Multidrug Resistance Reversal Effects of Aminated Thioxanthones and Interaction with Cytochrome P450 3A4

- Purpose. Aminated thioxanthones have recently been described as dual-acting agents: growth inhibitors of leukemia cell lines and P-glycoprotein (P-gp) inhibitors. To evaluate the selectivity profile of thioxanthones as inhibitors of multidrug resistance (MDR), their interaction with other ABC transporters, which were found to have a strong correlation with multidrug resistance, such as multidrug resistant proteins 1 (MRP1), 2 (MRP2) and 3 (MRP3) and breast cancer resistance protein (BCRP) was also evaluated. The interaction of thioxanthones with cytochrome P450 3A4 (CYP3A4) together with the prediction of their binding conformations and metabolism sites was also investigated. Methods . The UIC2 monoclonal antibody-labelling assay was performed using P-gp overexpressing leukemia cells, K562Dox, incubated with eight thioxanthonic derivatives, in order to confirm their P-gp inhibitory activity. A colorimetric-based ATPase assay using membrane vesicles from mammalian cells overexpressing a selected human ABC transporter protein (P-gp, MRP1, MRP2, MRP3, or BCRP) was performed. To verify if some of the thioxanthonic derivatives were substrates or inhibitors of CYP3A4, a luciferin-based luminescence assay was performed. Finally, the in silico prediction of the most probable metabolism sites and docking studies of thioxanthones on CYP3A4 binding site were investigated. Results . Thioxanthones interacted not only with P-gp but also with MRP and BCRP transporters. These compounds also interfere with CYP3A4 activity in vitro , in accordance with the in silico prediction. Conclusion . Thioxanthonic derivatives are multi-target compounds. A better characterization of the interactions of these compounds with classical resistance mechanisms may possibly identify improved treatment applications.


INTRODUCTION
Cancer is a major health concern worldwide and the emergence of cell populations resistant to multidrug-based chemotherapy constitutes a major obstacle to treatment (1). Multidrug resistance (MDR) involves cellular resistance to several structurally unrelated drugs and one of the possible reasons is the overexpression of ATPbinding cassette (ABC) transporters (2). These transporters pump the antitumor agents to the extracellular medium, removing them from the site of action, thereby rendering the treatment ineffective (3). The first ABC transporter whose overexpression was shown to confer a MDR phenotype was P-glycoprotein (P-gp), the product of the mdr1 (ABCB1) gene (4). P-gp transports neutral or positively-charged hydrophobic substrates, particularly anticancer drugs, consuming energy from ATP hydrolysis (1).
Drug discovery in the last century has largely been based on the concept of "one molecule -one target -one disease" but more recently there has been a growing recognition that molecules that modulate multiple targets simultaneously can be beneficial for treating several diseases (5).
The biological interest in thioxanthones started with the use of lucanthone and hycanthone as antischistosomal agents. However, their ability to intercalate into DNA (6) prompted these compounds to be used as promising antitumor agents. ________________________________________ The similarities of these compounds with other intercalating agents lay on the tricyclic moiety and the aminated side chain (6). However, their mutagenicity (which was associated with the presence of the methylene moiety directly linked to C-4 (7)) proscribed their use in cancer chemotherapy. SR 233377 and SSR271425, two more recent aminated thioxanthones, entered phase I clinical trials (8,9) but were found to cause cardiotoxicity. Therefore, it is of considerable importance to further investigate aminated thioxanthonic derivatives, in the search for more active and less toxic antitumor compounds.
We have previously shown that aminated thioxanthones may be dual-acting agents capable of simultaneously inhibiting cell growth and P-gp function (10). These compounds resulted from the hybridization between a thioxanthonic scaffold (described for its antitumor activity) (9,11) and pharmacophoric features for P-gp inhibition (an amine group) (12). Overall, results with a leukemia cell line (P-gp overexpressing) resistant to doxorubicin, showed that aminated thioxanthones (1-8, Figure 1) can be characterized as a new class of P-gp inhibitors with improved efficacy when compared to verapamil, a known Pgp inhibitor (10).
Indeed, an issue that affects all generations of Pgp inhibitors concerns their multiple actions towards various ABC transporters. Even the most recent and potent P-gp inhibitors such as tariquidar (14) (P-gp and BCRP inhibitor) and CBT-1 (15) (P-gp and MRP1 inhibitor) modulate more than one ABC transporter, and both have entered clinical trials. It is not surprising to find compounds which are multi-target inhibitors of ABC transporters, since these proteins share structural and functional similarity (16). Indeed, since the clinical MDR appears to be related to the overexpression of several transporters, compounds that inhibit multiple ABC transporters may possibly be preferable to the use of a combination of specific modulators, as they would reduce cumulative toxicities and drug-drug interaction (17)(18)(19).
One of the main issues regarding second generation P-gp inhibitors was the pharmacokinetic interactions due to overlapping substrate specificity between P-gp and cytochrome P450, which required an increase in the dose of the cytotoxic drug in order to achieve therapeutic efficacy (20).
The novelty of this work resides in the investigation of several potential targets of newly synthesized thioxanthonic derivatives. The aim was to better characterize the P-gp inhibitory activity of thioxanthones and to further explore the potential of interaction with other ABC transporters and CYP3A4, as the selectivity profile of these compounds may help to clarify their possible clinical usefulness. Finally, a computational prediction of the docking conformations of thioxanthones on CYP3A4 as well as the most probable metabolism sites was also performed.

