Introduction

Metal-catalyzed enantioselective transformations often employ chiral ligands to control the enantioselectivity of the products. Chiral ligands work by affecting the reaction progress and selectivity by tuning both their electronic and steric properties. The development of innovative chiral ligands, therefore, is always of strategic importance. The N-stereocenter containing chiral ligands have not been studied enough in metal-catalyzed asymmetric reactions when compared to phosphorus derivatives1,2,3,4,5,6,7,8 and organosulfurs9,10,11,12,13 (Fig. 1a). Substituted ureas are a relatively new class of chiral ligands that feature nitrogens with hydrogen atoms on nitrogen serving as key receptors for anions. Urea derivatives involving the formation of a stable conformation by the interaction of two hydrogen atoms on the nitrogens and a suitable anion have recently attracted significant attention for their remarkable efficiency, selectivity, and versatility in a range of synthetic transformations14,15,16,17,18 (Fig. 1b). The chiral, pseudo-cyclic structure that hinders the inversion of the N atom can work to discriminate between the enantiomers of a substrate based on their different sizes, shapes, and electronic properties. In general, chiral urea ligands act as notable hydrogen bond donors, which is key to their efficacy in promoting asymmetric reactions with high enantioselectivity19,20,21,22,23,24. In addition to the H-bonding, the lone pairs of the nitrogen and oxygen atoms of urea ligands have powerful coordination ability, while rarely been used25. In order to explore such potential, we designed an innovative chiral cyclic urea ligand that switches the hydrogen-bonding ability to nitrogen/oxygen coordination (Fig. 1c). In doing so, we successfully developed an innovative asymmetric catalytic intramolecular hydroamination of alkenes (intra-HA) and obtained pyrrole, piperidine, and indoline derivatives with excellent enantioselectivity and high yields.

Fig. 1: The design of a chiral cyclic urea ligand.
figure 1

a The easy inversion of the N-stereocenters make them less attractive in metal-catalyzed asymmetric reactions than phosphorus and sulfur. b Chiral urea ligands were shown to slow down the inversion through hydrogen bonding. c Our chiral cyclic urea ligand design.

Full size image

Chiral N-heterocyclic compounds incorporating stereocenters adjacent to the nitrogen atoms are commercially valuable chemicals prevalent in many biologically active natural products and drug candidates. Therefore, efficient methods for the preparation of chiral amines have been a hot research topic in the last several decades. Catalytic intra/inter-HA reaction received considerable attention, as this conversion would represent an attractive route to make amines derivatives26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42. Most intra-HA reactions target the synthesis of achiral pyrrole and piperidine derivatives, but only a few report enantioselective (>90% ee) products. Of particular note is the seminal work by Marks and colleagues reporting enantioselective intra-HA reactions enabled by chiral lanthanide metal complexes (Fig. 2a)26,43,44,45. Other elegant examples reported by Schafer46 and Sadow47,48,49 on chiral zirconium amidate complexes, providing good enantioselectivity for a range of terminal alkenes (Fig. 2b). Buchwald and co-workers described a highly enantioselective intra-HA reaction through Rh complexes (Fig. 2c)50. More recently, Dorta et al. reported chiral cationic NHC–iridium complexes as catalysts for asymmetric intra-HA reactions (Fig. 2d)51. To the best of our knowledge, there are no examples of earth abundant Mn metal-catalyzed enantioselective intra-HA reactions using chiral cyclic urea ligands. Herein, we disclose examples of chiral urea ligand-catalyzed asymmetric intra-HA reactions with excellent yields and enantioselectivity (up to 99% ee) (Fig. 2f).

Fig. 2: The introduction of asymmetric intra-HA reactions, previous work and this work.
figure 2

a–d Previous examples of enantioselective intra-HA reactions. e Our previous work. f This work.

Full size image

Results and discussion

Reaction conditions optimization

Our discovery started with a chiral (Salen)Mn(III) complex that we previously developed to facilitate enantioselective intramolecular halo-amination reactions of alkenes through a chiral aziridinium ion ring-opening sequence (Fig. 2e)52. Building on a similar design, we employed a combination of Mn (II) and chiral cyclic urea ligands to carry out intra-HA reactions, which resulted in the production of highly enantioselective pyrrole, piperidine, and indoline derivatives. This approach represents the utilization of the oxygen coordination ability of chiral cyclic urea in achieving asymmetric intra-HA reactions and introduces an innovative strategy and application of chiral cyclic urea in asymmetric catalysis.

