Accessing Nature’s diversity through metabolic engineering and synthetic biology

Diversity through scaffold derivatization

Chemical transformations of complex molecules often suffer from a lack of regioselectivity and stereoselectivity, poor discrimination between functional groups of similar reactivity, and an incompatibility with biological media. Enzymes, however, catalyze site-specific and stereoselective chemistries in water—often within a microorganism. Numerous enzyme-mediated chemical functionalizations of natural products are known, including scaffold alkylation^14–16, acylation¹⁷, oxidation^18,19, glycosylation^4,20, and halogenation²¹. Here we focus the discussion of enzyme-tailored scaffold derivatization on the mature cases of natural product tailoring by cytochrome P450 oxidases (P450s or CYPs) and GTs. It is worth noting the biosynthetic potential of the lesser-utilized bio-acylation and bio-halogenation reactions for natural product derivatization, as these reactions can introduce orthogonally reactive handles for late-stage library differentiation²¹.

A robust derivatization strategy employs naturally promiscuous P450s that have been engineered to harness multiple natural and novel chemistries in vivo^22–24. For example, Keasling and co-workers used rational enzyme mutagenesis of a plant-mimicking bacterial enzyme (P450-BM3) capable of epoxidizing the plant-derived taxane amorpha-4,11-diene (Figure 2 [4]) to obtain a more thermostable and selective epoxidation catalyst. P450-BM3 mutant G3A328L enabled the biosynthesis of the value-added compound artemisinic-11S,12-epoxide at 250 mg/L in E. coli, thereby improving the semi-synthesis of the antimalarial drug artemisinin (Figure 2 [10])¹⁸. Recently, McLean et al. evolved CYP105AS1 from Amycolatopsis orientalis to hydroxylate the pravastatin (Figure 2 [17]) precursor compactin (Figure 2 [16]) in the engineered Penicillium chrysogenum strain D550662, ultimately achieving titers of 6 g/L of the blockbuster drug after 200 h in a 10 L fed-batch fermentation (Figure 2D)²⁵.

P450-catalyzed metabolite derivatization is likewise offering avenues to explore chemical space that was previously unavailable in a biological setting. Frances Arnold’s lab has developed an impressive array of P450 catalysts including an engineered diazoester-derived carbene transferase (P450_BM3/CYP102A1) for the stereoselective cyclopropanation of styrenes, which have concomitantly become available biologically via the metabolic engineering of E. coli for styrene production from L-phenylalanine at a titer of 260 mg/L^26,27. Arnold’s team expanded the work to enable incorporation of the cyclopropane in vivo by engineering the electronics of the enzyme active site to accommodate NAD(P)H as an electron donor, and upon altering the active site architecture, they further engineered the catalyst for cyclopropanation of N,N-diethyl-2-phenylacrylamide, a putative intermediate in the formal synthesis of the serotonin and norepinephrine reuptake inhibitor levomilnacipran—marketed by Actavis Inc. as Fetzima for the treatment of clinical depression^28–30. On the chemical front, Wallace and Balskus have developed porphyrin-iron(III) chloride catalysts that function similarly to Arnold’s P450-BM3 mutants while presenting biocompatible reactions with living styrene-producing E. coli³¹. Such approaches highlight the potential to meld chemical and biological approaches for tailored molecule derivatization in engineered organisms^32,33.

GTs are also attracting attention in the derivatization of natural products, including polyketides, non-ribosomal peptides, and terpenoids, for the discovery of novel antimicrobial agents with tailored pharmacological properties, including augmented target recognition and improved bio-availability^4,20,34,35. In this regard, dNDP-glycosides (Figure 1) represent a biosynthetically viable class of saccharide donors for promiscuous and engineered GTs that exhibit substrate tolerances for both the saccharide and aglycone portions of the reaction products. For instance, Minami et al. exploited the broad substrate tolerance of vicenisaminyltransferase VinC from Streptomyces halstedii HC 34 in the discovery of 22 novel glycosides from 50 sets of reactions for the glycodiversification of natural polyketide scaffolds (Figure 2D)³⁶. More recently, Pandey et al. demonstrated the derivatization of clinically relevant resveratrol glycosides, producing ten different derivatives of the plant-derived metabolite, all accommodated by YjiC, a bacterial GT from Bacillus licheniformis^34,37. However, in order for GT-catalyzed glyco-derivatization to be realized in vivo, the prerequisite biosynthesis of NDP-glycosides as glycosyl donors and acceptors must be engineered from bacterial monosaccharide and nucleotide triphosphate pools. A key advance in the supply of glycosyl donors was the discovery of the reversibility of GT-catalyzed reactions whereby Thorson and co-workers were able to generate more than 70 analogs of the natural products calicheamicin and vancomycin (Figure 1) using various nucleotide sugars³⁸. Using OleD as the initial model enzyme, Thorson’s team evolved the substrate tolerance of GTs to enable glycosylation of not only natural products but also non-natural compounds and proteins^4,39. More recently, Gantt et al. enabled the rapid, colorimetric screening of engineered GTs, and subsequently evolved an enzyme (OleD Loki) for the combinatorial enzymatic synthesis of 30 distinct NDP-sugars that are putatively amenable to further enzymatic manipulation in common microbial hosts^4,40.