MDR1 UIC2 shift assay
The MDR1 UIC2 shift assay was performed using P-gp overexpressing leukemia cells, K562Dox, as previously described (21) with minor modifications. K562Dox is a cell line derived from K562 by doxorubicin stimulated overexpression of P-gp (kindly donated by Prof. J.P. Marie, Paris, France). P-gp expression on the K562Dox cell line (by Western blot) has previously been confirmed (22). K562Dox cells were washed with phosphate buffered saline (PBS) and re-suspended in UIC2 binding buffer (PBS + 1% BSA). Approximately 1 × 10 6 cells in 800 µL binding buffer were pre-warmed at 37ºC for 10 min, incubated with the compounds 1-8 (final concentration 200 μM) or Na 3 VO 4 (final concentration of 20 μM; positive control, noncompetitive inhibitor), verapamil, quinidine, and mibefradil (final concentration of 200 μM; negative control, competitive inhibitors) at 37ºC for 30 min, and the monoclonal antibody UIC2 was added (final concentration 12.5 μg.mL -1 ). After 30 min at 37ºC, 800 µL of ice-cold UIC2 buffer was added to stop the reaction. Cell samples were washed with cold buffer twice, resuspended in 500 µL ice-cold UIC2 buffer and goat anti-mouse IgG2a-phycoerithrin (PE) was added (final concentration of 5 μg.mL -1 ). After 45 min at 4ºC in the dark, samples were washed twice using UIC2 ice-cold buffer and finally resuspended in 250 µL ice-cold UIC2 binding buffer containing propidium iodide (final concentration of 40 µg.mL -1 ). The cellular suspension was analyzed using a flow cytometer (Epics XL-MCL, Beckman Coulter, USA) and FL2 and FL3 channels were used for indirect UIC2 staining and dead cell exclusion by propidium iodide, respectively. At least 5000-10000 events were collected. Fluorescence intensity associated with the cells was expressed on a log scale. Normal mouse immunoglobulin G2a (IgG2a) served as a negative control. Results are presented as the average of at least three independent experiments. Fluorescence intensity ratio (FL2 ratio ) was calculated in relation to the DMSO solvent control (cells treated with DMSO and medium only, and incubated with primary and secondary antibodies as described). It was calculated using the following formula that fits the blank control to zero for easier interpretation: FI2 ratio = (FI test -FI dmso ).FI dmso -1 .

