Computational Insights into the Structural Basis for Reduced Hepatotoxicity of Novel Nonopioid Analgesics

Abstract

Acetaminophen (ApAP) toxicity arises from the reactive intermediate N‐acetyl‐p‐benzoquinone imine (NAPQI), a degradation product known to cause significant liver damage and kidney injury. This toxicity is a major concern associated with the widespread use of ApAP, a commonly used nonsteroidal anti‐inflammatory drug. To address this important issue, a series of novel nonopioid analgesic candidates with reduced toxicity have been recently reported. However, the molecular and atomic‐level mechanisms underlying their decreased toxicity remain largely unexplored. In this study, computational analyses is performed to investigate the dynamic behavior, physicochemical properties, and ligand‐receptor interactions of these new chemical entities (NCEs). The findings provide a rational explanation for their differing toxicity profiles and contribute to a deeper understanding of their metabolic pathways. Based on these insights, compound 6 has emerged as a promising ApAP alternative and is currently under development. These investigations pave the way for designing novel hepatotoxicity‐free NCE analgesics with improved drug metabolism and pharmacokinetic properties.

Docking and molecular dynamics simulations within CYP2E1 clarify the structural basis for the reduced hepatotoxicity of novel non‐opioid analgesics developed as safer alternatives to acetaminophen. The results reveal that hydrophobicity and substituent size critically influence binding conformations, thereby modulating toxicity. Compounds 6 and 6o adopt distinct conformations, which account for their divergent toxicity profiles.

1. Introduction

The treatment of acute and chronic pain remains one of the most prevalent and costly public health challenges worldwide. In the United States alone, pain affects more adults than diabetes, heart disease, and cancer combined,^{[ 1 ]} with an estimated annual cost of $635 billion to the healthcare system.^{[ 2 ]} Similarly, in Europe, chronic pain affects ≈19% of the adult population, significantly impairing quality of life and imposing a substantial economic burden on healthcare services and productivity.^{[ 1 , 3 ]} Current analgesics carry potential risks, including abuse liability (e.g., opioids), hepatotoxicity (e.g., acetaminophen), and nephrotoxicity (e.g., nonsteroidal anti‐inflammatory drugs; NSAIDs). The widespread reliance on opioids—especially following work‐related injuries—and the high overdose rates in the U.S. further emphasize the urgent need for safer, effective, nonopioid analgesics. In addition to NSAIDs, N‐acetyl‐para‐aminophenol (acetaminophen, also known as paracetamol or ApAP) (Figure  1 ) remains the most commonly used over‐the‐counter analgesic and antipyretic worldwide.^{[ 4 ]}

Figure 1.

Chemical structures of ApAP and N‐arachidonoylaminophenol (1).

Interestingly, the comparable efficacy of ApAP and NSAIDs to opioids for the management of moderate to severe acute pain was demonstrated in a recent randomized clinical trial. In this study, 416 patients presenting with moderate to severe acute limb pain were randomized to receive either NSAIDs, ApAP, or an opioid/ApAP combination. Two hours after a single dose, no significant differences were observed in pain reduction among the treatment groups.^{[ 2 ]}

Although ApAP has been used for several decades, its precise mechanism of action remains unidentified.^{[ 5 , 6 ]} However, growing evidence suggests that N‐arachidonoylaminophenol (1) (Figure 1) may play a key role in ApAP's analgesic effects. In the liver, ApAP is metabolized into p‐aminophenol, which subsequently crosses the blood–brain barrier and is converted by fatty acid amide hydrolase into AM404 in the brain.^{[ 7 , 8 ]} AM404 may then exert analgesic effects through the endocannabinoid system by activating cannabinoid receptor type 1 (CB1),^{[ 7 , 9 ]} and by modulating transient receptor potential vanilloid 1 (TRPV1) signaling within the periaqueductal gray matter.^{[ 7 ]} Remarkably, the FAAH/CB1/TRPV1 triad—previously thought to be essential for ApAP‐induced analgesia—is colocalized in this specific brain region.^{[ 7 , 10 ]} Beyond these mechanisms, the central processes underlying ApAP‐induced analgesia remain largely unknown.

