Distilled water, ultrapure water, ethyl acetate, ethanol, FeCl₃, magnesium metal, Folin-Ciocalteu reagent, gallic acid, ascorbic acid, 10% AlCl₃, CH₃CO₂K 1M, quercetin, dimethyl sulfoxide, FeSO₄, 2,2-diphenyl-1-picrylhydrazyl (DPPH), 2,4,6, tripyridyl-s-triazine (TPTZ), 2,20-azo-bis (2-amidinopropane dichlorohydrate), chlorogenic acid, diosmetin, mexoticin, and daidzein with purity higher than 95% were purchased from Sigma-Aldrich Chem. Co. (St Louis, MO, United States), and analytical grade solvents were obtained from Merck^® (Santiago de Chile).

Fronds (leaves) of L. quadripinnata were collected in an area adjacent to the botanical garden of the Universidad Austral de Chile (Valdivia, Los Ríos region, Chile). Based on a specialized bibliography and the review of morphological characters, was identified by the botanist Javian Gallardo-Valdivia, and a specimen was deposited in the EIF Herbarium of the Faculty of Forestry Sciences and Nature Conservation of the Universidad de Chile (EIF 17430), acronym according to Thiers (continuously updated: http://sweetgum.nybg.org/science/ih/).

Fronds of L. quadripinnata (20 g) were ground manually and mixed with 250 mL of distilled water in ultrasound for 30 min at 40°C, after obtaining a homogeneous solution, the procedure was filtered and repeated three times. The filtered solution was deposited in a FreeZone LABCONCO benchtop lyophilizer; the aqueous material was stored under refrigeration for later use, and a final yield of 9% was obtained.

Using an Ultimate 3000 RS UHPLC instrument with Chromeleon 6.8 software, and a Bruker maXis ESI-QToF-MS with Data Analysis 4.0 software, the separation of the phytocompounds present in the aqueous extract of L. quadripinnata fern was performed. The extract (5 mg) was dissolved in 2 mL of methanol, filtered with a polytetrafluoroethylene (PTFE) filter and 10 µL were injected into the apparatus. Elution was performed with a binary gradient system with eluent (A) corresponding to 0.1% formic acid in water, and eluent (B) corresponding to 0.1% formic acid in acetonitrile: 1% B isocratic (0–2 min), 1%–5% B (2–3 min), 5% B isocratic (3–5 min), 5%–10% B (5–8 min), 10%–30% B (8–30 min), 30%–95% B (30–38 min) and 1% B isocratic (38–50 min). ESI-QToF-MS experiments were recorded in negative ion mode, and the scan range was between 100 and 12,000 m/z. Separation was performed with a Thermo 5 µm C18 80 Å column (150 mm × 4.6 mm) at a flow rate of 1.0 mL/min. Electrospray ionization (ESI) conditions were at a capillary temperature of 200°C, capillary voltage of 2.0 Kv, dry gas flow rate of 8 mL/min and nebulizer pressure of 2 bar, and the experiments were performed in automatic MS/MS mode. The final identification of the compounds was based on mass data, ion fragmentation patterns and comparison with specialized literature.

An Ultimate 3,000 RS UHPLC-DAD system equipped with Chromeleon 7.2 software was used to measure some of the main compounds in the extract. Peak 3, corresponding to atrovenetin, showed UV maxima at 218 nm and 383 nm; therefore, 383 nm was used for quantification, employing a calibration curve constructed with diosmetin at 383 nm. Peaks 7 and 8, identified as caffeoyl glucose derivatives, and peaks 13 and 14, corresponding to caffeoylquinic acid derivatives, were quantified using a calibration curve of caffeoylquinic acid (chlorogenic acid) at its UV maximum of 320 nm. Peak 10, a coumarin, and peak 11, mexoticin, were quantified using a standard curve prepared with the structurally similar coumarin daphnetin at 320 nm. Peak 12, daidzein, was quantified at its UV maximum of 254 nm. At least six calibration points were used, with concentrations ranging from 5.0 to 125 mg/mL, showing a high correlation coefficient (R ² > 0.998) and results are expressed as mean ± standard deviation (SD).

