Tyler J.
Finnegan‡
,
Christopher
Mortensen‡
and
Jovica D.
Badjić
*
The Ohio State University, Department of Chemistry & Biochemistry, 100 W 18th Avenue, Columbus, OH 43210, USA. E-mail: badjic.1@osu.edu
First published on 16th January 2024
Molecular basket 1^{6−} comprising a nonpolar cavity and an anionic nest of six carboxylates at its rim was found to form inclusion complexes with (1R, 2S)-ephedrine, (1R, 2R)-pseudoephedrine, and (1S, 2R)-tranylcypromine. Experimental results (NMR) and theory (MM/DFT) suggest the basket encapsulates phenethylamines in unique and predictable fashion.
Molecular basket 1 was obtained by following an earlier described synthetic protocol.^{12} This molecule can be dissolved in aqueous 30 mM phosphate buffer solution (PBS) at pH = 7.0 to give hexaanionic 1^{6−}. At concentrations of 5.0 mM or lower, amphiphilic 1^{6−} was shown to be monomeric with results from DLS, UV-Vis, MS, and NMR experiments all suggesting the absence of aggregation.^{12} Importantly, ^{1}H NMR spectrum of 1^{6−} (Fig. 3A) comprises a set of sharp signals corresponding to, on average, a C_{3} symmetric compound. A Monte Carlo conformational search (OPLS3) in implicit water solvent showed the presence of two conformers 1_{A/B}^{6−} within 1 kcal mol^{−1} (Fig. 2; Table S3, ESI†); each conformer was additionally optimized at a higher level of theory (DFT:B3LYP/6-31+G*). While 1_{A}^{6−} has all three α-carboxylates from glutamic acid residues pointing to the concave side of the host, 1_{B}^{6−} has one of these carboxylates at the convex face. Along with computational results, we reasoned that a rapid interconversion of 1_{A}^{6−} to 1_{B}^{6−} (Fig. 2), in addition to each molecule's conformational dynamics, contributed to the observed ^{1}H NMR spectrum of 1^{6−}.
An incremental addition of a standard solution of phenethylamines 2^{+}–4^{+} to 1^{6−} was monitored (30 mM PBS at pH = 7.0) with ^{1}H NMR spectroscopy (Fig. 3 and Fig. S1–S3, ESI†); each titration was repeated three times. Importantly, additions prompted a steady change (i.e., magnetic shielding) of the basket's resonances (Fig. 3A). We reasoned that noncovalent complexation ought to be taking place with the host residing in the shielding region of aromatic guests (Fig. 3B). To determine the stoichiometry of the complexation, the change in chemical shift of seven resonances from 1^{6−} as a function of the concentration of 2^{+}–4^{+} was subjected to nonlinear regression analyses using 1:1, 1:2 and 2:1 binding models (Fig. S4–S6, ESI†).^{16} In this regard, it is important to note that ^{1}H NMR resonances of drugs 2^{+}–4^{+} remained constant as a function of their variable concentrations (Fig. S7–S9, ESI†) so no competing dimerization or aggregation of drug molecules took place during supramolecular titrations. The curve fitting result for the titration of pseudoephedrine 3^{+} to 1^{6−} using 1:2 binding model (i.e., [3⊂1]^{5−} and [3_{2}⊂1]^{4−} complexes) is shown in Fig. 3B. The model is accepted given a somewhat random distribution of residuals^{17} and the small covariance of the fit (i.e., cov_{fit} = 3 × 10^{−4}; Fig. S5, ESI†).^{18,19} When the same criteria were used for evaluating the goodness of fit (GOF) for 1:1 and 2:1 binding models, we could disqualify both 1:1 (cov_{fit} = 3 × 10^{−3}) and 2:1 complexations (cov_{fit} = 6 × 10^{−4}).^{19} For all three drugs, the covariance of the fit slightly favoured 1:2 over 2:1 models of association (Fig. S4–S6, ESI†). Additionally, we decided to examine Akaike's information criterion (AIC).