ATPase activity assays
ATPase activity was measured according to the described procedure (23). The predeasy MDR1-ATPase kit, BCRP-HAM-ATPase kit, MRP1, MRP2 and MRP3 ATPase kits were used for determining the effect of compounds 1 and 2 on the MDR1-, BCRP-, MRP1-, MRP2-and MRP3-ATPase activities respectively, according to the manufacturer's instructions. In brief, membrane vesicles in TMEP solution (50 mM Tris, 50 mM mannitol, 2 mM ethylene glycol tetraacetic acid (EGTA), 8 μg.mL -1 aprotinin, 10 μg.mL -1 leupeptin, 50 μg.mL -1 phenylmethanesulfonylfluoride (PMSF), 2 mM dithiothreitol (DTT), pH 7.0) were incubated at a concentration of 20 μg per well with an assay mix (40 mM 3-morpholinopropanesulfonic acid with (MRP assay) or without glutathione (the remaining assays)), 5 mM ATP and four concentrations of test compounds 1 and 2 (1.6, 8.0, 40.0, 200.0 μM). Two protocols were used: (a) incubation with or without 1.2 mM Na 3 VO 4 (activation study) or (b) incubation with 200.0 μM of a known substrate (verapamil for MDR1 assay, N-ethylmaleimide glutathione (NEM-GS) for MRP1 assay, sulfasalazine for MRP2 and BCRP assay, and benzbromarone for MRP3 assay) and with or without 1.2 mM sodium orthovanadate (inhibition study), for 10 min at 37°C. Controls without treatment and with a control inhibitor at 40.0 μM (cyclosporine A for MDR1 and MRP3 assay, benzbromarone for MRP1 and MRP2, and Hoechst 33342 for BCRP assay) were also used. The reaction was stopped by adding a developer solution (ethanol 95%) to each well, followed by a blocker solution (4 mM sodium azide) and an additional 30 min incubation at 37ºC. Optical density (OD) was read at 630nm. The amount of phosphate (Pi) liberated by the transporter is proportional to the activity of the transporter (24). Transporter activity was measured as the vanadate sensitive portion of total ATPase activity required to eliminate background ATPase activity (25). Liberated Pi was quantified based on the calibration line established with Pi standards (0, 4, 8 nmol per well). Results are presented as % of control (% of the nontreated control). The IC 50 concentrations were calculated from the inhibition studies. Similar assays were also performed using the predeasy defMDR1, defBCRPHAM and defMRP ATPase control kits, using the manufacturer protocols (a).
The IC 50 value (concentration that inhibits the metabolism of the known substrate luciferin-BE by 50%) of compounds 1 and 2 was determined with the same protocol but testing eight concentrations of 1 and 2 ranging from 2 nM to 200 μM (0.002, 0.013, 0.064, 0.32, 1.60, 8.00, 20.00, and 200.00 μM). The IC 50 value was calculated after scaling the data to 100% activity.

Modelling the metabolism of compounds 1 and 2
All calculations were performed on a commodity PC Microsoft Windows 7 with an Intel(R) Core(TM) i5 processor. Compounds 1-8 were drawn and minimized using the AM1 semiempirical method (Hyperchem v8.07, Hypercube, USA) (26).
Compounds 1 and 2 were submitted to the MetaSite v3.0.4 program (MolDiscovery, Middlesex, United Kingdom) (27) which is a fully automated procedure that considers structural complementarity between the enzyme active site and ligand and provides the best orientation. Both the protein active site and ligand are represented by selected distance-based descriptors using molecular interaction fields computed in GRID. Metabolism site prediction was performed for CYP3A4. The most probable sites of metabolism (SOM) or "hot spots" were plotted according to their scores. The metabolism site is described by a probability index which is a product of similarity between ligand and protein. In this study, the top six averaged rankings with the reactivity component were considered. SOM was used to calculate those atoms that contribute most to orienting the metabolism site towards the heme group.

STATISTICAL ANALYSIS
Data was expressed as the mean ± SE of at least three independent experiments, except in the case of vanadate-sensitive ATPase activity (mean of two experiments only) and analyzed by the Student's t test. P-values below 0.05 were considered statistically significant when data was analysed for statistical meaningfulness (student's t test).

MDR1 UIC2 shift assay
In an effort to confirm the previously-found direct interaction of thioxanthones 1-8 with P-gp (10), a MDR1 UIC2 shift assay was performed ( Figure  2). Therefore, two competitive inhibitors (2 and 3) and six noncompetitive inhibitors (1, 4-8) were selected from our previous work to perform the MDR1 shift assay. This assay uses the monoclonal antibody (mAb) UIC2, a conformation-sensitive antibody which preferentially binds to an external epitope of P-gp which is in the process of substrate transportation (31,32). UIC2 is also unique among the mAbs directed against P-gp because it blocks drug pumping efficiently, probably by trapping P-gp in a certain conformational state of its enzymatic cycle (32). Based on this conformation-sensitive competition assay, it is possible to distinguish between compounds that are being transported (substrates or competitive inhibitors) from compounds that block the P-gp transportation cycle (noncompetitive inhibitors) (33). Figure 2 shows that mibefradil, verapamil and quinidine (all P-gp substrates used as controls) increased UIC2 labelling of K562Dox cells, as expected. The blocking of the ATP site by sodium vanadate decreased UIC2 labelling, also as expected. Compounds 1, 4, 6-8 had UIC2 binding reactivity inferior to the DMSO control while compound 2 had a UIC2 binding reactivity greater than that of the DMSO control ( Figure 2).

