Interplay Between Ionization and Tautomerism in Bioactive β- Enamino Ester-Containing Cyclic Compounds: Study of Annulated 1,2,3,6-Tetrahydroazocine Derivatives

Interplay Between Ionization and Tautomerism in Bioactive βEnamino Ester-Containing Cyclic Compounds: Study of Annulated 1,2,3,6-Tetrahydroazocine Derivatives Antonio Viayna, Salvatore G. Antermite, Modesto de Candia, ‡ Cosimo D. Altomare and F. Javier Luque † Department of Nutrition, Food Science and Gastronomy, Faculty of Pharmacy and Food Sciences, Institute of Biomedicine (IBUB) and Institute of Theoretical and Computational Chemistry (ITQCUB), University of Barcelona, Av. Prat de la Riba 171, E-08921 Santa Coloma de Gramenet, Spain


INTRODUCTION
Finding the bioactive species that mediates the functional role of small molecules is crucial to gain insight into many biochemical processes of the cell. 1 This knowledge permits to disclose the molecular determinants implicated in the recognition and binding of the ligand to specific macromolecular targets, hence providing a basis to understand the mechanism of action of small (bio)organic compounds. Furthermore, it may be valuable for disclosing the relationships between chemical structure and biological activity, highlighting clues for the design of novel compounds with improved functional profile.
The identification of the bioactive species may often be challenging, especially for ionizable small molecules, as the chemical features of a given compound can modulate the balance between distinct ionization and tautomeric states. [2][3][4][5][6][7][8][9] Moreover, depending upon the number of rotatable bonds in the chemical skeleton of the compound, the ionization and tautomeric preferences may influence the population of distinct conformational wells, reflecting the balance between intramolecular interactions, such as hydrogen bonds or stacking interactions, and intermolecular interactions (i.e., with solvent molecules and co-solutes present in condensed media). [10][11][12][13] This scenario may be further complicated by the presence of stereocenters in the chemical structure of the compound. Overall, the activity of the compound is dictated by the interplay between ionization, tautomerism and conformational flexibility. On the other hand, these chemical features will also modulate the physicochemical behavior of the compound, affecting properties such as solubility, partitioning in aqueous/organic phases and biodistribution in the organism. 14,15 Dissecting the interplay between the aforementioned factors may be important for better characterizing the physicochemical and biological profile of a compound. This could be illustrated by a series of β-enamino ester derivatives of annulated (e.g., 1H-pyrrole-, 1H-indole-and pyrimidine-fused) 1,2,3,6-tetrahydroazocines, which were previously synthesized and assayed by some of us as acetylcholinesterase inhibitors with potential application for the treatment of mild Alzheimer's disease. 16,17 The main structural variations affecting acid-base equilibria of the β-enamino ester-containing heterocyclic compounds are exemplified for the two 1H-indole-fused tetrahydroazocine scaffolds shown in Table 1, namely the fusion isomers 2,3,6,11-tetrahydro-1H-azocino [4,5b]indole (set I) and 2,3,6,7-tetrahydro-1H-azocino [5,4-b]indole (set II). Within each molecular set, major variations pertain to the presence of a second ester group (R 2 = H or COOMe) α to the enamino N and the size-increasing alkyl groups at C 6 in set II (R 4 = Me, iPr, Bn), whereas the differences in the alkyl groups in R 1 and R 3 appear to not significantly affect the basicity of N3. Table 1. General structures, numbering and experimental pK a of the investigated β-enamino ester derivatives of (I) 2,3,6,11-tetrahydro-1H-azocino [4,5-b]indole and (II) 2,3,6,7-tetrahydro-1H-azocino [5,4-b]indole. The β-enamino ester function is colored in blue.

