Thioxanthone-Mediated Cytoprotection Against Cisplatin Toxicity: Exploring the Potential Involvement of P-Glycoprotein Through Computational and Experimental Approaches

Abstract

P-glycoprotein (P-gp), an efflux transporter highly expressed in renal tubules, plays a crucial role in the detoxification and protection of barrier/excretory tissues from harmful xenobiotics. Xanthones and thioxanthones (TXs) are known for their antimicrobial and antitumor activities and for their ability to modulate membrane transporters such as P-gp. Previous studies have reported that (thio)xanthonic derivatives enhance P-gp expression and/or activity in intestinal cells, reducing the intracellular accumulation of toxic substrates; however, their capacity to modulate P-gp in renal cells remains poorly explored. This study aimed to predict, in silico, TXs’ binding sites within P-gp and to evaluate, in vitro, in human kidney (HK)-2 cells, the effects of selected TXs (TX1–5) on P-gp activity and expression, and protection against cisplatin-induced cytotoxicity. Computational studies identified preferential TX1–5 binding to the drug-binding pocket, particularly the rhodamine 123 (R) or modulator (M) sites, and to nucleotide-binding domain 1. In vitro, rhodamine 123 accumulation assays revealed increased P-gp transport activity after 120 min or 24 h exposure to TX1–5, except TX4. TX2 elicited the strongest effect (141% increase, p < 0.0001), upregulated P-gp expression (24 h, p < 0.0001), and significantly protected HK-2 cells from cisplatin-induced cytotoxicity (increased IC₅₀, p < 0.0001). Altogether, these findings position thioxanthones as promising scaffolds for the development of P-gp-targeted strategies to mitigate drug-induced nephrotoxicity.

1. Introduction

P-glycoprotein (P-gp), an efflux transporter belonging to the adenosine triphosphate (ATP)-binding cassette (ABC) superfamily, was discovered and isolated from hamster ovary cells by Juliano and Ling in 1976 [1]. P-gp, derived from the term “permeability-glycoprotein” and alternatively referred to as multidrug resistance 1 (MDR1), is encoded by the MDR1/ABCB1 gene in humans [2,3,4,5]. P-gp has been extensively studied not only for its crucial role in detoxifying and protecting major barrier tissues/organs (intestine, liver, kidney and blood–brain barrier) from toxic compounds, but also for its involvement in multidrug resistance (MDR) phenomena in cancer cells [4,5,6,7]. In fact, P-gp can transport a broad range of structurally unrelated compounds that differ in size, structure, and pharmacological class. Its substrates range from small molecules of approximately 330 Da up to 4000 Da, reflecting its non-specific transport capacity [8,9,10]. Examples of P-gp substrates include anticancer agents, cardiac glycosides (e.g., digoxin), β-adrenoreceptor antagonists, Ca²⁺ channel blockers, HIV protease inhibitors, steroids, immunosuppressants, antiemetic drugs, antibiotics, antimicrobials, antiretrovirals, and histamine H1-receptor antagonists [11,12,13,14,15].

Early studies predominantly focused on P-gp inhibition as a strategy to overcome MDR in cancer by preventing the efflux of anticancer agents from tumor cells, thereby increasing/prolonging their therapeutic effect [6,16,17,18]. On the other hand, P-gp induction/activation has been recently proposed to prevent toxicity caused by certain harmful P-gp substrates. Noteworthy, while P-gp inducers enhance P-gp overall activity by upregulating protein expression, P-gp activators directly stimulate efflux pump activity without increasing protein expression, resulting in a faster process. In fact, enhancing P-gp activity offers a promising antidotal strategy in intoxication scenarios by facilitating the efficient efflux of harmful xenobiotics from cells, thereby reducing their intracellular concentrations and associated toxicity [6,19,20,21].

Cisplatin (cis-diamminedichloroplatinum II) was the first platinum-based antineoplastic agent discovered and remains widely used in the treatment of solid tumors [22,23,24,25]. However, a major therapeutic limitation of cisplatin remains its pronounced nephrotoxicity, emerging from its mechanism of toxicity, causing DNA damage, oxidative stress induction, mitochondrial dysfunction, endoplasmic reticulum stress, activation of apoptotic signaling pathways, and stimulation of inflammatory responses [25,26,27,28,29,30]. Consequently, several strategies have been investigated to mitigate cisplatin-induced renal damage, including promoting cisplatin clearance from renal cells by efflux pumps [31,32,33].

In this context of P-gp modulation through induction and activation, xanthones have emerged as prominent P-gp modulators extensively studied for their diverse pharmacological properties. Chemically, xanthone (9H-xanthen-9-one) is an aromatic oxygenated heterocyclic compound with a dibenzo-pirone scaffold and the molecular formula C₁₃H₈O₂. The wide variety of xanthonic derivatives, obtained either from natural sources (e.g., plants, lichens, fungi, and marine organisms) or through synthetic methods involving a wide variety of substitutions at various positions, exhibits a wide spectrum of pharmacological activities [34]. Closely related, thioxanthones (9H-thioxanthen-9-one) are synthetic isosteric analogues of xanthones, distinguished by an S-heterocyclic dibenzothiopyrone scaffold. These compounds also exhibit a broad spectrum of biological activities, including antibiotic properties, monoamine oxidase inhibition, antitumor properties (inhibiting cancer cell growth, modulating mechanisms of action, and, in some cases, sensitizing tumor cells to conventional chemotherapeutics) [35,36]. Furthermore, thioxanthones have also demonstrated a well-established ability to modulate membrane/barrier transporters, namely P-gp activation [37,38].

To understand the interaction of (thio)xanthonic derivatives, previously reported as P-gp activators, with P-gp-binding sites, it is notable that the currently accepted P-gp structural model consists of two homologous functional units, the N- and C-terminal halves, with pseudo-2-fold symmetry. Each half comprises one transmembrane domain (TMD), containing six transmembrane helices (TMHs) and one nucleotide-binding domain (NBD), where ATP binding and hydrolysis occur [10,39]. Likewise, P-gp is considered to be a full transporter because it contains two TMDs and two NBDs [9,38,40,41]. A linker, a small peptide, connects the N- and C-terminal halves [42]. On the other hand, each TMH is directly connected to its respective NBD by intracellular loops (TMH6—NBD1; TMH12—NBD2), and also interacts non-covalently through short intracellular coupling helices (ICHs). These ICHs, located between TMHs 2/3 (ICH1-NBD1), 4/5 (ICH2-NBD2), 8/9 (ICH3-NBD2) and 10/11 (ICH4-NBD1), are involved in signal transmission between the TMDs and NBDs, and have an important role in the maturation and folding of this efflux pump [8,10]. Additionally, a large cavity is formed by the TMHs of both N- and C-terminal P-gp halves, constituting a drug-binding pocket (DBP) where distinct molecules can interact and/or be transported [10,42]. Currently, three different drug-binding sites (DBSs) have been described in the literature. The first two, proposed in the late 1990s by Shapiro and Ling, were designated as the H-site and R-site due to their distinct drug specificities for Hoechst 33342 and rhodamine-123 (RHO123), respectively [43]. A third site, also located within the DBP, was later identified as the M-site (subsequently referred to as Modulators-site) [42].

Thus, in this study, we aimed to elucidate, in silico, the P-gp binding sites and characterize the interaction types of several (thio)xanthonic derivatives previously reported as P-gp activators. Furthermore, an in vitro analysis was conducted in renal proximal tubular epithelial cells to assess P-gp transport activity, focusing on P-gp activation and/or induction by five selected thioxanthones (TX1–5). The protective effects of these compounds against cisplatin-induced cytotoxicity were also evaluated. This integrative approach thus aimed to contribute to understanding the molecular mechanisms of P-gp modulation by thioxanthones, highlighting their therapeutic potential in mitigating drug-induced nephrotoxicity.

2. Materials and Methods

2.1. In Silico Analysis

Ligands were initially designed in two dimensions (2D) using ChemDraw^® software (version 16.0.1.44). Subsequently, their three-dimensional (3D) minimum energy conformations were generated in Molecular Operating Environment (MOE) software (version 2022.02) using the SMILES (Simplified Molecular Input Line Entry System) obtained in ChemDraw and conducting energy minimization using default parameters. Table 1 lists the (thio)xanthonic derivatives previously described as P-gp activators at the intestinal level and selected as ligands for this study.

Table 1.

List of the studied (thio)xanthonic derivatives, including their International Union of Pure and Applied Chemistry (IUPAC) names, Simplified Molecular Input Line Entry System (SMILES) code, and chemical structure. All compounds selected for the present study were previously reported as P-gp activators at the intestinal level, in Caco-2 and/or SW480 cells (20 µM, for 45–60 min).