However, ApAP has a narrow therapeutic index, and hepatotoxicity is a significant risk associated with both intentional (e.g., suicide attempts) and unintentional overdose, as well as use in patients with compromised liver function. In fact, ≈30 000 patients are hospitalized each year in the United States due to ApAP‐induced hepatotoxicity.^{[ 11 ]} In Europe, data from the study of acute liver transplantation revealed substantial intercountry variability in the incidence of ApAP‐induced acute liver failure leading to liver transplantation. Between 2005 and 2007, the average event rate across seven European countries was one case per 6 million inhabitants per year.^{[ 12 ]} This toxicity is primarily linked to the formation of the electrophilic metabolite N‐acetyl‐p‐benzoquinone imine (NAPQI, compound 2), via oxidative metabolism mediated by cytochrome P450 enzymes (mainly CYP2E1 and CYP3A4). Under normal conditions, NAPQI is detoxified through a Phase II conjugation process involving glutathione (GSH),^{[ 13 ]} and is excreted as a mercapturic acid conjugate (Scheme  1 , Pathway A). In overdose scenarios, this Phase II detoxification capacity becomes saturated, leading to GSH depletion. As a result, NAPQI reacts with nucleophilic macromolecules, initiating a cascade of cellular events that culminates in hepatocellular death (Scheme 1, Pathway B).^{[ 14 , 15 ]}

Scheme 1.

Metabolic pathway of ApAP leading to hepatotoxic byproducts in overdoses.^{[ 14 , 15 ]}

Due to the narrow therapeutic index of ApAP and the clinical demand for safer nonopioid analgesics, Bazán et al. initiated a research effort to develop ApAP analogs devoid of hepatotoxicity.^{[ 16 – 17} , ^{18 ]} They synthesized more lipophilic derivatives by introducing a 1,1‐dioxo‐1,2‐benzothiazol‐3‐one moiety (hereafter referred to as “imide”) into the methyl group of ApAP.^{[ 18 ]} Although increased lipophilicity is often associated with higher hepatotoxicity, the goal was to use steric hindrance and modified electronic properties to block metabolic oxidation and thereby prevent the formation of NAPQI The resulting compound 4 (Table  1 ) retained the in vivo analgesic profile of ApAP while significantly reducing hepatotoxicity; however, compound 4 is rapidly metabolized in vivo into the major metabolite 5. Compound 5 is a comparatively hydrophilic carboxylic acid, facilitating its rapid renal excretion. To modulate the pharmacokinetic profile of compound 4, a novel series of analgesics was described and evaluated.^{[ 17 ]} These compounds were obtained via ring‐opening of the imide moiety in 4, yielding the corresponding N‐substituted amides 6.^{[ 17 ]} The synthesis and characterization of this new class of analogs aimed to identify chemical entities that lack hepatotoxicity while maintaining analgesic and antipyretic properties (Table 1). Compared to compound 2, the new 2‐(benzenesulfonamide)‐N‐(4‐hydroxyphenyl)acetamide analogs exhibited enhanced stability, increased lipophilicity, and slower amide hydrolysis. Nonetheless, despite reporting a series of N‐substituted amides that did not form NAPQI, results clearly indicated that not all N‐substituted compounds were free from such toxicity.^{[ 17 ]} Interestingly, studies demonstrated that compound 1 is primarily metabolized to NAPQI by CYP2E1,^{[ 19 ]} despite the cytochrome P450 system being a highly conserved protein family with a common ancestor.^{[ 20 ]} Furthermore, the catalytic cycle of cytochrome P450 is well established,^{[ 21 ]} and the mechanism leading to NAPQI and other metabolites has been investigated using quantum mechanical approaches.^{[ 22 ]}

Table 1.

Chemical structure of ApAP and compounds 4, 5, 6 and 6a‐6u. Predicted logP values. Mean value and standard deviation of the radius of gyration (Rgyr) measured in Å of all compounds during the 100 ns simulation. Compound toxicity measured as LDH release after 3 h of treatment with 500 μM (See Table S2 and S3, Supporting Information for full toxicity data).^{[ 18 ]}

ApAPlogP = 0.9Rgyr = 2.61 (±0.03)LDH = 0.9627

4logP = 1.3Rgyr = 3.86 (±0.15)

5logP = 1.1Rgyr = 4.10 (±0.30)

