CA2350189A1 - Recombinant methods and materials for producing epothilone and epothilone derivatives - Google Patents

Abstract

Recombinant nucleic acids that encode all or a portion of the epothilone polyketide synthase (PKS) of sorangium cellulosum are used to express recombinant PKS genes in host cells for the production of epithilones, epothilone derivatives, and polyketides that are useful as cancer chemotherapeutics, fungicides, and immunosuppressants.

Description

RECOMBINANT METHODS AND MATERIALS FOR PRODUCING
EPOTHILONE AND EPOTHILONE DERIVATIVES
Reference to Government Funding This invention was supported in part by SBIR grant 1843-CA79228-O1. The U.S.
government has certain rights in this invention.
Field of the Invention The present invention provides recombinant methods and materials for producing epothilone and epothilone derivatives. The invention relates to the fields of agriculture, chemistry, medicinal chemistry, medicine, molecular biology, and pharmacology.
Background of the Invention The epothilones were first identified by Gerhard Hofle and colleagues at the National Biotechnology Research Institute as an antifungal activity extracted from the myxobacterium Sorangium cellulosum (see K. Gerth et al., 1996, J. Antibiotics 49: 560-563 and Germany Patent No. DE 41 38 042). The epothilones were later found to have activity in a tubulin polymerization assay (see D. Bollag et al., 1995, Cancer Res.
55:2325-2333) to identify antitumor agents and have since been extensively studied as potential antitumor agents for the treatment of cancer.
The chemical structure of the epothilones produced by. Sorangium cellulosum strain So ce 90 was described in Hofle et al., 1996, Epothilone A and B -novel 16-membered macrolides with cytotoxic activity: isolation, crystal structure, and conformation in solution, Angew. Chem. Int. Ed. Engl. 35(13/14): 1567-1569, incorporated herein by reference. The strain was found to produce two epothilone compounds, designated A (R = H) and B (R = CH3), as shown below, which showed broad cytotoxic activity against eukaryotic cells and noticeable activity and selectivity against breast and colon tumor cell lines.

WO 00!31247 - 2 - PCT/US99/27438 R
S ~d~".. 10 I -~ .; 15 14 $ .~~C
a OFi The desoxy counterparts of epothilones A and B, also known as epothilones C (R
=
S
H) and D (R = CH3), are known to be less cytotoxic, and the structures of these epothilones are shown below.
R
S
J
y s~
a' Two other naturally occurring epothilones have been described. These are epothilones E and F, in which the methyl side chain of the thiazole moiety of epothilones A and B has been hydroxylated to yield epothilones E and F, respectively.
Because of the potential for use of the epothilones as anticancer agents, and because of the low levels of epothilone produced by the native So ce 90 strain, a number of research teams undertook the effort to synthesize the epothilones. This effort has been successful (see Balog et al., 1996, Total synthesis of (-}-epothilone A, Angew. Chem. Int.
Ed. Engl. 35(23/24): 2801-2803; Su et al., 1997, Total synthesis of (-)-epothilone B: an extension of the Suzuki coupling method and insights into structure-activity relationships of the epothilones, Angew. Chem. Int. Ed. Engl. 36(7): 757-759; Meng et al., 1997, Total syntheses of epothilones A and B, JACS 119(42): 10073-10092; and Balog et al., 1998, A
novel aldol condensation with 2-methyl-4-pentenal and its application to an improved total synthesis of epothilone B, Angew. Chem. Int. Ed. Engl. 37(19): 2675-2678, each of which is incorporated herein by reference). Despite the success of these efforts, the chemical synthesis of the epothilones is tedious, time-consuming, and expensive.
Indeed, the methods have been characterized as impractical for the full-scale pharmaceutical development of an epothilone.
A number of epothilone derivatives, as well as epothilones A - D, have been studied in vitro and in vivo (see Su et al., 1997, Structure-activity relationships of the epothilones and the first in vivo comparison with paclitaxel, Angew. Chem.
Int. Ed. Engl.

36(19): 2093-2096; and Chou et al., Aug. 1998, Desoxyepothilone B: an efficacious microtubule-targeted antitumor agent with a promising in vivo profile relative to epothilone B, Proc. Natl. Acad. Sci. USA 95: 9642-9647, each of which is incorporated herein by reference). Additional epothilone derivatives and methods for synthesizing epothilones and epothilone derivatives are described in PCT patent publication Nos.
99/54330, 99/54319, 99/54318, 99/43653, 99/43320, 99/42602, 99/40047, 99/27890, 99/07692, 99/02514, 99/01124,98/25929, 98/22461, 98/08849, and 97/19086; U.S.
Patent No. 5,969,145; and Germany patent publication No. DE 41 38 042, each of which is incorporated herein by reference.
There remains a need for economical means to produce not only the naturally occurring epothilones but also the derivatives or precursors thereof, as well as new epothilone derivatives with improved properties. There remains a need for a host cell that produces epothilones or epothilone derivatives that is easier to manipulate and ferment than the natural producer Sorangium cellulosum. The present invention meets these and 1 S other needs.
Summary of the Invention In one embodiment, the present invention provides recombinant DNA compounds that encode the proteins required to produce epothilones A, B, C, and D. The present invention also provides recombinant DNA compounds that encode portions of these proteins. The present invention also provides recombinant DNA compounds that encode a hybrid protein, which hybrid protein includes all or a portion of a protein involved in epothilone biosynthesis and all or a portion of a protein involved in the biosynthesis of another polyketide or non-ribosomal-derived peptide. In a preferred embodiment, the recombinant DNA compounds of the invention are recombinant DNA cloning vectors that facilitate manipulation of the coding sequences or recombinant DNA expression vectors that code for the expression of one or more of the proteins of the invention in recombinant host cells.
In another embodiment, the present invention provides recombinant host cells that produce a desired epothilone or epothilone derivative. In one embodiment, the invention provides host cells that produce one or more of the epothilones or epothilone derivatives at higher levels than produced in the naturally occurring organisms that produce epothilones.
In another embodiment, the invention provides host cells that produce mixtures of epothilones that are less complex than the mixtures produced by naturally occurring host cells. In another embodiment, the present invention provides non-Sorangium recombinant host cells that produce an epothilone or epothilone derivative.
In a preferred embodiment, the host cells of the invention produce less complex mixtures of epothilones than do naturally occurring cells that produce epothilones.
Naturally occurring cells that produce epothilones typically produce a mixture of epothilones A, B, C, D, E, and F. The table below summarizes the epothilones produced in different illustrative host cells of the invention.
Cell Type Epothilones Produced ~othilones Not Produced 1 A, B, C, D, E, F _____ 2 A,C,E B,D,F

3 B,D,F A,C,E

4 A, B, C, D E, F

B, D, E, F

C A, B, D,E,F

B,D A, C,E,F

D A, B, C,E,F

In addition, cell types may be constructed which produce only the newly discovered epothilones G and H, further discussed below, and one or the other of G and H
or both in combination with the downstream epothilones. Thus, it is understood, based on the present invention, that the biosynthetic pathway which relates the naturally occurring epothilones is, respectively, G --~ C --> A --~ E and H -> D --. B -~ F.
Appropriate enzymes may also convert members of each pathway to the corresponding member of the other.
Thus, the recombinant host cells of the invention also include host cells that produce only one desired epothilone or epothilone derivative.
In another embodiment, the invention provides Sorangium host cells that have been modified genetically to produce epothilones either at levels greater than those observed in naturally occurnng host cells or as less complex mixtures of epothilones than produced by naturally occurnng host cells, or produce an epothilone derivative that is not produced in nature. In a preferred embodiment, the host cell produces the epothilones at equal to or greater than 20 mg/L.

In another embodiment, the recombinant host cells of the invention are host cells other than Sorangium cellulosum that have been modified genetically to produce an epothilone or an epothilone derivative. In a preferred embodiment, the host cell produces the epothilones at equal to or greater than 20 mg/L. In a more preferred embodiment, the recombinant host cells are Myxococcus, Pseudomonas, or Streptomyces host cells that produce the epothilones or an epothilone derivative at equal to or greater than 20 mg/L.
In another embodiment, the present invention provides novel compounds useful in agriculture, veterinary practice, and medicine. In one embodiment, the compounds are useful as fungicides. In another embodiment, the compounds are useful in cancer chemotherapy. In a preferred embodiment, the compound is an epothilone derivative that is at least as potent against tumor cells as epothilone B or D. In another embodiment, the compounds are useful as immunosuppressants. In another embodiment, the compounds are useful in the manufacture of another compound. In a preferred embodiment, the compounds are formulated in a mixture or solution for administration to a human or animal.
These and other embodiments of the invention are described in more detail in the following description, the examples, and claims set forth below.
Brief Description of the Figures Figure 1 shows a restriction site map of the insert Sorangium cellulosum genomic DNA in four overlapping cosmid clones (designated 8A3, 1 A2, 4, and 85 and corresponding to pKOS35-70.8A3, pKOS35-70.1A2, pKOS35-70.4, and pKOS35-79.85, respectively) spanning the epothilone gene cluster. A functional map of the epothilone gene cluster is also shown. The loading domain (Loading, epoA), the non-ribosomal peptide synthase (NRPS, Module l, epoB) module, and each module (Modules 2 through 9, epoC, epoD, epoE, and epoF~ of the remaining eight modules of the epothilone synthase gene are shown, as is the location of the epoK gene that encodes a cytochrome P450-like epoxidation enzyme.
Figure 2 shows a number of precursor compounds to N-acylcysteamine thioester derivatives that can be supplied to an epothilone PKS of the invention in which the NRPS-like module 1 or module 2 KS domain has been inactivated to produce a novel epothilone derivative. A general synthetic procedure for making such compounds is also shown.

