Evaluation for Potential Drug–Drug Interaction of MT921 Using In Vitro Studies and Physiologically–Based Pharmacokinetic Models

MT921 is a new injectable drug developed by Medytox Inc. to reduce submental fat. Cholic acid is the active pharmaceutical ingredient, a primary bile acid biosynthesized from cholesterol, endogenously produced by liver in humans and other mammals. Although individuals treated with MT921 could be administered with multiple medications, such as those for hypertension, diabetes, and hyperlipidemia, the pharmacokinetic drug–drug interaction (DDI) has not been investigated yet. Therefore, we studied in vitro against drug-metabolizing enzymes and transporters. Moreover, we predicted the potential DDI between MT921 and drugs for chronic diseases using physiologically-based pharmacokinetic (PBPK) modeling and simulation. The magnitude of DDI was found to be negligible in in vitro inhibition and induction of cytochrome P450s and UDP-glucuronosyltransferases. Organic anion transporting polypeptide (OATP)1B3, organic anion transporter (OAT)3, Na+-taurocholate cotransporting polypeptide (NTCP), and apical sodium-dependent bile acid transporter (ASBT) are mainly involved in MT921 transport. Based on the result of in vitro experiments, the PBPK model of MT921 was developed and evaluated by clinical data. Furthermore, the PBPK model of amlodipine was developed and evaluated. PBPK DDI simulation results indicated that the pharmacokinetics of MT921 was not affected by the perpetrator drugs. In conclusion, MT921 could be administered without a DDI risk based on in vitro study and related in silico simulation. Further clinical studies are needed to validate this finding.