ATPase assay
As our previous data showed that compounds 1 and 2 were the most potent P-gp noncompetitive and competitive inhibitors, respectively (10), they were further investigated with ATPase screening assays for various ABC transporters (including Pgp), which are based on purified membrane vesicles from mammalian cells overexpressing a selected human ABC transporter protein. The transport function of ABC transporters is tightly coupled to its ATPase activity, as ATP hydrolysis drives the transport process (34). Substrates transported by P-gp can therefore be detected by hydrolysis of ATP and production of the inorganic phosphate detected by a colorimetric reaction (35). The assay is composed of two different tests: an activation and inhibition test, both of which are complementary assays. The assay is a modification of the method of Sarkadi et al (36). In the activation test, the transported substrates increase baseline vanadate sensitive ATPase activity while inhibitors decrease the baseline vanadate sensitive ATPase activity. A stimulation detected on the activation assay indicates that the compound is a substrate of the transporter being studied. In the inhibition test, carried out in the presence of a known substrate/activator (which varies according to the ABC transporter under study), inhibitors or transported compounds inhibit the maximum vanadate sensitive ATPase activity (37). Interactions detected in the inhibition test indicate interactions of the test compounds with the transporter. The obtained results are graphically represented in Figures 3 and 4 and the IC 50 values are indicated in Table 1. Compounds 1 and 2 did not affect levels of Pi liberation on membranes lacking P-gp, BCRP, or MRP (results not shown).

CYP3A4 interaction assay
Luciferin-BE is used as a CYP3A4 substrate which is converted by the enzyme to luciferin product, which in turn reacts with luciferase to produce an intensity of light proportional to CYP activity (38). Changes in the luminescence signals from samples treated with test compound compared to untreated control samples, represent modulation of CYP3A4 activity by the test compound (39).

Figure 2.
Effect of eight thioxanthonic derivatives on mAb UIC2 binding reactivity to P-gp, presented as a FL2-ratio. Normal IgG2a was used as negative control. Control luminescence (DMSO) is represented as zero for easier interpretation. An FL2-ratio less than zero corresponds to inhibitors of the P-gp transportation cycle, i.e., noncompetitive P-gp inhibitors. An FL2-ratio greater than zero corresponds to compounds that stimulate the P-gp transportation cycle i.e., substrates of P-gp or competitive P-gp inhibitors. Results are expressed as the mean ± SE of at least three independent experiments. Na 3 VO 4 , verapamil, mibefradil and quinidine were used as controls. Statistical significance was tested by the paired t-test using DMSO as control. *** indicates P < 0.001; ** indicates 0.001 < P ≤ 0.01; * indicates 0.01 < P ≤ 0.05. ns indicates not significant, i.e., P > 0.05 (n=4).   Figure 5 shows that the eight thioxanthonic derivatives (1-8) interfere with CYP3A4 activity (inhibitors or substrates) as they decreased the levels of luciferin-associated luminescence when tested at concentrations of 200 µM.
To calculate the CYP3A4 IC 50 of our best Pgp inhibitors, the same luciferin luminescencebased assay was used to test several concentrations of compounds 1 and 2 ( Figure 6). It was found that these compounds presented IC 50 values for CYP3A4 inhibition of approximately 1-2 μM (Figure 6). The CYP3A4 inhibition is low for compounds 1 and 2 at concentrations below 0.01 μM, but is almost total for concentrations close to 100 μM.

