1-Azinyl-1’-Alkenylferrocenes with Anticholinesterase, Antioxidant, and Antiaggregating Activities as Multifunctional Agents for Potential Treatment of Alzheimer’s Disease

1. Introduction

A promising scaffold for the synthesis of new biologically active compounds is ferrocene (Fc). Its unique physical and chemical properties, such as inherent stability, resilience in biological environments, strong redox characteristics, and low toxicity, make it widely applicable in medicinal chemistry [1,2,3,4].

Although Fc itself is not biologically active, its presence can significantly influence the properties of organic molecules into which it is integrated [5,6]. Adding an Fc fragment to parent compounds can induce substantial alterations in physicochemical characteristics such as solubility, lipophilicity, and enhanced biological activity. Noteworthy examples include ferrocifen, an Fc-based analog of the anticancer drug tamoxifen, and ferroquine, an Fc derivative of the antimalarial drug chloroquine [6,7].

A functionalized ferrocenyl fragment with different heteroaromatic substituents, such as oxazolines, quinolines, indoles, triazoles, imidazoles, and dihydropyrimidines, has been explored recently for its potential biological applications [2,3,4,7,8,9,10]. Among Fc derivatives, compounds exhibiting antitumor, antiviral, antimalarial, antibacterial, antioxidant, and antifungal activities have been identified.

Despite these advances in many areas of therapeutics, the application of Fc derivatives as agents for neurodegenerative disease treatment, particularly Alzheimer’s disease (AD), has been less thoroughly investigated. Given the critical role of oxidative stress in AD pathogenesis, Fc’s reversible oxidation capability, which can convert the electron-donating Fc group into an electron-deficient ferrocenium cation (Fc⁺), is an attractive property [1,6]. In addition, certain Fc derivatives have shown remarkable antiaggregating effects against β-amyloid [10,11,12]. Moreover, examples of Fc derivatives that inhibit cholinesterases are available, demonstrating moderate to strong inhibitory activity against acetylcholinesterase (AChE) and butyrylcholinesterase (BChE) [13,14], as well as selective BChE inhibition [15,16].

Additionally, Fc moieties are frequently seen as efficient carriers [1], that can cross the blood-brain barrier (BBB), enhancing targeted drug delivery to the brain [5].

Nitrogen-based heterocyclic compounds have been widely adopted in the development of AD drugs [17,18,19,20]. These include derivatives of pyridine [21,22,23,24], pyridazine [25,26], bipyridine [24,27], quinoline [28,29,30,31,32], and acridine [33], among other heterocycles.

On the [28,29,30,31,32] theme of developing multifunctional agents enhanced by the presence of an Fc moiety, combining an azinyl group, a redox-active Fc scaffold, and a highly conjugated two-dimensional π-electron system of an olefin fragment within one molecule offers a promising route for developing compounds with notable antioxidant properties [34,35,36,37]. Several foundational methods have been developed for synthesizing olefin-containing Fcs. For instance, metal-catalyzed alkenylation via C-H bond activation allows for the creation of planar chiral Fc olefins [38,39,40]. The Wittig reaction, which involves phosphonium salts reacting with aldehydes and ketones, is a conventional approach leading to vinylferrocenes [41,42]. The McMurry cross-coupling reaction, in the presence of Lewis acids, has been successfully employed for synthesizing analogs of the antitumor drug tamoxifen [43,44,45,46,47,48]. Another pathway for vinylferrocenes synthesis involves reducing aldehydes or ketones to the corresponding alcohols followed by dehydration [49,50,51]. Previous work from our group has established synthetic strategies for incorporating azinyl groups into metallocenes and related compounds [52,53,54,55,56,57,58], as well as proposing a regioselective acetylation approach for azinylferrocenes [59].

The present study endeavored to synthesize alkenyl derivatives of azinylferrocenes and evaluate their potential as anti-AD agents. New azinyl derivatives of alkenylferrocenes were synthesized. The chemical diversity and biological activity of synthesized 1-azinyl-, 1-azinyl-1’-vinyl, and 1-azinyl-1’-isopropenylferrocenes was investigated by the following endpoints and methods (see Methods for details): (1) Calculation of Tanimoto similarity coefficients (Tc) and normalized molecular volumes (Vn) of ligands; (2) Calculation of volumes of overlap of docked ligands; (3) Esterase profiles — the ability to inhibit target cholinesterases (AChE and BChE) and structurally related off-target carboxylesterase (CES); (4) Potential to block AChE-induced β-amyloid aggregation assessed by displacement of propidium from the AChE PAS; (5) Inhibition of self-aggregation of Aβ₄₂ (β-amyloid 1-42); (6) Antioxidant activity in ABTS and FRAP assays; (7) Predictions of protonation states, dimerization tendencies, and structure-antioxidant relationships employing QC calculations; (8) Mechanistic insights into protein binding and enzyme inhibition using enzyme kinetics and molecular docking with protein targets generated by Protenix, a reproduction code of AlphaFold3; and (9) Cytotoxicity of the lead compounds in reprogrammed human neuronal cells (MSC-Neu) and human neuroblastoma cells (SH-SY5Y) using the MTT assay.

2. Results and Discussion

2.1. Synthesis and Characterization of Products

Compounds 1a-e were synthesized following the procedure described by Chupakhin [52], with compound 1-(pyridazin-4-yl)ferrocene (1e) obtained in a yield of 53% (Table 1).

1-Azinyl-1’-acetylferrocenes 2a-e were produced via the established Friedel-Crafts acetylation protocol (Scheme 1), yielding satisfactory results as shown in Table 1 [59]. Next, these 1-azinyl-1’-acetylferrocenes 2a-e were subjected to a Wittig reaction at room temperature under an argon atmosphere using ylide 4, generated in situ from triphenylmethylphosphonium iodide 3 without the presence of n-BuLi (Scheme 1, Path A). This reaction resulted in the formation of 1-azinyl-1’-isopropenylferrocenes 5a-e in good yields [59]. Additionally, reducing acetylferrocenes 2a-e with NaBH₄ in methanol (MeOH) at 0 °C yielded alcohols 6a-e within 20 min and with high efficiency (Scheme 1, Path B).

It was observed that the synthesized 1-(α-hydroxyethyl)-1’-azinylferrocenes 6a-e could undergo dehydration in an inert atmosphere in the presence of Et₃N as a base, catalytic amounts of 4-dimethylaminopyridine, and methanesulfonyl chloride (mesyl chloride). This reaction successfully afforded the desired vinylferrocenes (7a-d) with up to 95% yield. However, it was not possible to isolate derivative 7e due to the formation of a mixture of unidentifiable compounds.

In the ¹H NMR spectra for 6a-e, Fc proton signals appeared between δ 3.90–5.08 ppm. There was an upfield shift for the CH₃-group proton signal (from δ 2.03–2.30 ppm to δ 1.23–1.36 ppm), along with a one-proton doublet at δ 1.76–5.32 ppm and a one-proton quartet or multiplet at δ 4.30–4.57 ppm, which corresponded to the OH and CH groups, respectively. For compounds 7a-d, the ¹H NMR spectra showed characteristic signals for 1,1’-disubstituted ferrocenes at δ 4.11–5.01 ppm, with additional proton signals for heterocyclic fragments between δ 7.03–9.06 ppm. Notably, the ¹H NMR spectra of 7a-d displayed a singlet corresponding to the CH group of the double bond at δ 5.35–6.17 ppm, and CH₂-group signals at δ 4.85–5.38 ppm.

In the ¹³C NMR spectra for 6a-e, no C=O signal was observed, while a CH group signal appeared at δ 65.10–65.65 ppm. The IR spectra for 6a-e featured a characteristic absorption band at ν = 3416–3312 cm⁻¹, indicating valence vibrations of the alcohol fragment, along with vibrations corresponding to secondary alcohols (ν = 1100–1038 cm⁻¹). Mass spectra of derivatives 6a-e and 7a-d showed a peak for the molecular ion.

We developed a high-yield method for synthesizing azinyl-containing alkenylferrocenes. Additionally, isopropenylferrocene (8) [60] and vinylferrocene (9) [51] were synthesized through established methods to facilitate comparative analyses of biological properties.

Quantum chemical (QC) calculations (see Supplemental Information, Section S4; Table S1) predicted that under the conditions used to assess the endpoints summarized in Table 2, the Fc derivatives with pyridine, bipyridine, quinolone and acridine ligands would be fully or partially protonated, while those with pyridazine ligands would be neutral. In addition, QM calculations predicted that the Fc derivatives with pyridine, quinoline, or acridine ligands would display varying degrees of dimer formation stabilized by a HPO₄^2- anion in phosphate buffer for all endpoints in Table 2 except propidium displacement, for which the dimers would be stabilized by two chloride anions in Tris-HCl buffer.

The chemical diversity of the Fc derivatives and reference compounds was quantified in two ways: (1) Calculating their Tanimoto similarity coefficients (Tc) based on 3D alignments of shape and pharmacophores relative to the lead compound 7b; and (2) Calculating normalized van der Waals volumes (Vn) relative to the compound with the largest volume in each group. These results are shown in Figure 1.

Note that in Figure 1A for the monomeric forms of the molecules, the compound with the largest volume was propidium. In contrast, when the dimeric forms of the molecules were considered, the acridine Fc derivatives 5c2p, 7c2p, and 1c2p (Figure 1B) and 5c2t, 7c2t, and 1c2t (Figure 1C) displayed the largest volumes.

When ranking the compounds by their Tc combo scores relative to the 7b monomer (Figure 1A), 7b2p dimer (Figure 1B), or 7b2t dimer (Figure 1C), the quinoline derivatives were grouped together either as monomers or dimers, the pyridine derivatives were grouped together as dimers, and the acridine derivatives were grouped together as dimers in Tris-HCl buffer. As expected, the reference compounds (BNPP, donepezil, myricetin, propidium, and tacrine) exhibited extremely low Tc color values, reflecting their lack of pharmacophoric similarity to the Fc-containing molecules.

2.2. Biological Investigations

2.2.1. Inhibition Studies of HsAChE, EcBChE, and SsCES. Structure-Activity Relationships

For all synthesized azinylferrocenes (azinyl-Fc) 1a-e, 5a-e, 7a-d, and their basic pharmacophores — Fc (Fc), 1-alkenylferrocenes: 8 (Fc–C(CH₃)=CH₂, isopropenylferrocene) and 9 (Fc––CH=CH₂, vinylferrocene) — the esterase profile was evaluated. These determinations and analyses indicate the relative capability of the compounds to inhibit three serine esterases: acetylcholinesterase (AChE, EC 3.1.1.7) and butyrylcholinesterase (BChE, EC 3.1.1.8) — both cholinergic targets — as well as carboxylesterase (CES, EC 3.1.1.1), a structurally related off-target enzyme. This profiling allows for the assessment of primary pharmacological effects based on AChE and BChE inhibition, alongside potential adverse effects due to CES inhibition, as CES is known to hydrolyze many ester-containing drugs [61,62,63].

In these experiments, human (Homo sapiens) erythrocyte AChE (HsAChE), equine (Equus caballus) serum BChE (EcBChE), and porcine (Sus scrofa) liver CES (SsCES) were used because of their ready availability, relatively low cost, and high sequence homology of EcBChE and SsCES to the corresponding human enzymes [62]. Note that for the experimental studies with porcine liver CES, a specific isozyme was not used; hence, it is designated as SsCES. However, for the molecular docking studies, a specific isoform was used, which is designated as SsCES1.