Through investigation of enantioselective intra-HA reactions, we anticipated that structural variety in chiral urea could reveal vastly different effects on reactivity (Fig. 3a, Cat1-Cat12). In our systematic evaluation, the reaction did not occur with the addition of chiral acyclic urea or cyclic urea with hydrogen-bond donor ability, such as Cat1, Cat2, Cat7, Cat9, and Cat11, while the chiral acyclic urea and cyclic urea without hydrogen-bonding ability displayed excellent reactivities (Cat3-Cat6, Cat8, Cat10, and Cat12). The consistent experimental observations have led our investigation into uncovering the underlying mechanisms. We, therefore, performed quantitative molecular surface electrostatic potential (ESP)53,54 analysis on the reactions of Cat7, Cat8, 1a, and MnBr2 (Fig. 3b). It was revealed that the electrostatic potential minimum of Cat8 is lower than that of Cat7 (Fig. 3b, −35.8 kcal/mol vs −30.2 kcal/mol). Moreover, due to the presence of multiple conformations in Cat7, the corresponding positions in other tautomers of Cat7 exhibited electrostatic potentials much higher than that of Cat8 (Fig. 3b, −5.44 kcal/mol and −12.1 kcal/mol vs −35.8 kcal/mol). According to the principle of electrostatic complementarity, it could be inferred that the binding between the oxygen of Cat7 and MnBr2 is weaker compared to that between the oxygen of Cat8 and MnBr2, which results in the inability of the reaction to proceed smoothly when employing Cat7 as the catalyst ligand. In contrast, chiral aprotic urea ligands, exemplified by Cat8 with highly structural analogy to Cat7, exhibit unique oxygen coordinative propensity unimpeded by pronounced hydrogen bonding or tautomerism. Cat8 readily coordinated to MnBr2 to catalyze an intra-HA reaction. Encouraged by the above-mentioned results, we conducted a vigorous screening of chiral urea ligands, which displayed remarkable effects on the outcome of the ee value. The aprotic chiral cyclic urea ligands furnished chiral 2a in 75–85% yield and 85–91% ee (Cat4-Cat6). The reaction of aprotic chiral acyclic urea ligands, on the other hand, exhibited superior conversion, affording 2a with 91–98% yields, albeit in racemic form (Cat3, Cat8, Cat10, and Cat12). The results with vastly different enantioselectivity clearly indicated that the aprotic chiral cyclic urea ligands can effectively hinder the inversion of the N atoms on urea, which can successfully discriminate between the enantiomers of 2a. Thus, we determined that the optimal chiral urea ligand is Cat5, achieving 2a in 85% yield with 91% ee.

Fig. 3: Optimization of chiral urea ligands.
figure 3

a Intra-HA of 1a using different chiral urea ligands. b The analysis of molecular surfaces with ESP of Cat7, Cat8, and MnBr2.

Full size image

With the optimized chiral catalyst Cat5, we further examined other reaction parameters and found that the solvent and temperature had great influence on the reactivity and ee values (Table 1, entries 1-6). Good to high yields and excellent enantioselectivities were obtained for the solvents with a benzene ring, such as toluene and PhCl (Table 1, entries 3–4). When the temperature was below 105 °C, trace or no conversion of 1a was observed (Table 1, entry 7). When the temperature was at 107 °C (close to the boiling point of toluene), the reaction provided the best combination of reactivity and enantioselectivity (87% yield, 98% ee value, Table 1, entries 8–11). See supplementary Table 1 for more details.