Figure 3 shows restriction site and function maps of plasmids pKOS35-82.1 and pKOS35-82.2.
Figure 4 shows restriction site and function maps of plasmids pKOS35-154 and pKOS90-22.
Figure 5 shows a schematic of a protocol for introducing the epothilone PKS
and modification enzyme genes into the chromosome of a Myxococcus xanthus host cell as described in Example 3.
Figure 6 shows restriction site and function maps of plasmids pKOS039-124 and pKOS039-1248.
Figure 7 shows a restriction site and function map of plasmid pKOS039-1268.
Figure 8 shows a restriction site and function map of plasmid pKOS039-141.
Figure 9 shows a restriction site and function map of plasmid pKOS045-12.
Detailed Description of the Invention The present invention provides the genes and proteins that synthesize the epothilones in Sorangium cellulosum in recombinant and isolated form. As used herein, the tenor recombinant refers to a compound or composition produced by human intervention, typically by specific and directed manipulation of a gene or portion thereof.
The team isolated refers to a compound or composition in a preparation that is substantially free of contaminating or undesired materials or, with respect to a compound or composition found in nature, substantially free of the materials with which that compound or composition is associated in its natural state. The epothilones (epothilone A, B, C, D, E, and F) and compounds structurally related thereto (epothilone derivatives) are potent cytotoxic agents specific for eukaryotic cells. These compounds have application as anti-fungals, cancer chemotherapeutics, and immunosuppressants. The epothilones are produced at very low levels in the naturally occurring Sorangium cellulosum cells in which they have been identified. Moreover, S. cellulosum is very slow growing, and fermentation of S. cellulosum strains is difficult and time-consuming. One important benefit conferred by the present invention is the ability simply to produce an epothilone or epothilone derivative in a non-S cellulosum host cell. Another advantage of the present invention is the ability to produce the epothilones at higher levels and in greater amounts in the recombinant host cells provided by the invention than possible in the naturally _7_ occurring epothilone producer cells. Yet another advantage is the ability to produce an epothilone derivative in a recombinant host cell.
The isolation of recombinant DNA encoding the epothilone biosynthetic genes resulted from the probing of a genomic library of Sorangium cellulosum SMP44 DNA. As described more fully in Example 1 below, the library was prepared by partially digesting S. cellulosum genomic DNA with restriction enzyme SauIIIAl and inserting the DNA
fragments generated into BamHI-digested SupercosTM cosmid DNA (Stratagene).
Cosmid clones containing epothilone gene sequences were identified by probing with DNA probes specific for sequences from PKS genes and reprobing with secondary probes comprising nucleotide sequences identified with the primary probes.
Four overlapping cosmid clones were identified by this effort. These four cosmids were deposited with the American Type Culture Collection (ATCC), Manassas, VA, USA, under the terms of the Budapest Treaty, and assigned ATCC accession numbers.
The clones (and accession numbers) were designated as cosmids pKOS35-70.1A2 (ATCC
203782), pKOS35-70.4 (ATCC 203781), pKOS35-70.8A3 (ATCC 203783), and pKOS35-79.85 (ATCC 203780). The cosmids contain insert DNA that completely spans the epothilone gene cluster. A restriction site map of these cosmids is shown in Figure 1.
Figure 1 also provides a function map of the epothilone gene cluster, showing the location of the six epothilone PKS genes and the epoK P450 epoxidase gene.
The epothilone PKS genes, like other PKS genes, are composed of coding sequences organized to encode a loading domain, a number of modules, and a thioesterase domain. As described more fully below, each of these domains and modules corresponds to a polypeptide with one or more specific functions. Generally, the loading domain is responsible for binding the first building block used to synthesize the polyketide and transfernng it to the first module. The building blocks used to form complex polyketides are typically acylthioesters, most commonly acetyl, propionyl, malonyl, methylmalonyl, and ethylmalonyl CoA. Other building blocks include amino acid-like acylthioesters.
PKSs catalyze the biosynthesis of polyketides through repeated, decarboxylative Claisen condensations between the acylthioester building blocks. Each module is responsible for binding a building block, performing one or more functions on that building block, and transfernng the resulting compound to the next module. The next module, in turn, is responsible for attaching the next building block and transferring the growing compound to the next module until synthesis is complete. At that point, an enzymatic thioesterase (TE) activity cleaves the polyketide from the PKS.
Such modular organization is characteristic of the class of PKS enzymes that synthesize complex polyketides and is well known in the art. Recombinant methods for manipulating modular PKS genes are described in U.S. Patent Nos. 5,672,491;
5,712,146;
5,830,750; and 5,843,718; and in PCT patent publication Nos. 98/49315 and 97/02358, each of which is incorporated herein by reference. The polyketide known as 6-deoxyerythronolide B (6-dEB) is synthesized by a PKS that is a prototypical modular PKS
enzyme. The genes, known as eryAl, eryAll, and eryAlll, that code for the mufti-subunit protein known as deoxyerythronolide B synthase or DEBS (each subunit is known as DEBS1, DEBS2, or DEBS3) that synthesizes 6-dEB are described in U.S. Patent Nos.
5,712,146 and 5,824,513, incorporated herein by reference.
The loading domain of the DEBS PKS consists of an acyltransferase (AT) and an acyl carrier protein (ACP). The AT of the DEBS loading domain recognizes propionyl CoA (other loading domain ATs can recognize other acyl-CoAs, such as acetyl, malonyl, methylmalonyl, or butyryl CoA) and transfers it as a thioester to the ACP of the loading domain. Concurrently, the AT on each of the six extender modules recognizes a methylmalonyl CoA (other extender module ATs can recognize other CoAs, such as malonyl or alpha-substituted malonyl CoAs, i.e., malonyl, ethylmalonyl, and 2-hydroxymalonyl CoA) and transfers it to the ACP of that module to form a thioester. Once DEBS is primed with acyl- and methylmalonyl-ACPs, the acyl group of the loading domain migrates to form a thioester (trans-esterification) at the KS of the first module; at this stage, module one possesses an acyl-KS adjacent to a methylmalonyl ACP.
The acyl group derived from the DEBS loading domain is then covalently attached to the alpha-carbon of the extender group to form a carbon-carbon bond, driven by concomitant decarboxylation, and generating a new acyl-ACP that has a backbone two carbons longer than the loading unit (elongation or extension). The growing polyketide chain is transferred from the ACP to the KS of the next module of DEBS, and the process continues.
The polyketide chain, growing by two carbons for each module of DEBS, is sequentially passed as a covalently bound thioester from module to module, in an assembly line-like process. The carbon chain produced by this process alone would possess a ketone at every other carbon atom, producing a polyketone, from which the name polyketide arises. Commonly, however, additional enzymatic activities modify the beta keto group of each two carbon unit just after it has been added to the growing polyketide chain but before it is transferred to the next module. Thus, in addition to the minimal module containing KS, AT, and ACP necessary to form the carbon-carbon bond, modules may contain a ketoreductase (KR) that reduces the keto group to an alcohol.
Modules may also contain a KR plus a dehydratase (DH) that dehydrates the alcohol to a double bond. Modules may also contain a KR, a DH, and an enoylreductase (ER) that converts the double bond to a saturated single bond using the beta carbon as a methylene function. The DEBS modules include those with only a KR domain, only an inactive KR
domain, and with all three KR, DH, and ER domains.
Once a polyketide chain traverses the final module of a PKS, it encounters the releasing domain or thioesterase found at the carboxyl end of most PKSs. Here, the polyketide is cleaved from the enzyme and, for most but not all polyketides, cyclized. The polyketide can be modified further by tailoring or modification enzymes; these enzymes add carbohydrate groups or methyl groups, or make other modifications, i.e., oxidation or reduction, on the polyketide core molecule. For example, 6-dEB is hydroxylated, methylated, and glycosylated (glycosidated) to yield the well known antibiotic erythromycin A in the Saccharopolyspora erythraea cells in which it is produced naturally.
While the above description applies generally to modular PKS enzymes and specifically to DEBS, there are a number of variations that exist in nature.
For example, many PKS enzymes comprise loading domains that, unlike the loading domain of DEBS, comprise an "inactive" KS domain that functions as a decarboxylase. This inactive KS is in most instances called KSQ, where the superscript is the single-letter abbreviation for the amino acid (glutamine) that is present instead of the active site cysteine required for ketosynthase activity. The epothilone PKS loading domain contains a KSY domain not present in other PKS enzymes for which amino acid sequence is currently available in which the amino acid tyrosine has replaced the cysteine. The present invention provides recombinant DNA coding sequences for this novel KS domain.
Another important variation in PKS enzymes relates to the type of building block incorporated. Some polyketides, including epothilone, incorporate an amino acid derived building block. PKS enzymes that make such polyketides require specialized modules for incorporation. Such modules are called non-ribosomal peptide synthetase (NRPS) WO 00/31247 - 1 ~ - PCTNS99/27438 modules. The epothilone PKS, for example, contains an NRPS module. Another example of a variation relates to additional activities in a module. For example, one module of the epothilone PKS contains a methyltransferase (MT} domain, a heretofore unknown domain of PKS enzymes that make modular polyketides.
The complete nucleotide sequence of the coding sequence of the open reading frames (ORFs) of the epothilone PKS genes and epothilone tailoring (modification) enzyme genes is provided in Example 1, below. This sequence information together with the information provided below regarding the locations of the open reading frames of the genes within that sequence provides the amino acid sequence of the encoded proteins.
Those of skill in the art will recognize that, due to the degenerate nature of the genetic code, a variety of DNA compounds differing in their nucleotide sequences can be used to encode a given amino acid sequence of the invention. The native DNA sequence encoding the epothilone PKS and epothilone modification enzymes of Sorangium cellulosum is shown herein merely to illustrate a preferred embodiment of the invention. The present invention includes DNA compounds of any sequence that encode the amino acid sequences of the polypeptides and proteins of the invention. In similar fashion, a poIypeptide can typically tolerate one or more amino acid substitutions, deletions, and insertions in its amino acid sequence without loss or significant loss of a desired activity and, in some instances, even an improvement of a desired activity. The present invention includes such polypeptides with alternate amino acid sequences, and the amino acid sequences shown merely illustrate preferred embodiments of the invention.
T7he present invention provides recombinant genes for the production of epothilones. The invention is exemplified by the cloning, characterization, and manipulation of the epothilone PKS and modification enzymes of Sorangium cellulosum SMP44. The description of the invention and the recombinant vectors deposited in connection with that description enable the identification, cloning, and manipulation of epothilone PKS and modification enzymes from any naturally occurring host cell that produces an epothilone. Such host cells include other S. cellulosum strains, such as So ce 90, other Sorangium species, and non-Sorangium cells. Such identification, cloning, and characterization can be conducted by those of ordinary skill in accordance with the present invention using standard methodology for identifying homologous DNA sequences and for identifying genes that encode a protein of function similar to a known protein.
Moreover, the present invention provides recombinant epothilone PKS and modification enzyme genes that are synthesized de novo or are assembled from non-epothilone PKS
genes to provide an ordered array of domains and modules in one or more proteins that assemble to form a PKS that produces epothilone or an epothilone derivative.
The recombinant nucleic acids, proteins, and peptides of the invention are many S and diverse. To facilitate an understanding of the invention and the diverse compounds and methods provided thereby, the following discussion describes various regions of the epothilone PKS and corresponding coding sequences. This discussion begins with a general discussion of the genes that encode the PKS, the location of the various domains and modules in those genes, and the location of the various domains in those modules.
Then, a more detailed discussion follows, focusing first on the loading domain, followed by the NRPS module, and then the remaining eight modules of the epothilone PKS.
There are six epothilone PKS genes. The epoA gene encodes the 149 kDa loading domain (which can also be referred to as a loading module). The epoB gene encodes module l, the 158 kDa NRPS module. The epoC gene encodes the 193 kDa module 2.
The epoD gene encodes a 765 kDa protein that comprises modules 3 through 6, inclusive. The epoE gene encodes a 405 kDa protein that comprises modules 7 and 8. The epoF
gene encodes a 257 kDa protein that comprises module 9 and the thioesterase domain.
Immediately downstream of the epoF gene is epoK, the P450 epoxidase gene which encodes a 47 kDa protein, followed immediately by the epoL gene, which may encode a 24 kDa dehydratase. The epoL gene is followed by a number of ORFs that include genes believed to encode proteins involved in transport and regulation.
The sequences of these genes are shown in Example 1 in one contiguous sequence or contig of 71,989 nucleotides. This contig also contains two genes that appear to originate from a transposon and are identified below as ORF A and ORF B. These two genes are believed not to be involved in epothilone biosynthesis but could possibly contain sequences that function as a promoter or enhancer. The contig also contains more than 12 additional ORFs, only 12 of which, designated ORF2 through ORF 12 and ORF2 complement, are identified below. As noted, ORF2 actually is two ORFs, because the complement of the strand shown also comprises an ORF. The function of the corresponding gene product, if any, of these ORFs has not yet been established. The Table below provides the location of various open reading frames, module-coding sequences, and domain encoding sequences within the contig sequence shown in Example 1.
Those of skill in the art will recognize, upon consideration of the sequence shown in Example 1, that the actual start locations of several of the genes could differ from the start locations shown in the table, because of the presence in frame codons for methionine or valine in close proximity to the codon indicated as the start codon. The actual start codon can be confirmed by amino acid sequencing of the proteins expressed from the genes.
Start Stop Comment 3 992 transposase gene ORF A, not part of the PKS

989 1501 transposase gene ORF B, not part of the PKS

1998 6263 epoA gene, encodes the loading domain 2031 3548 KSY of the loading domain 3621 4661 AT of the loading domain 4917 5810 ER of the loading domain, potentially involved in formation of the thiazole moiety 5856 6155 ACP of the loading domain 6260 10493 epoB gene, encodes module 1, the NRPS module 6620 6649 condensation domain C2 of the NRPS
module 6861 6887 heterocyclization signature sequence 6962 6982 condensation domain C4 of the NRPS
module 7358 7366 condensation domain C7 (partial) of the NRPS

module 7898 7921 adenylation domain A 1 of the NRPS
module 8261 8308 adenylation domain A3 of the NRPS
module 8411 8422 adenylation domain A4 of the NRPS
module 8861 8905 adenylation domain A6 of the NRPS
module 8966 8983 adenylation domain A7 of the NRPS
module 9090 9179 adenylation domain A8 of the NRPS
module 9183 9992 oxidation region for forming thiazole 10121 10138 Adenylation domain A10 of the NRPS
module 10261 10306 Thiolation domain (PCP) of the NRPS
module 10639 16137 epoC gene, encodes module 2 10654 12033 KS2, the KS domain of module 2 12250 13287 AT2, the AT domain of module 2 13327 13899 DH2, the DH domain of module 2 14962 15756 KR2, the KR domain of module 2 15763 16008 ACP2, the ACP domain of module 2 16134 37907 epoD gene, encodes modules 3-6 WO 00/31247 _ 13 _ PCT/US99/27438 Start Stop Comment 37912 49308 epoE gene, encodes modules 7 and 45262 45810 DH (inactive) 46072 47172 MTB, the rnethyltransferase domain of module 8 48103 48636 KR8, this domain is inactive 49323 56642 epoF gene, encodes module 9 and the TE domain 52173 53414 DH (inactive) 54747 SS3~13 KR9 SS878 56600 TE9, the thioesterase domain 56757 58016 epoK gene, encodes the P450 epoxidase Start Stop Comment 58194 58733 epoL gene (putative dehydratase) 59405 59974 ORF2 complement, complement of strand shown 60271 60738 ORF3, complement of strand shown 61730 62647 ORF4 (putative transporter) 66237 67472 ORF7 (putative oxidoreductase) 67572 68837 ORF8 (putative oxidoreductase membrane subunit) 69993 71174 ORF 10 (putative transporter) With this overview of the organization and sequence of the epothilone gene cluster, one can better appreciate the many different recombinant DNA compounds provided by the present invention.
The epothilone PKS is multiprotein complex composed of the gene products of the epoA, epoB, epoC, epoD, epoE, and epoF genes. To confer the ability to produce epothilones to a host cell, one provides the host cell with the recombinant epoA, epoB, epoC, epoD, epoE, and epoF genes of the present invention, and optionally other genes, capable of expression in that host cell. Those of skill in the art will appreciate that, while the epothilone and other PKS enzymes may be referred to as a single entity herein, these enzymes are typically multisubunit proteins. Thus, one can make a derivative PKS (a PKS
that differs from a naturally occurring PKS by deletion or mutation) or hybrid PKS (a PKS
that is composed of portions of two different PKS enzymes) by altering one or more genes that encode one or more of the multiple proteins that constitute the PKS.
The post-PKS modification or tailoring of epothilone includes multiple steps mediated by multiple enzymes. These enzymes are referred to herein as tailoring or modification enzymes. Surprisingly, the products of the domains of the epothilone PKS
predicted to be functional by analysis of the genes that encode them are compounds that have not been previously reported. These compounds are referred to herein as epothilones G and H. Epothilones G and H lack the C-12-C-13 ~-bond of epothilones C and D
and the C-12-C-13 epoxide of epothilones A and B, having instead a hydrogen and hydroxyl group at C-13, a single bond between C-12 and C-13, and a hydrogen and H or methyl group at C-12. These compounds are predicted to result from the epothilone PKS, because the DNA and corresponding amino acid sequence for module 4 of the epothilone PKS
does not appear to include a DH domain.
As described below, however, expression of the epothilone PKS genes epoA, epoB, epoC, epoD, epoE, and epoF in certain heterologous host cells that do not express epoK or epoL leads to the production of epothilones C and D, which lack the C-13 hydroxyl and have a double bond between C-12 and C-13. The dehydration reaction that mediates the formation of this double bond may be due to the action of an as yet unrecognized domain of the epothilone PKS (for example, dehydration could occur in the next module, which possesses an active DH domain and could generate a conjugated dime precursor prior to its dehydrogenation by an ER domain) or an endogenous enzyme in the heterologous host cells (Streptomyces coelicolor) in which it was observed. In the latter event, epothilones G
and H may be produced in Sorangium cellulosum or other host cells and, to be converted 1 S to epothilones C and D, by the action of a dehydratase, which may be encoded by the epoL
gene. In any event, epothilones C and D are converted to epothilones A and B
by an epoxidase encoded by the epoK gene. Epothilones A and B are converted to epothilones E
and F by a hydroxylase gene, which may be encoded by one of the ORFs identified above or by another gene endogenous to Sorangium cellulosum. Thus, one can produce an epothilone or epothilone derivative modified as desired in a host cell by providing that host cell with one or more of the recombinant modification enzyme genes provided by the invention or by utilizing a host cell that naturally expresses (or does not express) the modification enzyme. 'Thus, in general, by utilizing the appropriate host and by appropriate inactivation, if desired, of modification enzymes, one may interrupt the progression of G -> C -~ A --~ E or the corresponding downstream processing of epothilone H at any desired point; by controlling methylation, one or both of the pathways can be selected.
Thus, the present invention provides a wide variety of recombinant DNA
compounds and host cells for expressing the naturally occurring epothilones A, B, C, and D and derivatives thereof. The invention also provides recombinant host cells, particularly Sorangium cellulosum host cells that produce epothilone derivatives modified in a manner similar to epothilones E and F. Moreover, the invention provides host cells that can produce the heretofore unknown epothilones G and H, either by expression of the epothilone PKS genes in host cells that do not express the dehydratase that converts epothilones G and H to C and D or by mutating or altering the PKS to abolish the dehydratase function, if it is present in the epothilone PKS.
The macrolide compounds that are products of the PKS cluster can thus be modified in various ways. In addition to the modifications described above, the PKS
products can be glycosylated, hydroxylated, dehydroxylated, oxidized, methylated and demethylated using appropriate enzymes. Thus, in addition to modifying the product of the PKS cluster by altering the number, functionality, or specificity of the modules contained in the PKS, additional compounds within the scope of the invention can be produced by additional enzyme-catalyzed activity either provided by a host cell in which the polyketide synthases are produced or by modifying these cells to contain additional enzymes or by additional in vitro modification using purified enzymes or crude extracts or, indeed, by chemical modification.
The present invention also provides a wide variety of recombinant DNA
1 S compounds and host cells that make epothilone derivatives. As used herein, the phrase "epothilone derivative" refers to a compound that is produced by a recombinant epothilone PKS in which at Least one domain has been either rendered inactive, mutated to alter its catalytic function, or replaced by a domain with a different function or in which a domain has been inserted. In any event, the "epothilone derivative PKS" functions to produce a compound that differs in structure from a naturally occurring epothilone but retains its ring backbone structure and so is called an "epothilone derivative." To faciliate a better understanding of the recombinant DNA compounds and host cells provided by the invention, a detailed discussion of the loading domain and each of the modules of the epothilone PKS, as well as novel recombinant derivatives thereof, is provided below.
The loading domain of the epothilone PKS includes an inactive KS domain, KSY, an AT domain specific for malonyI CoA (which is believed to be decarboxylated by the KSY domain to yield an acetyl group), and an ACP domain. The present invention provides recombinant DNA compounds that encode the epothilone loading domain.
The loading domain coding sequence is contained within an ~8.3 kb EcoRI
restriction fragment of cosmid pKOS35-70.8A3. The KS domain is referred to as inactive, because the active site region "TAYSSSL" of the KS domain of the loading domain has a Y
residue in place of the cysteine required for ketosynthase activity; this domain does have WO 00/31247 - 1 ~ - PCT/US99/27438 decarboxylase activity. See Witkowski et al., 7 Sep. 1999, Biochem. 38(36):

