Structure- and isoform-specific glucuronidation of six curcumin analogs.

Abstract 1. In the present study, we aimed to characterize the glucuronidation of six curcumin analogs (i.e. RAO-3, RAO-8, RAO-9, RAO-18, RAO-19, and RAO-23) derived from galangal using human liver microsomes (HLM) and twelve expressed UGT enzymes. 2. Formation of glucuronide was confirmed using high-resolution mass spectrometry. Single glucuronide metabolite was generated from each of six curcumin analogs. The fragmentation patterns were analyzed and were found to differ significantly between alcoholic and phenolic glucuronides. 3. All six curcumin analogs except one (RAO-23) underwent significant glucuronidation in HLM and expressed UGT enzymes. In general, the methoxy group (close to the phenolic hydroxyl group) enhanced the glucuronidation liability of the curcumin analogs. 4. UGT1A9 and UGT2B7 were primarily responsible for the glucuronidation of two alcoholic analogs (RAO-3 and RAO-18). By contrast, UGT1A9 and four UGT2Bs (UGT2B4, 2B7, 2B15 and 2B17) played important roles in conjugating three phenolic analogs (RAO-8, RAO-9, and RAO-19). Interestingly, the conjugated double bonds system (in the aliphatic chain) was crucial to the substrate selectivity of gastrointestinal UGTs (i.e. UGT1A7, 1A8 and 1A10). 5. In conclusion, glucuronidation of six curcumin analogs from galangal were structure- and isoform-specific. The knowledge should be useful in identifying a curcumin analog with improved metabolic property.


Introduction
Diarylheptanoids are a class of natural products that share the 1,7-diphenylheptane skeleton. They are mainly distributed in the roots and rhizomes of Alpinia, Curcuma and Zingiber species (Lv & She, 2010). Curcumin from turmeric (Curcuma longa) is the one of the most famous diarylheptanoids for putative anti-inflammatory, anti-carcinogenic and anti-Alzheimer effects (Epstein et al., 2010). Galangal, the rhizome of Alpinia officinarum Hance (Zingiberaae), is a spice widely used in Europe and China for over 1000 years. Phytochemical studies of this medicinal plant have led to the isolation of many diarylheptanoids (curcumin analogs) that possess diverse pharmacological effects such as antiemetic (Shin et al., 2002;Yang et al., 2002), anti-inflammatory (Li et al., 2011;Matsuda et al., 2006), anti-tumor (Chen et al., 2015;Lee et al., 2012;Matsuda et al., 2009), and anti-oxidant activities (Kose et al., 2015;Riethmuller et al., 2015).
Curcumin and its analogs are a type of agents with diverse therapeutic potential. Thus, it is of great value to establish the structure-metabolism relationships in an attempt to identify a structure with favorable pharmacokinetics. In the present study, we aimed to determine the glucuronidation potential of six curcumin analogs using human liver microsomes (HLM) and twelve expressed UGT enzymes. Glucuronidation rate was measured by incubation assays of the analogs with UDPGA-supplemented enzyme preparations. Formation of glucuronide metabolite was confirmed using high-resolution mass spectrometry. Kinetic parameters were derived by fitting an appropriate model to the data of metabolic rates versus a wide range of substrate concentrations. Based on the metabolic rates and kinetic parameters, we analyzed the structureand isoform-specific glucuronidation of curcumin analogs.

Glucuronidation assay
A modified method was used to determine the glucuronidation activity of the tested enzymes toward six curcumin analogs as described previously (Lu et al., 2016). In brief, the incubation medium contained liver microsomes or expressed UGT enzyme (25 mg/ml), MgCl 2 (0.88 mM), saccharolactone (4.4 mM), alamethicin (20 mg/ml), and curcumin analog (dissolved in DMSO) in a total volume of 0.2 mL of 50 mM potassium phosphate (pH 7.4). After pre-incubation at 37 C for 5 min, the reaction was initiated by the addition of UGPDA (2 mM) and subsequently incubated at 37 C for 60 min. Later 100 ml ice-cold acetonitrile was added to stop the reaction, followed by vortex and centrifugation (10 min; 15 000g), and 10-ml aliquot of the supernatant was subjected to UPLC-QTOF/MS analysis. The organic solvent (DMSO) concentration in the incubation mixture was kept at 1%. All experiments were performed in triplicate. Preliminary experiments showed that the glucuronide formation rate was linear with incubation time (up to 180 min) and protein concentration (up to 0.2 mg/ml).