Metabolic engineering strategies for in vivo combinatorial glyco-derivatization of secondary metabolites have also demonstrated success by combining heterologous saccharide biosynthesis genes into non-natural pathways. In 1998, Madduri et al. demonstrated the fermentation of the antitumor drug epirubicin (Figure 1) in Streptomyces peucetius and sparked intense interest in the role of metabolic engineering and combinatorial biosynthesis for the discovery and production of glyco-pharmaceuticals^20,41–44. These efforts have begun to impact glyco-engineering in more common microbial hosts, such as an in vivo small molecule “glyco-randomization” study in E. coli by Thorson and co-workers⁴ or the variable derivatization of erythromycin by Pfeifer and co-workers⁴⁵; however, much success for in vivo glyco-derivatization remains in Streptomyces^4,41,46–48.

Diversity through scaffold diversification

Scaffold derivatization enables the fine-tuning of compound activity by increasing compound resolution in a defined chemical space. The production of novel secondary metabolites through scaffold diversification, on the other hand, is a common theme of biosynthesis in plants and fungi that enables the exploration of completely new areas of chemical space. In order to generate beneficial molecules, it has been proposed that microbes and plants generate a diverse library of small molecules. Many liken these broad ranges of natural products to the host’s chemical “immune system”, where producing compounds with no known target could allow for resistance to an as-yet unencountered pathogen and provide evolutionary fitness of organisms with more diverse natural product portfolios^49–51. Natural metabolite diversification has recently inspired diversity-oriented chemical syntheses that emulate the biological reaction cascades in the generation of new, drug-like scaffolds^1,52,53. Others have attempted to simplify metabolite archetypes to common core structures that may serve as starting points for discovery through derivatization; however, metabolite profiling of novel compounds from marine life and fungi continues to produce novel scaffold core structures, suggesting that to limit discovery to known scaffolds would severely curb the biosynthetic potential of evolutionary pressure^2,54–56. Engineering whole cells for scaffold diversification, on the other hand, was recently demonstrated by Wang et al., who combined the biosynthetic potential of plant, fungal, and bacterial enzymes for the production of 12 novel phenylpropanoid derivatives from L-tyrosine and L-phenylalanine in E. coli⁵⁷. Evolva reported the discovery of novel antiviral scaffolds using a heterologous flavonoid biosynthesis platform in S. cerevisiae^3,58. By consolidating biosynthesis and screening into a single cell, the team was able to synthesize, validate, and structurally characterize 74 new compounds—28 of which showed activity in a secondary Brome Mosaic Virus assay—in less than nine months³.

Aiding the discovery of new scaffolds, non-ribosomal peptide synthetases (NRPS) and polyketide synthases (PKS) comprise equally useful, and often interconnected, classes of “assembly line” enzymes for in vivo scaffold diversification. The utility of NRPS/PKS enzymes for complex scaffold synthesis and elaboration emerges from the simplicity and modularity of their catalytic domains⁵⁹. Core NRPS/PKS genes encode for ketosynthase (KS), acyl transferase (AT), acyl/peptidyl carrier protein (ACP/PCP), condensation domain (C), and adenyltransferase (A) that catalyze the elongation of the polyketide/ peptide skeleton, and a terminal thioesterase (TE) severs the formed macrolide from the multi-domain synthetase. Along with auxiliary ketoreductase (KR), dehydratase (DH), and enoyl reductase (ER) domains, the core domains allow for the programmable building of variable macrolide and macrolactone scaffolds from divergent pools of ketoacids and amino acids^17,59–61. Once formed, the core scaffolds are natively derivatized by so-called “tailoring enzymes” to introduce native and non-native functionality as per the discussion of engineered P450s and GTs (vide supra)⁵⁹.

