivo interaction between CTE and NGTS, the inhibitions of GRg1, GRb1, GRd, HSYA, NGR1, and GRe to CYP450s at their *C*max levels were calculated by their respective "dose–response curve". The results (Table S13) showed GRg1, GRb1, GRd, HSYA, NGR1, and GRe could not significantly inhibit CYP450s at their *C*max levels, which is in accordance with the results that the combination use of CTE and NGTS can only increase the system exposures of these components to some extent.
