An enzymatic Alder-ene reaction

Abstract

An ongoing challenge in chemical research is to design catalysts that select the outcomes of the reactions of complex molecules. Chemists rely on organocatalysts or transition metal catalysts to control stereoselectivity, regioselectivity and periselectivity (selectivity among possible pericyclic reactions). Nature achieves these types of selectivity with a variety of enzymes such as the recently discovered pericyclases-a family of enzymes that catalyse pericyclic reactions¹. Most characterized enzymatic pericyclic reactions have been cycloadditions, and it has been difficult to rationalize how the observed selectivities are achieved^2-13. Here we report the discovery of two homologous groups of pericyclases that catalyse distinct reactions: one group catalyses an Alder-ene reaction that was, to our knowledge, previously unknown in biology; the second catalyses a stereoselective hetero-Diels-Alder reaction. Guided by computational studies, we have rationalized the observed differences in reactivities and designed mutant enzymes that reverse periselectivities from Alder-ene to hetero-Diels-Alder and vice versa. A combination of in vitro biochemical characterizations, computational studies, enzyme co-crystal structures, and mutational studies illustrate how high regioselectivity and periselectivity are achieved in nearly identical active sites.

Extended Data Figure 1.. Density functional theory calculations for non-enzymatic Alder-ene and hetero-Diels–Alder reactions from (Z)-QM and (E)-QM and PdxI theozyme.

a, Transition states, products, and energies for eight hetero-Diels–Alder and Alder-ene reactions are shown. In the transition states, the Alder-ene reactions adopt a conformation where the pyridone and forming cyclohexane are perpendicular to each other compared to the hetero-Diels–Alder reactions that are more co-planar in geometry. The Alder-ene reactions are synchronous and the hetero-Diels–Alder reactions are asynchronous, but concerted. TS-1 and TS-2 lead to 8’ and 9 with barriers of 25.1 and 23.2 kcal·mol⁻¹, respectively. The structures 8a’, b’, c’, 9b, c, and 11 are isomers of natural product scaffolds with barriers greater than TS-1. b, Alder-ene theozyme of (Z)-quinone methide complex leading to Alder-ene adduct (8’), O4- and O2-hetero-Diels–Alder adducts (9 and 11) with energies reported as enthalpies and Gibbs free energies.

Extended Data Figure 2.. Homologous biosynthetic gene clusters of pyridoxatin (1) and fusaricide (3) and the functions of PdxI and EpiI homologs.

a, Putative biosynthetic gene clusters of 1 and 3 (and 2), and their homologous biosynthetic gene clusters found in NCBI database. b, Key active site residues shown in an alignment with those from PdxI and EpiI homologs. Key residues involved in PdxI and EpiI catalysis are colored. c, In vitro analysis of PdxG and selected pericyclases using 5 as the starting substrate. The periselectivity can be correlated with the identity of the amino acid at position 413 (in PdxI, indicated in red dashed box). If valine occupies this position, the enzyme catalyzes the Alder-ene reaction. On the other hand, if methionine occupies the position, the enzyme catalyzes the hetero-Diels–Alder reaction.

Extended Data Figure 3.. Biochemical characterization of the ketoreductase PdxG.

a, In vitro reaction of 3 μM PdxG with 2 mM NADPH using 600 μM 5 as the substrate. 60% conversion from 5 to 6 was observed within 20 min. b, Kinetic analysis of PdxG-catalyzed reduction of 5. Reaction mixtures containing 3 μM PdxG, 2 mM NADPH and different concentrations of 5 (10 μM to 1.2 mM) were incubated at 30 °C for 20 min. Error bars indicate s.d. of three independent replicates. c, Formation of 10 from 6 in the presence of 2 mM NADPH can be observed both with and without PdxG. Compound 6 was obtained from chemical reduction of 5 with NaBH₄. Since 10 can be formed in the presence of only NADPH, we conclude NADPH can nonenzymatically reduce the QM to 10, which accounts for the result in Fig. 1f.

