Synthetic microbial sensing and biosynthesis of amaryllidaceae alkaloids

Abstract

A major challenge to achieving industry-scale biomanufacturing of therapeutic alkaloids is the slow process of biocatalyst engineering. Amaryllidaceae alkaloids, such as the Alzheimer's medication galantamine, are complex plant secondary metabolites with recognized therapeutic value. Due to their difficult synthesis they are regularly sourced by extraction and purification from low-yielding plants, including the wild daffodil Narcissus pseudonarcissus. Engineered biocatalytic methods have the potential to stabilize the supply chain of amaryllidaceae alkaloids. Here, we propose a highly efficient biosensor-AI technology stack for biocatalyst development, which we apply to engineer amaryllidaceae alkaloid production in Escherichia coli . Directed evolution is used to develop a highly sensitive (EC50= 20 uM) and specific biosensor for the key amaryllidaceae alkaloid branchpoint 4-O'Methylnorbelladine. A machine learning model (MutComputeX) was subsequently developed and used to generate activity-enriched variants of a plant methyltransferase, which were rapidly screened with the biosensor. Functional enzyme variants were identified that yielded a 60% improvement in product titer, 17-fold reduced remnant substrate, and 3-fold lower off-product regioisomer formation.

Snowprint: a predictive tool for genetic biosensor discovery

“…In the future, pairing this discovery workflow with directed evolution to refine the specificities of biosensors identified by screening [25][26][27] provides a new paradigm research and development teams can adopt to leverage biosensor-enabled screens for a wider diversity of small molecules. In particular, the biosensors already identified herein can be used as relevant evolutionary starting points for important pharmaceutical biomanufacturing efforts, as olivetolic acid is a precursor to all cannabinoids 32 , geraniol is a precursor to all monoterpene indole alkaloids 41 , tetrahydropapaverine is a precursor to three licensed nondepolarizing muscle relaxants 25 , and ursodiol is directly used in the clinic to dissolve gallstones and treat liver diseases 34 .…”

“…To demonstrate utility for synthetic biology applications, we used Snowprint to extract and "domesticate" a panel of transcription factors. We targeted TetR-family transcription factors that regulate multidrug efflux pumps in particular, since they are likely to promiscuously bind to a wide range of structurally diverse ligands and may serve as excellent starting points for directed evolution of effector specificity [25][26][27] . To generate designs, TetR-family regulators were downloaded from the UniRef50 database, clustered into 30% identity groups using CD-HIT, and filtered for regulators with sequence lengths between 140-260 amino acids, typical for the TetR-family 28 (Supplementary Figure 3).…”

Asymmetric Synthesis of α-Chloroamides via Photoenzymatic Hydroalkylation of Olefins

“…With the goal of identifying mutations distal to the protein active site, we used MutComputeX, a 3D self-supervised machine learning (ML) framework trained to identify residues where the wild-type amino acid is chemically incongruent with its local environment, for ML-guided rational design of surface mutations based on the GluER-T36A crystal structure. − We anticipated that these mutations could enhance the stability or catalytic performance of the protein. Among the residues predicted, we found that mutating alanine (originally threonine) at position 36 to aspartic acid (T36E) and phenylalanine at position 68 to tyrosine (F68Y) led to variants that provided product in comparable yields and enantioselectivities but with decreased formation of the hydrodehalogenated product 4 .…”

“…Among the residues predicted, we found that mutating alanine (originally threonine) at position 36 to aspartic acid (T36E) and phenylalanine at position 68 to tyrosine (F68Y) led to variants that provided product in comparable yields and enantioselectivities but with decreased formation of the hydrodehalogenated product 4 . Contrary to our initial hypothesis for these ML mutations, they did not thermally stabilize the protein, as has been previously observed (Figure S13), but rather served to decrease formation of the hydrodehalogenated product. − This observation is aligned with the learning objective of self-supervised protein ML frameworks: they are trained to predict the extant wild-type amino acids of natural proteins, which have been evolutionarily selected to optimize overall fitness rather than a particular phenotype. Thus, residues where the wild-type amino acid is poorly predicted are primed for identifying gain-of-function mutations, and it is up to the experimentalist to screen these ML predictions and select for the phenotype to optimize, such as decreased shunt production formation.…”