In Vitro Pharmacokinetic Profiling of Thiourea Derivatives of Naproxen With Anti‐Inflammatory and Anticancer Activity

ABSTRACT

Pharmacokinetic deficiencies account for nearly 40% of clinical drug candidate failures, emphasising the importance of early in vitro evaluation. Following the 3Rs principle, this study assessed the in vitro pharmacokinetic properties of thiourea derivatives of naproxen (1–14). Permeability assessment using the parallel artificial membrane permeability assay method revealed that derivatives 4 (‐logPe = 4.13, p = 0.0859) and 7 (‐logPe = 3.94, p = 0.2291) exhibited the highest passive gastrointestinal absorption potential, comparable to naproxen. Evaluation of binding affinity to human serum albumin (HSA) using the high‐performance affinity chromatography method identified two groups of compounds: high‐binding aromatic amine derivatives (except derivative 5) and low‐binding aromatic amino acid ester derivatives. Microsomal stability assays indicated that compound 7 underwent the most extensive metabolic degradation (65.83% decrease). These findings suggest that further structural optimisation of compound 7, such as esterification with higher or branched alcohols, should lead to the improvement of HSA binding and reduction of metabolic susceptibility. Conversely, compound 4 combines favorable biological activity with a pharmacokinetic profile closely resembling that of naproxen. Although limited by the exclusive use of in vitro models and the absence of in vivo validation, this study provides valuable insights for the rational design of naproxen‐based thiourea derivatives with enhanced anti‐inflammatory and anticancer potential.

This in vitro study indicated that thiourea derivatives of naproxen exhibit a significant potential for passive gastrointestinal absorption, among which derivatives 4 and 7 demonstrated the highest permeability potential. Aromatic amine derivatives without polar groups showed the strongest interaction with human serum albumin. The highest metabolic rate was observed for compound 7, while compounds 3 and 4 demonstrated greater microsomal stability.

1. Introduction

Non‐steroidal anti‐inflammatory drugs (NSAIDs) are among the most widely used medications due to their analgesic, antipyretic, and anti‐inflammatory effects [1, 2]. Naproxen (2‐(6‐methoxynaphthalen‐2‐yl)propionic acid), a propionic acid derivative, acts as a non‐selective inhibitor of cyclooxygenase isoforms (COX‐1 and COX‐2), thereby reducing prostaglandin, prostacyclin, and thromboxane synthesis [3]. It is administered orally, with a bioavailability of approximately 50% [4, 5] and an elimination half‐life of 12–17 h, allowing for sustained therapeutic plasma levels and convenient dosing [6].

Since oral administration remains the most common route for drug delivery, evaluation of passive gastrointestinal absorption is essential in early drug development [7]. Drug absorption is influenced by gastrointestinal conditions and the physicochemical properties of the compound. Therefore, assessing these properties of new naproxen derivatives is crucial for optimizing their bioavailability and therapeutic potential [8].

The Three Rs (3Rs) principle (Replacement, Reduction, and Refinement) seeks to ensure the rational and respectful use of laboratory animals and plays a crucial role in the development of new drugs. In vitro studies, strongly supported by the 3Rs principle, have become essential tools in modern drug development, particularly in its early phases. By integrating innovative in vitro approaches, researchers can obtain more human‐relevant data, improve the predictive accuracy of drug responses, and accelerate the development of safer and more effective therapeutics [9, 10]. A commonly used in vitro method for the estimation of passive gastrointestinal permeability is the parallel artificial membrane permeability assay (PAMPA). The PAMPA model simulates passive drug diffusion across a lipid membrane that mimics the gastrointestinal barrier. It relies on a concentration gradient between donor and acceptor compartments, allowing assessment of permeability under physiologically relevant pH conditions. The obtained permeability coefficient provides an estimate of a compound's potential for passive gastrointestinal absorption [11, 12].

Pharmacological activity, safety, and tissue distribution of a drug are determined by the concentration of its free (unbound) form in plasma. Plasma protein binding is a critical physiological factor affecting the pharmacokinetics and pharmacodynamics of drugs [13]. The main plasma proteins responsible for drug binding are human serum albumin (HSA) and α1‐acid glycoprotein (AGP), with HSA preferentially binding neutral and acidic drugs, while AGP binds basic drugs. Several in vitro techniques, including equilibrium dialysis, ultrafiltration, and ultracentrifugation, are commonly used to measure plasma protein binding. However, chromatography‐based methods, such as high‐performance affinity chromatography (HPAC), offer advantages in terms of simplicity, specificity, and speed. In the HPAC method, stationary phases immobilized with plasma proteins allow drugs with a higher affinity for the protein to elute more slowly than those with a lower affinity [13, 14, 15].

Drug metabolism studies are pivotal for optimizing lead compounds to achieve favorable pharmacokinetic (PK) and pharmacodynamic (PD) profiles, identifying active metabolites, and minimizing safety risks associated with reactive or toxic metabolites. An ideal drug candidate should exhibit high oral bioavailability (for oral dosing) or adequate aqueous solubility (for intravenous administration), an elimination half‐life compatible with the intended dosing regimen, and balanced clearance pathways [16]. Drug metabolism involves a complex biotransformation process in which drugs are structurally modified into metabolites through the action of various metabolizing enzymes. This process consists of two phases: Phase I, involving oxidation, reduction, and hydrolysis reactions, and Phase II, comprising conjugation reactions. The liver serves as the primary site of drug metabolism, with hepatic microsomes (subcellular structures derived from the endoplasmic reticulum) playing a critical role due to their abundance of drug‐metabolizing enzymes. Microsomes from human or animal sources are extensively used in drug discovery to investigate metabolic pathways and evaluate potential drug interactions [17].