The selection of tacrine, an effective inhibitor of both AChE and BChE, and bis-4-nitrophenyl phosphate (BNPP), a selective CES inhibitor, as reference compounds, helped contextualize the activity of the test compounds.

As shown in Table 2, across the three groups of Fc derivatives, compounds with quinoline (1b, 5b, 7b) and bipyridine (1d, 5d, 7d) substituents demonstrated effective inhibition of BChE within the micromolar range. Bipyridine derivatives were also found to be potent AChE inhibitors across all groups — compounds 1d, 5d, and 7d. Quinoline derivatives exhibited significant anti-AChE activity in the isopropenyl-Fc (5b) and vinyl-Fc (7b) groups, with 10 times less activity observed for compound 1b, which lacked an alkenyl substituent.

All compounds studied displayed minimal inhibition of CES, a structurally similar enzyme to the other esterases in Table 2, but functionally distinct by being responsible for hydrolyzing a wide range of ester-containing drugs. This low CES inhibition suggests a lower risk of adverse drug interactions during potential clinical applications [61].

2.2.2. Kinetic Studies of HsAChE and EcBChE Inhibition

The mechanism of HsAChE and EcBChE inhibition by the synthesized compounds was further examined with the most active compound 7b. To determine the relationship between enzyme activity and inhibitor concentration at different substrate levels, we conducted kinetic assays. The kinetic data analysis, using Lineweaver–Burk plots, is shown in Figure 2A,B for HsAChE and EcBChE, respectively.

It can be seen that when compound 7b binds to HsAChE and EcBChE, V_max and K_m both change, which is typical for a mixed type of inhibition. The competitive (K_i) and noncompetitive (αK_i) components of inhibition constant by compound 7b were 1.92±0.13 µM and 3.05±0.18 µM for HsAChE and 2.77±0.22 µM and 4.58±0.27 µM for EcBChE.

2.2.3. Molecular Docking of Fc Derivatives and Reference Compounds into HsAChE

As shown in Table 2, the lead compound 7b, a quinoline derivative, potently inhibited HsAChE, while compound 7c, an acridine derivative, was a poor inhibitor of this esterase. To gain mechanistic insight into these contrasting inhibitory behaviors, molecular docking into HsAChE was performed with the monomeric and dimeric forms of these compounds in separate virtual experiments.

As can be seen in Figure 3, the dimers of both 7b and 7c bind at the mouth of the HsAChE gorge. Residue Y72 forms a hydrogen bond with a phosphate oxygen in dimeric 7b, and Y72 and W286 have hydrophobic interactions with one of the quinoline rings in dimeric 7b (Figure 3A). The phosphate group in dimeric 7c accepts a hydrogen bond from T75 and one of the acridine rings in 7c has hydrophobic interactions with Y72 and W286 (Figure 3B).

In contrast to the peripheral binding of the dimers, the monomers of 7b and 7c bind within the catalytic active site (Figure 3). The quinoline ring in the 7b monomer has a π-T interaction with residue Y124 and hydrophobic interactions with W286, F297, and Y337. The 7c monomer forms a hydrogen bond with Y341, π-T stacking interactions with Y337 and Y341, and hydrophobic interactions with D74, W86, F338, and Y341.

Both the 7b and 7c monomers had an Fc carbon atom within 4.5 Å of the CB carbon of the active site S203 residue, which is shown as a reference point (Figure 3A,B), but these proximities were not counted as hydrophobic interactions due to the 4.0 Å cutoff distance that was employed.

Experimental enzyme inhibition studies (Table 2) demonstrated that 7b is a potent inhibitor of HsAChE, while 7c is not. These relative inhibitory potencies, coupled with the docking poses obtained for the monomeric and dimeric forms of 7b and 7c (Figure 3), indicate that the monomer is the dominant species for 7b and the dimer is the dominant species for 7c.

It is apparent from visual inspection of the docking poses in Figure 3 that the monomeric forms of each of the ligands do not overlap with their respective dimeric forms. However, this case-by-case qualitative analysis can be extended to all possible pairs of a set of docked ligands and quantified by computing a matrix of their normalized molecular intersection volumes expressed as percent overlap.

Figure 4 shows such a matrix as a heatmap of percent overlap for a set of 21 ligands docked into HsAChE. The ligands included the following molecules: dimeric forms for the pyridine, quinoline, and acridine Fc derivatives; monomeric forms for the bipyridine and pyridazine Fc derivatives; the three types of parent Fc moieties, and four reference compounds. It can readily be seen from the shaded blocks that the dimers have substantial overlap with dimers, and monomers have substantial overlap with monomers. Likewise, the unshaded or lightly shaded blocks show that dimers tend to have minimal or no overlap with monomers. Because this heatmap is a similarity matrix, the elements on the diagonal are all 100%.

Figure 5 is a mixed heatmap showing the percent overlap between two sets of ligands docked into HsAChE. One set consists of all monomers; the other set includes monomers and dimers. Thus, unlike Figure 4, which is a similarity matrix, Figure 5 is not a similarity matrix; therefore, the elements on the diagonal are not 100%. The mixed heatmap allows visualization of the degree to which a monomeric form of a ligand overlaps with the corresponding dimeric form of the ligand. For example, the monomers 7b and 7c have zero overlap with their corresponding dimers, 7b.2 and 7c.2, respectively, in agreement with visual inspection of the docking poses in Figure 3. On the other hand, the monomeric form of compound 7b has 32.8% and 21.2% overlap with donepezil and tacrine, respectively, known inhibitors of HsAChE, which is consistent with the relatively high inhibitory potency of compound 7b against HsAChE reported in Table 2 and in agreement with the conclusion that the dominant form of this molecule is monomeric.

2.2.4. Molecular Docking of Fc Derivatives and Reference Compounds into EcBChE

Figure 6 shows the 7b monomer and 7c dimer docked into EcBChE. As with HsAChE, the 7b monomer binds to the catalytic active site, and the 7c dimer binds to the peripheral rim of the gorge.

In the catalytic active site of EcBChE, the quinoline ring of monomeric 7b forms parallel π-stacking interactions with residue Y332 as well as hydrophobic interactions with residues F329 and Y332. In addition, the Fc moiety of 7b has a hydrophobic interaction with residue W82 (Figure 6).

On the peripheral rim of the EcBChE gorge, the phosphate group in the 7c dimer forms a hydrogen bond with residue D283, while the 1’-vinyl group on one of the Fc moieties has hydrophobic interactions with Q119. One of the acridine rings also has a hydrophobic interaction with V277 (Figure 6).

The heatmaps of percent overlap of ligands in EcBChE were similar to those for human AChE. However, in the case of EcBChE, the dimeric form of compound 1c exhibited overlaps ranging from 17.8% to 24.3% with monomeric species 1d, 5d, 7d, donepezil, and propidium (Figures S6.1 and S6.2).

2.2.5. Molecular Docking of Fc Derivatives and Reference Compounds into SsCES1

As shown in Table 2, BNPP was capable of potently inhibiting SsCES, but none of the Fc derivatives was an effective inhibitor of this enzyme. Consistent with the experimental results, molecular docking showed that BNPP binds to the catalytic active site of SsCES1, but both the monomeric and dimeric forms of Fc derivatives were found to bind near the enzyme surface, as illustrated in Figure 7 for BNPP and the monomeric and dimeric forms of compound 7c.

As can be seen in Figure 7, BNPP forms extensive interactions with the active site residues of SsCES1. These contacts include hydrogen bonds from G144 and S222 to phosphoryl oxygens of BNPP as well as hydrophobic interactions with T98, F102, L256, F318, L359, and F425.

The monomeric form of compound 7c entered into interactions with only three SsCES1 residues: hydrophobic interactions with K302 and D307, along with a hydrophobic interaction and a hydrogen bond with T305. The dimeric form of 7c exhibited only a single SsCES1 contact — a hydrophobic interaction with K302. Note that the docking poses of the monomeric and dimeric forms of compound 7c overlap with each other (Figure 7).

Heatmaps of percent overlap of docked ligands in SsCES1 differed markedly from those for HsAChE and EcBChE by showing fewer substantial overlaps (Figures S6.3 and S6.4). There were no overlaps between BNPP and other ligands except Fc (29.6%) and 9 (34.6%) — this might account for the 4.2% and 23.4% inhibition of SsCES that was produced by a 20 μM concentration of these compounds (Table 2).

2.2.6. Displacement of Propidium from the EeAChE PAS

To assess the potential of the synthesized compounds as inhibitors of the proaggregating activity of AChE, a fluorescence-based assay was employed to evaluate their ability to competitively displace propidium [64,65,66]. Propidium, a selective ligand of the AChE peripheral anionic site (PAS), is known to effectively inhibit AChE-induced β-amyloid aggregation [67].

As shown in Table 2, at a concentration of 20 µM, quinoline derivatives (1b, 5b, 7b), acridine derivatives (1c, 5c, 7c) and bipyridine derivatives (1d, 5d, 7d) from all three Fc groups, along with compounds 7a containing a pyridine substituent, were able to displace propidium. Their level of displacement was comparable to that of the reference compound, donepezil (11.9±0.9%), indicating that these compounds are able to bind to AChE PAS and may act as blockers of AChE-induced β-amyloid aggregation [67,68,69].

Taking the displacement level of donepezil as 100% and expressing the displacement level of the active Fc compounds as a percentage of the donepezil value, the mean ± SEM of the 10 Fc derivatives that had non-zero displacement values was 84.1 ± 3.33%. The Fc compounds that had displacement values of zero were 1a, 1e, Fc, 5a, 5e, 8, and 9. From this analysis, it appeared reasonable to hypothesize that the ability of a compound to displace propidium from the EeAChE PAS could be treated as an all-or-nothing phenomenon, suggesting a simple binary classification of compounds into active and inactive categories. Furthermore, it also seemed reasonable that the ability of a compound to displace propidium from the EeAChE PAS would depend upon two principal variables: (1) the compound partially occupying the same volume otherwise filled by the propidium molecule; and (2) the binding energy (-ΔG, where ΔG = free energy of binding) of the compound to the EeAChE PAS. These two variables are readily obtained from the results of molecular docking, described in the next section.

2.2.7. Molecular Docking of Fc Derivatives and Reference Compounds into EeAChE

Figure 8 shows the results for molecular docking of propidium, compound 9, monomeric compound 7b and dimeric compound 7b into EeAChE. Because the propidium displacement experiments were carried out in Tris-HCl buffer, the dimeric form of compound 7b stabilized by chloride anions was used for docking rather than the dimeric form stabilized by a phosphate anion as was done for docking into HsAChE, EcBChE, SsCES1, and HsAβ₄₂.

In Figure 8A-C, propidium binds to the PAS of EeAChE near landmark residues S287, G288, and Y71. Not shown are extensive contacts including parallel π-stacking interactions with W281 and Y336, a hydrogen bond with D73, and hydrophobic interactions with Y71, W261, L284, F292, F333, and Y336. Compound 9 binds to the CAS of EeAChE where it has only two hydrophobic interactions with a single residue, W85. In Figure 8B, propidium is overlaid with the monomeric form of compound 7b, which has a π-T stacking interaction with Y123, two parallel π-stacking interactions with Y336, and hydrophobic interactions with Y71, D73, F333, and Y336. Figure 8C shows propidium overlaid with the dimeric form of 7b, which has two π-T interactions with W281 and hydrophobic interactions with Y71, W281, and L282.