Modelling the metabolism of compounds 1 and 2
In view of these results, in silico analysis of the potential metabolism sites on the thioxanthonic derivatives 1 and 2 was performed. The metabolism site and molecular contribution to the exposure of a reactive atom to the heme moiety (activity contribution plot) were determined by MetaSite (40). The program has been described as being capable of predicting the most probable metabolism site within the top three predictions in 80% of the cases in structurally diverse compounds (27). MetaSite predicted that the methylene groups directly linked to the aniline at C-1 in compounds 1 and 2 were the top-ranked metabolism sites by CYP3A4 (Figure 7, dark red arrows). Other possible metabolism sites are also depicted in Figure 7 (light red arrows). To further predict the binding affinity of compounds 1 and 2 to the CYP3A4 catalytic site, a docking study using Autodock Vina was performed on those compounds, ketoconazole, a known CYP3A4 inhibitor (41) and propofol, which does not interact with CYP3A4 (29). The structure of the CYP3A4-ketoconazole is available at 3.8 Å resolution (42,43). As expected, ketoconazole registered a high affinity to the CYP3A4 catalytic site, as predicted by the highly negative docking score (Table 2). Besides, the docked ketoconazole and crystallographic ketoconazole are almost perfectly superimposed, with a RMSD of 0.14 Å. The negative control, propofol, as expected, had the worst docking score. Compounds 1-8 had docking scores ranging from -7.4 to -10.2 kJ.mol -1 , values similar to the ketoconazole control (Table 2). A B Figure 7. Metabolism sites and contribution: the groups in compound 1 and 2 that most likely will be metabolized by CYP3A4 are marked with dark red arrows; the darker the colour the greater the probability of metabolism occurring. The atoms in compounds 1 and 2 with the highest molecular contribution to the exposure of the metabolic "hot spot" (blue circle) to the heme moiety of CYP3A4 are marked with yellow arrows. inhibitor of CYP3A4 not metabolized by CYP3A4 *binding conformations represented in Figure 8

DISCUSSION
Although thioxanthones have been extensively studied for their P-gp modulation activity (10), there is a lack of more detailed information regarding their interaction with other ABC transporters. This study represents the first investigation of the potential of thioxanthonic derivatives to interact with MRP1, MRP2, MRP3, BCRP and CYP3A4.

Thioxanthones are competitive or noncompetitive P-gp inhibitors
Concerning the UIC2 labelling assay, compounds 1, 4, 6-8 significantly decreased UIC2 reactivity, a result similar to that produced by orthovanadate, behaving as non-competitive inhibitors. On the other hand, compound 2 significantly increased UIC2 reactivity in the same order of magnitude as verapamil and quinidine, behaving as competitive inhibitors. These results are in accordance with those obtained by the luminescence-based ATPase assay(10), thereby confirming our previous data.