Table 1 Optimization of the reaction conditionsa
Full size table

Substrate scope

We explored, next, the scope of the asymmetric transformation, focusing initially on chiral pyrrolidine derivatives (Fig. 4a). High to excellent yields and enantioselectivities were obtained for the aminoalkenes bearing para-substituted phenyl rings, regardless of the electronic properties (2a–2i, 82–98% yields and 89–98% ee). But when the position of the R group on the benzene ring at the N-substituted position of the aminoalkenes changed, yields were affected. The p-Cl substituted compounds (e.g., 2h) showed better yields than the o- and m-Cl substituted ones (2h vs 2j-2k, 88% vs 82%-83%). Notably, substituents at the meta or ortho positions of the aromatic rings were also suitable for the reaction with little effect on the enantioselectivity (2h and 2j−2k, 97–98% ee). In addition, naphthyl- and pyridyl-substituted compounds performed just as well in this asymmetric reaction to give 2l-2n in 89–90% yields and 93–98% ee, respectively. The alkyl or allyl groups at the N-substituted position of the aminoalkenes (2o−2q) also resulted in high yields (81–89%) and excellent enantioselectivities (96–98% ee). Remarkably, the Thorpe-Ingold effect was exhibited in the studied reactions, and diverse substituents on the principal chain of the substrates imparted pronounced influence on the reactivity (the yields of 2a and 2r-2t, from 87% to 62%). On the other hand, the substituents appended to the main chain of the substrates displayed minimal impact on the enantioselectivities (the ee of 2a and 2r-2t, 93%-99%). The aforementioned results indicated that the substituents appended to the main chain govern the conversion and did not exert an influence on the ee values, as evidenced by the pronounced variance in yields of 2a and 2r-2t and excellent ee of 2a and 2r-2t. As in the case of terminal aminoalkanes, the 1,2-disubstituted aminoalkane (1 u), and 1,1-disubstituted aminoalkane (1v) were smoothly transformed into the corresponding pyrrolidines in high yields (75–89%) and excellent ee values (2 u: 99% ee). Optimization of the reaction of 1,2,2-trisubstituted aminoalkane 1w yielded an unexpected result (Fig. 4b). The desired hydroamination product 2w was not obtained. Instead, the product 2w’ was likely formed through a β-elimination reaction of the C-Mn bond. This implies the possible existence of a reactive intermediate C containing a C-Mn bond during the reaction pathway (Fig. 4b).

Fig. 4: Substrate scope of enantioselective intra-HA reactions.
figure 4

a Substrate scope. b Beta-elimination route.

Full size image

We then attempted to extend our methodology to the synthesis of chiral piperidine derivatives. The results are summarized in Fig. 4 (4a−4x). Substrates 3 cyclized in 73%-88% yields and 90%-99% ee providing piperidine derivatives 4-S (Fig. 4, 4a−4k). Aminoalkenes 3a-3i bearing substitutions with different electronic properties (electron-deficient, -neutral, or -rich) at the para-position of the benzene ring completed the reaction under identical conditions, achieving 4a−4i in high to excellent yields and ee values. The substituents on the benzene ring at the N-substituted position of the aminoalkenes (ortho-, meta-, and para-) were well tolerated, and the corresponding products were isolated in 81–84% yields with 96–98% ee (Fig. 4, 4h, 4j, 4k). The presence of dialkyl substituents on the main chain, exemplified by phenyl, cyclohexane, methyl, and hydrogen substrates, exhibited a discernible Thorpe-Ingold effect on reactivity (the yield of 4a and 4r-4t). However, it is noteworthy that the Thorpe-Ingold effect minimally influences the enantioselective control of the product (the ee value of 4a and 4r-4t). Under the optimized conditions, the standard protocol was extended to 2-allylanilines (3x-3z), and they were converted, via an intra-HA reaction, to chiral indoline derivative 4x-4z with 71–78% yield and 96–98% ee (Fig. 4). It was observed that the primary amine substrates could undergo the reaction to afford the desired products 5c-S and 6b-S. Due to the absence of functional groups on the nitrogen atom of primary amine substrates, the chiral induction effect is diminished, resulting in 85% ee value. We introduced two additional types of olefin substrates, exocyclic olefin (3aa) and endocyclic olefin (1ab). Both substrates proceeded well to deliver the target products 4aa-R and 2ab-(S,S). Remarkably, the exclusive formation of the absolute configuration of product 2ab-(S,S) further corroborated that the final step of the reaction mechanism was achieved through the backside attack of HBr to the C-Mn-Br intermediate. 1ac and 1ad were implemented under the standard conditions, producing the chiral 2ac-cis/trans-S and 2ad-cis/trans-S with diastereomeric ratio of 1:1, 83–85% yield and 98–99% ee value. We also studied the effects of the concentrations of 1ac on the reaction and found that it can hardly influence the diastereomeric ratio. This experiment suggests that the reaction pathway is distinct from the reaction mechanism of Sadow’s Zr complex55.