The percent overlap heatmaps for reference compounds as well as the monomeric and dimeric forms of the Fc derivatives docked into EeAChE are shown in Figures S6.5 and S6.6. The data in these matrices are especially important for understanding and making predictions about propidium displacement from the EeAChE PAS by the study compounds. In this regard, it should be noted that significant differences exist between the heatmaps for HsAChE and EeAChE, particularly with respect to the overlap patterns between propidium and monomeric vs dimeric forms of the Fc derivatives. For example, in HsAChE, propidium overlaps were greater with monomers than with dimers when molecular docking was carried out with phosphate-stabilized dimers of the a, b, and c types. However, when all compounds were docked as monomers into HsAChE, the propidium overlaps were more evenly distributed across the range of compounds (Figure 4 and Figure 5).

In contrast, when molecular docking was conducted against EeAChE with chloride-stabilized dimers of the a, b, and c types of compounds with the other types docked as monomers, the overlaps with propidium were markedly higher for the dimeric Fc compounds. However, when docking was performed with all ligands as monomers, propidium overlaps were low for 12 of the 21 compounds (0 – 4.1%), and higher for 1b, 5b, 7b; 5c, 7c; 5d, 7d; and donepezil (11.4 – 35.3%) (Figures S6.5 and S6.6).

To test the hypothesis that information gleaned from molecular docking, namely, propidium overlap volumes and binding energies, could be combined to predict the ability of compounds to displace propidium from the PAS of EeAChE, binary classification models were constructed using linear discriminant analysis (LDA). To do this, the “ground truth” classifications were assigned based on the experimental results, where a displacement based on the reference compound, donepezil, was taken as 100%. On this basis, a compound that had a displacement value > 50% was considered active. The inactive Fc compounds each had a displacement value of zero. The only reference compound other than donepezil for which there was an experimentally determined value was tacrine, which scored 26%; accordingly, this compound was deemed inactive. When running the models, the ground truth classifications and the propidium displacement data were not included in the analysis. The only data used were the propidium overlap volumes and affinities (binding energies) obtained from molecular docking. After the LDA predictions were made, the ground truth experimentally determined classifications were used to assess the prediction accuracy. Figure 9 shows the results of the LDA model.

By using only two variables obtained from molecular docking (percent overlap with propidium and binding energy), LDA was able to distinguish between active and inactive compounds for propidium displacement from the EeAChE PAS with 94.7% accuracy (Figure 9). One compound (7a) out of 19 was misclassified. Future work will determine the extent to which an LDA binary classifier is a successful predictor of propidium displacement by other classes of compounds. It is noteworthy that in this case, the best results were obtained by using the binding affinities for monomers rather than dimers. This suggests the possibility that the chloride-stabilized dimers were not as stable as the phosphate-stabilized ones, which is in agreement with the QC predictions (Table S3).

2.2.8. Inhibition of HsAβ₄₂ Self-Aggregation

The potential of compounds from all three groups (1a-e, 5a-e, 7a-d) to inhibit HsAβ₄₂ self-aggregation was evaluated using a Thioflavin T (ThT) fluorescence assay.

The results, presented in Table 2, demonstrate that the studied Fc derivatives inhibited HsAβ₄₂ self-aggregation within a range of 43–90%. Compounds 1 with the azinyl group displayed slightly lower activity compared to 1-azinyl-1’-isopropenyl-Fcs 5 and 1-azinyl-1’-vinyl-Fcs 7. Maximum inhibition (71–90%) was observed in quinoline derivatives (1b, 5b, 7b), acridine derivatives (1c, 5c, 7c), and bipyridine derivatives (5d, 7d), as well as compound 7a with a pyridine substituent. These compounds inhibited HsAβ₄₂ self-aggregation to a degree comparable to or surpassing that of the reference compound, myricetin (79.4±6.3%). To characterize further the interactions of the study compounds with HsAβ₄₂, molecular docking investigations were conducted, as described in the next section.

2.2.9. Molecular Docking of Fc Derivatives and Reference Compounds to HsAβ₄₂

Molecular docking to HsAβ₄₂ was done with both the monomeric and dimeric forms of the Fc compounds as well as with the reference compounds: donepezil, myricetin, propidium, and tacrine; the results are summarized in Figure 10 and the text below.

Figure 10 illustrates the three distinct binding sites on HsAβ₄₂ for monomers characterized by 7c.1, myricetin (myr), and tacrine (tac)/7b.1. When molecular docking included the a, b, and c types of compounds as dimers, there were again three distinct binding sites characterized by 7c.2, myricetin (myr)/7b.2, and tacrine (tac).

When monomers were employed as ligands, all compounds except 1c, 5a, 5c, 7b, 7c, and tacrine docked to the myricetin site of HsAβ₄₂, defined by contacts with residues F4, H6, D7, S8, G9, F20, A21, E22, D23, S26, N27, and K28. When the a, b, and c types of ligands were in dimeric form, all compounds docked to the myricetin site except 1c, 5a, 5c, 7c, and tacrine. These bindings to the myricetin site can easily be read from the percent overlap heatmaps (Figures S6.5 and S6.6).

It is noteworthy that the myricetin contacts identified here, using blind molecular docking to the entire HsAβ₄₂ peptide, recapitulated residues found in other studies using molecular docking [70], molecular dynamics (MD) [71,72,73] and NMR [74]. Moreover, the present molecular docking was carried out using a predicted peptide structure using Protenix, an AlphaFold3 reproduction code [75], which generated a structure quite different from those reported from NMR studies [76,77] that were used in our previous investigations [78,79]. Our decision not to use previously published structures solved by solution NMR in the present study was influenced the fact that the NMR structures were determined in solutions that included various concentrations of the highly apolar solvent, hexafluoroisopropanol (HFIP) [76,77], known to promote helix formation. In contrast, our predicted peptide structure in the current investigation contained beta strands and loops but no helices.

However, although the majority of our compounds docked to the myricetin site, binding to this site was apparently not necessary for blocking self-aggregation of HsAβ₄₂. For example, the most potent blocker of self-aggregation was compound 5c (90.1% inhibition; Table 2). This compound, as well as 1c and 7c, which were all predicted to exist mainly in dimeric form, docked congruently to the same site on HsAβ₄₂, defined by residues Y10, L17, V18, F19, A21, E22, D23, and V24. No other compounds docked to this site.

The tacrine binding site on HsAβ₄₂ included residues A2, E3, F4, E11, and F20. Tacrine itself was a poor inhibitor of HsAβ₄₂ self-aggregation (5.9%; Table 2). Among other study compounds, only the monomeric forms of 1c, 7b, and 7c docked to the tacrine site, but in the phosphate buffer used for the self-aggregation assays, the dimeric forms of these compounds were likely favored. The dimeric form of compound 5a docked to the tacrine site, but this compound also made contacts with residues in the myricetin site, and the potency of 5a was relatively modest at 62.4% (Table 2).

Whereas the binding locus on a peptide or protein is a vital determinant of the biological function of a ligand, the potency of the ligand also depends on its binding energy [80]. As noted above in the case of propidium displacement by ligands from the EeAChE PAS, there was an apparent binary response that prompted prediction from a combination of propidium overlap volume and binding energy using LDA (Table 2; Figure 9). In contrast, inhibition of HsAβ₄₂ self-aggregation exhibited a graded response over a wide range from zero to 90.1% (Table 2), and although correlations with myricetin overlap volumes were not significant (p = 0.86 for monomers; p = 0.25 for dimers), there were strong correlations with the binding energies obtained from molecular docking of the monomers (r = 0.800, p < 0.0001) and dimers (r = 0.789; p < 0.0001), as shown in Figure 11. Although the r-value for the monomers was slightly higher than the value for the dimers, the difference between them was not statistically significant. Moreover, based on the QC predictions for the compounds in phosphate buffer, the expectation is that there would be a monomer-dimer equilibrium among the Fc derivatives, with dimers predominating in the following order: acridines > quinolines > pyridines (see Supplemental Information, Section S5).

2.2.10. Antioxidant Activity (AOA)

The primary antioxidant activity of the newly synthesized azinyl-Fcs (1a-1e), azinyl- isopropenyl-Fcs (5a-5e), azinylvinyl-Fcs (7a-7d) and their basic pharmacophores Fc and 1-alkenyl-Fcs 8 and 9 was evaluated using two standard assays. The first is the ABTS test, which measures the ability of a compound to scavenge the model radical cation ABTS^•+. The second is the Ferric Reducing Antioxidant Power (FRAP) test, which assesses the ability of a compound to reduce a particular Fe³⁺ complex to the Fe²⁺ state. The results of these assays, which are displayed in Table 3, provide insight into the antioxidant efficacy of the compounds based on their radical-scavenging and iron-reducing capacities.

Table 3 illustrates that azinyl-Fcs exhibit significantly higher primary antioxidant activity compared to Fc. All the azinyl-Fc derivatives tested displayed high antioxidant activity, matching or surpassing the standard antioxidant Trolox in both the ABTS and FRAP tests.

Among all tested groups, quinoline derivatives demonstrated the highest activity across both assays. In the ABTS test, quinoline derivatives were up to four times more active than Trolox, while in the FRAP test, they showed activity levels 2–3 times greater than Trolox. Conversely, the basic pharmacophores—Fc, isopropenyl-Fc 8, and vinyl-Fc 9, — did not exceed Trolox’s antioxidant activity.

2.2.11. QC Analyses of Antioxidant Activity (AOA)

To investigate trends in antioxidant activity (AOA), representatives of all three series of the synthesized Fc derivatives with varied substituents on the second cyclopentadienyl (Cp) rings of Fc were examined: Fc (Fc) and compounds 1a, 1b, and 1c, each containing different azinyl substituents of varying sizes (pyridine, quinoline, and acridine) on a single Cp ring. Isopropenyl-Fc 8 and derivatives 5a, 5b, and 5c, with an iso-propenyl substituent on one Cp ring and the same azinyl substituents on the other. Vinyl-Fc 9 and derivatives 7a, 7b, and 7c, incorporating a vinyl substituent on one Cp ring and the azinyl substituents on the other. In all three series, a significant increase in AOA was observed when a heterocyclic aromatic substituent was added to the second Cp ring.

QC calculations (see Supplementary Information for details) indicated that, under the FRAP test conditions, all compounds containing an aromatic heterocyclic substituent are protonated at the nitrogen atom of this substituent. For the ABTS test, compounds with a pyridine substituent (1a, 5a, 7a) showed partial protonation at the nitrogen atom, while those with quinoline (1b, 5b, 7b) and acridine (1c, 5c, 7c) substituents were fully protonated at the heterocyclic nitrogen atom. Compounds with the protonated nitrogen atom are marked with the subscript “p” for clarity, and all antioxidant properties were calculated based on the protonation states relevant to the experimental conditions.

To evaluate AOA in the FRAP test, both ionization potential (IP) and bond dissociation enthalpy (BDE) values in water were calculated (see Table 4). It was found that compounds lacking azinyl substituents (Fc, 8, 9) have relatively similar and low IP values, about 74.5 kcal/mol. However, the BDE values for these compounds varied widely, ranging from 88.1 kcal/mol (compound 8) to 111.6 kcal/mol (Fc). Given the similar and moderate AOA of these compounds in the FRAP test (0.8 – 1.1 TE) a single electron transfer (SET) mechanism is likely responsible for their antioxidant action.

