Enhancement and Selective Production of Phoslactomycin B, a Protein Phosphatase IIa Inhibitor, through Identification and Engineering of the Corresponding Biosynthetic Gene Cluster*

Phoslactomycins (PLMs), potent and selective inhibitors of serine threonine phosphatases, are of interest for their antitumor and antiviral activity. Multiple analogs and low titers in the fermentation process have ham-pered the development of this class of natural products. The entire 75-kb PLM biosynthetic gene cluster of Streptomyces sp. HK-803 was cloned, sequenced, and analyzed. The loading domain and seven extension modules of the PLM polyketide synthase generate an unusual linear unsaturated polyketide chain containing both E-and Z -double bonds from a cyclohexanecarboxylic acid (CHC) primer. Hydroxylation of the CHC-derived side chain of the resulting PLM-B by PlmS 2 , and a subse- quent esterification, produces the remaining PLM analogs. A new PCR targeting technology allowed rapid and facile allelic replacement of plmS 2 . The resulting mutant selectively produced the PLM-B, at 6-fold higher titers than the wild type strain. This mutant and the biosynthetic gene cluster will facilitate engineered microbial production of hybrid PLMs with improved properties. Streptomyces produce a series of structurally novel antitumor agents of

breast, and ovarian cancer. The same activity likely also permits fostriecin to limit myocardial infarct size and to protect cardiomyocytes during ischemia (6 -8). Recent evidence suggests that PP2A has a role in regulating cell death by apoptosis, in the activation of natural killer cells and cytotoxic T-lymphocytes involved in tumor surveillance, and even in the transcription and replication of human immunodeficiency virus-1 (HIV1) (9). For these reasons selective protein phosphatase inhibition has been suggested as a clinically unexplored novel mechanism worthy of pursuit for the introduction of a new class of antitumor or even antiviral agents (1,5,9).
Phoslactomycins (PLMs) (also called phosphazomycins or phospholines) and leustroducsins ( Fig. 1) are structurally related natural products. The principal differences from fostriecin are replacement of the C-8 methyl substituent by ethylamine and the terminal allylic alcohol by a cyclohexane ring. Like fostriecin, these compounds are produced in multiple forms (Fig. 1). PLM-B (phospholine) contains a fully reduced cyclohexyl ring, whereas the other PLMs and the leustroducsins contain a hydroxyl substituent (C-18) esterified with a wide array of carboxylic acids (ranging in length from 4 to 9 carbons) (10 -13). The phoslactomycins and leustroducsins all exhibit antifungal activity (1,10,13,14). More importantly, these compounds like fostriecin are potent and highly selective inhibitors of PP2A (IC 50 values raging from 3.7 to 5.8 M) as compared with PP1 (IC 50 Ͼ 1 mM) (15). Most recently, the PP2A inhibition activity of phoslactomycin has been shown to inhibit tumor metastasis through augmentation of natural killer cells (16).
The unique and selective biological activity of this class of natural products has attracted considerable interest in recent years. Phase I clinical trials of fostriecin were suspended before either dose-limiting toxicities or therapeutic plasma levels were attained because of inherent drug instability and unpredictable purity in the clinical supply of the natural product (1,4). In an attempt to address these limitations and further develop this class of novel antitumor agents, no less than six elegant total syntheses of fostriecin have been developed over a short 2-year period (5,(17)(18)(19)(20)(21). A complementary approach for developing these agents can come through gene manipulation of the natural biosynthetic process. In all cases, the production of numerous analogs complicates the isolation process and the purity of the product, as well as negatively impacting yields (2,14,22). These analogs are based on variations in the level of hydroxylation and, in the case of the PLMs and leustroducsins, esterification of the resulting secondary alcohol with a series of different carboxylic acids (this modification does not significantly impact either antifungal activity or binding to PP2A). * This work was supported by a Grant AI51629-01 from the National Institutes of Health. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The This is an Open Access article under the CC BY license.