Compound (IUPAC Name; SMILE CODE)	Chemical Structure	Compound (IUPAC Name; SMILES CODE)	Chemical Structure
Thioxanthonic derivatives (TX1–TX5) [37]
TX1 1-[(3-hydroxypropyl)amino]-4-propoxy-9H-thioxanthen-9-one; O=C1C2=C(C(OCCC)=CC=C2NCCCO)SC3=CC=CC=C31		TX4 1-[(2-methylpropyl)amino]-4-propoxy-9H-thioxanthen-9-one; O=C1C(CNCC(C)C)=C(COCCC)SC2=CC=CC=C21
TX2 1-chloro-4-hydroxy-9H-thioxanthen-9-one; O=C1C2=C(C(O)=CC=C2Cl)SC3=CC=CC=C31		TX5 1-(propan-2-ylamino)-4-propoxy-9H-thioxanthen-9-one; O=C1C2=C(C(OCCC)=CC=C2NC(C)C)SC3=CC=CC=C31
TX3 1-{[2-(1,3-benzodioxol-5-yl)ethyl]amino}-4-propoxy-9H-thioxanthen-9-one; O=C1C2=C(C(OCCC)=CC=C2NCCC3=CC=C(OCO4)C4=C3)SC5=CC=CC=C51
Chiral aminated thioxanthones (ATX1–ATX8) [44]
ATX1 (S)-1-((1-Hydroxypropan-2-yl)amino)-4-propoxy-9H-thioxanthen-9-one; O=C1C2=C(C(OCCC)=CC=C2N[C@@H](C)CO)SC3=CC=CC=C31		ATX5 (S)-1-((1-Hydroxy-4-methylpentan-2-yl)amino)-4-propoxy-9H-thioxanthen-9-one; O=C1C2=C(C(OCCC)=CC=C2N[C@@H](CC(C)C)CO)SC3=CC=CC=C31
ATX2 (R)-1-((1-Hydroxypropan-2-yl)amino)-4-propoxy-9H-thioxanthen-9-one; O=C1C2=C(C(OCCC)=CC=C2N[C@H](C)CO)SC3=CC=CC=C31		ATX6 (R)-1-((1-Hydroxy-4-methylpentan-2-yl)amino)-4-propoxy-9H-thioxanthen-9-one; O=C1C2=C(C(OCCC)=CC=C2N[C@H](CC(C)C)CO)SC3=CC=CC=C31
ATX3 (S)-1-((2-Hydroxypropyl)amino)-4-propoxy-9H-thioxanthen-9-one; O=C1C2=C(C(OCCC)=CC=C2NC[C@H](C)O)SC3=CC=CC=C31		ATX7 (S)-1-((1-Hydroxy-3-methylbutan-2-yl)amino)-4-propoxy-9H-thioxanthen-9-one; O=C1C2=C(C(OCCC)=CC=C2N[C@@H](C(C)C)CO)SC3=CC=CC=C31
ATX4 (R)-1-((2-Hydroxypropyl)amino)-4-propoxy-9H-thioxanthen-9-one; O=C1C2=C(C(OCCC)=CC=C2NC[C@@H](C)O)SC3=CC=CC=C31		ATX8 (R)-1-((1-Hydroxy-3-methylbutan-2-yl)amino)-4-propoxy-9H-thioxanthen-9-one; O=C1C2=C(C(OCCC)=CC=C2N[C@H](C(C)C)CO)SC3=CC=CC=C31
Xanthones
Di-hydroxylated xanthones (X1–X5) [45]
X1 3,4-dihydroxy-9H-xanthen-9-one; O=C1C2=C(C(O)=C(O)C=C2)OC3=CC=CC=C31		X4 2,3-dihydroxy-9H-xanthen-9-one; O=C1C2=C(C=C(O)C(O)=C2)OC3=CC=CC=C31
X2 1,2-dihydroxy-9H-xanthen-9-one; O=C1C2=C(C=CC(O)=C2O)OC3=CC=CC=C31		X5 3,6-dihydroxy-9H-xanthen-9-one; O=C1C2=C(C=C(O)C=C2)OC3=CC(O)=CC=C31
X3 1,3-dihydroxy-9H-xanthen-9-one; O=C1C2=C(C=C(O)C=C2O)OC3=CC=CC=C31
Other oxygenated xanthones (OX1–OX6) [19,41]
OX1 3,4-dimethoxy-1-methyl-9H-xanthen-9-one; O=C1C2=C(C(OC)=C(OC)C=C2C)OC3=CC=CC=C31		OX5 3,4-dimethoxy-9-oxo-9H-xanthene-1-carbaldehyde; O=C1C2=C(C(OC)=C(OC)C=C2C=O)OC3=CC=CC=C31
OX2 1-(dibromomethyl)-3,4-dimethoxy-9H-xanthen-9-one; O=C1C2=C(C(OC)=C(OC)C=C2C(Br)Br)OC3=CC=CC=C31		OX6 1-(hydroxymethyl)-3,4-dimethoxy-9H-xanthen-9-one; O=C1C2=C(C(OC)=C(OC)C=C2CO)OC3=CC=CC=C31
OX4 4-hydroxy-3-methoxy-9-oxo-9H-xanthene-1-carbaldehyde; O=C1C2=C(C(O)=C(OC)C=C2C=O)OC3=CC=CC=C31

Simultaneously, a previously refined and complete human P-gp structure was used [10]. This homology modeled structure based on the murine inward-facing crystallographic structure conformation (PDB ID: 4Q9H), originally obtained by [46], ensured the usage of a current and most helpful human P-gp structural model. Molecular docking analyses were performed to estimate binding affinity energies and rank how the ligands bind to the P-gp model at three drug-binding sites (DBSs): the drug-binding pocket (DBP) and nucleotide-binding domains (NBDs) 1 and 2. These dockings employed Autodock Vina software and scoring function (version 1.2.3) [47], SMINA with Vinardo scoring function (12 February 2019; based on AutoDock Vina 1.1.2) [48], and GNINA (version 1.3.2) software [49], applying predefined docking boxes centered in each DBS location. For the NBDs, it also contains the intracellular helices (ICHs) in a box size of 28.5 Å in each axis, while for the DBP it contains the full pocket with a size of 32.25, 26.25 and 37.50 Å in each x, y and z direction, respectively (exhaustiveness increased to 50 for an increased sampling of this larger space). Stronger ligand-P-gp interactions correspond to lower (more negative) affinity energies (as given by the Vina and Vinardo scoring functions), whereas higher affinity energies indicate weaker binding [50]. In contrast, the convolutional neural network (CNN) Affinity (provided by GNINA) estimates the energy required to disrupt the complex.

Subsequently, the DBP and NBD1 were selected for more detailed investigation. Using MOE software, the top ranked ligand poses were visualized and superimposed onto the P-gp structural model for clear identification of the DBS and binding mode for each ligand. Specifically, within the DBP, three well-characterized DBSs described in the literature were specifically examined: the Hoechst 33342 H-site, the rhodamine 123 R-site, and the Modulators M-site [42]. Regarding NBD1, a comprehensive computational search was conducted to accurately identify ligand binding sites on the P-gp structural model. This approach enabled a comprehensive mapping of ligand interactions at both well-characterized and less-characterized binding surfaces.

For a more detailed examination of the specific interactions, BINding ANAlyzer (BINANA) software (version 2.0) [51] was employed to quantify and characterize the interactions occurring between the evaluated compounds and the specific residues of P-gp. BINANA software provided an exhaustive analysis detailing the interaction types, such as hydrogen bonds, halogen bonds, hydrophobic contacts, π-π stacking interactions, T-stacking (face-to-edge) interactions, cation-π interactions, and salt bridges, when applicable, with designated P-gp residues. This automated approach allows precise characterization of molecular contacts crucial to ligand binding and supports deeper insights into the structural basis of P-gp modulation by these compounds.

2.2. In Vitro Studies

2.2.1. Materials and Reagents

Reagents used in cell culture, including Roswell Park Memorial Institute (RPMI) 1640 medium, sodium bicarbonate, rhodamine 123 (RHO123), zosuquidar (ZOS), 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide (MTT), neutral red (NR) solution, and cisplatin were purchased from Sigma-Aldrich, Merck (Darmstadt, Germany). Triton™ X-100 detergent solution and dimethyl sulfoxide (DMSO) were acquired from Thermo Fisher Scientific (Waltham, MA, USA). Fetal bovine serum (FBS), antibiotic (10,000 U/mL penicillin, 10,000 µg/mL streptomycin), Hank’s balanced salt solution with or without calcium and magnesium [HBSS (+/+) or HBSS (−/−), respectively] and 0.25% trypsin/1 mM ethylenediamine tetraacetic acid (EDTA) were purchased from Pan Biotech (Aidenbach, Germany). P-gp mouse monoclonal antibody (UIC2) conjugated with phycoerythrin (PE) was purchased from Abcam (Cambridge, UK). All the reagents used were of analytical grade or of the highest grade available.

According to the results obtained in previous studies [37], and the present in silico results, five thioxanthones (TX1–5, Table 1) were selected for the following in vitro experiments. TX1–5 were synthetized as previously described [35,37], in which TX1, TX3, TX4, and TX5 were obtained via Ullmann cross-coupling between 1-chloro-4-propoxy-9H-thioxanthen-9-one and an amine in an alkaline medium under microwave heating. TX2 was synthesized through the dealkylation of 1-chloro-4-propoxy-9H-thioxanthen-9-one using boron bromide. TX1–5 were then characterized using spectroscopic methods, and their purity was confirmed by high-performance liquid chromatography with diode array detection (HPLC–DAD) analysis, with purity levels of at least 95%. Stock solutions of TX1–5 were prepared at 50.0 mM in DMSO, stored at −20 °C, and diluted in cell culture medium on the day of exposure. It is noteworthy that the final DMSO concentration in the exposure medium did not exceed 0.1%, reducing the risk of solvent-related cytotoxic effects.

2.2.2. Cell Culture

Human kidney (HK)-2 cells (ATCC^® CRL-2190), obtained from the American Type Culture Collection (ATCC; Manassas, VA, USA), were routinely cultured in 75 cm² flasks using RPMI 1640 medium supplemented with 2 g/L NaHCO₃, 10% FBS, and 1% antibiotic (100 U/mL penicillin and 100 µg/mL streptomycin). The cell culture medium was replaced every 2 to 3 days to maintain optimal cell growth. Cells were routinely maintained in a humidified atmosphere of 5% CO₂—95% air at 37 °C. Cell passages were performed when cultures reached confluence, using 0.25% trypsin/1 mM EDTA to maintain exponential growth and culture viability. Cells used in the experiments were between the 35th and 45th passages.

2.2.3. Cytotoxicity Assays

HK-2 cells were seeded at a density of 50,000 cells/cm² into 96-well plates and exposed, 24 h post-seeding, to the tested TX1–5 (0, 5, 10 and 20 μM), RHO123 (0, 5 and 10 μM) and ZOS (0, 2.5, 5 and 10 μM), in fresh cell culture medium. Concentrations of TX1–5 (5, 10 and 20 μM) were selected based on preliminary cytotoxicity assays and previous research [37], ensuring non-cytotoxic yet pharmacologically relevant ranges. Fresh cell culture medium (0 μM TX1–5, RHO123 and ZOS) was used as the negative control, whereas 1% Triton™ X-100 served as the positive control to confirm assay sensitivity. Subsequently, after 24 h of exposure, TX, RHO123 and ZOS cytotoxicity were evaluated by MTT reduction and NR uptake assays.

MTT Reduction Assay

The MTT reduction assay is a colorimetric method commonly used to measure mitochondrial dysfunction. It relies on the enzymatic reduction of the yellow water-soluble tetrazolium dye, MTT, to a purple, water-insoluble formazan product by mitochondrial dehydrogenases. In viable cells, mitochondrial dehydrogenases cleave the tetrazolium ring, reducing the dye to purple formazan crystals, which can be easily dissolved in an organic solvent and measured at 550 nm. Therefore, the amount of formazan formed is directly proportional to the number of metabolically active cells in culture [52]. Accordingly, after 24 h of exposure, the cell culture medium containing TX1–5, RHO123 or ZOS was removed and replaced with fresh cell culture medium containing 0.5 mg/mL MTT. After 60 min incubation in a humidified atmosphere of 5% CO₂ and 95% air at 37 °C, the cell culture medium was removed, and the formed formazan crystals were dissolved in DMSO. The absorbance was then measured at 550 nm in a multi-well plate reader (Biotek Synergy HT Microplate Reader, BioTek^® Instruments Inc., Winooski, VT, USA). The percentage of MTT reduction compared to the control cells (0 µM) was used to assess cytotoxicity. Six independent experiments were performed, each in triplicate.