Scale up

Considering the prospective applications of our methodology in chiral synthesis, we proceeded to verify its scalability through large-scale synthesis. A significant amount of 1a (3.27 g) was effectively converted to give 2a-S in 84% yield (2.75 g) and 97% ee (Fig. 5a). Then, we sought to employ debenzylation reactions to further explore the practical potential of our methodology (Fig. 5b). The asymmetric intra-HA reaction of substrates1c, 3b, and 3x using Cat5 under optimized conditions afforded products 2c, 4b, and 4x, followed by debenzylation that furnished products 5c, 6b, and 6x in high yields without any loss of enantioselectivity.

Fig. 5: Application of the new synthetic methodology to pharmaceuticals.
figure 5

a Gram scale conversion of 1a. b Debenzylation follow up. c, d Successful synthesis of CEP-26401 and CSH-P25 with the asymmetric intra-HA reaction developed in this study.

Full size image

Pharmaceuticals’ synthesis

To demonstrate the viability of our asymmetric conversion in real pharmaceutical development, we chose CEP-26401 platform (Fig. 5c, d). CEP-26401 (irdabisant) stands as a remarkably potent H3 receptor inverse agonist, featuring a chiral 2-methylpyrrolidine moiety as the key pharmacophore. We listed its key pharmacological parameters including human H3 receptor (hH3R) and rat H3 receptor (rH3R) binding affinities as well as lipophilicity measure clogP (Fig. 5d)56. Leveraging the intra-HA reaction we developed, a concise synthetic route to CEP-26401 and a series of its derivatives were successfully executed with high yields and ee values (Fig. 5c, d, see Supplementary Fig. 11, and Supplementary section 15 for more details). Among these derivatives, CSH-P25 compound, a novel structure, emerges as the closest analog to CEP-26401 in terms of hH3, rH3, and clogP parameters. Notably, based on the comparable pharmacological profiles, we hypothesize that CSH-P25 may exhibit potent activity as a histamine H3 receptor inverse agonist, analogous to the reference compound CEP-26401. At present, structure-activity relationships (SAR), pharmaceutical properties and in vivo activity in the rat dipsogenia models are in progress to further characterize the therapeutic potential of these derivatives.

Mechanisms

DFT calculations and complimentary experiments were performed to decipher the nature of the intermediates involved in the intra-HA reaction with aprotic chiral cyclic urea ligands (Fig. 6). As evidenced by the preceding quantitative molecular surface ESP analysis (Figs. 3b and 6a), Cat5 and MnBr2 are proposed to generate a model catalyst A in situ, via weak intermolecular interactions, with the goal of constituting catalytic species operative in an intra-HA reaction. In a high-resolution mass spectrometry (HRMS) experiment, we detected trace amounts of the manganese complex 9, generated from catalyst A and substrate 8 under the standard reaction conditions (Fig. 6b). This observation implies that catalyst A likely engages in preferential interactions with the N atom of the substrate, furnishing a postulated Mn-N intermediate C and HBr. Quantitative molecular orbital energy analysis revealed that the absolute values of the alkene π orbital energy and Mn-N bond orbital energy in complex C1s, composed of Cat5, MnBr2 and substrate 1s, were higher than the corresponding orbital energies in complex C1s’ which was lacking Cat5 (Fig. 6c). This reveals that Cat5 plays a role in increasing the orbital energies at the reactive sites, thus functioning as a chiral catalyst. Control experiments showed negligible conversion under Cat5-deficient conditions, furnishing trace quantities of hydroamination product, with recovery of most substrates (Fig. 6d). Conversely, in the presence of Cat5, excellent yields were obtained. The experimental results provide validation for the hypothesized catalytic function of Cat5 in facilitating the intra-HA reaction. To demonstrate the formation of HBr during the reaction process, we conducted an HBr trapping experiment (Fig. 6e). DMAP was introduced into the reaction system, initially presenting in a fully dissolved state. After 6 h of the reaction, precipitation was observed in the reaction system. Through 1H NMR and 13C NMR analyses, we confirmed that the precipitate is the hydrobromide salt of DMAP (DMAP•HBr) (Supplementary Figs. 9 and 10). The formation of DMAP•HBr precipitate conclusively demonstrates the generation of HBr or quaternary ammonium bromide (NR3•HBr) during the reaction process. When subjected to the optimal conditions, substrate 1w generated product 2w’, analogous to β-elimination mechanism from C-metal bonded, implying the plausible involvement of a C-Mn-Br intermediate in the reaction pathway (Fig. 6f). Deuterium labeling experiment reveals that the H atom incorporated into the product originates from the N-H fragment of the substrate (Fig. 6f). Remarkably, the exclusive formation of the absolute configuration of product 2ab-(S,S) further corroborated that the final step of the reaction mechanism was achieved through the backside attack of HBr/(NR3•HBr) to the C-Mn-Br intermediate. On the basis of the above-mentioned mechanistic studies, we conclude that the intra-HA reaction proceeds through HBr/(NR3•HBr) addition to a C-Mn-Br intermediate to furnish the target product.