^{20} This information-theoretic approach was recently introduced^{20} to the field of host–guest chemistry as a superior alternative for evaluating the GOF. To calculate AIC, one needs to use the formula AIC = Nln (SSR/N) + 2k in which N is the number of fitted points, SSR is the sum of the squared residuals and k is the number of fitted parameters. The smaller the AIC the better the GOF. After computing AIC values for three independent titrations of each drug to 1^{6−} (Table 1 and Tables S1 and S2, ESI†), in addition to Akaike weights w_{i}, we compared the data. First, 1:1 complexations are uniformly showing the least favorable AIC scores (i.e., the most positive numbers). Between the other two models, 1:2 gave lower AIC values than 2:1 in two out of three titrations for each drug. We conclude that 1:2 association ought to be dominating under the experimental conditions. To corroborate statistical considerations of the fit, ESI† mass spectrometry measurements of a mixture of phenethylamine drugs 2^{+}–4^{+} with basket 1^{6−} were taken in 1:10, 1:1, and 10:1 ratios (Fig. S10–S12, ESI†). At all ratios, only the presence of 1:1 and 1:2 complexes were observed. While the electrostatic repulsion between two anionic 1^{6−} disfavors 2:1 complexation, we hypothesize that 1:2 complex comprises the first guest occupying the cavity of 1^{6−} while the second one undergoes an exo-complexation.
Model 1:1 | Model 1:2 | Model 2:1 | ||
---|---|---|---|---|
Titration I | AIC | −1039.7 | −. | −1113.3 |
w _{ i } | 4 × 10^{−67} | 2 × 10^{−26} | ||
Titration II | AIC | −1108.8 | −. | −1232.5 |
w _{ i } | 7 × 10^{−57} | 4 × 10^{−3} | ||
Titration III | AIC | −1541.3 | −1619.2 | −. |
w _{ i } | 1 × 10^{−190} | 9 × 10^{−157} |
After elucidating the binding stoichiometry, we went on to analyse the thermodynamic stability of binary [drug⊂1]^{5−} and ternary [drug_{2}⊂1]^{4−} complexes (Table 2; see also Fig. S4–S6, ESI†). Evidently, the formation of binary complexes dominates each equilibria with K_{1} ≫ K_{2}. This bodes well with the earlier notion that [drug⊂1]^{5−} is an inclusion and more stable complex than [drug_{2}⊂1]^{4−} in which the association takes place on the basket's outer side. The stability of 1:1 complexes are in the millimolar range, but interestingly the affinity of tranylcypromine 4^{+} toward 1^{6−} is an order of magnitude greater than ephedrine 2^{+} and pseudoephedrine 3^{+}.
Phenethylamine drug | K _{1} (M^{−1}) | K _{2} (M^{−1}) |
---|---|---|
(1R, 2S)-ephedrine 2^{+} | 7 ± 1 × 10^{3} | 54 ± 11 |
(1R, 2R)-pseudoephedrine 3^{+} | 6 ± 2 × 10^{3} | 58 ± 11 |
(1S, 2R)-tranylcypromine 4^{+} | 2 ± 1 × 10^{4} | 39 ± 14 |
After quantifying the thermodynamic parameters characterizing the binding, we sought out to learn more about the mode of complexation and intermolecular interactions driving the association. First, the magnetic perturbation of resonances from each drug undergoing complexation show phenyls’ H_{a/b/c} experiencing the largest magnetic shielding (Δδ = δ_{bound} − δ_{free}, Fig. 4A; Fig. S1–S3, ESI†). Inclusion of the phenyl ring into the basket's cavity with H_{a} reaching for the cavity's bottom would explain the observation.^{13} Along with such docking of the aromatic, the ammonium site from 2–4^{+} ought to reside in the anionic nest (Fig. 1C). If so, what is the exact role of α- and γ-carboxylates in the complexation and what is the dominant pose of each guest with respect to the basket host? To answer these questions, we developed a computational protocol to estimate a pose which best represents the observed NMR shielding effect. First, the magnetic environments of 1_{A}^{6−} and 1_{B}^{6−} were mapped by computing nucleus independent chemical shift values (NICS) for a grid around the molecules (Fig. S14 and S15, ESI†).