Structural characterization of metabolites by UPLC-QTOF/MS
To identify and characterize UGT metabolites, samples containing curcumin analogs and their metabolites were injected to an UPLC-QTOF/MS system (Waters Corporation, Milford, MA). Chromatographic separation was performed using the Waters ACQUITY UPLC system (Waters Corporation, Milford, MA) equipped with a Kinetex C18 column (2.1 Â 50 mm, 2.6 mm, Phenomenex, Torrance, CA). The mobile phase was formic acid (0.1%) in water (mobile phase A) and formic acid (0.1%) in acetonitrile (mobile phase B). The flow rate was set at 0.35 ml/min. The gradient elution program was 5% B at 0 to 1 min, 5 to 90% B at 1 to 3.3 min, and 90 to 5% B at 3.3 to 4 min. Mass spectrometry analysis was performed on the Xevo G2 Q-TOF/MS (Waters Corporation, Milford, MA) using the electrospray ionization source (under negative ion mode). The capillary, sampling cone, and extraction cone voltages were 2500, 25 and 4 V, respectively. The source and desolvation temperature were 110 and 350, respectively. The cone gas and desolvation gas was set to 30 and 600 l/h, respectively. To obtain informative fragment ions, collision energies ranging from 10 $ 30 eV were used for MS/ MS scanning. The spectra were acquired and processed using MassLynx software (Waters Corporation, Milford, MA).

Quantification of glucuronides
Glucuronides were quantified with the abovementioned Waters UPLC-QTOF/MS system using home-made reference standards (Waters Corporation, Milford, MA). The procedures for preparing glucuronide standards have been described elsewhere (Lu et al., 2016;Sun et al., 2015). In brief, the target   Figure 1. Structures, chemical names and coding names (in bracket) of six curcumin analogs from galangal. The arrows refer to the potential conjugating sites for UGTs. DOI: 10.1080DOI: 10. /00498254.2016 glucuronides were synthesized according to the glucuronidation assay protocol. Samples containing the glucuronides were subjected to the UPLC system equipped with a photo-diode array detector (Waters Corporation, Milford, MA). The elution fractions containing the metabolites were collected in vials and then dried using Eppendorf Concentrator Plus (Hamburg, Germany). The residue was reconstituted in phosphate buffer (pH 7.4) to make a stock solution. The concentration of stock solution was determined through complete conversion of glucuronide to parent molecule (by b-glucuronidase, 25 U) and quantitation of the parent molecule using its reference standard (measured at 280 nm). A series of working standards (0.005-10 mM) of glucuronides were prepared by serial dilution of the stock solution. These working solutions were verified to be stable during the analytical period. Quantification of glucuronides was performed using the exacted ion chromatograms of their respective precursor ions with a mass window of ± 0.05 Da (Table 1).

Absolute quantification of UGT protein concentrations in BD supersomes
The absolute amount of UGT isoform in BD supersomes was determined in order to make an accurate comparison of UGT activities. The enzymes of supersomes were trypsin-digested under the optimized condition and quantified using their signature peptides as described (Fallon et al., 2013a, b) with minor modifications. In brief, 10 ml of each recombinant enzyme (5-mg protein) was added to 30 ml of 100 mM ammonium bicarbonate in a PCR tube. Samples were reduced with the addition of 10 ml of 50 mM dithiothreitol and denatured by incubation at 60 C for 40 min. After cooling to room temperature, 10 ml of 60 mM iodoacetamide was added and the tubes were placed in the dark at room temperature for 30 min. Then, 5 ml of 0.1 mg/ml trypsin (prepared in 50 mM acetic acid) was added and the samples were digested at 37 C for 4 h. To stop the reaction, 35-ml of ice-cold acetonitrile was added and the tubes were placed on ice for 10 min. After vortex and centrifugation (13 000g, 10 min), 10-ml aliquot of the supernatant was injected to the UPLC-QTOF/MS system (Waters, Milford, MA) Samples were separated by a Waters BEH C18 column (1.7 mm, 2.1 Â 50 mm) maintained at 40 C. The mobile phase, consisting of phase A (water containing 0.1% (v/v) formic acid) and phase B (acetonitrile containing 0.1% (v/v) formic acid), was set at a flow rate of 0.45 ml/min. A gradient elution program was applied: 5% B at 0 to 1 min, 5 to 90% B at 1 to 3 min, 90% B at 3 to 3.3 min, 90 to 5% B at 3.5 to 4 min. Signature peptides were detected by the Waters Xeno G2 QTOF mass spectrometer with a electrospray ionization source (ESI) (positive ion mode). The capillary, sampling cone, and extraction cone voltages were 3000, 25 and 1 V, respectively. The source and desolvation temperature were 120 and 400 C, respectively. The cone gas and desolvation gas was set to 0 and 800 l/h, respectively. Peak areas of the extracted ion based on the full scan analysis were used to calculate the absolute concentration of each UGT isoform.

