9-Phosphoryl-9,10-dihydroacridines 1a–f and 9-phosphorylacridines 2a–f were obtained according to Figure 1 as described in our recent publication (Shchepochkin et al., 2021).

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

Plots, linear regressions, and IC₅₀ values were determined using Origin 6.1 for Windows, OriginLab (Northampton, MA, United States). All 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).

Currently, 9-substituted acridines can be obtained by several approaches. The classical method for the preparation of acridine derivatives is based on high-temperature condensation of diphenylamine and carboxylic acid in the presence of zinc chloride (Bernthsen, 1884) (Pathway A, Figure 2). Another method consists of preliminary functionalization of acridine and subsequent substitution of the introduced group by a nucleophile in the presence of transition-metal salts (Liu et al., 2013; Zhilyaev et al., 2022) (Pathway B, Figure 2). The presented strategies do not exclude a number of shortcomings. For example, the desire to introduce auxiliary groups, the use of catalysts and additional ligands, the employment of high temperature processes, and the need to separate reaction products from metal-containing byproducts complicate the preparation of target acridines. In addition, such accompanying negative factors as the high cost of the reaction process and, in some cases, low yields of the final products make the method inefficient and unprofitable. However, an alternative approach to functionalized acridines is nucleophilic aromatic substitution of hydrogen (S_N ^H) (Chupakhin et al., 1994; Charushin and Chupakhin, 2005; Makosza, 2010; Mąkosza, 2011; Mąkosza and Wojciechowski, 2014). This method makes it possible to obtain 9-substituted acridines efficiently and directly, avoiding the disadvantages mentioned above.

The S_N ^H reactions proceed in two stages. The first one is nucleophilic addition to an aromatic substrate to form an intermediate (ϭ^H-adduct). The ϭ^H-adducts, e.g., in the case of acridine, are stable compounds, which makes it possible to comprehensively study their physicochemical properties and biological activities (Chupakhin et al., 2017; Volkov et al., 2018). The second stage is characterized by an oxidative process with formation of the target-functionalized acridine (Pathway C, Figure 2).

We used the S_N ^H approach described in detail in a recent publication (Shchepochkin et al., 2021) to obtain the 9-phosphoryl-9,10-dihydroacridines 1a–f and 9-phosphorylacridines 2a–f studied in this work (Figure 1). Briefly, acridine and the corresponding phosphonate were stirred without solvent under an argon atmosphere at 75°C–80°C for 2–7 h. 9-Phosphoryl-9,10-dihydroacridines 1a-f were isolated in 85%–95% yields. The corresponding aromatized compounds 2a-f were obtained in 81%–89% yields by electrochemical oxidation of 1a–f.

Figure 3 illustrates the structures of the previously described compounds, which were the subject of the research in the present work.

Table 1 lists the components of the esterase profile of our compounds, i.e., their inhibitory potencies against the enzymes of cholinergic neurotransmission, AChE and BChE, and a structurally related off-target enzyme, CES (Makhaeva et al., 2016; Makhaeva et al., 2019b).

As can be seen from Table 1, all tested compounds weakly inhibited AChE, while a number of compounds demonstrated effective inhibition of BChE.

The most effective BChE inhibitors were found among the series of phosphorylated dihydroacridines. Moreover, the influence of the structure of the phosphorus-containing fragment on the inhibitory activity is clearly seen: dialkylphosphonates weakly inhibited BChE, but the introduction of aryl fragments increased the activity. Thus, diphenylphosphonate 1c was a moderate inhibitor (IC₅₀ = 48.0 ± 3.1 µM), while dibenzyloxy derivative 1d and its diphenethyl bioisostere 1e were effective inhibitors of BChE with IC₅₀ = 2.90 ± 0.23 µM and 3.22 ± 0.25 µM, respectively. The introduction of a chlorine atom into the para-position of the phenethyl fragment of 1e (compound 1f) reduced tenfold the anti-BChE activity (IC₅₀ = 21.7 ± 1.7 µM).