Recent efforts in drug discovery emphasize early pharmacokinetic assessment to predict in vivo performance and reduce late‐stage attrition. Although thiourea derivatives of naproxen have demonstrated both anti‐inflammatory and anticancer activities, no comprehensive pharmacokinetic evaluation of these compounds has been conducted. The introduction of an aryl‐substituted thiourea moiety can significantly influence pharmacokinetic properties in ways that cannot be reliably predicted without experimental validation. This modification increases both the number of hydrogen bond‐forming atoms and the overall lipophilicity of the molecule. Understanding how thiourea substitution affects passive permeability, serum albumin binding, and microsomal stability is essential for translating these promising scaffolds into viable therapeutic leads. The present study, therefore, addresses this unmet need by providing a detailed in vitro pharmacokinetic profiling of naproxen thiourea derivatives. This approach aims to bridge the gap between their chemical structure and drug‐like behavior.

The aim of this study was to estimate the in vitro pharmacokinetic properties of thiourea derivatives of naproxen previously synthesized in our laboratory, which have demonstrated anti‐inflammatory and anticancer activities [18, 19]. The potential for passive gastrointestinal absorption of the investigated compounds was assessed using the PAMPA model. Based on the experimental data, quantitative structure‐permeability relationship (QSPR) models were developed to support the design of derivatives with improved gastrointestinal absorption. Drug interactions with HSA were evaluated using HPAC, while incubation with CD‐1 mouse liver microsomes was performed to assess microsomal stability.

2. Results and Discussion

2.1. Evaluation of Passive Gastrointestinal Absorption

Passive gastrointestinal absorption of thiourea derivatives of naproxen and the parent drug was evaluated using the PAMPA model with a hexadecane/hexane mixture as an artificial membrane. This model has several advantages, such as the ability to determine membrane retention (R), as well as a short incubation time of 5 h. Membrane retention is important to determine, since it can have a significant impact on the permeability [20, 21].

The primary parameter that describes permeability across the artificial membrane is the Pe value, which represents the effective passive permeability coefficient. For clarity of discussion, Pe ​is expressed as a ‐logPe value, where lower values indicate a higher potential for passive gastrointestinal absorption. Additional parameters, such as membrane retention (R) and transport percentage (%T), further describe the compound behavior in the hexadecane/hexane model system. The results of PAMPA parameters for the tested compounds are outlined in Table 1.

TABLE 1.

Parallel artificial membrane permeability assay (PAMPA) permeability and retention parameters of 14 previously synthesized compounds (1–14) and naproxen. Data are presented as mean ± standard deviation (SD), based on three independent measurements (triplicates). Pe denotes the effective permeability coefficient (cm/s), %T represents the percentage of transport through the artificial membrane, and C_A(t)/C_D(0) indicates the ratio between the compound concentration in the acceptor well at time t (µM) and its initial concentration in the donor well at time 0 (µM). p‐Values (two‐sided Student's t‐test) indicate the statistical significance of the difference between each derivative and naproxen (p < 0.05 indicates a statistically significant difference).

Compound	CA(t)/CD(0)	Ре (·10−6)	‐logPe	%T
1	0.11 ± 0.01	40.14 ± 3.10	4.40 ± 0.04 (p = 0.0027)	10.83 ± 1.09
2	0.11 ± 0.01	46.52 ± 6.92	4.34 ± 0.07 (p = 0.0046)	10.92 ± 0.66
3	0.11 ± 0.03	53.08 ± 3.65	4.28 ± 0.03 (p = 0.0134)	10.76 ± 2.55
4	0.15 ± 0.01	74.78 ± 6.54	4.13 ± 0.04 (p = 0.0859)	15.02 ± 1.07
5	0.12 ± 0.00	44.78 ± 2.27	4.35 ± 0.02 (p = 0.0096)	11.92 ± 0.35
6	0.11 ± 0.03	67.74 ± 17.24	4.18 ± 0.11 (p = 0.1050)	10.53 ± 2.92
7	0.13 ± 0.01	115.21 ± 9.63	3.94 ± 0.04 (p = 0.2291)	13.03 ± 0.76
8	0.10 ± 0.04	44.72 ± 7.37	4.35 ± 0.07 (p = 0.0041)	9.89 ± 3.75
9	0.11 ± 0.00	39.38 ± 1.56	4.41 ± 0.02 (p = 0.0080)	10.89 ± 0.38
10	0.11 ± 0.00	38.93 ± 1.44	4.41 ± 0.02 (p = 0.0081)	10.78 ± 0.35
11	0.11 ± 0.30	50.30 ± 4.02	4.30 ± 0.04 (p = 0.0088)	10.71 ± 3.00
12	0.12 ± 0.01	43.72 ± 2.61	4.36 ± 0.03 (p = 0.0077)	11.90 ± 0.65
13	0.07 ± 0.02	29.62 ± 0.43	4.53 ± 0.01 (p = 0.0060)	6.92 ± 1.58
14	0.10 ± 0.00	34.17 ± 0.48	4.47 ± 0.01 (p = 0.0078)	9.60 ± 0.12
Naproxen	0.18 ± 0.03	98.63 ± 15.70	4.01 ± 0.07	18.09 ± 3.44

The ‐logPe values of the tested compounds ranged from 3.94 to 4.53, and all of them exhibited significant potential for passive gastrointestinal absorption compared to naproxen (‐logPe = 4.01). All compounds for which logPe values determined in the hexadecane/hexane PAMPA model were greater than ‐4.70 (i.e., ‐logPe < 4.70) are expected to exhibit a fraction of absorption higher than 30% and are therefore considered to have a high passive gastrointestinal absorption potential [21]. Derivatives 13 and 14 exhibited the highest ‐logPe values, indicating the lowest passive gastrointestinal absorption, while derivatives 4, 6, and 7 showed the lowest ‐logPe values, suggesting the highest passive gastrointestinal absorption. Based on the presented results, it can be concluded that the highest permeability, comparable to naproxen, was observed for derivatives of m‐methoxyaniline (4), L‐phenylalanine ethyl ester (6), and N‐methyl‐L‐tryptophan methyl ester (7), whereas the lowest permeability was recorded for derivatives of L‐phenylalanine (13) and L‐tryptophan methyl esters (14). Therefore, the design of novel thiourea derivatives of naproxen should focus on the incorporation of ethyl or higher alkyl esters of aromatic amino acids, as well as m‐methoxyaniline‐based moieties, to enhance passive gastrointestinal absorption. Previous findings have shown that derivatization of naproxen with hydroxamic acid groups leads to a reduction in passive gastrointestinal absorption, as evidenced by the higher ‐logPe value of the hydroxamic acid derivative (3.81) compared to that of naproxen (3.46). These results further underscore the sensitivity of passive permeability to structural modifications at the carboxyl group of naproxen [22].