I II
The biological properties of these compounds, including ADMET (Absorption, Distribution, Metabolism, Excretion and Toxicity) profiles, could be affected by the susceptibility of the nitrogen atom to be protonated. This may depend on the COOR 3 substituents α or β to the enamino N and on the conformational flexibility of the fused tetrahydroazocine ring, which in turn may differ in the scaffold chemotypes I and II. As a matter of fact, the enamino esters investigated herein form two clusters of experimental pK a values, namely α,β-diesters (Ia, Ib and IIa), with values ranging within 6.1-7.3, and β-esters (Ic, IIb, IIc and IId) ranging within 8.1-9.0 (Table 1).
Remarkably, this implies that the predominant species in aqueous solution at physiological pH will be neutral or protonated, respectively, which is relevant not only for identifying the bioactive species, but also affects the feasibility to cross biological barriers, such as the blood-brain barrier for compounds acting at the central nervous system.
In this context, this study aims to examine the interplay between ionization, tautomerism and conformational preferences, in order to interpret the behavior of the enamino ester derivatives of 1H-indole-fused tetrahydroazocines. At this point, it can be hypothesized that the stability of the enamine species may be altered upon protonation, favoring the transition to alternative tautomeric species where the double bond between C 4 and C 5 is shifted to C 5 and C 6 (Scheme 1). This process will switch the electron delocalization of the double bond with the azocino nitrogen N3 by the conjugation with the π-electron density of the indole ring. Moreover, the description of this process may be influenced by the changes in conformational flexibility arising from the different skeleton of the azocino ring in the two tautomers, and by the substituents attached to the ring. Accordingly, this study reports the results of high-level ab initio quantum mechanical (QM) calculations performed to examine the mutual influence between ionization and tautomerism, in conjunction with the Multilevel methodology, 11,13 to carry out the conformational sampling of the compounds. The results highlight the subtle influence of these factors on the bioactive species of these compounds.

Scheme 1.
Representation of the tautomeric equilibria in the two sets of β-enamino esters of 1H-indole-fused tetrahydroazocines.

METHODS
Molecular systems. The theoretical study was performed for the generic compounds I (R 1 = Me; R 2 = COOMe; R 3 = Me) and II (R 1 = Me; R 2 = H; R 3 = R 4 = Me) shown in Scheme 1, which are strictly related to Ia and IIc, respectively (Table 1). Compared to Ia, compound I differs by the replacement of the ethyl group (R 1 ) with methyl in N3.
Besides the Et→Me replacement, the COOEt at C5 in IIc was replaced by COOMe in II. While these changes simplify the conformational space of the molecules, and reduce the expensiveness of computations, they are not expected to have a significant influence on the pK a of compounds Ia and IIc, which was determined to be 6.12 (±0.04) and 8.76 (±0.12), respectively (values in parentheses represent standard deviations of the mean pK a s, as obtained from the fitting to potentiometric curves of at least three independent determinations). Finally, the choice of these compounds was also motivated by the structural information available for the same (Ia) or related (IIb) compounds in the Cambridge Crystallographic Data Centre (CCDC code 218170 and 229769, respectively; Figure 1). 18 Note that the CCDC structures correspond to the neutral tautomers I-A N for Ia, and II-B N for IIb, respectively.  Finally, the pK a of the compound was determined using the standard relationship shown in Eq. 3. ( In this computational framework, the free energies of the protonated and neutral species were determined from the ensemble of conformations sampled by using the Multilevel strategy (see below). The free energy in aqueous solution of the proton was estimated by combining the experimental data of and , which are -6.28 kcal/mol and -265.9 kcal/mol. 19,20 A correction term of +1.89 kcal/mol was also introduced to account for the change in the reference states, which correspond to 1 atm and 24.46 L in the gas phase and 1 M in aqueous solution at 298.15 K.

