4″-Oxo-avermectin is a key intermediate in the manufacture of the insecticide emamectin benzoate from the natural product avermectin. Seventeen Streptomyces strains with the ability to oxidize avermectin to 4″-oxo-avermectin in a regioselective manner have been discovered, and the enzymes responsible for this reaction were found to be CYPs (cytochrome P450 mono-oxygenases). The genes for these enzymes have been cloned, sequenced and compared to reveal a new subfamily of CYPs. The biocatalytic enzymes have been overexpressed in Escherichia coli, Streptomyces lividans and solvent-tolerant Pseudomonas putida strains using different promoters and vectors. FDs (ferredoxins) and FREs (ferredoxin:NADP+ reductases) were also cloned from Streptomyces coelicolor and biocatalytic Streptomyces strains, and tested in co-expression systems to optimize the electron transport. Subsequent studies showed that increasing the biocatalytic conversion levels to commercial relevance results in the production of several side products in significant amounts. Chimaeric Ema CYPs were created by sequential rounds of GeneReassembly™, a proprietary directed evolution method, and selected for improved substrate specificity by high-throughput screening.
Microbial oxidations have a long and successful history in industrial processes [1,2]. Regioselective microbial oxidations utilize dehydrogenases, oxidases, laccases and CYPs (cytochrome P450 mono-oxygenases) [3,4]. Among these biocatalysts, CYPs are proficient in highly enantio- and regio-specific biosynthetic reactions [5,6]. Streptomyces- or Pseudomonas-derived CYPs have been shown to catalyse the regiospecific oxidation of disparate chemicals including natural products [1,7,8] and the activation and catabolism of xenobiotics [9–11].
Emamectin benzoate  is the active ingredient in the commercial insecticides Proclaim® and Affirm® marketed by Syngenta for the control of many agriculturally important pests. Emamectin benzoate is chemically synthesized from the natural product mixture avermectin B1a and B1b via the key intermediate 4″-oxo-avermectin (Figure 1). Direct regiospecific chemical oxidation of the 4″-hydroxyl of avermectin to 4″-oxo-avermectin is precluded by the high reactivity of the allylic secondary alcohol at position 5 in the molecule, necessitating a protection–deprotection strategy. Avoiding these additional steps would greatly reduce the complexity of the production process along with the final cost of emamectin benzoate, prompting us to search for a cost-effective, regioselective, direct biological oxidation of the 4″-hydroxyl of avermectin [13,14].
(Bio)synthetic conversion of avermectin into emamectin benzoate
Our initial attempts to establish the regioselective oxidation of avermectin to 4″-oxo-avermectin using commercially available dehydrogenases, oxidases and laccases were unsuccessful. Thus we have undertaken a screening programme to discover biocatalytically competent microbes in public and in-house strain collections . Since cultivating microbial strains in liquid media was not feasible for even a medium-throughput screen, we have established a method employing micro-organisms growing on avermectin-containing agar media. Among the 2246 actinomycetes, 243 non-actinomycetes bacteria and 845 fungal strains tested, 17 closely related streptomycetes, all belonging to the Streptomyces lydicus and Streptomyces tubercidicus clusters were identified to be capable of oxidizing the 4″-carbinol of avermectin.
Identification of the genes encoding the biocatalytic enzymes
In vitro enzyme assays using crude cell extracts were found to require FD (ferredoxin) and FRE (ferredoxin:NADP+ reductase), together with NADPH as the source for electrons, indicating that the biocatalytic enzymes (collectively named the Ema enzymes) are CYPs. We proposed that the Ema CYPs hydroxylate the 4″ secondary alcohol of avermectin, and the resultant 4″-gem-diol then collapses to the corresponding ketone by dehydration (Figure 1). 4″-Oxo-avermectin hydrate was subsequently identified by LC-MS and shown to equilibrate with the 4″-ketone in water-containing solvents . The CYP Ema1 was purified from S. tubercidicus R-922 and characterized as a novel soluble CYP that requires FD, FRE and NADPH for electron supply, and its crystal structure has been determined at 2.2 Å (1 Å=0.1 nm) resolution without substrate (V. Jungmann and J.P. Pachlatko, unpublished work).
To identify the gene encoding the Ema2 biocatalytic enzyme, we have screened more than 16000 clones of S. lividans expression libraries carrying random genomic DNA fragments of the Ema2-producer strain, S. tubercidicus I-1529. No biocatalytically competent transformants emerged: we have later found that co-expression of a suitable FD-encoding gene is necessary to yield detectable biocatalytic activity in ema2-expressing S. lividans strains . Next, we have cloned eight CYP-homologous genes (five from strain I-1529 and a further three from the Ema1-producer strain, S. tubercidicus R-922) using heterologous CYP gene probes, but none of the corresponding CYP enzymes catalysed the oxidation of the 4″-carbinol of avermectin. Streptomyces strains, however, are known to produce many different CYP enzymes, with the genomes of S. coelicolor A3(2) and Streptomyces avermitilis MA-4680 encoding 18 and 33 CYPs respectively [15–17].
