Prospective Prediction of Dapaconazole Clinical Drug–Drug Interactions Using an In Vitro to In Vivo Extrapolation Equation and PBPK Modeling

This study predicted dapaconazole clinical drug–drug interactions (DDIs) over the main Cytochrome P450 (CYP) isoenzymes using static (in vitro to in vivo extrapolation equation, IVIVE) and dynamic (PBPK model) approaches. The in vitro inhibition of main CYP450 isoenzymes by dapaconazole in a human liver microsome incubation medium was evaluated. A dapaconazole PBPK model (Simcyp version 20) in dogs was developed and qualified using observed data and was scaled up for humans. Static and dynamic models to predict DDIs following current FDA guidelines were applied. The in vitro dapaconazole inhibition was observed for all isoforms investigated, including CYP1A2 (IC50 of 3.68 µM), CYP2A6 (20.7 µM), 2C8 (104.1 µM), 2C9 (0.22 µM), 2C19 (0.05 µM), 2D6 (0.87 µM), and 3A4 (0.008–0.03 µM). The dynamic (PBPK) and static DDI mechanistic model-based analyses suggest that dapaconazole is a weak inhibitor (AUCR > 1.25 and <2) of CYP1A2 and CYP2C9, a moderate inhibitor (AUCR > 2 and <5) of CYP2C8 and CYP2D6, and a strong inhibitor (AUCR ≥ 5) of CYP2C19 and CYP3A, considering a clinical scenario. The results presented may be a useful guide for future in vivo and clinical dapaconazole studies.

Current concern about the use of azole antifungals in clinical practice is the many drug-drug interactions (DDIs) related to these drugs, mainly as moderate/strong inhibitors of cytochrome P450 (CYP) isoenzymes [8,9]. The investigation of a potential DDI of a new azole antifungal in the early drug development process is critical to allow the safe use of the new molecular entity during clinical trials and later during clinical practice [10]. Therefore, understanding and identifying the enzymes that are inhibited in the presence of the drug under study is directly related to deciding whether to proceed or not in the development phase [11]. Mechanistic approaches such as static (in vitro to in vivo extrapolation equations, IVIVE) or dynamic models (physiologically-based pharmacokinetic (PBPK) models) incorporating in vitro data of human systems are being used increasingly to predict the clinical DDI potential associated with new chemical entities to help in the drug development process [12,13].

Development of PBPK Model in Dogs
A PBPK model was first developed in dogs using a hybrid PBPK approach employing the in vivo clearance reported in dogs [7]. The input parameters are described in Table 4. First, we used the raw data from Palo et al. [7] to obtain the dapaconazole compartmental model employing Phoenix ® WinNonlin ® , version 6.3 (Certara). The intravenous dapaconazole exposure was described as a bicompartmental model (data not shown). The PBPK model was first developed using a single 2 mg/kg intravenous dose of dapaconazole and employing the reported in vivo clearance of 591.7 L/min [7] in a hybrid approach. Then, the full PBPK model, which used the method 2 (Rodgers and Rowland [15]) to predict the volume of distribution at steady-state conditions (V ss ) was evaluated. The sensitivity analysis estimated the tissue-to-plasma partition coefficient (K p ) value of 0.01 to best fit the observed values. Despite that, the full PBPK model did not describe the dapaconazole bicompartmental profile. Considering this issue, in addition to the low volume of distribution (V d ) observed in dogs, the minimal PBPK plus a single adjusting compartment (SAC) was selected. The V sac (apparent volume of SAC), k in (rate constant from systemic compartment to SAC), and k out (rate constant from SAC compartment to the systemic compartment) values were determined from the parameter estimation approach selecting the values of 3.88 L/kg, 0.026 h −1 , and 0.016 h −1 , respectively, that better described the observed PK profile. The predicted V ss using Rodgers and Rowland's method [15] was 6.36 L/kg. The developed PBPK model described the observed dapaconazole PK profile in dogs reasonably well after intravenous single doses of 1 mg/kg, 2 mg/kg, or 20 mg/kg ( Figure 1); furthermore, the PK parameters' observed/predicted ratios were between 0.5-fold to 2-fold (Table 5).  Figure 1); furthermore, the PK parameters' observed/predicted ratios were between 0.5fold to 2-fold (Table 5).    To predict the role of CYP450 liver enzymes on the elimination of dapaconazole in dogs, the intrinsic clearance value obtained from dog liver microsomes (257.97 µL/min/mg) [14] was incorporated as the single elimination pathway in dogs. The predicted systemic clearance obtained after a single intravenous dose of 20 mg/kg dapaconazole from the model incorporating intrinsic clearance from dog liver microsomes in an in vitro model (164 mL/min) was 3.3-fold lower than the predicted systemic clearance (454.7 mL/min) from the model incorporating in vivo clearance in dogs. Considering this, the in vivo clearance value in dogs, and not intrinsic clearance from the microsome, was applied in the allometric scaling formula to predict clearance in humans.