Multilevel conformational sampling.
To explore the conformational space of the protonated (A P , B P ) and neutral (A N , B N ) tautomers of compounds I and II (Scheme 1), we adopted the computational strategy outlined in Figure 3.
pK a = ΔG aq RT ln(10) For each tautomeric species, five distinct conformers were generated with the Avogadro software 21 as starting structures to perform the conformational sampling with the Multilevel (ML) strategy. In the case of tautomer A, one of these structures was taken from the CCDC crystallographic data (see Figure 1). The ML strategy relies on the use of two computational levels to determine the conformational preferences of flexible compounds: a low-level (LL) method is chosen to perform an exhaustive conformational search, and a high-level (HL) technique is subsequently used to determine the relative stability of the conformational wells In this work, the LL method involved the use of classical molecular dynamics performed with AMBER16, 22 and the HL refinement was performed at the MP2 level with Gaussian16. 23 According to the predominant-states approximation adopted by Gilson and coworkers, 24,25 the ML strategy partitions the conformational space sampled at the LL into M local energy wells, and the free energy ( ) is approximated from the freeenergy contributions ( ) of these wells (Eq 4).
For a given well m, the free energy is expressed by combining the contribution of the minimum energy conformer ( ) within the well, and the contribution due to the rest of conformers ( ) that populate the conformational family (Eq. 5; the reader is addressed to ref. 24 for a detailed derivation of this expression).
where is the volume of configurational space for the conformational well m, stands for the number of conformational states in the well, and the contribution due to the local curvature, which reflects the width of the potential energy surface around the minimum energy conformer of the well, is given as (6) where .
Let us note that if one assumes a uniform sampling of the conformational volume associated to distinct wells m and n, the contribution due to the last term in Eq. 5 will cancel, and the relative stability between two conformational families will be determined by the difference between the energy of the minimum-energy conformers and the local curvature of the wells.
Computational details. In this work, the LL sampling was performed via classical simulations of the tetrahydroazocino compounds immersed in a box (with edges located at 12 Å from the ligand) of TIP3P water 26 molecules. The parametrization of the annulated tetrahydroazocines was made using the gaff force field, 27 and the atomic point charges were determined using the RESP 28 procedure by fitting the HF/6-31G(d) electrostatic potential. A chloride anion was added to the simulation systems corresponding to protonated compounds. After energy minimization, the temperature of the system was raised from 50 to 298 K in 200 ps at constant volume, and the density of E m _min The application of this procedure to the set of structures that define the curvature of the well would be computationally cumbersome. Therefore, its contribution was estimated from the LL sampling by projecting the conformers onto a k-dimensional grid, with k being the number of active torsions associated to the flexibility of the azocino ring and attached substituents. For each grid element j, an effective energy was estimated from the conformer population, and the local curvature contribution was determined from the ratio (Eq. 8) between the population of sampled structures assigned to grid element j ( ) and the grid point with the largest population ( ).

(8)
Experimental determination of pK a . Ionization constants (pK a ) were determined through potentiometric titration using a Sirius GLpKa instrument (Sirius Analytical Instruments Ltd.