Introduction
MT921 is a new injectable drug developed by Medytox Inc. intended to reduce submental fat, commonly referred to as double chin, through a minimally invasive medical procedure known as injection adipolysis. Its active pharmaceutical ingredient is cholic acid (CA), a primary bile acid synthesized from cholesterol and endogenously produced in the liver of humans and other mammals [1]. In adult humans, the synthesis of CA primarily occurs via the classical pathway of bile acids (BA) biosynthesis [2], inside hepatocytes. It starts with hydroxylation of cholesterol to 7α-hydroxycholesterol by the cholesterol 7α-hydroxylase (CYP7A1) enzyme [3]. Most BAs are reabsorbed in the ileum and sent to the liver by active transport, called enterohepatic circulation. This process involves various transporters, including organic anion transporting polypeptides (OATPs), organic anion Pharmaceuticals 2021, 14, 654 2 of 17 transporters (OATs), a Na + -taurocholate cotransporting polypeptide (NTCP), an apical sodium-dependent bile acid transporter (ASBT), and a bile salt export pump (BSEP) [4][5][6][7].
CA is an amphipathic molecule with surfactant properties capable of dissolving lipid bilayers. CA is particularly effective in adipolysis, a process that involves destabilizing and disintegrating the lipid bilayer of adipocytes, the major cell found in adipose tissue [8][9][10][11]. A single subcutaneous injection of MT921 into the submental region at strategic intervals results in a significant reduction of submental fat (the data is not shown due to confidentiality issues). Individuals treated with MT921 are likely to have a high body mass index (BMI), so they may presumably be afflicted with some form of metabolic syndrome, such as hypertension, diabetes, and hyperlipidemia, among others [12,13]. Therefore, there is a high probability that these individuals would be on some medication to treat their existing medical condition. Hence, it is imperative to determine the pharmacokinetic drug-drug interaction (DDI) of MT921 to avoid any antagonistic or adverse effect in these patients.
The present study assessed the in vitro pharmacokinetic interaction potential of MT921 following the US Food and Drug Administration (FDA) guidelines [14]. In addition, the effect of MT921 on pharmacokinetics of drugs for metabolic diseases was also investigated by performing physiologically-based pharmacokinetic (PBPK) modeling and simulation studies. H]taurocholate for ASBT). Uptake rate was determined at two different concentrations of MT921 (10 µM and 100 µM) in transporter-expressing cells and mock cells. The OATP1B1 transporter showed less than two-fold of uptake ratio, and thus was not involved in MT921 transport, according to US FDA guidelines [14], whereas OATP1B3, OAT3, NTCP, and ASBT showed greater than two-fold of uptake ratio, and thus could be involved in transport of MT921. The uptake rate of MT921 was higher in 10 µM of MT921 than 100 µM of MT921, indicating that the transporters may get saturated, thus plateauing the uptake (Table 1).
We confirmed the concentration-dependent uptake of MT921 by OATP1B3, OAT3, NTCP, and ASBT and calculated their kinetic values (K m , V max ) in HEK 293-OATP1B3, -OAT3, -NTCP, -ASBT, and mock cells, with increasing MT921 concentration (5-100 µM for OATP1B3, NTCP, and ASBT, 5-300 µM for OAT3) (Figure 1). Kinetic parameters of OATP1B3-mediated MT921 uptake were K m 27.4 µM, V max 94.88 pmol/mg protein/min, intrinsic clearance (V max /K m ) 3.46 µL/mg protein/min, and CL diffusion 1.33 µL/mg protein/min ( Figure 1A). In the OAT3-mediated MT921 uptake, K m , V max , intrinsic clearance, and CL diffusion were calculated as 133.33 µM, 170.76 pmol/mg protein/min, 1.28 µL/mg protein/min, and 1.14 µL/mg protein/min, respectively ( Figure 1B). For NTCP, K m , V max , intrinsic clearance, and CL diffusion were 30.54 µM, 409.28 pmol/mg protein/min, 13.4 µL/mg protein/min, and 1.27 µL/mg protein/min, respectively ( Figure 1C). In the case of ASBT, K m , V max , intrinsic clearance, and CL diffusion were calculated as 34.9 µM,  Figure 1D). In the final PBPK model, total hepatic clearance, ASBT, NTCP, OAT3, OATP1B3, Glomerular Filtration Rate (GFR), and enterohepatic circulation were utilized to describe the pharmacokinetic properties of MT921. Drug-dependent parameters are summarized in Table S8. Predicted plasma concentration of the simulated population was compared to plasma concentration-time profile of three doses of MT921 from a clinical trial (mentioned in the method section). Observed data of all doses were appropriately predicted in simulation. Semilogarithmic and linear plots of the simulated population for three dosages are shown in the Figures S4 and S5. A MT921 goodness-of-fit plot for the predicted plasma concentrations versus observed plasma concentrations of all studies was made. All observed data were included in 2-fold deviation ( Figure S6). The mean relative deviation (MRD) value for all predicted plasma concentration-time profiles was 1.13 (1.08-1.24) and geometric mean fold error (GMFE) values of AUC and C max were 1.10 (1.01-1.17) and 1.09 (1.04-1.16) (Tables S9 and S10). All predicted AUC and C max values were within the tw-fold compared with the observed AUC and C max . Detailed values of each dose are listed in Tables S9 and S10. Sensitivity was analyzed with a simulation of 150 mg MT921. The MT921 model in the 150 mg was sensitive to the properties of the lipophilicity (−2.35), the fraction unbound (−0.71), permeability (−0.07), and total hepatic clearance (−0.05). The result of sensitivity analysis is illustrated in Figure S7.
In the final PBPK model, total hepatic clearance, ASBT, NTCP, OAT3, OATP1B3, Glomerular Filtration Rate (GFR), and enterohepatic circulation were utilized to describe the pharmacokinetic properties of MT921. Drug-dependent parameters are summarized in Table S8. Predicted plasma concentration of the simulated population was compared to plasma concentration-time profile of three doses of MT921 from a clinical trial (mentioned in the method section). Observed data of all doses were appropriately predicted in simulation. Semilogarithmic and linear plots of the simulated population for three dosages are shown in the Figures S4 and S5. A MT921 goodness-of-fit plot for the predicted plasma concentrations versus observed plasma concentrations of all studies was made. All observed data were included in 2-fold deviation ( Figure S6). The mean relative deviation (MRD) value for all predicted plasma concentration-time profiles was 1.13 (1.08-1.24) and geometric mean fold error (GMFE) values of AUC and Cmax were 1.10 (1.01-1.17) and 1.09 (1.04-1.16) (Tables S9 and S10). All predicted AUC and Cmax values were within the twfold compared with the observed AUC and Cmax. Detailed values of each dose are listed in Tables S9 and S10. Sensitivity was analyzed with a simulation of 150 mg MT921. The MT921 model in the 150 mg was sensitive to the properties of the lipophilicity (−2.35), the fraction unbound (−0.71), permeability (−0.07), and total hepatic clearance (−0.05). The result of sensitivity analysis is illustrated in Figure S7.    Table S12. We compared population simulation to the observed plasma concentrationtime profile from 19 clinical studies. All population simulations adequately predicted the results from the clinical studies. Semilogarithmic and linear plots of population simulation of 19 clinical studies are shown in the Figures S8 and S9, respectively. The goodnessof-fit of AMLO confirmed that most of the observed values were included within twofold deviation lines ( Figure S10). The MRD of AMLO was 1.29. MRD values for all predicted plasma concentration-time profiles were within the two-fold deviation. GMFE values of AUC and Cmax were 1.16 and 1.19, repectively, also within the two-fold deviation (Tables S13 and S14). Sensitivity analysis results of a simulation of 10 mg AMLO are illustrated in Figure S11. The final model of AMLO was sensitive to the unbound fraction (−0.82), total hepatic clearance (−0.64), lipophilicity (0.46), and permeability (−0.24).