Total phenolic content was determined by the Folin-Ciocalteu method, using 7% Na₂CO₃ solution and gallic acid calibration curve; the solutions were incubated for 90 min in the dark and absorbance was measured at 750 nm; the result was expressed as mg of gallic acid per gram of dried fern, and gallic acid was used as the reference compound (Singleton et al., 1999). The determination of total flavonoids was performed with the reaction between the samples and AlCl₃ in the presence of NaNO₂ with alkaline medium and quercetin calibration curve (Zhishen et al., 1999); the solutions were incubated, and the absorbance was measured at 510 nm; the result was expressed as mg of quercetin per gram of dried fern, and quercetin was used as the reference compound.

It was performed by assaying the reduction of the ferric 2,4,6-tripyridyl-s-triazine complex (TPTZ) 10 mM dissolved in ethanol, using CH₃COONa-3H₂O 3.1%/glacial CH₃COOH 16% dissolved in deionized water, FeCl₃ 20 mM in 0.02 M HCl aqueous solution, trolox calibration curve and reading the corresponding absorbances of each sample at 593 nm. The result was expressed as micromoles of trolox equivalents per gram of dried fern, and trolox was used as the reference compound (Benzie and Strain, 1996).

The compound 2,2′-azobis (2-amidinopropane) dihydrochloride (AAPH) was used in a readout kinetics every 2 minutes for one and a half hours, measuring each solution at an excitation and emission wavelength of 485 and 530 nm respectively. The result was expressed as micromoles of trolox equivalents per gram of dried fern, and trolox was used as the reference compound (Cao and Prior, 1999).

A 400 μM solution of 2,2-diphenyl-1-picrylhydrazyl (DPPH) radical in absolute ethanol was used to evaluate antioxidant activity. A gallic acid calibration curve was prepared in the same solvent and used as a standard. The samples were incubated for 30 min, and absorbance was measured at 515 nm. Results were expressed as IC50 values (µg of fern extract/mL), with gallic acid serving as the positive control (Brand-Williams et al., 1995).

An in silico molecular docking was performed to evaluate the binding capacity of L. quadripinnata compounds to various crystallographic structures of enzymes and proteins: Torpedo californica acetylcholinesterase (TcAChE; PDB ID: 1DX6, resolution 2.30 Å, Greenblatt et al., 1999), human butyrylcholinesterase (hBChE; PDB ID: 4BDS, resolution 2.10 Å, Nachon et al., 2013), tyrosinase from Agaricus bisporus (PDB ID: 2Y9X, resolution 2.78 Å, Ismaya et al., 2011), and human Keap1 (PDB ID: 4L7B, resolution 2.10 Å). All protein structures were obtained from the RCSB PDB database. Optimization of the enzymes was carried out using the UCSF Chimera software (v1.16, San Francisco, California, United States), each of the enzymes had water molecules and ligands removed from the active sites of the enzymes. All polar hydrogen atoms were added at pH = 7.4, and the appropriate ionization states were considered for each of the basic and acidic amino acid residues. The molecular docking box was set to a cube for each of the enzymes with sides 30 Å long (Table 1).