Fig. 6: Mechanistic study.
figure 6

a Cat5 and MnBr2 are proposed to generate the catalyst A. b The HRMS experiment. c Molecular orbital energy analysis. d The control experiment without cat5. e The HBr trapping experiment. f The existence of C-Mn-Br intermediate. g Deuterium labeling experiment.

Full size image

In consideration of all the experimental results, we propose a plausible reaction pathway that is further corroborated by DFT calculations in Fig. 7. Initially, complex A is formed via weak (electrostatic) interactions between Cat5 and MnBr2. Subsequently, complex A reacts with substrate 1a to generate intermediate B and releases a molecule of HBr to generate intermediate C. Then, intermediate C undergoes a transformation with the alkene and amine moieties in a specific orientation to furnish chiral intermediate D or D’ through transition state TS1a-favored or TS1a-disfavored as substantiated by calculations. Analysis of the Gibbs free energy differences, revealed that the location of TS1a-favored was 3.9 kcal/mol more stable than that of TS1a-disfavored. For other types of substrates, such as 1s and 1p, the Gibbs free energy differences were also calculated, being 6.2 kcal/mol and 3.9 kcal/mol respectively (Fig. 7b and Supplementary Fig. 5). The difference is in accordance with the preferred enantiomer of the product. Furthermore, a reduced density gradient (RDG)57,58 analysis on the structure of TS1s-favored reveals weak interactions (T-shaped form)59 between the benzyl group on the nitrogen atom of the substrate and the chiral benzene ring of Cat5. This observation provides a crucial insight into the origins of the high ee values achieved in the intra-HA reactions by elucidating the predominant factors governing stereoselectivity. Finally, the chiral intermediate D/D’ reacts with HBr/(NR3•HBr) to form the target compound 2a.

Fig. 7: The proposed mechanism.
figure 7

a Proposed mechanism. b DFT calculations. c The reduced density gradient (RDG) analysis of the TS1s-favored path.

Full size image

In conclusion, we have developed chiral aprotic cyclic urea ligand-based catalytic asymmetric intra-HA reactions. Through chiral urea catalyst design, the reactivity of substrates and the enantioselectivity of the reaction were controlled. The products included a range of enantioenriched chiral pyrrolidine, piperidine, and indoline derivatives as representatives of five and six-membered N-heterocycles (89%-99% ee). DFT calculations and the complimentary experiments including high-resolution mass spectrometry (HRMS), trapping experiments, controls, substrate 1w, and isotope labeling were performed to decipher the nature of the intermediates involved in the intra-HA reaction. The studies of the Gibbs free energies revealed that the Gibbs free energy of TS1a-favored is 3.9 kcal/mol lower than that of TS1a-disfavored. The RDG analysis further showed noticeably weak interactions (T-shaped form) between the phenyl ring on the chiral carbon of Cat5 and the phenyl ring of the substrate in TS1a-favored, which accounts for the energy difference. This weak interaction leads to a lower barrier in TS1a-favored, making the reaction to proceed through lower energy. Utilizing this new catalyst in asymmetric intra-HA reactions, we have successfully synthesized a series of novel compounds with excellent enantioselectivity and prospective pharmaceutical applications as potent H3 receptor inverse agonists, including CSH-P25. The structure-activity relationships (SAR), pharmaceutics properties, in vivo activity in the rat dipsogenia models, and other pharmacological profiles of these promising drug candidates are currently under investigation.