Membrane retention values ​​ranged from 13% to 34%, except for derivatives 1, 5, 7, 9, 10, 12, and 14, whose retention values ​​were below 5%. Derivatives 3, 6, and 11 showed the highest retention within the artificial membrane (24%, 34%, and 32%, respectively), which limited their permeability and prevented them from achieving even lower ‐logPe values.

According to the Biopharmaceutics Classification System, naproxen belongs to Class II, characterized by high permeability and low solubility [23]. Chemical modification of naproxen did not substantially improve this profile, as the thiourea derivatives showed permeability properties similar to the parent drug.

The PAMPA model employed in this study is designed to simulate passive transcellular diffusion across the gastrointestinal lipid membrane. It is therefore highly suitable for comparative assessment of intrinsic permeability among structurally related compounds. However, this in vitro system does not capture the contribution of active transport mechanisms or efflux mediated by intestinal transporters such as P‐gp (P‐glycoprotein), BCRP (Breast Cancer Resistance Protein), or MRP2 (Multidrug Resistance‐Associated Protein 2), which are known to influence the absorption of many NSAIDs, including naproxen. Consequently, the reported logPe values should be interpreted as an indicator of passive permeability potential rather than a predictor of total in vivo absorption. To address this limitation, future studies should include cell‐based models such as Caco‐2 monolayers or MDCK‐MDR1 assays to better delineate the contribution of active transport to the intestinal uptake of selected derivatives. Evaluating passive permeability, the factors affecting it, and potential strategies for its optimization are essential in the early stages of drug development. Enhanced passive absorption can substantially contribute to the overall gastrointestinal drug uptake.

2.2. QSPR Analysis Results

The QSPR study was conducted with the aim of examining the relationship between the ‐logPe values ​​obtained from the PAMPA test (dependent variable) and the calculated molecular descriptors (independent variables). Partial least squares (PLS), ordinary least squares (OLS), and support vector machine (SVM) models were developed. The statistical parameters of the resulting QSPR models are shown in Table 2.

TABLE 2.

Statistical parameters of quantitative structure‐permeability relationship (QSPR) models developed using partial least squares (PLS), ordinary least squares (OLS), and support vector machine (SVM) methods. Model performance was assessed using root mean squared error of estimation (RMSEE), root mean squared error of prediction (RMSEP), correlation coefficient (r), coefficient of determination (R²), and cross‐validated coefficient of determination (Q²) values. Descriptors that form the created QSPR models are presented in regression equations.

Model	Regression equation	RMSEE	RMSEP	Q2 test	r	R2	Q2 training
OLS(‐logPe)	‐logPe = 4,469 + 0,063·F06[C‐C] ‐0,637·WHALES100_Rem	0.080	0.166	−1.685	0.333	0.111	0.483
PLS(‐logPe)	‐logPe = f(RDCHI, VE2_D/Dt, SM4_Dz(i), Wi_B(e), MATS6e, GGI3, SpMax4_Bh(m), SpMax4_Bh(i), P_VSA_p_2, Chi1_AEA(bo), SM06_AEA(dm), Eig07_EA(ri), Eig10_AEA(ed), RDF040v, RDF155p, E3v, L2p, ISH, H7m, HATS2e, R1p+, SP19, SsF, NssO, gmin, F05[C‐S], CATS3D_03_DL, CATS3D_10_AA, CATS3D_02_AL, s34_phRelSize)	0.055	0.072	0.495	0.747	0.559	−4.674
SVM(‐logPe)	‐logPe = f(SM07_EA(dm), CATS3D_02_AL)	0.077	0.069	0.536	0.956	0.914	0.543

A total of 30 descriptors were selected for the PLS (‐logPe) modeling, namely RDCHI, VE2_D/Dt, SM4_Dz(i), Wi_B(e), MATS6e, GGI3, SpMax4_Bh(m), SpMax4_Bh(i), P_VSA_p_2, Chi1_AEA(bo), SM06_AEA(dm), Eig07_EA(ri), Eig10_AEA(ed), RDF040v, RDF155p, E3v, L2p, ISH, H7m, HATS2e, R1p+, SP19, SsF, NssO, gmin, F05[C‐S], CATS3D_03_DL, CATS3D_10_AA, CATS3D_02_AL, and s34_phRelSize. With the SVM algorithm implemented in the alvaModel software, the hyperplane was found by solving a set of linear equations instead of the convex quadratic programming problem, as is the case with classical SVM. The descriptors forming this model were SM07_EA(dm) and CATS3D_02_AL, while the OLS model was constructed using the descriptors F06[C‐C] and WHALES100_Rem. The PLS(‐logPe) and OLS(‐logPe) models should not be taken into further consideration since they did not pass all validation tests (Q² = ‐4.674 and 0.483, respectively). The calculated parameters showed that the SVM model is the most robust and reliable (Q²> 0.5 and R²> 0.5), so the descriptors that form it were interpreted in detail.

