The 3D structure of the target protein was developed using homology modelling. The query sequence was obtained from Crassostrea gigas AChE (XP_011454985.2). A template for this sequence was identified using Basic Local Alignment Search Tool (BLAST) (Altschul et al., 1990). The crystal structure for AChE (PDB ID: 6G1U, 1.79Å, Tetronarce californica) (Galdeano et al., 2018) was selected as the template and had an overall sequence similarity of 45% and an E-value of 7 × 10^–166. A sequence alignment was performed using the ClustalW server tool (Larkin et al., 2007). Homology models were generated using MODELLER 9.17 (Fiser and Šali, 2003), the model with the lowest MODELLER objective function was chosen for refinement and further validation. The SciGRESS FJ 2.9 program (Stewart, 2009) was used to prepare the homology model for docking by adding hydrogens, resolving any clashes, correcting missing valences and hydrogen energy minimisation using the MM2 protocol (Allinger, 1977). Validation of the model was conducted using the Ramachandran plot and evaluation from the ProSA web server (Wiederstein and Sippl, 2007).

For the first screening run, a library of compounds was prepared using the SciGRESS FJ program using the MM2 force field (Allinger, 1977) to optimise the structures and remove any salts. The library selection was from the Chembridge EXPRESS screening library, which comprises small compounds with a mean molecular weight of 355 Da, mean cLogP of 1.6 and mean total polar surface area of 66.8 Å (Schultz et al., 2011). A majority of the compounds within this library are defined by drug-like and lead-like descriptors, making them suitable targets for further development. The centre of the modelled binding pocket was defined by the positioning of the catalytic triad of AChE, (x = –58.4790, y = 58.3950, z = 19.8190) with 10 Å radius. The compounds were initially screened at 50% search efficiency, and 10 runs per compound. The basic amino acids lysine and arginine were defined as protonated and aspartic and glutamic acids were assumed to be deprotonated. The GoldScore (GS) (Jones et al., 1997), ChemScore (CS) (Eldridge et al., 1997; Verdonk et al., 2003), ChemPLP (PLP) (Korb et al., 2009) and ASP (Mooij and Verdonk, 2005) scoring functions were implemented to validate the predicted binding modes and relative energies of the ligands using the GOLD v5.4 software suite. The output for the scoring functions is dimensionless, a greater score predicts better likelihood of binding success. A higher screening efficiency of 100% and 100 runs per compound were used for the second round of screening. A consensus model was used to evaluate the pose of each compound, compounds that had similar poses determined by at least three of the four scoring functions were considered as possible candidates (Charifson et al., 1999). The final step of the virtual screen was the visual inspection in which the position of the compound within the pocket, and all protein ligand interactions, were evaluated for binding potential. Compounds with unfavourable poses or interactions were deemed unsuitable for selection.

Cholinesterase activities were measured by the Ellman method (Ellman et al., 1961) adapted for microtiter plates, as described by Ristovski et al. (2018) using electric eel acetylcholinesterase (eeAChE), human recombinant AChE (hrAChE) or horse serum butyrylcholinesterase (BChE) (all Sigma-Aldrich, St. Louis, Missouri, United States), all dissolved in 100 mM potassium phosphate buffer (pH 7.4) to 0.0075 U/mL. Stock solutions of the tested compounds (1 mg/mL) were prepared in 100% ethanol and progressively diluted in 100 mM potassium phosphate buffer (pH 7.4) to a final volume of 50 μL. Acetylthiocholine chloride and 5,5′-dithiobis-2-nitrobenzoic acid were dissolved in the same buffer in the respective final concentrations of 1 mM and 0.5 mM, and added in 100 μL volumes to the wells of the microtiter plates. Aliquots of 50 μL of each of the cholinesterases were added to start the reactions, which were followed spectrophotometrically at 405 nm and 25°C over 5 min using a kinetic microplate reader (Dynex Technologies Inc, Chantilly, Virginia, United States). Blank reactions without the inhibitors were run in the presence of the appropriate dilution of ethanol in 100 mM potassium phosphate buffer (pH 7.4). The concentrations of the compounds inducing 50% inhibition of enzyme activity (IC₅₀) were determined as mean values ± SEM of three independent measurements. The data were analysed using the OriginPro software (OriginPro, 2020, OriginLab Corporation, Northampton, Massachusetts, United States). Active compounds were classified as either potent (IC₅₀ < 10 μM), moderate (IC₅₀ 10–100 μM) or weak inhibitors (IC₅₀ > 100 μM) based on their IC₅₀ values (Moodie et al., 2019).

