Pharmacokinetics and pharmacodynamics of three oral formulations of curcumin in rats.

Abstract

Curcumin (CUR) is a major component of turmeric Curcuma longa, which is often used in food or as a dietary supplement. The purpose of this preclinical study is to investigate the acute pharmacokinetic and pharmacodynamic (PK/PD) profiles of two commercially marketed CUR products (GNC and Vitamin Shoppe) and a CUR powder from Sigma in female rats. Plasma samples were collected at specific time points and analyzed for CUR and its metabolite curcumin-O-glucuronide (COG). RNA was extracted from leukocytes and analyzed for the expression of Nrf2-mediated antioxidant genes Nrf2, Ho-1, and Nqo1 by qPCR as selected PD markers. CUR PK was characterized by a 2-compartment model (2CM) after intravenous (IV) or oral administrations. Compared to IV CUR, the absolute bioavailability (F) of CUR for GNC (GC) is 0.9%, Vitamin Shoppe (VC) is 0.6% and Sigma (SC) is 3.1%. Pharmacodynamically, all three formulations showed induction of antioxidant Nrf2, Ho-1 and Nqo1 gene expression in rat leucocytes. PK/PD modeling of CUR’s effect on antioxidant gene expression was well captured by an indirect response (IDR) model. Physiologically based PK (PBPK) modeling and simulation using GastroPlus described the observed PK data reasonably well. In summary, our current study shows that the absolute oral bioavailability of the parent CUR was very low for all three formulations. However, despite the low CUR plasma concentrations, all three oral CUR formulations displayed PD response in the induction of Nrf2-mediated antioxidant genes, suggesting the potential of oral CUR contributing to the overall health beneficial effects of oral CUR.

Introduction

Curcumin (diferuloylmethane, CUR) is a yellow pigment presents in the dried rhizomes of Curcuma longa (turmeric), and turmeric is also commonly used as a dietary spice in some Asian countries. Furthermore, turmeric (Curcuma longa) is often used in traditional Indian and Chinese medicines for the treatment of various diseases [1] and commonly used in the US as botanical dietary supplement. CUR has been shown as the significant phytochemical responsible for the beneficial effects of turmeric. Many studies have investigated the important role of CUR in the prevention and treatment of many diseases, including cancer [2] [3], anxiety, cardiovascular disease [4], osteoarthritis, and metabolic disturbances such as diabetes mellitus and obesity [5] [6]. Also, many in vivo studies have demonstrated that CUR can be involved in various cellular signal transduction pathways and inhibits carcinogenesis such as the GI tract, including the colon, esophagus, stomach and liver [7]. The wide range of biological activities of turmeric has been attributed to CUR [8]. However, its experimental biological benefits in clinical trials have faced many barriers since the low bioavailability of CUR after oral administration might have contributed and confounded its pharmacodynamic (PD) responses in human subjects [9] and in rodents [10]. Previously, scientists have performed preclinical and clinical studies with CUR. For instance, UC-MS/MS study showed that maximum levels of CUR in the plasma of rats receiving a single oral dose of 500 mg CUR per kg body weight were around 0.06 μg/mL [11,12]. By comparing with an intraperitoneal (IP) dosing, the oral bioavailability of CUR was estimated as about 1%. Clinical trials in humans indicate that the systemic bioavailability of orally administered CUR is relatively low [13–15] and that mostly the metabolites of CUR, instead of CUR itself, are detected in the plasma or serum following oral administration [16,17]. These results are similar with our previous study showing , the plasma levels of parent CUR were below the detection limit of 0.1 ng/ml by HPUC-ITMS/MS/MS and only the metabolite, CUR glucuronide COG was detected as early as 30 minutes after oral administration [18]. This poor bioavailability can be explained by its poor absorption due to low water solubility (<0.1 mg/ml), limited tissue distribution and extensive metabolism in the intestines and liver [19]. Therefore, the poor bioavailability of CUR appears to be the principal barrier in achieving adequate blood circulating levels to give rise to desirable PD effects.

Glucuronidation is the major phase II conjugation reaction of many xenobiotics, including polyphenolic CUR [20,21]. With the addition of the glucuronic acid group of uridine-5’-diphosphoglucuronic acid (UDPGA), CUR molecules become more hydrophilic and are therefore more readily eliminated. This glucuronidation process is catalyzed by a superfamily of UDP-glucuronosyltransferases (UGTs), which are membrane-bound enzymes localized at the luminal side of the endoplasmic reticulum (ER). The glucuronidation pathway of CUR contributed greatly to the overall poor oral bioavailability observed in both preclinical models and human subjects and many formulations of CUR have been developed and studied. While many studies on CUR have been focused on improving its bioavailability and investigating its potential effects on disease, to date, few studies have compared the PKs and PDs of commercially marketed CUR. Additionally, no studies have been performed to study the PK/PD of CUR of different marketed formulations and its pharmacological effect on antioxidant gene activity.