RESULTS AND DISCUSSION
A remarkable difference in the experimental pK a values of the cyclic enamine derivatives investigated herein, with pK a values ranging from 6.1 to 7.3 (median 6.7) for α,β-diester derivatives, and from 8.1 to 9.0 (median 8.6) for β-ester derivatives, is observed irrespective of the fusion isomerism of the 1H-indole-fused tetrahydroazocine scaffolds I and II (Table 1). Even considering the involvement of the tautomeric equilibrium shown in Scheme 1, a consistent description of the differences in pK a values cannot be gained from chemoinformatic tools. When the pK a of these compounds was estimated with MarvinSketch ® software, 40 the results predicted for the amino species (tautomer B) were similar to the experimental values, whereas the pK a of the enamino species (tautomer A) differed by more than 8 pK a units (Table 2). However, the results obtained using ACDLabs 41 showed a slight sensitivity to the tautomeric species, as the pK a values determined for tautomers A and B differed by less than 1 pK a unit ( Table 2).
These differences are surprising, as several studies have reported that these methods exhibit a close agreement in the pK a prediction of ionizable sites for a variety of compounds, leading to pK a estimates that accurately reproduce the experimental values (generally with mean average deviations < 0.7 log units). [42][43][44][45] However, these studies have often found discrepancies in the predicted pK a values that appear to be related to the presence of specific chemical features, 45 which likely originate from the different nature of the formalisms used by these tools in the pK a prediction and the chemical diversity and structural complexity of the compounds included in the datasets. Overall, these data pose the question about the origin of the chemical factors that determine the experimental pK a of the two sets of tetrahydroazocino enamino ester derivatives. Here to follow, we examine separately the interplay between ionization and tautomerism and its impact on the bioactive species of α,β-diester and β-ester of the fused heterocyclic scaffolds I and II, respectively. Prediction of pK a in the Multilevel framework. Before undertaking this study, the chemical accuracy of the HL method used in the ML strategy for predicting the pK a was checked for a series of nitrogen-containing heterocyclic compounds ( Table 3). The pK a was estimated from Eqs. 1-3, taking advantage of the lack of relevant conformational effects in these compounds. in Table 3.
There is a nice agreement between experimental and predicted results. The calculated pK a values tend to be, on average, around 0.5 pK a units lower/higher than the experimental ones at the ML(MP2) and ML(CCSD(T)) levels, respectively, although this is mainly due to the deviation in the predicted value of pyridine at the ML(MP2) level, and to morpholine in the ML(CCSD(T)) case. Nevertheless, the two methods lead to a similar root-mean square deviation (RMSD), which amounts to ca. 1 pK a unit.
Remarkably, the similar accuracy attained from these computations is achieved at a much lower cost at the ML(MP2) level. Overall, this analysis supports the suitability of the QM level of theory implemented in the ML strategy for the computation of pK a in the series of tetrahydroazocino ester derivatives.   The conformational sampling performed for I-A P and I-A N reveals the preference for structures characterized by a folded conformation of the azocino ring ( Figure 5). The folded arrangement is more pronounced in the protonated structure, I-A P , as it favors the electrostatic interaction between the positive charge of the protonated amine and the π-electron cloud of the indole ring. The conformational flexibility primarily reflects the orientations adopted by the ester groups, leading to two main conformational states corresponding to 71.4% (A P 1 ) and 27.4% (A P 2 ) respectively, for the protonated structure ( Figure 5 and Supporting Information Figure S1), and four conformers for the neutral one, with a population ranging from 55.6% to 10.2% ( Figure 5 and Supporting Information Figure S2). It is worth noting that the conformation of the tricyclic ring is highly similar in the four conformers of I-A N , whose chemical stability lie in a close range (≤ 1 kcal/mol), and that the differences mainly concern the relative arrangement of the ester groups. In fact, there is a nice overlay between the crystallographic structure of compound Ia and the first conformation (population close to 56%), as the differences in the orientation of the ester groups may arise from the replacement of the N-ethyl moiety (X-ray) by N-methyl (computational model) and the packing with other molecules in the X-ray structure ( Figure 6). With regard to the alternative tautomer B, the protonated species I-B P is found in three main conformational states (Supporting Information Figure S3). In contrast to the results obtained for I-A P , the shift of the double bond in the azocino moiety favors an extended arrangement of the fused indole-azocino rings, where the most stable conformation (B P 1 ) has a population of 63.6% ( Figure 5). Finally, the conformational space of tautomer I-B N is represented by four major species (Supporting Information Figure S4), which reflect different arrangements of the ester substituents, albeit the main conformer contributes 62.2% of the conformational population ( Figure 5).