The centroid of the selected residue was chosen based on the putative catalytic site in each of the enzymes, considering its known catalytic amino acids: Ser200 and His440 for acetylcholinesterase (TcAChE) (Castillo et al., 2016; Roy et al., 2019) Ser198 for butyrylcholinesterase (BChE) (Liu et al., 2020a; Amat-ur-Rasool et al., 2022), for Nrf2-Keap1 Tyr334, Ser363, Arg380, Asn414, Arg415, Ser508, Ser555, Tyr572 and Ser602 (Hassanein et al., 2020a) and for tyrosinase Asn81, His244 and His263 (Ismaya et al., 2011; da Silva et al., 2017). The two-dimensional structures of the ligands were processed in ChemDraw 8.0 software (PerkinElmer Informatics, Waltham, MA, United States) where they were saved in. mol format. Subsequently, they were imported into Avogadro software (https://avogadro.cc, accessed April 20, 2022) to perform geometric optimization of each of the reference ligands and inhibitors using the MMFF94 force field function (Torres-Benítez et al., 2022). Molecular docking was performed using each of the respective rigid crystallographic structures of the enzymes/protein (acetylcholinesterase, butyrylcholinesterase, tyrosinase and Nrf2-Keap1) docking them with the flexible ligands (5C3M, 5GDC and irifloside) whose torsion angles were identified (for 10 independent runs per ligand), in addition to the fact that they did not present any toxicological risk and no more than a violation in the Lipinski’s parameters.

To perform the molecular docking, the UCSF Chimera software was used (Volkamer et al., 2012; Chowdhury et al., 2014) where the galantamine compound was used as a reference inhibitor for acetylcholinesterase and butyrylcholinesterase; as a reference inhibitor for Nrf2-Keap1 to the compound (S, R, S) (Hassanein et al., 2020b) and a reference for tyrosinase to the compound kojic acid (Ismaya et al., 2011; da Silva et al., 2017). Polar hydrogens and partial Gasteiger charges were added, and a grid box was created using the Autodock Vina tools at UCSF Chimera. Docking and analysis results were visualized using Discovery Studio Visualizer (BIOVIA, 2023). Once the molecular docking was completed, the best conformation in each of the ligands for hydrogen bonds or π interactions was analyzed, including the binding energy of the free ligand (kcal/mol) (Torres-Benítez et al., 2022). After a successful docking process with the protein, the lowest binding energy for each ligand was determined. The binding site and docked conformations were then visualized to detect interactions. After performing the molecular docking, the inhibition constant was determined for each of the ligands with each of the crystallographic enzymes/protein (Kumar and Sarkar, 2023). The inhibition constant (Ki) was calculated using the following Equation 1: Ki = e ∆ G / R T (1) Where ∆G represents the minimum binding energy of the docked conformations, R is the universal gas constant (R = 1.985 × 10⁻³ kcal mol⁻³K⁻³) and T is the temperature (T = 298.15K) (Kumar and Sarkar, 2023).

To ensure the reliability of the molecular docking protocol, a redocking validation analysis was also performed. The co-crystallized ligands of each protein target—galantamine for acetylcholinesterase (PDB ID: 1DX6) and butyrylcholinesterase (PDB ID: 4BDS), (S, R, S) for the Nrf2-Keap1 complex (PDB ID: 4L7B), and tropolone for tyrosinase (PDB ID: 2Y9X)—were re-docked into their respective active sites using the same docking settings described above. The resulting poses were compared to the crystallographic ligand conformations by calculating the Root Mean Square Deviation (RMSD) values using PyMOL v2.5.0. The docking was considered valid when the RMSD was below the accepted threshold of 2.0 Å. The RMSD values obtained ranged from 0.533 Å to 1.728 Å, confirming high reproducibility of the native poses and validating the accuracy of the docking methodology used in this study.

To determine the pharmacokinetic properties and toxicological risks of the compounds obtained in L. quadripinnata, the Osiris Data Warrior computational tool is used. The molecular descriptors that were calculated are the participation coefficient (cLogP), the number of hydrogen donor bonds (HBD), the number of hydrogen acceptor bonds (HBA), the total number of rotatable bonds (n-ROTB), topological polar surface area (TPSA) and violations of Lipinski’s rule of five. Using the TPSA value, the percentage of absorption (% ABS) was calculated using the following Equation 2 (Torres-Benítez et al., 2022): %  ABS = 109 − 0.345 ×  TPSA (2)

For each of the molecules, the possible toxicity risks they could present were evaluated according to their structure and chemical fragments. The toxicity risks that were evaluated were mutagenicity, tumorigenicity, irritation and reproductive effect Torres-Benítez et al., 2022).