11650, incorporated herein by reference.
The presence of the Y residue in place of a Q residue (which occurs typically in an inactive loading domain KS) may make the KS domain less efficient at decarboxylation.
The present invention provides a recombinant epothilone PKS loading domain and corresponding DNA sequences that encode an epothilone PKS loading domain in which the Y residue has been changed to a Q residue by changing the codon therefor in the coding sequence of the loading domain. The present invention also provides recombinant PKS enzymes comprising such loading domains and host cells for producing such enzymes and the polyketides produced thereby. These recombinant loading domains include those in which just the Y residue has been changed, those in which amino acids surrounding and including the Y domain have been changed, and those in which the complete KSY domain has been replaced by a complete KSQ domain. The latter embodiment includes but is not limited to a recombinant epothilone loading domain in which the KSY domain has been replaced by the KSQ domain of the oleandolide PKS or the narbonolide PKS (see the references cited below in connection with the oleandomycin, narbomycin, and picromycin PKS and modification enzymes).
The epothilone loading domain also contains an AT domain believed to bind malonyl CoA. The sequence "QTAFTQPALFTFEYALAALW. . . GHSIG" in the AT
domain is consistent with malonyl CoA specificity. As noted above, the malonyl CoA is believed to be decarboxylated by the KSY domain to yield acetyl CoA. The present invention provides recombinant epothilone derivative loading domains or their encoding DNA sequences in which the malonyl specific AT domain or its encoding sequence has been changed to another specificity, such as methylmalonyl CoA, ethylmalonyl CoA, and 2-hydroxymalonyl CoA. When expressed with the other proteins of the epothilone PKS, such loading domains lead to the production of epothilones in which the methyl substituent of the thiazole ring of epothilone is replaced with, respectively, ethyl, propyl, and hydroxymethyl. The present invention provides recombinant PKS enzymes comprising such loading domains and host cells for producing such enzymes and the polyketides produced thereby.
Those of skill in the art will recognize that an AT domain that is specific for 2-hydroxymalonyl CoA will result in a polyketide with a hydroxyl group at the corresponding location in the polyketide produced, and that the hydroxyl group can be methylated to yield a methoxy group by polyketide modification enzymes. See, e.g., the patent applications cited in connection with the FK-520 PKS in the table below.
Consequently, reference to a PKS that has a 2-hydroxymalonyl specific AT
domain herein similarly refers to polyketides produced by that PKS that have either a hydroxyl or methoxyl group at the corresponding location in the polyketide.
The loading domain of the epothilone PKS also comprises an ER domain. While, this ER domain may be involved in forming one of the double bonds in the thiazole moiety in epothilone (in the reverse of its normal reaction), or it may be non-functional. In either event, the invention provides recombinant DNA compounds that encode the epothilone PKS loading domain with and without the ER region, as well as hybrid loading domains that contain an ER domain from another PKS (either active or inactive, with or without accompanying KR and DH domains) in place of the ER domain of the epothilone loading domain. The present invention also provides recombinant PKS enzymes comprising such loading domains and host cells for producing such enzymes and the polyketides produced thereby.
The recombinant nucleic acid compounds of the invention that encode the loading domain of the epothilone PKS and the corresponding polypeptides encoded thereby are useful for a variety of applications. In one embodiment, a DNA compound comprising a sequence that encodes the epothilone loading domain is coexpressed with the proteins of a heterologous PKS. As used herein, reference to a heterologous modular PKS (or to the coding sequence therefor) refers to all or part of a PKS, including each of the multiple proteins constituting the PKS, that synthesizes a polyketide other than an epothilone or epothilone derivative (or to the coding sequences therefor). This coexpression can be in one of two forms. The epothilone loading domain can be coexpressed as a discrete protein with the other proteins of the heterologous PKS or as a fusion protein in which the loading domain is fused to one or more modules of the heterologous PKS. In either event, the hybrid PKS formed, in which the loading domain of the heterologous PKS is replaced by the epothilone loading domain, provides a novel PKS. Examples of a heterologous PKS
that can be used to prepare such hybrid PKS enzymes of the invention include but are not limited to DEBS and the picromycin (narbonolide), oleandolide, rapamycin, FK-506, FK-520, rifamycin, and avermectin PKS enzymes and their corresponding coding sequences.
In another embodiment, a nucleic acid compound comprising a sequence that encodes the epothilone loading domain is coexpressed with the proteins that constitute the remainder of the epothilone PKS (i.e., the epoB, epoC, epoD, epoE, and epoF
gene products) or a recombinant epothilone PKS that produces an epothilone derivative due to an alteration or mutation in one or more of the epoB, epoC, epoD, epoE, and epoF genes.
As used herein, reference to an epothilone or a PKS that produces an epothilone derivative (or to the coding sequence therefor) refers to all or any one of the proteins that comprise the PKS (or to the coding sequences therefor).
In another embodiment, the invention provides recombinant nucleic acid compounds that encode a loading domain composed of part of the epothilone loading domain and part of a heterologous PKS. In this embodiment, the invention provides, for example, either replacing the malonyl CoA specific AT with a methylmalonyl CoA, ethylmalonyl CoA, or 2-hydroxymalonyl CoA specific AT. This replacement, like the others described herein, is typically mediated by replacing the coding sequences therefor to provide a recombinant DNA compound of the invention; the recombinant DNA is used to prepare the corresponding protein. Such changes (including not only replacements but also deletions and insertions) may be referred to herein either at the DNA or protein level.
The compounds of the invention also include those in which both the KSY and AT
domains of the epothilone loading domain have been replaced but the ACP and/or linker regions of the epothilone loading domain are left intact. Linker regions are those segments of amino acids between domains in the loading domain and modules of a PKS that help form the tertiary structure of the protein and are involved in correct alignment and positioning of the domains of a PKS. These compounds include, for example, a recombinant loading domain coding sequence in which the KSY and AT domain coding sequences of the epothilone PKS have been replaced by the coding sequences for the KSQ
and AT domains of, for example, the oleandolide PKS or the narbonolide PKS.
There are also PKS enzymes that do not employ a KSQ domain but instead merely utilize an AT
domain that binds acetyl CoA, propionyl CoA, or butyryl CoA (the DEBS loading domain) or isobutyryl CoA (the avermectin loading domain). Thus, the compounds of the invention also include, for example, a recombinant loading domain coding sequence in which the KSY and AT domain coding sequences of the epothilone PKS have been replaced by an AT domain of the DEBS or avermectin PKS. The present invention also provides recombinant DNA compounds encoding loading domains in which the ACP
domain or any of the linker regions of the epothilone loading domain has been replaced by another ACP or linker region.

WO 00/31247 - 2~ - PCT/US99/27438 Any of the above loading domain coding sequences is coexpressed with the other proteins that constitute a PKS that synthesizes epothilone, an epothilone derivative, or another polyketide to provide a PKS of the invention. If the product desired is epothilone or an epothilone derivative, then the loading domain coding sequence is typically expressed as a discrete protein, as is the loading domain in the naturally occurring epothilone PKS. If the product desired is produced by the loading domain of the invention and proteins from one or more non-epothilone PKS enzymes, then the loading domain is expressed either as a discrete protein or as a fusion protein with one or more modules of the heterologous PKS.
The present invention also provides hybrid PKS enzymes in which the epothilone loading domain has been replaced in its entirety by a loading domain from a heterologous PKS with the remainder of the PKS proteins provided by modified or unmodified epothilone PKS proteins. The present invention also provides recombinant expression vectors and host cells for producing such enzymes and the polyketides produced thereby.
In one embodiment, the heterologous loading domain is expressed as a discrete protein in a host cell that expresses the epoB, epoC, epoD, epoE, and epoF gene products.
In another embodiment, the heterologous loading domain is expressed as a fusion protein with the epoB gene product in a host cell that expresses the epoC, epoD, epoE, and epoF
gene products. In a related embodiment, the present invention provides recombinant epothilone PKS enzymes in which the loading domain has been deleted and replaced by an NRPS
module and corresponding recombinant DNA compounds and expression vectors. In this embodiment, the recombinant PKS enzymes thus produce an epothilone derivative that comprises a dipeptide moiety, as in the compound leinamycin. The invention provides such enzymes in which the remainder of the epothilone PKS is identical in function to the native epothilone PKS as well as those in which the remainder is a recombinant PKS that produces an epothilone derivative of the invention.
The present invention also provides reagents and methods useful in deleting the loading domain coding sequence or any portion thereof from the chromosome of a host cell, such as Sorangium cellulosum, or replacing those sequences or any portion thereof with sequences encoding a recombinant loading domain. Using a recombinant vector that comprises DNA complementary to the DNA including and/or flanking the loading domain coding sequence in the Sorangium chromosome, one can employ the vector and homologous recombination to replace the native loading domain coding sequence with a recombinant loading domain coding sequence or to delete the sequence altogether.
Moreover, while the above discussion focuses on deleting or replacing the epothilone loading domain coding sequences, those of skill in the art will recognize that the present invention provides recombinant DNA compounds, vectors, and methods useful in deleting or replacing all or any portion of an epothilone PKS gene or an epothilone modification enzyme gene. Such methods and materials are useful for a variety of purposes. One purpose is to construct a host cell that does not make a naturally occurring epothilone or epothilone derivative. For example, a host cell that has been modified to not produce a naturally occurring epothilone may be particularly preferred for making epothilone derivatives or other polyketides free of any naturally occurring epothilone.
Another purpose is to replace the deleted gene with a gene that has been altered so as to provide a different product or to produce more of one product than another.
If the epothilone loading domain coding sequence has been deleted or otherwise rendered non-functional in a Sorangium cellulosum host cell, then the resulting host cell will produce a non-functional epothilone PKS. This PKS could still bind and process extender units, but the thiazole moiety of epothilone would not form, leading to the production of a novel epothilone derivative. Because this derivative would predictably contain a free amino group, it would be produced at most in low quantities. As noted above, however, provision of a heterologous or other recombinant loading domain to the host cell would result in the production of an epothilone derivative with a structure determined by the loading domain provided.
The loading domain of the epothilone PKS is followed by the first module of the PKS, which is an NRPS module specific for cysteine. This NRPS module is naturally expressed as a discrete protein, the product of the epoB gene. The present invention provides the epoB gene in recombinant form. The recombinant nucleic acid compounds of the invention that encode the NRPS module of the epothilone PKS and the corresponding poiypeptides encoded thereby are useful for a variety of applications. In one embodiment, a nucleic acid compound comprising a sequence that encodes the epothilone NR.PS
module is coexpressed with genes encoding one or more proteins of a heterologous PKS.
The NRPS module can be expressed as a discrete protein or as a fusion protein with one of the proteins of the heterologous PKS. The resulting PKS, in which at least a module of the heterologous PKS is replaced by the epothilone NRPS module or the NRPS module is in WO 00/31247 _ 22 _ PCT/US99/27438 effect added as a module to the heterologous PKS, provides a novel PKS. In another embodiment, a DNA compound comprising a sequence that encodes the epothilone NRPS
module is coexpressed with the other epothilone PKS proteins or modified versions thereof to provide a recombinant epothilone PKS that produces an epothilone or an epothilone derivative.
Two hybrid PKS enzymes provided by the invention illustrate this aspect. Both hybrid PKS enzymes are hybrids of DEBS and the epothilone NRPS module. The first hybrid PKS is composed of four proteins: (i) DEBS1; (ii) a fusion protein composed of the KS domain of module 3 of DEBS and all but the KS domain of the loading domain of the epothilone PKS; (iii) the epothilone NRPS module; and (iv) a fusion protein composed of the KS domain of module 2 of the epothilone PKS fused to the AT domain of module 5 of DEBS and the rest of DEBS3. This hybrid PKS produces a novel polyketide with a thiazole moiety incorporated into the macrolactone ring and a molecular weight of 413.53 when expressed in Streptomyces coelicolor. Glycosylated, hydroxylated, and methylated derivatives can be produced by expression of the hybrid PKS in Saccharopolyspora erythraea.
Diagrammatically, the construct is represented:
-~- DEBS epo DEBS ->
DEBS1 KS3 ATo NRPS KS2 AT5 The structure of the product is:
O
The second hybrid PKS illustrating this aspect of the invention is composed of five proteins: (i) DEBS 1; (ii) a fusion protein composed of the KS domain of module 3 of DEBS and all but the KS domain of the loading domain of the epothilone PKS;
(iii) the epothilone NRPS module; and (iv) a fusion protein composed of the KS domain of module WO 00/31247 _ 23 _ PCT/US99/27438 2 of the epothilone PKS fused to the AT domain of module 4 of DEBS and the rest of DEBS2; and (v) DEBS3. This hybrid PKS produces a novel polyketide with a thiazole moiety incorporated into the macrolactone ring and a molecular weight of 455.61 when expressed in Streptomyces coelicolor. Glycosylated, hydroxylated, and methylated derivatives can be produced by expression of the hybrid PKS in Saccharopolyspora erythraea.
Diagrammatically, the construct is represented:
DEBS epo DEBS
DEBS1 KS3 ATo NRPS KS2 AT4 KS2 AT5 The structure of the product is:
H( H
In another embodiment, a portion of the NRPS module coding sequence is utilized in conjunction with a heterologous coding sequence. In this embodiment, the invention provides, for example, changing the specificity of the NRPS module of the epothilone PKS from a cysteine to another amino acid. This change is accomplished by constructing a coding sequence in which all or a portion of the epothilone PKS NRPS module coding sequences have been replaced by those coding for an NRPS module of a different specificity. In one illustrative embodiment, the specificity of the epothilone NRPS module is changed from cysteine to serine or threonine. When the thus modified NRPS
module is expressed with the other proteins of the epothilone PKS, the recombinant PKS
produces an epothilone derivative in which the thiazole moiety of epothilone (or an epothilone derivative) is changed to an oxazole or 5-methyloxazole moiety, respectively.
Alternatively, the present invention provides recombinant PKS enzymes composed of the products of the epoA, epoC, epoD, epoE, and epoF genes (or modified versions thereof) without an NRPS module or with an NRPS module from a heterologous PKS. The heterologous NRPS module can be expressed as a discrete protein or as a fusion protein with either the epoA or epoC genes.