CUR exerts both direct and indirect antioxidant effects by sweeping away reactive oxygen species (ROS) and stimulating the expression of cytoprotective proteins through the nuclear factor (erythroid-derived 2)-like 2 (Nrf2)-Keap1 signaling pathway. Nrf2 is an important regulator of the cellular response to oxidative stress. In this context, CUR can activate Nrf2-mediated downstream antioxidant genes such as heme oxygenase-1 (HO-1) [21,22], NAD(P)H dehydrogenase [quinone] 1 (NQO1) [10] and glutathione S-transferase P1 (GSTP1) (Fig 1). The ability of CUR to suppress inflammatory genes has also been widely reported in both human and rat studies. Knockout of Nrf2 attenuates CUR’s antioxidant and anti-inflammatory activities in mouse macrophages [23], suggesting the important role Nrf2 is playing in CUR’s biological responses. Hence, in our current study, we aimed to examine the PK/PD of commercially marketed CUR products (GNC and Vitamin Shoppe) and CUR from Sigma on the antioxidant response in preclinical rat model integrated with PK/PD modeling approach. In addition, PBPK modeling, which is a compartmental and flow-based type of PK model which could facilitate the simultaneous description of drug concentration changes over time in each organ [24]. The compartment may not be limited to entire organs, and often PBPK models may contain nested compartments that represent different cell types within an organ, and even different organelles within a cell. These levels of hierarchical complexity permit modeling of molecularly-driven events, such as specific metabolic pathways. In our current study, the Advanced Compartmental Absorption and Transit (ACAT^™) physiologically based PK (PBPK) model in GastroPlus will be utilized to help examine and enhance the understanding of the oral absorption process and PK of CUR. This kind of simulation with preclinical PK parameters could be potentially extrapolatable from rats to humans, and may inform dosing recommendation and the design of clinical trials.

Fig 1. Nrf2 transcription factor regulates the expression of antioxidant and Phase II detoxifying genes.

CUR will disrupt Keap1-Nrf2 binding and activates Nrf2, which then migrates into the cell nucleus and binds to the Antioxidant Response Element (ARE) present in the 5’ flanking region of many Nrf2-mediated genes. ARE then upregulates a variety of antioxidant enzymes and Phase II detoxifying genes.

Materials and Methods

Chemicals and reagents

CUR (98% purity as standard for quantification; Uot C7727), ethanol (99%), and formic acid (98%) were purchased from Sigma-Aldrich (St. Uouis, MO). CUR-d6 (>98%, Santa Cruz Biotechnology, TX), acetonitrile (ACN) and pure water for the UC mobile phase were purchased from Honeywell Burdick & Jackson (Muskegon, MI). Deionized water was obtained from a Milli-Q system (EMD MilliporeTM Direct-QTM3, Millipore, Bedford, MA). Heparin sodium injections (1000 U/ml) were purchased from Baxter Healthcare Corporation (Deerfield, IU), and sodium chloride injections (0.9%) were purchased from Hospira Inc. (Uake Forest, IU). GNC Herbal Plus®Turmeric Curcumin capsules (GC, Item #181212) was purchased from GNC, Vitamin Shoppe Curcumin C3 Complex capsules (VC, Item # DB-7025) was purchased from Vitamin Shoppe and Sigma curcumin powder (SC) was purchased from Sigma (Uot C1386). HPLC analysis of commercial formulations showed the content of individual curcuminoids mainly contains CUR, demethoxycurcumin, and bisdemethoxycurcumin. This was in close agreement with the manufacturer’s product data sheet that listed CUR content of 65 to 70% with the major impurities being demethoxycurcumin and bisdemethoxycurcumin. The CUR purity of three formulations was shown in Supplementary Table 1 (S1).

Animal treatment and sample collection

Four adult female Sprague-Dawley (SD) rats (250–300 g) per group with cannulated jugular veins were purchased from Hilltop Laboratories (Scottsdale, PA). The rats were housed and acclimatized for three days in the Laboratory Animal Service facility at Rutgers University and had full access to food and water before the PK experiment. This PK study was carried out according to the animal protocol approved by the Institutional Animal Care and Use Committee of Rutgers University (01–016). Heparinized saline (50 U/ml) was used to flush the inserted cannulas and tubes. CUR was suspended in vehicle consisting of Cremophor, Tween 80, ethanol and water at a volume ratio of 1:1:1:7. CUR at 250 mg/kg body weight was administered through oral gavage (PO). For the intravenous (IV) administration group, the same vehicle-suspended compound as the PO group was given as CUR 40 mg/kg. Blood samples (0.3 ml) were withdrawn through an exteriorized cannula at 0, 10, 20, 30, 40, 50, 60, 120, 240, 360, 480 and 720 min after the oral and IV administration of CUR. The cannula was flushed with one volume (0.3 ml) of heparinized saline solution immediately after each collection. All blood samples were centrifuged at 2,500 rpm and 4°C for 10 min to isolate plasma samples and then stored at −80°C for subsequent triple-stage ion trap mass spectrometry coupled with high-performance liquid chromatography (HPLC-ITMS/MS/MS) analysis.

