Theoretical Investigation of the Biogenetic Pathway for Formation of Antibacterial Indole Alkaloids from Voacanga africana.

The energetic viability of the previously proposed biogenetic pathway for the formation of two unique monoterpenoid indole alkaloids, voacafricine A and B, which are present in the fruits of Voacanga africana, was investigated using density functional theory computations. The results of these calculations indicate that not only is the previously suggested pathway not energetically viable but also that an alternative biosynthetic precursor is likely.

Introduction

Antibiotic resistance has contributed to an increasing interest in the development of plant-based antibiotics. Plant-derived natural products have been used in traditional medicine since ancient times. For example, Voacanga africana (Apocynaceae) is a flowering small tropical tree native to West Africa. Its root bark is used to treat diarrhea in Kinshasa,[1] while its stem bark has been used in the treatment of leprosy, diarrhea, ulcers, generalized edema, and microbial infections in Côte d’Ivoire, Ghana, Cameroon, and Congo.[2,3] Additionally, in Cameroon, the fruits, leaf extracts, and seed extracts are used to treat orchitis, gonorrhea, and tooth decay, respectively.[4−8] Most of the applications of V. africana are linked to its antimicrobial properties.[9,10] Pharmacological studies have found that the main bioactive compounds that are responsible for its success as a traditional medicine are monoterpenoid indole alkaloids (MIAs).[11] In a recent study,[12] the fruits of V. africana were investigated and two MIAs, called voacafricine A and voacafricine B (Scheme ), were isolated and shown to possess antibacterial activity against Staphylococcus aureus and Salmonella typhi bacteria. This study suggested a plausible biogenetic pathway (Scheme ) involving 19-epi-voacristine as a precursor. This 6/5/7/6/6 pentacycle is a major indole alkaloid present in V. africana.[13] Curious about the geometric constraints these polycyclic structures would impose on a rearrangement such as that proposed, we evaluated the energetic viability of the proposed process using density functional theory calculations to determine which elementary steps, if any, would require enzymatic intervention.[14−16]

Scheme 1

Proposed Biogenetic Pathway to Voacafricine A and B Starting From 19-epi-voacristine

In the proposed pathway,[12] dehydration of 19-epi-voacristine generates 4,20-didehydro-voacangine (Scheme ). The introduction of hydroxyl groups to the C-18 and 20 positions of 4,20-didehydro-voacangine forms a diol intermediate, which could be protonated to form A in anticipation of water loss. Loss of water and cleavage of the C-16–C-21 bond leads to proposed intermediate B, which could reclose (via N-4 attack on C-16) to form intermediate C. The conversion of A to C was our focus. Intermediate C could then be transformed to voacafricine A via epoxidation, intramolecular nucleophilic substitution, and ester hydrolysis and voacafricine B via deoxygenation.

Computational Methods

All calculations were performed with Gaussian16.[17] Geometry optimizations were performed using mPW1PW91/6-31+G(d,p), a level of theory that has been used to model many carbocation rearrangements.[18] Calculations were performed in the gas phase, water, and chloroform. Calculations with implicit solvents were performed using the Polarizable Continuum Model (PCM).[19] Continuum dielectric environments corresponding to two solvents were used: water (ε = 78.36), a very polar solvent, and chloroform (ε = 4.71), a nonpolar solvent with a dielectric constant that falls within the range of estimates for enzyme active sites.[20] Structures also were optimized at the B3LYP-D3(BJ)/6-311+G(d,p)[21] and M06-2X/6-311+G(d,p)[22] levels to compare geometries and relative energies. For key transition structures (TSs), which were confirmed to have single imaginary frequencies, intrinsic reaction coordinates (IRCs) were computed to verify the minima to which they are connected.[23] All energies given are free energies at room temperature, except those in the IRC plots, which are electronic energies. CylView2.0[24] was used to generate three-dimensional molecular images.

Results and Discussion

The computed free energy profile for the conversion of A to C in water (see the SI for results in the gas phase and chloroform) is given in Figure , and TS geometries are shown in Figure . On the basis of our results, the conversion of A to B cannot occur in a single step, but rather it would involve three discrete chemical steps.

Figure 1

Relative free energies (PCM(water)-mPW1PW91/6-31+G(d,p), kcal/mol) of species involved in the A → B → C pathway.