To further validate the docking results and assess the stability of the predicted protein-ligand complexes, molecular dynamics (MD) simulations were performed using YASARA Dynamics v.18.4.24 (Yasara GmbH, Vienna, Austria; Krieger and Vriend, 2015) with the AMBER14 force field (Case et al., 2014). Initial structures were derived from the top-scoring docking poses of the multi-target ligands with each protein. Each complex was solvated in a cubic water box (100 Å × 100 Å × 100 Å) with periodic boundary conditions (PBC) to mimic a bulk environment. The system was maintained at 298 K and 0.997 g/cm³ water density. Sodium (Na+) and chloride (Cl-) ions were added for charge neutrality, simulating physiological conditions (pH 7.4, 0.9% NaCl). The particle-mesh Ewald (PME) algorithm with a cut-off radius of 8 Å was employed for electrostatic interactions. Simulations were run for 100 ns, recording snapshots every 250 ps with a timestep of 2.5 fs. Equilibrium was established based on Root Mean Square Deviation (RMSD) fluctuations below 2 Å for at least 15 ns. The resulting trajectories were analyzed using YASARA’s built-in tools, including RMSD, Root Mean Square Fluctuation (RMSF), and radius of gyration (Rg) variations over time as well as ligand binding energy (LBE) calculations using the MM-PBSA method. LBE calculations were performed on snapshots from the last 75 ns of the simulations. In Yasara protocols, the more positive the binding energy, the more favorable the interaction (Ahmed et al., 2021). PLIP web tool (Adasme et al., 2021) was used to determine the interactions of final ligand - protein complexes after 100 ns of simulation, and compared with those obtained after the docking process. ProteinsPlus (https://proteins.plus) PoseView was used to create 2D pose depictions of protein - ligand complexes after 100 ns of simulation (Stierand and Rarey, 2010).

For each in vitro test solution, three measurements were performed, and the results were expressed as the mean ± standard deviation (SD) using Microsoft Excel 2019 (Microsoft Corporation, Redmond, WA, United States). Subsequently, with a one-way analysis of variance with Tukey’s test, statistical significance was corroborated (p < 0.05).

The chemical profile of the aqueous extract of the fern L. quadripinnata was obtained by high-performance ultra-high chromatography coupled with mass spectrometry (UHPLC-MS). The LC-MS/MS raw data were processed using Bruker’s DataAnalysis software, which enabled noise reduction, spectral deconvolution, and mass peak visualization. For compound structural characterization, we considered accurate mass data and fragmentation patterns, complemented by searches in digital spectral libraries (MoNA–MassBank of North America, SDBS–Spectral Database for Organic Compounds, and SpectraBase), and relevant scientific literature. Fourteen peaks were tentatively identified in negative ionization mode (Figure 2), corresponding to carbohydrate and aromatic compounds, as shown in Table 2, Figure 3 and Supplementary Figures S1–S6. This ionization method, based on the formation of negatively charged ions, is suitable for the detection of compounds such as phenolic acids and flavonoids, which contain functional groups that are easily deprotonated (Straub and Voyksner, 1993).

Phenolic acids and phenylpropanoid derivatives: Peak 3 was tentatively identified as atrovenetin (C₁₉H₁₇O₆), with a molecular anion m/z 341.09808 and with diagnostic peaks at m/z 191.0513 and 683.20954, and peak 4 was tentatively identified as atrovenetin isomer. This isomer was identified based on its matching molecular formula, theoretical mass, and mass fragments with atrovenetin, along with the low differences in measured mass and mass accuracy between the two. However, considering that both compounds share the same retention time, future chromatographic and isolation studies should include a commercial standard and additional spectroscopic data such as nuclear magnetic resonance. Peak 7 and 8 were tentatively identified as 3-O-caffeoyl glucose (C₁₅H₁₇O₉), with an [M-H]- ion at m/z 341.07924 and diagnostic peaks at m/z 327.09700, 281.05861 and 191.0513, and 5-O-caffeoyl glucose (C₁₅H₁₇O₉), with an [M-H]^- ion at m/z 341.07929, respectively. Peak 10 was tentatively identified as 5-O-glucoside-6,7-dimethoxycoumarin (C₁₇H₁₉O₁₄) with a molecular anion m/z 447.08330 and with diagnostic peaks at m/z 393.04092, 311.15992 and 255.01204. Peaks 13 and 14 were tentatively identified as 5-O-caffeoyl-3-O-malonylquinic acid (C₁₉H₁₉O₁₂) with a molecular anion at m/z 439.0823, and 3-O-caffeoyl-5-O-malonylquinic acid (C₁₉H₁₉O₁₂) with a molecular anion at m/z 439.1034, respectively.