Sample preparation

CUR samples in rat plasma were prepared by a novel protein precipitation extraction (PPE) coupled with size exclusion chromatography (SEC) method according to our previously published method [25]. Briefly, PPE was carried out with cold ACN (−20°C). Aliquots of the original rat plasma samples (100 μl) and CUR standard samples in blank SD rat plasma (100 μl, Innovative Research, MI) spiked with CUR-d6 as an internal standard (IS) were mixed with 300 μl of cold ACN and whirled for 30 s on a bench vortex. The resulting mixture was subject to an SEC procedure. It was centrifuged at 4°C and 16,000 g for 25 min. Then, 300 μl of sample supernatant was carefully transferred to an OMEGA NANOSEP 3 K size exclusion cut-off tubes (cut-off 10 kD, Pall Corporation, NY) and centrifuged at 4°C and 16,000 g for 30 min. Filtered CUR with internal standard (IS; curcumin-d6) samples were then added to 300 μl auto sample vials (Life Technologies, NY). Calibration standards were prepared by spiking a series of CUR working standard solutions and the IS into 100 μL of blank rat plasma, and then the same extraction procedure as that used for the plasma samples above was followed. We have previously reported this novel HPLC–ITMS/MS/MS analysis method performed by using a Thermo Finnigan LTQ mass spectrometer coupled with a Thermo Finnigan Surveyor HPLC System (Thermo Fisher Scientific Inc., San Jose, CA) [25]. Chromatographic separation was performed by a Zorbax Eclipse XDB Cl8 column (3.5 _m, 4.6 × 50 mm, Agilent Technologies, Santa Clara, CA). Briefly, aqueous mobile phase A was composed of 0.1% formic acid in water, and organic mobile phase B was ACN. The column oven and autosampler temperatures were 40°C and 4°C, respectively. The mobile phase was 25:75 (A: B, v/v) for the first 0.8 min at a flow rate of 250 μL/min followed by 15:85 (A: B, v/v) for 1.1 min at a flow rate of 250 μl/min and then was transitioned to 5:95 (A: B, v/v) at a flow rate of 400 μL/min in 2 min. The mobile phase was maintained at 5:95 (A:B, v/v) at 400 μl/min for an additional 4 min and then returned to 25:75 (A: B, v/v) at 250 μl/min for 3 min. The total run time for the present analysis was 11 min. The injection volume of the sample was 25 μl. The ITMS system was operated in a positive ESI ion source. The analytes were detected using consecutive reaction monitoring (CRM) mode. Data acquisition, peak integration and quantitation were achieved by the Xcalibur Data System (Thermo Electron, San Jose, CA, USA). The lower limit of detection was 1 ng/mL and upper limit of detection was 3000 ng/mL. The plasma calibration curve was linear, with r²>0.99.

Measurement of mRNA expression in leukocytes

Total RNA was extracted from leukocytes according to the protocol of the PicoPure® RNA Isolation Kit. RNA concentrations were quantified using an Infinite M200 NanoQuant spectrophotometer (Tecan, Mannedorf, Switzerland). First-strand cDNA was synthesized from 300 ng of RNA using TaqMan reverse transcription reagents (Applied Biosystems, Foster City, CA, USA). Quantitative polymerase chain reaction (qPCR) was conducted using SYBR Green reagent on a QuantStudio5 Real-Time PCR system (Thermo Fisher Scientific, Rockford, IL). The relative mRNA levels were calculated using the ΔΔCt approach [26], and GAPDH was used as a reference for gene expression normalization. PD data were presented as the gene fold change at each time point against their respective expression in the vehicle control arm.

PK/PD model development

This study investigated the PK of three different CUR formulation preparations in 4 groups of Female Sprague-Dawley rats 0-12 h after a single oral gavage or IV dose. CUR was administered at a dose of 250 mg/kg (oral) or 40 mg/kg (IV) to each rat. Plasma samples were quantified for free CUR and its glucuronidation conjugate, curcumin-O-glucuronide (COG), since CUR predominantly undergoes glucuronidation in the gut wall [27]. Noncompartmental analysis (NCA) was first conducted on the time course of CUR plasma concentration with Phoenix WinNonlin (Certara USA, Inc., Princeton, NJ, USA). And then the NCA PK parameters such as the area under the curve (AUC), clearance (CL), the mean residence time (MRT) and the volume of distribution (Vss) were estimated using Phoenix WinNonlin. The oral and IV CUR plasma concentrations versus time t were analyzed by a two-compartment (2CM) PK models (Fig 2a) and plasma levels of the COG were fitted with one-compartment (1CM) PK model. The parent CUR and metabolite COG PK model differential equations are as follows:

Fig 2. (A) Scheme of PKPD model and deconvolution of observed PK profile.

Schematic of two-compartment model that describes CUR PK following both IV and oral doses and PK-PD model of CUR-mediated induction of antioxidant/phase II detoxifying genes. Vc is the central compartment; Vp is the peripheral compartment; CLd is inter-compartmental clearance between the central and peripheral compartment; CL is the total clearance from the central compartment. Cp represents the plasma concentration; ka is First-order absorption rate constant; F is bioavailability; kin is the zero-order rate constant for the production of antioxidant genes’ mRNA, kout is the first-order rate constant for the degradation; Emax is the maximum ability of CUR to stimulate signal initiation; EC50 represents the concentration of CUR resulting in 50% of the maximum induction. (B) Average fraction input vs. time profiles for GC, VC, SC formulations estimated by deconvolution. Cumulative CUR input v.s. time, single oral dose.