Kinetic evaluation
The formation rates of glucuronides were determined for six curcumin analogs according to the glucuronidation assay protocol. All substrates were evaluated at nine concentrations ranging from 0.39 to 75 mM (except for RAO-8). For RAO-8, substrate concentrations were 0.04-12.5 mM for UGT1A9, 0.04-25 mM for UGT2B4, and 0.2-75 mM for UGT2B15, respectively. The kinetic model Michaelis-Menten equation (Equation (1)) or substrate inhibition equation (Equation (2)) or Hill equation (Equation (3)) was fitted to the data of formation rates versus substrate concentrations using the Graphpad Prism V5 software (GraphPad Software, Inc., La Jolla, CA). Model selection was based on visual inspection of the characteristic Eadie-Hofstee plots (Hutzler & Tracy, 2002).
Michaelis-Menten equation: where V is the formation rate of product, V max is the maximal velocity, K m is the Michaelis constant and [S] is the substrate concentration. Substrate inhibition equation: where K si is the substrate inhibition constant. Hill equation: where S 50 is the substrate concentration resulting in 50% of V max , and n is the Hill coefficient. For the Michaelis-Menten and substrate inhibition models, the intrinsic clearance (CL int ) was derived by V max /K m ; For the Hill model, the maximal clearance (CL max ) was obtained using Equation (4).
Statistical analysis Data are presented as mean ± SD. Statistical differences were analyzed by one-way analysis of variance or Student's t test as appropriate, and the level of significance was set at p50.05.

Results and discussion
Structural characterization of six curcumin analogs and their glucuronides by UPLC-QTOF/MS The six curcumin analogs were introduced into the QTOF mass spectrometer. The analogs containing a phenolic hydroxyl group (i.e. RAO-8,  were detected, whereas those containing alcoholic hydroxyl (i.e. RAO-3, RAO-18, and RAO-23) were not (  (Table 1). They produced an identical product ion at m/z 173.43 (Table 1), suggestive of a similar fragmentation pattern. RAO-19 reduced the ethanol group to the chemical skeleton shared by RAO-8 and RAO-9 ( Figure 2). Further, the bond between C6 and C7 was cleaved, and the typical fragment ion at m/z 173.43 was generated upon a chemical rearrangement (Figure 2). HLM generated a single metabolite from each of six curcumin analogs. The metabolites have an increase of 176 Da in mass compared with their respective parent compounds, indicated that mono-glucuronides were generated (Table 1). This was corroborated by the fact that the metabolites cannot be generated in the absence of UDPGA, the glucuronide acid donor. Interestingly, the alcoholic and phenolic glucuronides differed in the fragmentation pattern (Table 1 and Figure 3). The alcoholic glucuronides (of RAO-3, RAO-18 and RAO-23) generated a fragment ion at m/z 193 Da, whereas the phenolic glucuronides (of RAO-8, RAO-9 and RAO-19) formed a fragment ion at m/z 175 Da (Figure 3). A difference of 16 Da in the molecular weight corresponded exactly to the addition/lack of an oxygen atom, revealing a distinct fragmentation between the alcoholic and phenolic glucuronides (Figure 3).
The product ion of m/z 193 Da was not observed in the alcoholic glucuronides such as the glucuronides of estradiol (17-glucuronide) and 3 0 -azido-3 0 -deoxythymidine (Barbier et al., 2000;Diaz-Cruz et al., 2003). However, this product ion was observed in the alcoholic glucuronide of nabumetone (Nobilis et al., 2004), as well as the acyl glucuronide of phenylacetic acids (Karlsson et al., 2010). Moreover, the curcumin analog hexahydro-O,O-dimethyl-curcumin also produced such a characteristic ion . The results from previous and current study suggested this ion might be a characteristic fragment for alcoholic glucuronides of curcumin analogs (in which the conjugated double bond system in the aliphatic chain of curcumin is reduced). Therefore, our findings provided a useful MS/MS approach for quick differentiation of alcoholic and phenolic glucuronides of curcumin analogs.