NR Uptake Assay

The NR uptake assay relies on viable cells’ ability to incorporate and retain supravital NR within their lysosomes. Consequently, the amount of NR dye incorporated serves as a quantitative indicator of lysosomal function, thereby providing an estimate of the number of viable cells in a culture. The dye, extracted from the viable cells using an acidified ethanol solution, can then be measured in a spectrophotometer at 540 nm [53]. After the exposure period (24 h), the cell culture medium was removed and replaced by fresh cell culture medium containing 50 μg/mL NR. Following 90 min incubation in a humidified atmosphere of 5% CO₂–95% air at 37 °C, the cell culture medium was removed and the dye was extracted from the living cells using a solution of absolute ethyl alcohol/distilled water (1:1) containing 5% acetic acid. The absorbance of the extract dye was then measured at 540 nm using a multi-well plate reader. The percentage of NR uptake compared to the control cells (0 µM) was used to assess cytotoxicity. Six independent experiments were performed, each in triplicate.

2.2.4. Evaluation of P-gp Transport Activity

The alterations in P-gp transport activity caused by TX1–5 were evaluated using the RHO123 accumulation assay, where RHO123 serves as a well-known P-gp fluorescent substrate and ZOS is used as a specific third-generation P-gp inhibitor. This assay has been widely employed to directly evaluate P-gp activity by measuring RHO123 intracellular fluorescence in the presence of diverse (thio)xanthonic derivatives in several other cell lines, including Caco-2 [38,41,44,45], and SW480 [19]. Two different experimental protocols were performed: (i) an RHO123 accumulation assay in cells exposed to TX1–5 for 120 min; and (ii) an RHO123 accumulation assay in cells pre-exposed to TX1–5 for 24 h.

RHO123 accumulation assay in HK-2 cells exposed to TX1–5 for 120 min: For this protocol, HK-2 cells were seeded on 24-well plates at a density of 50,000 cells/cm², 24 h before the experiment. Twenty-four hours after seeding, the cell culture medium was aspirated, and HK-2 cells were then exposed to TX1–5 (0, 5, 10 and 20 µM), in the presence or absence of ZOS (5 µM), both prepared in HBSS (+/+). Thirty minutes later, RHO123 [5 µM, in HBSS (+/+)] was added, followed by incubation for 90 min in a humidified 5% CO₂—95% air atmosphere at 37 °C. After this incubation, the RHO123-containing medium was aspirated, and HK-2 cells were washed with HBSS (+/+) to remove extracellular RHO123. The cells were then lysed with Triton™ X-100 [1%, in HBSS (+/+), for 10 min in the dark and at room temperature] and RHO123 fluorescence was measured at excitation/emission wavelengths of 485/528 nm, using a multi-well plate reader. This protocol was designed to promptly detect the immediate effects of xenobiotics on P-gp transport activity, whether through direct activation or inhibition of the pump. Thus, in this setup, the measured intracellular fluorescence of RHO123 directly reflects P-gp activation (decreased intracellular RHO123 fluorescence) or inhibition (increased intracellular RHO123 fluorescence), excluding potential influences from changes in pump expression, which require longer incubation periods to manifest.

RHO123 accumulation assay in HK-2 cells pre-exposed to TX1–5 for 24 h: HK-2 cells were seeded in 24-well plates at a density of 50,000 cells/cm² and, 24 h later, exposed to TX1–5 (0, 5, 10 and 20 µM) for 24 h. Following incubation, the cell culture medium containing TXs was removed and replaced with HBSS (+/+) with(/out) ZOS (5 µM). Thirty minutes later, RHO123 [5 µM, in HBSS (+/+)] was added, followed by incubation for 90 min in a humidified 5% CO₂—95% air atmosphere at 37 °C. After this incubation period, the 24-well plates were aspirated, HK-2 cells were washed with HBSS (+/+), and cells were lysed with Triton™ X-100 [1%, in HBSS (+/+)] for 10 min in the dark and at room temperature. RHO123 fluorescence was then measured at excitation/emission wavelengths of 485/528 nm using a multi-well plate reader. Accordingly, evaluating intracellular RHO123 fluorescence after 24 h of exposure to TXs enables assessment of potential increases in P-gp activity resulting from alterations in protein expression levels.

Subsequently, the results obtained for both experimental designs were calculated through the ratio of the fluorescence intensity (FI) after RHO123-inhibited accumulation (IA, inhibited accumulation in the presence of the P-gp inhibitor ZOS) to the FI of RHO123 normal accumulation (NA, in the absence of the ZOS) (Equation (1)) and presented as the percentage of control cells (0 μM). Five independent experiments were performed, each in triplicate. When P-gp transport activity increases, there is a marked increase in the efflux of RHO123 from cells, leading to a corresponding decline in intracellular RHO123 levels and, consequently, a reduced FI. Therefore, a higher accumulation ratio of RHO123 is a direct consequence of reduced FI during NA conditions, driven by increased P-gp activity that exports the dye during the accumulation phase. Conversely, when P-gp activity diminishes, the FIIA/FINA ratio declines, reflecting a higher FINA due to reduced RHO123 efflux.

$$\mathrm{R}\mathrm{H}\mathrm{O}123 \mathrm{A}\mathrm{c}\mathrm{c}\mathrm{u}\mathrm{m}\mathrm{u}\mathrm{l}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}={\displaystyle \frac{\mathrm{F}\mathrm{I} \mathrm{o}\mathrm{f} \mathrm{R}\mathrm{H}\mathrm{O}123 \mathrm{i}\mathrm{n}\mathrm{h}\mathrm{i}\mathrm{b}\mathrm{i}\mathrm{t}\mathrm{e}\mathrm{d} \mathrm{a}\mathrm{c}\mathrm{c}\mathrm{u}\mathrm{m}\mathrm{u}\mathrm{l}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\left(\mathrm{I}\mathrm{A}\right)}{\mathrm{F}\mathrm{I} \mathrm{o}\mathrm{f} \mathrm{R}\mathrm{H}\mathrm{O}123 \mathrm{n}\mathrm{o}\mathrm{r}\mathrm{m}\mathrm{a}\mathrm{l} \mathrm{a}\mathrm{c}\mathrm{c}\mathrm{u}\mathrm{m}\mathrm{u}\mathrm{l}\mathrm{a}\mathrm{t}\mathrm{i}\mathrm{o}\mathrm{n}\left(\mathrm{N}\mathrm{A}\right)}}$$ (1)

2.2.5. Evaluation of P-gp Expression

The alterations in P-gp expression caused by TX1–5 were evaluated by flow cytometry using a P-gp monoclonal antibody (UIC2) conjugated with phycoerythrin (PE). HK-2 cells were seeded in 24-well plates (50,000 cells/cm²) and exposed the following day to TX1–5 (20 µM) in fresh cell culture medium for 24 h. After incubation, cells were washed with phosphate-buffered saline (PBS) buffer (pH 7.4), detached by trypsinization, and centrifuged (2000 rpm, for 3 min). Cells were then resuspended in PBS buffer containing 10% heat-inactivated FBS and the UIC2-PE antibody (according to the manufacturer’s instructions), and then incubated for 1 h at room temperature, protected from light, with gentle agitation. To account for any potential autofluorescence interference, unlabeled cells (both treated and untreated with TX1–5) were included in every experiment. After incubation, cells were washed with PBS, centrifuged (2000 rpm for 3 min), and kept on ice until analysis. Immediately before the cytometer analysis, cells were resuspended in ice-cold PBS buffer. The fluorescence of the UIC2-PE antibody in each sample was measured using a BD AccuriTM C6 flow cytometer (BD Biosciences, San Jose, CA, USA) equipped with a 585 ± 40 nm band-pass filter (FL2 detector). Logarithmic fluorescence values were recorded and presented as single-parameter histograms based on data from 30,000 cells per sample. The mean fluorescence intensity (MFI) was used to quantify P-gp expression, with values calculated and presented as a percentage of the control cells (0 µM). Seven independent experiments were performed in duplicate.

2.2.6. Evaluation of Thioxanthones’ Protective Effects Against Cisplatin-Induced Cytotoxicity

The potential protective effects of TX1–5 against cisplatin-induced cytotoxicity were evaluated using the MTT reduction assay (described in 2.2.3.1. MTT reduction assay) with HK-2 cells. Cells were seeded in 96-well plates (density of 50,000 cells/cm²) and, after 24 h, were exposed to increasing concentrations of CP (0, 1.25, 2.5, 5, 10, 25, 50, 100, 250, and 500 µM), in the presence or absence of TX1–5 (0, 5, 10, and 20 µM), for 24 h. Subsequently, the MTT reduction assay was performed to assess cellular viability, as previously described. Results were expressed as percentages of control cells (0 µM CP, 0 µM TX1–5). At least four independent experiments were performed, each in triplicate.

2.3. Statistical Analysis

All statistical analyses were performed using GraphPad Prism 8^® (GraphPad Software, San Diego, CA, USA). Normality of molecular docking results was first confirmed by the Anderson–Darling test, and results were compared using one-way ANOVA followed by Tukey’s multiple comparisons test. Experimental data were first assessed for normality using the Shapiro–Wilk test. For normally distributed data, one-way ANOVA followed by Dunnett’s or Tukey’s multiple comparison test was applied. Concentration–response curves were fitted using the least squares method as a fitting method, and curve parameters (Bottom, Top, LOG IC₅₀ and Hill Slope) were compared using the extra sum of squares F test. Detailed statistical analysis information is provided in each figure legend. Differences were significant for p values lower than 0.05.

3. Results

3.1. In Silico Results

The computational study began with molecular docking analysis to assess the affinity energy of several (thio)xanthonic derivatives, previously identified as P-gp activators at the enterocyte level, for interaction with a human P-gp structural model. The estimated binding affinities for these derivatives are presented in Table 2. Additionally, Figure 1 provides a graphical comparison of the three evaluated binding sites, highlighting an evident preference of P-gp activators for the DBP and NBD1 regions of the P-gp model.

Table 2.

Estimated affinity energy values (kcal/mol) or convolutional neural network (CNN) affinity values (expressed in pK units: e.g., 9 corresponds to 1 nM) obtained from the molecular docking analysis at the best binding pose of 23 (thio)xanthonic derivatives, including xanthones [X1–5, and other oxygenated xanthones (OX1–2 and 4–6)] and thioxanthones [TX1–5 and other chiral aminated thioxanthones (ATX1–8)], all known as P-gp activators at the enterocyte level. These interactions were evaluated against a human P-gp structural model at three distinct binding sites: the drug-binding pocket (DBP), nucleotide-binding domain 1 (NBD1) and nucleotide-binding domain 2 (NBD2). The final column specifies the drug-binding site (DBS) specifically within the DBP where each derivative achieved the best binding pose. Results were obtained using Autodock Vina, SMINA (using the Vinardo scoring function) and GNINA docking software. The best binding poses are highlighted in bold, corresponding to the lowest-energy poses (Autodock Vina and SMINA) or to the CNN affinity ranked by the highest CNN pose score.