The active compounds from the first screen and their analogs were evaluated against five microfouling organisms, including marine bacteria, Vibro harveyi (DSM 19623, Halomonas aquamarina (DSM 30161) and Pseudoalteromonas citrea (DSM 8771), and the two microalgae Porphyridium purpureum (AC122) and Exanthemachrysis gayraliae (AC15). The adhesion and growth inhibiting potential against the microfoulers was investigated in a 96-well format at 10 and 100 μg/mL employing the methodology described by Hellio et al. (2015) Briefly, stock solution of the selected compounds (2 mg/mL) were prepared in 20% ethanol and diluted to 10 and 100 μg/mL. Aliquots (100 μL) of diluted compound were distributed in microplates and evaporated at 25°C under vacuum. For the marine bacteria, 100 μL of stock culture at 2 × 10⁸ cells/mL in peptone marine broth were added to each well. Bacterial growth was monitored at 620 nm after incubation for 48 h at 25°C. The bacterial adhesion was quantified at 545 nm after crystal violet staining of residual biofilm after careful washing of the wells. For microalgae, 100 μL of algal culture (0.1 mg chlorophyll a/L) in F/2 medium were added to the wells as described for the bacterial growth experiments above. The culture was maintained for 5 days at 20°C and the microalgal growth was quantified via fluorescent analysis (excitation at 485 nm, emission at 645 nm) of chlorophyll a released upon addition of 100 μL of methanol to the wells. Chlorophyll extraction was also performed on the adhered cells, after using a multichannel pipette to eliminate all the non-attached cells, and quantified by fluorescence measurement to determine the algal adhesion.

The ability of the rescreened leads to inhibit the settlement and metamorphosis of the Pacific transparent sea squirt Ciona savignyi larvae was investigated. Compounds were screened for effects following methods described in Cahill et al. (2016). Adult brood stock were harvested from Nelson Marina, Nelson, New Zealand and held in a recirculating seawater system (18 ± 1°C, 33 ± 1 PSU) and fed bulk-cultured Isochorysis glabana until ready to spawn (as indicated by full egg and sperm ducts). Three individuals were surgically spawned, cross-fertilised in artificial seawater, and left for 24 h to hatch. Freshly hatched larvae were diluted in additional artificial seawater to yield final larval concentration of 3 ± 1 larvae/mL. Aliquots (7.1 mL) of this larval suspension were added to wells of 12-well tissue culture plates (Corning Costar) which contained serial dilutions of the test compound in small amounts (70 μL) of 20% ethanol. The range of concentrations of each compound assessed was 0.25, 0.5, 2.5, 5, 12.5, 25, 50, and 100 μg/mL. Blank and solvent controls were also included, and three replicates were performed in all cases (N = 3). Plates were held at 18 ± 1°C for 5 days, after which the number of successfully settled and metamorphosed individuals were counted. Dose-responses were modelled using Weibull curve fitting (as dictated by model fit) and nominal concentration estimates that reduced settlement and metamorphosis relative to the controls by 50% (IC₅₀) using R statistical software (R Core Team, 2020).

Fifty models were built using the MODELLER software package (Fiser and Šali, 2003). The model with the lowest objective function was chosen for further refinement and validation. Stereochemistry of the model was evaluated using a Ramachandran plot. The plot identified that 97.9% of the amino acids were in their highly preferred regions while the remaining 2.1% were preferred observations. No amino acids were positioned in questionable locations. The ProSA server result generated a Z-score value of –9.49 indicating that the overall model quality is more comparable to an NMR structure rather than an X-ray structure. Given the low sequence similarity a lower Z-score was expected. However, given the limited options the homology model was used for the virtual screen.