2 CM oral and IV

$$\frac{dAa}{dt}=-ka·Aa;A{a}_0=\text{oral dose}$$ (1)

$$\frac{dAc}{dt}=ka·Aa·F+\frac{CL}{Vc}·Ap-\left(\frac{CL}{Vc}+\frac{CLd}{Vc}\right)·Ac-k_{met}·Ac;A{c}_0=IV\text{dose or}0$$ (2)

$$\frac{dAp}{dt}=-\frac{CLd}{Vp}·Ap+\frac{CLd}{Vp}·Ac;A{p}_0=0$$ (3)

1 CM metabolite (COG)

$$\frac{d{A}_{met}}{dt}=k_{met}·Ac-\frac{C{L}_{met}}{V_{met}}·A_{met}{A}_{met0}=0$$ (4)

ka is the first-order absorption rate constant, and F is the oral CUR bioavailability. The oral and IV datasets were fitted simultaneously with a two compartment model. Eqs. 1–3 show the differential equations for the 2CMIV and oral dosing, where Vc represents the volume of distribution in the central compartment and Vp is the volume of distribution in the peripheral compartment. CL represents the total clearance from the central compartment and CLd represents the clearance from inter-compartmental distribution clearance. Ac represents the amount of CUR in the central compartment and Ap represents the amount of CUR in the peripheral compartment; thus, the plasma concentration Cp equals Ac/Vc. Eq 4 described the differential equation for the 1 CM metabolite (COG), where kmet is the rate constant of COG input, Vmet is the volume of distribution of COG, CLmet represents the clearance of COG, A met represents the amount of COG in the compartment.

Gene expression of antioxidant genes (Nrf2, Ho-1, Nqo1) was measured as PD response (E) and described by an indirect response (IDR) model with baseline (Fig 2a) [28,29]. The plasma CUR concentration is linked to the rate of change of the signal induction by the Hill function. The initial biologic signal effector parameter mRNA₀ relates to the rise or fall of plasma CUR concentrations to the time course of changes in antioxidant gene expression, where mRNA₀ is the initial condition and is defined as mRNA0 = 1, which represents the gene expression before CUR administration. Finally, an IDR model with the stimulation of input (kin) by CUR was applied to describe mRNAgene formation (PD) [30]. The differential equations are shown in Eqs. (5–6.

$$\frac{dmRNA}{dt}=kin·\left(1+\frac{\mathit{Emax}}{EC50+C}·C\right)-kout·mRNA;mRN{A}_0=mRNA0$$ (5)

$$mRN{A}_0(t=0)=\frac{kin}{kout}=1$$ (6)

kin is the zero-order rate constant for the production of antioxidant genes’ mRNA, kout is the first-order rate constant for the degradation of mRNA, and mRNA0 is set as 1 as the initial condition before CUR administration. Emax is the maximum ability of CUR to stimulate signal initiation. EC50 represents the concentration of CUR resulting in 50% of the maximum induction. Cp is the CUR plasma concentration driven by the PK parameters obtained from the final 2CM PK model used as input variables to predict drug concentrations of CUR (C) in plasma for PD modeling. PK/PD models were conducted using R software (Version 3.5.2) with Ubiquity modeling platform. The code of the modeling differential equations are provided in the Appendix (S2). Each of the oral formulation datasets was fitted using the same structural model and separate sets of data were estimated.

The above PK/PD models were performed using R software (Version 3.5.2) installed with the PK/PD Model Development and Deployment package (Ubiquity) which tested with Strawberry Perl. All the PK parameters were estimated using the maximum likelihood method. The variance model was defined as ${\mathit{VAR}}_i={\left({\sigma }_1+{\sigma }_2\cdot Y(\theta ,t_i)\right)}^2$, where VAR_i is the variance of the i^th data point, σ₁ and σ₂ are the variance model parameters, and Y (θ, t_i) is the i^th predicted value from the PK model. The goodness of fit was assessed by system convergence, Akaike Information Criterion, estimator criterion value for the maximum likelihood estimation method, and visual inspection of residuals and fitted curves.

PBPK model simulation by GastroPlus

To simulate intestinal absorption and metabolism, GastroPlus^™ software (Simulations Plus, Lancaster, CA) embedded with the Advanced Compartmental Absorption Transit (ACAT) model was used to obtain the simulated plasma concentration-time profile. In our study, the permeability values were predicted by using different permeability models (patented by Affymax, Inc.US patent 6,043,027) for human jejunum effective permeability (Peff) based on CUR molecular structure with ADMET® Predictor. Then Peff was used to convert the permeability in rat in GastroPlus. All the physicochemical parameters (pKa, LogP, diffusion coefficient, solubility and permeability), as well as the systemic PK properties (central volume of distribution, blood to plasma ratio (B:P ratio) and unbound fraction (%Fu)) used in rat PBPK model building were obtained using ADMET® Predictor (Simulations Plus), which is a built-in module within GastroPlus®. The total clearance (CL) value of formulated CUR was from the estimated PK parameters from the observe data. The predicted parameters as shown in Table 4 were used as input parameters to perform the simulation. The physiologically based pharmacokinetic (PBPK) model for simulation of oral absorption was performed using Advanced Compartmental Absorption and Transit (ACAT) model to predict the Cp vs time PK profile of CUR. This process was validated by comparing the simulated PK profile and properties of CUR to our animal experimental data after oral absorption. Lastly, to better explain the variance of PK profile and evaluate the absorption kinetics among the three oral formulations, the deconvolution of the in vivo PK data was performed with Phoenix WinNonlin. The objective of Phoenix WinNonlin’s deconvolution is to estimate the cumulative amount and fraction absorbed over time for the individual formulation utilizing the PK profile dataset and dose for each profile.