The gene encoding Ema1 was ultimately cloned from S. tubercidicus R-922 by a PCR-based approach using oligonucleotides whose design was based on the determined amino acid sequence of the purified enzyme. The ema1 gene was used as a probe to clone the encoding genes of the remaining 16 biocatalytic CYPs . The 17 Ema enzymes show mutual sequence identities exceeding 60%. The amino acid sequences of Ema4 and Ema10, and Ema2 and Ema14 were found to be identical respectively despite originating from apparently different Streptomyces spp. The closest homologue of the Ema enzymes with identities of less than 49% is a CYP from the carbomycin biosynthetic gene cluster (CYP107C1). The Ema enzymes belong to the CYP107 family that also harbours many other Streptomyces CYPs involved in xenobiotic or secondary metabolism. They, however, form a newly established subfamily, CYP107Z, with no other known members (D.R. Nelson, personal communication). The eight non-Ema CYPs cloned from strains I-1529 and R-922 are members of divergent CYP subfamilies.
Kinetic characterization of Ema1–Ema16
Sixteen of the 17 Ema enzymes were overproduced as their C-terminal His-tagged versions in Escherichia coli BL21(DE3). Purifications yielded >95% pure, red-coloured CYPs that regioselectively oxidized avermectin. The kinetic parameters of the purified His-tagged enzymes were determined in the presence of FD and FRE from spinach.
The Km values of the purified His-tagged Ema enzymes were in the micromolar range, while their Vmax varied between 0.001 and 0.060 μmol·min−1·mg−1 Ema CYP protein, with Ema1 showing the highest specific activity. The calculated catalytic-centre activity of Ema1 was 0.048 s−1 . In comparison, 10–1000-fold higher catalytic-centre activities were recorded for NADPH-dependent CYPs acting on their native substrates [18–20]. Since avermectin is unlikely to be the natural substrate for the Ema enzymes, the interaction of this fortuitous substrate is probably suboptimal with the active sites of the enzymes, leading to the observed lower catalytic-centre activities.
Heterologous expression of ema genes
CYP-catalysed reactions require NAD(P)H as a source of electrons. Although enzymatic [21,22] or direct  regeneration of this cofactor is feasible, metabolically active cells were expected to provide a more cost-effective solution in an industrial setting. Thus we investigated different Ema expression systems to produce resting cells with high biocatalytic activity. S. lividans ZX7 and two solvent-tolerant Pseudomonas putida strains were chosen for primary expression hosts. The constitutive PermE*  and the thiostrepton-inducible PtipA promoter  were used for S. lividans, while the Ptac promoter  that is constitutive in Pseudomonas and the alkane-inducible Palk promoter  were utilized for Ps. putida.
Expression of the ema1 gene from the different promoters provided biocatalytically active S. lividans and Ps. putida cells (Figure 2), indicating that Ema1 is able to partner with FD and FRE electron transport proteins from these strains. Ema2 on the other hand, despite having similar a kcat/Km to that of Ema1 in vitro, showed biocatalytic activity only in Ps. putida, but not in S. lividans.
Biocatalytic conversion in heterologous expression systems
Both ema1 and ema2 contain near their 5′-end a TTA codon that is extremely rare in Streptomyces and is used as a post-transcriptional regulatory device that restricts the translation of the corresponding mRNA to late growth phases [28,29]. We attempted to release Ema1 and Ema2 from this regulation in S. lividans by either truncating the corresponding gene (ema1A) or by creating a silent replacement of the TTA codon (ema1CTG and ema2CTG). No improvements were achieved with the ema2 gene variants in S. lividans. A significant increase in Ema1 biocatalytic activity was, however, recorded in the inducible, but not in the constitutive, S. lividans expression system (Figure 2). The inducible PtipA promoter is active only in the early phases of the cultures  when TTA codons restrict translation, while the constitutive promoter is still active in the stationary phase when TTA-containing genes such as ema1 are actively translated. Thus we hypothesize that removing the TTA codon allows the ema1A- or the ema1CTG-expressing strains to capitalize on the high early-stage activity of the PtipA promoter.
Co-expression of electron transport components
Biotransformation reactions with bacterial Class I CYPs require an electron transport chain consisting of a small iron–sulfur protein FD and a FAD-containing FRE, although neither of these partners need to be exclusive for a given CYP . Tellingly, the 18 CYPs of S. coelicolor rely on six FDs and three to five FREs for electron transport , while the two CYPs of Ps. putida KT2440 utilize five FDs and five FREs . Specific CYPs, however, do show higher activity with specific FDs [6,32], and certain FREs are also known to be more efficient electron donors for certain FDs than others . Whole-cell biocatalysts thus need to express suitable FD and FRE partners for a heterologous CYP enzyme.
While none of the 17 ema genes was found to be clustered with FD- or FRE-encoding genes, several of the non-biocatalytic CYP genes cloned in the early phase of the project displayed such associations. Thus we identified genes for four different low-potential monocluster [3Fe-4S] FDs  from S. tubercidicus strains R-922 and I-1529 . Co-expression of Fd233 with Ema1 or Ema1A from artificial operons led to approx. 2-fold increases in the bioconversion rates in S. lividans (Figure 2). Fd232 (derived from R-922) was found to be a less effective electron donor for Ema1 in spite of its near identity with Fd233 (derived from I-1529). Increased biocatalytic performance with the ema1-fd233 artificial operon, compared with ema1 alone, was also manifest in Ps. putida ATCC 17453, but less evident in Ps. putida ATCC 700801 (Figure 2). Identification of the cognate FD partner for Ema1 might further improve the efficiency of the recombinant biocatalysts.