In contrast to non-substituted Fc derivatives, protonated compounds with azinyl substituents (1a_p, 1b_p, 1c_p, 5a_p, 5b_p, 5c_p, 7a_p, 7b_p, 7c_p) exhibit higher IP values (about 80 kcal/mol) and substantially lower BDE values (67–69 kcal/mol). The observed increase in AOA upon the addition of azinyl substituents indicates an alternative mechanism, where AOA is determined by BDE values. Taking into account that Fe³⁺[TPTZ]₂ is reduced to Fe²⁺[TPTZ]₂ during the antioxidant reaction and that AOA correlates with BDE values, a concerted mechanism can be assumed. In this mechanism, the antioxidant molecule loses a hydrogen atom, with the electron being transferred to the cation Fe³⁺[TPTZ]₂ and the proton being transferred to the solvent molecule. A similar mechanism is realized for phenolic antioxidants in the FRAP test [81].

The lowest AOA of acridine-based compounds likely stems from the greater propensity of acridine compounds to dimerize [82]. QC calculations support this, showing that dimer formation increases with increasing ligand size in the series 1a_p, 1b_p, 1c_p (with pyridine, quinolone and acridine substituents, respectively) (see Supplementary Information, Section S5, Table S3).

For AOA evaluation in the ABTS test, IP and BDE values were calculated in ethanol (refer to Table 4). QC analyses indicated that changes in BDE and IP values in ethanol followed the same trends observed in water, aligning with similar AOA results in the FRAP and ABTS assays. The mechanism of antioxidant action of each compound is the same in both FRAP and ABTS tests. Compounds without azinyl substituents (Fc, 8, 9) showed relatively low IP values (about 82 kcal/mol, see Table 4) and reduced the ABTS^•+ radical through the SET mechanism.

In contrast, protonated compounds with azinyl substituents (1a_p, 1b_p, 1c_p, 5a_p, 5b_p, 5c_p, 7a_p, 7b_p, 7c_p) have low BDE values (67–69 kcal/mol, Table 4) and operate via the alternative mechanism similar to that in the FRAP test. In this process, an electron transfers to the ABTS^•+ radical while a proton is simultaneously transferred to the SO₄^2- anion in the solvent. Similar mechanism was realized in the ABTS test with conjugates of amiridine and salicylic acid derivatives [79]. The change in the AOA mechanism explains the significant increase in AOA with the addition of azinyl substituents.

QC calculations further revealed that, in the ABTS test, compounds with a pyridine substituent are only partially protonated, while compounds with quinoline and acridine substituents are fully protonated. This protonation difference explains the relatively lower TEAC values of pyridine-based compounds compared to quinoline-based ones. The decreased TEAC values observed for acridine-based compounds are attributed to their strong tendency toward dimerization, consistent with results obtained in the FRAP test (see Supplementary Information, Section S5, Table S3).

2.2.12. Cytotoxicity of Fc Derivatives

The cytotoxicity of lead compounds — 1b, 1d, 5b, 5d, 7b, 7d as well as reference compounds 10 and 11 [45]— was assessed using reprogrammed human neuronal cell cultures (MSC-Neu) (see Supplementary Information) and human neuroblastoma cell cultures (SH-SY5Y). Previously studied compounds 10 and 11 [45] were selected as reference compounds for cytotoxicity assessment as the most structurally similar to ferrocifen type anticancer agents [83,84]. Cytotoxic activity was assessed using the MTT test [85]. The range of compound concentrations was from 1 to 1024 μM and the incubation time was 24 h. For each compound, the half-lethal concentration parameter LC₅₀ (μM) was calculated. The results are presented in Table 5.

The results showed that the study compounds generally exhibited relatively low cytotoxicity, being substantially less toxic than the control ferrocifen-type compounds (Figure 12), with a more pronounced cytotoxic effect on human neuroblastoma tumor cells (SH-SY5Y) compared to neuronal cells (MSC-Neu).

3. Materials and Methods

3.1. Chemistry

The dehydration reactions and Wittig reactions were conducted under an argon atmosphere using standard Schlenk techniques [88]. ¹H-NMR (600 MHz) and ¹³C-NMR (151 MHz) spectra were recorded on a Bruker AVANCE NEO NMR spectrometer, while ¹H-NMR (400 MHz) and ¹³C-NMR (100 MHz) spectra were recorded on a Bruker AVANCE II NMR spectrometer. Chemical shifts are reported in δ values (ppm) using tetramethylsilane (TMS) as the internal standard, with CDCl₃ and DMSO-d₆ as solvents.

High-resolution electrospray ionization mass spectra were obtained for positive ions within the m/z range of 50–1300 Da on a Bruker maXis Impact HD high-resolution Q-TOF LC-MS/MS spectrometer. Samples were prepared in methanol and analyzed after partial chromatographic separation on a ZORBAX SB-C18 column (2.1 × 50 mm, 1.8 μm, 40 °C, methanol-water 9:1, 0.25 mL/min flow rate). The mass scale was calibrated using a calibration solution of HCOONa (10 mmol) in a water-propan-2-ol mixture (1:1).

IR spectra were recorded on a Perkin Elmer Spectrum One B Fourier-transform infrared spectrometer equipped with a diffuse reflection attachment. Melting points were determined using a Boetius heating stage.

Preparative column chromatography was performed on Merck silica gel 60 (particle size 0.063–0.200 mm) and Macherey-Nagel aluminum oxide 90 neutral. Reaction progress was monitored using thin-layer chromatography (TLC) on 0.2 mm aluminum oxide plates with a fluorescent indicator (Polygram^® Alox N/UV254, Macherey-Nagel) and 0.20 mm silica gel plates coated with fluorescent indicator F254 (ALUGRAM^® Xtra SIL G/UV254, pre-coated TLC sheets).

All solvents were purified according to standard procedures. Pyridazine, methanesulfonyl chloride, 4-(dimethylamino)pyridine, sodium tetrahydroborate, acetic anhydride, aluminum chloride, n-BuLi, Fc, acetylferrocene were purchased from Aldrich. Starting methyl(triphenyl)phosphonium iodide [89], bromoferrocene [90] azinylferrocenes 1a-d [52], isopropenylferrocene 8 [60], vinylferrocene 9 [51], compounds 2a-d and 5a-d [59], 10 [87] and 11 [45] were prepared according to the published procedures.

3.1.1. Synthesis of 1-(pyridazin-4-yl)ferrocene 1e

To a freshly prepared suspension of lithioferrocene, obtained from the solution bromoferrocene (7.58 mmol, 2 g) in dry diethyl ether (15 mL) и from a 1.6 M nBuLi solution in hexane (7.1 mL, 11.3 mmol) in Schlenk tube under an atmosphere of argon at 0 °C within 30 minutes, added pyridazine solution (15.16 mmol, 1.1 mL) in dry diethyl ether. Then reaction mixture was heating at 40 °C for 2 h. The product was obtained by column chromatography on Al₂O₃ using a mixture of hexane:EtOAc (1:1) as an eluent. The eluate was concentrated to dryness in vacuo and the residue was recrystallized from hexane.

Orange solid. Yield of 1e 1.06 g (53%). M.p. 147 °C. ¹H NMR (600 MHz, CDCl₃): δ – 4.09 (s, 5H, Cp), 4.52 (s, 2H, C₅H₄), 5.08 (s, 2H, C₅H₄), 7.35 (s, 1H, 5′-H), 7.51 (s, 1H, 3′-H), 8.99 (s, 1H, 6′-H) ppm. ¹³C NMR (151 MHz, CDCl₃): δ – 67.6 (C-Cp); 69.9 (CH-Cp); 70.9 (C-Cp′); 80.2, 123.5, 126.1, 148.9, 162.2 (pyridazine) ppm. IR (ν, cm^–1): 3085, 2961, 2857, 1787, 1668, 1581, 1484, 1415, 1377, 1327, 1288, 1258, 1110, 1006, 887, 793, 746, 707, 514. HRMS (ESI) m/z calcd for C₁₄H₁₃FeN₂⁺ [M+H]⁺: 265.0422; found 265.0423.

3.1.2. Synthesis of 1-acetyl-1’-(pyridazin-4-yl)ferrocene 2e

Acetic anhydride (3 mmol, 0.37 ml) was added to the solution of 1-(pyridazin-4-yl)ferrocene (1 mmol) 3e in dry CH₂Cl₂ (20 ml) at 0 °C, then for 2 hours AlCl₃ (3 mmol) was added in equal portions approximately every 15 min. The reaction was stirring for 24 h, then the reaction mass was poured into cold water and extracted with CH₂Cl₂ (3 ×25 ml). The extract was dried over Na₂SO₄, decanted and concentrated at reduced pressure. The residue was purified by flash column chromatography on Al₂O₃ (hexane:EtOAc 7:3, R_f = 0.1). The eluate was concentrated to a dry state in vacuo.

Red oil. Yield of 2e: 177 mg (58%). M.p. 88 ^oC. ¹H NMR (600 MHz, CDCl₃): δ – 2.12 (s, 3H, CH₃); 4.41-4.42 (m, 2H, C₅H₄); 4.52 (m, 2H, C₅H₄); 4.66-4.67 (m, 2H, C₅H₄); 5.08 (m, 2H, C₅H₄); 7.41-7.46 (m, 2H, 3’-H, 5’-H); 9.03 (d, 1H, J = 3 Hz, 6’-H) ppm. ¹³C NMR (151 MHz, CDCl₃): δ – 27.6 (CH₃); 68.9, 71.1, 72.2, 73.8 (CH-Cp); 80.6, 82.2 (C-Cp); 123.9, 126.3, 149.4, 160.1 (pyridazine); 201.6 (C=O) ppm. IR (ν, cm^–1): 2924, 1660, 1582, 1487, 1455, 1375, 1278, 1116, 889, 818, 750, 620, 538, 522, 501. HRMS (ESI) m/z calcd for C₁₆H₁₅FeN₂O⁺ [M+H]⁺: 307.0528; found 307.0527.

3.1.3. General Procedure for Formation of 1-isopropenyl-1’-(pyridazin-4-yl)ferrocene 5e

1.6 M n-BuLi solution in hexane (0.71 ml, 1.14 mmol) was added to a suspension of methyltriphenylphosphonium iodide 3 (1.25 mmol) in dry diethyl ether in a Schlenk tube under argon at room temperature. After 30 min stirring, a solution of 1-acetyl-1’-(pyridazin-4-yl)ferrocene 2e (1 mmol) in anhydrous diethyl ether (20 ml) was added. After stirring at room temperature for 24 h, reaction mixture was purified by flash column chromatography on Al₂O₃ (hexane:EtOAc 7:3, R_f = 0.25). The eluate was concentrated to dryness in vacuo.

Red solid. Yield of 5e: 191 mg (63%). M.p. 90 ^oC. ¹H NMR (600 MHz, DMSO-d₆): δ – 1.72 (s, 3H, CH₃), 4.15 (t, 2H, J = 2.0 Hz, C₅H₄), 4.32 (t, 2H, J = 2.0 Hz, C₅H₄), 4.47 (t, 2H, J = 2.0 Hz, C₅H₄), 4.62 (s, 1H, =CH₂), 4.94 (s, 1H, =CH₂), 5.04 (t, 2H, J = 2.0 Hz, C₅H₄), 7.57 (br. s, 1H, 5′-H), 7.75 (s, 1H, 3′-H), 8.98 (br.s, 1H, 6′-H) ppm. ¹³C NMR (151 MHz, DMSO-d₆): δ – 20.9 (CH₃); 67.0, 68.1, 70.0, 71.3 (CH-Cp); 81.3, 87.2 (C-Cp); 109.5 (CH₂); 124.1, 126.2, 149.0, 160.1 (pyridazine); 139.5 (C) ppm. IR (ν, cm^–1): 3075, 2918, 1665, 1662, 1579, 1486, 1375, 1288, 1216, 1181, 1112, 1081, 1020, 987, 885, 815, 746, 586, 517. HRMS (ESI) m/z calcd for C₁₇H₁₆FeN₂⁺ [M+H]⁺: 304.0657; found 304.0656.