Among the isoflavonoids, peak 9 was tentatively identified as irifloside (C₂₃H₂₁O₁₂), showing a molecular anion at m/z 489.09280 and diagnostic fragments at m/z 447.08263 and 341.08066. For the aromatic derivatives, peak 5 was assigned to 6″-O-acetylgenistin (C₂₃H₂₁O₁₁), peak 11 to mexoticin (C₁₆H₁₉O₆), and peak 12 to daidzein 6″-O-acetate (C₂₃H₂₁O₁₀), with a [M-H]^- ion at m/z 457.01110 and diagnostic fragments at m/z 439.09962 and 311.16356. Regarding carbohydrates, peak 6 was identified as sucrose succinate (C₁₆H₂₃O₁₅), presenting a molecular anion at m/z 455.10424 and characteristic fragments at m/z 359.08599 and 279.10013. Peaks 1 and 2 were detected but remained unidentified, while peak 15 corresponded to sodium formate (C₄H₂O₄) used as the internal standard, for calibration of the mass spectrometer which improves the precision and accuracy in determining the masses of the compounds present in the extract.

These compounds present in the extract of L. quadripinnata are part of the group of phytocompounds also reported in other species from Blechnaceae such as Parablechnum chilense (ex Blechnum chilense), Blechnum hastatum, Lomariocycas magellanica (ex Blechnum magellanicum), Austroblechnum penna-marina (ex Blechnum penna-marina), Blechnopsis orientalis (ex Blechnum orientale), Parablechnum novae-zelandiae (ex Blechnum novae-zelandiae), Blechnum occidentale and Lomaridium binervatum (ex Blechnum binervatum) (Yusuf, 1994; Fons et al., 2010; Lai et al., 2010; Gasper et al., 2016; Andrade et al., 2017; Waswa et al., 2022; Torres-Benítez et al., 2023).

On the other hand, other triterpenoid-type compounds have been identified and isolated from L. quadripinnata such as fern-9 (11)-ene, dryocrassol, adipedatol, diplopterol, and a hydroxymethyl derivative of dryocrassol, which are also reported in fern genera such as Adiantum, Polypodium, and Dryopteris (Tanaka et al., 1992). The presence of the isoflavonoid irifloside should also be highlighted, as it has been widely reported in rhizomes of plants from the Iridaceae family, showing high cytotoxic activity (Qin et al., 2007; Xie et al., 2013; Amin et al., 2021). In short, the study of its chemical composition at the compound level allows validating the traditional use of this species, which in the case of the south of the American continent has been used as a homeostatic remedy (Looser and Rodríguez Ríos, 2004).

In general, the identification of the fourteen compounds by UHPLC-ESI-QToF-MS was focused on qualitative characterization, based on accurate mass values, MS/MS fragmentation patterns, comparison with spectral databases, and specialized literature. However, a semi-quantitative approximation was initially performed from the chromatogram using the relative abundances of the compounds selected for the in silico study, yielding a percentage estimation relative to the total of 60.8% for irifloside (peak 9), 58.6% for 5GDC (peak 10), and 89.1% for 5C3M (peak 13).