Multilevel analysis of compound
Taking into account the conformational preferences of protonated and neutral species, the protonated tautomer I-B P is favored by 15.4 kcal/mol compared to I-A P (Figure 7), indicating that the population of I-A P is negligible in aqueous solution, and that the protonated form can be represented by I-B P . The difference in stability between the two tautomers is much lower in the neutral state, and in fact the tautomer I-A N is favored by almost 2 kcal/mol. According to Figure 7, the pK a estimated for the deprotonation of the acidic species I-A P (pK a = -6.1) agrees with the value estimated using MarvinSketch ® (pK a of -5.7 for the enamino species of compound Ia in Table 2). Nevertheless, since protonation markedly favors the stability of tautomer I-B P , the titration of compound I can be described as the process that converts I-B P into I-B N , leading to a predicted pK a of 6.4, which agrees with the experimental pK a determined for compound I (pK a = 6.1).
Let us note, however, that the conversion between I-B N and I-A N should be kinetically impeded, because a fast conversion from I-B N to I-A N , which is found to be the thermodynamically favored species, would lead to an estimated pK a of 5.2. Multilevel analysis of compound II. The same methodological approach was used to explore the protonated and neutral species of compound II (Figure 8). The protonated species II-A P presents up to 8 distinct conformational states, including either structures where the azocino ring is folded toward the indole ring, reflecting the stabilization afforded by the positive charge of the protonated nitrogen with the π-electron density of the indole ring, or alternatively found in more extended arrangement. Indeed, the most populated folded and extended conformers account for almost 22% (A P 1 ) and 19% (A P 2 ) of the conformational space (Figure 9; see also Supporting Information Figure   S5). The flexibility of the azocine ring is also found in the neutral species, II-A N , which populates four major conformational states, the main species with a population close to 40% (Figure 9 and Supporting Information Figure S6). It is worth noting that the skeleton of the most populated conformation superposes well the crystallographic structure of compound IIb ( Figure 10).   In contrast to these results, the conformational distribution of the protonated species II-B P is primarily reduced to a single conformational state, where the azocino ring adopts an extended arrangement stabilized by the intramolecular interaction between the protonated nitrogen and the carbonyl oxygen of the ester in position β (population close to 96%; Figure 9 and Supporting Information Figure S7). Finally, deprotonation increases the conformational flexibility of the neutral species II-B N , which can be found in four conformations that are similarly populated (Figure 9 and Supporting Information Figure S8).
By combining the relative stabilities of the conformations sampled for protonated species II-A P and II-B P , this latter tautomer is favored by 10.9 kcal/mol, thus mimicking the behaviour found for compound I (see Figures 7 and 11). The analysis of the neutral forms reveals that II-A N was more stable than II-B N by 2.6 kcal/mol ( Figure   11). If one assumes that deprotonation of II-B P leads to the thermodynamically favored species II-A N , the predicted pK a would be 6.8, which is almost 2 pK a units lower than the experimental value (pK a = 8.8). In contrast, a nice agreement is found when the pK a is determined for the conversion of the protonated species II-B P into the neutral form II-B N , as the predicted pK a amounts to 8.7 ( Figure 11). This suggests that the conversion of the neutral tautomer II-B N into the thermodynamically more stable tautomeric form II-A N should be kinetically impeded. In turn, this raises the questions about the most feasible mechanism involved in the prototropic tautomerism and the role of solvent molecules in assisting the proton transfer, which will be addressed in future studies. Finally, let us also remark that the predicted value for the conversion of the protonated species II-A P to II-A N leads to a pK a value of -1.3, which matches the empirical value provided by MarvinSketch ® (pK a of -0.8 for the enamino species of compound IIc in Table 2).

Effect of ester substituents on the tautomerism of protonated and neutral species.
Comparison of the results shown in Figures 7 and 11 allow us to identify key factors that justify the differences in basicity of the two series of compounds represented by compounds I and II.
In the two protonated compounds there is a marked preference for the tautomeric species B P , which is stabilized by more than 10 kcal/mol relative to tautomer A P .
Comparison of the hydration free energies of A P and B P reveals small differences (< 2 kcal/mol), thus showing that the solvation of the positive charge has a minor effect in the relative stability between protonated tautomers. Therefore, the preference for B P can be attributed to the extended π-electron conjugation triggered upon shifting of the double bond between positions 4 and 5 in A P to positions 5 and 6 in B P , enabling the conjugation with the electron density of the indole ring. In contrast the slight preference found for tautomer A N in the neutral species suggests a larger sensitivity to the specific pattern of the COOR 3 group(s) in α and β positions to nitrogen N3.
To examine the influence of the ester groups, the relative stability of the two tautomers was determined for a simplified molecular system containing only the azocino ring with different ester substitutions, replacing the other chemical groups by hydrogen atoms (Supporting Information Table S1). For the model compound I, the presence of the ester group in position α or β changes the relative stability from -3.9 to +2.7 kcal/mol, although there is a large cancellation of their influence when both esters are added simultaneously. For the model compound II, the tautomeric preference for the unsubstituted azocino ring is changed by 5.8 kcal/mol upon addition of the ester in position β, which tends to stabilize tautomer A N . Overall, the results highlights the influence exerted by the electron-withdrawing ester groups on the tautomerism of tetrahydroazocino ester derivatives.
Effect of ester groups on the pK a . The preceding discussion points out that the experimental pK a describes the deprotonation of tautomers I-B P and II-B P to I-B N and II-B N , respectively (Scheme 2). This suggests that the carbomethoxy group in position β exerts would increase the basicity of the nitrogen atom by ca. 2 pK a units.
The preceding trends were further confirmed by experimental measurements of the pK a of two additional pairs of enamino ester-containing annulated medium-sized cyclic derivatives (Chart 1), namely β-CO 2 Me and α,β-(CO 2 Me) 2 derivatives of 4-ethyl-1,4,5,6,7,8-hexahydroazonino [5,6-b]indole (IIIa-b) 46,47 and 3-ethyl-1,2,3,6-tetrahydro- 48 The hexahydroazonino [5,6-b]indole scaffold III is a ring-expanded (azocine → azonine) analog of I, whereas the dimethoxybenzo moiety in IV replaces the indole ring in II. The pK a values were determined by potentiometric titration, but titrations were carried out in water-methanol mixtures due to the low water solubility of the examined compounds ( Figure 12). The apparent pK a values enabled an accurate extrapolation to the 100% aqueous pK a ( Table 4) according to 38 Scheme 2. Representation of the deprotonation process for the ester derivatives of azocino compounds I and II. Chart 1. Enamino (di)ester derivatives whose pK a values have been measured in this study to confirm theoretical predictions: (III) 1,4,5,6,7,8-hexahydroazonino [5,6-b]indole, and (IV) 1,2,3,6-tetrahydro-8,9-dimethoxybenzo[d]azocine. Diester (R 2 = CO 2 Me) and monoester (R 2 = H) species are denoted a and b in the text.   Figure 7). For the α,β-diester derivative IIIa, the observed pK a is 6.5, which is close to the values reported in Table 1 for both Ia (pK a = 6.1) and Ib (pK a = 6.7). The pK a of the diester compound IVa follows the same trend (pK a = 6.0). On the other hand, the absence of the α-carbomethoxy group does increase the basicity of the enamino group by more than 2 pK a units: from 6.5 (IIIa) to 8.6 (IIIb) and from 6.0 (IVa) to 8.3 (IVb), following the behavior observed for Ic (pK a = 8.2). These results can be understood if one assumes that the measured pK a corresponds to the deprotonation of the tautomeric species B P .