Ligands	Autodock Vina (Affinity [kcal/mol])				SMINA (Affinity [kcal/mol])				GNINA (CNN Affinity [pK Units])
	NBD1	NBD2	DBP	DBS	NBD1	NBD2	DBP	DBS	NBD1	NBD2	DBP	DBS
ATX1	−7.4	−7.0	−8.0	M-site	−7.1	−6.8	−8.4	M-site	5.907	4.173	5.907	M-site
ATX2	−7.6	−7.2	−8.0	M-site	−6.7	−7.2	−8.3	M-site	5.520	4.399	7.289	M-site
ATX3	−7.5	−7.2	−8.1	M-site	−6.9	−7.0	−8.2	M-site	5.612	4.904	5.988	M-site
ATX4	−7.4	−7.2	−8.1	M-site	−6.6	−7.3	−8.0	M-site	5.425	4.936	7.609	M-site
ATX5	−7.8	−7.6	−7.9	M-site	−7.6	−6.9	−8.0	M-site	6.105	5.689	7.257	M-site
ATX6	−7.8	−7.5	−7.4	M-site	−6.8	−7.2	−8.2	M-site	6.171	5.741	7.515	M-site
ATX7	−7.9	−7.1	−8.0	M-site	−6.9	−6.9	−8.5	M-site	5.845	5.842	7.130	M-site
ATX8	−7.7	−7.5	−7.7	M-site	−6.7	−7.1	−8.1	M-site	5.971	5.473	7.791	M-site
OX1	−7.8	−7.1	−7.5	R-site	−5.5	−6.2	−7.7	R-site	4.111	4.281	4.830	R-site
OX2	−7.9	−7.1	−7.4	M-site	−5.6	−6.1	−6.8	M-site	4.742	4.236	5.193	R-site
OX4	−7.9	−7.1	−8.0	M-site	−5.8	−6.6	−7.7	R-site	4.282	4.227	5.989	M-site
OX5	−7.9	−7.1	−7.1	M-site	−5.8	−5.9	−6.8	M-site	3.884	4.133	6.139	M-site
OX6	−7.8	−6.9	−7.3	R-site	−5.8	−5.9	−7.6	R-site	4.127	4.097	6.028	M-site
TX1	−7.2	−6.8	−7.7	M-site	−6.9	−7.5	−8.0	M-site	5.562	4.178	5.984	M-site
TX2	−7.2	−6.7	−7.7	R-site	−5.6	−6.2	−7.7	R-site	5.034	4.778	5.321	R-site
TX3	−8.7	−7.8	−9.3	M-site	−8.2	−7.3	−10.9	M-site	5.954	5.313	7.375	M-site
TX4	−6.6	−6.4	−7.0	M-site	−6.6	−6.3	−7.7	M-site	4.750	4.396	6.855	M-site
TX5	−7.6	−6.9	−8.1	M-site	−6.2	−5.8	−8.1	M-site	5.209	4.843	7.219	M-site
X1	−7.6	−7.0	−7.7	M-site	−6.3	−6.6	−7.8	R-site	4.151	3.244	5.280	H/R-site
X2	−7.3	−7.3	−7.4	R-site	−6.4	−6.7	−7.4	R-site	4.077	4.032	4.019	M-site
X3	−7.5	−7.0	−7.3	R-site	−6.2	−6.6	−7.5	R-site	4.026	3.417	3.935	M-site
X4	−7.6	−7.1	−7.6	R-site	−6.3	−6.8	−7.6	R-site	4.307	4.056	3.671	M-site
X5	−7.3	−6.8	−7.3	R-site	−6.1	−6.7	−7.4	M-site	3.875	3.558	4.778	R-site

Figure 1.

Estimated affinity energy values (kcal/mol) or convolutional neural network (CNN) affinity values (pk units) for interactions between 23 (thio)xanthonic derivatives, including xanthones [X1–5, and other oxygenated xanthones (OX1–2 and 4–6)] and thioxanthones [TX1–5 and other chiral aminated thioxanthones (ATX1–8), known as P-glycoprotein (P-gp) activators, and the human P-gp structural model, evaluated at three distinct locations: the drug-binding pocket (DBP), nucleotide-binding domain 1 (NBD1) and nucleotide-binding domain 2 (NBD2). Affinity values were obtained using Autodock Vina (A), SMINA (B) and GNINA (C) software, and are presented as mean ± SD (standard deviation). Statistical comparisons were performed using one-way ANOVA followed by Tukey’s multiple comparisons test (* p < 0.05; **** p < 0.0001). Triangles (▲), squares (■), and spheres (●) represent individual affinity energy values of the tested (thio)xanthonic derivatives at the DBP, NBD1, and NBD2 binding sites, respectively.

Subsequently, the DBP and NBD1 were selected for further analysis, including the visualization of the binding sites of (thio)xanthonic derivatives within the P-gp model and a detailed evaluation of the ligand interactions with the P-gp residues. Within the DBP, as previously described, three distinct binding sites—H, R, and M—were examined. Table 2 summarizes the preferred binding locations (best poses) of each derivative, specifically within the DBP, as identified by Autodock Vina, SMINA or GNINA, and visualized using MOE software. Notably, within the DBP, most compounds have a preference to bind to the M-site, except for OX1, OX6, TX2, and X2–5, which exhibited preferences for the R-site. The spatial arrangement of these binding sites is illustrated in Figure 2. Although compounds were not always exclusively associated with either the M-site or R-site, those binding to the same designated binding site shared an identical spatial location. This pattern is evident in Figure 2, where ATX1 and TX5 represent the common binding location for all derivatives interacting with the M-site, while OX1 and X5 exemplify the binding position of the remaining derivatives that preferentially associate with the R-site.

Figure 2.

Representation of the preferred binding sites (location of the pose with lowest affinity energy) of (thio)xanthonic derivatives (light pink), within the drug-binding pocket (DBP): aminated thioxanthone 1 (ATX1) and thioxanthone 5 (TX5) on the M-site (yellow residues), and oxygenated xanthone 1 (OX1) and hydroxylated xanthone 5 (X5) on the R-site (green residues). Images were obtained with MOE software using the P-gp structural model from [10].

On the other hand, Figure 3 illustrates the binding locations of all studied derivatives interacting with NBD1. Although limited evidence exists detailing specific binding sites within the NBDs [54], a consistent binding pattern was observed among all analyzed P-gp activators. From the visual analysis, it was evident that all tested compounds bound to the same site located on the upper and intrinsic side of both NBDs, in proximity to the TMHs. This observation was further corroborated by the evaluation of specific P-gp binding residues, confirming a new shared binding hotspot for activators within the NBDs. This uniformity is exemplified by ATX1 and TX5, which serve as representative models for derivatives binding to NBD1. Subsequently, an evaluation of the number and types of interactions between the studied P-gp activators and specific P-gp residues was performed, encompassing hydrogen bonds, halogen bonds, hydrophobic contacts, π–π stacking interactions, T-stacking (face-to-edge) interactions, cation–π interactions, and salt bridges. Table S1 (Supplementary Material) lists the specific P-gp residues involved, along with their respective interaction types for each ligand included in the present study. The analysis of the interactions among P-gp activators and P-gp residues revealed a recurring interaction pattern, as several ligands shared similar hydrophobic contacts and other non-covalent interactions (hydrogen bonds, π–π and t-stacking, and cation–pi interactions) with specific P-gp residues.

Figure 3.

Representation of the preferred binding sites (location of the pose with lowest affinity energy) of aminated thioxanthone 1 (ATX1, dark pink) and thioxanthone 5 (TX5, light pink) on the NBD1. This binding location is common to all the (thio)xanthonic derivatives included in the present study. Images were obtained with MOE software, using the P-gp structural model from [10].

3.2. Cytotoxicity Evaluation

TX1–5 were selected for experimental testing because they previously demonstrated the strongest ability to boost P-gp activity in the intestine, particularly TX5 [37]. Furthermore, computational models predicted a valuable contrast within this group—showing that TX3 has the strongest binding affinity while TX4 has the weakest—making them ideal candidates to validate the range of possible interactions.

This study initially assessed the safety of the selected thioxanthones (TX1–5; 5, 10 and 20 μM), RHO123 (5 and 10 μM) and ZOS (2.5, 5 and 10 μM) using MTT reduction and NR uptake assays. The results showed slight discrepancies between the two assays. The MTT reduction test demonstrated that none of the compounds (TX1–5) exhibited significant cytotoxicity to HK-2 cells after 24 h of exposure (Figure 4). In contrast, the NR uptake assay revealed small though significant decreases in dye uptake to 94.5% and 88.3% of control values following 24 h exposure to 10 and 20 µM TX3, respectively, suggesting a slight, sub-toxic effect on lysosomal function that did not translate into mitochondrial impairment or loss of cell viability (Figure 5). For RHO123 and ZOS, although no significant cytotoxicity was detected in the NR uptake assay at any of the tested concentrations, significant cytotoxic effects were detected at 10 µM in the MTT reduction assay (MTT reduction significantly decreased to 87% 24 h after exposure to 10 µM RHO123 and 10 µM ZOS). Therefore, to accurately assess the effects of TXs on P-gp activity, both the P-gp substrate and specific inhibitor were used at a safer concentration of 5 µM.

Figure 4.

Cytotoxicity of TX1–5 (5, 10 and 20 µM), RHO123 (5 and 10 μM) and ZOS (2.5, 5 and 10 μM), evaluated in HK-2 cells by the MTT reduction assay, 24 h after exposure. Results are presented as mean + SD from 6 independent experiments performed in triplicate. Statistical comparisons were performed using one-way ANOVA followed by Dunnett’s multiple comparison test [**** p < 0.0001 versus control cells (0 μM)].

Figure 5.

Cytotoxicity of TX1–5 (5, 10 and 20 µM), RHO123 (5 and 10 μM) and ZOS (2.5, 5 and 10 μM), evaluated in HK-2 cells by the neutral red uptake assay 24 h after exposure. Results are presented as mean + SD from 6 independent experiments performed in triplicate. Statistical comparisons were performed using one-way ANOVA followed by Dunnett’s multiple comparison test [* p < 0.05; **** p < 0.0001 versus control cells (0 μM)].

