Mechanism of the Stereoselective Catalysis of Diels–Alderase PyrE3 Involved in Pyrroindomycin Biosynthesis

Abstract

The biosynthesis of pyrroindomycins A and B features a complexity-building [4 + 2] cycloaddition cascade, which generates the spirotetramate core under the catalytic effects of monofunctional Diels–Alderases PyrE3 and PyrI4. We recently showed that the main functions of PyrI4 include acid catalysis and induced-fit/conformational selection. We now present quantum mechanical and molecular dynamics studies implicating a different mode of action by PyrE3, which prearranges an anionic polyene substrate into a high-energy reactive conformation at which an inverse-electron-demand Diels–Alder reaction can occur with a low barrier. Stereoselection is realized by strong binding interactions at the endo stereochemical relationship and a local steric constraint on the endo-1,3-diene unit. These findings, illustrating distinct mechanisms for PyrE3 and PyrI4, highlight how nature has evolved multiple ways to catalyze Diels–Alder reactions.

Graphical Abstract

1. INTRODUCTION

The Diels–Alder (DA) reaction is a powerful transformation capable of constructing complex cyclic structures within a single synthetic step with remarkable regio- and stereocontrol.¹ In the past decades, great attention has been paid to the discovery of natural Diels–Alderases (DAases),^2–4 which was paralleled by the development of artificial enzymes⁵ and biomimetic catalysts⁶ for DA reactions. Mechanistic insights have been proven essential for understanding biocatalytic DA reactions^2,3,7,8 and motivating the design of artificial DAases.⁵ However, the studies of DAases are often difficult because of enzyme multifunctionality,⁹ the multiplicity of mechanistic scenarios,¹⁰ and the occurrence of formal DA reactions that are stepwise in nature^11,12 or display ambimodal reactivity.^13–16 Because of these challenges, monofunctional DAases have received much interest. SpnF,^14,15 the first monofunctional DAase discovered in nature,¹⁶ is also the most studied DAase. In 2015, two new monofunctional DAases, known as PyrE3 and PyrI4, were identified in the biosynthetic gene cluster of pyrroindomycins A and B.¹⁷ The products belong to the bacterial spirotetronate/spirotatramate family and display potent antimicrobial activities for MRSA and vancomycin-resistant bacteria .¹⁸ PyrE3 was found to catalyze the first step of a [4 + 2] cycloaddition cascade, converting a linear polyene precursor SM into a multi-substituted trans-decalin system P-1 in a highly endo-selective fashion (Scheme 1a). It is therefore one of many enzymes that catalyze decalin formation by intramolecular DA reactions.^2,3d,19 Another DAase, PyrI4, then catalyzes the second cycloaddition to construct the spirotetramate core.¹⁷

Scheme 1.

DAases, PyrE3 and PyrI4, in Pyrroindomycin Biosynthesis

The mechanism of catalysis by PyrI4 was recently studied by our group.²⁰ However, PyrE3 is a structurally distinct FAD-dependent DAase, and a different mechanism is possible because the two enzymes exhibit contrasting active-site electrostatic environments^21,28 and catalyze different types of DA reactions (Scheme 1b). This apparent contradiction presents a puzzle for the biosynthetic mechanism of pyrroindomycin. Although the structure of PyrE3 was disclosed by X-ray crystal analysis,²¹ it remains unclear how PyrE3 performs its function with a structure significantly different from PyrI4. Here, we present quantum mechanical and molecular dynamics (MD) explorations for the mechanism of the stereoselective catalysis of PyrE3. Unlike the case of PyrI4, in which acid catalysis and induced-fit/conformational selection comprise the main catalytic effects,²⁰ our computations show that PyrE3, with a rigid enzyme scaffold, utilizes electrostatic effects, H-bonds, and shape complementarity to stereoselectively preorganize the deprotonated linear substrate into the endo reactive conformation (RC). These factors allow an inverse-electron-demand DA reaction to be facile at room temperature.