3.1.4. Synthesis of 1-azinyl-1’-(α-hydroxyethyl)ferrocenes 6a-e (General Procedure)

To a solution of 1-azinyl-1’-acetylferrocene 2a-e (1 mmol) in MeOH (15 mL), solid NaBH₄ (5 mmol) was added in one portion at 0 °C. The reaction mixture was stirred at 0 °C for 20 min. Formed solution was poured into ice-cold water (20 mL), then extracted with ethyl acetate (3 × 30 mL). The collected organic layers were washed with brine (100 mL), dried over Na₂SO₄, decanted, solvent was removed under reduced pressure to afford the crude product. Isolated alcohols 6a-e were purified by flash column chromatography on SiO₂. The eluate was concentrated to dryness in vacuo.

1-(Pyridin-2-yl)-1’-(α-hydroxyethyl)ferrocene (6a).

Using 312 mg (1 mmol) 1-(pyridin-2-yl)-1’-acetylferrocene and 190 mg (5 mmol) NaBH₄, 1-(pyridin-2-yl)-1’-(α-hydroxyethyl)ferrocene was obtained as orange crude oil in 93% yield (285 mg, 0.928 mmol). R_f = 0.4 (TLC: hexane:EtOAc, 9:1). ¹H NMR (600 MHz, CDCl₃): δ – 1.32 (d, 3H, J = 6.5 Hz, CH(OH)CH₃), 3.90 (m, 1H, C₅H₄), 3.96 (m, 1H, C₅H₄), 4.10-4.11 (m, 1H, C₅H₄), 4.15-4.16 (m, 1H, C₅H₄), 4.38-4.39 (m, 1H, C₅H₄), 4.44-4.45 (m, 1H, C₅H₄), 4.56 (q, 1H, J = 6.5 Hz, CH(OH)CH₃ ), 4.80-4.83 (m, 2H, C₅H₄), 5.29 (s, 1H, CH(OH)CH₃), 7.10-7.13 (m, 1H, 5’-H), 7.30-7.31 (d, 1H, J = 7.0 Hz, 3’-H), 7.57-7.60 (td, 1H, J = 7.0 Hz, J = 2.0 Hz, 4’-H), 8.53-8.54 (m, 1H, 6’-H) ppm. ¹³C NMR (151 MHz, CDCl₃): δ – 25.5 (CH₃); 65.5 (CH); 67.2, 67.3, 68.2, 68.3, 69.6, 69.9, 70.2 (CH-Cp); 84.8, 96.9 (C-Cp); 120.5 (C3), 121.3 (C5), 136.7 (C4), 149.5 (C6), 159.0 (C2) ppm. IR (ν, cm^-1): 3086, 2968, 2924, 2857, 1720, 1588, 1564, 1499, 1425, 1368, 1278, 1151, 1100, 1024, 924, 875, 758, 742, 698, 624, 521. HRMS (ESI) m/z calcd for C₁₇H₁₈FeNO⁺ [M+H]⁺: 308.0732; found 308.0731.

1-(Quinolin-2-yl)-1’-(α-hydroxyethyl)ferrocene (6b).

Using 362 mg (1 mmol) 1-(quinolin-2-yl)-1’-acetylferrocene and 190 mg (5 mmol) NaBH₄, 1-(quinolin-2-yl)-1’-(α-hydroxyethyl)ferrocene was obtained as orange solid in 98% yield (350 mg, 0.980 mmol). R_f = 0.5 (TLC: hexane:EtOAc, 9:1). M.p. 79-81 °C. ¹H NMR (600 MHz, CDCl₃): δ – 1.34 (d, 3H, J = 6.5 Hz, CH(OH)CH₃), 3.93 (d, 2H, J = 12.0 Hz, C₅H₄), 4.14 (d, 2H, J = 25.0 Hz, C₅H₄), 4.50 (d, 2H, J = 30.0 Hz, C₅H₄), 4.56 (q, 1H, J = 6.4 Hz, CH(OH)CH₃), 5.02 (d, 2H, J = 18.0 Hz, C₅H₄), 5.32 (s, 1H, CH(OH)CH₃), 7.45 (d, 1H, J = 8.5 Hz, 4’-H), 7.48 (t, 1H, J = 7.6 Hz, 7’-H), 7.70 (t, 1H, J = 7.6 Hz, 6’-H), 7.75 (d, 1H, J = 7.5 Hz, 5’-H), 8.05 (d, 1H, J = 8.5 Hz, 3’-H), 8.10 (d, 1H, J = 7.5 Hz, 8’-H) ppm. ¹³C NMR (151 MHz, CDCl₃): δ – 25.4 (CH₃); 65.3 (CH); 67.6, 68.22, 68.24, 68.30, 68.33, 70.0, 70.4, 70.6 (CH-Cp); 84.4, 97.1 (C-Cp); 119.3 (C4); 125.9 (C7); 126.9 (C8a); 127.6 (C5); 128.5 (C8); 130.3 (C6); 136.5 (C3); 148.1 (C4a); 159.8 (C2) ppm. IR (ν, cm^-1): 3082, 2964, 1616, 1594, 1557, 1508, 1442, 1422, 1376, 1245, 1142, 1125, 1092, 1003, 826, 747, 519. HRMS (ESI) m/z calcd for C₂₁H₂₀FeNO⁺ [M+H]⁺: 358.0888; found 358.0891.

1-(Acridin-9-yl)-1’-(α-hydroxyethyl)ferrocene (6c).

Using 406 mg (1 mmol) 1-(acridin-9-yl)-1’-acetylferrocene and 190 mg (5 mmol) NaBH₄, 1-(acridin-9-yl)-1’-(α-hydroxyethyl)ferrocene was obtained as red solid in 96% yield (390 mg, 0.958mmol). R_f = 0.4 (TLC: hexane:EtOAc, 1:1). M.p. 59-60 °C. ¹H NMR (600 MHz, CDCl₃): δ – 1.36 (d, 3H, J = 6.0 Hz, CH(OH)CH₃), 1,76 (s, 1H, CH(OH)CH₃), 4,18 (s, 2H, C₅H₄), 4.25 (s, 1H, C₅H₄), 4.30 (s, 1H, C₅H₄) 4.57 (q, 1H, J = 6.4 Hz, CH(OH)CH₃), 4.66 (s, 2H, C₅H₄), 4.90 (s, 2H, C₅H₄), 7.51-7.53 (m, 2H, 3’-H, 6’-H), 7.73-7.75 (m, 2H, 2’-H, 7’-H), 8.22 (d, 2H, J = 9.0 Hz, 4’-H, 5’-H), 9.05 (d, 2H, J = 8.8 Hz, 1’-H, 8’-H) ppm. ¹³C NMR (151 MHz, CDCl₃): δ – 24.0 (CH₃); 65.6 (CH); 67.9, 68.0, 69.6 (2C), 71.0, 71.1, 73.9, 74.0 (CH-Cp); 81.8, 95.3 (C-Cp); 124.7 (C3, C6); 125.5 (C4a,C5a); 127.5 (C1, C8); 129.7 (C2, C7); 130.2 (C4,C5); 143.6 (C8a, C9a); 149.0 (C9) ppm. IR (ν, cm^-1): 3091, 2969, 2922, 1723, 1621, 1553, 1447, 1400, 1259, 1038, 912, 871, 827, 760, 672, 638, 484. HRMS (ESI) m/z calcd for C₂₅H₂₂FeNO⁺ [M+H]⁺: 408.1045; found 408.1045.

1-(2,2’-Bipyridin-6-yl)-1’-(α-hydroxyethyl)ferrocene (6d).

Using 383 mg (1 mmol) 1-(2,2’-bipyridin-6-yl)-1’-acetylferrocene and 190 mg (5 mmol) NaBH₄, 1-(2,2’-bipyridin-6-yl)-1’-(α-hydroxyethyl)ferrocene was obtained as orange solid in 98% yield (376 mg, 0.979 mmol). R_f = 0.45 (TLC: hexane:EtOAc, 6:4). M.p. 60-61 °C. ¹H NMR (600 MHz, CDCl₃): δ – 1.23 (d, 3H, J = 6.5 Hz, CH(OH)CH₃), 3.19 (s, 1H, CH(OH)CH₃), 3.99-4.00 (m, 1H, C₅H₄), 4.05-4.06 (m, 1H, C₅H₄), 4.11-4.12 (m, 2H, C₅H₄), 4.30 (q, 1H, J = 6.4 Hz, CH(OH)CH₃), 4.46-4.48 (m, 2H, C₅H₄), 4.98-4.99 (m, 1H, C₅H₄), 5.07-5.08 (m, 1H, C₅H₄), 7.31-7.33 (m, 1H, 5’’-H), 7.39 (dd, 1H, J = 8.0 Hz, J = 1.0 Hz, 5’-H), 7.73 (t, 1H, J = 8.0 Hz, 4’-H), 7.87 (td, 1H, J = 8.0 Hz, J =1.0 Hz, 4’’-H), 8.26 (dd, 1H, J = 8.0 Hz, J = 1.0 Hz, 3’-H), 8.52 (d, 1H, J = 8.0 Hz, 6’’-H), 8.67-8.69 (m, 1H, 3’’-H) ppm. ¹³C NMR (151 MHz, CDCl₃): δ – 24.5 (CH₃); 65.1 (CH); 67.5, 67.6, 67.8, 67.9, 68.6, 68.8, 70.2, 70.3 (CH-Cp); 84.2, 96.9 (C-Cp); 118.6 (C3); 120.1 (C5); 121.3 (C6’); 123.9 (C4’); 137.1 (C4); 137.6 (C4); 149.2 (C3’); 155.9 (C2’); 156.1 (C2); 158.4 (C6) ppm. IR (ν, cm^-1): 3085, 2968, 2924, 2855, 1970, 1733, 1666, 1563, 1491, 1426, 1368, 1311, 1258, 1230, 1155, 1091, 923, 903, 870, 821, 781, 743, 679, 474. HRMS (ESI) m/z calcd for C₂₂H₂₁FeN₂O⁺ [M+H]⁺: 385.0997; found 385.0998.

1-(Pyridazin-4-yl)-1’-(α-hydroxyethyl)ferrocene (6e).