Extended Data Figure 4.. Biochemical characterization of PdxI and EpiI.

a, LC/MS analyses of chemically denatured PdxI and EpiI show no trace of SAM after purification. b, HPLC analyses of in vitro reaction of 150 μM 5 with 3 μM PdxG, 1 mM NADPH and 30 μM PdxI or 20 μM EpiI at 30 °C for 2 h in the presence or absence of cofactors. SAM or SAH does not alter the enzymatic activity of PdxI and EpiI. c, Since the interconversion of 8 and 9 could be envisioned by hydroalkoxylation/retro-hydroalkoxylation, we examined the possibility that PdxI and EpiI could catalyze the reaction of 9 to 8 and 8 to 9, respectively. The in vitro reactions of 100 μM 9 or 8 with 50 μM PdxI or 40 μM EpiI at 30 °C for 24 h were performed. However, no conversion of 9 to 8 and 8 to 9 was observed. d, In vitro reaction of PdxI and EpiI using 6 as the substrate. To obtain 6 for in vitro reaction, we chemically reduced 5 by NaBH₄. Since this reduction proceeds non-stereoselectively, 6 and diastereomer 6′ were formed. After isolation of 6 and 6′ by HPLC, fractions containing 6 were not concentrated because of the instability and were immediately used as the substrate for PdxI and EpiI. e, LC/MS analysis of in vitro reactions catalyzed by pericyclases using 6 as the substrate. Shown are compounds detected by selected ion monitoring at (M+H)⁺ of 248. In this mode, 6 is detected as the fragment ion. In the absence of either enzyme, 6 was converted to several products nonenzymatically, including 11. Minor compounds not isolated are indicated with *. In the presence of PdxI, 6 was nearly all converted to 8. In the presence of EpiI, 6 was nearly all converted to 9. Mutation of T232S in PdxI or T231A in EpiI changed selectivities of the enzymes to give other products, including 11. For additional mutagenesis data, please see Extended Data Figs. 7-8.

Extended Data Figure 5.. Overlays of crystal structures with transition state structures for Alder-ene and hetero-Diels–Alder reactions.

a, Overlay of Alder-ene TS-3 with 5 bound in PdxI. Note the extended conformation of the alkyl chain versus the folded transition state geometry. The pyridone is bound by hydrogen bonds from K337, H161, Q412, and water mediated hydrogen bonds from T232, D233 and H336. b, Overlay O4-hetero-Diels–Alder TS-6 with 5 bound in HpiI. The pyridone is bound by hydrogen bonds from H161, Q414, and water mediated hydrogen bonds from T232, D233 and H338. Note that the K339 hydrogen bond to the pyridone O4 is not present in this structure. c, Overlay of PdxI-5 and HpiI-5. Omit maps not shown for clarity. Both PdxI and HpiI bind the pyridone such that it is prone to a syn-dehydration assisted by K337 (PdxI) or water molecule W (HpiI) and water molecules surrounding C7. The inset shows how V413 (in PdxI) or M415 (in HpiI) affects the orientation of the lysine residue (K337 in PdxI or K339 in HpiI) and its ability to hydrogen bond to the 4-OH of the pyridone. d, Overlay of PdxI-5, HpiI-5, Alder-ene TS-3, and O4-hetero-Diels–Alder TS-6, and O2-hetero-Diels–Alder TS-5. Omit maps not shown for clarity. TS-3 and TS-6 bind in the active site sans disfavorable interactions whereas TS-5 clashes with T232. As both the calculated Alder-ene transition structure TS-3 and hetero-Diels–Alder TS-6 are quite similar in geometry and both easily fit into the PdxI active site, PdxI cannot solely rely on shape complementarity to catalyze the reaction with observed periselectivity. e, Chain B active site of PdxI-product (8) complex. Note the closer distances between K337 and the pyridone O4, the change in coordination of water mediated hydrogen bond from H336, and H161 shifting from a N1 hydrogen bond (in PdxI-5) to an O2 hydrogen bond. f, Overlay of Alder-ene TS-3 with 8 bound in PdxI. Note high similarity in structures of 8 and TS-3. This suggests that the enzyme distorts the product structure towards that of the Alder-ene transition state.

Extended Data Figure 6.. Molecular dynamic simulation of 7 in the PdxI active site.