Reaction kinetics for glucuronidation of curcumin analogs by HLM
Kinetic profiling revealed that formation of RAO-3 glucuronide in HLM followed the classical Michaelis-Menten kinetics ( Figure 4). The glucuronidation profiles of RAO-8, RAO-9, RAO-18, and RAO-19 were well modeled by the substrate inhibition equation (Figure 4). In contrast, the glucuronide formation of RAO-23 in HLM was too slow (52 pmol/min/mg protein), disallowing a full kinetic characterization. The V max , K m , and CL int values ranged from 230 to 2320 pmol/min/mg protein, from 3.04 to 12.0 mM, and from 30.5 to 650 ml/min/mg protein, respectively (Table 2). Three phenolic substrates (RAO-8, RAO-9 and RAO-19) had significantly higher CL int values than those of RAO-3 and RAO-18 (Table 2). In addition, the CL int values of RAO-8 and RAO-19 were more than 2-fold higher than that of RAO-9 (Table 2).

Absolute enzyme levels in BD supersomes
Signature peptides of twelve UGT isoforms were synthesized and their structures were confirmed by MS/MS spectra ( Figures S1 & S2). The absolute levels of various UGT enzymes were determined by UPLC-QTOF/MS using their respect signature peptides. The UGT level in their respect BD supersomes was as follow: 1371 (UGT1A1), 722 (UGT1A3), 644 (UGT1A4), 523 (UGT1A6), 1004 (UGT1A7), 456 (UGT1A8), 1135 (UGT1A9), 1770 (UGT1A10), 1347 (UGT2B4), 354 (UGT2B7), 689 (UGT2B15), and 1549 (UGT2B17) pmol/mg. There was a large difference between UGT concentrations with a maximal difference of 5.0-fold. Consistently, Fallon et al. (2013a) revealed a large variability (up to 6.9-fold) in enzyme levels of recombinant UGTs. The results suggested that glucuronidation rates per mg total protein in vitro will not accurately reflect the true specific activity of UGT enzymes for a particular compound (Fallon et al., 2013a). Absolute quantification of UGT levels in recombinant materials herein allowed determination of glucuronidation as per nmol of UGT isoform, facilitating precise evaluation of enzyme selectivity toward a compound.

Reaction phenotyping with UGT enzymes
To identify the enzymes involving in the glucuronidation of curcumin analogs, twelve UGT isoforms were screened for Figure 2. MS/MS spectrum and fragmentation pathway of RAO-19. DOI: 10.1080DOI: 10. /00498254.2016 their catalysis activities (expressed as pmol/min/nmol) at the substrate concentrations of 5 and 50 mM. The metabolic profiles were similar at two test substrate concentrations ( Figure 5). UGT1A9 and UGT2B7 were the two enzymes capable of glucuronidating all six curcumin analogs ( Figure 5). UGT1A6, 1A8, 1A10 and 2B15 did not show any activities   Table 2 shows the derived kinetics parameters, and the glucuronide formation of RAO-23 in HLM was too slow (52 pmol/min/mg protein), disallowing a full kinetic characterization. toward three alcoholic substrates (RAO-3, RAO-18 and RAO-23) ( Figure 5). RAO-3 and RAO-18 were primarily metabolized by UGT1A9 and UGT2B7, whereas three phenolic substrates (RAO-8,  were extensively glucuronidated by multiple UGT1As and UGT2Bs ( Figure 5). In general, UGT2B7 showed the highest glucuronidation activities, whereas UGT1A3, 1A4 and 1A7 showed negligible activities. Compared with other five curcumin analogs, RAO-23 was a very-poor substrate for all UGT isoforms. This was consistent with its exceedingly low glucuronidation activity in HLM.