Compound	R1	R2	log P	Rgyr	LDH
6	Et	Et	1.4	4.21 (±0.31)
6a	H	H	0.3	4.05 (±0.29)	0.3989
6b	CH6	H	0.5	4.05 (±0.32)	0.4541
6c	CH6	CH6	0.7	4.08 (±0.31)	0.4119
6d	CH6	CH2‐CH2‐OH	0.0	4.22 (±0.32)
6e	H	CH2‐CH2‐CH2‐CH6	0.8	4.32 (±0.37)	0.2024
6f	H	CH(CH6)2	1.2	4.04 (±0.39)	0.2190
6g	H	CH2‐CH(CH6)2	1.7	4.29 (±0.38)	0.2314
6h	H	CH2‐C*HOH‐CH2‐NH2	−1.5	4.32 (±0.37)
6i	H	Cyclopentyl	1.8	4.27 (±0.38)	0.2416
6j	H	Benzyl	2.2	4.45 (±0.38)	0.5099
6k	H	p‐methylbenzyl	2.7	4.40 (±0.44)	1.2540
6l	H	p‐methoxybenzyl	2.0	4.54 (±0.49)	0.6478
6m	H	p‐nitrobenzyl	2.15	4.53 (±0.47)	0.6009
6n	H	m‐chlorobenzyl	2.8	4.47 (±0.44)	0.7708
6o	H	6,4‐dichlorobenzyl	3.4	4.71 (±0.57)	1.4720
6p	H	pyrimidin‐2‐yl‐methyl	1.42	4.24 (±0.32)	0.3413
6q	H	Phenetyl	2.5	4.62 (±0.47)	0.8967
6r	H	6,4‐dihydroxiphenetyl	1.9	4.88 (±0.58)	0.2898
6s	H	2(pyrrolidin‐1‐yl)‐ethyl	0.5	4.16 (±0.37)	0.3915
6t logP = 0.5Rgyr = 4.19 (±0.37)			6u logP = 2.3Rgyr = 4.87 (±0.45)LDH = 0.803

Through comprehensive computational analyses, this study provides a mechanistic rationale that directly correlates the dynamic behavior of ApAP analogs in aqueous solvent and their complexation with CYP2E1 to the experimentally observed differences in toxicity profiles. These insights not only enhance our understanding of their metabolic pathways but also offer a predictive framework to guide the design of new ligand series with improved safety profiles and reduced toxicity.

2. Results and Discussion

The structure of CYP2E1 mainly consists of alpha helices that englobe the heme group that is held in position by the coordination of the sulfur atom of C437 to the Fe⁴⁺ ion and the reinforced ionic bonds and hydrogen bonds established between the two carboxylic acids and the side chains of R100, W122, R126, H370, and R435.

The ligand‐binding site on top of the heme group is made up of the side chains of hydrophobic amino acids L103, F106, I115, P116, F207, F298, L363, V364, L368, and F478. The access to the heme group is blocked by an alpha helix comprising positions 104–109 and linked in each side to the rest of the protein by loops.

Most alpha helices make up the back of the ligand‐binding site, whereas the only existing beta sheets are located at one side of the access of the binding site. The first two normal modes (NM) show a movement of the protein that leads to the opening and closing of the access tunnel to heme cavity and ligand binding‐site (Figure S1, Supporting Information).

The amino acid composition of the access tunnel is slightly more diverse than that of the heme cavity. While the former is made up of mainly bulky hydrophobic amino acids, in this case, smaller hydrophobic as well as hydrophilic amino acids line the access tunnel (Q216, N219, N220, P365, S366, N367, and G477) and the entrance to the cavity (I53, P54, F57, V72, P104, P369, E371, V388, V390, and I476) (Figure S2, Supporting Information).

The logP values for all compounds were predicted by means of Chemicalize [04/2024] (http://www.chemicalize.com), developed by ChemAxon. These results indicate that all compounds fall within a predicted logP range of less than 3.5. However, upon examining their structures and logP values, it was observed that the series of compounds can be classified into three groups based on their lipophilicity: 1) hydrophilic with a logP ranging from −1.5 to 1.9 (ApAP, 4, 5, 6, and compounds 6a‐6i, 6p, 6r, 6s, and 6t); 2) intermediate with a logP ranging from 2.0 to 2.5 (compounds 6j, 6l, 6m, 6q, and 6u); and 3) hydrophobic with a logP ranging from 2.5 to 3.4 (compounds 6k, 6n, and 6o) (Table 1).

Notably, compounds in the hydrophobic group exhibit the highest toxicity, while those in the hydrophilic group are the least toxic (Table 1, S2, and S3, Supporting Information).