The invention also provides methods and reagents useful in changing the specificity of a heterologous NRPS module from another amino acid to cysteine.
This change is accomplished by constructing a coding sequence in which the sequences that determine the specificity of the heterologous NRPS module have been replaced by those that specify cysteine from the epothilone NRPS module coding sequence. The resulting heterologous NRPS module is typically coexpressed in conjunction with the proteins constituting a heterologous PKS that synthesizes a polyketide other than epothilone or an epothilone derivative, although the heterologous NRPS module can also be used to produce epothilone or an epothilone derivative.
In another embodiment, the invention provides recombinant epothilone PKS
enzymes and corresponding recombinant nucleic acid compounds and vectors in which the NRPS module has been inactivated or deleted. Such enzymes, compounds, and vectors are constructed generally in accordance with the teaching for deleting or inactivating the epothilone PKS or modification enzyme genes above. Inactive NRPS module proteins and the coding sequences therefore provided by the invention include those in which the peptidyl carrier protein (PCP) domain has been wholly or partially deleted or otherwise rendered inactive by changing the active site serine (the site for phosphopantetheinylation) to another amino acid, such as alanine, or the adenylation domains have been deleted or otherwise rendered inactive. In one embodiment, both the loading domain and the NRPS
have been deleted or rendered inactive. In any event, the resulting epothilone PKS can then function only if provided a substrate that binds to the KS domain of module 2 (or a subsequent module) of the epothilone PKS or a PKS for an epothilone derivative. In a method provided by the invention, the thus modified cells are then fed activated acylthioesters that are bound by preferably the second, but potentially any subsequent, module and processed into novel epothilone derivatives.
Thus, in one embodiment, the invention provides Sorangium and non-Sorangium host cells that express an epothilone PKS (or a PKS that produces an epothilone derivative) with an inactive NRPS. The host cell is fed activated acylthioesters to produce novel epothilone derivatives of the invention. The host cells expressing, or cell free extracts containing, the PKS can be fed or supplied with N-acylcysteamine thioesters (MACS) of novel precursor molecules to prepare epothilone derivatives. See U.S. patent application Serial No. 60/117,384, filed 27 Jan. 1999, and PCT patent publication No.
US99/03986, both of which are incorporated herein by reference, and Example 6, below.

WO 00131247 - 25 - PC1'/US99/27438 The second (first non-NRPS) module of the epothilone PKS includes a KS, an AT
specific for methylmalonyl CoA, a DH, a KR, and an ACP. This module is encoded by a sequence within an 13.1 kb EcoRI-NsiI restriction fragment of cosmid pKOS35-70.8A3.
The recombinant nucleic acid compounds of the invention that encode the second module of the epothilone PKS and the corresponding polypeptides encoded thereby are useful for a variety of applications. The second module of the epothilone PKS
is produced as a discrete protein by the epoC gene. The present invention provides the epoC gene in recombinant form. In one embodiment, a DNA compound comprising a sequence that encodes the epothilone second module is coexpressed with the proteins constituting a heterologous PKS either as a discrete protein or as a fusion protein with one or more modules of the heterologous PKS. The resulting PKS, in which a module of the heterologous PKS is either replaced by the second module of the epothilone PKS
or the latter is merely added to the modules of the heterologous PKS, provides a novel PKS. In another embodiment, a DNA compound comprising a sequence that encodes the second module of the epothilone PKS is coexpressed with the other proteins constituting the epothilone PKS or a recombinant epothilone PKS that produces an epothilone derivative.
In another embodiment, all or only a portion of the second module coding sequence is utilized in conjunction with other PKS coding sequences to create a hybrid module. In this embodiment, the invention provides, for example, either replacing the methylmalonyl CoA specific AT with a malonyl CoA, ethylmalonyl CoA, or 2-hydroxymalonyl CoA specific AT; deleting either the DH or KR or both;
replacing the DH
or KR or both with a DH or KR or both that specify a different stereochemistry; and/or inserting an ER. Generally, any reference herein to inserting or replacing a PKS KR, DH, and/or ER domain includes the replacement of the associated KR, DH, or ER
domains in that module, typically with corresponding domains from the module from which the inserted or replacing domain is obtained. In addition, the KS and/or ACP can be replaced with another KS and/or ACP. In each of these replacements or insertions, the heterologous KS, AT, DH, KR, ER, or ACP coding sequence can originate from a coding sequence for another module of the epothilone PKS, from a gene for a PKS that produces a polyketide other than epothilone, or from chemical synthesis. The resulting heterologous second module coding sequence can be coexpressed with the other proteins that constitute a PKS
that synthesizes epothilone, an epothilone derivative, or another polyketide.
Alternatively, one can delete or replace the second module of the epothilone PKS with a module from a heterologous PKS, which can be expressed as a discrete protein or as a fusion protein fused to either the epoB or epoD gene product.
Illustrative recombinant PKS genes of the invention include those in which the AT
domain encoding sequences for the second module of the epothilone PKS have been altered or replaced to change the AT domain encoded thereby from a methylmalonyl specific AT to a malonyl specific AT. Such malonyl specific AT domain encoding nucleic acids can be isolated, for example and without limitation, from the PKS genes encoding the narbonolide PKS, the rapamycin PKS (i.e., modules 2 and 12), and the FK-(i.e., modules 3, 7, and 8). When such a hybrid second module is coexpressed with the other proteins constituting the epothilone PKS, the resulting epothilone derivative produced is a 16-desmethyl epothilone derivative.
In addition, the invention provides DNA compounds and vectors encoding recombinant epothilone PKS enzymes and the corresponding recombinant proteins in which the KS domain of the second (or subsequent) module has been inactivated or deleted. In a preferred embodiment, this inactivation is accomplished by changing the codon for the active site cysteine to an alanine codon. As with the corresponding variants described above for the NRPS module, the resulting recombinant epothilone PKS
enzymes are unable to produce an epothilone or epothilone derivative unless supplied a precursor that can be bound and extended by the remaining domains and modules of the recombinant PKS enzyme. Illustrative diketides are described in Example 6, below.
The third module of the epothilone PKS includes a KS, an AT specific for malonyl CoA, a KR, and an ACP. This module is encoded by a sequence within an ~8 kb BgIII-NsiI restriction fragment of cosmid pKOS35-70.8A3.
The recombinant DNA compounds of the invention that encode the third module of the epothilone PKS and the corresponding polypeptides encoded thereby are useful for a variety of applications. The third module of the epothilone PKS is expressed in a protein, the product of the epoD gene, which also contains modules 4, 5, and 6. The present invention provides the epoD gene in recombinant form. The present invention also provides recombinant DNA compounds that encode each of the epothilone PKS
modules 3, 4, 5, and 6, as discrete coding sequences without coding sequences for the other epothilone modules. In one embodiment, a DNA compound comprising a sequence that encodes the epothilone third module is coexpressed with proteins constituting a heterologous PKS. The third module of the epothilone PKS can be expressed either as a WO 00/31247 - 2~ - PCT/US99/27438 discrete protein or as a fusion protein fused to one or more modules of the heterologous PKS. T'he resulting PKS, in which a module of the heterologous PKS is either replaced by that for the third module of the epothilone PKS or the latter is merely added to the modules of the heterologous PKS, provides a novel PKS. In another embodiment, a DNA
compound comprising a sequence that encodes the third module of the epothilone PKS is coexpressed with proteins comprising the remainder of the epothilone PKS or a recombinant epothilone PKS that produces an epothilone derivative, typically as a protein comprising not only the third but also the fourth, fifth, and sixth modules.
In another embodiment, all or a portion of tlae third module coding sequence is utilized in conjunction with other PKS coding sequences to create a hybrid module. In this embodiment, the invention provides, for example, either replacing the malonyl CoA
specific AT with a methylmalonyl CoA, ethylmalonyl CoA, or 2-hydroxymalonyl CoA
specific AT; deleting the KR; replacing the KR with a KR that specifies a different stereochemistry; and/or inserting a DH or a DH and an ER. As above, the reference to inserting a DH or a DH and an ER includes the replacement of the KR with a DH
and KR
or an ER, DH, and KR. In addition, the KS and/or ACP can be replaced with another KS
and/or ACP. In each of these replacements or insertions, the heterologous KS, AT, DH, KR, ER, or ACP coding sequence can originate from a coding sequence for another module of the epothilone PKS, from a coding sequence for a PKS that produces a polyketide other than epothilone, or from chemical synthesis. The resulting heterologous third module coding sequence can be utilized in conjunction with a coding sequence for a PKS that synthesizes epothilone, an epothilone derivative, or another polyketide.
Illustrative recombinant PKS genes of the invention include those in which the AT
domain encoding sequences for the third module of the epothilone PKS have been altered or replaced to change the AT domain encoded thereby from a malonyl specific AT
to a methylmalonyl specific AT. Such methylmalonyl specific AT domain encoding nucleic acids can be isolated, for example and without limitation, from the PKS genes encoding DEBS, the narbonolide PKS, the rapamycin PKS, and the FK-520 PKS. When coexpressed with the remaining modules and proteins of the epothilone PKS or an epothilone PKS derivative, the recombinant PKS produces the 14-methyl epothilone derivatives of the invention.
Those of skill in the art will recognize that the KR domain of the third module of the PKS is responsible for forming the hydroxyl group involved in cyclization of WO 00/31247 - 2g - PCT/US99/27438 epothilone. Consequently, abolishing the KR domain of the third module or adding a DH
or DH and ER domains will interfere with the cyclization, leading either to a linear molecule or to a molecule cyclized at a different location than is epothiIone.
The fourth module of the epothilone PKS includes a KS, an AT that can bind either malonyl CoA or methylmalonyl CoA, a KR, and an ACP. This module is encoded by a sequence within an ~10 kb NsiI-HindIII restriction fragment of cosmid pKOS35-70.1A2.
The recombinant DNA compounds of the invention that encode the fourth module of the epothilone PKS and the corresponding polypeptides encoded thereby are useful for a variety of applications. In one embodiment, a DNA compound comprising a sequence that encodes the epothilone fourth module is inserted into a DNA compound that comprises the coding sequence for one or more modules of a heterologous PKS.
The resulting construct encodes a protein in which a module of the heterologous PKS is either replaced by that for the fourth module of the epothilone PKS or the latter is merely added to the modules of the heterologous PKS. Together with other proteins that constitute the 1 S heterologous PKS, this protein provides a novel PKS. In another embodiment, a DNA
compound comprising a sequence that encodes the fourth module of the epothilone PKS is expressed in a host cell that also expresses the remaining modules and proteins of the epothilone PKS or a recombinant epothilone PKS that produces an epothilone derivative.
For making epothilone or epothilone derivatives, the recombinant fourth module is usually expressed in a protein that also contains the epothilone third, fifth, and sixth modules or modified versions thereof.
In another embodiment, all or a portion of the fourth module coding sequence is utilized in conjunction with other PKS coding sequences to create a hybrid module. In this embodiment, the invention provides, for example, either replacing the malonyl CoA and methylmalonyl specific AT with a malonyl CoA, methylmalonyl CoA, ethylmalonyl CoA, or 2-hydroxymalonyl CoA specific AT; deleting the KR; and/or replacing the KR, including, optionally, to specify a different stereochemistry; and/or inserting a DH or a DH
and ER. In addition, the KS and/or ACP can be replaced with another KS and/or ACP. In each of these replacements or insertions, the heterologous KS, AT, DH, KR, ER, or ACP
coding sequence can originate from a coding sequence for another module of the epothilone PKS, from a gene for a PKS that produces a polyketide other than epothilone, or from chemical synthesis. The resulting heterologous fourth module coding sequence is incorporated into a protein subunit of a recombinant PKS that synthesizes epothilone, an WO 00/31247 - 29 _ PCT/US99/27438 epothilone derivative, or another polyketide. If the desired polyketide is an epothilone or epothilone derivative, the recombinant fourth module is typically expressed as a protein that also contains the third, fifth, and sixth modules of the epothilone PKS
or modified versions thereof. Alternatively, the invention provides recombinant PKS
enzymes for epothilones and epothilone derivatives in which the entire fourth module has been deleted or replaced by a module from a heterologous PKS.
In a preferred embodiment, the invention provides recombinant DNA compounds comprising the coding sequence for the fourth module of the epothilone PKS
modified to encode an AT that binds methylmalonyl CoA and not malonyl CoA. These recombinant molecules are used to express a protein that is a recombinant derivative of the epoD
protein that comprises the modified fourth module as well as modules 3, 5, and 6, any one or more of which can optionally be in derivative form, of the epothilone PKS.
In another preferred embodiment, the invention provides recombinant DNA compounds comprising the coding sequence for the fourth module of the epothilone PKS modified to encode an I S AT that binds malonyl CoA and not methylmalonyI CoA. These recombinant molecules are used to express a protein that is a recombinant derivative of the epoD
protein that comprises the modified fourth module as well as modules 3, 5, and 6, any one or more of which can optionally be in derivative form, of the epothilone PKS.
Prior to the present invention, it was known that Sorangium cellulosum produced epothilones A, B, C, D, E, and F and that epothilones A, C, and E had a hydrogen at C-12, while epothilones B, D, and F had a methyl group at this position.
Unappreciated prior to the present invention was the order in which these compounds were synthesized in S. cellulosum, and the mechanism by which some of the compounds had a hydrogen at C-12 where others had a methyl group at this position. The present disclosure reveals that epothilones A and B are derived from epothilones C and D by action of the epoK
gene product and that the presence of a hydrogen or methyl moiety at C-12 is due to the AT
domain of module 4 of the epothilone PKS. This domain can bind either malonyl or methylmalonyl CoA and, consistent with its having greater similarity to malonyl specific AT domains than to methylmalonyl specific AT domains, binds malonyl CoA more often than methylmalonyl CoA.
Thus, the invention provides recombinant DNA compounds and expression vectors and the corresponding recombinant PKS in which the hybrid fourth module with a methyhnalonyl specific AT has been incorporated. The methylmalonyl specific AT
coding WO 00/31247 _ 3~ _ PCTNS99/27438 sequence can originate, for example and without limitation, from coding sequences for the oleandolide PKS, DEBS, the narbonolide PKS, the rapamycin PKS, or any other PKS that comprises a methylmalonyl specific AT domain. In accordance with the invention, the hybrid fourth module expressed from this coding sequence is incorporated into the epothilone PKS (or the PKS for an epothilone derivative), typically as a derivative epoD
gene product. The resulting recombinant epothilone PKS produces epothilones with a methyl moiety at C-12, i.e., epothilone H (or an epothilone H derivative) if there is no dehydratase activity to form the C-12-C-13 alkene; epothilone D (or an epothilone D
derivative), if the dehydratase activity but not the epoxidase activity is present; epothilone B (or an epothilone B derivative), if both the dehydratase and epoxidase activity but not the hydroxylase activity are present; and epothilone F (or an epothilone F
derivative), if all three dehydratase, epoxidase, and hydroxylase activities are present. As indicated parenthetically above, the cell will produce the corresponding epothilone derivative if there have been other changes to the epothilone PKS.
If the recombinant PKS comprising the hybrid methylmalonyl specific fourth module is expressed in, for example, Sorangium cellulosum, the appropriate modifying enzymes are present (unless they have been rendered inactive in accordance with the methods herein), and epothilones D, B, and/or F are produced. Such production is typically carried out in a recombinant S. cellulosum provided by the present invention in which the native epothilone PKS is unable to function at all or unable to function except in conjunction with the recombinant fourth module provided. In an illustrative example, one can use the methods and reagents of the invention to render inactive the epoD
gene in the native host. Then, one can transform.that host with a vector comprising the recombinant epoD gene containing the hybrid fourth module coding sequence. The recombinant vector can exist as an extrachromosomal element or as a segment of DNA integrated into the host cell chromosome. In the latter embodiment, the invention provides that one can simply integrate the recombinant methylmalonyl specific module 4 coding sequence into wild-type S. cellulosum by homologous recombination with the native epoD gene to ensure that only the desired epothilone is produced. The invention provides that the S.
cellulosum host can either express or not express (by mutation or homologous recombination of the native genes therefor) the dehydratase, epoxidase, and/or oxidase gene products and thus form or not form the corresponding epothilone D, B, and F compounds, as the practitioner elects.