Using 307 mg (1 mmol) 1-acetyl-1’-(pyridazin-3-yl)ferrocene and 190 mg (5 mmol) NaBH₄, 1-(pyridazin-3-yl)-1’-(α-hydroxyethyl)ferrocene was obtained as a red crude oil in 82% yield (253 mg, 0.821 mmol). R_f = 0.3 (TLC: EtOAc). M.p. 73-74 °C. ¹H NMR (400 MHz, CDCl₃): δ – 1.32 (d, 3H, J = 6.3 Hz, CH(OH)CH₃), 3.60 (s, 1H, CH(OH)CH₃), 4.01 (s, 2H, C₅H₄), 4.10 (d, 2H, J = 10.0 Hz, C₅H₄), 4.46-4.53 (m, 3H, C₅H₄, CH(OH)CH₃), 4.99 (s, 2H, C₅H₄), 7.33-7.36 (m, 1H, 5’-H), 7.47 (br.s, 1H, 4’-H), 8.99 (s, 1H, 6’-H) ppm. ¹³C NMR (100 MHz, CDCl₃): δ – 24.9 (CH₃); 65.3 (CH); 67.6, 67.8, 68.1, 68.9 (2C), 69.0, 70.9, 71.0 (CH-Cp); 80.8, 97.1 (C-Cp); 123.7 (C5), 126.4 (C4), 149.1 (C6), 162.3 (C3) ppm. IR (ν, cm^-1): 3099, 2973, 2925, 1581, 1486, 1401, 1325, 1289, 1227, 1185, 1113, 1089, 983, 883, 869, 752, 711, 590, 522, 477. HRMS (ESI) m/z calcd for C₁₆H₁₇FeN₂O⁺ [M+H]⁺: 309.0684; found 309.0682.

3.1.5. Synthesis of 1-azinyl-1’-vinylferrocenes 7a-e (General Procedure)

Under an inert atmosphere methanesulfonyl chloride (1.05 mmol), then Et₃N (3 mmol) were added dropwise at 0 °C to a solution of 1-azinyl-1’-(α-hydroxyethyl)ferrocene 6a-e (1 mmol) and 4-dimethylaminopyridine (0.05 mmol) in anhydrous CH₂Cl₂ (10 mL) in tube Shlenk. Then the reaction mass was heated to room temperature and stirred for 5 hours. The reaction was quenched by addition of 10% solution of NaHCO₃ (10 mL), the mixture was extracted with ethyl acetate (3 × 20 mL). Collected organic layers were washed with brine (30 mL), dried over Na₂SO₄, decanted, solvent was removed under reduced pressure. The residue was purified by flash chromatography on Al₂O₃, the eluate was concentrated to a dry state in vacuo.

1-(Pyridin-2-yl)-1’-vinylferrocene (7a).

Using 308 mg (1 mmol) 1-(pyridin-2-yl)-1’-(α-hydroxyethyl)ferrocene, 0.08 ml (1.05 mmol) methanesulfonyl chloride, 0.41 ml (3 mmol) Et₃N, 6.2 mg (0.05 mmol) 4-dimethylaminopyridine, 1-(pyridin-2-yl)-1’-vinylferrocene was obtained as a red crude oil in 90% yield (261 mg, 0.903 mmol). R_f = 0.6 (TLC: hexane:EtOAc, 9:1). ¹H NMR (600 MHz, CDCl₃): δ – 4.10-4.11 (m, 2H, C₅H₄), 4.20-4.21 (m, 2H, C₅H₄), 4.34-4.35 (m, 2H, C₅H₄), 4.85-4.87 (m, 3H, C₅H₄ and -CH=CH₂), 5.17 (d, 1H, J = 17.0 Hz, -CH=CH₂), 6.15 (dd, 1H, $J_H^{trans}$ = 17.0 Hz, $J_H^{cis}$= 10.0 Hz -CH=CH₂), 7.03-7.05 (m, 1H, 5’-H), 7.35 (d, 1H, J = 8.0 Hz, 6’-H), 7.56-7.59 (td, 1H, J = 8.0 Hz, J = 1.8 Hz, 4’-H), 8.52 (d, 1H, J = 4.0 Hz, 3’-H) ppm. ¹³C NMR (151 MHz, CDCl₃): δ – 68.2, 68.3, 70.3, 71.2 (CH-Cp); 84.5, 84.6 (C-Cp); 120.4 (C6), 120.6 (C5), 135.9 (C4), 149.4 (C3), 158.5 (C2); 112.1 (-CH=CH₂), 133.4 (-CH=CH₂) ppm. IR (ν, cm^-1): 3084, 3004, 2923, 2852, 1682, 1628, 1563, 1496, 1453, 1423, 1387, 1288, 1240, 1200, 1149, 1046, 1028, 985, 891, 815, 784, 739, 696, 644, 518. HRMS (ESI) m/z calcd for C₁₇H₁₆FeN⁺ [M+H]⁺: 290.0626; found 290.0628.

1-(Quinolin-2-yl)-1’-vinylferrocene (7b).

Using 357 mg (1 mmol) 1-(quinolin-2-yl)-1’-(α-hydroxyethyl)ferrocene, 0.08 ml (1.05 mmol) methanesulfonyl chloride, 0.41 ml (3 mmol) Et₃N, 6.2 mg (0.05 mmol) 4-dimethylaminopyridine, 1-(quinolin-2-yl)-1’-vinylferrocene was obtained as red solid in 93% yield (316 mg, 0.932 mmol). R_f = 0.4 (TLC: hexane:EtOAc, 9:1). M.p. 59-61 °C. ¹H NMR (600 MHz, CDCl₃): δ – 4.12 (s, 2H, C₅H₄), 4.22 (s, 2H, C₅H₄), 4.42 (s, 2H, C₅H₄), 4.84 (d, 1H, J = 10.0 Hz, -CH=CH₂), 5.01 (s, 2H, C₅H₄), 5.17 (d, 1H, J = 18.0 Hz, -CH=CH₂), 6.19 (dd, 1H, $J_H^{trans}$ = 18.0 Hz, $J_H^{cis}$ = 10.0 Hz -CH=CH₂), 7.47 (t, 1H, J = 7.0 Hz, 7’-H), 7.52 (d, 1H, J = 9.0 Hz, 3’-H), 7.66 (t, 1H, J = 7.0 Hz, 6’-H), 7.74 (d, 1H, J = 8.0 Hz, 4’-H), 8.03-8.05 (m, 2H, 5’-H, 8’-H) ppm. ¹³C NMR (151 MHz, CDCl₃): δ – 68.3, 69.1, 70.4, 71.1 (CH-Cp); 84.73, 84.75 (C-Cp); 119.8 (C3), 125.5 (C7), 126.8 (C8a), 127.6 (C4), 129.1 (C5), 129.5 (C6), 135.4 (C8), 148.5 (C4a), 158.9 (C2); 112.1 (-CH=CH₂), 133.5 (-CH=CH₂) ppm. IR (ν, cm^-1): 3054, 2919, 1614, 1594, 1551, 1506, 1420, 1378, 1324, 1259, 1240, 1125, 1087, 1010, 978, 901, 816, 755, 619, 526. HRMS (ESI) m/z calcd for C₂₁H₁₈FeN⁺ [M+H]⁺: 340.0783; found 340.0785.

1-(Acridin-9-yl)-1’-vinylferrocene (7c).

Using 407 mg (1 mmol) 1-(acridin-9-yl)-1’-(α-hydroxyethyl)ferrocene, 0.08 ml (1.05 mmol) methanesulfonyl chloride, 0.41 ml (3 mmol) Et₃N, 6.2 mg (0.05 mmol) 4-dimethylaminopyridine, 1-(acridin-9-yl)-1’-vinylferrocene was obtained as a red solid in 88% yield (343 mg, 0.881 mmol). R_f = 0.5 (TLC: hexane:EtOAc, 9:1). M.p. 109-110 °C. ¹H NMR (600 MHz, DMSO-d₆): δ – 4.29 (s, 2H, C₅H₄), 4.52 (s, 2H, C₅H₄), 4.65 (s, 2H, C₅H₄), 4.84 (s, 2H, C₅H₄), 5.03 (dd, 1H, J = 10.0 Hz, J = 1.5 Hz, -CH=CH₂), 5.36 (dd, 1H, J = 18.0 Hz, J = 1.5 Hz, -CH=CH₂), 6.41 (dd, 1H, $J_H^{trans}$ = 18.0 Hz, $J_H^{cis}$ = 10.0 Hz -CH=CH₂), 7.64-7.66 (m, 2H, 3’-H, 6’-H), 7.80-7.83 (m, 2H, 2’-H, 7’-H), 8.14 (d, 2H, J = 9.0 Hz, 4’-H, 5’-H), 9.05 (d, 2H, J = 9.0 Hz, 1’-H, 8’-H) ppm. ¹³C NMR (151 MHz, DMSO-d₆): δ – 68.4, 71.2, 71.4, 74.1 (CH-Cp); 80.5, 84.4 (C-Cp); 124.6 (C4a, C5a), 124.9 (C3, C6), 127.3 (C1, C8); 129.6 (C4, C5), 129.8 (C2, C7); 143.3 (C9); 148.2 (C8a, C9a); 112.8 (-CH=CH₂), 133.7 (-CH=CH₂) ppm. IR (ν, cm^-1): 3069, 2925, 1625, 1563, 1516, 1459, 1404, 1335, 1262, 1101, 1032, 988, 898, 851, 802, 750, 670, 641, 519. HRMS (ESI) m/z calcd for C₂₅H₂₀FeN⁺ [M+H]⁺: 390.0939; found 390.0941.

1-(2,2’-Bipyridin-6-yl)-1’-vinylferrocene (7d).

Using 384 mg (1 mmol) 1-(2,2’-bipyridin-6-yl)-1’-(α-hydroxyethyl)ferrocene, 0.08 ml (1.05 mmol) methanesulfonyl chloride, 0.41 ml (3 mmol) Et₃N, 6.2 mg (0.05 mmol) 4-dimethylaminopyridine, 1-(2,2’-bipyridin-6-yl)-1’-vinylferrocene was obtained as an orange solid in 95% yield (348 mg, 0.950 mmol). R_f = 0.4 (TLC: hexane:EtOAc, 9:1). M.p. 48-49 °C. ¹H NMR (600 MHz, CDCl₃): δ – 4.12 (t, 2H, J = 2.0 Hz, C₅H₄), 4.20 (t, 2H, J = 2.0 Hz, C₅H₄), 4.37 (t, 2H, J = 2.0 Hz, C₅H₄), 4.85 (dd, 1H, J = 11.0, 1.0 Hz, -CH=CH₂), 4.96 (t, J = 2.0 Hz, 2H, C₅H₄), 5.17 (dd, 1H, J = 18.0 Hz, J = 1.5 Hz, -CH=CH₂), 6.15 (dd, 1H, $J_H^{trans}$ = 28.0 Hz, $J_H^{cis}$ = 7.0 Hz -CH=CH₂), 7.31-7.33 (m, 1H, 5’’-H), 7.37 (dd, 1H, J = 8.0 Hz, J = 1.0 Hz, 5’-H), 7.71-7.73 (t, 1H, J = 9.0 Hz, 4’-H), 7.85 (td, 1H, J = 7.0, 1.7 Hz, 4’’-H), 8.19 (dd, 1H, J = 8.0 Hz, J = 1.8 Hz, 3’-H), 8.58-8.59 (d, 1H, J = 8.0 Hz, 6’’-H), 8.68-8.69 (m, 1H, 3’’-H) ppm. ¹³C NMR (151 MHz, CDCl₃): δ – 68.3, 68.6, 70.2, 71.0 (CH-Cp); 84.7, 85.0 (C-Cp); 117.8 (C3); 120.3 (C5); 121.3 (C6’); 123.7 (C5’); 136.8 (C4), 136.9 (C4’), 149.1 (C3’), 155.4 (C2’), 156.8 (C2), 157.7 (C6); 112.1 (-CH=CH₂), 133.5 (-CH=CH₂) ppm. IR (ν, cm^-1): 3086, 3007, 2961, 2924, 2852, 1734, 1662, 1627, 1563, 1489, 1428, 1382, 1310, 1260, 1202, 1155, 1091, 1024, 983, 898, 853, 817, 779, 743, 681, 618, 564, 538, 515, 477. HRMS (ESI) m/z calcd for C₂₂H₁₉FeN₂⁺ [M+H]⁺: 367.0892; found 367.0892.