SM07_EA(dm) belongs to the group of edge adjacency index descriptors based on distance moments. These molecular descriptors are calculated from the adjacency matrix of the molecular edges. Such a matrix is derived from a molecular graph in which hydrogen atoms are removed, encoding the connection between the edges of the formed graph. It is a square symmetric matrix of dimensions BxB, where B is the number of bonds between pairs of non‐hydrogen atoms [24]. According to its definition, this descriptor represents the seventh‐order spectral moment of the edge adjacency, mathematically calculated by the dipole moment [25]. Descriptor SM07_EA(dm) quantifies the structural information of a molecule by focusing on the molecular topology and edge connectivity. In the molecular graph of this descriptor, atoms are represented as nodes and bonds as edges. The “dm” (distance moment) in the descriptor indicates that it considers distances between pairs of atoms across the molecular structure, weighted according to specific properties. This descriptor provides parameters that reflect the molecular shape, size, and branching by encoding information regarding the connectivity and atomic distances. [26] The analysis of SM07_EA(dm) descriptor values across tested compounds revealed a predominant value of 5.024 that is characteristic of aromatic amine derivatives (compounds 1–5 and 8–12). The value of this descriptor for aromatic amino acid ester derivatives (compounds 6, 7, 13, and 14) was 6.500, while for naproxen it was 6.050. The consistent presence of SM07_EA(dm) descriptor values in groups of aromatic amine derivatives and aromatic amino acid ester derivatives implies that compounds of these subsets of molecules share similar structural and electronic characteristics. Namely, the SM07_EA(dm) descriptor is associated with the electronic properties of the molecule and indicates its electron density and ability to gain an electron [27]. The presence of heteroatoms (oxygen and nitrogen) in aromatic amino acid ester derivatives contributes to significantly higher values of the analyzed descriptor compared to both aromatic amine derivatives and the parent drug.

CATS3D_02_AL is a member of the CATS3D (Chemically Advanced Template Search in 3D) group of descriptors. [28] CATS descriptors belong to the class of atomic pair topological descriptors. These descriptors do not characterize atoms by chemical element type. Instead, atoms are assigned to specific types of potential pharmacophore points. CATS3D descriptors describe the distribution of potential pharmacophore points in three‐dimensional space within a defined distance threshold, typically around 2 Å. The distances between all pairs of pharmacophore points are calculated, and the frequencies of the pairs belonging to the defined ranges are expressed as a correlation vector. [29] The CATS3D_02_AL descriptor links information about the lipophilicity of the molecule and the number of hydrogen bond acceptors. The CATS3D_02_AL descriptor, therefore, reflects how hydrogen bond acceptor and lipophilic features in a molecule are distributed within this specific spatial range. [30] The values ​​of the CATS3D_02_AL descriptor for the investigated compounds ranged from 3 to 7. The highest values of this descriptor were calculated for derivatives 3 and 5, while the lowest values were observed for derivatives 10 and 14. Based on the above, it can be concluded that the highest values ​​of CATS3D_02_AL were observed for derivatives of aromatic amines characterized by high lipophilicity. These derivatives contain two carbon moieties with one oxygen atom acting as a hydrogen bond acceptor in the para position of the benzene ring, namely an ethoxy group in derivative 3 and an acetyl group in derivative 5.

The QSPR analysis of thiourea derivatives of naproxen suggests that aromatic amino acid ester derivatives possess a higher potential for passive gastrointestinal absorption compared to aromatic amine derivatives. This absorption potential can be further enhanced by increasing electron density through the incorporation of heteroatom‐containing groups (nitrogen and oxygen) within their molecular structure. The introduction of hydrogen bond‐accepting functional groups at the para position may also improve the passive gastrointestinal absorption potential of aromatic amine thiourea derivatives of naproxen.

2.3. High‐Performance Affinity Chromatography

A broad spectrum of structurally different drugs can interact with HSA because it has multiple binding sites. NSAIDs demonstrate a particularly strong affinity for HSA, with binding rates reaching up to 99%. This high binding capacity can significantly influence the pharmacokinetics of concurrently administered drugs by displacing them from HSA, thereby increasing their free concentration in plasma. This displacement can enhance drug effects or toxicity, as observed in the interactions of NSAIDs with warfarin, where increased free warfarin levels heighten the risk of bleeding complications. Additionally, this type of drug‐drug interaction can affect the rate of drug metabolism, as a higher concentration of the drug becomes more accessible to enzymes [31, 32].

HPAC is a widely used technique for the evaluation of interactions between potential drug candidates and HSA [33]. The retention behavior of tested compounds on the HSA‐modified stationary phase was used to determine the degree of their binding. The value of the retention factor (k) was used for the evaluation of the tested compound‐HSA interaction, and those results are presented in Table 3.

TABLE 3.

Retention factor (k) values of 14 previously synthesized compounds (1–14) and naproxen obtained by the high‐performance affinity chromatography (HPAC), presented as mean ± standard deviation (SD) based on n = 2 measurements (duplicates). p‐Values (two‐sided Student's t‐test) indicate the statistical significance of the difference between each derivative and naproxen (p < 0.05 indicates a statistically significant difference).

Compound	Retention factor (k)	p‐value
1	62.84 ± 0.52	0.00324
2	35.22 ± 0.54	0.00243
3	20.08 ± 0.62	0.00243
4	33.23 ± 1.57	0.00694
5	9.54 ± 0.01	0.00002
6	12.26 ± 0.01	0.00004
7	11.66 ± 0.30	0.00109
8	109.17 ± 1.30	0.02253
9	102.22 ± 3.00	0.04112
10	57.43 ± 2.02	0.01173
11	24.03 ± 1.62	0.00659
12	> 135	/
13	7.63 ± 0.05	0.00019
14	10.35 ± 0.01	0.00005
Naproxen	> 135	/

For all tested compounds, the k values were in the range of 7.63–109.17, except for derivative 12 and naproxen, which showed the strongest HSA binding interaction, with k values higher than 135. Based on the results presented in Table 3, it can be concluded that thiourea derivatives of naproxen with esters of aromatic amino acids (compounds 6, 7, 13, and 14) exhibited weaker binding affinity for HSA compared to derivatives of aromatic amines. However, in addition to ester derivatives, compound 5, which belongs to the group of aromatic amine derivatives, displayed the second lowest k value. In line with this observation, it seems that the presence of any polar group in the side chain of aromatic amines reduces the k value. Conversely, the absence of a polar substituent within the aromatic amine side chain increases the k value, as evident from derivative 12, which demonstrated comparable binding affinity for HSA as the control drug naproxen. Therefore, a high k value and a high potential for drug‐drug interactions at the HSA level result either from high lipophilicity, as observed in aromatic amine derivatives lacking polar groups, or from the presence of a carboxyl group, as in naproxen.