3.3. Effects of TX1–5 on P-gp Transport Activity

P-gp activity was assessed by the RHO123 accumulation assay at two distinct incubation time-points. In the first experimental approach, HK-2 cells were incubated with TX1–5 (0, 5, 10, and 20 µM) for 120 min, either in the presence or absence of zosuquidar (5 µM), with simultaneous exposure to the fluorescent P-gp substrate RHO123 during the last 90 min of this incubation period. The results showed that all thioxanthonic derivatives except TX4 significantly increased P-gp transport activity at all concentrations tested (5, 10, and 20 µM) after 120 min of incubation (Figure 6). Specifically, TX1 enhanced P-gp activity to 122%, 117%, and 126% at 5, 10, and 20 µM, respectively. Higher levels of P-gp activation were observed in the presence of TX2, with activity values of 141%, 131%, and 140% at 5, 10, and 20 µM, respectively. TX3 also promoted a significant increase in P-gp activity, reaching 125%, 121%, and 132% at the same concentrations. Lastly, TX5 enhanced P-gp activity to 126% and 124% at 5 µM and 10–20 µM, respectively.

Figure 6.

P-gp activity evaluated in HK-2 cells in the presence of TX1–5 (5, 10 and 20 µM, for 120 min). Results are presented as mean + SD from 5 independent experiments performed in triplicate. Statistical comparisons were performed using one-way ANOVA followed by Tukey’s multiple comparisons test [* p < 0.05; ** p < 0.01; *** p < 0.001; **** p < 0.0001 versus control cells (0 μM)].

In the second experimental approach, HK-2 cells were preincubated with TX1–5 (0, 5, 10, and 20 µM) for 24 h, followed by removal of the compounds and subsequent incubation with HBSS+/+ or zosuquidar (5 µM, prepared in HBSS+/+) for 120 min, simultaneously with RHO123 (during the final 90 min of this incubation period). Assessment of P-gp activity after the 24 h pre-exposure to the compounds showed that the thioxanthonic derivatives significantly increased P-gp transport activity at all tested concentrations, except TX1, which significantly increased P-gp activity only at 20 µM (108%), and TX4, which did not show significant changes at any concentration (Figure 7). TX2 also promoted a significant enhancement in P-gp activity, with values of 112%, 109%, and 106% at 5, 10, and 20 µM, respectively. The highest P-gp activity levels were achieved with TX3, reaching 121%, 114%, and 113% across the tested concentrations (5, 10, and 20 µM, respectively). Lastly, TX5 produced moderate though significant increases in P-gp activity, with values of 116% and 118% at 5–10 µM and 20 µM, respectively.

Figure 7.

P-gp activity evaluated in HK-2 cells following a 24 h pre-exposure to TX1–5 (5, 10 and 20 µM). Results are presented as mean + SD from 5 independent experiments, performed in triplicate. Statistical comparisons were performed using one-way ANOVA followed by Tukey’s multiple comparisons test (* p < 0.05; ** p < 0.01; **** p < 0.0001).

3.4. Effects of TX1–5 on P-gp Expression

To specifically assess changes in P-gp expression in response to the selected TXs, flow cytometry analysis was performed using a P-gp monoclonal antibody (UIC2-PE). HK-2 cells were incubated with each of the five TXs (20 µM) for 24 h, after which P-gp expression levels were measured. Among the tested compounds, only TX2 significantly upregulated P-gp expression compared to the untreated control cells (Figure 8). Indeed, in the presence of TX2, P-gp expression in HK-2 cells increased to 129% compared with control cells (0 µM).

Figure 8.

P-gp expression evaluated in HK-2 cells after exposure to TX1–5 (20 µM) for 24 h. Results are presented as mean + SD from 7 independent experiments performed in duplicate. Statistical comparisons were performed using one-way ANOVA followed by Dunnet’s multiple comparisons test (**** p < 0.0001).

3.5. Thioxanthones’ Protective Effect Against Cisplatin-Induced Cytotoxicity

The protective effects of TX1–5 against cisplatin-induced nephrotoxicity were evaluated in HK-2 cells using an MTT reduction assay. Cells were exposed to increasing concentrations of cisplatin (0–500 µM) in the presence or absence of TX1–5 (5, 10, or 20 µM) for 24 h. Figure 9 depicts the concentration–response curves obtained for each concentration of the five thioxanthones selected for the study, and Table 3 presents the fitted parameters (IC₅₀, Top, Bottom, and Hill slope).

Figure 9.

Concentration–response curves of cisplatin (CP, 0–500 μM) assessed in human kidney (HK)-2 cells by the MTT reduction assay, 24 h after exposure to CP in the presence or absence of TX1–5 (5, 10, and 20 μM). The results are presented as mean ± 95% confidence interval (CI) of a minimum of 4 independent experiments (3 replicates each). The concentration–response curves (% MTT reduction) were fitted using the least squares method as a fitting method and the comparisons between curves (Bottom, Top, LOG IC₅₀ and Hill Slope) were made using the extra sum of squares F test.

Table 3.

IC₅₀, Top, Bottom and Hill Slope of Cisplatin concentration–response curves, assessed in human kidney (HK)-2 cells by the MTT reduction assay, 24 h after exposure to cisplatin in the presence or absence of thioxanthones (5, 10 and 20 μM). Results are presented as mean ± 95% confidence interval (CI) of a minimum of 4 independent experiments (3 replicates each). The concentration–response curves (% MTT reduction) were fitted using the least squares method as a fitting method and the comparisons between curves (Bottom, Top, LOG IC₅₀ and Hill Slope) were made using the extra sum of squares F test.

Best-Fit Curves	Global	IC50	Top	Bottom	Hill Slope
CP	-	31.2 (29.2–33.3)	99.5 (97.7–101)	15.2 (13.1–17.1)	−1.98 (−2.24–−1.76)
CP + TX1 (5 μM)	ns	36.3 * (31.5–41.6)	96.7 (92.7–101)	14.7 (9.98–18.8)	−1.94 (−2.56–−1.47)
CP + TX1 (10 μM)	*	37.2 (33.8–40.8)	97.8 (95.1–101)	18.3 (15.1–21.3)	−2.50 (−3.10–−2.02)
CP + TX1 (20 μM)	*	34.2 (31.2–37.5)	99.8 (97.0–103)	16.0 (12.7–19.2)	−2.60 * (−3.27–−2.07)
CP	-	31.9 (30.5–33.3)	97.8 (96.4–99.3)	10.3 (8.64–12.0)	−2.55 (−2.87–−2.28)
CP + TX2 (5 μM)	ns	31.9 (29.1–34.8)	102 * (98.4–105)	7.03 (3.10–10.7)	−2.39 (−3.02–−1.92)
CP + TX2 (10 μM)	****	37.9 **** (35.7–40.2)	101 * (98.6–103.)	9.48 (6.54–12.3)	−3.57 ** (−4.27–−2.99)
CP + TX2 (20 μM)	****	37.2 **** (35.3–39.2)	101 * (98.6–103)	13.9 * (11.5–16.1)	−3.32 ** (−3.83–−2.88)
CP	-	31.2 (29.7–32.7)	96.6 (95.0–98.3)	8.01 (6.15–9.81)	−2.44 (−2.77–−2.16)
CP + TX3 (5 μM)	****	32.3 (30.2–35.2)	101 ** (98.5–104)	8.61 (5.08–12.0)	−2.84 (−3.48–−2.32)
CP + TX3 (10 μM)	**	31.3 (29.1–33.6)	101 ** (98.2–104)	8.49 (5.18–11.6)	−2.89 (−3.54–−2.37)
CP + TX3 (20 μM)	*	33.7 (30.7–37.1)	96.4 (92.6–100)	9.64 (5.24–13.8)	−3.43 *** (−4.48–−2.61)
CP	-	31.3 (29.8–32.9)	98.4 (96.8–100)	8.54 (6.50–10.5)	−2.59 (−2.95–−2.28)
CP + TX4 (5 μM)	**	34.0 (31.5–36.6)	101 (97.5–104)	9.58 (5.98–13.0)	−3.01 (−3.72–−2.41)
CP + TX4 (10 μM)	**	34.6 (30.6–39.0)	102 (97.8–106)	7.59 (2.29–12.4)	−2.08 (−2.68–−1.62)
CP + TX4 (20 μM)	ns	34.3 (31.7–36.9)	99.6 (96.8–103)	8.95 (5.50–12.2)	−2.76 (−3.41–−2.22)
CP	-	31.1 (29.6–32.7)	96.9 (95.3–98.7)	7.28 (5.29–9.19)	−2.50 (−2.87–−2.18)
CP + TX5 (5 μM)	**	31.7 (29.5–34.0)	101 * (98.4–104)	8.01 (4.74–11.1)	−2.70 (−3.31–−2.20)
CP + TX5 (10 μM)	ns	30.2 (28.0–32.5)	98.6 (95.6–102)	6.93 (3.65–10.0)	−2.58 (−3.23–−2.08)
CP + TX5 (20 μM)	ns	31.4 (28.9–34.1)	97.3 (94.2–101)	7.99 (4.21–11.5)	−2.79 (−3.56–−2.19)

Legend: ns. p > 0.05; * p < 0.05; ** p < 0.01; *** p < 0.001; **** p < 0.0001, compared to each cisplatin (CP) curve.

Cisplatin alone had IC₅₀ values ranging from 31.1 to 31.9 µM. A significant rightward shift in the cisplatin concentration–response (MTT reduction, %) curves, accompanied by significant increases in the corresponding IC₅₀ values, was observed in the presence of TX2. Indeed, among the tested compounds, TX2 showed the strongest protective effect, significantly increasing the IC₅₀ to 37.9 µM and 37.1 µM (p < 0.0001) at 10 and 20 µM, respectively, suggesting a robust protective effect against cisplatin-induced cytotoxicity (Table 3). For the remaining compounds, although significant rightward shifts in the cisplatin concentration–response curves were observed (e.g., 10 and 20 µM TX1; 5, 10, and 20 µM TX3), no significant differences were detected in the IC₅₀ values of the fitted curves.

These findings indicate that thioxanthones can modulate cisplatin-induced cytotoxicity in HK-2 cells through a mechanism potentially involving P-gp modulation. Among the compounds tested, TX2 (particularly at 10 and 20 µM, and after 24 h of exposure) demonstrated the most consistent and robust protective profile, as evidenced by the significant rightward shifts in the concentration–response curves, resulting in significant increases in IC₅₀ values. This suggests that TX2 effectively reduces the cytotoxic effect of cisplatin, thereby enhancing cell viability and providing a protective effect against cytotoxicity.