Expression of an artificial operon of ema2 with fd233 led to a dramatic improvement in the biocatalytic performance in S. lividans (Figure 2). We speculate that Ema2 might not be receptive to S. lividans FDs, but co-expression of a cloned FD that is acceptable to both Ema2 and the FREs of S. lividans restores the electron flow and the activity of the enzyme.
To further optimize the biocatalyst, we have cloned three FREs from S. tubercidicus R-922, two from S. tubercidicus I-1529 and four FRE homologues from S. coelicolor, and used these to create artificial operons with ema1-fd233 or ema1A-fd233 that were expressed in S. lividans and Ps. putida . Fre12 and Fre16 provided for a nearly 2-fold higher 4″-oxo-avermectin yield in Ps. putida ATCC 17453 (results not shown). In S. lividans, significant increases were seen in the biocatalytic performance when FRE-encoding genes were co-expressed with Ema1 and Fd233 using the PtipA promoter, while no improvements were seen when Ema1A was used instead of Ema1 irrespective of the promoter choice (results not shown). We propose that FRE availability might have been a limiting factor in S. lividans in the early stages of the cultures when the PtipA promoter is active. Similarly, Lei et al.  found that FRE expression was restricted to older cultures, while the native CYP complement of S. coelicolor was expressed even in early-stage cultures.
Biocatalysis in the presence of solvents
The exceedingly low solubility of avermectin in water prompted us to investigate a solvent–water biphasic system to accommodate higher concentrations of substrate during biocatalysis. Solvent-tolerant Ps. putida strains  carrying different Ema expression constructs were cultivated in the presence of increasing concentrations of dioctyl phthalate. In spite of acceptable growth rates, the bioconversion capacities of these cells decreased rapidly (Figure 2B), apparently due to a dramatic decline of Ema1 expression .
Biotransformation reactions run to increasingly high substrate concentrations and conversion rates also yielded avermectin-derived side products in substantial amounts apart from the desired 4″-oxo-avermectin product. The two major side products were 4″-oxo-3″-desmethyl-avermectin and 4′-desoleandrosyl-avermectin. We propose that the side products are derived by the oxidative degradation of avermectin and 4″-oxo-avermectin (Figure 1), and the formation of some of the side products is catalysed by the Ema1 enzyme itself as no such products were formed from avermectin or 4″-oxoavermectin by S. tubercidicus R-922 with an inactivated ema1 gene (results not shown). Side products accumulation showed an exponential correlation to the conversion rate, precluding the commercial utilization of the native Ema1 enzyme. Similar results were recorded with the other Ema enzymes.
To optimize the Ema CYPs, we have utilized GeneReassembly™, a proprietary ligation-based recombination procedure where chimaeras are assembled from DNA fragments generated by PCR or oligonucleotide synthesis. Cross-over points can be designed for any given position of a gene. This extensive control over the content and the size of the library allowed the Ema molecular evolution programme to incorporate structural, functional and sequence information to construct high-quality, targeted, low-complexity libraries. Moderate library sizes were crucial since the throughput of this molecular evolution programme was limited by an HPLC-based screening method.
Four successive rounds of GeneReassembly™ were used to incorporate sequence variations from three different sources: (i) fragments from the wild-type Ema1, Ema6, Ema8, Ema12, Ema13 and Ema16 CYPs; (ii) designed artificial ‘consensus’ sequences representing further sequence diversity from other Ema CYPs; and (iii) point mutations from combinatorial mutagenesis of targeted positions of the substrate-binding pocket area. The best enzyme emerging from this programme, Ema-V5b (Figure 3), was expressed in S. tubercidicus R-922 after allele exchange with Ema1. This biocatalyst has repeatedly demonstrated a yield of >70% 4″-oxo-avermectin without the costly protection/deprotection of the highly reactive C5 allyl alcohol used in the chemical process (Figure 3; and I. Molnár, V. Jungmann, J. Stege, A. Trefzer and J.P. Pachlatko, unpublished work). Further optimization of the biocatalytic process parameters will lead to an economical biocatalytic alternative to a costly chemical synthesis step in the commercial production of the insecticide emamectin benzoate.
GeneReassembly™ of CYP10Z subfamily enzymes
8th International Symposium on Cytochrome P450 Biodiversity and Biotechnology: Independent Meeting held at Swansea Medical School, Swansea, Wales, U.K., 23–27 July 2006. Organized and Edited by D. Kelly, D. Lamb and S. Kelly (Swansea, U.K.).
This work was supported by Syngenta Biotechnology, Syngenta Crop Protection and Diversa Corporation.
Present address: Invitrogen Corporation, 1600 Faraday Avenue, Carlsbad, CA 92008, U.S.A.