Previously published research has shown that commercially available non‐steroidal anti‐inflammatory drugs such as naproxen, ibuprofen, indomethacin, and diclofenac bind to HSA with significantly higher affinity than their derivatives obtained by modifying the carboxyl group [22]. This phenomenon can be explained by the fact that ibuprofen binds to HSA with a significantly higher affinity in its anionic form compared to its non‐ionized form. Specifically, NSAIDs as weak acids establish an ionic interaction with HSA, which is crucial for their binding affinity [34]. Based on the above‐mentioned observations, we can assume why all tested compounds (except derivative 12) bind more weakly to HSA compared to the parent drug naproxen. Specifically, thiourea derivatives of naproxen do not have an ionization center; they do not have the possibility to form ionic interactions with the binding pocket of HSA.

Excessive molecular polarity may diminish the overall hydrophobic character required for efficient penetration into the nonpolar binding cavity of HSA. This could also explain the weaker binding affinity observed for aromatic amino acid ester derivatives and aniline derivatives bearing polar substituents. Their increased polarity may hinder optimal interactions within the hydrophobic pocket of HSA. For a more detailed understanding of the binding mode, molecular docking or X‐ray crystallographic studies could be employed in future work. In contrast, compounds featuring larger aromatic surfaces can insert more deeply into the hydrophobic pocket, thereby strengthening van der Waals and π‐π stacking interactions and enhancing overall binding affinity. The presence, size, and position of substituents on aromatic rings further modulate these interactions. Bulky or sterically demanding groups may impede optimal fitting within the pocket, leading to reduced affinity even when key polar functionalities, such as carboxylate groups, remain accessible for hydrogen bonding and electrostatic interactions. Consequently, optimal HSA binding is typically observed in molecules exhibiting a balanced distribution of polar and nonpolar regions. Classic examples include diclofenac and naproxen. Their moderately hydrophobic aromatic cores, complemented by one or two peripheral polar substituents, enable an efficient fit within the binding site and provide favorable enthalpic and entropic contributions to complex formation [30, 35, 36].

2.4. Microsomal Stability

Microsomal stability (i.e., susceptibility to metabolism) was evaluated using microsomes derived from CD‐1 mouse liver. Although microsomal stability can vary depending on the species origin of the enzymes, the use of mouse liver microsomes is well justified in the early phases of drug discovery. Numerous studies have demonstrated a reasonably high degree of similarity between mouse and human microsomal metabolic profiles under early screening conditions. For example, comparable metabolism of 5F‐APINACA has been observed in CD‐1 mouse and human liver microsomes [37]. Moreover, preliminary high‐throughput absorption, distribution, metabolism, and excretion (ADME) workflows commonly employ non‐human microsomes (e.g., rodent) as Tier I assays to prioritize compounds before subsequent evaluation in human microsomes [38, 39]. Therefore, the use of mouse microsomes in this study is consistent with standard early‐stage screening practices, whereas confirmatory studies with human liver microsomes or hepatocytes could be considered in future research.

The highest decrease in peak area (and the highest metabolic rate) was recorded for compound 7 (65.83% decrease). This compound contains a methyl ester of N‐methyl‐L‐tryptophan in its side chain. Since esters are generally highly susceptible to enzymatic hydrolysis, this structural feature likely accounts for the observed result. Compounds 3 and 4 exhibited considerably lower decreases in peak area (14.76% and 25.56%, respectively), indicating higher microsomal stability. The decrease in peak area for positive control propranolol was 48.48%.

Overall, the markedly lower microsomal stability of compound 7 can be attributed primarily to its ester functionality, as esters are prone to rapid degradation in liver microsomes through enzymatic hydrolysis. Such structural features may lead to a rapid degradation of the parent compound in the presence of liver microsomal enzymes, thereby reducing its stability and potential bioavailability.

In a previously published study investigating the microsomal stability of selected NSAIDs and their derivatives, it was demonstrated that N‐hydroxyurea and hydroxamic acid derivatives exhibit a markedly lower degree of microsomal stability compared to their parent compounds [22]. This reduced stability was attributed to the increased susceptibility of these functional groups to metabolism, leading to faster biotransformation and clearance. These findings highlight the significant impact that specific structural modifications can have on metabolic behavior and emphasize the importance of early microsomal stability assessment during the optimization of NSAID derivatives.

3. Experimental

3.1. Chemicals

S‐naproxen, TWEEN80, microsomes from liver (pooled, from male (CD‐1) mice, catalogue number: M9441), NADPH tetrasodium salt, and uridine 5'‐diphosphoglucuronic acid trisodium salt (UDP‐GlcUA) were purchased from Sigma‐Aldrich (St. Louis, MO, USA), while sodium dihydrogen phosphate monohydrate (NaH₂PO₄·H₂O), disodium hydrogen phosphate (Na₂HPO₄), and orthophosphoric acid (85%) were acquired from Merck (Darmstadt, Germany). Hexane, sodium hydroxide (NaOH), ammonium acetate, acetonitrile (high‐performance liquid chromatography [HPLC] gradient grade), and dimethyl sulfoxide (DMSO) were purchased from Fisher (Loughborough, UK), while hexadecane was obtained from Acros Organics (Geel, Belgium). Deionized water was prepared using a TKA water‐purification system (Niederelbert, Germany).