Extrapolation of the PBPK Model Developed in Dogs to Humans
The allometry tools provided by Simcyp were employed to extrapolate the V ss and clearance (CL) from dogs to humans. Simple allometry with one species (dog) considered V ss, human = V ss,animal (L/kg) [16], and CL human = CL animal (mL/day) [17]. The input parameters included in the dapaconazole PBPK model for humans are described in Table 4.
To predict the role of CYP450 liver enzymes in the elimination of dapaconazole in humans, the intrinsic clearance value obtained from human liver microsomes (118.5 µL/min/mg) [14] was incorporated as the single elimination pathway in humans.
The elimination half-life (t 1/2 ) observed after simulating a single 20 mg/kg intravenous dose of (in an adult of 73 kg corresponding to a 1.460-mg dose) dapaconazole was 7.9 h. As an exercise in the simulation of potential DDI, we selected an intravenous dose of 500 mg every 8 h as a dosing regimen.

DDI Prediction of Dapaconazole as an Inhibitor
First, dapaconazole inhibition potential was evaluated by calculating the R1 value for reversible inhibition, demonstrating that dapaconazole has clinical inhibition potential in all CYP isoforms tested, except CYP2C8. To assess the clinical potential interaction of dapaconazole to the exposure (area under the curve ratio-AUC) of CYP substrate drugs, dapaconazole was evaluated as an inhibitor using static and dynamic models to predict DDI in clinical scenarios (Table 6) [18]. Considering both results from static and dynamic models, the worst-case scenario (higher AUCR value) was considered to classify the potential of dapaconazole as an inhibitor according to FDA classification [13]. Dapaconazole is a potential clinically weak inhibitor of CYP1A2 and CYP2C9, a moderate inhibitor of CYP2D6 and CYP2C8, and a strong inhibitor of CYP2C19 and CYP3A4. , inhibitor concentration that is the total plasma maximum concentration (C max ); Ki, inhibition constant; and AUCR, area under the curve ratio between AUC with inhibitor and AUC without inhibitor. Notes: Ki was corrected to the unbound value using the in vitro unbound fraction (f u,inc ) of 0.94 Ki = IC 50 /2; the mean C max of 9.5 µM in humans was obtained from simulations of 500 mg every 8 h and corrected by multiplying the ratio of unbound fraction in plasma (fu) with the blood-to-plasma ratio (Rb); the fm of nifedipine, midazolam, phenacetin, S-mephenytoin, and bufuralol were adapted from Simcyp. Paclitaxel was extracted from Hua et al. [19]; and diclofenac from Siu and Lai [20]. AUCR > 1.25 and <2: weak inhibitor; AUCR > 2 and <5: moderate inhibitor; and AUCR ≥ 5: strong inhibitor.

