*2.8. In Vitro Metabolism and Protein Binding Study*

An in vitro microsomal metabolism study was conducted using Corning® GentestTM pooled male RLM (from Sprague-Dawley rats) and HLM (from more than 5 male donors) as previously described [29,30], with slight modifications and in accordance with the manufacturer's protocol. To assess the possibility of metabolic interaction between REP and CEL, a microsomal reaction mixture was prepared as follows (total volume: 0.2 mL): RLM or HLM (0.5 mg/mL), 50 mM phosphate buffer, 1 mM NADPH, 10 mM MgCl2, 1 µM substrate, and various concentrations of inhibitor (1–100 µM). The disappearance rates of REP (as a substrate) in the absence or presence of CEL (as an inhibitor), and vice versa, were determined. At 0 and 15 min (REP) or 0 and 45 min (CEL) after starting the metabolic reaction, a 50 µL aliquot of microsomal incubation mixture was sampled and transferred into a clean 1.5 mL microcentrifuge tube containing 100 µL cold ACN containing IS (50 ng/mL) to stop the metabolic reaction. After vortex mixing and centrifugation at 15,000× *g* for 10 min, a 100 µL aliquot of the supernatant was stored at −80 ◦C until HPLC analysis.

The fractions of unbound REP and CEL (fu) in rat and human plasma were measured using the rapid equilibrium dialysis (RED) device (Thermo Fisher Scientific, Inc.) as previously described [31]. The plasma was spiked with REP alone, CEL alone, and both drugs, yielding final concentration of 10 µM. A 0.2-mL spiked plasma was placed into the 'sample' chamber, and a 0.35 mL isotonic phosphate buffered saline was placed into the adjacent 'buffer' chamber. The fraction unbound was calculated as the ratio of the drug concentrations in the 'buffer' compartment to those in the 'sample' compartment.

#### *2.9. Data Analysis*

The IC<sup>50</sup> of CEL for the inhibition of the metabolism of REP was determined by GraphPad Prism 5.01 (GraphPad Software, San Diego, CA, USA) according to the following Hill equation:

$$\mathbf{y} = \mathbf{M}\mathbf{n} + \frac{\mathbf{M}\mathbf{x} - \mathbf{M}\mathbf{n}}{1 + \left(\frac{\mathbf{x}}{\mathbf{IC}\_{\\$0}}\right)^{-\mathrm{P}}}.$$

Analytical data were acquired and processed using the LC Solution Software (Version 1.25; Shimadzu Co.). Non-compartmental analysis was conducted to estimate pharmacokinetic parameters such as total area under plasma concentration versus time curve from time zero to infinity (AUCinf), total area under plasma concentration versus time curve from time zero to time of last sampling (AUClast), and terminal half-life (*t*1/2) using the NCA200 and 201 models of WinNonlin software (Version 3.1; Certara USA Inc., Princeton, NJ, USA) [32]. Peak plasma concentration (*C*max) and time to reach *C*max (*T*max) were directly read from the observed data.

#### *2.10. Statistical Analysis*

A *p*-value below 0.05 was considered statistically significant by using *t*-test for comparison between two unpaired means or by using analysis of variance (ANOVA) with post-hoc Tukey's honestly significant difference test for comparison among three unpaired means. Unless indicated otherwise, all data except *T*max were expressed as mean ± standard deviation (median (ranges) for *T*max). All data numbers were rounded to three significant figures.

#### **3. Results and Discussion**

#### *3.1. Method Development*

In this study, various chromatographic conditions were evaluated for sufficient sensitivity and good separation of analytes from endogenous substances of biological matrix within an appropriate run time. Several experiments were performed to choose suitable stationary phase, mobile phase, sample preparation procedure, and IS.

To choose a stationary phase, several types of HPLC columns including AccucoreTM HILIC LC column (150 × 4.6 mm, 2.6 µm; Thermo Fisher Scientific, Waltham, MA, USA), XTerra Shield RP18 column (150 × 3.9 mm, 5 µm; Waters Co., Milford, MA, USA), and Kinetex® C8 and C18 columns (250 × 4.6 mm, 5 µm; Phenomenex) were tested. Our analysis found that Kinetex® C18 column showed higher peak resolution and intensity than other columns (data not shown). Therefore, Kinetex® C18 column was chosen as a stationary phase for the analytes.