Thiourea derivatives of naproxen were synthesized as previously described [18, 19]. Their chemical structures are presented in Figure 1.

FIGURE 1.

Chemical structures of previously synthesized 14 thiourea derivatives of naproxen (compounds 1–14). The pharmacokinetic properties of these derivatives were evaluated in vitro in the present study.

3.2. PAMPA Test

The membrane solution was prepared by mixing hexadecane and hexane (5:95, v/v). Phosphate buffer solutions (pH 5.5 and 7.4) were prepared by dissolving 0.5292 g NaH₂PO₄·H₂O and 0.0234 g Na₂HPO₄ in purified water (500 mL). This solution was divided into two parts, after which the pH value of one part of the solution was adjusted to 5.5 using orthophosphoric acid, while the pH value of the other part was adjusted to 7.4 by adding NaOH solution. Test solutions (50 µM) were prepared by diluting stock solutions of tested compounds and naproxen in DMSO (5 mM). One drop of TWEEN80 (0.2%) was added to 100 µl of each test solution and supplemented to 10 mL with phosphate buffer solution pH 5.5. The acceptor solution consisted of 5% DMSO in phosphate buffer (pH 7.4) containing 0.2% TWEEN80.

The PAMPA assay was conducted using 96‐well polycarbonate filter plates (Merck Millipore, Carrigtwohill, Ireland), according to a previously published protocol [40, 41]. Each well of the plate was impregnated with 15 µl of hexadecane/hexane (5:95, v/v) solution and allowed to dry for 20 min at room temperature. A 300 µl portion of 5% DMSO solution in phosphate buffer (pH 7.4) was added to each well of the acceptor plate (lower plate). Then, the acceptor plate was overlapped with the donor plate (upper plate) to create a PAMPA sandwich, and 300 µl of donor solution was transferred to the wells of the donor plate (Figure 2). Each compound was tested in triplicate. The system was covered to prevent the evaporation of the solvent from donor wells, after which it was placed on a vibratory mixer and incubated at room temperature for 5 h.

FIGURE 2.

Schematic representation of the parallel artificial membrane permeability assay (PAMPA) enabling quantitative evaluation of passive permeability through an artificial membrane. Polycarbonate membrane (96‐well format) was impregnated with hexadecane/hexane mixture (5:95, v/v) to form an artificial membrane separating the donor and acceptor compartments. 300 µL of test solutions (compounds dissolved in 5% DMSO/phosphate buffer pH 5.5, with addition of TWEEN80) were added into the upper (donor) plate, while the corresponding acceptor wells contained 300 µL of 5% DMSO/phosphate buffer pH 7.4, with the addition of TWEEN80. The system was incubated for 5 h at room temperature. All compounds were tested in triplicate (n = 3). DMSO—dimethyl sulfoxide, TWEEN80 – Registered trademark of polyoxyethylene (80) sorbitan monooleate.

3.2.1. HPLC Method

Concentrations of the analyzed compounds and naproxen in the starting solutions, donor wells, and acceptor wells were determined using HPLC with diode‐array detection (HPLC‐DAD). The analysis was performed on a Dionex Ultimate 3000 chromatograph (Thermo Fisher Scientific, Waltham, MA, USA), equipped with a Dionex Ultimate 3000 quaternary pump, autosampler, and DAD detector. The mobile phase composition was acetonitrile/water 75:25, v/v for the tested compounds and 50:50, v/v for naproxen. The analysis was performed using a ZORBAX Eclipse Plus C8 column (150 × 4.6 mm, 5 µm particle size). The column temperature was set to 25°C, and the mobile phase flow rate was set to 1 mL/min. Detection was performed at 230 nm for the tested compounds and 280 nm for naproxen. Validation of the method was performed according to the ICH guidelines [42], and results are presented in the Supporting Information.

3.2.2. Calculation of PAMPA Parameters

Using Equations (1)–(3), values of membrane retention factor (R), permeability coefficients (logPe), transport percentage (T%), and permeation parameters (C_A(t)/C_D(0)) of the tested compounds were calculated.

$$\mathrm{R}=1-\frac{{\mathrm{C}}_{\mathrm{D}}\left(\mathrm{t}\right)}{{\mathrm{C}}_{\mathrm{D}}\left(0\right)}-\frac{{\mathrm{V}}_{\mathrm{A}}{\mathrm{C}}_{\mathrm{A}}(\mathrm{t})}{{\mathrm{V}}_{\mathrm{D}}{\mathrm{C}}_{\mathrm{D}}(0)}$$ (1)

$$\begin{array}{c}\hfill \mathrm{logPe}=\log \left[-\frac{2.303{\mathrm{V}}_{\mathrm{D}}}{\mathrm{A}(\mathrm{t}-{\mathrm{t}}_{\mathrm{LAG}})}\left(\frac{{\mathrm{V}}_{\mathrm{A}}}{{\mathrm{V}}_{\mathrm{A}}+{\mathrm{V}}_{\mathrm{D}}}\right)\log \left[1-\left(\frac{{\mathrm{V}}_{\mathrm{A}}+{\mathrm{V}}_{\mathrm{D}}}{{\mathrm{V}}_{\mathrm{D}}(1-\mathrm{R})}\right)\frac{{\mathrm{C}}_{\mathrm{A}}(\mathrm{t})}{{\mathrm{C}}_{\mathrm{D}}(0)}\right]\right]\end{array}$$ (2)

$$\%\mathrm{T}=100·\frac{{\mathrm{A}}_{\mathrm{A}}·{\mathrm{V}}_{\mathrm{A}}}{{\mathrm{A}}_{\mathrm{D}}(0)·{\mathrm{V}}_{\mathrm{D}}}$$ (3)

V_A—the volume in the acceptor wells (mL); V_D—the volume in the donor wells (mL); A—the filtration area (cm²); t—the incubation time (s); t_LAG—the steady‐state time (s); C_D(t)—concentration of the tested compound in the donor well after incubation time (µM); C_A(t)—concentration of the tested compound in acceptor well after incubation time (µM); C_D(0)—concentration of the tested compound in donor well at the beginning of the incubation—starting solution (µM); A_A—peak area of ​​the tested compound in the receiving wells; A_D(0)—peak area of ​​the tested compound in the donor wells before incubation—starting solution [40].