Thioxanthones are multi-ABC transporter inhibitors
Results regarding the effect of compound 1 on Pgp ATPase activity confirmed that compound 1 is not transported by P-gp ( Figure 3A, activation study) as it does not interfere with the basal P-gp ATPase activity. Additionally, compound 1 was found to inhibit verapamil transport ( Figure 3A, inhibition study) in a concentration-dependent manner up to 40 µM, before reaching a "plateau" thereafter. The IC 50 value for inhibition of P-gp by compound 1 was close to 20 μM (Table 1).
Regarding the BCRP ATPase assay, compound 1 was shown to not be a substrate for BCRP ( Figure  3B, activation assay) as the levels of Pi detected were close to the baseline. Also, compound 1 inhibited the efflux of sulfasalazine ( Figure 3B, inhibition assay) in a concentration-dependent manner, with an IC 50 value of 20 μM. In addition, compound 1 was shown to be transported by MRP1 at low concentrations but at increasing concentrations (> 10 μM), the levels of transport decreased ( Figure 3C, activation assay). Compound 1 also inhibited the transport of a known MRP1 substrate, NEM-GS, particularly at concentrations above 8 μM and in a concentration-dependant manner ( Figure 3C, inhibition assay), presenting an IC 50 of 25 μM. Compound 1 was a weak substrate ( Figure 3D, activation assay) and weak inhibitor of sulfasalazine transport by MRP2 ( Figure 3D, inhibition study). Compound 1 was revealed as a weak MPR-3 substrate at concentrations below 40 μM ( Figure 3E, activation study). Compound 1 inhibited benzbromarone transport by MRP3 which only becomes more evident at concentrations above 40 μM ( Figure 3E, inhibition assay), with the IC 50 being reached only at 200 μM. Hence, compound 1 is not a substrate (or is a very low-affinity substrate) for P-gp, BCRP and MRP2 but is a substrate for MRP3 and mainly for MRP1 (as shown by values above zero in the activation studies of Figures 3C and 3E).
Besides, the decreasing curves in the inhibition studies (values below 100 %) revealed that compound 1 interferes with the transport of these known ABC transporter substrates (Figure 3, inhibition studies). The results from the effect of compound 2 on P-gp ATPase activity confirmed that compound 2 is transported by P-gp at concentrations above 10 μM ( Figure 4A, activation study) and revealed that compound 2 inhibits verapamil transport at concentrations above 8 μM, in a concentrationdependent manner ( Figure 4A, inhibition study). The IC 50 value for the inhibition of P-gp by compounds 2 is close to 20 μM (Table 1). Regarding the BCRP ATPase assay, compound 2 was shown to not be a substrate for BCRP ( Figure  4B, activation assay) as the levels of Pi detected were close to the baseline. In addition, compound 2 inhibited the efflux of sulfasalazine ( Figure 4B, inhibition assay) in a concentration dependent manner up to concentrations of approximately 10 μM and stabilizing thereafter,, revealing an IC 50 value of 2 μM. Compound 2 was actively transported by MRP1 at low concentrations (~1 μM) but the levels of transport decreased with increasing concentrations of this compound ( Figure 4C, activation study). Indeed, data showed that compound 2 stimulated ATPase activity at low concentrations (<50 μM) but at higher concentrations (>50 μM), ATPase stimulation was found to be decreased. Compound 2 strongly inhibited the transport of NEM-GS ( Figure 4C, inhibition study) with an IC 50 of 4 μM. In addition, compound 2 was a weak MRP2 substrate ( Figure 4D, activation assay) and decreased the level of Pi release to values lower than the baseline, suggesting that it blocks MRP2 basal ATPase activity at concentrations above 100 μM. Compound 2 inhibited the transport of sulfasalazine by MRP2 ( Figure 4D, inhibition study), reaching the IC 50 at 145 μM. Finally, compound 2 was transported by MRP3 but ATPase activity decreases in a concentration-dependent manner until complete loss of ATPase function becomes evident at concentrations above 100 μM ( Figure 4E, activation study). Compound 2 inhibited MRP3 mediated transport of benzbromarone in a concentration-dependent manner ( Figure 4E, inhibition study) with an IC 50 value of 13 μM.
Based on the ATPase results as a whole, the IC 50 of compounds 1 and 2 to inhibit P-gp is approximately 20 μM (for a P-gp specific substrate). However, compound 1 is also a potent BCPR and MRP1 inhibitor with IC 50 values of 20 and 25 μM, respectively. Compound 2 is a strong inhibitor of BCRP, MRP1 and MRP3, with IC 50 values of 2, 4 and 13 μM, respectively.
ABC transporters are co-localized in protective tissues throughout the human body with important roles in restricting the intestinal absorption of foreign compounds and mediating the secretion of drugs and metabolites to bile (44). As a consequence, inhibition of P-gp, BCRP, and MRP1, MRP2 and MRP3 can lead to serious adverse effects, for example through the accumulation of toxic xenobiotics (45). Nonetheless, MRP1 and MRP3 are localized in basolateral membranes. P-gp, BCRP and MRP2 are localized in apical membranes. Thus, the efflux direction will be different for those pumps. For example, in an enterocyte, P-gp, BCRP, and MRP2 extrude substrates to the intestinal lumen while MRP1 and -3 pump their substrates to the systemic circulation (46,47). Besides, even if these transporters are partially blocked, other intestinal transporters such as peptide transporter 1 (PEPT1), organic anion transporters (OAT), organic cation transporters (OCT), as well as other members of the ABC family, are available for transport across the membrane (48). Therefore, as far as MDR reversal is concerned, these thioxanthonic derivatives could possibly represent further study candidates for treating cancer cells that express the ABC transporters (49). However, this could only be done by careful analysis of the phenotype of the cancer cells of each patient, using for example PET radiotracers (50).