Prediction of Potential DDI between MT921 and Chronic Disease Drugs
With the developed PBPK model of MT921, AMLO, simvastatin (SIMV), and pioglitazone (PIO), potential DDI was tested when these chronic disease drugs were coadministered with MT921. The highest doses of single or multiple drugs were administered once a day for 9 days. At 10 days, 150 mg of MT921 was co-administered with chronic disease drugs. The plasma concentration of MT921 was predicted up to 24 h after administration. In all scenarios, the concentration of MT921 alone (purple line) was within the range of 5-95% concentration of MT921 with chronic disease drugs (pink shade), regardless of which drugs were taken ( Figure 3). This means that MT921 concentration did

Amlodipine Model
The final model of amlodipine (AMLO) contains information about hepatic clearance, GFR, and K i values of ASBT. The drug-dependent parameters of AMLO are listed in Table S12. We compared population simulation to the observed plasma concentration-time profile from 19 clinical studies. All population simulations adequately predicted the results from the clinical studies. Semilogarithmic and linear plots of population simulation of 19 clinical studies are shown in the Figures S8 and S9, respectively. The goodness-of-fit of AMLO confirmed that most of the observed values were included within two-fold deviation lines ( Figure S10). The MRD of AMLO was 1.29. MRD values for all predicted plasma concentration-time profiles were within the two-fold deviation. GMFE values of AUC and C max were 1.16 and 1.19, repectively, also within the two-fold deviation (Tables S13 and S14). Sensitivity analysis results of a simulation of 10 mg AMLO are illustrated in Figure S11. The final model of AMLO was sensitive to the unbound fraction (−0.82), total hepatic clearance (−0.64), lipophilicity (0.46), and permeability (−0.24).

Prediction of Potential DDI between MT921 and Chronic Disease Drugs
With the developed PBPK model of MT921, AMLO, simvastatin (SIMV), and pioglitazone (PIO), potential DDI was tested when these chronic disease drugs were coadministered with MT921. The highest doses of single or multiple drugs were administered once a day for 9 days. At 10 days, 150 mg of MT921 was co-administered with chronic disease drugs. The plasma concentration of MT921 was predicted up to 24 h after administration. In all scenarios, the concentration of MT921 alone (purple line) was within the range of 5-95% concentration of MT921 with chronic disease drugs (pink shade), regardless of which drugs were taken ( Figure 3). This means that MT921 concentration did not change with co-administration to AMLO, SIMV, and PIO. Linear and semilogarithmic graphs for not change with co-administration to AMLO, SIMV, and PIO. Linear and semilogarithmic graphs for comparing only plasma concentration of MT921 alone and MT921 co-administration are shown in Figures S12 and S13. Chronic disease drugs were administered for 9 days, once a day. At 10 days, 150 mg of MT921 was co-administered with chronic disease drugs. After administration, MT921 concentration was predicted. The purple line is a MT921-only population plasma concentration. The pink shade area is the 5-95% range of MT921 co-administration population plasma concentration. The green line is AMLO. The blue line is SIM. The gray line is PIO.