Molecular dynamics (MD) simulations in free water solvent highlighted that all compounds have a greater radius of gyration (Rgyr) than ApAP, attributed to the presence of the sulfonamide substituent, which makes them nonlinear with different conformations arising from the rotation around the methylene between the amide and sulfonamide moieties (Table 1). However, a linear correlation between the Rgyr and the toxicity of the compound was not encountered.

However, the main difference amongst all compounds lies in the conformations in free water solvent. Those conformations depend on the substituent attached to the sulfonamide moiety (Table S4, Supporting Information).

Based on their chemical structures, the set of compounds can be divided into three groups depending on the nature of the substituent: small or small‐polar (4, 5, 6, and compounds 6a‐6i), bulky‐hydrophobic (compounds 6k‐6l and 6n‐6q) and bulky‐polar (6m and 6r‐6u).

Interestingly, compounds that present either a small, small‐polar or bulky‐polar sulfonamide substituent present a major conformation in which the sulfonamide substituent moiety remains far from the phenol moiety. Exception to this behavior are compounds 6h and 6s, that are predicted to be protonated at physiological pH, and flexible enough to establish an intramolecular π‐cation interaction between the protonated substituent attached to the sulfonamide moiety and the phenol aromatic ring. However, this π‐cation interaction cannot be established in compound 6u as it lacks the necessary flexibility to bend enough to approach both functional groups.

Compounds with a bulky‐hydrophobic sulfonamide substituent adopt a major conformation in which the compound is folded and establishes intramolecular π–π stacking interactions between the phenol and the group attached to the sulfonamide moiety. Exceptions to this behavior are compound 6m that despite having a p‐nitrophenol moiety, presents a folded conformation probably due to the electrostatic interaction brought about by the electron withdrawing characteristics of the nitro group, and compound 6q, which exceptionally presents an ethylene linker between the sulfonamide and the aromatic ring that increases length and flexibility and prevents the π‐π stacking interaction to take place, thus leading to a nonfolded major conformation.

It is noteworthy that bulky‐hydrophobic compounds that present a major folded conformation such as compounds 6k, 6l, 6m and 6o are the most toxic, whereas small, small‐polar, and bulky‐polar compounds that present a nonfolded conformation are the least toxic. Compound 6m is an exception due to the hydrophilic nature of the nitro moiety, which prevents it from accessing the hydrophobic pocket and stabilizing binding to CYP2E1.

Although a pattern of behavior is already apparent upon analysis of the dynamic behavior in water of the set of compounds previously reported by us,^{[ 16 , 18 ]} their binding to CYP2E1 was studied by means of MD simulations in water. Thus, three of them were chosen to represent each group and their interaction with CYP2E1: 6 (small and small‐polar), 6o (bulky‐hydrophobic and the most toxic of the described compounds), and 6r (bulky‐polar and nontoxic albeit with a very similar structure to 6o). For comparison purposes and to give a plausible explanation to the different toxicity profiles, the binding mode of ApAP to CYP2E1 was first analyzed using the same methodologies (Figure  2 ).

Figure 2.

Top. PyMOL stick and cartoon representation of the representative structure of the most populated conformation of ApAP (cyan) bound to CYP2E1. The heme group is shown in yellow, the Fe⁴⁺‐O²⁻ system as spheres, only polar hydrogens are shown for the sake of clarity, hydrogen bonds are shown by dashed lines and amino acids that establish hydrogen bonds are tagged in bold. Bottom. Graphical representation of the per residue energy decomposition (Kcal/mol) of the binding of ApAP to CYP2E1. Error bars show standard deviation during the 20 ns MD simulation.

According to our computational study, ApAP (a rigid molecule) binds CYP2E1 in the same conformation that it presents in the free aqueous solvent. Interestingly, after the initial first ns of MD simulation the binding mode was stabilized within the deep hydrophobic pocket of CYP2E1 without interacting with the access tunnel. The overall conformation adopted by the compounds both in free water solvent and in complex with CYP2E1, is in agreement with the expected conformation of a benzamide moiety. Although slight distortions can be appreciated in the conformations, they fall within the range of normal behavior of molecules under the effect of different interactions, as well as for the distortion suffered by a solvated dynamic system simulated at 300K. The initial accommodation of ApAP to the binding site was mainly due to the π–π interaction with the side chain of F478; additionally, the binding mode was stabilized by three hydrogen bonds: the backbone carbonyl of F478 and NH of L368 established hydrogen bonding interactions with the amide moiety, which favored the hydrogen bonding interaction between the side chain OH of T303 and the phenolic OH (Figure S3, Supporting Information). This triggers formation of a hydrogen bond between the phenol polar hydrogen and the oxygen of the Fe⁴⁺‐O²⁻ system, which is a plausible starting orientation for the NAPQI formation. During this MD simulation and as ApAP is small and only interacts with the main binding site on top of the heme group and does not establish any interactions with the amino acids lining the access tunnel and cavity entrance (Figure S4, Supporting Information).