At the same time, aromatized analogs 2 of compounds 1 were largely inactive as BChE inhibitors. An exception was the acridine analog of the most effective dihydroacridine 1d–dibenzyloxy derivative 2d, which inhibited BChE with IC₅₀ = 6.90 ± 0.55 µM, i.e., it was only 2.4 times less effective than the corresponding dihydroacridine.

Weak inhibition of CES, an enzyme that catalyzes the hydrolysis of numerous ester-containing drugs, is a positive property of the studied compounds because this decreases the likelihood of unwanted drug-drug interactions in the event of their use in AD therapy (Tsurkan et al., 2013; Makhaeva et al., 2019b).

The most active BChE inhibitor 1d was selected for inhibition kinetics studies. Lineweaver-Burk plots-double reciprocal representations of Michaelis-Menten parameters-were used to assess the type of inhibition. Graphical analysis of the kinetic data on BChE inhibition by the tested compound (Figure 4) demonstrates changes in both K _m and V _max values; this result attests to a mixed type of inhibition. The values obtained for the competitive component (K _i) and the non-competitive component (αK _i) of the constants of BChE inhibition by compound 1d were 2.22 ± 0.19 µM and 11.5 ± 0.9 µM, respectively.

To investigate the structural origins of the observed differences in inhibitory activity, we performed molecular docking to AChE and BChE for the most active dihydroacridine derivatives, namely, the bioisosteric compounds 1d and 1e and their aromatized analogs 2d and 2e.

According to the results of molecular docking, compounds with benzyloxy and phenethyl substituents 1d,e, and 2d,e were able to bind to the AChE active site gorge in a similar manner (Figure 5). However, in this case, the bulky dihydroacridine/acridine and benzyloxy/phenethyl groups were located below amino acid residues Tyr341 and Tyr124, which are the part of a narrow bottleneck (Sussman et al., 1991; Wlodek et al., 1997). This makes trafficking of the ligands to the bottom of the gorge kinetically hindered (Lushchekina and Masson, 2020), thus reducing their inhibitory efficacy.

With respect to BChE docking (Figure 6), a significant difference of binding to the enzyme active site at the bottom of the gorge was observed between dihydroacridine/acridine compounds 1d,e and 2d,e with benzyloxy/phenethyl substituents. Due to the wider active site gorge of BChE compared to AChE (Saxena et al., 1997), no kinetic difficulties of such binding were expected.

According to the docking results, the dihydroacridine derivatives 1d,e form direct hydrogen bonds with the Tyr332 and D70 side chains (Figure 6A). There is also a theoretical possibility of interaction of the phosphoryl group with the Thr120 side chain oxygen atom through a bridging water molecule (Figure 6A). According to X-ray crystallography (Koellner et al., 2000; Bourne et al., 2010; Berg et al., 2011), and molecular modeling (Nascimento et al., 2017; Makhaeva et al., 2021; van der Westhuizen et al., 2022), such interactions are typical for protein-ligand interactions in the active site gorge of cholinesterases.

Regarding acridines 2d and 2e, no specific interactions were observed, which agrees with their lower experimentally observed inhibitory activity. However, interactions of the acridine fragment of both compounds with the Trp82 main chain oxygen atom are possible (Figure 6B). We observed such a water-mediated interaction for the similar amiridine fragment and the homologous AChE residue Trp86 in molecular dynamics simulations (Makhaeva et al., 2021). As observed with dihydroacridine derivatives, the phosphoryl group of the acridine compounds was located in such a way to suggest that a similar water molecule-bridged interaction occurred with the oxygen atom of the Thr120 side chain (Figure 6B). Regarding the dibenzyloxy group of compound 2d, such interactions are also possible for the oxygen atom (Figure 6B), which is consistent with its higher inhibitory activity compared to compound 2e (Table 1).