Discussion
MT921 is a fat-solubilizing drug, injected subcutaneously into the adipose tissue lying anterior to the platysma muscle to reduce submental fat. Since this drug is administered directly into topical sites, its bioavailability is expected to be very low. This was confirmed by a Phase I clinical trial of MT921, where a minimal amount of the drug was found in the systemic circulation of healthy volunteers administered MT921 subcutaneously. Nonetheless, we decided to investigate the in vitro pharmacokinetic DDI potential of MT921 when we had enough of the intended patients, many of whom were expected to be on multiple medications. Furthermore, we could simulate the change of MT921 AUC and Cmax when patients took drugs for their pre-existing metabolic conditions with MT921. Chronic disease drugs were administered for 9 days, once a day. At 10 days, 150 mg of MT921 was co-administered with chronic disease drugs. After administration, MT921 concentration was predicted. The purple line is a MT921-only population plasma concentration. The pink shade area is the 5-95% range of MT921 co-administration population plasma concentration. The green line is AMLO. The blue line is SIM. The gray line is PIO.

Discussion
MT921 is a fat-solubilizing drug, injected subcutaneously into the adipose tissue lying anterior to the platysma muscle to reduce submental fat. Since this drug is administered directly into topical sites, its bioavailability is expected to be very low. This was confirmed by a Phase I clinical trial of MT921, where a minimal amount of the drug was found in the systemic circulation of healthy volunteers administered MT921 subcutaneously. Nonetheless, we decided to investigate the in vitro pharmacokinetic DDI potential of MT921 when we had enough of the intended patients, many of whom were expected to be on multiple medications. Furthermore, we could simulate the change of MT921 AUC and C max when patients took drugs for their pre-existing metabolic conditions with MT921. Our in vitro study and in silico PBPK modeling predicted that there would be no potential DDI between MT921 and chronic disease drugs.
Our results revealed that MT921 seemed to have no significant inhibitory effect on cytochrome P450 (CYP) and the UDP-glucuronosyltransferase (UGT) enzyme or inductory effect on CYP isoforms and UGT isoforms (Tables S1, S2, S5 and S6) (IC 50 > 100 µM on CYP isoforms and UGT isoforms; less than two-fold change in the mRNA expression level of CYPs and UGTs compared with vehicle control). We confirmed that MT921 had an extremely rare chance of DDI with the metabolic enzymes recommended by FDA. In addition, MT921 was determined to not be a substrate and did not show inhibitory effects on efflux (MDR1, BCRP, MRP2, and BSEP) or uptake (OATP1B1, OATP2B1, OAT1, OCT1, OCT2, MATE1, and MATE2K) transporters (Table 1, Tables S3 and S4, Figures S1-S3). However, OATP1B3, OAT3, NTCP, and ASBT seemed to be involved in MT921 transport (Figures 1 and 2). Therefore, in our MT921 model, CYP and UGT-associated metabolic processes were not considered, but ASBT, NTCP, OAT3, and OATP1B3 were included in our final model to determine the potential drug interaction with MT921 (Figures 1 and 2).
In the PBPK model development, transporter K m values obtained from in vitro tests were used to describe pharmacokinetic characteristics, after adjusting the scale used in PKsim. K i values of the transporter reflected auto-inhibition and DDI and were calculated from the in vitro IC 50 value. Total hepatic clearance, lipophilicity, unbound fraction, first-order absorption, permeability, and intestinal specific permeability were optimized. Total hepatic clearance of [24-14 C] CA was included in metabolizing enzymes to cover unknown hepatic clearance [33]. Since the lipophilicity and unbound fraction data were sparse, they were optimized and found to be within the reported range of values proving its accuracy [34,35]. There was no data for absorption or distribution of MT921 after subcutaneou administration. Therefore, first-order absorption, permeability, and intestinal specific permeability were optimized to predict absorption and distribution of MT921. Three dosages of observed data were utilized in the evaluation step. Despite the lack of numbers of clinical trials in evaluation, MRDs and GMFEs of AUC and C max were within two-fold deviation. The GOF and population simulation results also verified correct predictions of observed data using the final PBPK model. Nonlinear PK of MT921 for different doses was also described well.