2. RESULTS AND DISCUSSION

2.1. DFT Studies for the Uncatalyzed Reaction.

To reveal the intrinsic DA reactivity of the polyene substrate (SM) and provide a basis for understanding PyrE3 catalysis, we first studied the uncatalyzed DA reaction using density functional theory (DFT).²² We employed the M06–2X method,²³ and solvent effects of water were accounted for with the SMD implicit model²⁴ (see Figure 1 for the full level of theory and Supporting Information for more details). As shown in Scheme 2, the DA reaction of SM can give four stereoisomers (P-1 through P-4) depending on the relative positions and orientations of the diene with the dienophile. To study the mechanism of catalysis, we concentrate on the endo-1 stereochemical pathway that leads to P-1. The stereoselectivity will be discussed later in the paper.

Figure 1.

Free-energy profile and transition state geometries for the uncatalyzed intramolecular DA reaction at the endo-1 stereochemical relationship.

Scheme 2.

Possible DA Adducts

The substrate features a 3-acyltetramic acid structure as a part of the dienophile fragment. DFT calculations predict a pK_a of 1.1 for SMn (the neutral form of SM, see Supporting Information for computations), which agrees with reported acidities of 3-acyltetramic acids (2.3–3.5).²⁵ Therefore, the substrate is deprotonated and exists in the form of a tetramate anion SMa at a physiological pH. As shown in Figure 1, the formation of Pa-1 from SMa is exergonic by −23.4 kcal/mol. The barrier is 27.7 kcal/mol in the anionic manifold via TSa-1. TSa-1 is an asynchronous transition state; the shorter C2–C6 bond (2.0 Å) is much shorter than the longer partial bond C1–C3 (2.6 Å, see Scheme 1a for C atom numbering). In view of the high barrier of the anionic DA reaction, we also explored the effect of protonation. The cycloaddition barrier for the neutral pathway is only 21.1 kcal/mol (bidirectional blue arrow in Figure 1), 7 kcal/mol below that of the anion. However, the cost of protonation at pH 7.0 raises the overall barrier to 29.1 kcal/mol for TSn-1 in aqueous solution. The high barriers of both mechanisms explain the inertness of SM toward cycloaddition.¹⁷

2.2. Mechanism of Catalysis.

We next studied the catalytic mechanism of PyrE3. Although experimental efforts failed to co-crystallize PyrE3 with the substrate/product, the active site was identified by site-specific mutations to be a cavity between the middle domain and the FAD-binding domain (Figure 2a).²¹ How the substrate binds to the catalytic site is yet unknown. To study substrate binding, we docked RCa-1, the RC before TSa-1 (and thus it resembles TSa-1), into the crystal structure of apo PyrE3.²¹ The docked structure of RCa-1 was relaxed by 500 ns MD simulations to allow for induced-fit motions and to later evaluate the stability of the complex. Protein–ligand docking and MD simulations employed the AutoDock Vina and AMBER 16 programs,²⁶ respectively. The protonation states were assigned by PROPKA,²⁷ while the side-chain proton on H77 was manually moved from δ-N (i.e., HID) to ε-N (i.e., HIE) to facilitate hydrogen bonding. The predicted binding mode is presented in Figure 2b,c. At this pose, RCa-1 can bind PyrE3 with abundant hydrophobic contacts and hydrogen bonds (see Figure S4 for distance vs time plots). These binding forces, together with the shape complementarity between RCa-1 and PyrE3,²¹ cause a low average rmsd of 0.833 Å for the entire substrate and 0.185 Å for C1 to C6 (i.e., the reaction center) in the 500 ns MD trajectory, indicating the stability of RCa-1 in the bound state. In contrast, the other three stereoisomeric RCs RCa-2 to RCa-4 show poorer fits to the pocket, which will be discussed later.

Figure 2.

Structure of PyrE3. (a) Apo PyrE3. (b) Substrate–enzyme complex. (c) Substrate binding at endo RC.

2.2.1. Substrate Preorganization.

The folding of the substrate can significantly reduce the activation barrier.^{15e,20,29–31} In fact, DFT calculations predict that the folding of SMa into RCa-1 requires a free energy of 7.9 kcal/mol (8.9 kcal/mol by enthalpy) in solution, whereas PyrE3 overcomes the folding energy by binding in the active site. The RC is enthalpically higher in energy due to the necessity to introduce a variety of gauche and sterically hindered conformers. In addition, the free substrate SMa can explore many unreactive conformers, as revealed by 500 ns MD simulations in solution (Figure 3). The flexible conformation of SMa implies that a decrease in entropy is required for it to reach TSa-1, in which the rigid trans-decalin ring system is partially formed. This mixture of conformations will lead to a mixed entropy³² that increases the barrier by ca. 2.1 kcal/mol at room temperature (see Supporting Information). PyrE3 can alleviate this negative entropic effect by preorganizing the substrate into one specific pose.³³

Figure 3.