WO 00/31247 _ 31 _ PCTNS99/27438 Sorangium celluloSUm modified as described above is only one of the recombinant host cells provided by the invention. In a preferred embodiment, the recombinant methylmalonyl specific epothilone fourth module coding sequences are used in accordance with the methods of invention to produce epothilone D, B, and F (or their corresponding derivatives) in heterologous host cells. Thus, the invention provides reagents and methods for introducing the epothilone or epothilone derivative PKS and epothilone dehydratase, epoxidase, and hydroxylase genes and combinations thereof into heterologous host cells.
The recombinant methylmalonyl specific epothilone fourth module coding sequences provided by the invention afford important alternative methods for producing desired epothilone compounds in host cells. Thus, the invention provides a hybrid fourth module coding sequence in which, in addition to the replacement of the endogenous AT
coding sequence with a coding sequence for an AT specific for methylmalonyl Co A, coding sequences for a DH and KR for, for example and without limitation, module 10 of the rapamycin PKS or modules 1 or 5 of the FK-520 PKS have replaced the endogenous KR coding sequences. When the gene product comprising the hybrid fourth module and epothilone PKS modules 3, S, and 6 (or derivatives thereof) encoded by this coding sequence is incorporated into a PKS comprising the other epothilone PKS
proteins (or derivatives thereof) produced in a host cell, the cell makes either epothilone D or its traps stereoisomer (or derivatives thereof), depending on the stereochemical specificity of the inserted DH and KR domains.
Similarly, and as noted above, the invention provides recombinant DNA
compounds comprising the coding sequence for the fourth module of the epothilone PKS
modified to encode an AT that binds malonyl CoA and not methylmalonyl CoA. The invention provides recombinant DNA compounds and vectors and the corresponding recombinant PKS in which this hybrid fourth module has been incorporated into a derivative epoD gene product. When incorporated into the epothilone PKS (or the PKS for an epothilone derivative), the resulting recombinant epothilone PKS produces epothilones C, A, and E, depending, again, on whether epothilone modification enzymes are present.
As noted above, depending on the host, whether the fourth module includes a KR
and DH
domain, and on whether and which of the dehydratase, epoxidase, and oxidase activities are present, the practitioner of the invention can produce one or more of the epothilone G, WO 00/31247 _ 32 _ PCT/US99/27438 C, A, and E compounds and derivatives thereof using the compounds, host cells, and methods of the invention.
The fifth module of the epothilone PKS includes a KS, an AT that binds malonyl CoA, a DH, an ER, a KR, and an ACP. This module is encoded by a sequence within an 12.4 kb NsiI-NotI restriction fragment of cosmid pKOS35-70.1A2.
The recombinant DNA compounds of the invention that encode the fifth module of the epothilone PKS and the corresponding polypeptides encoded thereby are useful for a variety of applications. In one embodiment, a DNA compound comprising a sequence that encodes the epothilone fifth module is inserted into a DNA compound that comprises the coding sequence for one or more modules of a heterologous PKS. The resulting construct, in which the coding sequence for a module of the heterologous PKS is either replaced by that for the fifth module of the epothilone PKS or the latter is merely added to coding sequences for the modules of the heterologous PKS, can be incorporated into an expression vector and used to produce the recombinant protein encoded thereby.
When the recombinant protein is combined with the other proteins of the heterologous PKS, a novel PKS is produced. In another embodiment, a DNA compound comprising a sequence that encodes the fifth module of the epothilone PKS is inserted into a DNA compound that comprises coding sequences for the epothilone PKS or a recombinant epothilone PKS that produces an epothilone derivative. In the latter constructs, the epothilone fifth module is typically expressed as a protein comprising the third, fourth, and sixth modules of the epothilone PKS or derivatives thereof.
In another embodiment, a portion of the fifth module coding sequence is utilized in conjunction with other PKS coding sequences to create a hybrid module coding sequence and the hybrid module encoded thereby. In this embodiment, the invention provides, for example, either replacing the malonyl CoA specific AT with a methylmalonyl CoA, ethylmalonyl CoA, or 2-hydroxymalonyl CoA specific AT; deleting any one, two, or all three of the ER, DH, and KR; and/or replacing any one, two, or all three of the ER, DH, and KR with either a KR, a DH and KR, or a KR, DH, and ER, including, optionally, to specify a different stereochemistry. In addition, the KS and/or ACP can be replaced with another KS and/or ACP. In each of these replacements or insertions, the heterologous KS, AT, DH, KR, ER, or ACP coding sequence can originate from a coding sequence for another module of the epothilone PKS, from a coding sequence for a PKS that produces a polyketide other than epothilone, or from chemical synthesis. The resulting hybrid fifth WO 00/31247 _ 33 _ PCT/US99/27438 module coding sequence can be utilized in conjunction with a coding sequence for a PKS
that synthesizes epothilone, an epothilone derivative, or another polyketide.
Alternatively, the fifth module of the epothilone PKS can be deleted or replaced in its entirety by a module of a heterologous PKS to produce a protein that in combination with the other proteins of the epothilone PKS or derivatives thereof constitutes a PKS that produces an epothilone derivative.
Illustrative recombinant PKS genes of the invention include recombinant epoD
gene derivatives in which the AT domain encoding sequences for the fifth module of the epothilone PKS have been altered or replaced to change the AT domain encoded thereby from a malonyl specific AT to a methylmalonyl specific AT. Such methylmalonyl specific AT domain encoding nucleic acids can be isolated, for example and without limitation, from the PKS genes encoding DEBS, the narbonolide PKS, the rapamycin FKS, and the FK-520 PKS. When such recombinant epoD gene derivatives are coexpressed with the epoA, epoB, epoC, epoE, and epoF genes (or derivatives thereof), the PKS
composed thereof produces the 10-methyl epothilones or derivatives thereof. Another recombinant epoD gene derivative provided by the invention includes not only this altered module 5 coding sequence but also module 4 coding sequences that encode an AT domain that binds only methylmalonyl CoA. When incorporated into a PKS with the epoA, epoB, epoC, epoE, and epoF genes, the recombinant epoD gene derivative product leads to the production of 10-methyl epothilone B and/or D derivatives.
Other illustrative recombinant epoD gene derivatives of the invention include those in which the ER, DH, and KR domain encoding sequences for the fifth module of the epothilone PKS have been replaced.with those encoding (i) a KR and DH domain;
(ii) a KR domain; and (iii) an inactive KR domain. These recombinant epoD gene derivatives of the invention are coexpressed with the epoA, epoB, epoC, epoE, and epoF genes to produce a recombinant PKS that makes the corresponding {i) C-11 alkene, (ii) C-hydroxy, and (iii) C-11 keto epothilone derivatives. These recombinant epoD
gene derivatives can also be coexpressed with recombinant epo genes containing other alterations or can themselves be further altered to produce a PKS that makes the corresponding C-11 epothilone derivatives. For example, one recombinant epoD
gene derivative provided by the invention also includes module 4 coding sequences that encode an AT domain that binds only methylmalonyl CoA. When incorporated into a PKS
with the epoA, epoB, epoC, epoE, and epoF genes, the recombinant epoD gene derivative WO 00/31247 _ 34 _ PCT/US99/27438 product leads to the production of the corresponding C-11 epothilone B and/or D
derivatives.
Functionally similar epoD genes for producing the epothilone C-11 derivatives can also be made by inactivation of one, two, or all three of the ER, DH, and KR
domains of the epothilone fifth module. However, the preferred mode for altering such domains in any module is by replacement with the complete set of desired domains taken from another module of the same or a heterologous PKS coding sequence. In this manner, the natural architecture of the PKS is conserved. Also, when present, KR and DH or KR, DH, and ER
domains that function together in a native PKS are preferably used in the recombinant PKS. Illustrative replacement domains for the substitutions described above include, for example and without limitation, the inactive KR domain from the rapamycin PKS
module 3 to form the ketone, the KR domain from the rapamycin PKS module 5 to form the alcohol, and the KR and DH domains from the rapamycin PKS module 4 to form the alkene. Other such inactive KR, active KR, and active KR and DH domain encoding nucleic acids can be isolated from, for example and without limitation, the PKS genes encoding DEBS, the narbonolide PKS, and the FK-520 PKS. Each of the resulting PKS
enzymes produces a polyketide compound that comprises a functional group at the C-11 position that can be further derivatized in vitro by standard chemical methodology to yield semi-synthetic epothilone derivatives of the invention.
The sixth module of the epothilone PKS includes a KS, an AT that binds methylmalonyl CoA, a DH, an ER, a KR, and an ACP. This module is encoded by a sequence within an 14.5 kb HindIII-NsiI restriction fragment of cosmid pKOS35-70. I A2.
The recombinant DNA compounds of the invention that encode the sixth module of the epothilone PKS and the corresponding polypeptides encoded thereby are useful for a variety of applications. In one embodiment, a DNA compound comprising a sequence that encodes the epothilone sixth module is inserted into a DNA compound that comprises the coding sequence for one or more modules of a heterologous PKS. The resulting protein encoded by the construct, in which the coding sequence for a module of the heterologous PKS is either replaced by that for the sixth module of the epothilone PKS or the latter is merely added to coding sequences for the modules of the heterologous PKS, provides a novel PKS when coexpressed with the other proteins comprising the PKS. In another embodiment, a DNA compound comprising a sequence that encodes the sixth module of WO 00/3124? _ 35 _ PCT/US99/2743$
the epothilone PKS is inserted into a DNA compound that comprises the coding sequence for modules 3, 4, and 5 of the epothilone PKS or a recombinant epothilone PKS
that produces an epothilone derivative and coexpressed with the other proteins of the epothilone or epothilone derivative PKS to produce a PKS that makes epothilone or an epothilone derivative in a host cell.
In another embodiment, a portion of the sixth module coding sequence is utilized in conjunction with other PKS coding sequences to create a hybrid module. In this embodiment, the invention provides, for example, either replacing the methylmalonyl CoA
specific AT with a malonyl CoA, ethylmalonyl CoA, or 2-hydroxymalonyl CoA
specific AT; deleting any one, two, or all three of the ER, DH, and KR; and/or replacing any one, two, or all three of the ER, DH, and KR with either a KR, a DH and KR, or a KR, DH, and ER, including, optionally, to specify a different stereochemistry. In addition, the KS and/or ACP can be replaced with another KS and/or ACP. In each of these replacements or insertions, the heterologous KS, AT, DH, KR, ER, or ACP coding sequence can originate from a coding sequence for another module of the epothilone PKS, from a coding sequence for a PKS that produces a polyketide other than epothilone, or from chemical synthesis. The resulting heterologous sixth module coding sequence can be utilized in conjunction with a coding sequence for a protein subunit of a PKS that makes epothilone, an epothilone derivative, or another polyketide. If the PKS makes epothilone or an epothilone derivative, the hybrid sixth module is typically expressed as a protein comprising modules 3, 4, and 5 of the epothilone PKS or derivatives thereof.
Alternatively, the sixth module of the epothilone PKS can be deleted or replaced in its entirety by a module from a heterologous PKS to produce a PKS for an epothilone derivative.
Illustrative recombinant PKS genes of the invention include those in which the AT
domain encoding sequences for the sixth module of the epothilone PKS have been altered or replaced to change the AT domain encoded thereby from a methylmalonyl specific AT
to a malonyl specific AT. Such malonyl specific AT domain encoding nucleic acids can be isolated from, for example and without limitation, the PKS genes encoding the narbonolide PKS, the rapamycin PKS, and the FK-520 PKS. When a recombinant epoD
gene of the invention encoding such a hybrid module 6 is coexpressed with the other epothilone PKS genes, the recombinant PKS makes the 8-desmethyl epothilone derivatives. This recombinant epoD gene derivative can also be coexpressed with recombinant epo gene derivatives containing other alterations or can itself be further altered to produce a PKS that makes the corresponding 8-desmethyl epothilone derivatives. For example, one recombinant epoD gene provided by the invention also includes module 4 coding sequences that encode an AT domain that binds only methylmalonyl CoA. When incorporated into a PKS with the epoA, epoB, epoC, epoE, and epoF genes, the recombinant epoD gene product leads to the production of the 8-desmethyl derivatives of epothilones B and D.
Other illustrative recombinant epoD gene derivatives of the invention include those in which the ER, DH, and KR domain encoding sequences for the sixth module of the epothilone PKS have been replaced with those that encode (i) a KR and DH
domain; (ii) a KR domain; and (iii) an inactive KR domain. These recombinant epoD gene derivatives of the invention, when coexpressed with the other epothilone PKS genes make the corresponding (i) C-9 alkene, (ii) C-9 hydroxy, and (iii) C-9 keto epothilone derivatives.
These recombinant epoD gene derivatives can also be coexpressed with other recombinant epo gene derivatives containing other alterations or can themselves be further altered to produce a PKS that makes the corresponding C-9 epothilone derivatives. For example, one recombinant epoD gene derivative provided by the invention also includes module 4 coding sequences that encode an AT domain that binds only methylmalonyl CoA.
When incorporated into a PKS with the epoA, epoB, epoC, epoE, and epoF genes, the recombinant epoD gene product leads to the production of the C-9 derivatives of epothilones B and D.
Functionally equivalent sixth modules can also be made by inactivation of one, two, or all three of the ER, DH, and KR domains of the epothilone sixth module. The preferred mode for altering such domains in any module is by replacement with the complete set of desired domains taken from another module of the same or a heterologous PKS coding sequence. Illustrative replacement domains for the substitutions described above include but are not limited to the inactive KR domain from the rapamycin PKS
module 3 to form the ketone, the KR domain from the rapamycin PKS module 5 to form the alcohol, and the KR and DH domains from the rapamycin PKS module 4 to form the alkene. Other such inactive KR, active KR, and active KR and DH domain encoding nucleic acids can be isolated from for example and without limitation the PKS
genes encoding DEBS, the narbonolide PKS, and the FK-520 PKS. Each of the resulting PKSs produces a polyketide compound that comprises a functional group at the C-9 position that WO 00/31247 _ 37 _ PCT/US99/27438 can be further derivatized in vitro by standard chemical methodology to yield semi-synthetic epothilone derivatives of the invention.
The seventh module of the epothilone PKS includes a KS, an AT specific for methylmalonyl CoA, a KR, and an ACP. This module is encoded by a sequence within an ~8.7 kb BgIII restriction fragment from cosmid pKOS35-70.4.
The recombinant DNA compounds of the invention that encode the seventh module of the epothilone PKS and the corresponding polypeptides encoded thereby are useful for a variety of applications. The seventh module of the epothilone PKS
is contained in the gene product of the epoE gene, which also contains the eighth module.