Discussion
Azole antifungals are used as a primary treatment for fungal infections, and to support the treatment of immune-suppressed patients, for example, organ transplant patients and patients with acquired immunodeficiency syndrome, who also use other medications. Azole antifungals can change the exposure of these medications due to DDIs caused by inhibiting drug-metabolizing enzymes [9,21]. Drug interactions represent a major problem in drug therapy. Therefore, prior knowledge of these interactions during the development process of new drugs through in vitro enzyme inhibition studies is of great value, as this helps in avoiding adverse reactions generated by the interactions [22].
Dapaconazole is metabolized by CYP450 enzymes in the liver as previously demonstrated by our research group using in vitro studies with liver microsomes from humans, dogs, and rats [14]. Other azole antifungals such as itraconazole and voriconazole also are substrates of CYP450 enzymes. In addition, azole antifungals are well-known CYP450 inhibitors, and the inhibition of CYP3A4 enzymatic activity is considered the main source of DDIs by these drugs [23][24][25][26][27].
Considering that one important application of the PBPK model during early drug development is to predict drug exposure prior to in vivo studies (mainly clinical trials), we employed this valuable tool to build a dapaconazole PBPK model in dogs, and recently it was extrapolated to humans in order to predict potential DDI scenarios in humans. The PBPK model developed in dogs, using a middle-out approach considering the observed in vivo clearance previously reported by Palo et al. [7], was qualified with the observed data through a visual predictive check and 0.5-to 2-fold ratio difference observed to predict the pharmacokinetic parameter value (Figure 1 and Table 5).
After developing and qualifying the PBPK model in dogs, we extrapolated the model to humans, assuming the same CL and V ss values from dogs. Other input parameters for the dapaconazole PBPK model in humans are described in Table 4. Later, we extrapolated this PBPK model to humans to perform a prospective clinical DDI prediction obtaining the AUCR dynamic ( Table 6).
The DDI potential analysis from Ki and R1 values ( Table 6) indicates that dapaconazole has a low potential to inhibit CYP2C8 (Ki: 52.1 µM; R1: 1.01), while it has a moderate potential to inhibit CYP1A2 (Ki: 1. The results of the DDI static and dynamic mechanistic models indicated that dapaconazole is a weak inhibitor of CYP1A2 and CYP2C9, a moderate inhibitor of CYP2C8 and CYP2D6, and a strong inhibitor of CYP2C19 and CYP3A, according to FDA guidelines [13]. In general, the AUCR static and dynamic values for each CYP isoform evaluated correlated reasonably well. To critically evaluate these results, it is important to highlight that the static model considers the worst-case scenario, assuming that the inhibitor concentration is maintained at the maximum plasma concentration (C max ) throughout the entire timed course, while the dynamic model uses the concentration versus time profiles of both inhibitor and substrate drugs, giving a more realistic prediction of a clinical DDI. Considering the AUCR ≥ 1.25 values obtained, we indicate that a clinical DDI study using a sensitive index substrate should be further performed for all CYP isoforms evaluated in the current work. Clinical DDI studies demonstrated that fluconazole, itraconazole, ketoconazole, posaconazole, and voriconazole are moderate to strong CYP3A inhibitors, and fluconazole, itraconazole, ketoconazole, miconazole, and voriconazole are moderate CYP2C9 inhibitors [9]. The findings of the present study indicate that dapaconazole has the same characteristics as other azoles as an inhibitor of many CYP isoforms with a moderate to strong inhibition in CYP3A, but a weak inhibition in CYP2C9.
Solutions (n = 3) containing 10 µL of probe substrate of each CYP450 isoform with or without 10 µL dapaconazole (0.01, 0.1, 1, 10, and 100 µM) or specific inhibitors were added to 1.5 mL propylene tubes and evaporated to dryness. Buffer solution (0.1 mol/L aqueous phosphate buffer, pH 7.4, 69 µL), NADPH-regenerating solutions (10 µL solution A: 26 mmol/L NADP+, 66 mmol/L glucose-6-phosphate, and 66 mmol/L MgCl 2 ; 2 µL solution B: 40 U/mL glucose-6-phosphate dehydrogenase in 5 mmol/L sodium citrate), and deionized water (69 µL) were added, in a total volume of 150 µL. The tubes were vortex-mixed and preincubated with continuous gentle shaking for 5 min at 37 • C in a shaking water bath. Reactions were initiated by the addition of 50 µL of HLM solution in the tubes which were gently mixed by hand and incubated with continuous gentle shaking at 37 • C. Aliquots of 200 µL ice-cold solvent (Table 1) containing IS (1 µg/mL) were added. The samples were vortexed for 5 min at 2000 rpm in a VXR basic Vibrax ® (Staufen, Germany) and then centrifuged at 16,000× g (Hettich ® MIKRO 185, Tuttlingen, Germany) for 15 min at 25 • C. The supernatants were transferred to glass autosampler vials and submitted to LC-MS/MS analysis (5 µL injection volume) to monitor the substrate metabolite formation.