Distances over time of hydrogen bonds to the various positions of the pyridone are tracked in chain A (a) and chain B (b) of the active site. Left panels show H336 and K337 form hydrogen bonds to the 4-position substituent on the pyridone ring. Right panel shows Q412 and H161 remain hydrogen bonded to 2-position substituent and pyridone nitrogen N1, respectively, for the majority of the simulation. c, Molecular dynamic simulations were initiated from an extended conformation (dihedral = ~180°). Over time, we monitored this conformation to see if the alkyl chain could spontaneously fold to a reactive conformation (dihedral = −20°). Indeed, for short durations of the simulations we observe the chain folding into a reactive conformation for a pericyclic reaction.

Extended Data Figure 7.. HPLC analysis of in vitro reaction of PdxI and mutants.

Mutation of the catalytic base K337A abolished the activity, while K337R mutant could retain approximately 10% activity. Individual substitution of H336A, Q412A, and H161A all completely abolished the activity. Mutation of D233A or D233N completely abolish enzymatic activity. In contrast, in the D233E mutant 60% of activity and the original periselectivity were retained. This suggests that the carboxylate group of D233 in PdxI is important for enzyme function. A single mutation, V413M is sufficient to change the periselectivity from Alder-ene (>98:2, 8:9) to hetero-Diels–Alder reaction (40:60, 8:9). Further, mutation of T232 to either alanine or serine, but not valine, can lead to the formation of the O2-hetero-Diels–Alder product 11 along with the Alder-ene product 8. The data show one representative experiment from at least three independent replicates. Reaction conditions: 150 μM 5 with 3 μM PdxG, 1 mM NADPH and 50 μM PdxI (wild type or mutant) at 30 °C for 2 h.

Extended Data Figure 8.. HPLC analysis of in vitro reaction of EpiI and mutants.

In contrast to PdxI, substitution of K338 to alanine did not abolished and retained the activity (83%) (Extended Data Fig. 9b). H336A, H161A, and Q410A (corresponding to Q412 in PdxI) mutants were highly insoluble and cannot be assayed. Although D232A and D232N mutations completely abolished the enzymatic activity, the D233E mutation retained 53% of activity and maintained the original periselectivity. This suggests that the carboxylate group of D232 in EpiI is also important for enzyme function. Mutation of Y205F retained 89% activity and maintained the original periselectivity, suggesting the hydroxy group of Y205 is not essential for catalysis. The M411V (corresponding to V413 in PdxI) and M411C mutations increased the Alder-ene product ratio compared to the wild type of EpiI. The T232A and T232S mutations but not T232V mutation, generated the O2-hetero-Diels–Alder product 11 and the Alder-ene product 8 as the minor products, with the hetero-Diels–Alder product 9 as a major product. The double mutation M411V/T231A of EpiI reversed the periselectivity from the native hetero-Diels–Alder reaction (<5:95, 8:9) to the energetically disfavored Alder-ene reaction (2:1, 8:9), although the enzymatic activity is only moderately decreased (Extended Data Fig. 9). In the double mutant, 11 was formed due to the mutation of T231. Other double mutants such as M411V/T231S, M411T/T231A, M411C/T231A, and M411G/T231A also reversed periselectivity. The data shown are that of one representative experiment from at least three independent replicates. Reaction condition: 300 μM 5 with 3 μM PdxG, 1 mM NADPH and 40 μM EpiI (wild type or mutant) at 30 °C for 2 h.

Extended Data Figure 9.. Relative activities of PdxI, EpiI, and mutants.

The activity of each mutant is compared to that of wild-type PdxI or EpiI quantified by the formation of 8, 9 and 11. Error bars indicate s.d. of three independent replicates. Asterisks indicate mutants with no measurable activity. a, The relative enzymatic activity of PdxI mutants. Reaction conditions: 150 μM 5 with 3 μM PdxG, 1 mM NADPH and 50 μM PdxI mutants at 30 °C for 2 h. b, The relative activity of EpiI mutants. Reaction condition: 300 μM 5 with 3 μM PdxG, 1 mM NADPH and 40 μM EpiI mutants at 30 °C for 2 h.