Additionally, a more formal quantification was carried out, revealing that the total content of phenolic compound peaks analyzed in our L. quadripinnata fern extract amounts to 146.783 mg/g of extract. Peaks 10 and 11 were the most abundant compounds (64.271 mg/g), followed by the caffeoyl glucosides (peaks 7 and 8) at 37.283 mg/g, the biologically active compound atrovenetin (peak 3) at 18.038 mg/g, the caffeoylquinic acid derivatives (peaks 13 and 14) at 15.281 mg/g, and the lowest content was observed for daidzein 6″-O-acetate (peak 12), with 11.910 mg/g (Table 3).

The absence of commercial standards for the major compounds and the exploratory nature of the study limited the possibility of absolute quantifying the bioactive candidates. As a continuation of this work, we plan to carry out the isolation, purification, and structural elucidation (NMR, IR, UV-Vis) of key constituents such as 5C3M, 5GDC, and irifloside. This will allow for more accurate quantification of their concentrations in the extract and evaluation of their therapeutic potential through in vitro and in vivo models.

Table 4 shows the values obtained in the assays of total phenols, total flavonoids, and antioxidant activity for the aqueous extract of L. quadripinnata. The result of the phenol (49.029 ± 0.015 mg GAE/g) and flavonoid (91.307 ± 0.059 mg QE/g) contents indicate a medium-high concentration of polyphenolic compounds that trigger significant results in FRAP (2,110.051 ± 0.986 µmol Trolox/g), ORAC (2,631.432 ± 0.998 µmol Trolox/g) and especially DPPH tests with an IC₅₀ of 78.137 ± 0.010 μg/mL inferring a high antioxidant potential that would allow the decrease of high concentrations of reactive oxygen species.

For phenolic compound content, L. quadripinnata is similar with the data reported for A. penna-marina (88.846 ± 0.020 mg GAE/g) and P. chilense (34.078 ± 0.010 mg GAE/g) species and in flavonoids with A. penna-marina (128.662 ± 0.065 mg QE/g) (Torres-Benítez et al., 2023). For antioxidant activity in FRAP, ORAC and DPPH assays, L. quadripinnata extract shows high potential which is close to the outstanding activity reported in A. penna-marina (3,301.847 ± 1.050 μmol Trolox/g, 2,677.519 ± 0.096 μmol Trolox/g and 41.818 ± 0.005 μg/mL, respectively) (Torres-Benítez et al., 2023). On the other hand, the flavonoid content is also comparable with those obtained in Pyrrosia petiolosa (80.6 mg/g), Lepisorus carnosus (ex Lemmaphyllum carnosum) (125.9 mg/g), Goniophlebium amoenum (ex Polypodiodes amoena) (71.4 mg/g), Diplopterygium glaucum (ex Hicriopteris glauca) (77.6 mg/g), Davallodes squamata (ex Araiostegia imbricata) (105.9 mg/g) and Drynaria coronans (ex Pseudodrynaria coronans) (78.9 mg/g) (Cao et al., 2015). With the Asian species W. unigemmata reporting a maximum flavonoid content of 151 ± 11.44 mgQE/g and 768 ± 10.4 mg AAE/g for FRAP, the extract of L. quadripinnata maintains comparable ranges, however, its phenolic content and IC₅₀ in DPPH results lower compared to 873 ± 6.01 mgGAE/g and 6.07 ± 1.4 μg/mL, respectively in Woodwardia unigemmata (Takuli et al., 2020).

The polyphenol and flavonoid content, as well as the antioxidant activity of L. quadripinnata, differ significantly from those reported for Athyrium niponicum and Dryopteris erythrosora. These variations are influenced by abiotic factors such as salinity, sunlight exposure, and environmental stress (Pietrak et al., 2022b). In the case of D. erythrosora, flavonoid content and antioxidant activity also vary depending on the type of drying pretreatment, such as shade or heat (Zhang et al., 2019). Overall, the antioxidant potential of ferns highlights their promising applications in pharmaceutical and industrial fields, particularly for the development of photosynthesized nanoparticles with diverse uses and impacts (Fierascu et al., 2021).