3.1.6. Synthesis of Isopropenylferrocene 8 [60]

n-BuLi solution in hexane (1.6 M, 0.71 ml, 1.14 mmol) was added to a suspension of methyltriphenylphosphonium iodide 3 (0.505 g, 1.25 mmol) in dry diethyl ether in a Schlenk tube under argon at room temperature. After 30 min stirring, a solution of the acetylferrocene (0.228 g, 1 mmol) in anhydrous diethyl ether (20 ml) was added. The mixture was refluxed at 40 °C for 1 h. Then reaction mixture was purified by chromatography on Al₂O₃ (hexane/EtOAc, 9:1; R_f = 0.6) as eluent. The eluate was concentrated to dryness in vacuo.

Isopropenylferrocene 8. Dark orange solid. Yield: 147 mg (65%). M.p. 77-78 °C. ¹H NMR (600 MHz, CDCl₃): δ – 2.06 (s, 3H, CH₃), 4.10 (s, 5H, C₅H₄), 4.22 (s, 2H, C₅H₄), 4.39 (s, 2H, C₅H₄), 4.84 (s, 1H, =CH₂), 5.13 (s, 1H, =CH₂) ppm. ¹³C NMR (151 MHz, CDCl₃): δ – 21.7 (CH₃), 65.9, 68.7, 69.3 (CH-Cp), 86.6 (C-Cp), 108.4 (CH₂), 141.6 (C) ppm. IR (ν, cm^-1): 3093, 3080, 2973, 2941, 2917, 2848, 1705, 1672, 1407, 1386, 1332, 1291, 1260, 1148, 1102, 1029, 998, 935, 871, 837, 819, 736, 703, 686, 673, 663, 623, 593, 583, 566, 524, 514. HRMS (ESI) m/z calcd for C₁₃H₁₄Fe⁺ [M+H]⁺: 226.0439; found 226.0438.

3.1.7. Synthesis of Vinylferrocene 9 [51]

To a solution of acetylferrocene (1.00 g, 4.39 mmol) in MeOH (25 mL), solid NaBH₄ (0.834 g, 21.95 mmol) was added in five equal portions at r.t during 10 min. The reaction mixture was stirred at r.t. for 20 min. Formed solution was poured into ice-cold water (60 mL) and extracted with CH₂Cl₂ (3 × 60 mL). The collected organic layers were washed with brine (60 mL), dried over Na₂SO₄, filtrated, and solvent was removed under reduced pressure to afford the α-methylferrocenemethanol. Then under an inert atmosphere in Schlenk flask, to a solution of α-methylferrocenemethanol. (0.959 g, 4.17 mmol) and 4-dimethylaminopyridine (25.4 mg, 0.21 mmol) in anhydrous CH₂Cl₂ (10 mL) was added dropwise Et₃N (1.8 mL, 12.51 mmol) at 0 °C, followed by addition of methanesulfonyl chloride (0.34 mL, 4.38 mmol). The reaction mixture was then stirred at r.t. for 4 h. The reaction was quenched by addition of 5% solution of NaHCO₃ (25 mL), and the mixture was extracted with CHCl₃ (3 × 50 mL). Collected organic layers were washed with brine (50 mL), dried over Na₂SO₄, filtrated, and solvent was removed under reduced pressure to afford the crude product. The crude product was purified by chromatography on Al₂O₃ (hexane/EtOAc, 9:1; Rf = 0.8) to afford product 9.

Vinylferrocene 9. Orange solid. Yield: 530 mg (60%). M.p. 51-52 °C. NMR spectra are in agreement with the literature [51].

3.2. Biological Investigations

Experiments were performed in alignment with the standard operating procedures approved by IPAS RAS as described below.

3.2.1. In Vitro Inhibition of HsAChE, EcBChE, and SsCES1 Activities

Enzymes, substrates, and reference compounds were sourced from Sigma-Aldrich (St. Louis, MO, United States). Enzyme activity was measured spectrophotometrically according to [91] using ATCh iodide, BTCh iodide, and 4-NPA as substrates for HsAChE, EcBChE, and SsCES, respectively. Experimental conditions included a K/Na-phosphate buffer (100 mM), at 25 °C, with pH 7.5 for AChE and BChE assays, and pH 8.0 for CES assays. Measurements were conducted on a BioRad Benchmark Plus microplate spectrophotometer (Hercules, CA, United States). Test compounds were dissolved in DMSO with a final solvent concentration of 2% (v/v) in the incubation mixture. Initial inhibition screening was conducted at 20 μM for each compound, with IC₅₀ values determined for compounds showing inhibition ≥35%.

3.2.2. Kinetic Study of HsAChE and EcBChE Inhibition: Determination of Steady-State Inhibition Constants

To investigate the mechanisms of HsAChE and EcBChE inhibition, enzyme kinetics were analyzed (detailed in [91] ). After a 5-min equilibration at 25 °C with three different inhibitor concentrations and six substrate concentrations, residual enzyme activity was measured as described. Lineweaver–Burk plots (1/V vs. 1/[S]) were used to determine competitive (K_i) and non-competitive (αK_i) inhibition constants.

3.2.3. Propidium Displacement Studies

A fluorescence-based assay assessed the ability of test compounds to competitively displace propidium iodide from AChE’s PAS [64,92]. Donepezil and tacrine were used as references, with electric eel AChE (EeAChE, type VI-S, lyophilized powder, Sigma-Aldrich) as the enzyme. The 3D alignment of EeAChE with human AChE confirmed structural similarity [93]. EeAChE (7 μM) was incubated with 20 μM test compounds in Tris-HCl buffer (pH 8.0) for 15 minutes, followed by addition of propidium iodide (8 μM) as described in detail in [94]. Fluorescence readings (530 nm excitation, 600 nm emission) were obtained using a FLUOStar Optima microplate reader (LabTech, Ortenberg, Germany). Displacement (%) was calculated.

3.2.4. Inhibition of β-Amyloid (1-42) (HsAβ₄₂) Self-Aggregation

Inhibition of HsAβ₄₂ self-aggregation by test compounds was studied using the thioflavin T (ThT) fluorescence method [67,95,96] with minor modifications as described in detail in [78]. Lyophilized HFIP-pretreated HsAβ₄₂ (AnaSpec Inc., 0.5 mg) was dissolved in DMSO to obtain a stable 500 μM solution. The samples of 50 μM HsAβ₄₂ in 215 mM Na-phosphate buffer pH 8.0 were incubated for 24 h at 37 °C in the absence or presence of 100 μM test compounds. Myricetin in the same concentration was used as a reference. After incubation, 5 μM ThT in 50 mM glycine-NaOH buffer pH 8.5 was added, and the fluorescence was measured at 440 nm (exc.) and 485 nm (emis.) with a FLUOStar Optima microplate reader. Blanks consisted of 215 mM Na-phosphate buffer, pH 8.0, 20% (v/v) DMSO or test compounds, respectively. The inhibition (%) of HsAβ₄₂ self-aggregation by the test compounds was calculated.

3.2.5. Antioxidant Activity (AOA)

3.2.5.1. ABTS Radical Cation Scavenging Activity Assay

Radical scavenging activity of the compounds was evaluated using the ABTS radical cation (2,2ʹ-azinobis-(3-ethylbenzothiazoline-6-sulfonic acid), ABTS^•+) decolorization assay [97], with minor modifications described in detail in [98]. by mixing 10 μL of compound solution (concentration range 1 – 100 μM) in DMSO with 240 μL of ABTS^•+ working solution in ethanol (100 μM final concentration). The reduction in ABTS^•+ absorbance was measured spectrophotometrically at 734 nm using a xMark UV/VIS microplate spectrophotometer (Bio-Rad, Hercules, CA, United States) for 1 h.

Trolox was used as a reference, and results were reported as Trolox Equivalent Antioxidant Capacity (TEAC) as the ratio of the slopes of the concentration−response curves, test compound/Trolox. The IC₅₀ values for the test compounds were also determined.

3.2.5.2. FRAP Assay

The ferric reducing antioxidant power (FRAP) assay [99,100] modified to be performed in 96-well microplates as described in detail in [94] was used. 10 μL (0.5 mM) of the test or reference compounds were mixed with 240 μL of FRAP reagent (2.5 mL of 10 mM TPTZ (2,4,6-tris(2-pyridyl)-s-triazine) solution in 40 mM HCl, 2.5 mL of 20 mM FeCl₃ in distilled water and 25 mL of 0.3 M acetate buffer pH 3.6); the absorbance of the mixture was measured spectrophotometrically at λ = 593 nm after a 1 h incubation at 37 °C against a blank with a SPECTROStar Nano microplate reader (BMG Labtech, Ortenberg, Germany). Trolox was used as a reference compound. The results were expressed as Trolox equivalents (TE)—the ratio of the concentrations of Trolox and the test compound resulting in the same effect on ferric reducing activity.

3.2.6. Cytotoxicity Study

This study employed SH-SY5Y neuroblastoma cells and dental pulp-derived mesenchymal stem cells (MSC-DP). Cells were cultured in DMEM/F12 with 10% (v/v) FBS under standard conditions. MSC-DP were reprogrammed into neuronal-like cells (MSC-Neu) using Neurobasal medium with addition of Neuromax and 3% (v/v) FBS for 5 days. Neuronal identity was confirmed via RT-PCR analysis of key markers (i.e., β3-tubulin, NeuN, and MAP2).

RNA isolation, cDNA synthesis, and RT-PCR followed standard protocols using SYBR Green chemistry. Primer sequences are listed (Table S1), and amplicons were validated via melt curves. Fold changes were calculated using BioRad software. MSC-Neu verification methods were previously reported [101]

The cytotoxic effect of Fcs was assessed using the Mossman test for cellular dehydrogenase activity (MTT test). Cytotoxicity was measured in the concentration range from 1 to 1024 μM, the incubation time was 24 h. For each compound, the half-lethal concentration parameter (LC₅₀) in μM was calculated as described previously [45].

3.3. Statistical Analyses

3.3.1. Experimental Data Presentation

All experimental tests were performed at least in triplicate in three independent experiments. Results are presented as mean ± SEM calculated using GraphPad Prism version 6.05 for Windows (San Diego CA, United States).

3.3.2. Experimental Data Analyses

Plots, linear regressions, and IC₅₀ values from experimental data were determined using Origin 6.1 for Windows, OriginLab (Northampton, MA, United States). The LC₅₀ values in the cytotoxicity assays were calculated with GraphPad Prism 8 for Windows (GraphPad Software, Boston, Massachusetts, USA).

3.4. Molecular Modeling Studies

3.4.1. Preparation of Ligands for Molecular Docking, Tanimoto Similarity Coefficients (Tc), and van der Waals Volumes (Vn)

SDF files of the reference compounds bis(4-nitrophenyl) phosphate (BNPP; bnp), donepezil (don), myricetin (myr), propidium (prm), and tacrine (tac) were downloaded from PubChem [102], loaded into YASARA-Structure 25.07.15 for Linux [103] (hereinafter referred to as YASARA), energy-minimized in water at pH 7.4 using the AMBER14 force field [104], and exported as PDB, SDF, and YASARA object (YOB) files, all in 3D format.