Extended Data Figure 10.. Proposed mechanisms of PdxI- and EpiI-catalyzed reactions.

a, The catalytic cycle of PdxI-catalyzed reaction is initiated by the deprotonation of the 4-hydroxy group by K337 followed by the syn-dehydration to 7 assisted by the extend water hydrogen bonding network mediated by H336. Subsequently, protonated K337 serves as the general acid catalyst and forms the strong hydrogen bonding with 4-carbonyl oxygen of 7 to set the stage for the periselective Alder-ene reaction. Note that the steric effect of T232 inhibits the formation of the O2-hetero-Diels–Alder product 11 to further control regioselectivity. The alkyl chain folds to a reactive conformation and readily undergoes an Alder-ene reaction. After this, the tautomerization is facilitated by K337 and possibly water mediated by H336 to form and release 8. Then, the next catalytic cycle initiates. b, The catalytic cycle of EpiI-catalyzed reaction, in contrast to PdxI, is initiated by the deprotonation of the hydroxy group by an alternative general base, possibly water followed by the syn-dehydration to 7. Since the key lysine residue does not form hydrogen bonding with 4-carbonyl oxygen of 7 due to the bulkier side chain of M411 (corresponding to V413 in PdxI), the favored hetero-Diels–Alder reaction takes place to form and release 9. As same as PdxI, the steric effect of T231 inhibits the formation of the O2-hetero-Diels–Alder product 11 to further control regioselectivity. Then, the next catalytic cycle initiates.

Figure 1 ∣. Pericyclic reactions in natural product biosynthesis.

a, Known and unknown enzymatic examples of pericyclic reactions. b, The biosynthesis of leporin B involves a multifunctional O-methyltransferase-like pericyclase LepI. c, Theoretical investigations indicate that hetero-Diels–Alder TS-2 is nonenzymatically favored by 1.9 kcal·mol⁻¹ over Alder-ene TS-1 from a common intermediate 7. Further transformations lead to natural products pyridoxatin 1, cordypyridones, asperpyridone 2 and fusaricide 3. d, The biosynthetic gene cluster (lep) of leporin B from Aspergillus flavus, the putative biosynthetic gene cluster (adx) of pyridoxatin from Albophoma yamanashiensis, the putative biosynthetic gene cluster (pdx) of pyridoxatin from Aspergillus bombycis, and the putative biosynthetic gene cluster (epi) of fusaricide from Epicoccum sorghinum FT1062. PKS-NRPS, polyketide synthase-nonribosomal peptide synthetase; TF, transcription factor; MCT, monocarboxylate transporter; P450, cytochrome P450; SDR, short-chain dehydrogenase/reductase; ER, enoylreductase; PC, pericyclase. e, The proposed biosynthesis of the Alder-ene product (8) and the hetero-Diels–Alder product (9) from the common intermediate 7. f, One-pot in vitro tandem assay of 5 with PdxG and in the presence or absence of selected pericyclases.

Figure 2 ∣. Crystal structures of PdxI and HpiI.

Simulated annealing omit map shown in grey mesh and contoured at 1.0 σ. Hydrogen bond interactions are indicated with black dashed lines. a, Cartoon representation of apo-PdxI tertiary structure and binding-cavity (magenta). The C-terminal catalytic domain is shown in green and the N-terminal dimerization domain in lime. b, Overlay of interlocking homodimer structures of apo-PdxI and apo-HpiI. c, Active-site view of co-crystal structure of PdxI with substrate analogue ketone 5. d, Active-site view of co-crystal structure of HpiI with substrate analogue ketone 5. In c and d, M76 from Chain B is indicated in different colors.

Figure 3 ∣. Mechanism of periselective and regioselective pericyclic reactions.

a, Alder-ene theozyme model based on PdxI structure. See supplementary information for further details. b, O4-Hetero-Diels–Alder theozyme model confirmed by HpiI structure. c, Analysis of the relative production ratio of the O4-hetero-Diels–Alder adduct 9, Alder-ene adduct 8, and the O2-hetero-Diels–Alder adduct 11 from in vitro reaction of 5 with PdxG, NADPH, and selected pericyclases. To quantify the ratio, the reaction time of control sample without pericyclases (Cont.) was 12 hours, and reaction times with enzyme were 2 hours. Error bars indicate s.d. of three independent replicates. *Putative cyclized products other than 8, 9, and 11 were detected in the control reaction (Extended Data Fig. 4e).