The pharmacokinetic parameters and toxicity risks of the compounds present in the aqueous extract of L. quadripinnata were predicted using Osiris DataWarrior software (Figure 3; Tables 5, 6). Table 5 summarizes the ADME properties of each compound, compared to the reference inhibitors galantamine and the standard inhibitor (S, R, S). The results indicate that none of the tested compounds violated more than one of Lipinski’s rules, which are commonly used to assess drug-likeness. Specifically, 5GDC, atrovenetin, D6OA, mexoticin, SUSC, along with both reference inhibitors, fully complied with Lipinski’s criteria, suggesting potential oral bioavailability (Williams et al., 2022).

However, toxicity predictions in Table 6 revealed that atrovenetin, D6OA, mexoticin, and SUSC posed potential toxicological risks, such as mutagenicity or tumorigenicity. Consequently, these compounds were excluded from subsequent molecular docking simulations. The remaining compounds—5GDC, 5C3M, and irifloside—met the pharmacokinetic criteria and exhibited no predicted toxicity, and were therefore selected as potential inhibitors for crystallographic enzyme/protein targets.

Bioavailability was further analyzed using the Topological Polar Surface Area (TPSA), a parameter that reflects a molecule’s ability to permeate membranes and is inversely related to intestinal absorption and blood-brain barrier penetration (Table 5) (Prasanna and Doerksen, 2009). TPSA values were used to estimate the theoretical absorption percentage (Equation 2), providing insight into how much of each compound could potentially be absorbed. Among the tested compounds, atrovenetin (72.01%) and mexoticin (79.60%) showed the highest predicted absorption, with values comparable to those of the reference inhibitors (Table 5). This suggests good membrane permeability, in line with literature indicating that compounds with TPSA < 140 Å² are more likely to be absorbed orally (Veber et al., 2002).

However, one compound exhibited low predicted absorption, and according to the toxicological analysis, its chemical structure may fragment, potentially generating reactive or unstable metabolites that increase toxicity risk (Table 6; Thompson et al., 2016). This highlights the importance of integrating both pharmacokinetic and toxicity data during early drug discovery stages.

In summary, only compounds that did not violate Lipinski’s rules and showed no predicted toxicity were selected as candidate inhibitors for acetylcholinesterase, butyrylcholinesterase, and Nrf2-Keap1. Accordingly, 5C3M, 5GDC, and irifloside were identified as the most promising compounds and were further evaluated via molecular docking.

Although in silico predictions using Osiris DataWarrior provided preliminary insights into the toxicity profiles of the identified compounds, no experimental validation of biological safety was performed. This represents a limitation of the current study, as computational predictions may not fully reflect biological responses under physiological conditions (Wang et al., 2022). Future research should incorporate in vitro toxicity assessments in human-derived cell lines such as HepG2, which are widely used for hepatotoxicity and metabolism studies (Godoy et al., 2013). Cytotoxicity assays like MTT or LDH (Fotakis and Timbrell, 2006), genotoxicity evaluations via the comet assay (Collins, 2004), and mitochondrial viability tests such as JC-1 staining (Smiley et al., 1991; Huang et al., 2018) are recommended to confirm the safety and therapeutic relevance of the compounds. These analyses will be essential to validate the in silico predictions and support their translational potential.

To evaluate the potential of L. quadripinnata compounds as enzyme inhibitors, a consolidated molecular docking analysis was performed using only the most promising candidates. These included the compounds 5C3M, 5GDC, and irifloside, selected based on their compliance with essential pharmacokinetic criteria: none showed more than one violation of Lipinski’s rules, and none presented toxicity risks according to in silico predictions (Table 7 and Table 8). These criteria are widely accepted in early-stage drug discovery to increase the likelihood of bioactivity, safety, and oral bioavailability (Veber et al., 2002; Baillie and Jordan, 2015).