Molecular docking has been employed in several studies to develop an understanding of novel cholinesterase inhibitors for neurological applications (Correa-Basurto et al., 2007; Farrokhnia and Nabipour, 2014; Moodie et al., 2019). To probe a virtual screening approach for generating antifouling leads, compounds with theoretical binding to the active site of an AChE homology model of a marine fouling organism (C. gigas) were searched for from 10,000-compound library. The compounds from the screening library were selected, prepared, and docked to the AChE homology model to identify potentially active compounds for further analysis.

Selection of compounds to proceed to the next round of screening was based on the scoring threshold generated for each scoring function. The first-round thresholds of 31 (GS), 41(CS), 43(PLP), and 21(ASP) resulted in a total of 1,058 compounds continuing for the second round of screening. The selected compounds were then redocked to the active site of AChE with more stringent scoring thresholds set at 51 (GS), 25(CS), 60(PLP), 35(ASP) resulting in 201 compounds with sufficient binding scores. These compounds were visually inspected and further evaluated based on their pose and bonding interactions within the binding pocket. Finally, 39 compounds displaying good theoretical fit were purchased from the commercial vendor for initial testing in the biological assays (Figure 1).

The initial leads were tested for their ability to inhibit electric eel AChE. Compounds 3, 12, 14, 17, 21, 22, 26, 28, 35, 36, and 39 were shown experimentally to be too hydrophobic to allow dissolution in either water, ethanol or DMSO which excluded them from the in vitro evaluation. Of the remaining 28 compounds, five displayed either weak (IC₅₀ > 400 uM, compounds 6 and 9) or moderate inhibitory activity (IC₅₀ 10–100 μM, compounds 20, 29, and 37). The remainder of compounds were inactive. The five active compounds (6, 9, 20, 29, and 37) were used as the basis for further activity refinement via analog searches of the active scaffolds. A search for all commercially available analogs was conducted and 19 additional compounds were subsequently purchased for testing after taking compound hydrophobicity into account (ClogP < 4.0) as a key selection criterion (Figure 2).

When the 19 compounds were rescreened for eeAChE, hrAChE, and BCh inhibitory activities, seven analogs displayed moderate or high inhibitory activity in a dose-response dependent manner (Figure 3 and Table 1). IC₅₀ values against eeAChE ranged from 3.36 ± 0.39 μM for 20e to 169.54 ± 0.91 μM for 37a. Most compounds were active against eeAChE only and only two compounds (20c and 37) were active inhibitors of all three enzymes.

The activity seen for the most active compounds is high (Moodie et al., 2019) and shows how the virtual screening is able to guide the discovery of novel structures. For this study, the analog design/selection was based on commercially available compounds and no chemical modifications or synthetic procedures were employed to modify the leads. Given the inhibitory activities observed, these lead structures identified could form the foundation for future rational design and structure-activity relationship studies.

No suitable analogs of 6 and 29 were commercially available but several analogs of 9 with structural diversity in the N-alkyl chain were included in the second screen. Of these analogs (9 – 9j), only 9j displayed pronounced cholinesterase inhibition with a potent inhibitory activity against eeAChE of 9.44 ± 0.85 μM. This experimental observation was supported by subsequent docking studies. Docking predictions indicated consistent poses for the non-active compounds based on the four scoring functions used to generate results. Compound 9j had a different predicted pose (Figure 4), showing the opposite pattern, predicting a hydrogen bonding interaction with SER124 and Pi Stacking interactions with TRP83, and TRP 331.

Four of the five chromenone analogs of compound 20 displayed positive results in the cholinesterase bioassays with only terminal amide 20a being inactive. Interestingly compound 20b (49 μM) and 20e (3 μM) appears to be selective for eeAChE while 20 and 20d and showed activity against both eeAChE and hsBChE indicating they may be a less selective inhibitors. Finally, 20c showed a high activity against all enzymes illustrating a high general cholinesterase inhibitory profile.

Selected analogs were redocked to investigate the difference in activity seen between evaluated enzymes. Compound 20b showed a consistent predicted binding pose by all four scoring functions. The chromenone moiety was positioned deep in the pocket with the tail facing the surface (Figure 5). The chromenone carbonyl moiety has a predicted hydrogen bonding interaction with GLY119. The ester also has a predicted interaction with SER124 positioned in the middle of the pocket. The chromenone moiety also features multiple hydrophobic interactions with TRP83.