Analysis by LC-MS/MS
The liquid chromatography (HPLC) system (Agilent Technologies, Santa Clara, CA, USA) consisted of a 1290 binary LC pump, a 1290 Infinity II Series TM autosampler, and an MCT 1290 column oven. Dapaconazole was separated in a Luna TM Omega polar C18 column (150 × 2.1 mm, 5-µm particle size; Phenomenex, Torrance, CA, USA) held at 40 • C using deionized water + 0.1% formic acid as mobile phase A and acetonitrile + 0.1% formic acid as mobile phase A at a flow rate of 0.3 mL/min. The applied gradient program consisted of 10% B, followed by a linear change to 100% B over 3 min. Mobile phase percentage B was then kept at 100% for 1 min and returned to initial conditions over 0.2 min (total run time of 5 min). The temperature of the autosampler was maintained at 5 • C.
Analytes and ISs were monitored in an API 4000™ triple quadrupole mass-spectrometer (AB Sciex, Concord, ON, Canada) with positive heated ion spray (Positive TurboIonSpray, MH + ) for analyte detection. Source conditions were gas (high-purity nitrogen) temperature of 300 • C, collision gas of 3 mTorr, and IonSpray voltage of 5000 V. The analysis was performed in multiple reaction monitoring (MRM) mode. The MRM transitions, collision energy (CE), and collision cell exit potential (CXP) are presented in Table 2. The data acquisition and quantification were performed using Analyst TM version 1.3.2 (AB Sciex, Concord, ON, Canada).

IC 50 Determination
The remaining CYP450 activity was calculated by comparing samples in the presence and absence of dapaconazole or selective inhibitors, according to Equation (1): where %REA is the percentage of remaining enzymatic activity, Ai corresponds to the metabolite-to-IS peak area ratio in the presence of dapaconazole or selective inhibitors, and A 0 corresponds to the metabolite-to-IS peak area ratio in the absence of dapaconazole or selective inhibitors. IC 50 values were determined by a nonlinear regression of the %REA of each CYP450 isoform versus the logarithm of inhibitor concentration, using GraphPad Prism version 5.01 software (GraphPad Software, San Diego, CA, USA).

PBPK Model Strategy
All modeling was conducted using Simcyp modeling software (v. 20, Certara, Princeton, NJ, US). A PBPK model was constructed to describe the pharmacokinetic profiles of intravenous dapaconazole in dogs, and it was extrapolated to predict drug-drug interactions in humans using a bottom-up approach (Figure 2). All of the input parameters are described in Table 4.