4. Discussion

The present study integrated in silico and in vitro approaches to investigate the interactions and functional modulation of human P-gp by five thioxanthonic derivatives (TX1–5). Complementing this, computational analyses were expanded to include a broader range of (thio)xanthonic derivatives, comprising aminated thioxanthones (ATX1–8), oxygenated xanthones (OX1–2 and 4–6), and hydroxylated xanthones (X1–5), all previously reported as P-gp activators [19,37,41,44,45].

The computational study began with molecular docking analysis using Autodock Vina software to estimate the binding affinities of ligands to the P-gp structural model at three selected binding sites: the DBP, NBD1, and NBD2, which were previously described as key for interpreting P-gp modulation [42]. At the DBP, binding affinity values ranged from −7.0 kcal/mol (TX4) to −9.3 kcal/mol (TX3), with an average value of −7.7 kcal/mol, with no significant differences being observed among ligand groups. Similarly, NBD1-binding affinity values spanned from −6.6 kcal/mol (TX4) to −8.7 kcal/mol (TX3), averaging −7.6 kcal/mol, again with no significant differences among groups. At NBD2, slightly weaker binding affinities were observed, ranging from −6.4 kcal/mol (TX4) to −7.8 kcal/mol (TX3), with an overall average of −7.1 kcal/mol, and no significant differences were observed across groups, making it difficult to establish any structure–activity relationship. Consistently, TX3 showed the strongest binding (lowest binding affinity energy) across all sites, indicating higher complex stabilization energy while binding to the DBP. In contrast, TX4 consistently exhibited the weakest binding (highest energy values), suggesting that is less effective in stabilizing the ligand–P-gp complex (Table 2, Figure 1).

SMINA and GNINA were then employed to validate the Autodock Vina docking results. SMINA (Vinardo) produced affinity values comparable to those obtained with Autodock Vina. At the DBP, affinity values ranged from −6.8 kcal/mol (OX2 and OX5) to −10.9 kcal/mol (TX3), with an average of −7.9 kcal/mol. NBD1 and NBD2 showed weaker affinity values, ranging from −5.5 kcal/mol (OX1) to −8.2 kcal/mol (TX3), averaging −6.5 kcal/mol, and from −5.8 kcal/mol (TX4) to −7.54 kcal/mol (TX1), averaging −6.7 kcal/mol, respectively. Consistently, GNINA revealed that at the DBP, CNN affinities (pK) were highest, ranging from 3.671 (X3) to 7.791 (ATX8), with an average of 6.048. At NBD1, CNN affinity values ranged from 3.875 (X5) to 6.171 (ATX6), averaging 4.985, while at the NBD2, affinity values were weaker, ranging from 3.244 (X1) to 5.842 (ATX7), with an average of 4.519. Across all docking software applied, ligands consistently showed the greatest preference for the DBP, followed by NBD1.

Following the identification of the best docking poses from the molecular docking analysis, an evaluation of the location and type of interactions was performed. This involved a detailed examination of how and where each ligand binds within the P-gp structure, considering the spatial orientation and specific residue contacts in order to elucidate the molecular basis of ligand–P-gp interactions. Initially, the MOE software was used to visually inspect and characterize each ligand’s binding site within the DBP, specifically in the H-, R- or M-sites (Table 2, Figure 2), as well as to perform a broader search for potential novel binding locations within the NBD1 (Figure 3). The SMINA and GNINA docking software were used to provide insights into the validity of the Vina docking results, and generally confirmed the best docking poses of each ligand. SMINA showed similar binding site preferences to Autodock Vina, except for OX4, X1, X5. Regarding GNINA, although greater discrepancies in site preferences were observed, this software relies on CNN-based scoring functions that were primarily trained on typical protein–ligand complexes [49]. Consequently, its performance may be less reliable for large and highly flexible membrane transporters such as P-gp. Subsequently, a more detailed analysis was performed using BINANA software to examinate the specific types of interactions occurring at these sites and to identify the specific P-gp residues involved, using the Autodock Vina docking results, as similar outcomes were obtained with the other docking software. The DBP and NBD1 sites were selected for detailed study due to the observed binding preferences of the (thio)xanthonic derivates for these regions, supported by their lower binding energy values from the molecular docking analysis compared to those at NBD2.

Starting with the DBP, the superposition of the P-gp structural model with the ligands’ best docking poses revealed two predominant binding regions: the M-site, occupied by ATXs, OX2, OX4, OX5, TX1, TX3–5, and X1, and the R-site, occupied by OX1, OX6, TX2, and X2–5, as detailed in Table 2. This binding preference for the M/R binding sites was further corroborated by BINANA analysis, which identified the specific P-gp residues interacting with the ligands as belonging to these two distinct regions (Table S1). Accordingly, the comparative analysis at the DBP focused on categorizing ligands into two groups: ligands binding at the M-site vs. the R-site.

All compounds binding at the M-site consistently interacted with PHE983 via hydrophobic contacts and, in some cases (ATX5, ATX7, ATX8, OX4, and TX5), through T-stacking interactions. Notably, over 80% of the 16 ligands binding at the M-site shared hydrophobic interactions with several key P-gp residues: ALA729 (except TX4), ILE340 (except OX2 and OX4), PHE728 (except OX2), PHE732 (except ATX4, X1 and X5), and VAL982 (except OX2, TX1 and X1). At PHE732, diverse types of interactions were observed, including T-stacking interactions (ATX1, ATX2, ATX3, ATX4, OX4, and TX1), cation–π interactions (ATX7, ATX8, and TX5), and π–π stacking interactions (TX3). Additionally, although below the 80% threshold, since ATX1, ATX3, ATX5, and OX4 lacked hydrophobic contacts, PHE343 engaged in T-stacking interactions with OX2, OX5, TX3 and X1, highlighting its role in ligand binding at this site.

On the other hand, ligands binding at the R-site primarily interacted with ILE700, ILE836, and TYR998 through hydrophobic contacts. Additionally, ligands connecting at the R-site also created other types of interactions with TYR998, including π–π stacking interactions observed with OX1, OX6, TX2, and X3; as well as hydrogen bonds formed by X4. This highlights TYR998 as a key residue mediating diverse ligand–P-gp interactions within the R-site.

Overall, compounds exhibit a preference for binding to two distinct regions within the drug-binding pocket (DBP) of P-gp: the M-site and the R-site, although with a much higher preference for the M-site. Ligands binding the M-site predominantly interact with conserved hydrophobic residues such as PHE983 and PHE732 through various non-covalent interactions, while those associating with the R-site mainly engage hydrophobic contacts and specialized interactions with TYR998. Slight structural variations within the structure (when comparing TX3–5) had influences on distinct binding patterns, underscoring site-specific residue interactions among this library of (thio)xanthonic derivative activators of P-gp. Importantly, in the interpretation of these docking results, the most probable interaction site should be considered as the one associated with the lowest-energy binding pose. Therefore, when comparing the M and R sites, TX2 is the only derivative that demonstrates a clear energetic preference for the R-site. All other compounds exhibit more favorable interactions with either the M-site or NBD1. Notably, X4 and X5 present identical affinity energies for sites R and NBD1, whereas X2 displays only a negligible difference between these sites, indicating the absence of a definitive binding site preference for this derivative.

The computational investigations focusing on P-gp activation remain, to the best of our knowledge, very scarce. Sterz and colleagues reported in 2009 that a majority of the 27 newly synthesized imidazobenzothiazoles and imidazobenzimidazoles (compounds structurally related to the known P-gp activators QB102 and QB11) modulated P-gp activity in a substrate-dependent manner. Specifically, these compounds, by binding to the H-site, directly compete with H-site substrates such as vinblastine and colchicine, increasing the intracellular accumulation and toxicity of these substrates. Concurrently, they stimulated the efflux of R-site substrates like rhodamine 123 and daunorubicin in a concentration-dependent manner, thereby modulating P-gp activity through site-specific interactions. These results support the hypothesis that within the same group of compounds, molecules are capable of binding to the H-site of P-gp, competitively inhibiting H-type substrates, while simultaneously exerting positive cooperative activation of R-site substrates efflux, all without affecting ATP hydrolysis or its affinity for P-gp [55].

Despite the limited findings on P-gp activation, it is important to note that most of the research has predominantly focused on P-gp inhibition, as evidenced by several notable studies. For instance, using molecular docking and molecular dynamics approaches, researchers evaluated the binding energies and inhibitory potential of 512 drug candidates in their clinical or investigational stages for overcoming the MDR phenomena. Employing the AutoDock4.2.6 software, they identified five promising P-gp inhibitors, namely valspodar, dactinomycin, elbasvir, temsirolimus, and sirolimus, which were confirmed in molecular dynamics simulations in a membrane-water environment, demonstrating favorable ADMET (Absorption, Distribution, Metabolism, Excretion, and Transport) profiles [56]. Similarly, Tombline et al. investigated a set of tetramethylrosamine analogues targeting the P-gp R-site, assessing the binding specificity and its coupling to the distal ATP catalytic site via ATPase stimulation. This work revealed up to ~1000-fold variation in ATPase specificity constants among analogues, with further analyses using ADP-Vi trapping and ATP binding assays demonstrating that drug binding can influence ATP hydrolysis transition-state stabilization and ATP occlusion at a single site. These findings highlight how structural variations within a drug-like scaffold can dissect the coupling between drug binding and ATP hydrolysis, offering valuable insights for designing novel P-gp modulators targeting the R-site [57].

With the same goal of overcoming the MDR phenomenon, Mora Lagares and group developed a 3D model of the human P-gp (PDB ID: 6QEX), constructed using mouse P-gp structural templates from the PDB repository. They applied this model to perform molecular docking analysis of 13 compounds (including some known P-gp substrates, inhibitors, or both), using two different algorithms: CDOCKER and GOLD. The in silico results revealed a high number of hydrogen bonds and diverse π interactions (π–sigma, π–alkyl, and π–π types of interactions), suggesting a stabilization of the ligand–protein complex at the binding site. The inhibitory potential of the compounds was validated through an in vitro approach, where nine compounds from the docking set (cyclosporine A, amiodarone, doxorubicin, verapamil, valproic acid, busulfan, gentamicin, pamidronate, and paraquat) significantly increased the accumulation of RHO123 (10 µM, for 120 min) in Caco-2 cells [58].

Also focusing on the DBP, Isca and colleagues’ group used molecular docking to demonstrate that benzoylated royleanone derivatives interact with the P-gp DBP. These compounds interacted either at the R/H site, leading to competitive inhibition (without specifying particular residues), or at the M-site, where noncompetitive inhibition was associated with interactions involving several defined residues, including LEU64, MET67, MET68, PHE71, THR75, PHE728, TYR949, LEU971, PHE974, ILE977, and MET982 [59]. These observations are consistent with the interaction patterns revealed in our study.