Methods

Synthesis of ligand Cat5

(S)-α-phenylethylamine (11.0 g, 90.9 mmol) was dissolved in methanol (60 mL), then the solution was added to a round-bottomed flask (125 mL) with a magnetic stirrer. To this solution was added acrylonitrile (6.95 mL, 100 mmol) and the resulting mixture was heated at reflux for 18 h. Concentration in a rotary evaporator afforded the crude product that was distilled in a Kugelrohr at 110–113 °C/ 0.1 mmHg to give 8S (15.0 g, 95% yield).

A solution of LiAlH4 (4.36 g, 114.8 mmol) in dry THF (100 mL) was added drop-wise to the 8S (5.0 g, 28.7 mmol). The mixture was stirring at 0 °C for 48 h. Subsequently, the reaction mixture was successively quenched with dropwise addition of EtOAc, 20% NaOH aqueous solution (10 mL) and stirred until the slurry changed colour from grey to white. Finally, the mixture was extracted with CH2Cl2 (20 mL × 3) and the organic mixture was then dried with MgSO4 and concentrated in vacuo to obtain crude product 9S (4.35 g, 85% yield).

To a solution of the diamine 9S (0.34 g, 2.0 mmol) in DCM (4 mL) was added Et3N (1.2 equiv) and CH3I (1.0 equiv). Stirring was continued at room temperature overnight. The mixture was concentrated in vacuo to removed solvent and the product was purified by flash column chromatography on silica gel (MeOH/DCM = 1:30, v/v) to obtain 10S (0.29 g, 75% yield).

The diamine (10S) (4.85 g, 25.2 mmol), Et3N (7.02 mL, 50.4 mmol), and 250 mL of dry CH2Cl2 were placed in a round-bottom flask, and the resulting mixture was cooled to 0 °C. Then a solution of the triphosgene (2.65 g, 8.9 mmol) in CH2Cl2 (150 mL) was added. Stirring was continued at 0 °C for 3 h and then at room temperature for 2 d. Then, 200 mL of 1 N HCl was added at 0 °C and the resulting mixture was extracted with CH2Cl2 (100 mL × 2), the combined organic phases were washed with brine, dried with MgSO4, and concentrated in vacuo. The product was purified by flash column chromatography on silica gel (petroleum ether/ethyl acetate = 9:1, v/v) to obtain Cat5.

Catalytic experiments

Substrate (0.25 mmol), Cat5 (0.025 mmol), and MnBr2 (0.25 mmol) were charged in a 40 mL sealed tube equipped with a stir bar in Ar atmosphere. Toluene (2.5 mL) was added to the sealed tube by syringe and the mixture was stirred at 107 °C for 48 h. Upon completion, the mixture was extracted with CH2Cl2 (30 mL × 3) after cooling to room temperature. The combined organic mixture was dried with anhydrous magnesium sulfate, concentrated in vacuo, and purified over silica gel via flash column chromatography with PE and EA as the eluent to obtain the desired product.

DFT method

All DFT60 calculations were carried out using Gaussian 16 ES64L-G16RevB.0161 suite of programs and some calculated results were processed by Multiwfn53. Geometries of intermediates and transition states were optimized using the PBE062,63,64 level of theory in combination with the DFT-D3 dispersion corrections with the Becke-Johnson damping scheme (D3BJ)65,66. The def2-SVP basis set67,68 was used for all atoms. Vibrational frequency calculations were performed for all stationary points to confirm if each optimized structure was a local minimum or a transition state structure. All optimized transition state structures have only one imaginary (negative) frequency, and all minima (reactants, products, and intermediates) have no imaginary frequencies. The single-point energies and solvent effects were computed with the PBE069,70/def2-TZVPP basis sets by using the SMD solvation model71 (Solvent = Toluene). The corrections of free energies72 were applied for entropy calculations with a frequency cut-off of 100 cm−1 using the Shermo73 program. The relative Gibbs free energies are given in kcal/mol, which were calculated by adding the gas-phase thermal and non-thermal corrections at 380 K to the single-point energies. The electrostatic potential calculations were conducted at B3LYP/6-31 G(d,p) level. Localized molecular orbitals (LMO) were obtained by Hirshfeld method. Quantitative molecular orbital energy analysis was achieved by Multiwfn program. The 3D structures of optimized geometries were constructed with the VMD software74.