Thioxanthones interact in vitro with CYP3A4
As the overlapping substrate specificity between P-gp and CYP3A4 is a common concern, namely concerning hydrophobic chemotherapeutic agents (51), the thioxanthonic derivatives 1-8 were screened for CYP3A4 interaction ( Figure 5). CYP3A is the most prominent CYP in humans, comprising about 30 and 70% of total CYP in the liver and intestine respectively (52) and is responsible for phase I metabolism of 50% of drugs administered to humans (52). CYP3A has a very broad substrate specificity and may deactivate several anticancer drugs within the tumor cells (53).
The initial screen for CYP3A4 activity modulation demonstrated that all of the thioxanthones inhibit the metabolism of BEluciferin by CYP3A4, reducing the production of a luciferin product that generates light when compared to the untreated control ( Figure 5). Thioxanthones 1-8 inhibit CYP3A4 in the same order of magnitude as a known CYP3A4 inhibitor, ketoconazole (28) at the tested concentration (200 μM). Compound 1 was the compound with lower CYP3A4 inhibition ability, followed by compounds 2 and 3. Although thioxanthones are able to inhibit several ABC transporters, their further development as clinical MDR modulators could fail due to their doselimiting effects (54,55). Indeed, at 1-2 μM concentration ( Figure 6), compounds 1 and 2 exhibit strong potency for binding CYP3A4, and activities comparable to ketoconazole in binding that enzyme (56). Thus, the competition between cytotoxic agents and these P-gp modulators for cytochrome P450 3A4 activity requires monitoring of pharmacokinetic interactions (57).
It has been previously reported that (thio)xanthonic derivatives can be substrates or inhibitors of multiple P450 isoforms (58). Therefore, the interference of compounds 1 and 2 with CYP3A4 determined experimentally using the P450-Glo kit, may result from a blockage of the CYP3A4 catalytic site or metabolism. However, this distinction would require further experiments, such as the analysis of the metabolite profile on incubation with CYP3A4 or with human liver microsomes (several CYPs) (59).
The metabolism prediction software MetaSite predicted that aliphatic carbon atoms directly linked to the nitrogen or two bonds away, the aniline, and the sulphur from the thioxanthonic ring as the most probable sites of metabolism ( Figure 7). Other less relevant metabolism sites would be the carbons from the propoxy side chain (Figure 7). It may be hypothesized that the aniline will probably suffer an anilinic hydroxylation or a N-dealkylation and the aliphatic carbons, an aliphatic hydroxylation. The sulphur will probably be oxidized to a sulfoxide or sulfone (60).
Concerning the docking results on CYP3A4, the ketoconazole molecule has its imidazole nitrogen bound to the heme iron group ( Figure  8A). The ketoconazole keto group is located in a polar pocket lined by the side chains of Arg372, Arg106 and Glu374. It has also been described that the side chain of Ser119 is important for the establishment of hydrogen bonds with some binding conformations of ketoconazole (43). Both compounds 1 and 2 are predicted to be bound in the same binding pocket as ketoconazole and with similar docking scores ( Table 2). As ketoconazole, compound 1 also establishes hydrogen bounds with residue Ser119 ( Figure  8B), and residue Ile301. These interactions put the "hot spot", the aliphatic carbon of the alkylarylamine, 3.0 Å from the ferryl group, in accordance with the predicted site of metabolism for compounds 1 (Figure 7A and 8B). In an effort to arrive at a more metabolically stable compound, one could for example, modify the carbonyl group from the thioxanthone. On the other hand, compound 2 has its aniline group facing the heme group of CYP3A4 ( Figure 8C). This is in accordance with the prediction which pointed to tetrahydroisoquinolinamine as the main target of metabolism ( Figure 7B). Again, the carbon from thioxanthonic carbonyl on compound 2 binds non-covalently to Ser119.
Although the use of computational methods as tools for predicting the likelihood of interference with CYP3A4 are inconclusive, the study demonstrated that they could be useful in both predicting, explaining the observed data obtained from the in vitro experiments, and possibly guiding further molecular modifications.

CONCLUSION
Thioxanthone 1 was found to be a multi-target inhibitor of P-gp, BCRP and MRP1. Thioxanthone 2 is a multi-target inhibitor of P-gp, BCRP, MRP1 and MRP3. Therefore, they could be further tested as potential MDR reversal agents in tumor cell lines that overexpress those transporters. Further in vivo work should be performed to assess the toxicity of these thioxanthones. Special attention should be taken in order to avoid pharmacokinetic interactions between co-administered CYP3A4-metabolizable drugs, which could lead to increased accumulation and toxicity. Therefore, the future direction of this work will focus on the pharmacokinetic interaction between thioxanthones and anticancer agents.