Compounds 6, 6o and 6r insert the phenol moiety in the binding site and orient it towards the Fe⁴⁺‐O²⁻ system, establishing van der Waals interactions with the side chain of F478 in a similar manner to ApAP (Figure  3 – 4 , 5 , respectively). However, compounds 6 and 6r do not establish hydrogen bonds with either the NH of L368 and side chain OH of T303 that respectively stabilize the binding of ApAP and favor a stable orientation of the OH towards the Fe⁴⁺‐O²⁻ system. (Figure 3 and 5). The sulfonamide substituent of the small and small‐polar compound representative 6, interacts with the amino acids of the access tunnel and due to its geometry and bulkiness, the complete access of the phenol moiety to the Fe⁴⁺‐O²⁻ and hydrophobic pocket is impaired.

Figure 3.

Top. PyMOL stick and cartoon representation of the representative structure of the most populated conformation of compound 6 (orange) bound to CYP2E1. The heme group is shown in yellow, the Fe⁴⁺‐O²⁻ system as spheres, only polar hydrogens are shown for the sake of clarity, hydrogen bonds are shown by dashed lines and amino acids that establish hydrogen bonds are tagged in bold. Bottom. Graphical representation of the per residue energy decomposition (Kcal/mol) of the binding of 6 to CYP2E1. Error bars show standard deviation during the 20 ns MD simulation.

Figure 4.

Top. PyMOL stick and cartoon representation of the representative structure of the most populated conformation of compound 6o (magenta) bound to CYP2E1. The heme group is shown in yellow, the Fe⁴⁺‐O²⁻ system as spheres, only polar hydrogens are shown for the sake of clarity, hydrogen bonds are shown by dashed lines and amino acids that establish hydrogen bonds are tagged in bold. Bottom. Graphical representation of the per residue energy decomposition (Kcal mol⁻¹) of the binding of 6o to CYP2E1. Error bars show standard deviation during the 20 ns MD simulation.

Figure 5.

Top. PyMOL stick and cartoon representation of the representative structure of the most populated conformation of compound 6r (lilac) bound to CYP2E1. The heme group is shown in yellow, the Fe⁴⁺‐O²⁻ system as spheres, only polar hydrogens are shown for the sake of clarity, hydrogen bonds are shown by dashed lines and amino acids that establish hydrogen bonds are tagged in bold. Bottom. Graphical representation of the per residue energy decomposition (Kcal/mol) of the binding of 6r to CYP2E1. Error bars show standard deviation during the 20 ns MD simulation.

Contrary to the latter, the bulky‐hydrophobic compound 6o has full access to the ligand binding site above the heme group and the Fe⁴⁺‐O²⁻ system (Figure S5, Supporting Information). The binding mode of compound 6o involves the entire molecule fitting inside the hydrophobic cavity, with the phenol and dichlorophenyl moieties close together, establishing strong π‐π stacking interactions with F478. Additionally, as it occupies the entire hydrophobic pocket, it establishes hydrophobic interactions with all the amino acids that line the cavity, a strong hydrogen bonding interaction with backbone carbonyl of F478, and between the phenol and the side chain OH of T303, thus favoring the orientation towards the Fe⁴⁺O²⁻ system (Figure 4).

Interestingly, the binding pose predicted for 6o was very similar to that of the major conformation predicted in the free state in water, which aligns with the higher toxicity experimentally shown by 6o, compared to 6 and 6r (Figure S6, Supporting Information).

To account for the non‐toxicity of 4 and 5, their binding mode to CYP2E1 was also analyzed. Both compounds were found to interact with the protein in a very similar manner as that of 6, establishing similar electrostatic and van der Waals interactions (Figure S7, Supporting Information) and interacting with the entrance to the pocket, thus not allowing a full access to the Fe⁴⁺O²⁻ system. This can also be assessed by analyzing the interaction pattern for these compounds, which is very similar to that of 6 with low interactions with T303 and F106.

