Ancestral Sequence Reconstruction to Enable Biocatalytic Synthesis of Azaphilones

Abstract

Biocatalysis can be powerful in organic synthesis but is often limited by enzymes' substrate scope and selectivity. Developing a biocatalytic step involves identifying an initial enzyme for the target reaction followed by optimization through rational design, directed evolution, or both. These steps are time consuming, resource-intensive, and require expertise beyond typical organic chemistry. Thus, an effective strategy for streamlining the process from enzyme identification to implementation is essential to expanding biocatalysis. Here, we present a strategy combining bioinformatics-guided enzyme mining and ancestral sequence reconstruction (ASR) to resurrect enzymes for biocatalytic synthesis. Specifically, we achieve an enantioselective synthesis of azaphilone natural products using two ancestral enzymes: a flavin-dependent monooxygenase (FDMO) for stereodivergent oxidative dearomatization and a substrate-selective acyltransferase (AT) for the acylation of the enzymatically installed hydroxyl group. This cascade, stereocomplementary to established chemoenzymatic routes, expands access to enantiomeric linear tricyclic azaphilones. By leveraging the co-occurrence and coevolution of FDMO and AT in azaphilone biosynthetic pathways, we identified an AT candidate, CazE, and addressed its low solubility and stability through ASR, obtaining a more soluble, stable, promiscuous, and reactive ancestral AT (AncAT). Sequence analysis revealed AncAT as a chimeric composition of its descendants with enhanced reactivity likely due to ancestral promiscuity. Flexible receptor docking and molecular dynamics simulations showed that the most reactive AncAT promotes a reactive geometry between substrates. We anticipate that our bioinformatics-guided, ASR-based approach can be broadly applied in target-oriented synthesis, reducing the time required to develop biocatalytic steps and efficiently access superior biocatalysts.

Figure 1.

Common strategies for biocatalyst development and the targeted biocatalytic synthesis in the study expected to diversify the azaphilone chemical space. (A) Strategies include screening wild-type enzymes, directed evolution, and using computational tools such as consensus sequence, ProteinMPNN, enzyme design, and bioinformatics-guided ancestral sequence reconstruction. (B) Stereocomplementary chemoenzymatic synthesis of linear tricyclic azaphilones enabled by the identification of a suitable flavin-dependent monooxygenase and acyltransferase. (C) Based on the versatility of groups R¹, R², and R³ on the azaphilones, stereocomplementarity doubles the number of combinations and thus chemical diversity.

Figure 2.

Acyltransferase mining based on the co-localization and the upstream-downstream relationship of flavin-dependent monooxygenase (FDMO) and acyltransferase (AT) in azaphilone biosynthetic gene clusters (BGCs). (A) Well-studied azaphilone related BGCs. (B) Enzymatic cascade featuring a sequential relationship between FDMO and AT. (C) Agreement between the FDMO and AT clusterings ex-emplified by three FDMO-AT clustering pairs and quantitated by the adjusted mutual information (AMI) score. (D) Workflow for AT mining starting from the BLAST search of AfoD and subsequent identification of ATs co-occurring with an adjacent homolog of AfoD. The following mining workflow diverges into an ancestral sequence reconstruction (ASR) based approach and a protein engineering-based approach using a wild-type enzyme as a template.

Figure 3.

Bioinformatics-guided sampling of ancestral sequence space of CazE and resurrection of ancestral acyltransferases (ATs). (A) Reported CazL/CazE (FDMO/AT) sequence in the biosynthetic gene cluster of chaetoviridin A. (B) Targeted two-enzyme sequence to access a linear tricyclic azaphilone. (C) Unrooted phylogenetic tree comprising 240 extant ATs reconstructed using the LG maximum likelihood model. The CazE clade was zoomed in, and all the ATs experimentally sampled were highlighted in cyan for ancestors and green for extant enzymes. (D) Preliminary study of resurrected ATs. AT-expressing E. coli BL21(DE3) pellets were lysed, and the soluble expression level of ATs was evaluated using SDS-PAGE. The preliminary screening of ATs’ reactivities was performed following the conditions: 2.5 mM substrate 13, 10 μM flavin-dependent monooxygenase (FDMO, AncFDMO), 1 mM NADP⁺, 5 mM glucose-6-phosphate (G6P), 2 U·mL⁻¹ glucose-6-phosphate dehydrogenase (G6PDH), 50 mM potassium phosphate buffer, pH 8.0, 30 °C, 1 h. 1.1 equiv of thioester 14 was added into 50 μL of the reaction mixture, mixed with 50 μL of the freshly-lysed clarified cell lysate, and incubated at 30 °C for another 1 h. (E) Melting temperatures of two extant and three ancestral ATs determined by differential scanning fluorimetry. (F) AlphaFold2 model of AncAT4 with an incorporated phosphopantetheine C5-thioester. The active site of AT4 was defined using the geometric center of His166 and colored based on the identity to AncAT5 (light green) and AT6 (green).

Figure 4.

Substrate scope of the acyltransferases (ATs) to access linear tricyclic azaphilones. (A) Reaction conditions for the first step from 3 to 4: 2.5 mM substrate 3, 10 μM flavin-dependent monooxygenase (FDMO, AzaH or AncFDMO), 1mM NADP⁺, 5mM glucose-6-phosphate (G6P), 2 U·mL⁻¹ glucose-6-phosphate dehydrogenase (G6PDH), 50 mM potassium phosphate buffer, pH 8.0, 30 °C, 1 h; conditions for the second step from 4 to 6: 1.1 equiv of thioester 5 and 250 μM AT (2 mol %) were added to the reaction mixture from the step 1. Conversion was calculated by the consumption of the substrate 4 and only reported when the product 6 was formed and detected by UPLC-UV or UPLC-MS. (B) Activity of ATs between (S)-12 and (R)-12. (C) Activity of AncAT4 reacting with various thioester chain lengths in the range from C1 to C9. (D) Substrates with modifications on R¹.

Figure 5.

Substrate docking and molecular dynamics (MD) simulations. (A) Near attack complex of the substrate (S)-12 in the active site of AncAT4. Near attack complex was selected by finding smallest (S)-12 and C5-thioester carbonyl carbon distance with a Bürgi-Dunitz angle within 90° and 107.5° from 40 ns simulation. (B) Substrate OH and C5-thioester carbonyl carbon distance of top cluster poses from rigid receptor docking (Rigid CDOCKER), flexible receptor docking (Flexible CDOCKER), and Flexible CDOCKER with a flexible C5-thioester group. (C) Joint plot of the OH-carbonyl carbon distance of (S)-12 with the Bürgi-Dunitz angle obtained from 40 ns simulation. The black dashed lines indicate an OH-carbonyl carbon distance of less than 6 Å and Bürgi-Dunitz angle of greater than 90°. (D) Plot of the histidine-OH distance obtained from 40 ns simulation.