Molecular docking simulations were conducted to assess the interaction of these selected compounds with the active sites of key therapeutic targets: acetylcholinesterase, butyrylcholinesterase, tyrosinase, and the Nrf2-Keap1 protein complex. The analysis focused on identifying binding interactions with amino acid residues known to be critical for enzymatic inhibition, thereby predicting potential efficacy. Figures and tables associated with this section summarize the specific binding residues and affinity scores for each compound.

To ensure analytical coherence and avoid fragmentation, only compounds that met both ADME and toxicity criteria were included in the docking study. While additional compounds from the extract presented interesting chemical structures, they were excluded due to predicted toxicity or poor pharmacokinetic profiles. Including them in the docking simulations would not be justifiable without first addressing their safety risks.

Future studies may consider performing docking analyses for these excluded compounds—such as atrovenetin, D6OA, mexoticin, and SUSC—after chemical modifications, derivatization, or formulation strategies that could mitigate their predicted toxicity. This stepwise approach aligns with rational drug design principles, avoiding premature conclusions based on compounds unlikely to progress to in vivo validation stages.

Despite the promising in silico and in vitro antioxidant findings, an important limitation of this study is the absence of biological validation in cellular or animal models. This lack of validation restricts the physiological extrapolation of the observed results (Godoy et al., 2013; Wang et al., 2022). To fully confirm the neuroprotective potential of the identified compounds, future studies should incorporate in vitro assays using neuronal cell lines such as SH-SY5Y or PC12, which are widely used in neutoxicity and neuroprotection models (Xie et al., 2020). These models will allow the evaluation of functional endpoints, including cytotoxicity (e.g., MTT and LDH assays), oxidative stress responses (e.g., DCFH-DA assay), and activation of the Nrf2 pathway via expression of target genes such as HO-1 and NQO1 (Loboda et al., 2016; Zhou et al., 2022). IIn vivo studies may also be warranted to explore pharmacokinetics, biodistribution, and therapeutic efficacy under physiological conditions (Xie et al., 2022). The implementation of these assays in future research will provide critical support to validate the computational predictions and further the translational potential of L. quadripinnata compounds in the context of neurodegenerative disease models.

It is important to acknowledge that this study did not include a re-docking validation step to assess the accuracy of the molecular docking protocol. Re-docking of native ligands followed by Root Mean Square Deviation (RMSD) analysis is a widely accepted method for verifying that a docking algorithm can reproduce the experimentally determined binding poses. RMSD values below 2.0 Å are generally considered acceptable for validation purposes (Kurze et al., 2022). Moreover, all docking calculations were conducted using AutoDock Vina, which—despite being efficient and widely used—has known limitations in scoring accuracy and conformational sampling (Pagadala et al., 2017). Therefore, no cross-platform validation was performed. Future studies should address this limitation by including re-docking of co-crystallized ligands, reporting RMSD values to ensure pose reliability, and performing comparative docking using at least one alternative tool such as Glide, GOLD, or MOE. This would enhance the robustness and reproducibility of the results, and help ensure that the docking outcomes are not platform-dependent (Chang et al., 2023). Additionally, combining docking with more accurate scoring functions or machine learning-based re-scoring could further refine the prediction of binding affinities (Cichonska et al., 2021).

The molecular dynamics experiments were carried out using Yasara Software, with 100 ns of simulation. After the simulation time, we wanted to analyze the stability of the complex of 5C3M, 5GDC and irifloside in the Acetylcholinesterase, Butyrylcholinesterase, Tyrosinase and Nrf2-Keap1 and determine their ligand binding energy via Molecular Mechanics Poisson-Boltzmann Surface Area (MM-PBSA) calculations. The set of plots in Figure 9 show the RMSD variations along the simulation time. The ligand binding energies (LBE) of the complexes after stabilization time are shown in Table 10. After a short equilibration time of 15 ns, all complexes were stable, with small changes in the RMSD values, except in Nrf2-Keap1 complexes.