Table 4:

Physicochemical and metabolism related in vitro and in vivo characteristics of CUR used in GastroPlus simulation

Parameters		Value	Source
MW (g/mol)	Molecular Weight	368.4	ADMET® Predictor
pKa	Ionization coefficient	9.5, 8.96, 8.32	ADMET® Predictor
logP	Partition coefficient	3.29	ADMET® Predictor
B:P ratio	Blood to plasma ratio	1.09	ADMET® Predictor
%Fu in rat plasma	Unbound fraction	6.29	ADMET® Predictor
CL (L/h/kg)	Elimination clearance	172.4	Experiment Data
Dose (mg/kg, Oral)	Dose	250	Experiment Design
Diff. Coeff (cm^2/s x 10^5)	Hayduk-Laudie infinite dilution diffusiona coefficient	0.67	ADMET® Predictor
S+Peff (cm/s*10^4)	Effective human jejunal permeability	5.14	ADMET® Predictor
Solubility (mg/ml)@PH=6.11	Solubility	0.045	ADMET® Predictor

Results

CUR Pharmacokinetics (PK)

The PK of CUR was first analyzed by noncompartmental analysis (NCA) and the results are shown in Table 1. The Cmax of SC formulation is 17.79 ng/ml, which was higher than the GC (12.6 ng/ml) and VC (9.92 ng/ml) formulations. The plasma concentration (Cp) versus time (t) profiles of CUR and its metabolite COG following oral dosing of 250 mg/kg and IV administration of 40 mg/kg parent CUR are presented in Fig 3. The Cp vs. t profiles after the oral and IV CUR administrations showed a biexponential decay that could be well captured by the 2CM models, as shown in Fig 3. By co-fitting the oral and IV datasets, the IV dose informed the parameters of 2 CM model, whereas the oral route provided the necessary data for estimating the parameters associated with the oral absorption (ka) part of the model. The two compartment oral and IV PK parameter estimations are summarized in Table 2. The PK profile of the metabolite COG fitted reasonably well with 1CM model and the PK parameters are summarized in Supplementary Table 2 (S3).

Table 1:

Parameters estimates of CUR pharmacokinetics with noncompartment model in rat plasma by Phoenix WinNonlin

PK parameters	Description	Estimation
		IV	Oral
			GC	VC	SC
Cmax (ng/ml)	Maximum concentration observed	-	12.6	9.92	17.79
Tmax (h)	Time to reach maximum concentration	-	0.33	0.33	0.83
AUC0-12h (ng/ml*h)	Area under the curve 0 to 12h	268.9	20.0	10.74	45.59
AUC0-∞ (ng/ml*h)	Area under the curve 0 to infinity	268.9	20.1	12.75	51.75
AUMC0-12h (ng/ml*h2)	Area under the first moment curve 0 to 12h	456.9	40.2	8.67	129.7
AUMC0-∞(ng/ml*h2)	Area under the first moment curve 0 to infinity	457.1	45.9	14.50	235.9
MRT(h)	Mean residence time	1.7	2.00	0.80	2.80
CL (L/h/kg)	Total clearance	112.4		--
Vss (L/kg)	Volume of distribution in steady state	62.8		--

Fig 3. Plasma concentration-time profile of oral, IV curcumin and CUR-O-glucuronide (COG).

Concentration-time profiles of PO CUR (solid line) and IV CUR (dash line) as fitted by a two compartment model. CUR-O-glucuronide (COG) of oral administration PK profile described by the one compartment model (dot line). GC, VC and SC represent CUR from GNC, Vitamin Shoppe and Sigma, respectively. Experimental observation data are shown as the mean +/− SD of plasma concentrations in 4 rats. The solid line represents the R software predicted curves after 250 mg/kg (oral) while the dash line represents the 40 mg/kg (IV) doses of CUR.

Table 2:

Pharmacokinetics parameters predicted from Phoenix WinNonlin compartment models.