At the NBD1, all ligands interact in the same region, as depicted in Figure 3. Further detailed analysis with BINANA revealed that all the compounds shared interactions with common P-gp residues, including similar interaction types in some cases. Notably, all ligands interacted with ASP555, predominantly through hydrophobic contacts and, in some cases, through salt bridges (ATX4, ATX7, ATX8, TX1, and TX5). Moreover, several other P-gp residues were identified for binding to at least 20 of the 23 ligands included in the study (>80%). Specifically, all ligands interact, through hydrophobic contacts, with: ILE585, except ATX1; with LEU553, except ATX1, TX4, and X5; with THR906, except ATX2, ATX4, and TX1; and with VAL437, except ATX4. Furthermore, with THR906, ATX1 and ATX5 also established hydrogen bonds. Unlike other prominent residues, ARG659 showed limited hydrophobic interactions, engaging with only ATX1, ATX3, and TX3. Nevertheless, ARG659 played a significant role in ligand binding within NBD1 by forming hydrogen bonds with multiple ligands, including TX3, TX4, TX7, OX4, TX1, and X1–5, highlighting its importance in stabilizing ligand interactions despite fewer hydrophobic contacts.

Studies suggest that P-gp substrates primarily interact with binding sites located within the TMDs, whereas ATP binds at the interface between the NBDs, specifically between the opposing Walker A motifs and LSGGQ signature sequences, highlighting the critical role of these domains in transporter function [60,61]. The functional relevance of NBD1 has been recognized for years, given its key role in driving the conformational changes required for ATP hydrolysis. Gabriel et al. (2003) advanced this understanding by introducing single cysteine residues at specific positions within NBD1 and applying molecular modeling, concluding that specific residues located in the α-helical subdomain of NBD1 were altered by the conformational changes that occur during ATP hydrolysis [62]. Similarly, the group led by Loo focused on elucidating how NBD–TMD interactions contribute to the transport cycle, proposing that a conformation involving close association of the NBDs may mimic the structural state induced by drug binding, thereby stimulating ATPase activity. Their cross-linking experiments further supported this model by demonstrating that bringing the central portion of NBD1 near the N-terminal region of NBD2 significantly enhances ATPase activity, highlighting dynamic domain movements during the catalytic cycle [60].

The TX1–5 series of (thio)xanthonic derivatives was selected for subsequent experimental analysis based on two considerations: (i) this group previously demonstrated the highest percentage of increase in P-gp activity, particularly TX5, at the intestinal level [37]; and (ii) the current computational study revealed that, within this group, TX3 and TX4 exhibited the highest and lowest binding affinities to P-gp, respectively, representing a spectrum of interaction profiles that warrant further experimental investigation and validation. Initial safety assessments of TX1–5 were conducted to evaluate their impact on HK-2 cell viability using the MTT reduction and NR uptake assays. The MTT assay demonstrated that none of the compounds exhibited cytotoxicity under the tested conditions (Figure 4). Consistently, the NR uptake assay confirmed the lack of significant cytotoxicity for all compounds except TX3, which induced a minor though statistically significant reduction in NR uptake at 10 and 20 μM (Figure 5). However, despite this reduction, mean cell viability upon exposure to TX3 remained high (95% and 89% 24 h after exposure to 10 and 20 μM, respectively). These findings, alongside the absence of significant effects in the MTT reduction assay, suggest that TX3 does not pose significant cytotoxic risk to HK-2 cells under the experimental conditions used. The discrepancy between assays possibly indicates a mild, lysosome-specific perturbation, including alterations of lysosomal membrane dynamics or pH, without compromising cell viability and leaving mitochondrial function intact.

The cytotoxic effects of these thioxanthonic compounds have not yet been evaluated in the HK-2 cell line or in any other renal-derived cell lines. However, their safety profile was previously assessed in intestinal cell models. Silva and colleagues investigated the cytotoxicity of TX1–5 at concentrations ranging from 0 to 100.0 μM in Caco-2 cells following 24 h of exposure, using both the MTT reduction and NR uptake assay. They reported no significant cytotoxic effects under the tested conditions, supporting the compounds’ safety [37].

Following the selection of TX1–5 non-cytotoxic concentrations, direct activation of the transporter mediated by these derivatives was assessed using the RHO123 accumulation assay following 120 min of exposure. The results indicated that all tested TXs significantly increased P-gp activity at all tested concentrations (5, 10, and 20 µM), except for TX4, which did not produce significant changes (Figure 6), which correlates with in silico data. The maximal increase in P-gp transport activity was observed for TX2, at 5 µM, reaching 141% compared to control cells (0 µM). These findings identify TX1, 2, 3 and 5 as clear P-gp activators. Additionally, after 24 h of pre-exposure to TX1–5, the RHO123 accumulation assay revealed that TX1 (20 µM), and TX2, TX3 and TX5 (5, 10, and 20 µM) were again capable of significantly increasing P-gp transport activity. Correspondingly, flow cytometry using the UIC2-PE antibody was used to evaluate the potential effects of TXs on P-gp protein expression. Notably, TX2 (20 µM) significantly upregulated P-gp protein expression, compared to the control cells (0 µM), underscoring its potential as a P-gp inducer (Figure 8). Given the absence of effects of TX1, TX3, and TX5 on P-gp expression after 24 h of exposure, their significant enhancement of P-gp transport activity at this timepoint is likely attributable to direct activation of the transporter rather than upregulation of P-gp protein expression.

Overall, these in vitro findings suggest that TX2 functions as both a P-gp activator and inducer, as demonstrated by its ability to increase P-gp transport activity after 120 min and sustain this effect for up to 24 h of exposure, alongside significant upregulation of P-gp expression after 24 h of incubation. Therefore, the increased P-gp transport activity observed in the RHO123 accumulation assay after 24 h of pre-exposure to TX2 may be attributed to ABCB1 gene upregulation and/or prolonged direct activation of P-gp. Relatively to TX1, TX3 and TX5, the enhanced transport activity observed in the RHO123 accumulation assay after 24 h of pre-exposure likely results from sustained P-gp activation rather than gene induction-related effects, as no significant changes in P-gp expression were observed. Therefore, TX1, TX3 and TX5 should be classified only as potential P-gp activators. In contrast, TX4 did not induce any significant changes in either P-gp transport activity or expression and should therefore not be considered as a P-gp activator or inducer.

The final experimental approach evaluated the protective effects mediated by TX against the cytotoxic P-gp substrate cisplatin (0–500 µM, for 24 h). Using the MTT reduction assay, TX2 (10 and 20 µM, for 24 h) significantly protected HK-2 cells from cisplatin-induced cytotoxicity (Figure 9, Table 3). This protective effect can be partially attributed to the previously demonstrated role of TX2 as both a P-gp activator and inducer. In the presence of TX2 for 24 h, P-gp transport activity was significantly increased through direct activation and ABCB1 upregulation, leading to a higher and more efficient elimination of cisplatin from HK-2 cells and reduced intracellular accumulation. Consequently, cisplatin-induced cytotoxicity was attenuated, leading to increased viability of HK-2 cells in the presence of TX2.

TX1–5, as previously stated, possess a dibenzo-γ-pyrone tricyclic scaffold and have previously been described as P-gp inducers and/or activators at the intestinal level, with the additional ability to protect enterocytes from the cytotoxicity caused by toxic P-gp substrates. Indeed, Silva et al. evaluated the potential of these thioxanthonic derivatives to act as P-gp inducers and/or activators, focusing on their capability to prevent paraquat-induced toxicity in Caco-2 cells. Their findings demonstrated that TX1–5 at 20 µM significantly increased P-gp expression and activity, measured by flow cytometry using the UIC2 antibody and the RHO123 accumulation assay, respectively. Additionally, TX2–5 significantly reduced paraquat-induced cytotoxicity (0–7500 µM), as evidenced by significant increases in the EC50 values of the paraquat + TXs curves, comparing with paraquat alone [37]. TX5 (20 µM) was further evaluated ex vivo and in vivo using rat everted intestinal sacs, where it was confirmed to act both as a P-gp inducer—evidenced by increased protein expression—and as a P-gp activator, demonstrated by the ability to enhance RHO123 efflux [38].

These results partially align with our findings, as all tested thioxanthonic derivatives significantly enhanced P-gp transport activity at 20 µM, specifically at the kidney level, apart from TX4. However, when examining P-gp expression, only TX2 significantly upregulated P-gp expression, contrasting with the data at the intestinal level, where all the thioxanthonic derivatives TX1–5 significantly increased P-gp protein expression [37]. Importantly, in the present study, TX2 was also the only derivative that significantly protected kidney cells against the cytotoxic effects of cisplatin. The apparent discrepancies in the effects of TX1–5 on P-gp transport activity, P-gp expression, and cytoprotection between kidney and intestinal cells may potentially arise from differences in the experimental models used, particularly the cellular origin, as P-gp expression levels differ substantially between organs, which can greatly influence the outcomes. Mechanistically, intestinal epithelial cells express high levels of pregnane X receptor (PXR) and constitutive androstane receptor (CAR), major transcriptional regulators of ABCB1 [63], which likely explains the strong responsiveness of Caco-2 cells to all thioxanthonic derivatives. In contrast, HK-2 cells exhibit minimal PXR/CAR expression [64], and instead rely on alternative pathways such as nuclear factor erythroid 2-related factor 2 (Nrf2) and peroxisome proliferator-activated receptor (PPAR) isoforms for ABCB1 regulation [65,66]. Consequently, TX2 may preferentially activate one of these kidney-relevant pathways, accounting for its unique capacity to induce P-gp expression in HK-2 cells while the other derivatives were inactive (despite their efficacy in Caco-2 cells). Moreover, in the specific context of protection against P-gp–dependent cytotoxicity, the nature of the toxic substrate applied is a critical factor. Variations in substrate affinity or interaction with P-gp may explain the divergent findings across studies.

Cisplatin is a standard chemotherapy agent and a benchmark for studying kidney toxicity, though its classification as a P-gp substrate remains a subject of debate. While some researchers argue that cisplatin is poorly transported by P-gp, there is substantial evidence linking P-gp overexpression to cisplatin resistance in various cancers, including lung cancer and osteosarcoma. Multiple studies [67,68,69,70] have demonstrated that inhibiting P-gp—whether through genetic silencing, chemical inhibitors, or nanoparticles—significantly increases cisplatin accumulation and efficacy. This suggests that even if the transport mechanism is debated, P-gp activity effectively reduces the drug’s impact on cancer cells. Beyond direct transport, P-gp appears to mitigate cisplatin toxicity through cell signaling and protective mechanisms. Gibalová et al. suggest that P-gp upregulation stabilizes anti-apoptotic pathways, helping cells survive cisplatin exposure even without active efflux [71]. Although several studies have reported P-gp-mediated resistance mechanisms that occur independently of drug efflux, other findings support a possible direct involvement of P-gp in cisplatin transport. Therefore, the cytoprotective effect of TX2 observed in our study may reflect enhanced P-gp expression/activity, either through increased cisplatin efflux or via P-gp-associated pro-survival signaling pathways.