Using a PLM-producing strain, Streptomyces sp. HK-803, we have carried out an incorporation study of [7-13 C]CHC, which suggests that PLM-A and PLM C-F are derived from a hydroxylation of PLM-B. The entire 75 kb of the phoslactomycin biosynthetic gene cluster has been identified and sequenced, including a gene, plmS 2 , encoding the cyctochrome P-450 monooxygenase responsible for this hydroxylation. The application of a new PCR targeting technology (23) in this strain allowed rapid gene replacement of plmS 2 . The resulting mutant strain, NP1, produces only PLM-B, at 6-fold higher levels than the wild type strain. This mutant will serve as a starting point for engineering strains capable of producing fostriecin and PLM analogs.

EXPERIMENTAL PROCEDURES
Strains and Culture Conditions-All Escherichia coli strains used in this study were grown following standard protocols (24). The PLM producer Streptomyces sp. HK-803 (13) was grown at 28°C in SY medium (1.0% soluble starch, 0.1% yeast extract, 0.1% N-Z-amine, type A) at pH 7.0. The PLM production medium was 2.0% glucose, 0.1% beef extract, 1.0% soybean flour, 0.2% NaCl, 0.005% K 2 HPO 4 , and 0.2% L-phenylalanine at pH 7.0. A loop-full of Streptomyces sp. HK-803 spores (from an SY agar plate) was inoculated into a baffled 500-ml flask containing 70 ml of the production medium and was incubated at 28°C in the dark for 96 h on a rotary shaker at 170 rpm.
Analyses of PLM Production-The culture filtrates and mycelial extracts of Streptomyces sp. HK-803 wild type and the mutant NP1 were worked up as described previously (13). HPLC was performed on a system equipped with a Waters TM 600 pump and a Waters 996 photodiode array detector under the following conditions: column, Senshu Pak PEGASIL ODS (4.6 ϫ 250 mm for analytical HPLC and 20 ϫ 250 mm for preparative HPLC); solvent, acetonitrile, 0.05% HCO 2 H (40/60); flow rate, 1 ml/min for analytical HPLC and 9 ml/min for preparative HPLC; detection at UV 235 nm. Pure PLM standards were used to generate a standard curve for determining PLM production in both the wild type and NP1 mutant. The PLMs were also identified by co-injections with the corresponding standards and by LC-MS analyses. LC-MS spectra (positive turbo-ion spray ionization mode; HPLC, Hewlett-Packard Series 1100; column, 150 ϫ 2.1 mm RP 183.5-mm column from Waters; mobile phase, 40:60 CH 3 CN-water (containing 0.05% HCO 2 H) at a flow rate 0.2 ml/min) were taken on a PerkinElmer Life Sciences SCIEX API 2000 pneumatically assisted electrospray triple quadrupole mass spectrometer.
Incorporation of [7-13 C]CHC into PLMs-[7-13 C]CHC was synthesized from cyclohexanone and K 13 CN via cyclohexanecarbonitrile as described previously (25,26) , the abundance of the molecular ion (m/z), was determined. A similar analysis was carried out with unlabeled PLMs and allowed the contributions of naturally abundant 13 C and [7-13 C]CHC to be separated in the standard manner. The amount of each 13 C-labeled PLM derived from [7-13 C]CHC was determined as a percentage of that entire analog pool. The location of the 13 C label was confirmed by 13 C NMR analyses of HPLC purified PLM-B, -E, and -F (see "Experimental Procedures") using previously published 13 C assignments (10).
Cloning and Sequencing of the PLM Biosynthetic Gene Cluster-The total genomic DNA of Streptomyces sp. HK-803 (grown on SY medium) was prepared according to the cetyltrimethylammonium bromide procedure for the isolation of genomic DNA in the manual of Practical Streptomyces genetics (27). A genomic library was constructed by using Supercos-1 cosmid vector as recommended in the manufacturer's protocol (Stratagene). About 2000 cosmid clones were probed by using the DNA of a digoxigenin-labeled chcA gene (28). Preparation of the digoxigenin probes and the subsequent hybridization and detection were performed as recommended in the manufacturer's protocol (Roche Applied Science). The overlapping cosmid clones identified by the chcA probe were sequenced to completion using the TOPO shotgun subcloning kit (Invitrogen). Automated DNA sequencing was performed on an ABI Prism 3700 DNA sequencer at the DNA core facility of the Medical College of Virginia, Virginia Commonwealth University. The DNA sequences were assembled using SeqMan II (DNAStar Inc.). The assembled DNA and deduced protein sequences were analyzed with MacVector 7.0 software (Accelrys) and FramePlot (29) and were compared with sequences in the public data bases using the BLAST suite of programs (30).