Change of substrate conformation by the PyrE3 active site. (a) Conformation clusters of the solvated substrate. (b) Conformation ensemble in the active site.

Since substrate prearrangement plays a critical role in the catalysis of PyrE3, it becomes important to understand the mode of substrate recognition by the active site. To this end, we carried out MM/GBSA binding energy calculations on the basis of MD simulations. The method combines molecular mechanics energies with generalized Born/surface area continuum solvation, and per-residue contributions can be identified to shed light on their individual roles.³⁴ The MM/GBSA calculations show that the strongest electrostatic attraction results from three arginine residues around the pocket: R211 (−16.7, all in kcal/mol, see Figure 2c for the residues), R350 (−13.0), and R42 (−10.9). Neutral residues H74 (−6.0), H77 (−4.9), L283 (−4.8), and G284 (−3.9) also provide notable electrostatic attractions. Among them, H74, H77, and G284 form H-bonds to anchor and orient the 3-acyltetramate unit. These results explain the increased K_m for the R211A (95), H74A (223), H77A (80), and H74A/H77A (477) mutants (52 for wild-type PyrE3).²¹ Because of the large non-polar interface, 12 residues display strong van der Waals (vdW) interactions with the substrate (larger than −1.0 kcal/mol). Moreover, the FAD co-factor not only acts as a structural motif to maintain the protein backbone and the shape of the active site but also provides non-covalent interactions with the bound substrate to enhance binding. The involvement of FAD and multiple residues was noted earlier in the X-ray crystallographic analysis of the PyrE3 active site.²¹

To reinforce substrate prearrangement, PyrE3 also possesses a rigid scaffold that can maintain the binding interactions and the shape of the pocket. The backbone of PyrE3 is largely unmoved before and after binding (Figure S3). Also, the substrate–enzyme complex of RCa-1 shows a low average rmsd of 1.63 Å (1.25 Å for the backbone atoms). These are contrasted against PyrI4, for which the entrance of the substrate triggers motions of the lid-like α0-helix to enclose the pocket,^20,28 which reflects higher scaffold flexibility.

2.2.2. Enzymatic DA Reaction.

To gain more insights into the enzymatic DA reaction, we constructed a theozyme to carry out quantum mechanical studies at the level of SMD-M06–2X/6–311++G(d,p)//M06–2X/6–31G(d).³⁵ As shown in Figure 4a, the theozyme includes the side chain of H74, the (simplified) flavin group of FAD, and the truncated P282-L283-G284-G285 and G44-A45-I46 loops. All truncation sites were capped with H atom(s). The model construction can reflect the main polar interactions in the pocket. Thermal fluctuations of the enzyme backbone were considered by evenly taking 10 snapshots from a 100 ns MD trajectory as initial geometries, giving rise to an ensemble of 10 theozymes (denoted by T1 to T10, see Figure 4b). Selected atoms were constrained to maintain the positions of the enzyme fragments and better reflect the rigid active pocket (marked with asterisks in Figure 4a). The remaining background of the enzyme is represented by a continuum model of diethyl ether.

Figure 4.

Theozyme model for the PyrE3 active site. (a) Illustration of the theozyme. Constrained atoms are marked with an asterisk. (b) Schematic description for the ensemble of 10 theozymes.

On account of the strong acidity of SM and the positive ESP (Scheme 1b) of the pocket, we propose that the anionic mechanism can be operative with PyrE3.²¹ This is supported by the agreement between experimental kinetics data (k_cat = 223.2 ± 5.4 min^−1,17 implying a 17.0 kcal/mol barrier) and the computed barrier (19.0 ± 1.4 kcal/mol, see Table S2 for details). The agreement shows again that PyrE3 differs from PyrI4, as acid catalysis plays an essential role in the latter case.²⁰ The hypothesis thus provides a possible explanation for the distinct electrostatic environments of PyrE3 and PyrI4.