^{21} Next, we ran Monte-Carlo conformational searches (OPLS3) for inclusion complexes [2–4⊂1_{A}]^{5−} and [2–4⊂1_{B}]^{5−}. During MC searches, we froze conformational motions of 1_{A/B}^{6−} but allowed guests to change their position and conformation, akin to docking protocols. Finally, we algorithmically assigned shielding values (Δδ_{comp}) for proton nuclei of all MC conformers of [2–4⊂1_{A}]^{5−} and [2–4⊂1_{B}]^{5−} using the NICS maps from 1_{A}^{6−} and 1_{B}^{6−}. Having computed (Δδ_{comp}) and experimental (Δδ_{exp}) values of proton shifts, we analysed the data in two different ways. For the first approach, we determined root mean squared error for each computed pose and the pose with the lowest RMSE was assumed to be the best representation of the experimental result (Table 3, Fig. 4B; see also Fig. S19–S21, ESI†). As for the second approach, we Boltzmann weighted Δδ_{comp} values for each proton from all computed poses and summed each poses contribution to the ensemble average 〈Δδ_{comp}〉. Next, RMSEs for [2–4⊂1_{A}]^{5−} and [2–4⊂1_{B}]^{5−} were determined using the 〈Δδ_{comp}〉 values (Table 3). From Table 3, one notes that the ensemble of poses is, in more cases than not, giving lower RMSEs than single structures. It follows that the ensemble is more effective than a single pose at describing the experimental result. Furthermore, the values for RMSEs from complexes of 1_{A}^{6−} and 1_{B}^{6−} are comparable to indicate that both conformers likely participate in the binding; however, we noted that 1_{A}^{6−} showed a closer agreement to experimental NMR shielding values (Fig. S16–S18, ESI†). For (1R, 2S)-ephedrine 2^{+} occupying 1_{A}^{6−}, RMSE values for the ensemble of poses (0.52) and a single pose (0.79) are similar (Table 3).
Phenethylamine drug | Type | 1 _{A} ^{6−} (RMSE) | 1 _{B} ^{6−} (RMSE) |
---|---|---|---|
(1R, 2S)-ephedrine 2^{+} | Ensemble | 0.52 | 0.43 |
Single | 0.79 | 0.89 | |
(1R, 2R)-pseudoephedrine 3^{+} | Ensemble | 0.86 | 0.88 |
Single | 0.65 | 0.70 | |
(1S, 2R)-tranylcypromine 4^{+} | Ensemble | 0.50 | 0.78 |
Single | 0.95 | 1.0 |
With an assumption that this single pose presents an average picture of the ensemble, the structure of [2⊂1_{A}]^{5−} in Fig. 4B shows ephedrine anchoring its benzene in the basket's cavity (via C–H⋯π contacts) while using hydroxyl and ammonium groups to form a hydrogen bond and salt bridge with α-carboxylates. Importantly, all of the best fit single poses of 2–4^{+} bound to 1_{A/B}^{6−} show the benzene from phenethylamines occupying the basket's cavity to contribute to non-classical hydrophobic effect^{22} with ΔH° < 0 (Fig. S13, ESI†) while OH and NH^{+} groups hydrogen bond/salt bridge α- and to a smaller degree γ-carboxylates (Fig. S19–S21, ESI†).
In conclusion, molecular basket 1^{6−} forms inclusion complexes with phenethylamines in water. The results of both experimental and computational studies suggest that the probed drug molecules anchor their phenyl ring in the basket's cavity as driven by C–H⋯π contacts and the hydrophobic effect while forming hydrogen bonds and/or salt bridges with primarily α-carboxylates at the rim. With a well characterized mode of binding, the opportunity to install alternative amino acids or peptides atop our basket scaffold is clear. In addition to the investigation of structure function relationships, further derivatization may indeed furnish a better agent for the preparation of chemosensors or sequesters of phenethylamines.
This study was financially supported with funds obtained from the National Science Foundation under CHE-2304883.
Footnotes |
† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d3cc05485c |
‡ These two authors contributed equally to this study. |
This journal is © The Royal Society of Chemistry 2024 |