Targeted Disruption of plmS 2 -The plmS 2 gene of the PLM biosynthetic gene cluster was disrupted by using the recently developed PCRtargeted Streptomyces gene replacement method (23). The aac (3)IV resistance marker and oriT were amplified from the pIJ 773 disruption cassette (23) using the primers PLMS 2 -F 5Ј-CCGCCGCCCCGAGCAC-GA-AGAGACGGCGTCCTCCGCATGattccggggatccgtcgacc-3Ј and PLMS 2 -R 5Ј-CCAGG-ACGACCGGGCCGGTTCAGCGTGCGGGAACG-CTCAtgtaggctggagctgcttc-3Ј (pIJ773 homologous sequence is in lowercase letters). The resulting PCR product was used to replace plmS 2 first in cosmid clone 3A11 and then in Streptomyces sp. HK-803 following the established methodologies (23), replacing mannitol soy agar with SY agar. Allelic replacement of the plmS 2 gene in the NP1 mutant was confirmed by Southern hybridization analysis, and PCR amplification and sequencing.

Incorporation Studies Establish CHC as a Precursor to all
PLMs-Analysis of the PLM-B structure suggested that it is likely assembled by a type I modular polyketide synthase (PKS) using a CHC starter unit. The other PLMs and the leustroducsins may be derived by hydroxylation of this PLM-B or by use of a cis-3-hydroxy-CHC starter unit. A CHC-derived moiety has previously been observed as a side chain in the antifungal agent ansatrienin and shown to be derived from shikimic acid (31,32). LC-MS analysis of the PLM mixture A pathway from shikimic acid to CHC-CoA has been delineated in the ansatrienin producer, Streptomyces collinus (31). The corresponding biosynthetic genes have recently been identified (33,34).
Isolation and Characterization of the PLM Biosynthetic Gene Cluster-The PLM biosynthetic gene cluster from the Streptomyces sp. HK-803 was identified from a cosmid library of genomic DNA using the chcA gene (28) from CHC-CoA biosynthetic gene cluster of S. collinus as a probe. A cosmid clone, 3A11, thus identified was used as a probe to identify two overlapping cosmids, 10B4 and 3E5 (Fig. 2). These three cosmids spanning 97 kb of the Streptomyces sp. HK-803 genome were sequenced and analyzed. This sequence analysis revealed a 75-kb region containing 29 open reading frames (ORFs) that appear to be associated with PLM biosynthesis. The flanking region contained a series of ORFs that are highly homologous (Ͼ80% identity at the amino acid sequence level) with and have an organization similar to the primary metabolic genes identified through sequencing of the S. coelicolor genome (35), which aided in delimiting the PLM biosynthetic gene cluster. The sequence analysis revealed the existence of the complete set of highly conserved CHC-CoA biosynthetic genes (plmJK, plmL, chcA, plmM) (28,33) as well as plmI (encoding a 3-deoxy-Darabino-heptulosonate-7-phosphate synthase) responsible for formation of the CHC-CoA starter unit. As in S. collinus, the chcB gene encoding the ⌬ 2,3 -enoyl CoA isomerase responsible for catalyzing the penultimate step in the CHC-CoA pathway does not appear to be located close to the other CHC-CoA biosynthetic genes or within the antibiotic biosynthetic gene cluster (36). A total of six large ORFs encoding the core PLM PKS were also identified (Fig. 3). PLM1 contains an initiation domain, presumably responsible for loading the CHC-CoA, and the first extension module (Fig. 3), whereas PLM2-3 has two modules containing the appropriate predicted catalytic activitities for catalyzing the second and third extension steps. Each of the remaining components, PLM4 -PLM7, contains a single module. PLM7 has a thioesterase domain, consistent with a role in catalyzing the last extension step and subsequent release of the polyketide chain by formation of the ␣,␤-unsaturated ␦-lactone (formation of a six-member lactone is unusual, as most thioesterase domains studied to date catalyze formation of large 12-14 macrolactone structures under natural conditions) (37,38). Two of the PKS polypeptides PLM4 and PLM6, contain the same catalytic domains, which precludes unambiguous assignment in this process. One of many interesting features of the PLM PKS cluster is that it contains only three dehydratase domains, and yet it produces a compound with three Z-and one E-double bonds (PLM7 does not contain the predicted dehydratase domain). Virtually all known cases in which a PKS modules has a ketoreductase and a dehydratase activity result in the formation of E-double bonds (39). It has been suggested that a stereochemical configuration of the product derived from the ketoreductase domain may control the double bond geometry. A diagnostic Asp residue has been proposed for predicting the stereospecificity of PKS ketoreductase domains and, by extension, the double bond configuration for products from cognate dehydratase domains (39). We constructed a multiple alignment of the seven ketoreductase domains of the PLM PKS and found that this model alone was only partially predictive of the stereochemical configuration of the PLM product.