Steady‐state time (t_LAG) represents the time required to saturate the membrane with the tested substance in the PAMPA system, which is relatively short compared to the incubation time (typically about 20 min for PAMPA experiments performed without shaking on a vibratory shaker). The concentration of the tested substance in the donor well at the beginning of the incubation (C_D(0)) corresponds to the concentration of the compound in the initial solution [43].

To further support the interpretation of permeability data, statistical analysis was performed using the Student's two‐sided t‐test in Microsoft Excel 2016 (Microsoft Corporation, Redmond, WA, USA). p‐Values were calculated to identify compounds whose permeability values did not differ significantly from that of naproxen, indicating that these derivatives exhibit similar membrane permeability.

3.3. QSPR Analysis

Geometric optimization of all tested compounds was initially performed using the semi‐empirical PM3 method, followed by the Hartree‐Fock method with the 3‐21G basis set [44]. A total of 5666 descriptors were calculated using alvaDesc software [45]. Prior to model development, descriptor selection was performed in alvaModel software [46] on the entire dataset of tested compounds, without separating them into training and test sets. For this purpose, the full set of descriptors was screened over 10 000 iterations using PLS, OLS, and SVM methods. As a result, sets of descriptors with the strongest influence on ‐logPe were selected: 10 descriptors for OLS and SVM models and 50 descriptors for the PLS model. Subsequently, the tested compounds were divided into training and test sets. The test set consisted of four derivatives (1, 3, 4, and 11), selected to ensure a uniform distribution of ‐logPe values across all compounds. The remaining compounds, along with naproxen, comprised the training set. Finally, alvaModel software was used to develop PLS(‐logPe), OLS(‐logPe), and SVM(‐logPe) models [46, 47].

The quality of the developed QSPR models was assessed using the following statistical parameters: RMSEE, RMSEP, F ratio, p‐value, correlation coefficient (r), coefficient of determination (R²), and cross‐validated coefficient of determination (Q²). RMSEE was computed for the training set, whereas the RMSEP was calculated for the test set. The Q² parameter, an internal validation metric calculated using Leave‐One‐Out (LOO) cross‐validation, reflects the model's predictive performance for compounds similar to those in the training set. In this procedure, each compound in the training set is excluded one at a time to build a model with the remaining compounds, which then predicts the dependent variable of the excluded compound. The process is repeated until each compound in the training set has been excluded once. Q² is calculated using Equation (4), where Y_training ​is the mean value, and Y_{obs(training)}​ is the experimentally determined ‐logPe value for the training set [48].

After completing the LOO procedure, the Predicted Residual Error Sum of Squares (PRESS) is calculated using Equation (5), where e_(i) represents the difference between the experimentally obtained and predicted logPe values [48, 49].

$${\mathrm{Q}}^2=1-\frac{\mathrm{PRESS}}{\sum {({\mathrm{Y}}_{\mathrm{obs}\left(\mathrm{training}\right)}-{\overline{\mathrm{Y}}}_{\mathrm{training}})}^2}$$ (4)

$$\mathrm{PRESS}=\stackrel{\mathrm{n}}{\sum _{\mathrm{i}-1}}{\mathrm{e}}_{(\mathrm{i})}^2$$ (5)

Models with strong predictive capabilities typically have Q² values higher than 0.5.

The R² parameter, an external validation metric, evaluates the model's predictive accuracy for compounds structurally distinct from those in the training set. This parameter is calculated using Equation (6), where Y_obs(test) is the experimentally determined value of ‐logPe for compounds from the test set [50].

$${\mathrm{R}}_{\mathrm{pred}}^2=1-\frac{\mathrm{PRESS}}{\sum {({\mathrm{Y}}_{\mathrm{obs}\left(\mathrm{test}\right)}-{\overline{\mathrm{Y}}}_{\mathrm{training}})}^2}$$ (6)

An R² value exceeding 0.5 indicates high predictive potential for structurally diverse compounds [48, 49, 51]. The correlation coefficient r reflects the degree of correlation between predicted and observed ‐logPe values for the test set [49].

3.4. High Performance Affinity Chromatography

Chromatographic analysis was performed using an Agilent 1200 Series chromatograph (Agilent Technologies, Santa Clara, CA, USA) equipped with a binary pump, a manual injector (20 µL sample loop), and a DAD detector. The retention behavior of the tested compounds and naproxen was investigated on an immobilized CHIRALPAK HSA HPLC column (4 × 100 mm, 5 µm particle size) purchased from Daicel Chiral Technologies (Illkirch‐Graffenstaden, France). The column temperature was maintained at 25°C, and the flow rate was set to 0.9 mL/min. UV detection was performed at 230 nm for the tested compounds and 280 nm for naproxen. The mobile phase consisted of a 10 mM ammonium acetate buffer (pH 7) and acetonitrile (80:20, v/v). Tested compounds were dissolved in DMSO (5 mM) and further diluted with the mobile phase to obtain working solutions (0.2 mg/mL). All analyzed solutions were injected in duplicate. The HPAC conditions were selected to optimize the retention characteristics of all synthesized compounds while adhering to the column manufacturer's recommendations. Retention factors (k) were calculated using Equation (7), where t₀​ (the column dead time) corresponds to the retention time of DMSO at 220 nm, and t_R represents the retention time of the tested compound.

$$\mathrm{k}=\frac{({\mathrm{t}}_{\mathrm{R}}-{\mathrm{t}}_0)}{{\mathrm{t}}_0}$$ (7)

To assess the statistical significance of differences in HSA‐binding affinity, p‐values were determined using the Student's two‐sided t‐test in Microsoft Excel 2016 (Microsoft Corporation, Redmond, WA, USA). This analysis allowed the identification of compounds that differed significantly from naproxen, specifically those exhibiting markedly lower binding constants (k), indicating reduced interaction with HSA.