3. Conclusions

In this work we have performed computational analyses to investigate the dynamic behavior, physicochemical properties, and ligand‐receptor interactions between CYP2E1 and ApAP together with the previously described novel analgesic candidates 4, 5, 6, and 6a‐6u.

Interestingly, ccompounds with higher logP values exhibit greater CYP2E1‐related toxicity, consistent with the hydrophobic ligand‐binding site over the heme group, which favors interactions with hydrophobic compounds and reduces desolvation penalties. In this line, the lack of measured values of GSH and released LDS for candidates 6d, 6h and 6t can be correlated to the lowest logP values due to the hydrophilic or ionizable nature of their substituents.

According to our studies, the toxicity of the most hydrophobic compounds is also related to their predominant conformation in water, which is in most cases very similar to the predicted bound ligand conformation. Less toxic compounds interact with the access tunnel, impairing stable interactions between the phenol and Fe⁴⁺‐O²⁻ system, reducing the likelihood of the radical reaction leading to NAPQI formation to occur.

The addition of a substituent to the acetamide of ApAP alters the overall shape of the molecules, making them larger and more flexible, with conformations unsuitable for accessing the heme site of CYP2E1. The nature of the substituent is crucial, as the most hydrophobic compounds find a favorable binding site within CYP2E1, increasing toxicity.

Based on previous experimental results and in complete agreement with the computational study carried out in this work, compound 6 is advancing as a promising novel nonopioid analgesic candidate.

4. Experimental Section

4.1.

4.1.1.

4.1.1.1.

All compounds (ApAP, 4, 5, 6, and 6a‐6u) were prepared with the LigPrep module of Maestro in both the phenol and phenoxy forms; and the pharmacokinetic properties had been calculated for the phenol forms using the Chemicalize [04/2024] (http://www.chemicalize.com), developed by ChemAxon web server. Chemicalize predicted compounds 5, 6h, 6s and 6u to be ionized under physiological conditions (Table S1, Supporting Information), therefore these compounds were modeled in the predicted ionization states.

One of the main issues of ligand‐protein docking was that the initial protein structure used for the docking will affect the results, as complete full protein flexibility was not achieved by any docking program. Therefore, the selection of the initial protein structure was crucial, as it will set the protein conformation that is going to be worked on. The first approach, however not always possible, was to use a protein bound to ligands which have a chemical structure somewhat similar to the ligands to be studied. The crystal structures of CYP2E1 deposited in the Protein Data Bank (PDB) are bound to ligands that are structurally very different to the set of compounds under investigation in this work. Nevertheless, the structure of CYP2E1 bound to 10‐(1H‐imidazol‐1‐yl)decanoic acid deposited in the PDB under code 3GPH, was the one selected for this study as the ligand is linear and it does not only interacted with the heme group, but also with the pocket entrance thus favoring an induced fit (IF) that leaves it open.^{[ 23 ]} Since there was a significant structural difference between the ligand bound to the selected protein structure and those studied in this work, and with the aim of exploring more possible conformations of the protein, the possible NM of the CYP2E1 were analyzed using the elNèmo web server.^{[ 24 ]} The first two NMs, NM1 and NM2, were selected, each of which were decomposed into eleven protein conformations that represented the range of movement of each NM. It was important to note that the first NM results issued from the calculations were modes 7 and 8, the first six were rotations and translations in the x, y and z axes. Thus, the 22 protein conformations were then prepared with the Protein Preparation Wizard module of the Schrödinger.^{[ 25 ]}

The proposed state of the CYP2E1 that leads to the formation of the NAPQI consists of the form Fe⁴⁺‐O²⁻, which according to the work presented by Li Ji and Gerrit Schuüuürmann,^{[ 22 ]} would be the form that would start the reaction upon interaction with the phenol form of ApAP.

For docking purposes, and given the limitations of the OPLS3 forcefield, the heme group in each protein conformation was modeled with a coordinated water molecule to the iron center and the phenoxy form of the set of compounds was screened into the 22 protein conformations. The docking of the phenoxy form of the set of compounds into the 22 protein conformations extracted from NM1 and NM2 allowed only binding of compound 6 as it was less bulky and could fit within the protein, but not small enough to remain unstabilized by any of the 22 open conformations of CYP2E1 used as receptors for docking studies. The protein conformations that achieved ligand binding in this first docking were subsequently used for IF docking using the Schrödinger software,^{[ 26 ]} a docking strategy that after each docking run carries out a full protein optimization to allow a better coupling of both ligand and receptor. Following this procedure, a binding pose for compound 6o was obtained.