We performed MD simulations for the complex of the most active compound 1d with BChE. During the MD trajectory, the ligand remained bound to the enzyme, though it had reoriented, and the resulting binding pose (Figure 7) differed from the docked one (Figure 6A) in specific details. While molecular docking showed that the dihydroacridine group interacted with side chains of the PAS residues Asp70 and Tyr332, these interactions were disrupted shortly after the start of the MD simulation (Figure 8A). Instead, the inhibitor reoriented so that the dihydroacridine group established new interactions with the side chain of Trp82 after 15 ns of simulation (Figure 8B). This residue is located deeper in the gorge than Asp70 and Tyr332 and forms the active site as a part of the cation-binding compartment. The newly established π-cation stacking or T-stacking interactions of the dihydroacridine fragment were maintained during the remainder of the MD trajectory. In addition, T-stacking interactions of one of the ligand’s phenyl rings with the Trp231 indole ring were observed during the entire MD trajectory (Figure 8C). The reorientation of the ligand in the active site gorge of BChE took place during the first 15 ns of the simulation and included temporary displacement of the phosphate group. After this, the phosphate group returned to the position outside the oxyanion hole and close to the Thr120 side chain (Figure 7; Figure 8D). The MD simulation confirmed that their interaction was maintained mainly through frequently exchanging bridging water molecules (Figure 9). However, occasional direct hydrogen bonding is also possible due to thermal motion (Figure 8D).

The in vitro inhibitory activity of the test compounds against Aβ₄₂ self-aggregation was determined using the thioflavin T (ThT) fluorimetric assay (LeVine, 1999; Bartolini et al., 2003; Munoz-Ruiz et al., 2005; Bartolini et al., 2011). This widely used procedure is based on specific binding of the fluorescent dye ThT to the β-sheets of assembled amyloid fibrils leading to a significant increase in fluorescence signal (Biancalana and Koide, 2010). Therefore, the decrease in ThT fluorescence in the presence of the studied compound correlates with its activity to inhibit the formation of amyloid aggregates.

As can be seen from Table 1, the ability of the studied compounds to inhibit the self-aggregation of Aβ₄₂ depends both on the structure of their acridine moiety and the substituents at the phosphorus atom; moreover, the structure-activity relationship largely coincides with that observed for BChE inhibition.

Indeed, acridine derivatives 2 inhibit the self-aggregation of Aβ₄₂ very weakly or do not inhibit at all. The ability of dihydroacridines 1 to block Aβ₄₂ self-aggregation depends on the structure of the phosphoryl fragment: dialkoxy derivatives are weak inhibitors, but the introduction of aryl substituents increases activity, which increases upon going from diphenoxy (1с) to dibenzyloxy (1d) substituents. Compound 1d exhibits the maximum inhibitory activity and suppresses the formation of amyloid aggregates by 58.9% ± 4.7%. Its bioisosteric analog, the diphenethyl derivative 1e, also has high activity, inhibiting Aβ₄₂ self-aggregation by 46.9% ± 4.2%. Adding a chlorine atom to the para-position of the phenethyl fragment 1e (compound 1f) slightly reduces the inhibitor activity.

Molecular docking did not reveal a direct explanation of the experimentally observed differences in the ability of acridine/dihydroacridine compounds to inhibit Aβ₄₂ self-aggregation. There were slight differences in estimated binding energies (Supplementary Table S1) and number of contacts (Supplementary Table S2) of dihydroacridine derivatives 1d,e compared to acridine derivatives 2d,e with Aβ₄₂ conformers obtained from the NMR structure (PDB: 1IYT) (Crescenzi et al., 2002) [see details in (Makhaeva et al., 2022)]. Although these differences could be considered minor, they are in line with different conformations of the Aβ₄₂ peptide used as molecular docking targets.

Another plausible reason for the experimentally observed substantial differences between active dihydroacridine inhibitors and practically inactive acridine compounds could be related to the different ability of planar aromatic acridines to self-associate through π-π stacking interactions compared to non-planar dihydroacridines. Such self-stacking of acridine derivatives is well known (Evstigneev et al., 2006; Franco Pinto et al., 2022) and could compete with binding to the Aβ₄₂ peptide (Gao et al., 1991; Buurma and Haq, 2008).