To estimate the catalytic effect of the active site, the barrier at the same pose without the theozyme was computed to be 22.5 kcal/mol, which increased by 3.5 kcal/mol. The theozyme can provide the necessary auxiliary to support shorter C1–C3 and C2–C6 distances in the reactant state (from 4.1 and 3.3 Å in relaxed RC to 3.6 ± 0.3 and 3.2 ± 0.1 Å in the theozyme, respectively). We reasoned that the active site of PyrE3 can result in an optimized diene–dienophile alignment and thereby promote the DA reaction,³¹ which further emphasizes the importance of substrate prearrangement and destabilization. On the other hand, although the dienophile moiety also forms H-bonds with the H74 side chain and (the N–H unit of) the L283-G284 peptide bond, the theozyme (and likely the PyrE3 enzyme) does not involve H-bond activation of the dienophile.²¹ In fact, removal of H74 or the L283-G284 peptide bond causes minor changes of 0.0 ± 0.3 and −0.2 ± 0.7 kcal/mol, respectively, to the reaction barrier. The absence of H-bond activation is ascribed to the inverse-electron-demand nature of the anionic DA pathway due to the elevated energy of π-HOMO (dienophile) (Figure 5). The electron-rich dienophile is not likely to accumulate electron density in the transition state, hence no enhancement of H-bonds with the enzyme.

Figure 5.

Frontier orbitals for diene and dienophile fragments.

2.2.3. Discussion.

To summarize, the main mode of catalysis by PyrE3 is the prearrangement of a linear anionic substrate into a near-attack RC with an ideal diene–dienophile alignment. The associated reactant destabilization effect renders an inherently inert inverse-electron-demand DA reaction facile under ambient conditions. The insights explain why most reported mutations undermine enzyme activity by increasing K_m rather than decreasing k_cat.²¹ These mutations (e.g., A45L, F70Q, and I213F) might only result in a local shape perturbation that affects substrate binding to some extent. However, when a mutation (e.g., H74A) removes a H-bond with the anionic 3-acyltetramte unit, a significant damage to both substrate binding and diene–dienophile alignment would cause simultaneous changes in K_m and k_cat (e.g., a 329% increase and a 15% decrease, respectively, for H74A).

2.3. Origin of Stereoselectivity.

With the plausible catalytic mechanism established, we explored the origin of stereoselectivity of PyrE3, which forms the endo-1 trans-decalin product P-1 with a high degree of stereocontrol.¹⁷ We remind the reader that four possible stereoisomeric DA adducts, namely P-1 through P-4, can be generated from the corresponding TSs TS-1 through TS-4 (1: endo-1, 2: exo-1, 3: endo-2, and 4: exo-2, see Scheme 2). To probe the intrinsic stereochemical preference, we carried out DFT computations for the uncatalyzed reaction (Figure 6). As presented in Figure 6a, We considered both protonation states because of the uncertainty of pK_a calculations. The results suggest that the endo-1 transition states TSa-1 and TSn-1 (27.8 and 29.1 kcal/mol) are not the most favorable in both mechanisms. Instead, TSa-2 (26.0 kcal/mol) and TSn-2/TSn-3 (27.9 and 27.8 kcal/mol) display the lowest barriers, contradicting the stereochemical outcome in pyrroindomycin biosynthesis. These results rely on the lowest-energy conformers in the absence of PyrE3. Since PyrE3 can alter substrate conformation, we also calculated ΔΔG^‡ between endo-1 and exo-1 adopting the pose in the bound state (i.e., as shown in Figure 2c) with the enzyme implicitly represented by a continuum model of diethyl ether (Figure 6b). The resulting ΔΔG^‡ (0.6 kcal/mol), despite favoring endo-1 over exo-1, is inadequate for explaining the exclusive formation of Pa-1. In sum, these calculations show that Pa-1 is the only product of biosynthesis¹⁷ because the enzyme stabilizes the endo-1 pathway. This feature is explored in the following two subsections.

Figure 6.

DFT studies on stereoselectivity. (a) Free-energy profile for the uncatalyzed reaction at different stereochemical relationships. (b) Transition states for the endo-1 and exo-1 pathways at the prearranged pose.