3.5. Microsomal Stability

CD‐1 mouse liver microsomes (catalogue number: M9441; Sigma Aldrich, Darmstadt, Germany) were used to estimate the microsomal stability of the synthesized compounds, following the protocol described by Knights and co‐workers [52]. NADPH and UDP‐GlcUA were used as cofactors to activate the first (oxidation, reduction, and hydrolysis) and second (conjugation) phases of metabolism, respectively. Microsomal stability was evaluated for the most promising drug candidates—compounds exhibiting the strongest anti‐inflammatory (compounds 4 and 7) and cytotoxic (compound 3) potential [18, 19]. Propranolol, an antihypertensive agent, was used as a positive control according to the protocol. Test solutions (0.1 mM) were prepared by diluting stock solutions of the tested compounds in DMSO (5 mM) with 0.1 M potassium phosphate buffer. The compounds were incubated at 37°C for 30 min, with both metabolic phases activated by the addition of corresponding cofactors. Incubation was halted, and peak areas were measured for the best drug candidates (derivatives 3, 4, and 7), as well as for the control drug propranolol, immediately after cofactor addition (t₀) and after the 30‐min incubation period by mixing 100 µL of each incubation mixture with 200 µL of cold acetonitrile. All compounds were analyzed in duplicate at two time points (0 and 30 min), and their concentrations were determined using the HPLC‐tandem mass spectrometry (HPLC‐MS/MS) method.

3.5.1. HPLC‐MS/MS Method

The HPLC‐MS/MS analysis was performed on a Vanquish Core chromatograph (Thermo Fisher Scientific Inc., Madison, WI, USA) coupled to a TSQ Fortis Plus triple quadrupole mass spectrometer (Thermo Fisher Scientific Inc., Madison, WI, USA) equipped with a heated electrospray ionization interface. Determination was achieved using a Zorbax Extend C18 column (150 × 4.6 mm, 5 µm; Agilent Technologies) maintained at 30°C. The mobile phase consisted of acetonitrile and 0.1% formic acid (80:20, v/v) with a flow rate of 500 µL/min. Positive SRM mode was used to detect and quantify compounds 3 (m/z: 409.2–185.0), compound 4 (m/z: 395.2–185.1), compound 7 (m/z: 504.3–185.2), and propranolol (m/z: 260.2–116.2). However, due to poor ionization, naproxen could not be reliably detected or quantified.

4. Conclusion

This study advances the understanding of the pharmacokinetic behavior of naproxen‐based thiourea derivatives by identifying the structural features that most strongly govern passive permeability, HSA binding, and metabolic stability. The results demonstrate that thiourea substitution can meaningfully modulate drug‐like properties and that even subtle structural variations lead to distinct pharmacokinetic behaviors. Derivatives 4 and 7 exhibited markedly enhanced passive gastrointestinal absorption, comparable to naproxen, highlighting the importance of appropriate thiourea positioning and the surrounding substituent environment. The SVM‐based QSPR model further supported this observation by accurately predicting permeability from structural descriptors. Despite the excellent permeability of compound 7, introducing an esterified amino‐acid moiety increased metabolic vulnerability and diminished HSA binding affinity. These insights suggest that more hydrolysis‐resistant ester groups (such as branched or higher alcohol derivatives) may better balance permeability with metabolic stability and HSA binding. In contrast, compound 4 emerged as a promising drug candidate, combining high expected passive gastrointestinal absorption with moderate HSA binding and improved microsomal stability. Its profile closely aligns with that of naproxen while offering potential advantages in metabolic resilience, indicating that certain thiourea substitutions can enhance pharmacokinetic performance without compromising favorable NSAID attributes. Collectively, these findings clarified how specific structural motifs influence key ADME properties in thiourea‐modified NSAIDs and provided a framework for rational design of future derivatives. Nonetheless, the conclusions remain limited by the exclusively in vitro nature of the data and the narrow structural scope examined. Expanding structural diversity and validating these results in in vivo systems will be essential to fully assess the translational potential of these derivatives.

Author Contributions

Nikola Nedeljković: formal analysis, visualization, and writing – original draft. Vladimir Dobričić: conceptualization, data curation, methodology, software, validation, and writing – review and editing. Marina Vesović: formal analysis, visualization, and writing – original draft. Ratomir Jelić: methodology, supervision, and writing – review and editing. Bojan Marković: data curation, methodology, supervision, and writing – review and editing. Zorica Vujić: methodology, supervision, and writing – review and editing. Miloš Nikolić: conceptualization, formal analysis, visualization, validation, and writing – original draft.

Funding

This research was supported by the Science Fund of the Republic of Serbia, 7739840, Utilization of interplay between inflammation and cancer in the development of compounds with anticancer activity—INFCANPLAY. The research was also supported by the Ministry of Science, Technological Development and Innovation, Republic of Serbia, through Grant Agreements with the University of Belgrade—Faculty of Pharmacy (No. 451‐03‐136/2025‐03/200161 and No. 451‐03‐137/2025‐03/200161), the Faculty of Medical Sciences, University of Kragujevac No. 451‐03‐137/2025‐03/200111, and the Faculty of Medical Sciences, University of Kragujevac (Junior Project 11/20).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Supporting File 1: cbdv70865‐sup‐0001‐SuppMat.docx

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.