Table 2 lists the results of our determinations of the AOA of compounds 1a–f and 2a–f using the ABTS and FRAP assays. In the ABTS test, the radical cation ABTS^•+ reacts with an antioxidant via one or both of the following pathways: single-electron transfer (SET) or hydrogen-atom transfer (HAT). In the FRAP test, the ferric 2,4,6-tripyridyl-s-triazine complex [Fe (TPTZ)₂]³⁺ is reduced to the corresponding ferrous complex [Fe (TPTZ)₂]²⁺ by the SET mechanism. In these tests, we used Trolox and ascorbic acid as the reference and positive control compounds, respectively.

All derivatives of 9,10-dihydroacridine 1 demonstrated high radical-scavenging activity in the ABTS test at levels near or slightly above that of Trolox, the standard antioxidant. In contrast, the 9-phosphoryl-acridines 2 were not active in the ABTS test. It should also be noted that compounds 1a–d exhibited a high initial rate of the ABTS^•+ binding reaction, comparable to Trolox (the maximum degree of radical binding was achieved within 1 min). This fact is consistent with the SET mechanism for the antiradical activity of these compounds. A slight decrease in the radical-scavenging activity was observed for the phenethyl derivative 1e and its para-Cl-substituted analog 1f, which was also accompanied by a decrease in the reaction rate (the time to reach the maximum degree of radical binding increased from 1 to 30 min).

All dihydroacridine derivatives 1a–f also showed high iron-reducing activity in the FRAP test, at the level or above that of Trolox. Moreover, these compounds were relatively rapid antioxidants, reaching the maximum effect within 4 min. The high activity of dihydroacridines 1a–f in the Fe³⁺-reducing test is probably directly related to the ease of their electrochemical oxidation to acridines (Shchepochkin et al., 2021). The dimethoxy derivative 1a (1.60 ± 0.01 TE) was the lead compound in this series, and the least active compound was para-Cl-substituted diphenethylphosphoryl dihydroacridine 1f (0.81 ± 0.02 TE). In general, the results of the ABTS and FRAP tests were in good agreement with each other.

Phosphoryl acridines 2a–f had almost no activity in the FRAP test. Only dimethoxy and diethoxy derivatives 2a,b, which are the aromatic analogs of the most active compounds 1a,b in the dihydroacridine series, showed detectable activity.

The FRAP assay is a typical SET method (Gulcin, 2020; Spiegel et al., 2020; Florin Danet, 2021), where [Fe (TPTZ)₂]³⁺ is reduced to [Fe (TPTZ)₂]²⁺ by electron transfer from an antioxidant molecule. Unlike [Fe (TPTZ)₂]³⁺, ABTS^•+ can be reduced by acquiring both a hydrogen atom and an electron, depending on the antioxidant structure, solvent type, and pH (Ilyasov et al., 2020). In particular, SET is facilitated at acidic pH (Gulcin, 2020). In the ABTS method used in this work, the main solvent is ethanol at pH∼4.5. Under these conditions, the ABTS^•+ molecule is deprotonated at both SO₃ groups; it is reduced by acquiring an electron from the antioxidant, similarly to [Fe (TPTZ)₂]³⁺ (Ivanova et al., 2020).

In both methods, reduction can occur by either the direct SET mechanism characterized by the ionization potential (IP), or by a hydrogen atom abstraction from the antioxidant molecule, when an electron is transferred to the radical and a proton to a solvent molecule. In the latter case, the actual antioxidant mechanism may be a complex multistage process, but the overall reaction can be characterized by the enthalpy of a hydrogen atom abstraction, i.e., the bond dissociation enthalpy, also known as the bond dissociation energy (BDE).