The present invention provides the epoE gene in recombinant form, but also provides DNA compounds that encode the seventh module without coding sequences for the eighth module as well as DNA compounds that encode the eighth module without coding sequences for the seventh module. In one embodiment, a DNA compound comprising a sequence that encodes the epothilone seventh module is inserted into a DNA
compound that comprises the coding sequence for one or more modules of a heterologous PKS. The resulting construct, in which the coding sequence for a module of the heterologous PKS is either replaced by that for the seventh module of the epothilone PKS or the latter is merely added to coding sequences for the modules of the heterologous PKS, provides a novel PKS coding sequence that can be expressed in a host cell. Alternatively, the epothilone seventh module can be expressed as a discrete protein. In another embodiment, a DNA
compound comprising a sequence that encodes the seventh module of the epothilone PKS
is expressed to form a protein that, together with other proteins, constitutes the epothilone PKS or a PKS that produces an epothilone derivative. In these embodiments, the seventh module is typically expressed as a protein comprising the eighth module of the epothilone PKS or a derivative thereof and coexpressed with the epoA, epoB, epoC, epoD, and epoF
genes or derivatives thereof to constitute the PKS.
In another embodiment, a portion or all of the seventh module coding sequence is utilized in conjunction with other PKS coding sequences to create a hybrid module. In this embodiment, the invention provides, for example, either replacing the methylmalonyl CoA
specific AT with a malonyl CoA, ethylmalonyl CoA, or 2-hydroxymalonyl CoA
specific AT; deleting the KR; replacing the KR with a KR that specifies a different stereochemistry; and/or inserting a DH or a DH and an ER. In addition, the KS
and/or ACP can be replaced with another KS and/or ACP. In each of these replacements or WO 00/3124? _ 3g _ PCT/US99/27438 insertions, the heterologous KS, AT, DH, KR, ER, or ACP coding sequence can originate from a coding sequence for another module of the epothilone PKS, from a coding sequence for a PKS that produces a polyketide other than epothilone, or from chemical synthesis. The resulting heterologous seventh module coding sequence is utilized, optionally in conjunction with other coding sequences, to express a protein that together with other proteins constitutes a PKS that synthesizes epothilone, an epothilone derivative, or another polyketide. When used to prepare epothilone or an epothilone derivative, the seventh module is typically expressed as a protein comprising the eighth module or derivative thereof and coexpressed with the epoA, epoB, epoC, epoD, and epoF
genes or derivatives thereof to constitute the PKS. Alternatively, the coding sequences for the seventh module in the epoE gene can be deleted or replaced by those for a heterologous module to prepare a recombinant epoE gene derivative that, together with the epoA, epoB, epoC, epoD, and epoF genes, can be expressed to make a PKS for an epothilone derivative.
Illustrative recombinant epoE gene derivatives of the invention include those in which the AT domain encoding sequences for the seventh module of the epothilone PKS
have been altered or replaced to change the AT domain encoded thereby from a methylmalonyl specific AT to a malonyl specific AT. Such malonyl specific AT
domain encoding nucleic acids can be isolated from for example and without limitation the PKS
genes encoding the narbonolide PKS, the rapamycin PKS, and the FK-520 PKS.
When coexpressed with the other epothilone PKS genes, epoA, epoB, epoC, epoD, and epoF, or derivatives thereof, a PKS for an epothilone derivative with a C-6 hydrogen, instead of a C-6 methyl, is produced. Thus, if the genes contain no other alterations, the compounds produced are the 6-desmethyl epothilones.
The eighth module of the epothilone PKS includes a KS, an AT specific for methylmalonyl CoA, inactive KR and DH domains, a methyltransferase (MT) domain, and an ACP. This module is encoded by a sequence within an ~10 kb NotI
restriction fragment of cosmid pKOS35-79.85.
The recombinant DNA compounds of the invention that encode the eighth module of the epothilone PKS and the corresponding polypeptides encoded thereby are useful for a variety of applications. In one embodiment, a DNA compound comprising a sequence that encodes the epothilone eighth module is inserted into a DNA compound that comprises the coding sequence for one or more modules of a heterologous PKS.
The WO 00/31247 - 39 - PC'f/US99/27438 resulting construct, in which the coding sequence for a module of the heterologous PKS is either replaced by that for the eighth module of the epothilone PKS or the latter is merely added to coding sequences for modules of the heterologous PKS, provides a novel PKS
coding sequence that is expressed with the other proteins constituting the PKS
to provide a novel PKS. Alternatively, the eighth module can be expressed as a discrete protein that can associate with other PKS proteins to constitute a novel PKS. In another embodiment, a DNA compound comprising a sequence that encodes the eighth module of the epothilone PKS is coexpressed with the other proteins constituting the epothilone PKS or a PKS that produces an epothilone derivative. In these embodiments, the eighth module is typically expressed as a protein that also comprises the seventh module or a derivative thereof.
In another embodiment, a portion or all of the eighth module coding sequence is utilized in conjunction with other PKS coding sequences to create a hybrid module. In this embodiment, the invention provides, for example, either replacing the methylmalonyl CoA
specific AT with a malonyl CoA, ethylmalonyl CoA, or 2-hydroxymalonyl CoA
specific AT; deleting the inactive KR and/or the inactive DH; replacing the inactive KR
and/or DH
with an active KR and/or DH; and/or inserting an ER. In addition, the KS
and/or ACP can be replaced with another KS and/or ACP. In each of these replacements or insertions, the heteroIogous KS, AT, DH, KR, ER, or ACP coding sequence can originate from a coding sequence for another module of the epothilone PKS, from a coding sequence for a PKS
that produces a polyketide other than epothilone, or from chemical synthesis.
The resulting heterologous eighth module coding sequence is expressed as a protein that is utilized in conjunction with the other proteins that constitute a PKS that synthesizes epothilone, an epothilone derivative, or another polyketide. When used to prepare epothilone or an epothilone derivative, the heterologous or hybrid eighth module is typically expressed as a recombinant epoE gene product that also contains the seventh module.
Alternatively, the coding sequences for the eighth module in the epoE gene can be deleted or replaced by those for a heterologous module to prepare a recombinant epoE gene that, together with the epoA, epoB, epoC, epoD, and epoF genes, can be expressed to make a PKS for an epothilone derivative.
The eighth module of the epothilone PKS also comprises a methylation or methyltransferase (MT) domain with an activity that methylates the epothilone precursor.
This function can be deleted to produce a recombinant epoD gene derivative of the invention, which can be expressed with the other epothilone PKS genes or derivatives WO 00/31247 - 40 _ PCT/US99/27438 thereof that makes an epothilone derivative that lacks one or both methyl groups, depending on whether the AT domain of the eighth module has been changed to a malonyl specific AT domain, at the corresponding C-4 position of the epothilone molecule. In another important embodiment, the present invention provides recombinant DNA
compounds that encode a polypeptide with this methylation domain and activity and a variety of recombinant PKS coding sequences that encode recombinant PKS
enzymes that incorporate this polypeptide. The availability of this MT domain and the coding sequences therefor provides a significant number of new polyketides that differ from known polyketides by the presence of at least an additional methyl group. The MT
domain of the invention can in effect be added to any PKS module to direct the methylation at the corresponding location in the polyketide produced by the PKS. As but one illustrative example, the present invention provides the recombinant nucleic acid compounds resulting from inserting the coding sequence for this MT activity into a coding sequence for any one or more of the six modules of the DEBS enzyme to produce a recombinant DEBS
that synthesizes a 6-deoxyerythronolide B derivative that comprises one or more additional methyl groups at the C-2, C-4, C-6, C-8, C-10, and/or C-12 positions. In such constructs, the MT domain can be inserted adjacent to the AT or the ACP.
The ninth module of the epothilone PKS includes a KS, an AT specific for malonyl CoA, a KR, an inactive DH, and an ACP. This module is encoded by a sequence within an 14.7 HindIII-BgIII kb restriction fragment of cosmid pKOS35-79.85.
The recombinant DNA compounds of the invention that encode the ninth module of the epothilone PKS and the corresponding polypeptides encoded thereby are useful for a variety of applications. The ninth module of the epothilone PKS is expressed as a protein, the product of the epoF gene, that also contains the TE domain of the epothilone PKS. The present invention provides the epoF gene in recombinant form, as well as DNA
compounds that encode the ninth module without the coding sequences for the TE
domain and DNA compounds that encode the TE domain without the coding sequences for the ninth module. In one embodiment, a DNA compound comprising a sequence that encodes the epothilone ninth module is inserted into a DNA compound that comprises the coding sequence for one or more modules of a heterologous PKS. The resulting construct, in which the coding sequence for a module of the heterologous PKS is either replaced by that for the ninth module of the epothilone PKS or the latter is merely added to coding sequences for the modules of the heterologous PKS, provides a novel PKS
protein coding WO 00/31247 _ 41 _ PCT/US99/27438 sequence that when coexpressed with the other proteins constituting a PKS
provides a novel PKS. The ninth module coding sequence can also be expressed as a discrete protein with or without an attached TE domain. In another embodiment, a DNA compound comprising a sequence that encodes the ninth module of the epothilone PKS is expressed as a protein together with other proteins to constitute an epothilone PKS or a PKS that produces an epothilone derivative. In these embodiments, the ninth module is typically expressed as a protein that also contains the TE domain of either the epothilone PKS or a heterologous PKS.
In another embodiment, a portion or all of the ninth module coding sequence is utilized in conjunction with other PKS coding sequences to create a hybrid module. In this embodiment, the invention provides, for example, either replacing the malonyl CoA
specific AT with a methylmalonyl CoA, ethylmalonyl CoA, or 2-hydroxy malonyl CoA
specific AT; deleting the KR; replacing the KR with a KR that specifies a different stereochemistry; and/or inserting a DH or a DH and an ER. In addition, the KS
and/or 1 S ACP can be replaced with another KS and/or ACP. In each of these replacements or insertions, the heterologous KS, AT, DH, KR, ER, or ACP coding sequence can originate from a coding sequence for another module of the epothilone PKS, from a coding sequence for a PKS that produces a polyketide other than epothilone, or from chemical synthesis. The resulting heterologous ninth module coding sequence is coexpressed with the other proteins constituting a PKS that synthesizes epothilone, an epothilone derivative, or another polyketide. Alternatively, the present invention provides a PKS for an epothilone or epothilone derivative in which the ninth module has been replaced by a module from a heterologous PKS or has been deleted in its entirety. In the latter embodiment, the TE domain is expressed as a discrete protein or fused to the eighth module.
The ninth module of the epothilone PKS is followed by a thioesterase domain.
This domain is encoded in the 14.7 kb HindIII-BgIII restriction comprising the ninth module coding sequence. The present invention provides recombinant DNA compounds that encode hybrid PKS enzymes in which the ninth module of the epothilone PKS is fused to a heterologous thioesterase or one or more modules of a heterologous PKS are fused to the epothilone PKS thioesterase. Thus, for example, a thioesterase domain coding sequence from another PKS can be inserted at the end of the ninth module ACP coding sequence in recombinant DNA compounds of the invention. Recombinant DNA compounds encoding this thioesterase domain are therefore useful in constructing DNA compounds that encode a protein of the epothilone PKS, a PKS that produces an epothilone derivative, and a PKS
that produces a polyketide other than epothilone or an epothilone derivative.
In one important embodiment, the present invention thus provides a hybrid PKS
and the corresponding recombinant DNA compounds that encode the proteins constituting those hybrid PKS enzymes. For purposes of the present invention a hybrid PKS
is a recombinant PKS that comprises all or part of one or more modules, loading domain, and thioesterase/cyclase domain of a first PKS and all or part of one or more modules, loading domain, and thioesterase/cyclase domain of a second PKS. In one preferred embodiment, the first PKS is most but not all of the epothilone PKS, and the second PKS is only a portion or all of a non-epothilone PKS. An illustrative example of such a hybrid PKS
includes an epothilone PKS in which the natural loading domain has been replaced with a loading domain of another PKS. Another example of such a hybrid PKS is an epothilone PKS in which the AT domain of module four is replaced with an AT domain from a heterologous PKS that binds only methylmalonyl CoA. In another preferred embodiment, the first PKS is most but not all of a non-epothilone PKS, and the second PKS
is only a portion or all of the epothilone PKS. An illustrative example of such a hybrid PKS
includes an erythromycin PKS in which an AT specific for methylmalonyl CoA is replaced with an AT from the epothilone PKS specific for malonyl CoA. Another example is an erythromycin PKS that includes the MT domain of the epothilone PKS.
Those of skill in the art will recognize that all or part of either the first or second PKS in a hybrid PKS of the invention need not be isolated from a naturally occurring source. For example, only a small portion of an AT domain determines its specificity. See U.S. patent application Serial No. 09/346,860 and PCT patent application No.
WO
US99/15047, each of which is incorporated herein by reference. The state of the art in DNA synthesis allows the artisan to construct de novo DNA compounds of size sufficient to construct a useful portion of a PKS module or domain. For purposes of the present invention, such synthetic DNA compounds are deemed to be a portion of a PKS.
The following Table lists references describing illustrative PKS genes and corresponding enzymes that can be utilized in the construction of the recombinant PKSs and the corresponding DNA compounds that encode them of the invention. Also presented are various references describing polyketide tailoring and modification enzymes and corresponding genes that can be employed to make the recombinant DNA compounds of the present invention.
Avermectin U.S. Pat. No. 5,252,474 to Merck.
MacNeil et al., 1993, Industrial Microorganisms: Basic and Applied Molecular Genetics, Baltz, Hegeman, & Skatrud, eds. (ASM), pp. 245-256, A Comparison of the Genes Encoding the Polyketide Synthases for Avermectin, Erythromycin, and Nemadectin.
MacNeil et al., 1992, Gene 115: 119-125, Complex Organization of the Streptomyces avermitilis genes encoding the avermectin polyketide synthase.
Ikeda and Omura, 1997, Chem. Res. 97: 2599-2609, Avermectin biosynthesis.
Candicidin (FR008) Hu et al., 1994, Mol. Microbiol. 14: 163-172.
Erythromycin PCT Pub. No. 93/13663 to Abbott.
US Pat. No. 5,824,513 to Abbott.
Donadio et al., 1991, Science 252:675-9.
Cortes et al., 8 Nov. 1990, Nature 348:176-8, An unusually large multifunctional polypeptide in the erythromycin producing polyketide synthase of Saccharopolyspora erythraea.
Glycosylation Enzymes PCT Pat. App. Pub. No. 97/23630 to Abbott.