PBPK Model Strategy
All modeling was conducted using Simcyp modeling software (v. 20, Certara, Princeton, NJ, US). A PBPK model was constructed to describe the pharmacokinetic profiles of intravenous dapaconazole in dogs, and it was extrapolated to predict drug-drug interactions in humans using a bottom-up approach (Figure 2). All of the input parameters are described in Table 4. First, a dog PBPK model was developed to describe the observed dapaconazole PK profile from a published study from our research group [7]. The Simcyp minimal PBPK, which considers all organs other than the intestine and liver as a single compartment [36], plus a SAC distribution model were selected. This model better described the bicompartmental distribution profile of dapaconazole [7]. The Vss was predicted using Rodgers and Rowland's equations (method 2) [15]. A Kp scalar was determined to best describe the observed dog data using a sensitivity analysis approach. The final values of Vsac, kin, and kout were determined from the parameter estimation approach selecting the best values to describe the shape of the observed PK profile. For the elimination model development, the in vivo systemic clearance values reported in dogs [7] were employed in a middle-out approach. The intrinsic clearance value from dog liver microsomes obtained from a previous study by our research group [14] was employed to test the role of CYP-mediated hepatic clearance in the systemic elimination of dapaconazole in dogs. The PBPK model was evaluated through a visual comparison of observed in vivo plasma concentrationtime profiles with the concentrations predicted in dogs, and the PK parameters within observed/predicted ratios between 0.5-fold to 2-fold were considered acceptable.
The PBPK model developed in dogs considering the in vivo systemic clearance values reported in dogs [7] (middle-out approach) was extrapolated to predict the plasma concentration-time profile in humans and it was used in a prospective prediction of DDI.
The single species allometric scaling tool provided by Simcyp was used to extrapolate the CL and Vss in dogs to humans based on Equation (2): First, a dog PBPK model was developed to describe the observed dapaconazole PK profile from a published study from our research group [7]. The Simcyp minimal PBPK, which considers all organs other than the intestine and liver as a single compartment [36], plus a SAC distribution model were selected. This model better described the bicompartmental distribution profile of dapaconazole [7]. The V ss was predicted using Rodgers and Rowland's equations (method 2) [15]. A K p scalar was determined to best describe the observed dog data using a sensitivity analysis approach. The final values of V sac , k in , and k out were determined from the parameter estimation approach selecting the best values to describe the shape of the observed PK profile. For the elimination model development, the in vivo systemic clearance values reported in dogs [7] were employed in a middle-out approach. The intrinsic clearance value from dog liver microsomes obtained from a previous study by our research group [14] was employed to test the role of CYP-mediated hepatic clearance in the systemic elimination of dapaconazole in dogs. The PBPK model was evaluated through a visual comparison of observed in vivo plasma concentration-time profiles with the concentrations predicted in dogs, and the PK parameters within observed/predicted ratios between 0.5-fold to 2-fold were considered acceptable.
The PBPK model developed in dogs considering the in vivo systemic clearance values reported in dogs [7] (middle-out approach) was extrapolated to predict the plasma concentration-time profile in humans and it was used in a prospective prediction of DDI.
The single species allometric scaling tool provided by Simcyp was used to extrapolate the CL and V ss in dogs to humans based on Equation (2): CL hum is the intravenous clearance in humans, CL dog is the intravenous dog clearance, BW hum is the human body weight, BW dog is a dog body weight of 10 kg, and a and b are the allometric coefficient and exponent for dapaconazole, respectively.
In Equation (3), V ss,hum is the steady-state volume of distribution in humans, and V ss,dog is the steady-state volume of distribution in the dog.
V sac , K in , and K out values were kept as estimated for the dog model. The intrinsic clearance from HLM was also tested as the main elimination route. The Simcyp healthy volunteer population was considered, selecting individuals aged from 20 to 50 years old and a gender ratio of 1:1 for all the simulations in humans. The mean demographic parameters were 29.5 years old, 73 kg body weight, and 168.3 cm height. Simulations of 10 trials with 10 subjects were conducted with the dosing regimen of intravenous 20 mg/kg single dose or intravenous 500 mg every 8 h.