Docking results of compound 20d showed a predicted shift in position, aligning the chromenone moiety closer to allow for hydrogen bonding interactions with both SER198 and SER121 (Figure 6). The shift in positioning now indicates that the hydrophobic interactions of this moiety are taking place with TRP331. The interaction with SER198 may be the reason for increased activity in the eeAChE assay and new activity in the hsBChE assay, as the amino acid is known to play a part in catalytic activity.

Compound 20e has structural similarities to 20d, featuring a trifluromethyl group on the chromenone in place of the phenyl ring. This alternation reduced activity against hsBChE but increased activity for eeAChE. The pose prediction for 20e compound varies from 20d with the chromenone moiety rotated and predicted to have hydrogen bonding interactions with GLY119 and TYR132. The pi-pi interactions appear to stack more compared to 20d. The impact of the trifluoromethyl can be observed when comparing 20 to 20e, as the only difference is the substitution of the methyl with trifluoromethyl, resulting in a large increase of activity against eeAChE. This observation implies that the change has either altered the polarity of the compound, resulting in a less stable energy conformation within the pocket, or the halogen interactions and size had been more favourable. Chromenone class compounds are known for their activity in human neurotransmission (Piazzi et al., 2003; Najafi et al., 2016) so it is interesting that the virtual screen has generated compounds that are not active against the hrAChE which implies that these compounds may be worth further exploring.

Of the two analogs of compound 37 evaluated, only 37a displayed any activity, representing a reduction in comparison to the generally active 37. Both compounds share a very similar pose, indicating that the difference in activity could be due to the loss of the nitro group.

Having established a wide range of inhibitory activities against three cholinesterase enzymes, the compounds were subsequently experimentally assessed in vivo against five microfoulers and one macrofouler. Microfouling can cause microbially induced corrosion and it also leads to loss of hydrodynamic properties despite being caused by microscopic organisms. The microfouling screen assessed settlement/adhesion and growth of selected marine bacteria and microalgae. Both the active compounds from the second cholinesterase screen as well as the inactive ones were included. No reduction in either the settlement or growth of the five microfouling organisms was observed over the course of the experiments (data not shown) suggesting that the compounds are inactive at the two concentrations evaluated (10 and 100 μg/mL). While no specific antifouling activity was observed, the lack of effect on the growth of the microorganisms also indicate that these compounds are not generally toxic.

The 11 active compounds (Table 1) were also evaluated as inhibitors for the metamorphosis of larvae of the seasquirt C. savignyi, an invasive macrofouler that is problematic in scenarios including ship hull fouling and aquaculture. Out of the 11 compounds, 9j was shown to be a potent inhibitor with an IC₅₀ of 4.60 ± 0.39 μg/mL as determined from the dose-response curve (Figure 7). None of the other compounds displayed any activity against C. savignyi at the employed concentrations.

While 9j is clearly a potent novel inhibitor of both C. savignyi larval metamorphosis and eeAChE, none of the other active cholinesterase inhibitor analogs were active in the antifouling screens. The initial lead structure 9 was only weakly active and it is likely that 9j would not have been discovered without inclusion of analogs after the initial in vitro screen. Several of the other inactive antifouling compounds display similar or stronger cholinesterase inhibitory potential in vitro (20c, 20d, and 20e) and cholinesterase inhibition at low micromolar concentrations and our study shows that this does not automatically equal a strong antifouling effect even though several natural products have been reported with these dual bioactivities. Key structural requirements are at play and these warrants further research. In addition, our study illustrates the importance of performing in vivo experiments to verify effects predicted from in silico and/or in vitro screening experiments.

The fact that none of the compounds were active against the microfoulers was more expected as these primitive microorganisms lack the nervous system of the eukaryotes. Prokaryotic cholinesterases exist, but despite catalysing the same chemical transformation, they share little structural homology with the eukaryotic counterparts and overlapping binding preferential of compounds from the virtual hrAChE screen is not expected (To et al., 2020). Nevertheless, this study highlights that the cholinesterase enzyme family represent a valid target for several types of antifouling organisms.