Motamedi et al., 1998, The biosynthetic gene cluster for the macrolactone ring of the immunosuppressant FK-506, Eur. J. Biochem. 256: 528-534.
Motamedi et al., 1997, Structural organization of a multifunctional polyketide synthase involved in the biosynthesis of the macrolide immunosuppressant FK-506, Eur. J.
Biochem. 244: 74-80.
Methyltransferase US 5,264,355, issued 23 Nov. 1993, Methylating enzyme from Streptomyces MA6858. 31-O-desmethyl-FK-506 methyltransferase.

Motamedi et al., 1996, Characterization of methyltransferase and hydroxylase genes involved in the biosynthesis of the immunosuppressants FK-506 and FK-520, J.
Bacteriol. 178: 5243-5248.

S U.S. patent application Serial No. 09/154,083, filed 16 Sep. 1998.
U.S. patent application Serial No. 09/410,55 l, filed 1 Oct. 1999.
Nielsen et al., I991, Biochem. 30:5789-96.
Lovastatin U.S. Pat. No. 5,744,350 to Merck.
Narbomycin U.S. patent application Serial No. 60/107,093, filed S Nov. 1998.
Nemadectin MacNeil et al., 1993, supra.
Niddamycin 1 S Kakavas et al., 1997, Identification and characterization of the niddamycin polyketide synthase genes from Streptomyces caelestis, J. Bacteriol. 179: 7S 1 S-7522.
Oleandomycin Swan et al., 1994, Characterisation of a Streptomyces antibioticus gene encoding a type I polyketide synthase which has an unusual coding sequence, Mol. Gen.
Genet. 242:
358-362.
U.S. patent application Serial No. 60/120,254, filed 16 Feb. 1999, Serial No.
09/ , filed 28 Oct. 1999, claiming priority thereto by inventors S. Shah, M.
Betlach, R. McDaniel, and L. Tang, attorney docket No. 30063-20029.00.
Olano et al., 1998, Analysis of a Streptomyces antibioticus chromosomal region 2S involved in oleandomycin biosynthesis, which encodes two glycosyltransferases responsible for glycosylation of the macrolactone ring, Mol. Gen. Genet.
259(3): 299-308.
Picromycin PCT patent application No. WO US99/11814, filed 28 May 1999.
U.S. patent application Serial No. 09/320,878, filed 27 May 1999.
U.S. patent application Serial No. 09/141,908, filed 28 Aug. 1998.

WO 00/31247 _ 45 _ PCT/US99/27438 Xue et al., 1998, Hydroxylation of macrolactones YC-17 and narbomycin is mediated by the pikC-encoded cytochrome P450 in Streptomyces venezuelae, Chemistry & Biology 5(11): 661-667.
Xue et al., Oct. 1998, A gene cluster for macrolide antibiotic biosynthesis in Streptomyces venezuelae: Architecture of metabolic diversity, Proc. Natl.
Acad. Sci. USA
95: 12111 12116.
Platenolide EP Pat. App. Pub. No. 791,656 to Lilly.
Pradimicin PCT Pat. Pub. No. WO 98/11230 to Bristol-Myers Squibb.
Rapamycin Schwecke et al., Aug. 1995, The biosynthetic gene cluster for the polyketide rapamycin, Proc. Natl. Acad. Sci. USA 92:7839-7843.
Aparicio et al., 1996, Organization of the biosynthetic gene cluster for rapamycin in Streptomyces hygroscopicus: analysis of the enzymatic domains in the modular polyketide synthase, Gene 169: 9-16.
Rifamycin PCT Pat. Pub. No. WO 98/07868 to Novartis.
August et al., 13 Feb. 1998, Biosynthesis of the ansamycin antibiotic rifamycin:
deductions from the molecular analysis of the rif biosynthetic gene cluster of Amycolatopsis mediterranei 5669, Chemistry & Biology, 5(2): 69-79.
Sorangium PKS
U.S. patent application Serial No. 09/144,085, filed 31 Aug. 1998.
Soraphen U.S. Pat. No. 5,716,849 to Novartis.
Schupp et al., 1995, J. Bacteriology 177: 3673-3679. A Sorangium cellulosum (Myxobacterium) Gene Cluster for the Biosynthesis of the Macrolide Antibiotic Soraphen A: Cloning, Characterization, and Homology to Polyketide Synthase Genes from Actinomycetes.
Spiramycin U.S. Pat. No. 5,098,837 to Lilly.
Activator Gene U.S. Pat. No. 5,514,544 to Lilly.

WO 00/31247 _ 46 _ PCT/US99/27438 Tylosin U.S. Pat. No. 5,876,991 to Lilly.
EP Pub. No. 791,655 to Lilly.
Kuhstoss et al., 1996, Gene 183:231-6., Production of a novel polyketide through the construction of a hybrid polyketide synthase.
Tailoring enzymes Merson-Davies and Cundliffe, 1994, Mol. Microbiol. 13: 349-355. Analysis of five tylosin biosynthetic genes from the tylBA region of the Streptomyces fradiae genome.
As the above Table illustrates, there are a wide variety of PKS genes that serve as readily available sources of DNA and sequence information for use in constructing the hybrid PKS-encoding DNA compounds of the invention. Methods for constructing hybrid PKS-encoding DNA compounds are described without reference to the epothilone PKS in U.S. Patent Nos. 5,672,491 and 5,712,146 and U.S. patent application Serial Nos.
09/073,538, filed 6 May 1998, and 09/141,908, filed 28 Aug 1998, each of which is I 5 incorporated herein by reference. Preferred PKS enzymes and coding sequences for the proteins which constitute them for purposes of isolating heterologous PKS
domain coding sequences for constructing hybrid PKS enzymes of the invention are the soraphen PKS
and the PKS described as a Sorangium PKS in the above table.
To summarize the functions of the genes cloned and sequenced in Example 1:
Gene Protein Modules Domains Present epoA EpoA Load KsY mAT ER ACP

epo8 EpoB 1 NRPS, condensation, heterocyclization , adenylation, thiolation, PCP

epoC EpoC 2 KS mmAT DH KR ACP

epoD EpoD 3 KS mAT KR ACP

4 KS mAT KR ACP
5 KS mAT DH ER KR ACP
6 KS mmAT DH ER KR ACP
epoE EpoE 7 KS mmAT KR ACP
8 KS mmAT MT DH* KR* ACP
epoF EpoF 9 KS mAT KR DH* ACP TE

WO 00/31247 _ q,~ _ PCT/US99/27438 NRPS - non-ribosomal peptide synthetase; KS - ketosynthase; mAT - malonyl CoA
specifying acyltransferase; mmAT - methylmalonyl CoA specifying acyltransferase; DH -dehydratase; ER - enoylreductase; KR - ketoreductase; MT - methyltransferase;
TE
thioesterase; * - inactive domain.
The hybrid PKS-encoding DNA compounds of the invention can be and often are hybrids of more than two PKS genes. Even where only two genes are used, there are often two or more modules in the hybrid gene in which all or part of the module is derived from a second (or third) PKS gene. Illustrative examples of recombinant epothilone derivative PKS genes of the invention, which are identified by listing the specificities of the hybrid modules (the other modules having the same specificity as the epothilone PKS), include:
(a) module 4 with methylmalonyl specific AT (mm AT) and a KR and module 2 with a malonyl specific AT (m AT) and a KR;
(b) module 4 with mM AT and a KR and module 3 with mM AT and a KR;
(c) module 4 with mM AT and a KR and module 5 with mM AT and a ER, DH, and KR;
(d) module 4 with mM AT and a KR and module 5 with mM AT and a DH and KR;
(e) module 4 with mM AT and a KR and module 5 with mM AT and a KR;
(f) module 4 with mM AT and a KR and module 5 with mM AT and an inactive KR;
(g) module 4 with mM AT and a KR and module 6 with m AT and a ER, DH, and KR;
(h) module 4 with mM AT and a KR and module 6 with m AT and a DH and KR;
(i) module 4 with mM AT and a KR and module 6 with m AT and a KR;
(j) module 4 with mM AT and a KR and module 6 with m AT and an inactive KR;
(k) module 4 with mM AT and a KR and module 7 with m AT;
(1) hybrids (c) through (f), except that module 5 has a m AT;
(m) hybrids (g) through (j) except that module 6 has a mM AT; and (n) hybrids (a) through (m) except that module 4 has a m AT.
The above list is illustrative only and should not be construed as limiting the invention, which includes other recombinant epothilone PKS genes and enzymes with not only two hybrid modules other than those shown but also with three or more hybrid modules.