The modularity of the biosynthetic machinery of NRPS/PKS megasynthases allows for the rational engineering of combinatorial biosyntheses to access novel chemical space^7,62. Combinatorial NRPS/PKS systems have enabled predictable changes to the scaffold core, derived from three programmable inputs into the biosynthesis. The inputs include the following: 1) variable use of organic building blocks such as short-chained acyl-coenzyme A (CoA) molecules or amino acids for the chain elongation step of scaffold synthesis^61,63–68; 2) chain length variations originating from KS and TE engineering^69–70; and 3) alterations in the reduction program of the scaffold as a result of DH, KR, and ER engineering^71–72. Yan et al. exemplified the biosynthetic potential to diversify antimycin (ANT) scaffolds through the metabolic engineering of promiscuous NRPS/PKS enzymes in the ANT-producing Streptomyces sp. NRRL 2288 (Figure 3)¹⁷. Following the in vivo production of ANT scaffolds with variable fluorination at C5’ and alkylation at C7, the authors further derivatized the ANT library at C8 with a promiscuous acylating protein (AntB) and various acyl-CoAs in vitro, generating 380 total and 356 novel ANT variants. Chemler et al. recently exploited the biosynthetic prowess of homologous recombination—a natural paradigm of NRPS/PKS evolution—to create PKS libraries for the programmable biosynthesis of engineered polyketide chimeras of known macrolide and macrolactone antibiotics pikromycin and erythromycin (Figure 1)⁶⁰. Similarly, Sugimoto et al. demonstrated that engineering of an artificial PKS pathway by domain swapping in Streptomyces albus allowed reprogramming of the aureothin (Figure 1) system for production of luteoreticulin and novel derivatives thereof⁷³. Despite significant challenges for NRPS/PKS engineering^5,71,74, recent successes with homologous recombination and structure-guided domain swapping of NRPS/PKS’s, coupled to the increased efficiency of Cas9-accelerated gene editing, forecast a time when functional NRPS/PKS variation may be routine^{6,60,64,66,73,75–76}.

Figure 3. Engineered diversification and derivatization of antimycin (ANT) scaffolds by a promiscuous PKS/NRPS (adapted from Yan et al., 2013)¹⁷.

Domain key: T = thioylation (e.g. acyl/peptidyl carrier protein ACP/PCP), C = condensation, A = adenylation, KR = ketoreductase, KS = ketosynthase, AT = acyl transacylase, TE = thioesterase. Green ball represents variable use of H or F-modified starting material. Red vs. blue star depicts unsaturated and carboxylated acyl substituent at C7, respectively. Orange triangle depicts variation in the alkyl chain of the C8 acyl substituent.

Isoprenoids, enumerating over 55,000 compounds, comprise perhaps the richest source of diversity among secondary metabolites⁷⁷. The ability to emulate the natural evolution of this diversity will likely allow the access to known and new plant-derived isoprenoids (Figure 2). Isoprenoid biosynthesis is characterized by four reactions of the five-carbon units isopentenyl pyrophosphate (IPP) and dimethylallyl pyrophosphate (DMAPP): chain elongation, branching, cyclopropanation, and cyclobutanation⁷⁷. Metabolic engineers have manipulated microbial pathways around the chain elongation reaction to build up terpenoid precursors of different lengths and stereochemistries, including IPP/DMAPP (Figure 2 [1a–b]), geranyl diphosphate (Figure 2 [1c]), farnesyl diphosphate (Figure 2 [1d]), and others^78–83. The strategy for introducing diversity as well as directing flux to a desired metabolite then comes from the subsequent pathway and enzyme engineering of terpene synthases that cyclize these building blocks, forming various scaffolds amenable to derivatization with downstream enzymes (Figure 2C). A natural paradigm of terpene synthases and cyclases is the combination of substrate specificity with structural plasticity—a pairing of characteristics that enables rapid evolution of the enzyme for the production of product profiles that meet the environmental demands of the host organism⁸⁴. In support of this hypothesis, multiple groups have confirmed that through evolution and rational engineering, diterpene synthase activities can be altered to produce multiple, non-native terpenes (Figure 2C)^9,84–88. Salmon et al. demonstrated that a convergent point mutation from a library of the Artemisia annua amorpha-4,11-diene synthase (Y420L) enabled the production of numerous cyclized products without compromising catalytic activity⁸⁹. Rising et al. discovered the serendipitous conversion of a non-natural substrate of tobacco 5-epi-aristolochene synthase, anilinogeranyl diphosphate, to the novel paracyclophane terpene alkaloid 3,7-dimethyl-trans,trans-3,7-aza[9]paracyclophane-diene, which they dubbed “geraniline”⁹⁰. The finding demonstrates that terpene precursor diversity and bioavailability, in addition to terpene synthase engineering, are key inputs for programmable scaffold diversification. The explicit application of terpene diversification to diversity-oriented molecule discovery is gaining interest, but to realize the full biosynthetic potential of terpenes will likely require more insight into the mechanism of terpene synthases and the directed biosynthesis of terpene precursors^9,91.