Consistent with the present study, Gao et al. studied the protective effects of omeprazole against cisplatin-induced nephrotoxicity both in vitro, using the HK-2 cell line, and in vivo, in male Sprague Dawley rats. Researchers observed that cell viability, assessed by the MTT reduction assay, was significantly improved in the presence of omeprazole (0.4–4000 μg/mL), compared to cisplatin alone (0.02–200 nM), across 12, 24, and 48 h exposures. In rats, co-administration of omeprazole (1.8 or 3.6 mg/kg/day for 5 days, i.p.) alongside a single cisplatin dose (15 mg/kg, one dose, i.p.) reduced plasma blood urea nitrogen and creatinine levels relative to cisplatin treatment alone. Mechanistically, omeprazole decreased OCT2 expression, leading to a lower cisplatin uptake, while also increasing P-gp levels in HK-2 cells co-exposed to both drugs, when compared to cisplatin alone, resulting in an increased cisplatin efflux, and preventing cisplatin-induced cytotoxicity [31].

Computational methods play a critical role in early drug development, particularly for predicting compound interactions with transporters like P-gp. These in silico tools accelerate the discovery process, reduce costs, and reduce reliance on animal testing by simulating molecular interactions and aiding in the identification of potential modulators [72,73,74]. Despite their usefulness, computation techniques present notable limitations in their predictions: compounds may act as both substrates and inhibitors [75,76]; the presence of multiple binding sites and/or allosteric mechanisms complicates data interpretation [8,77]; and transport often involves multiple pathways or can be influenced by passive diffusion, adding further complexity [78,79,80]. In the context of this study, computational evaluations of positive P-gp modulation specifically refer to direct transporter activation through ligand binding, as the modulators directly interact with P-gp itself. Conversely, P-gp induction involves changes in gene expression, where modulators act at the DNA level rather than binding to the transporter. As a result, the potential of a compound to act as a P-gp inducer cannot be predicted using the computational approaches presented here. Nonetheless, modulation of P-gp expression was not the sole focus of this work. Accordingly, computational techniques were employed to provide additional insight into P-gp activation mediated through direct ligand–transporter interactions, since the molecular mechanisms underlying P-gp activation remain insufficiently clarified in the available literature.

These challenges highlight the importance of using a thoughtful, integrated approach combining computational and experimental methods to achieve reliable and comprehensive evaluation of P-gp interactions. Nevertheless, experimental assays, whether as part of in vitro, ex vivo and/or in vivo studies, remain essential for validating computational predictions and investigating complex biological behaviors [81,82,83]. In vitro methodologies are especially valuable in transporter-related research due to their multiple advantages, including cost-effectiveness, time efficiency, lack of significant ethical constraints, capability to screen a high number of compounds under standardized conditions, and, specifically in this field, their unique ability to assess both the expression and functional activity of transport proteins. Nonetheless, these in vitro models have inherent limitations, including variability in transporter expression among different cell lines and the lack of systemic physiological context, which can compromise the accuracy and predictive value of the results [84,85,86].

When comparing the in silico predictions with the in vitro findings obtained in the present study, notable consistencies as well as a few limitations emerge. The computational analysis revealed that TX2 binds to a distinct region within the P-gp structural model, interacting with different residues specifically within the DBP, compared to the other tested compounds. Notably, TX2 was the only thioxanthonic derivative in this series that lacks an amine moiety in its structure, which may influence its polarity, hydrogen-bonding capacity, and preferred orientation within the DBP, potentially contributing to its distinct binding mode and functional profile. This difference in DBP binding may help explain the superior P-gp activation observed in vitro for TX2, which also showed protective effects against cisplatin-induced cytotoxicity in HK-2 cells. Furthermore, TX2 was also the only derivative that significantly up-regulated P-gp expression in HK-2 cells. However, P-gp induction, which involves upstream signaling and gene expression changes at the nuclear level rather than direct physical interaction between P-gp and the inducer, cannot be reliably predicted by in silico approaches that focus solely on direct transporter–ligand interactions. Therefore, the observed protective effects, potentially resulting from both P-gp activation and induction, should not be attributed exclusively to TX2’s distinct DBP binding mode. On the other hand, TX4 displayed higher binding affinity thresholds for interactions with the P-gp model at all evaluated regions (DBP, NBD1 and NBD2) and formed fewer overall contacts, particularly at the DBP, where it established only hydrophobic interactions compared with the richer interaction network observed for the other TXs evaluated in vitro. In line with these in silico findings, TX4 failed to activate P-gp in vitro, suggesting that its weaker and less diverse binding pattern, especially at the DBP, may potentially underlie its lack of functional activity. Regarding NBD1, all tested compounds engaged the rarely explored hotspot on this domain, and TX4 established contacts of similar general nature (polar and hydrophobic) to those observed for TX2 and the other activators, indicating that occupancy of this site alone is not sufficient to predict functional activation. Therefore, our NBD1 analysis primarily served to identify a previously underappreciated binding region rather than to discriminate the functional behavior of individual ligands.

Overall, these findings underscore the complementary yet distinct roles of in silico and in vitro approaches, highlighting the need for integrated strategies to fully elucidate compound–P-gp interactions. Therefore, while in vitro models provide a valuable first-line experimental platform, relying on them alone is insufficient. For more comprehensive and translationally relevant insights, it is crucial to complement them with in vivo or ex vivo studies, which capture the complexity of whole biological systems and account for processes such as absorption, distribution, metabolism, and excretion.

5. Conclusions

By combining computational modeling with experimental data, this study confirms that thioxanthonic derivatives effectively activate P-gp in renal cells, serving as a defense mechanism against kidney toxicity. Computational analysis indicates that these compounds function as direct P-gp activators by binding to specific P-gp sites (DBP and NBD1), while in vitro experiments identify TX2 as the most potent derivative, successfully protecting HK-2 cells from cisplatin-induced damage. These findings validate the strategy of increasing P-gp expression/activity to achieve cytoprotection, positioning thioxanthonic derivatives as promising candidates for future drug development, pending further in vivo safety and efficacy validation.

Abbreviations

The following abbreviations are used in this manuscript:

2D	2-Dimensions
3D	3-Dimensions
ABC	ATP-Binding Domain
ADMET	Absorption, Distribution, Metabolism, Excretion, and Transport
ADP	Adenosine Diphosphate
ATCC	American Type Culture Collection
ATP	Adenosine Triphosphate
ATX	Chiral Aminated Thioxanthone
BINANA	BINding ANAlyzer
CAR	Constitutive androstane receptor
CNN	Convolutional neural network
CI	Confidence Interval
DBP	Drug-Binding Pocket
DBS	Drug-Binding Site
DMSO	Dimethyl sulfoxide
DNA	Deoxyribonucleic acid
EDTA	Ethylenediamine tetraacetic acid
FBS	Fetal Bovine Serum
FI	Fluorescence Intensity
FIIA	FI of RHO123 Inhibited Accumulation
FINA	FI of RHO123 Normal Accumulation
HBSS	Hank’s balanced salt solution
HK-2	Human Kidney 2
HPLC-DAD	High-Performance Liquid Chromatography—Diode Array Detection
IC50	Half-Maximal Inhibitory Concentration
ICH	Intracellular Coupling Helices
IUPAC	International Union of Pure and Applied Chemistry
MDR	Multidrug Resistance
MDR1	Multidrug Resistance 1
MFI	Mean Fluorescence Intensity
MOE	Molecular Operating Environment
MTT	3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyl tetrazolium bromide
NBD	Nucleotide Binding Domain
Nrf2	Nuclear factor erythroid 2-related factor 2
NR	Neutral Red
OX	Oxygenated Xanthone
PBS	Phosphate-Buffered Saline
P-gp	P-glycoprotein
PPAR	Peroxisome proliferator-activated receptor
PXR	Pregnane X receptor
RHO123	Rhodamine 123
RPMI	Roswell Park Memorial Institute
SMILES	Simplified Molecular Input Line Entry System
TMD	Transmembrane Domain
TMH	Transmembrane Helice
TX	Thioxanthone
X	Xanthone
ZOS	Zosuquidar

Supplementary Materials

The following supporting information can be downloaded at https://www.mdpi.com/article/10.3390/jox16020055/s1, Table S1: P-gp residues interacting with each ligand included in the present study, all previously reported as P-gp activators.

Author Contributions

Conceptualization, F.R. and R.S.; methodology, J.V.-M., D.J.V.A.d.S., E.S. and R.S.; software, J.V.-M. and D.J.V.A.d.S.; formal analysis, J.V.-M. and R.S.; investigation, J.V.-M., A.P. and R.S.; writing—original draft preparation, J.V.-M.; writing—review and editing, D.J.V.A.d.S., E.S., A.I.M., M.P., F.R. and R.S.; supervision, A.I.M., M.P., F.R. and R.S. All authors have read and agreed to the published version of the manuscript.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

The original contributions presented in this study are included in the article/Supplementary Material. Further inquiries can be directed to the corresponding authors.

Conflicts of Interest

The authors declare no conflicts of interest.

Funding Statement

This work was supported by FCT—Fundação para a Ciência e Tecnologia, I.P., in the scope (J.V.-M., F.R. and R.S.) of the project UID/04378/2025 (https://doi.org/10.54499/UID/04378/2025), and UID/PRR/04378/2025 (https://doi.org/10.54499/UID/PRR/04378/2025), of the Research Unit on Applied Molecular Biosciences—UCIBIO and the project LA/P/0140/2020 (https://doi.org/10.54499/LA/P/0140/2020) of the Associate Laboratory Institute for Health and Bioeconomy—i4HB. Additionally, the work of A.P. and E.S. was supported by FCT and by the European Commission’s Recovery and Resilience Facility in the scope of the projects UID/04423/2025 (https://doi.org/10.54499/UID/04423/2025), UID/PRR/04423/2025 (https://doi.org/10.54499/UID/PRR/04423/2025), and LA/P/0101/2020 (https://doi.org/10.54499/LA/P/0101/2020). The work of D.J.V.A.d.S. was supported in the scope of the project UID/04567/2025 (CBIOS) and Seed Funding Project (COFAC/ILIND/CBIOS/2/2023). J.V.-M. was supported by a PhD grant (number 2021.06177.BD, https://doi.org/10.54499/2021.06177.BD), through national funds.