The PLM gene cluster also contains genes that encode a separate type II TE protein (plmT 8 ), recently established to have an important role in the editing and proofreading of PKSs (40), and a crotonyl-CoA reductase (plmT 7 ), presumably required for providing the ethylmalonyl-CoA extender unit used in the fourth and sixth extension steps. The plmT 5 gene product has some sequence identity with homoserine kinases and is likely responsible for phosphorylation of the C-9 hydroxyl group and for providing biological activity to PLM. Analysis of the gene cluster also predicts an oxidoreductase (plmT 8 ) and aminotransferase (plmT 1 ), likely involved in the replacement of a hydroxyl group of a C-8 hydroxyethyl side chain with an amine. The gene cluster also contains numerous other genes FIG. 2. The phoslactomycin biosynthetic gene cluster. The genes plm1-plm8 encode the polyketide synthase (see Fig. 3 for more details). Genes proposed to be involved in providing the precursors CHC-CoA and butyryl CoA (plmT 7 ) are shown in light blue. Regulatory genes (dark blue) and an ABC transporter (green) are shown. Genes depicted in yellow encode enzymes proposed to be involved in postpolyketide modification and include two cytochrome P-450 monooxygenases (plmS 2 , plmT 4 ), an oxidoreductase (plmT 8 ), an aminotransferase (plmT 1 ), an a kinase (plmT 5 ). The plmS 2 mutant (strain NP1) produces only PLM-B (see Fig. 4). The sequence of the entire gene cluster and the proposed function for individual ORFs is summarized in GenBank TM under accession number AY354515. involved in regulation and resistance. A detailed analysis of the entire PLM gene cluster will be described elsewhere.
PCR-targeted Gene Disruption of plmS 2 -Analysis of the PLM biosynthetic gene cluster, together with the results of the biosynthetic incorporation study, indicated that three hydroxylation steps (C-18, C-8, and the C-8 ethyl side chain) occur after assembly of the polyketide chain from the CHC-derived starter unit. Analysis of the PLM biosynthetic gene cluster revealed two ORFs, plmT 4 and plmS 2 , that both encode proteins with high sequence similarity to each other (38% identity and 53% similarity) as well as cytochrome P-450 monooxygenases from several microorganisms, including those involved in the modification of the natural products of streptomycetes (41). We reasoned that replacement of one of these genes would provide a strain blocked in hydroxylation of C-18 of PLM. With the caveat that the other hydroxylations are catalyzed by the second monooxygenase, the mutant strain should produce PLM-B exclusively.