After that, MT921 model sensitivity was analyzed to identify the parameter that could have an impact on the AUC of this model [36]. If the sensitivity value is −1.0, a 10% increase in the parameters leads to a 10% decrease of the PK parameter value. Similarly, if a sensitivity value is 0.5, a 10% increase in the parameters leads to a 5% increase in the PK parameter value. The results indicated that the AUC of MT921 is sensitive to lipophilicity and the unbound fraction, but not K cat and K i of the transporter. It is speculated that MT921 DDI would not be associated with those transporters.
To predict ADME properties of amlodipine, physicochemical properties, Weibull, total hepatic clearance, GFR, permeability, and intestine specific permeability were implemented. Amlodipine is majorly metabolized by CYP3A4 and CYP3A5 [37]. Although it could be more accurate to implement these metabolizing enzymes, total hepatic clearance was utilized as the dominant metabolic process in our model. As Weibull, permeability, and intestinal specific permeability could not be extracted from the published paper, these values were optimized in our model. Amlodipine model was evaluated using 19 clinical trials. Lastly, K i value of ASBT was added to develop the amlodipine model to study DDI. The MRD of amlodipine was 1.29. GMFEs of AUC and C max were 1.16 and 1.19, respectively, and MRD and GMFEs were within two-fold. Sensitivity of the amlodipine model was analyzed to investigate parameters that affect drug exposure. As mentioned above, if the sensitivity value was bigger, the parameter greatly affected the PK parameter. The results show that the amlodipine model was sensitive to unbound fraction, total hepatic clearance, lipophilicity, and permeability. Other parameters decreased drug exposure, while lipophilicity increased drug exposure.
In the SIMV and PIO final model, the reproducibility of the model was first confirmed, and then these models were used. The K i value for transporter was added to inhibit the transporter in DDI simulation [38][39][40][41].
In the simulation step, scenarios of DDI were set, as well as the period in which inhibition could occur. The highest doses of three chronic disease drugs were administered, along with the MT921. K i values of MT921 were obtained from the in vitro test. AMLO, SIMV, and PIO K i values were taken from literature. All of these drugs were assumed to competitively inhibit each other [42]. Before DDI simulation, drug interaction among chronic disease drugs was investigated. DDI between AMLO and SIMV was reported [43]. When SIMV was co-administered with AMLO, AUC and C max of SIMV increased 1.8-and 1.9-fold, respectively. DDI between AMLO and SIMV was reflected in our DDI simulation by adjusting the SIMV final model. In DDI prediction simulation, MT921 concentration with chronic disease drugs was not significantly different from those with MT921 administration alone. In other words, MT921 PK was not affected by transporter inhibition. Since the concentration of MT921 did not change, we checked whether inhibition functions would work well in DDI simulation. Considering the sensitivity value of the MT921 transporter, it is confirmed that the concentration of MT921 changed when the inhibition was changed by the dose or K i value. Therefore, the inhibition of the transporter was well implemented in DDI simulation. These results matched with inhibitor concentration and its IC 50 . C max of AMLO and SIM in the clinic was 2000 times lower than their IC 50 . For PIO, C max of ASBT was 30 times lower and NTCP was two times lower than its IC 50 , though C max of OAT3 was two times higher than its IC 50 . Since MT921 is rarely distributed in the kidney, where OAT3 is located (data not shown), PIO would not inhibit MT921.
Due to the lack of observed data, the MT921 model development and DDI simulation had a limitation. Although the MT921 model showed good results in the model evaluation, results may not be highly reliable because the same data was used for model development and evaluation, due to the lack of clinical trials. Moreover, since DDI simulation had no real-world data, all potential DDI simulations were predicted models, not developed models. These limitations should be considered while interpreting these models.

Materials
The active ingredient of MT921, cholic acid (CA), was synthesized from Medytox Inc.
The The PBPK model of MT921 and AMLO were developed using PK-sim ® (open systems pharmacology site 9.1 www.open-systems-pharmacology.org (accessed on 21 January 2021)). Model parameter optimization (Monte-Carlo algorithm) and sensitivity analysis were performed using PK-sim ® . Plasma concentration-time profiles from the literature were digitized with WebPlotDigitizer Version 4.4 [53]. Calculation of quantitative model evaluation, PK parameter analysis, and graph plotting were accomplished with R 4.0.2 (the R foundation for statistical computing) and R studio 1.4.1103 (R studio, Inc, Boston, MA, USA).