CONCLUSIONS
The analysis of the results presented in this work points out that the presence of tautomeric forms that convert the enamine species into a secondary amine has a key influence on the experimental pK a of the ester derivatives of annulated tetrahydroazocino (and hexahydroazonino as well) compounds. The results reveal that the tautomeric form B is predicted to be the most stable species for the protonated form in the two representative compounds considered in our study. Therefore, rather than the enamine species, titration of these azocino-based compounds reflects the deprotonation of a secondary amine, even though the experimental pK a also reflects the synergy between different factors, such as the pattern of ester substituents attached to the azocino ring.
If one assumes that deprotonation is controlled by thermodynamic stability, the predicted pK a for compounds I and II would underestimate the experimental value by 1- Keeping in mind that tautomer B represents the protonated form of these compounds, this study provides a basis to rationalize the differential effects of the two ester groups on the modulation of the pK a . Thus, whereas a stabilization of the π-electron delocalization with the indole ring can be expected from the ester group in position β, the ester group in position α exerts an inductive effect that decreases the basicity of the azocino nitrogen by almost 2 pK a units. Accordingly, the apparently minor chemical modification arising upon deletion of this ester group may have a drastic impact on the nature of the most populated species at physiological pH, affecting the pharmacodynamics and pharmacokinetics profiles of these compounds.
Finally, this study exemplifies the suitability of the Multilevel strategy to take into account the influence exerted by the substituents on the interplay between ionization, tautomerism, and conformational flexibility, which enables a subtle regulation of the bioactive species. This approach can then be valuable to explore the relationships between biological activity and the bioactive species of drug-like compounds at two levels. First, exploring the complementarity between the chemical groups present in a specific compound, which depend on the ionization, tautomerism and conformational preferences, and the 3D arrangement of residues in the binding pocket of the target.
to estimate a priori the suitability of the pharmacokinetics profile of the compound.

ASSOCIATED CONTENT (S) Supporting Information
The Supporting Information is available free of charge on the ACS Publications website at DOI: • Figures S1-S8 with conformational preferences of the tautomers of compounds I and II • Table S1 with the relative energies of tautomers for simplified models of of the 1,2,3,6-tetrahydroazocino moiety.

Notes
The authors declare no competing financial interest.