The quantum mechanical (QM) characteristics of possible antioxidant reactions for dihydroacridines and acridines were calculated taking two items into account: 1) the protonation state of the investigated structures in water and ethanol for the FRAP and ABTS tests, respectively; and 2) the method of synthesizing acridines from dihydroacridines by two-electron transfer electrochemical oxidation (Shchepochkin et al., 2021). The second item suggested a two-stage mechanism for the interaction of the antioxidant with ABTS^•+ or FRAP molecules.

The calculated IP and BDE values for acridines were much higher than the corresponding values for dihydroacridines. Moreover, the characteristics of acridines were close to each other, as were the characteristics of dihydroacridines, which agrees with similar experimental results for compounds 2a and 1a, respectively. A detailed description of the approach that was used and all calculated characteristics are given in Supplementary Materials. Here, we briefly summarize the results.

The obtained QM characteristics for the two-stage mechanism of interaction of the antioxidant and ABTS^•+ or FRAP molecules allowed us to suggest that each dihydroacridine molecule donates two electrons, yielding the corresponding acridine molecule, and thereby reducing two molecules of ABTS^•+ or FRAP. At the same time, the reduction of the second radical molecule was characterized by low values of the BDE, which was consistent with the high AOA of dihydroacridines in both tests. The poor AOA of acridines agrees with high values of their QM characteristics.

The suggested two-stage mechanism of the dihydroacridine antioxidant action is illustrated in Figure 10 for compound 1a in the ABTS test as an example. In path A, the antioxidant (dihydroacridine, AH_CH_N) sequentially loses an electron and a hydrogen atom to generate the protonated form of the corresponding acridine (AH_N ⁺). The latter dissociates to acridine (A); this process is characterized by the proton affinity (PA). In path B, the antioxidant sequentially loses two H-atoms to form the corresponding acridine.

In both cases, the reaction enthalpy of the second stage was low, indicating that the reduction of the second ABTS^•+ molecule occurred immediately after the reduction of the first one. For comparison, the IP and the enthalpy of a hydrogen atom abstraction of the phenol group of Trolox, calculated in ethanol, were 98.0 and 73.0 kcal/mol, respectively.

A noticeably higher AOA of compound 1a with methoxy substituents compared to other compounds 1 in the FRAP test and the similar AOA in the entire series of compounds 1 in the ABTS test could be explained by the different access of antioxidant molecules to the ABTS^•+ radical versus the [Fe (TPTZ)₂]³⁺ cation. Indeed, [Fe (TPTZ)₂]³⁺, with a three-dimensional propeller structure, is bulkier than the quasi-one-dimensional structure of ABTS^•+ (Supplementary Figure S1), which hinders access for bulky antioxidants.

Table 3 shows the predicted ADMET properties and physico-chemical characteristics that we computed for compounds 1a–f and 2a–f. As can be seen, with the exception of the phenethyl and chlorophenethyl derivatives 1e–f and 2e–f, we obtained favorable values for intestinal absorption, indicating that the compounds would be expected to be bioavailable when administered orally. Note that it is possible that the unfavorable numbers for compounds 1e–f and 2e–f were likely due to their being outside of the applicability domain of the model. Given the high BBB permeability values, the compounds would be expected to enter the brain to exert effects on CNS targets; this is especially so for the benzyl and phenethyl derivatives. Parameters for cardiac toxicity risk, hERG pK _i and pIC₅₀, were within 5.0–7.9 log units, near the center of the possible range of 3–9 log units. Lipophilicities, as assessed by computed LogP values, were in some cases outside of the classical Lipinsky rule of 5 range, but these values were deemed unreliable owing to their being outside of the applicability domain of the model. Otherwise, the molecular weights and predicted aqueous solubility values were near or within acceptable limits for drug-like molecules. Integral quantitative estimates of drug likeness (QED) fell within a range of 0.2–0.9; the most active compounds, 1d,e, scored 0.3 for this parameter. Finally, no alerts were identified by the Pan Assay INterference compoundS (PAINS) filter.

Taken together, the results of the computed ADMET, physico-chemical, and PAINS profiles indicate that the lead compounds were acceptable for being advanced to further optimization and experimental studies of efficacy and safety.