For the MD simulations of the ligand‐protein complex the heme group in the MD simulations was modeled in the form Fe⁴⁺‐O²⁻,^{[ 22 ]} therefore the phenoxy form of the compounds was reverted to the phenol form and the water‐coordinated heme group was transformed into the Fe⁴⁺‐O²⁻ form. The metal site (the heme group, the side chain of C437 and the coordinated oxygen) was then parametrized with the MCPB module embedded into the MTK++ software package in AMBER16, a methodology widely used to facilitate the modeling of metal centers on metalloproteins.^{[ 27 ]} The optimized coordination sphere was obtained by geometry optimization in the gas phase using Gaussian09 at B3LYP/6‐31++G** level, that led to obtaining the equilibrium values of bond lengths, angles and force constants of the atoms coordinated to the metal.

For the MD simulations, the geometry optimization and charge distributions of all ligands were calculated quantum mechanically (RHF/6‐31+G**) with Gaussian 09 Revision A.1 (Gaussian, Inc., Wallingford, CT) and the general AMBER force field 2 (GAFF2) was used to assign bonded and nonbonded parameters.^{[ 28 ]}

To account for the dynamic behavior of all compounds in water and to obtain the most populated water conformations of all ligands, they were all embedded in a TIP3P water octahedron of ≈1200 water molecules. Water molecules and counter ions were energy minimized and then the system was heated for 25 ps at constant volume to a target temperature of 300 K. To avoid the alteration of the overall conformation of the system, the protein was initially restrained using quadratic harmonic restraints with a constant force of 50 kcal mol⁻¹ Å⁻² and those restraints were gradually reduced until eliminated in 100 ps. In all systems, the hydrogen bond lengths were kept at their equilibrium distance by the SHAKE algorithm.^{[ 29 ]} Atom pair distance cutoffs to compute the van der Waals interactions were applied at 10.0 Å, while long‐range electrostatic interactions were computed by the particle‐mesh Ewald method.^{[ 30 ]} MD production was performed up to 20 ns using the thermostat NPT ensemble at target temperature, generating snapshots each 20 ps until a total simulation time of 100 ns for further analysis. The trajectories were collected and analyzed by the cpptraj module of AMBER16^{[ 31 ]} in order to obtain the root mean square deviation (RMSD) value of the atomic positions of the ligands as well as for the obtention of the most populated conformers and the radius of gyration that measures the bulkiness of a compound.^{[ 32 , 33 ]}

The docking poses of 6 and 6o were used as starting structures for the MD simulations after changing the protonation states, as mentioned above. Additionally, the system with compound 6 was used to model the complexes of CYP2E1 with 4, 5, and ApAP and the most populated water conformation of 6r based on the strong structural similarity with compound 6o. The classic ff14SB AMBER force field^{[ 34 ]} for the protein parametrization was applied together with previously assigned ligand and heme parameters. Systems were energy minimized at vacuum; neutrality was achieved in all systems by adding sodium or chlorine ions when necessary, and then all complexes were embedded in a TIP3P water octahedron of ≈15 000 water molecules. Classical MD simulations were performed on these ligand‐protein complexes using the same conditions as stated above until a total simulation time of 20 ns for further analysis. The trajectories of all complexes were collected and analyzed by the cpptraj module of AMBER16^{[ 31 ]} in order to obtain the RMSD value of the atomic positions of the ligands as well as for the obtention of the most populated conformer of the bound ligands and the protein. The per‐residue energy decomposition of the binding of the studied complexes was carried out with the MM‐ISMSA program.^{[ 35 ]} The analysis and visualization of tunnels and channels in the most populated protein conformers was carried out using the CAVER 3.0 software.^{[ 36 ]}

Conflict of Interest

The authors declare no conflict of interest.

Supporting information

Supplementary Material

Coderch Claire, Bazán Hernan A., Bazan Nicolas G., Surjyadipta Bhattacharjee, Alvarez‐Builla Julio, de Pascual‐Teresa Beatriz, ChemMedChem 2025, 0, e202500639. 10.1002/cmdc.202500639

Data Availability Statement

The data that support the findings of this study are available in the supplementary material of this article.