Glucuronidation kinetics by recombinant UGT enzymes
Based on the results of reaction phenotyping, we further performed kinetic analyses for active UGTs using a series of substrate concentrations ( Figures S3-S5 for kinetic profiles, Tables 3-5 for derived kinetic parameters). The V max , K m , and CL int values varied from 19.1 to 9681 pmol/ min/nmol, from 0.19 to 48.3 mM, and from 1.08 to 5797 ml/min/nmol, respectively. UGT1A9 and UGT2B7 were the main enzymes contributing to the glucuronidation of RAO-3 and RAO-18, whereas multiple enzyme (UGT1A9 and four UGT2Bs) were actively involved in glucuronidation of RAO-8, RAO-9 and RAO-19 ( Figure  6). Compared with alcoholic hydroxyl substrates, all UGT enzymes except UGT1A9 showed high affinity (with lower K m values) and high CL int /CL max values toward the phenolic substrates (Tables 3-5 and Figure 6).
The kinetics profiles of RAO-3 and RAO-18 by UGT1A9 (and UGT2B7) were well described by the Michaelis-Menten and substrate inhibition equation, respectively ( Figures S3 &  S5). These were in line with their glucuronidation profiles in HLM (Figure 4), suggesting that UGT1A9 and UGT2B7 were indeed two crucial enzymes responsible for hepatic glucuronidation of RAO-3 and RAO-18. UGT1A9 and four UGT2Bs were primary contributors to the glucuronidation of RAO-8 in HLM, as evidenced by the consistent substrate inhibition kinetics (Figures 4, S3 & S5). HLM-mediated glucuronidation of RAO-9 and RAO-19 followed the substrate inhibition kinetics (Figure 4), whereas glucuronidation of the two compounds by UGT2Bs (the main contributors) did not always followed the same kinetics ( Figure S5). RAO-23 was resistant to the metabolism by all UGTs, suggesting that a keto-enol tautomerism or a tight intramolecular hydrogen bond may preclude binding of the molecule to the UGT enzymes. Taken together, UGT1A9 and UGT2Bs played the most important roles in the glucuronidation of curcumin analogs.
RAO-8 and RAO-19, but not RAO-9, possess a methoxy group at the ortho-position of the phenolic hydroxyl group.  were much more efficiently glucuronidated compared with RAO-9. This clearly indicated that the addition of a methoxy group increased the metabolic susceptibility. It was interesting to note that the glucuronidation activities of three curcuminoids by human liver, intestine, and recombinant UGTs were in the order of curcumin > demethoxy-curcumin > bisdemethoxy-curcumin . Similar results were observed when hexahydro-curcuminoids were used as the substrates . Therefore, it may be a general rule that the methoxy group (close to the phenolic hydroxyl group) enhances the glucuronidation liability of the curcumin analogs.
The gastrointestinal UGTs (i.e. UGT1A7, UGT1A8, and UGT1A10) showed negligible or weak activities toward the six curcumin analogs (Table 4 and Figure 6). In line with this finding, glucuronidation activities of the gastrointestinal UGTs were significantly reduced when curcuminoids were transformed to their hexahydro derivatives . In contrast, the gastrointestinal UGTs (e.g. UGT1A8, and UGT1A10) showed high metabolic activities for curcuminoids Pfeiffer et al., 2007). Our results and previous one strongly indicated the Our results overall suggested that glucuronidation may be a determining factor to the pharmacokinetics of curcumin analogs from galangal. The oral bioavailability of these compounds may be significantly influenced by first-pass glucuronidation in the intestine and liver, highly expressing UGT1As and UGT2Bs. It was noteworthy that other metabolic enzymes such as cytochrome P450 and sulfotransferases might also contribute to the metabolism and pharmacokinetics of these curcumin analogs. On the other hand, it was uncovered that the glucuronidation was highly dependent on the chemical structure ( Figure 6). Therefore, it represented a promising approach to improve the pharmacokinetics of curcumin analogs through modification of their chemical structures. For example, an analog with a chemical structure of RAO-23 would be least glucuronidated thus is expected to have an improved oral bioavailability.

Conclusion
We have characterized the glucuronidation of six curcumin analogs by HLM and twelve expressed UGTs. It was found that the alcoholic and phenolic glucuronides differed in the generation of MS/MS product ions, allowing quick differentiation of these two types of glucuronides. Further, Figure 5. Comparisons of glucuronidation rates of six curcumin analogs by twelve expressed UGT enzymes at the substrate concentrations of 5 mM (A) and 50 mM (B).   glucuronidation was highly dependent on both the chemical structures and the isozymes. Compared with other analogs, RAO-23 was least glucuronidated by HLM and UGTs. The two alcoholic analogs (i.e. RAO-3 and RAO-18) were primarily glucuronidated by UGT1A9 and UGT2B7, whereas three phenolic analogs (i.e. RAO-8,  were mainly metabolized by UGT1A9 and four UGT2B enzymes. The knowledge should be useful in identifying a curcumin analog with improved metabolic property.