Figure 2

Geometries of optimized TSs from Figure . Selected distances shown on the three-dimensional molecular images are in Å.

First, the loss of water is facilitated by a nearby tertiary amine, which displaces the water to form an aziridinium ion (A1). The A → A1 reaction is predicted to have a low barrier (∼10 kcal/mol). Polycyclic natural product-derived aziridinium ions have been described previously.[25] Shown in Figure is the computed structure of intermediate A1. All four N–C distances in A1 are close to each other and are in the range expected for typical N–C single bonds (ranging from 1.47 to 1.53 Å). While some C–N–C bond angles in A1 are distorted, this structure does not show the geometric features expected for a nonclassical ion and instead resembles a classical aziridinium ion, similar to other polycyclic natural product-derived aziridium ions.[25]

Figure 3

Computed structure of intermediate A1 [mPW1PW91/6-31+G(d,p)]. Selected distances are provided in Å.

Second, cleavage of the C-16–C-21 bond coupled to a 1,2-shift of C-16 to C-20 and cleavage of the C-20–N-4 bond leads to A2. To characterize the synchronicity of these bond-forming/breaking events, the IRC for the A1 → A2 reaction (Figure ) was analyzed in detail (Figure ).[26−29] Before [TS2] is reached, the N-4–C-20 bond breaks, leading to a TSS that resembles a hyperconjugated but classical alkyl cation (Figures and 4). Subsequently, a 1,2-alkyl shift process ensues as the C-16–C-21 and C-16–C-20 bonds break and form synchronously. The resulting cation is in conjugation with the N-4 lone pair, leading to a shortening of the N-4–C-21 bond. The A1 → A2 reaction is predicted to have a barrier of ∼30 kcal/mol, which, based on the analysis above, appears to arise primarily from C–N bond cleavage. The magnitude of this barrier indicates that if this pathway is followed, enzyme-induced barrier lowering would be required, either by selective stabilization of [TS2], selective destabilization of A1, or both.

Figure 4

IRC plot obtained [PCM(water)-mPW1PW91/6-31+G(d,p)] for the formation of intermediate A2. Energies for IRC points are electronic energies in kcal/mol relative to the reactant.

Figure 5

Evolution of selected interatomic distances as the reaction progresses along the IRC (Figure ).

Third, cleavage of the C-16–C-20 bond, promoted by π-electron donation from the indole, generates intermediate B. This step is predicted to be endergonic and to have a barrier of ∼20 kcal/mol.

Conversion of B to C is predicted to proceed directly, as suggested previously, with a low barrier (∼7 kcal/mol). However, the overall conversion of A1 to [TS4] is predicted to involve a barrier of ∼26 kcal/mol, again pointing to the necessity of enzymatic intervention.

Thus, there is a problem with the A → C pathway: two barriers >25 kcal/mol are encountered en route to C. Reducing these through enzymatic intervention is not impossible but is a lot to ask of Nature. Nonetheless, we examined this possibility using “theozyme” calculations.[30,31] Since the A1 → A2 barrier was predicted to be the largest, we focused on its reduction. We examined complexes of A1 and [TS2] with one or two water molecules (also models of hydroxyl-containing enzyme sidechains), formate (a simple model of Asp and Glu), trifluoroacetate (a model of Asp and Glu that accounts for pKa lowering in an enzyme), and benzene (a model of aromatic enzyme sidechains). In none of these cases did the predicted A1 → A2 barrier drop below 24 kcal/mol (see the Supporting Information for details). While these results do not definitively rule out the possibility of extreme enzymatic barrier lowering, we consider it unlikely. In addition, alternative mechanisms connecting A to C were considered, but none were found to have lower barriers than that discussed above (see the Supporting Information for details). It is notable, however, that natural products with skeletons similar to that of B but derived from different precursors have been described.[32]

Conclusions

Using DFT calculations, a proposed rearrangement mechanism involved in the biosynthesis of voacafricine A and B was examined and found not to be energetically viable. Additional calculations addressing the potential for lowering the barrier for this process through selective transition state stabilization by an enzyme did not reveal an array of potential active site groups that would lower the barrier into a biologically reasonable range. Given these results, and our inability to find an alternative path connecting the same reactants and products, we suggest that alternatives to structures A and/or C be considered.