PK parameters	Description	Estimation
		IV (CV%)	oral (CV%)
			GC	VC	SC
Vc(L/kg)	Volume of distribution in central compartment	43.6(0.6)*
Vp (L/kg)	Volume of distribution in peripheral compartment	278.4(9.9)*
CL (L/h/kg)	Elimination clearance	172.4(8.9)*
CLd (L/kg)	Inter-compartment clearance	53.2 (6.6)*
ka (h−1)	First-order absorption rate constant	-	3.1 (11.9)	1.8 (16.7)	3.23 (10.25)
F (%)	Bioavailability	-	0.9 (3.5)	0.6 (8.0)	3.1 (21.7)

^*oral and IV shared parameter

The deconvolution of the three different oral CUR formulations with their in vivo data indicated different cumulative input amount (Fig 2b). The SC formulation which has the highest cumulative input and more sustained absorption correlated well with the higher AUC and Cmax PK profile. Comparison between the absorption rate (ka) and cumulative input profiles shows that the absorption rate correlates with the deconvolution cumulative input profile. Compared to GC and VC, SC shows higher absorption rate and also has faster cumulative input profile than the other two formulations as shown in Fig 2b. The bioavailability (F) of the three formulations is shown in Table 2, and the %CV of the predicted parameters is within a reasonable range. The F of CUR from Sigma, SC is 3.1%, which is much higher than that of GC (0.9%) and VC (0.6%). The relative importance of the blood levels of free CUR with respect to the pharmacological effects will be discussed below. The plasma CUR concentration (PK) is linked to the rate of change of the biomarker induction (PD) and the PK/PD modeling fitted simultaneously within R as described in the PD section below.

PD of antioxidant gene expression

The mRNA levels of the PD response genes were assessed by quantitative reverse-transcription polymerase chain reaction (qRT-qPCR). The time course of the Phase II/antioxidant genes Nrf2, Ho-1, and Nqo-1 in leukocytes for each formulation is shown in Fig 4. The mRNA expression levels increased over time, and the peak time for gene induction was approximately 1.5-3 h following the oral administration of 250 mg/kg CUR. Furthermore, the fold changes of gene expression level of Nrf2 was 1.5, Ho-1 was 1.6, and Nqo-lwas 2.1, approximately. Subsequently, the mRNA levels declined gradually to baseline (Fig 4). An indirect response (IDR) model [28] was used to estimate the quantitative aspects of the mRNA expression profiles. Antioxidant gene expression reached its highest point before 3 h after oral treatment with CUR, and the estimated PD parameters for phase II genes are presented in Table 3. mRNA0 is fixed as 1 as the initial condition before CUR administration. The maximum effect (Emax) of CUR to stimulate Nrf2 is 1.74, Holis 2.02 and Nqo1 is 2.24. The production antioxidant genes’ rate (kin) for the three oral CUR formulations are 4.24, 4.86 and 5.53.

Fig 4. PD response of CUR with induction of antioxidant gene expression and PKPD modeling with IDR model for GC (A), VC (B) and SC (C).

Phase II detoxifying/anti-oxidant gene expression including Nrf2, Ho-1, and Nqo1 induced by CUR and captured by IDR model. The red triangles represent the mean of the observed data in rat leukocytes, and the blue lines represent the model prediction.

Table 3:

Pharmacodynamic parameters of mRNA expression estimated from indirect response (IDR) model

Parameters	Definition	Nrf2 (CV%)*	Ho1 (CV%)	Nqo1 (CV%)
kin	zero-order rate constant for the production of mRNA	4.24 (2.4)	4.86 (4.1)	5.53 (1.5)
Emax	Maximum ability of CUR to stimulate signal initiation.	1.74(1.2)	2.02 (14.5)	2.24 (0.3)
EC50	Concentration of CUR gives 50% of the maximum response	3.87 (0.1)	4.32 (3.2)	4.03 (0.2)
mRNA0	Initial condition	1.00 (Fixed)	1.00 (Fixed)	1.00 (Fixed)
Calculated Parameter	Definition	Nrf2	Ho1	Nqo1
kout	First-order rate constant for the degradation of mRNA	4.24	4.96	5.53

^*Percentage coefficient of inter-individual variability of the estimation

PBPK model simulation by GastroPlus

The aim of GastroPlus study was to simulate the preclinical PK behavior of different oral CUR formulations using an Advanced Compartmental Absorption and Transit (ACAT) physiologically based PK (PBPK) model incorporating physicochemical properties from GastroPlus. All the parameters used for the simulation are listed in Table 4. For the total clearance (CL) of CUR used in this PBPK simulation was estimated from the observed PK data (Table 2). The observed GC and VC plasma concentration-time profiles appear to be well-captured by the proposed ACAT PBPK model as shown by the simulated curves (Fig 5). SC formulation shows higher observed Cp-time PK profile than GC and VC formulations. The observed and simulated PK parameters of the different oral CUR formulations are listed in Table 5. The simulated PK parameters reasonably predict the observed PK parameters.

Fig 5. PBPK model using Gastroplus® for GC, VC and SC CUR in rat.

Simulation of plasma profile for oral dosing with ADMET Predictor using physicochemical parameters such as pKa, logP, effective permeability (P_eff), solubility and estimated clearance from experimentally observed data from Table 1. The blue solid line represents the mean concentration for the simulated population by GastroPlus. The different symbols denote the mean values (n=4) of the observed data.

Table 5:

Observed and simulated pharmacokinetics parameters of different CUR formulations.