2.3.1. Binding Affinities.

In consideration of the favorable binding interactions and shape complementarity between RCa-1 and the PyrE3 active site, we envisaged that RCa-1 (and thus the structurally similar TSa-1) may exhibit a stronger binding affinity than RCa-2 to RCa-4, thus accounting for the stereopure formation of Pa-1. To assess the total binding affinities of these stereoisomeric RCs, we computed the free-energy cost to reach the RCs (i.e., from SMa to the best docked structures) and the MM/GBSA binding affinities at the resultant RC. Despite overly large numbers (referenced vs gas phase), the MM/GBSA affinities are good indicators for relative affinities.

The results shown in Figure 7a suggest that RCa-1 is in an ideal binding mode with the lowest folding energy (7.9 kcal/mol from SMa to RCa-1) and the strongest binding affinity with the pocket (−55.0 kcal/mol at RCa-1), giving rise to a total affinity of −47.1 kcal/mol. The other three stereoisomeric RCs are docked into somewhat similar poses at which the anionic 3-acyltetramte moiety is positioned downward in the figures (i.e., interfacing with the FAD-binding domain in Figure 2a). The resulting poses of RCa-2 to RCa-4 are substantially less favored, with the folding free energies exceeding 11 kcal/mol and the affinities ranging from −46.1 to −49.5 kcal/mol. Consequently, the total binding affinities are −33.9 to −35.1 kcal/mol, which are notably lower than those of RCa-1. Therefore, the conformation selection of PyrE3 is strongly stereoselective such that the generations of side products Pa-2 to Pa-4 are disfavored. Besides the complementarity of RCa-1, we speculate that the key His and Arg residues on the upper side of the pocket (H74, H77, R211, R350, and R254) can also contribute to the stereo-differentiation. As depicted in Figure 7b and detailed in Table S4, a comparison of MM/GBSA per-residue binding energies discloses that these residues provide stronger electrostatic attractions with RCa-1 than with RCa-2 through RCa-4. This reflects the substrate-His H-bonds and the enhanced Coulomb attractions when the anionic 3-acyltetramate is positioned toward the upper side of the pocket where R211, R350, and R254 exist. These factors can carry over to the respective transition states (which are similar in structure to the RCs), whose energies ultimately determine the stereoselectivity of the catalyzed reaction.

Figure 7.

Assessment of binding affinities. (a) Binding affinities for stereoisomeric RCs RCa-1 through RCa-4. (b) Average differences in MM/GBSA per-residue binding affinities (RCa-1 minus RCa-2-RCa-4).

2.3.2. Stereocontrolling Active Site Environment.

The previous discussion established the energetic differentiation in the favorable binding of RCa-1. Since the transition states of endo-1 and exo-1 share congruent arrangements except for the diene unit (Figure 8a), we envisioned that some feature of the binding site near the reaction center would exert extra control over the orientation of the diene unit. In fact, MD simulations show that the side chain of A45 is complementary to the endo-diene unit of RCa-1 (Figure 8b,c). The corresponding exo-diene, however, will overlay onto A45 and is thus disfavored. This precisely placed steric hindrance must have also enhanced the stability of the cis-1,3-diene, which would otherwise easily turn into the trans conformer (Figure 3a). With details provided in the Supporting Information, we note that the constraint by A45 is stabilized by an FAD-mediated HB network (HB-5 through HB-8 in Figure 2c) as well as favorable vdW interactions.

Figure 8.

Stereocontrolling active site environment. (a) Overlaid geometry of the endo-1 and exo-1 transition states in Figure 6b. (b,c) Stereocontrolling constraint exerted on the 1,3-diene unit.

3. CONCLUSIONS

Pyrroindomycin biosynthesis involves two monofunctional DAases, PyrE3 and PyrI4, that stereoselectively construct the spirotetramate ring system via a cycloaddition cascade. Unlike the acid catalysis by PyrI4, PyrE3 uses a positively charged pocket to promote an inverse-electron-demand DA reaction of the anion with substrate preorganization. The pocket strongly favors an endo RC and maintains the stereochemical information by imposing an exo-disfavoring constraint. The catalysis relies on a synergy of specific binding interactions, electrostatic background, shape complementarity, and scaffold rigidity. Strikingly, the mechanisms of the two DA pathways in pyrroindomycin biosynthesis involve significantly different modes of catalysis. This highlights how nature has evolved multiple ways to catalyze DA reactions.