PBPK Model Development
The PBPK model was developed using in vitro, in vivo, and clinical study data. Information of physicochemical properties, absorption, distribution, metabolism, and excretion (ADME) were used to reproduce compound characteristics. Clinical studies data (observed data) were used to create a data set, consisting of a training set and test set, for model development and evaluation. Demographic data of healthy individuals was used. If a clinical study did not provide demographic information, default values from the PK sim population database were used. Metabolic enzymes and transporter proteins were implemented using the PK-sim gene expression database [54]. Model parameters were then optimized by PK sim calculation methods for fitting predicted simulation to observed concentration of clinical study.

MT921 Model Development
MT921 in vitro data about the transporter and a clinical study containing demographic information (age, height, and weight) were provided by Medytox Inc. The rest of the in vitro data about physicochemical properties and ADME information of MT921 were retrieved from published literature. To develop a MT921 model, physicochemical properties, ASBT, NTCP, OAT3, OATP1B3, total hepatic clearance, GFR, and EHC recirculation was implemented. We found that MT921 is a substrate of ASBT, NTCP, OAT3, and OATP1B3, and MT921 can inhibit ABST, NTCP, OAT3, and OATP1B3. Experimental K m and V max values of ASBT, NTCP, OAT3, and OATP1B3 were utilized. The K i value was calculated from the IC 50 value using the Cheng-Prusoff equation where K i is the inhibition constant, IC 50 is half of the maximal inhibitory concentration of MT921, [S] is the concentration of substrate, and K m is the substrate concentration required for half of the maximum rate of transport. These values are shown in Figures 1 and 2. K cat of MT921 was calculated by the PK-sim-embedded Michaelis-Menten calculation method. To explain unknown clearance, total hepatic clearance was used. Total hepatic clearance was obtained from [24-14 C] CA clearance [33]. GFR and EHC recirculation values were assumed to be 1.
One clinical study was used as a training set; MT921 of 60 mg, 120 mg, and 150 mg were administered subcutaneously. If there was no information on sex in the clinical study, the population was assumed to be 100% male. A list of clinical studies is shown in Table S7. Partition coefficients and cellular permeability were taken from those calculated by Schmitt [55] and the PK-sim standard calculation method. Model parameters that could not obtain exact values from literature were optimized to observe the data of the training set.

Amlodipine Model Development
To develop the AMLO model, data on physicochemical properties, information about ADME, and clinical studies of AMLO were extracted from published literature. Total hepatic clearance and GFR were implemented to describe metabolism and excretion. Among the 19 clinical studies with two repeated doses and 17 single doses, 7 clinical studies are used as the training set and 12 as the test set. All AMLO was administered orally, 2.5-10 mg. Asian [56] demographic information was used for Korean and Chinese subjects whose demographic data was not provided. European [57] and Japanese (2015) demographic information was used for Caucasian and Japanese. All 19 clinical studies are shown in Table S11. Partition coefficients and cellular permeability was calculated using the Rogers and Rowland method [58,59] and PK-sim standard method. Model parameters whose exact values were not obtained from literature were optimized to fit the predicted simulation to observed data.

PBPK Model Evaluation
For model evaluation, several methods were used. As a visual comparison of the model performance, population simulations and goodness-of-fit were conducted. In total, 100 individuals were used in population simulations. If no demographic information was available, 20-50 years of age and male gender were assumed. The population predicted plasma concentration-time profile was compared to that observed in clinical data. Comparison of predicted versus observed plasma concentration of all studies was plotted in the goodness-of-fit plot. As a quantitative measure of performance of the model, the MRD of all PK parameters and the GMFE of all predicted PK parameters were calculated. If MRD and GMFE values were less than 2, it was considered to be adequate model performance.
The equation of MRD and GMFE are: where C predicted,i is predicted plasma concentration, C observed,i is observed data from literature, and n is the number of observed values.
where PK parameter predicted,i is predicted AUC and C max , PK parameter observed,i is observed AUC and C max from literature, and m is the number of studies.