Parameters	Observed Values			Simulated Values
	GC	VC	SC
AUC (ng/ml*h)	20.0	10.7	45.6	26.7
Cmax (ng/ml)	12.6	9.92	17.8	13.7
Tmax (h)	0.3	0.3	0.8	0.4
F (%)	0.9	0.6	3.1	1.2

Discussion

CUR is a common dietary phytochemical and botanical health supplement widely consumed because of its purported antioxidant and anti-inflammatory properties for a variety of diseases, such as arthritis, metabolic syndrome, and cancer [31]. CUR is pharmacologically more active than its conjugated metabolites and therefore it is assumed that the blood levels of free CUR would better reflect its pharmacological activity [27–31]. Previously, studies from our laboratory and from other scientists have shown that CUR could activate Nrf2, a master transcription factor regulating cellular protective gene expression as shown in Fig 1 [32]. Although there are a plethora of studies supporting CUR’s health benefits, its fairly low oral bioavailability presents a potential barrier for CUR to reach sufficient in vivo concentrations to achieve an adequate biological response. CUR undergoes substantial gut and hepatic first pass metabolisms after oral administration, which would contribute to the low systemic bioavailability [13,15]. To solve this problem, many scientists have developed and tested different formulations to hopefully achieve more favorable in vivo responses. This include administering CUR intravenously through liposomes [33] or micronized CUR [34]. Another common approach is to formulate CUR with piperine, which can enhance the bioavailability of CUR due to its potential in inhibiting intestinal and hepatic UGT metabolism [35]. In our current study, we compared 3 oral formulations, GC, VC and SC versus IV and all three oral formulations have measurable levels of free CUR with fairly high levels of CUR COG metabolite. Furthermore, the plasma concentration of CUR and the expression of phase II/antioxidant genes in leukocytes were measured simultaneously. Similar to the results of previous studies [9], the plasma levels of CUR in our current study were fairly low (0-18 ng/ml), and the CUR metabolite COG concentrations were fairly high. For many of later time points, the parent CUR concentrations were lower than the limit of quantification (LLOQ) (<= 0.1 ng/ml) are not shown in Fig 3. COG was detected at an earliest time point of 10 min after CUR administration and achieved maximal concentrations at approximately 1 h (Fig 3). Thus, the metabolism of CUR appears to occur very rapidly, forming the glucuronide conjugate. As shown in Table 1, the NCA PK parameters (and C_max and AUC of SC formulation are better than GC and VC. And the SC formulation of CUR tested in this current study demonstrated enhanced absorption compared to GC and VC formulations. To investigate the variation of CUR absorption phase, deconvolution of in vivo PK data was performed with Phoenix WinNonlin. The differences between each oral formulation as shown in Fig 2b could be due in part to disintegration/dissolution variation, potential impact of the different formulation on UGT metabolism and transport at the absorption site, or other potential factors that could contribute to this phenomenon. Unlike the SC powder, the commercial GC and VC capsules would contain additives (inactive ingredients such as magnesium stearate, microcrystalline cellulose and etc.) formulated with excipients which may affect the bioavailability of GC and VC as compared to SC. Some reports discussed that magnesium stearate interferes with the body’s ability to absorb the contents of the medication capsules [36]. Some other studies also reported that the dosage form related factors can affect drugs absorption. The major factors in this category are dissolution rate, particle size, polymorphism, amorphism, lipophilicity, ionization state, solvates, hydrates, and available surface area, and formulation factors including manufacturing, pharmaceutical ingredients, product age and storage conditions[37]. Setthacheewakul et al. showed that the formulations of CUR are more stable with a particle size of about 30 nm with 10 – 14 folds higher absorptions compared to the same oral dose of native CUR administrated in Wistar-strain rats [38]. In Fig 2b, the total cumulative absorption of the three oral CUR formulations is < 0.2 mg/kg, which is much less than the oral dose of 20 mg/kg. This result could be due in part to: a large portion of CUR may be unabsorbed; CUR molecules (parent and metabolites) are excreted unchanged in the feces either as unabsorbed or through biliary excretion; and or other factors [39]. Thus, we chose to present here the simplest approach where only the cumulative CUR input are simulated with the in vivo PK data used as input. Despite not representing a unique solution, we believe our analysis indicates that absorption/PBPK modeling could be considered as an approach to guide formulation development for some new drugs at early stages.

CUR metabolite COG can be explained by the important role that UGTs play in the metabolism of CUR. Glucuronidation is a major pathway in CUR metabolism in the rats, particularly in the intestine. Several isoforms, such as UGT1 Al, UGT1A7, UGT1A9, and UGT1A10, in human liver and human intestinal microsomes have been reported to be responsible, although the majority of UGT isoforms are able to catalyze CUR glucuronidation [21]. The PK profiles of oral, IV CUR and COG were well described by a 2CM and a 1CM (Fig 3) in agreement with previous works [34,40,41]. The bioavailability (F) of GC, VC, and SC was 0.9 %, 0.6% and 3.1%, suggesting the 2 commercially available products GC and VC are quite similar in their F. It would be interesting to find out the bioavailability and PK/PD of other commercially available CUR products currently on the market.