There are numerous different approaches for probing the secondary metabolic gene clusters through targeted gene disruptions. These approaches typically involve in vitro generation of suitable gene disruption vectors, introduction of these into the host using methods such as conjugation or transformation, and an extensive screening process to find the desired allelic replacement (42,43). This entire process can take several months, even when methodologies have been developed for the strain. For strains that have not previously been characterized or are not genetically amenable, the process can be substantially longer (44,45). We applied a new PCR-targeted gene replacement, developed for work with Streptomyces coelicolor (23), to efficiently and rapidly carry out targeted disrup-tion of plmS 2 in Streptomyces sp. HK-803. The plmS 2 in cosmid 3A11 was replaced in vivo by aac (3)IV (Apr R ) and the oriT (which allows for efficient transfer to Streptomyces by RP4mediated intergenic conjugation). This recombination was accomplished in E. coli BW25113/pIJ790 (expressing the -Red recombination functions) using a PCR product in which these two genes and a yeast Flp-recombinase target sequence were flanked by 39 nt of plmI and plmS 1 (genes surrounding plmS 2 ). The mutagenized cosmid, isolated from the resulting apramycin-resistant transformants, was introduced by electroporation into E. coli ET12567/pUZ8002 and was subsequently transferred into Streptomyces sp. HK-803 by conjugation. Approximately 80% of the resulting Apra R exconjugants were Kan S , indicating a double-crossover allelic exchange of plmS 2 , which was exhaustively confirmed by PCR and Southern blot analysis (data not shown). Utilization of the same approach by us and others (23) in S. coelicolor have shown a frequency of 30% or less double crossovers, suggesting that Streptomyces sp. HK-803 may have a higher rate of homologous recombination.
Enhanced and Selective Production of PLB by the plmS 2 Mutant-The plmS 2 mutant (NP1) was grown under standard conditions alongside the wild type PLM producer. After 4 days of fermentation both cultures were purified by ion exchange chromatography and analyzed by reverse-phase HPLC. The wild type strain produced a mixture of PLMs (A-F) with a total yield of PLMs of 7 mg/ml (Fig. 4). In stark contrast the NP1 mutant produced only PLM-B, at ϳ6-fold higher levels (10 mg/ml) than observed for the wild type (Ͻ1.6 mg/liter). The PLM-B titer in the NP1 mutant exceeds the cumulative yield of all PLM analogs in the wild type strain. The PLM-B was confirmed by coinjections with a standard and by LC-MS analyses, which demonstrated the correct mass ([MϩH] ϩ ϭ 514). These observations unequivocally demonstrate that the plmS 2 gene product is responsible for catalyzing hydroxylation at C-18 of PLM and is not required for hydroxylation at other positions. Hydroxylation at C-8 and the C-8 ethyl side chain are most likely catalyzed by the second cytochrome P-450 monooxygenase, encoded by plmT 4 . The efficient production of PLM-B by the NP1 strain suggests that hydroxylation at C-18 of PLM-B and the subsequent esterification step may be the final steps in the biosynthetic process. The possibility that these steps might also occur prior to other modifications of the PLM carbon backbone cannot be discounted.
Conclusion-This work represents the first cloning and analysis of a biosynthetic gene cluster for a potent and selective protein phosphatase inhibitor. Analysis and subsequent manipulation of this gene cluster has allowed for the rational engineering of a strain that produces only one analog, PLM-B, at significantly higher yields than the wild type strain. Some of the key drawbacks to continued development of this class of naturally occurring compounds, including low fermentation titers and resolution of the various analogs produced in the fermentation (both contributing to concerns about drug purity), have now been overcome. Cloning of the PLM biosynthetic gene cluster and creation of a strain producing only PLM-B also set the stage for the generation of novel compounds with improved stability, activity, and selectivity. As such, this work represents a complementary approach to the widespread synthetic approaches directed toward this class of compounds (5,(17)(18)(19)(20)(21). One of the key structural differences between PLMs and fostriecin is the C-8 ethylamine substituent, which may contribute to less potent activity (15). Inactivation of the plmT 1 gene predicted to be responsible for incorporation of this amino group into the PLM structure should allow for the generation of a novel hybrid structure bearing features of both fostriecin and PLM. Other hybrid natural products may be obtained by combining different PKS components from the two biosynthetic gene clusters. These avenues of research are currently being pursued and will be reported in due course.
This work has also demonstrated the utility of PCR-targeted gene replacement (23) for the rapid characterization and manipulation of polyketide biosynthetic gene clusters. This methodology, demonstrated for the strain S. coelicolor, has worked effectively for us not only in Streptomyces sp. HK-803 but also Streptomyces cinnamonensis. 2 The technique is potentially applicable to any other streptomycetes for which conditions for conjugation from E. coli either have or can be developed (43). As such, this new technology should revolutionize the way in which most natural product biosynthetic gene clusters are analyzed and manipulated. It represents a powerful new genetic tool for the generation of novel natural products through combinatorial biosynthesis.