Sensitivity Analysis
The sensitivity analysis of the final PBPK model was performed to determine how other parameters affect PK parameters. All parameters that could influence PK were included. Sensitivity analysis was performed for the simulation using the highest dose of the drug (MT921 150 mg, AMLO 10 mg) and conducted with the Sensitivity Analysis tool in PK-sim. The variation range and maximum number of steps were set to 10 and 9, respectively. The sensitivity of each parameter was calculated as the ratio of the relative change in the AUC to the relative change in the parameter. The equation is below: where S is the sensitivity of the area under the curve; AUC is the area under the curve; ∆AUC is the change in the area under the curve; ∆P is the change in the model parameter value; P is the original parameter value in final model.
cluded. Sensitivity analysis was performed for the simulation using the highest dose of the drug (MT921 150 mg, AMLO 10 mg) and conducted with the Sensitivity Analysis tool in PK-sim. The variation range and maximum number of steps were set to 10 and 9, respectively. The sensitivity of each parameter was calculated as the ratio of the relative change in the AUC to the relative change in the parameter.
The equation is below: where S is the sensitivity of the area under the curve; AUC is the area under the curve; Δ AUC is the change in the area under the curve; ΔP is the change in the model parameter value; P is the original parameter value in final model.
The equation of PK parameter ratio is below: DDI PK parameter ratio = PK parameter PK parameter (5) where PK parameter is AUC and Cmax.

Conclusions
To verify the DDI of MT921s with other drugs, we conducted various in vitro assays according to the DDI guidelines by the FDA, as well as in silico prediction using PBPK models. In summary, none of the results of in vitro assays indicated a significant DDI risk for metabolism or transport, considering the exposure from phase 1 clinical trials of MT921. Furthermore, we found no DDI possibility of MT921 with potential coadministrated drugs related to metabolic disease syndrome, using the PBPK modeling approach in various scenarios. Therefore, MT921 could be administered without a DDI risk based on in vitro study and related in silico simulation. Further clinical studies are needed to validate this finding. To estimate changes in PK parameter of MT921, DDI PK parameter ratio was calculated using PK parameters of MT921 administered alone and co-administered.
The equation of PK parameter ratio is below: DDI PK parameter ratio = PK parameter MT921 during co−administration PK parameter MT921 alone (5) where PK parameter is AUC and C max .

Conclusions
To verify the DDI of MT921s with other drugs, we conducted various in vitro assays according to the DDI guidelines by the FDA, as well as in silico prediction using PBPK models. In summary, none of the results of in vitro assays indicated a significant DDI risk for metabolism or transport, considering the exposure from phase 1 clinical trials of MT921. Furthermore, we found no DDI possibility of MT921 with potential co-administrated drugs related to metabolic disease syndrome, using the PBPK modeling approach in various scenarios. Therefore, MT921 could be administered without a DDI risk based on in vitro study and related in silico simulation. Further clinical studies are needed to validate this finding.  Figure S10: Goodness-of-fit plot of amlodipine, Figure S11: Amlodipine model sensitivity analysis, Figure S12: Linear graph comparing plasma concentration of MT921 administered alone and co-administered, Figure S13: Semilogarithmic graph comparing plasma concentration of MT921 administered alone and co-administered, Table S1: Summary of inhibitory potential of MT921 on nine CYP isoform catalytic activities in pooled human liver microsomal incubation system, Table S2: Summary of the inhibitory potential of MT921 on six UGT isoform catalytic activities in the pooled human liver microsomal incubation system, Table S3: Apical to Basal (A to B) and Basal to Apical (B to A) directional apparent permeability of MT921 (10 µM and 100 µM) in MDCKII-MDR1, -BCRP, -MRP2, and -BSEP, Table S4: The uptake transport rate of MT921 (10 µM and 100 µM) in HEK 293-OATP2B1, -OAT1, -OCT1, -OCT2, -MATE1, -MATE2K, and mock cells, Table S5: Cell viability of MT921 (0-100 µM) in cryopreserved human hepatocytes, Table S6: Effects of MT921 (1, 10, and 100 µM) on CYPs, UGTs, and transporter mRNA expression in cryopreserved human hepatocytes, Table S7: MT921 clinical studies, Table S8: Drug-dependent parameters of the MT921 PBPK model, Table S9: MRD values of predicted MT921 plasma concentration, Table S10: Observed and predicted MT921 AUC and Cmax values with mean GMFE values, Table S11: Summary of clinical studies used for the amlodipine model, Table S12: Drug-dependent parameters of the amlodipine PBPK model, Table S13: MRD values of predicted amlodipine plasma concentration, Table S14: Observed and predicted amlodipine AUC and Cmax values with mean GMFE values.