The beneficial effects of CUR could be explained in part via its actions on Nrf2 [41,42]. Numerous studies have observed the role of the activation of Nrf2 signaling by CUR in preventing diseases in animal models. For instance, in UVB-induced skin cytotoxicity models, CUR up-regulated Nrf2 and Ho-1 to reduce UVB cytotoxicity and oxidative stress [43]. CUR activates Nrf2 and increases antioxidant and anti-inflammatory activity to protect against cardiac injury [44]. CUR has also been observed to inhibit the transformation of cells from normal to tumor, and inhibit the synthesis of a protein thought to be instrumental in tumor formation [45]. By blocking the inflammatory molecule nuclear factor-kappaB (NF-κB), CUR blunts cancer-causing inflammation, slashing levels of inflammatory cytokines throughout the body [46,47]. CUR also interferes with production of dangerous advanced glycation end products that trigger inflammation, which can lead to cancerous mutation [48]. CUR alters cellular signaling to enhance healthy control over cellular replication, which tightly regulates the cellular reproductive cycle, helping to stop uncontrolled proliferation of new tissue in tumors [49]. To better understand the impact of different oral formulations of CUR on the biological responses and CUR’s antioxidant effects, Nrf2-mediated Phase II/antioxidant gene expression was measured. Nrf2, Ho-1, and Nqo1 expression level increased over time, and these changes were well-captured with an IDR model (Fig 4). Irrespective of the oral formulations, once CUR is absorbed, the systemic PK would drive the PD profile of each biomarker and can be modeled simultaneously with the shared PD model and the results are shown in Table 3. The levels of these three biomarkers Nrf2, Ho1 and NQO1 reached maximum peak values at approximately 1.5 to 3 h and then returned to baseline over time. It appears that all three CUR oral formulations exert similar PD response on the Nrf2-mediated antioxidant effects, and these PD responses would potentially contribute to the overall health beneficial effects of CUR.

This current PK/PD model represent out first attempt to characterize the anti-oxidant effects of CUR at the gene transcription level. Our current model can describe the observed dataset and provides certain flexibility to capture the dynamics of gene expression in different treatment groups. While this study used the relative gene fold change (normalized to each time point with the control time point (t=0)) as a PD marker, the circadian effects of mRNA expression may need to be considered in some situations. Although mechanistic consideration was incorporated in the current model, overall, it is still an empirical model. In the future, a more mechanistic based model could be further developed by including the intermediate molecules involved in the signal transduction at a series of time points, as reported in previous publication [50].

The ACAT physiologically based PK (PBPK) model in GastroPlus was utilized to enhance the understanding of oral absorption and PK of CUR. By inputting CUR’s physicochemical properties and the in vivo PK parameters (Table 4), the observed GC and VC PK data were fairly well-captured by the simulated PK profiles but not as well with SC (Fig 5). The reason for the differences are not clear, could be due in part to the vehicle (Cremophor, Tween 80, ethanol and water) used to suspend the CUR powder for SC versus capsules for GC and VC. Additional study with more marketed CUR products would be needed. Utilizing the simplified PBPK ACAT model, reasonably good predictions and better understanding of the role of physicochemical properties of CUR may play in the intestinal absorption kinetic processes resulting in the CUR PK profile as observed in the when compared to preclinical experimental study observations. Furthermore, many scientists have assessed the ability of PBPK models, more specifically the GastroPlus oral ACAT model, to predict oral absorption and exposure in preclinical species. For example, one study applied the GastroPlus oral ACAT model to a large dataset of 623 compounds from discovery projects. And the in vivo IV parameters were included into the model while the PO data served as observation control. Finally, the GastroPlus ACAT model provided reasonable oral absorption prediction [51]. The PBPK (ACAT) model for oral absorption, where the disposition and elimination were described by fitting a compartmental PK model to an in vivo IV profile, have also been studied in many other compounds: Parrott and Uave [52] and Jones et al. [53] evaluated such a model in rats respectively for 3 and 8 compounds. Generally, the predictions were reasonable, and when they were not, the model allowed hypothesis testing and an improved understanding of oral absorption.

This type of simulation exercise would inform dosing recommendations and or the design of clinical trials when evaluating CUR for a new indication. The model is readily implemented in GastroPlus and could be applied for drugs with similar metabolism pathways at various stages of drug development to evaluate if altered PK may occur in a gastrointestinal disorders population based on healthy subject data. In addition, a better and more ideal approach would be to conduct a comprehensive PBPK/PD study with the different organs and quantifying the PK and PD profiles.

To date, there are limited preclinical and clinical studies that have investigated the PK/PD of different commercially marketed formulations of CUR, a herbal medicinal product often taken as an over-the-counter botanical supplement. The presence of COG, the glucuronide metabolite of CUR, is due to intestinal metabolism by UGTs. Despite the very low concentrations of parent CUR in the rat blood/plasma, the antioxidant modulatory responses were observed in plasma for all 3 oral formulations, as shown by the increased gene expression levels of Nrf2, Ho1, and Nqo1. In summary, in our current study we conducted the PK and PD assessment after the IV and oral administrations of 3 different CUR products. The anti-oxidative properties of CUR as measured by the gene expression of Nrf2-mediated response in leukocytes would serve as potential surrogate PD biomarkers, implicating the potential health beneficial effects of the marketed CUR products, including the cancer chemopreventive effects. Moreover, the PK-PD modeling effort reasonably describes and links the PK and PK/PD of the three CUR products. Thus, this study presents a feasible approach to apply the PK/PD and PK/PD modeling and simulation in preclinical and clinical studies of botanical dietary supplements in general, including the understanding of a drug’s initial acute pharmacological response.