Dynamic Model Analysis Using PBPK for DDI Prediction
The potential of dapaconazole to inhibit CYP isoenzymes was predicted using the PBPK model extrapolated to humans. The Ki input values were estimated from the ratio IC 50 /2. Since the IC 50 experiments were designed considering the substrate concentration equal to Km, the simplification of Ki as IC 50 /2 was considered for a competitive inhibition reaction [37]. For the dynamic DDI model, the trial design included multiple intravenous dapaconazole 500 mg administrations every 8 h for 60 h and starting the protocol at 9 AM on day 1. Each CYP substrate (default substrate model provided by Simcyp) was administered at 9 AM on day 3 in a single oral dose in a fasted state: 150 mg phenacetin (CYP1A2), 0.25 mg repaglinide (CYP2C8), 500 mg tolbutamide (CYP2C9), 200 mg S-mephenytoin (CYP2C19), 20 mg bufuralol (CYP2D6), 5 mg midazolam, and 5 mg nifedipine (CYP3A).
The following Equation (4) was considered for the potential DDI evaluation: where AUCR is the area under the curve ratio, AUC with inhibitor is AUC in the presence of the inhibitor, and AUC without inhibitor is AUC in the absence of the inhibitor.

Static Model Analysis Using IVIVE for DDI Prediction
Initially, the ratio (R1) of intrinsic clearance values of a probe substrate for an enzymatic pathway in the absence and presence of the interacting drug (dapaconazole) for reversible inhibition was calculated according to Equation (5) [13]: where I max,u is the maximal unbound plasma concentration of the interacting drug at steady-state conditions, and K i,u is the unbound inhibition constant determined in vitro. The static DDI model employed Equation (6) to predict the AUCR: where [I] is the simulated Cmax , Ki is the inhibition constant, and fm is the fraction metabolized. Ki was corrected to the unbound value using the in vitro unbound fraction (f u,inc ) of 0.94; Ki = IC 50 /2. The mean dapaconazole C max of 9.5 µM in humans was obtained from simulations of 500 mg every 8 h and corrected by multiplying the ratio of unbound fraction in plasma (f u ) (0.037) [14] with the blood-to-plasma ratio (R b ) (6.08 predicted with Simcyp). Nifedipine, midazolam, phenacetin, S-mephenytoin, and bufuralol fm values were adapted from Simcyp; the paclitaxel fm value was extracted from Hua et al. [19]; and the diclofenac fm value was extracted from Siu and Lai [20].

Dapaconazole DDI Results Interpretation
Two recent guidelines provided by the FDA regarding in vitro [13] and clinical [38] DDI studies were considered for interpreting DDI results.
If R 1 ≥ 1.02, the potential DDI should be investigated further either using mechanistic (static and/or dynamic) models. The AUCRs obtained with static and dynamic models were evaluated according to the following criteria: AUCR > 1.25 and <2: weak inhibitor; AUCR > 2 and <5: moderate inhibitor; and AUCR ≥ 5: strong inhibitor. AUCR ≥ 1.25 based on static or dynamic mechanistic models indicates that a clinical DDI study using a sensitive index substrate should be further performed.

Conclusions
The isoforms CYP2C19, CYP2D6, and CYP3A4 were highly inhibited by dapaconazole, while CYP1A2 was moderately inhibited, CYP2C9 was weakly inhibited, and CYP1A6 was not inhibited when evaluated under in vitro inhibition studies with human liver microsomes. The hybrid intravenous dapaconazole PBPK model developed in dogs described the observed data reasonably well, and it was scaled up to humans. The dynamic (PBPK) and static DDI mechanistic model-based analysis suggest that dapaconazole is a weak inhibitor of CYP1A2 and CYP2C9, a moderate inhibitor of CYP2C8 and CYP2D6, and a strong inhibitor of CYP2C19 and CYP3A, considering a clinical scenario. The results presented may be a useful guide for future in vivo and clinical dapaconazole studies.