The X-ray crystal structure of 1-(pyrimidin-4-yl)Fc was previously determined by members of our group [52] and deposited into the Cambridge Crystallographic Data Centre [105] as code JETXUC and deposition number 282124. The CIF file of this structure was downloaded from the CCDC and imported into YASARA where the pyrimidine substituent was removed and each of the monomeric Fc derivatives shown in Table 2 were created using the 3D molecular builder tools in YASARA. Structures were energy-minimized in water at pH 7.4 using an AMBER14 force field that was parameterized in YASARA for Fe in Fc compounds, including the handling of coordinate-covalent (“dative”) bonds. Protonation states were adjusted in YASARA according to QC predictions described below. Structures were then reminimized in YASARA using the AMBER14 force field, keeping the QC-predicted protonation states intact. Reminimized structures were then exported as PDB, SDF, and YOB files, all in flexible 3D format.

Protonated dimeric forms of Fc compounds 1a-c, each stabilized by an HPO₄²⁻ anion, were predicted as described in Supplemental Information, Sections S4 and S5, subjected to geometry optimization using QC methods as described below, and saved as PDB files. A similar approach was used to generate the protonated dimeric forms of Fc compounds 1a-c, each stabilized by two Cl⁻ anions. These dimers were subjected to QC-based geometry optimization and exported as 3D PDB files. The dimeric PDB files were loaded into YASARA for preparation of the 5a-c and 7a-c versions of the dimers as described above for preparation of the monomers. However, because the monomeric constituents of the dimers were not held together by covalent bonds, to keep the dimers intact during reminimization in the AMBER14 force field, the positions of all atoms except those in the propenyl and vinyl substituents were fixed. For use in Tc and Vn calculations, the dimer structures were exported as PDB files. For use as ligands in molecular docking, the protonated dimeric structures were exported as semi-rigid 3D YOB files to maintain the core dimeric structures.

3.4.2. Quantum Chemical (QC) Calculations

QC calculations were performed using the DFT method in Gaussian 16 [106] and Priroda 20 [107] software. All energetics characteristics were calculated in the unrestricted B3LYP functional [108] with the 6-31++g(d,p) basis set [109] and the empirical Grimme correction [110]. The continuum SMD model [111] was used to account for solvent (water or ethanol) effects.

3.4.4. Tc and Vn Calculations

Tc values relative to the monomeric and dimeric forms of lead compound 7b were computed using ROCS 3.8.0.1 for Linux [112,113] using PDB files of the ligands as inputs.

Vn values were calculated using the Schrödinger 2025-2 utility script, phase_volCalc, after uploading the PDB monomeric or dimeric ligand files into Schrödinger Maestro and exporting them in MAE format to use as inputs [114].

3.4.5. Preparation of Protein Targets for Molecular Docking

In the present investigation, the workflow involved generation of the protein targets from their amino acid sequences using Protenix, an AlphaFold3 (AF3) reproduction code [75], followed by molecular docking with the YASARA implementation of Autodock Vina [103,115,116]. This procedure enabled the preparation of protein structures of similar quality to each other that matched the protein species employed in the experimental protocols. In contrast, high-quality X-ray crystal structures were not available for all the species of proteins used in the present study. Moreover, crystal structures of proteins can have variable resolution, missing residues, and biases introduced by structural adaptations to co-crystallized ligands [117]. Whereas co-folding of proteins with ligands might have been used, thereby obviating the docking step, preliminary trials showed that co-folding resulted in dissociation of the Fc dimers into their monomeric constituents.

Sequences for EcBChE, EeAChE, HsAβ₄₂, HsAChE, and SsCES1 were downloaded from UniProt [118] and imported into YASARA. Signal and/or anchor sequences determined by UniProt or NCBI-Protein [119] were deleted, and the remaining trimmed sequences were exported as FASTA files for importing into Protenix to generate the 3D structures of the proteins.

Protenix 0.5.0 [120] was installed in a conda python environment on a Steiger Dynamics Linux workstation equipped with a liquid-cooled AMD 5975WX CPU with 32 cores/64 threads, 256 GB RAM, and 2x Nvidia RTX A6000 GPUs. Each sequence was run in multiple sequence alignment (MSA) mode for 10 recycles and 200 sampling steps to generate 25 diffusion samples that were output as CIF files. Using the Protenix scoring function, the top-scoring model from each of the five sequences was chosen for further processing as protein targets for molecular docking.

Each of the top-scoring Protenix models was imported into YASARA, where the proteins were subjected to a docking preparation protocol in the AMBER14 force field using default settings. The protocol included setting of atom types, addition of hydrogens, and optimization of the hydrogen-bonding network [103,121].

For each of the enzyme proteins studied, a simulation cell was positioned to encompass amino acid residues known to comprise the catalytic active site as well as potential binding sites on the protein surface or lining the tunnel or gorge leading to the catalytic active site. These residues have been documented for each species either through direct determinations or by homology to the corresponding proteins in humans or other species as follows: EcBChE [122,123,124,125,126,127], EeAChE [128,129]; HsAChE [130,131,132]; SsCES1 [133]. For the HsAβ₄₂ peptide, the entire structure was enclosed within the simulation cell.

With the center of the local coordinates (x, y, z) of each protein set to (0, 0, 0), the dimensions of the cuboid simulation cell (in Å) for each protein were as follows: EcBChE: (20.0, 26.0, 23.0); EeAChE: (25.0, 35.0, 24.0); HsAβ₄₂: (74.0, 64.0, 52.0); HsAChE: (27.0, 36.0, 23.0); SsCES2: (26.0, 40.0, 25.0). Each protein and its accompanying simulation cell were exported as YASARA scene (SCE) files to serve as protein targets for molecular docking.

3.4.6. Molecular Docking Procedure

Molecular docking was conducted by the YASARA implementation of Autodock Vina 1.2.5 [103,115,116], using as ligands each of the reference compounds along with the monomeric and dimeric Fc derivatives prepared as described above. Likewise, the protein docking targets were prepared as described above. The number of runs per ligand was 500, and the resulting docking poses were sorted by binding energy (Eb = -ΔG, where ΔG = free energy of binding). The top-scoring pose was selected for further processing in images, calculations of overlap volumes and associated heatmaps, prediction of propidium displacement from the PAS of EeAChE, and correlations of binding energies to HsAβ₄₂ with percent inhibition of self-aggregation of HsAβ₄₂. Images of docking poses for all but one of the figures were produced with PyMOL 3.1.6.1 for Linux, Schrödinger, LLC. The docking poses depicted in Figure S7.1 were generated with YASARA 25.07.15 for Linux. Validation of the molecular docking method is described in Supplemental Information, Section S7, and illustrated in Figure S7.

3.4.7. Calculation of Overlap Volumes for Docked Ligands

Output from molecular docking included SDF files of the docked ligand coordinates. These files were concatenated into a multi-ligand SDF file, imported into Schrödinger Maestro 2025-2 [114], and exported as a multi-ligand MAE file. The MAE file was used as the input for the utility script, phase_volCalc. To generate a similarity matrix of ligand overlap volumes, the script was configured to compare a given MAE file to itself. To generate a non-similarity matrix of ligand overlap volumes, the script was configured to compare two different MAE files containing docking coordinates. By default, the script calculates molecular volumes based on the van der Waals radii of the constituent atoms.

3.4.8. Linear Discriminant Analysis (LDA) of Propidium Displacement from EeAChE PAS

LDA is a supervised procedure for dimension reduction and classification that finds a linear combination of variables to maximize the distance between the means of different classes and minimizes the within-class variance [134,135]. In the present study, data used as inputs for LDA comprised two outputs from molecular docking: (1) binding energies of the docked monomeric forms of the ligands; and (2) percent volume overlap of each monomeric ligand with propidium. LDA was performed using PAST 5.22 for Windows [136]. An exhaustive leave-one-out procedure (jackknifing) was applied [137], but the classification accuracy remained the same (94.7%) with or without jackknifing. Figure 9 depicting the LDA results was rendered by GraphPad Prism version 10.6.1 for Windows, GraphPad Software, Boston, Massachusetts USA, www.graphpad.com.

3.4.9. Correlation of HsAβ₄₂ Self-Aggregation % Inhibition and Docking Binding Energy

Verification of normal distributions of data by the D’Agostino and Pearson test, calculations of Pearson correlation coefficients and their statistical significance, and correlation plots were carried out using GraphPad Prism version 10.6.1 for Windows, GraphPad Software, Boston, Massachusetts USA, www.graphpad.com.

4. Conclusions

A synthetic method for producing azinyl-containing alkenylferrocenes has been developed. It was demonstrated that 1-azinyl-1’-acetylferrocenes can participate in the Wittig reaction to yield 1-azinyl-1’-isopropenylferrocenes. Additionally, sequential reduction and dehydration reactions of 1-azinyl-1’-acetylferrocenes can lead to the formation of 1-azinyl-1’-vinylferrocenes with excellent yields (up to 98%).

Quinoline and bipyridine derivatives of Fc inhibited cholinesterases within the micromolar range and showed minimal inhibition of the off-target enzyme CES. Enzyme kinetics studies demonstrated that the lead compound 7b exerted a mixed type of inhibition on the cholinesterases. These compounds, along with acridine derivatives, displaced propidium at a level comparable to the reference compound donepezil, indicating their potential to inhibit AChE-induced β-amyloid aggregation. They also exhibited strong inhibitory effects on HsAβ₄₂ self-aggregation, achieving efficacies similar to that of the reference compound myricetin. Among the synthesized compounds, quinoline derivatives displayed the highest antioxidant activity, being 4 times more potent than the standard antioxidant Trolox in the ABTS test and 2-3 times in the FRAP assay, which agrees with the results of QC calculations.

QC calculations have shown (see Table S3) that the advantage of dimer formation growths with increasing size of the aromatic heterocyclic ligand. At the same time, buffer anions increase the advantage of dimer formation for protonated compounds.

Calculations of Tanimoto similarity coefficients and molecular volumes quantified the chemical diversity of the study compounds and assisted with making inferences about the relative access of the monomeric and dimeric forms of the compounds to interior or exterior binding sites of proteins.

This study employed a novel approach to molecular docking by using Protenix, an AlphaFold3 reproduction code, to generate the 3D structures of the protein docking targets. Molecular docking results supported the QC predictions that the acridine derivatives in particular would exist in the conditions of the esterase assays as dimers, which would be too bulky to gain access to the catalytic active sites. However, either as dimers or monomers, the Fc derivatives could bind to the PAS of EeAChE to displace propidium. They could also bind to the surfaces of HsAβ₄₂ to block the self-aggregation of this peptide. Data derived from molecular docking, such as heatmaps of overlap volumes of docked ligands, proved helpful for making predictions about similar or dissimilar actions of compounds on target proteins, especially with respect to displacement of propidium or inhibition of HsAβ₄₂ self-aggregation. When considering propidium displacement as a binary response and classifying the study compounds as active or inactive, it was possible to predict the classifications with 94.7% accuracy using linear discriminant analysis and only two variables from molecular docking of the monomers: binding energy and propidium overlap volumes. Inhibition of HsAβ₄₂ self-aggregation showed a graded response that had strong correlations with the ligand binding energies to the protein.

Quinoline and bipyridine derivatives also exhibited relatively low cytotoxicity on neuronal cells MSC-Neu and human neuroblastoma tumor cells SH-SY5Y being less toxic for MSC-Neu cells.

Overall, these findings suggest that Fc derivatives with quinoline and bipyridine substituents are promising candidates for further investigation as potential multifunctional therapeutic agents for AD.