The composition of mobile phase was optimized with different buffer types, such as citrate buffer (pH 3–5) and phosphate buffer (pH 6–7), and various ACN contents. Changes in the pH of mobile phase considerably influenced the peak retention times of REP (acidic compound) and endogenous interferences; however, it exerted little influence on those of CEL and ketoconazole that are neutral compounds. As a result, the mobile phase of pH 6.0 containing 46.4% ACN achieved good separation from endogenous interference in plasma with acceptable peak resolution. Thus, we settled for this mobile phase in developing the present HPLC-FL method.

Sample preparation was performed using a solvent precipitation-reconstitution method which is an efficient and economical sample pretreatment procedure compared with a solid phase or liquid–liquid extraction method. For optimization, several organic solvents, such as acetone, methanol, trichloroacetic acid, ACN, and their mixtures, were evaluated. Among them, ACN yielded the lowest matrix effect and highest recovery for analytes following centrifugation at 15,000× *g* for a relatively short precipitation time of 5 min.

Several fluorescent compounds, such as diclofenac, diflunisal, doxorubicin, metoprolol, naproxen, propranolol, and quinidine, were tested as a potential IS. However, these were unsuitable as IS, due to poor separation from analytes and endogenous substances in biological matrix. As a result, ketoconazole was finally chosen, because it exhibited good separation with acceptable retention time, peak resolution, and fluorescence intensity at the same wavelength as REP and CEL.

#### *3.2. Method Validation: Selectivity, Linearity, Sensitivity, Precision, and Accuracy*

As shown in Figure 2, the analyte peaks were well separated from each other and from endogenous matrix peaks in the blank plasma. Thus, it appears that the present bioanalytical method could offer acceptable selectivity without endogenous interferences occurring at the retention times of the analytes. The calibration curves (REP-to-IS or CEL-to-IS peak area ratio versus REP or CEL concentration, respectively) for REP and CEL were observed to be linear from 10 to 2000 ng/mL in rat plasma samples. A representative equation for the calibration curves was constructed, as follows: *y* = 1.018*x* − 3.029 for REP and *y* = 2.093*x* − 0.723 for CEL, where *y* represents the ratio of the peak area of REP or CEL to that of IS, and *x* represents the ratio of nominal concentration of REP or CEL. The correlation coefficients (*r* 2 ) were over 0.999, showing good linearity of this method. Generally, the sensitivity of a bioanalytical method is represented by the LLOQ value, which, in the present study, was determined to be 10 ng/mL for both REP and CEL. Moreover, the present method offered good sensitivity for CEL, with LLOQ comparable to those reported by previous LC-MS/MS methods in human plasma (LLOQ: 5–10 ng/mL; plasma volume: 100–200 µL) [25,26,33]. The intra- and inter-day precision and accuracy of this method were determined for REP and CEL at the four different QC levels, as shown in Table 1. The precision was estimated to be 8.30% or less, and the accuracy ranged from 98.6% to 112%. These values are within a generally acceptable range, showing that the present method was precise, accurate, and reproducible.

**Figure 2.** *Cont*.

**Figure 2.** Representative chromatograms of repaglinide (REP), celecoxib (CEL), and ketoconazole (IS) in rat plasma: blank rat plasma (**A**); blank rat plasma spiked with REP, CEL (10 ng/mL, lower limit of quantification (LLOQ)), and IS (**B**); blank rat plasma spiked with REP, CEL (120 ng/mL, middle quality control (MQC)), and IS (**C**); plasma sample collected 120 min after concurrent oral administration of REP and CEL solution in rats, where calculated concentrations of REP and CEL were 53 and 968 ng/mL, respectively (**D**). EU: emission unit.


**Table 1.** Intra- and inter-day precision and accuracy of REP and CEL in rat plasma (*n* = 5). HQC: high quality control.