WO 00/31247 _ Qg _ PCT/US99/27438 Those of skill in the art will appreciate that a hybrid PKS of the invention includes but is not limited to a PKS of any of the following types: (i) an epothilone or epothilone derivative PKS that contains a module in which at least one of the domains is from a heterologous module; (ii) an epothilone or epothilone derivative PKS that contains a module from a heterologous PKS; (iii) an epothilone or epothilone derivative PKS that contains a protein from a heterologous PKS; and (iv) combinations of the foregoing.
While an important embodiment of the present invention relates to hybrid PKS
genes, the present invention also provides recombinant epothilone PKS genes in which there is no second PKS gene sequence present but which differ from the epothilone PKS
gene by one or more deletions. The deletions can encompass one or more modules and/or can be limited to a partial deletion within one or more modules. When a deletion encompasses an entire module other than the NRPS module, the resulting epothilone derivative is at least two carbons shorter than the compound produced from the PKS from which the deleted version was derived. The deletion can also encompass the NRPS
module and/or the loading domain, as noted above. When a deletion is within a module, the deletion typically encompasses a KR, DH, or ER domain, or both DH and ER
domains, or both KR and DH domains, or all three KR, DH, and ER domains.
The catalytic properties of the domains and modules of the epothilone PKS and of epothilone modification enzymes can also be altered by random or site specific mutagenesis of the corresponding genes. A wide variety of mutagenizing agents and methods are known in the art and are suitable for this purpose. The technique known as DNA shuffling can also be employed. See, e.g., U.S. Patent Nos. 5,830,721;
5,811,238;
and 5,605,793; and references cited therein, each of which is incorporated herein by reference.
Recombinant Manipulations To construct a hybrid PKS or epothilone derivative PKS gene of the invention, or simply to express unmodified epothilone biosynthetic genes, one can employ a technique, described in PCT Pub. No. 98/27203 and U.S. patent application Serial Nos.
08/989,332, filed 11 Dec. 1997, and 60/129,731, filed 16 April 1999, each of which is incorporated herein by reference, in which the various genes of the PKS are divided into two or more, often three, segments, and each segment is placed on a separate expression vector. In this manner, the full complement of genes can be assembled and manipulated more readily for heterologous expression, and each of the segments of the gene can be altered, and various altered segments can be combined in a single host cell to provide a recombinant PKS of the invention. This technique makes more efficient the construction of large libraries of recombinant PKS genes, vectors for expressing those genes, and host cells comprising those vectors. In this and other contexts, the genes encoding the desired PKS
are not only present on two or more vectors, but also can be ordered or arranged differently than in the native producer organism from which the genes were derived. Various examples of this technique as applied to the epothilone PKS are described in the Examples below. In one embodiment, the epoA, epoB, epoC, and epoD genes are present on a first plasmid, and the I 0 epoE and epoF and optionally either the epoK or the epoK and epoL genes are present on a second (or third) plasmid.
Thus, in one important embodiment, the recombinant nucleic acid compounds of the invention are expression vectors. As used herein, the term "expression vector" refers to any nucleic acid that can be introduced into a host cell or cell-free transcription and translation medium. An expression vector can be maintained stably or transiently in a cell, whether as part of the chromosomal or other DNA in the cell or in any cellular compartment, such as a replicating vector in the cytoplasm. An expression vector also comprises a gene that serves to produce RNA that is translated into a polypeptide in the cell or cell extract. Thus, the vector typically includes a promoter to enhance gene expression but alternatively may serve to incorporate the relevant coding sequence under the control of an endogenous promoter. Furthermore, expression vectors may typically contain additional functional elements, such as resistance-confernng genes to act as selectable markers and regulatory genes to enhance promoter activity.
The various components of an expression vector can vary widely, depending on the intended use of the vector. In particular, the components depend on the host cells) in which the vector will be used or is intended to function. Vector components for expression and maintenance of vectors in E. coli are widely known and commercially available, as are vector components for other commonly used organisms, such as yeast cells and Streptomyces cells.
In one embodiment, the vectors of the invention are used to transform Sorangium host cells to provide the recombinant Sorangium host cells of the invention.
U.S. Pat. No.
5,686,295, incorporated herein by reference, describes a method for transforming Sorangium host cells, although other methods may also be employed. Sorangium is a WO 00/31247 _ 5~ _ PCT/US99/27438 convenient host for expressing epothilone derivatives of the invention in which the recombinant PKS that produces such derivatives is expressed from a recombinant vector in which the epothilone PKS gene promoter is positioned to drive expression of the recombinant coding sequence. The epothilone PKS gene promoter is provided in recombinant form by the present invention and is an important embodiment thereof. The promoter is contained within an 500 nucleotide sequence between the end of the transposon sequences and the start site of the open reading frame of the epoA
gene.
Optionally, one can include sequences from further upstream of this 500 by region in the promoter. Those of skill in the art will recognize that, if a Sorangium host that produces epothilone is used as the host cell, the recombinant vector need drive expression of only a portion of the PKS containing the altered sequences. Thus, such a vector may comprise only a single altered epothilone PKS gene, with the remainder of the epothilone PKS
polypeptides provided by the genes in the host cell chromosomal DNA. If the host cell naturally produces an epothilone, the epothilone derivative will thus be produced in a mixture containing the naturally occurring epothilone(s).
Those of skill will also recognize that the recombinant DNA compounds of the invention can be used to construct Sorangium host cells in which one or more genes involved in epothilone biosynthesis have been rendered inactive. Thus, the invention provides such Sorangium host cells, which may be preferred host cells for expressing epothilone derivatives of the invention so that complex mixtures of epothilones are avoided. Particularly preferred host cells of this type include those in which one or more of any of the epothilone PKS gene ORFs has been disrupted, and/or those in which any or more of the epothilone modification enzyme genes have been disrupted. Such host cells are typically constructed by a process involving homologous recombination using a vector that contains DNA homologous to the regions flanking the gene segment to be altered and positioned so that the desired homologous double crossover recombination event desired will occur.
Homologous recombination can thus be used to delete, disrupt, or alter a gene.
In a preferred illustrative embodiment, the present invention provides a recombinant epothilone producing Sorangium cellulosum host cell in which the epoK gene has been deleted or disrupted by homologous recombination using a recombinant DNA
vector of the invention. This host cell, unable to make the epoK epoxidase gene product is unable to make epothilones A and B and so is a preferred source of epothilones C and D.

WO 00/31247 _ 51 _ PCTNS99/27438 Homologous recombination can also be used to alter the specificity of a PKS
module by replacing coding sequences for the module or domain of a module to be altered with those specifying a module or domain of the desired specificity. In another preferred illustrative embodiment, the present invention provides a recombinant epothilone producing Sorangium cellulosum host cell in which the coding sequence for the AT
domain of module 4 encoded by the epoD gene has been altered by homologous recombination using a recombinant DNA vector of the invention to encode an AT
domain that binds only methylmalonyl CoA. This host cell, unable to make epothilones A, C, and E is a preferred source of epothilones B, D, and F. The invention also provides recombinant Sorangium host cells in which both alterations and deletions of epothilone biosynthetic genes have been made. For example, the invention provides recombinant Sorangium cellulosum host cells in which both of the foregoing alteration and deletion have been made, producing a host cell that makes only epothilone D.
In similar fashion, those of skill in the art will appreciate the present invention 1 S provides a wide variety of recombinant Sorangium cellulosum host cells that make less complex mixtures of the epothilones than do the wild type producing cells as well as those that make one or more epothilone derivatives. Such host cells include those that make only epothilones A, C, and E; those that make only epothilones B, D, and F, those that make only epothilone D; and those that make only epothilone C.
In another preferred embodiment, the present invention provides expression vectors and recombinant Myxococcus, preferably M. xanthus, host cells containing those expression vectors that express a recombinant epothilone PKS or a PKS for an epothilone derivative. Presently, vectors that replicate extrachromosomally in M. xanthus are not known. There are, however, a number of phage known to integrate into M.
xanthus chromosomal DNA, including MxB, Mx9, Mx8l, and Mx82. The integration and attachment function of these phages can be placed on plasmids to create phage-based expression vectors that integrate into the M. xanthus chromosomal DNA. Of these, phage Mx9 and Mx8 are preferred for purposes of the present invention. Plasmid pPLH343, described in Salmi et al., Feb. 1998, Genetic determinants of immunity and integration of temperate Myxococcus xanthus phage Mx8, J. Bact. 180(3): 614-621, is a plasmid that replicates in E. coli and comprises the phage Mx8 genes that encode the attachment and integration functions.

The promoter of the epothilone PKS gene functions in Myxococcus xanthus host cells. Thus, in one embodiment, the present invention provides a recombinant promoter for use in recombinant host cells derived from the promoter of the Sorangium celludosum epothilone PKS gene. The promoter can be used to drive expression of one or more epothilone PKS genes or another useful gene product in recombinant host cells.
The invention also provides an epothilone PKS expression vector in which one or more of the epothilone PKS or epothilone modification enzyme genes are under the control of their own promoter. Another preferred promoter for use in Myxococcus xanthus host cells for purposes of expressing a recombinant PKS of the invention is the promoter of the pilA
gene of M. xanthus. This promoter, as well as two M. xanthus strains that express high levels of gene products from genes controlled by the pilA promoter, a pilA
deletion strain and a pilS deletion strain, are described in Wu and Kaiser, Dec. 1997, Regulation of expression of the pilA gene in Myxococcus xanthus, J. Bact. 179(24):7748-?758, incorporated herein by reference. Optionally, the invention provides recombinant Myxococcus host cells comprising both the pilA and pilS deletions. Another preferred promoter is the starvation dependent promoter of the sdcK gene.
Selectable markers for use in Myxococcus xanthus include kanamycin, tetracycline, chloramphenicol, zeocin, spectinomycin, and streptomycin resistance conferring genes.
The recombinant DNA expression vectors of the invention for use in Myxococcus typically include such a selectable marker and may further comprise the promoter derived from an epothilone PKS or epothilone modification enzyme gene.
The present invention provides preferred expression vectors for use in preparing the recombinant Myxococcus xanthus expression vectors and host cells of the invention.
These vectors, designated plasmids pKOS35-82.1 and pKOS35-82.2 (Figure 3), are able to replicate in E. coli host cells as well as integrate into the chromosomal DNA
of M. xanthus. The vectors comprise the Mx8 attachment and integration genes as well as the pilA promoter with restriction enzyme recognition sites placed conveniently downstream.
The two vectors differ from one another merely in the orientation of the pilA
promoter on the vector and can be readily modified to include the epothilone PKS and modification enzyme genes of the invention. The construction of the vectors is described in Example 2.
Especially preferred Myxococcus host cells of the invention are those that produce an epothilone or epothilone derivative or mixtures of epothilones or epothilone derivatives at equal to or greater than 20 mg/L, more preferably at equal to or greater than 200 mg/L, and most preferably at equal to or greater than 1 g/L. Especially preferred are M. xanthus host cells that produce at these levels. M. xanthus host cells that can be employed for purposes of the invention include the DZ1 (Campos et al., 1978, J. Mol. Biol.
119: 167-178, incorporated herein by reference), the TA-producing cell line ATCC 31046, (Hodgkin and Kaiser, 1979, MoI. Gen. Genet. 171: 177-191, incorporated herein by reference), and the DK1622 cell lines (Kaiser, 1979, Proc. Natl. Acad. Sci.
USA 76: 5952-5956, incorporated herein by reference}.
In another preferred embodiment, the present invention provides expression vectors and recombinant Pseudomonas fluorescens host cells that contain those expression vectors and express a recombinant PKS of the invention. A plasmid for use in constructing the P. fluorescens expression vectors and host cells of the invention is plasmid pRSF1010, which replicates in E. coli and P. fluorescens host cells (see Scholz et al., 1989, Gene 75:271-8, incorporated herein by reference). Low copy number replicons and vectors can also be used. As noted above, the invention also provides the promoter of the Sorangium cellulosum epothilone PKS and epothilone modification enzyme genes in recombinant form. The promoter can be used to drive expression of an epothilone PKS gene or other gene in P. fluorescens host cells. Also, the promoter of the soraphen PKS
genes can be used in any host cell in which a Sorangium promoter functions. Thus, in one embodiment, the present invention provides an epothilone PKS expression vector for use in P.
fluorescens host cells.
In another preferred embodiment, the expression vectors of the invention are used to construct recombinant Streptomyces host cells that express a recombinant PKS of the invention. Streptomyces host cells useful in accordance with the invention include S. coelicolor, S. lividans, S. venezuelae, S ambofaciens, S. fradiae, and the like. Preferred Streptomyces host cell/vector combinations of the invention include S.
coelicolor CH999 and S. lividans K4-114 and K4-155 host cells, which do not produce actinorhodin, and expression vectors derived from the pRMI and pRMS vectors, as described in U.S. Patent No. 5,830,750 and U.S. patent application Serial Nos. 08/828,898, filed 31 Mar. 1997, and 09/181,833, filed 28 Oct. 1998. Especially preferred Streptomyces host cells ofthe invention are those that produce an epothilone or epothilone derivative or mixtures of epothilones or epothilone derivatives at equal to or greater than 20 mglL, more preferably at equal to or greater than 200 mg/L, and most preferably at equal to or greater than 1 g/L.
Especially preferred are S. coelicolor and S. lividans host cells that produce at these levels.

Also, species of the closely related genus Saccharopolyspora can be used to produce epothilones, including but not limited to S. erythraea.
The present invention provides a wide variety of expression vectors for use in Streptomyces. For replicating vectors, the origin of replication can be, for example and without limitation, a low copy number replicon and vectors comprising the same, such as SCP2* (see Hopwood et al., Genetic Manipulation of Streptomyces: A Laboratory manual (The John Innes Foundation, Norwich, U.K., 1985); Lydiate et al., 1985, Gene 35: 223-235; and Kieser and Melton, 1988, Gene 65: 83-91, each of which is incorporated herein by reference), SLP1.2 (Thompson et al., 1982, Gene 20: 51-62, incorporated herein by reference), and pSGS(ts) (Muth et al., 1989, Mol. Gen. Genet. 219: 341-348, and Bierman et al., 1992, Gene 116: 43-49, each of which is incorporated herein by reference), or a high copy number replicon and vectors comprising the same, such as pIJ101 and pJV 1 (see Katz et al., 1983, J. Gen. Microbiol. 129: 2703-2714; Vara et al., 1989, J.
Bacteriol. 171:
5782-5781; and Servin-Gonzalez, 1993, Plasmid 30: 131-140, each of which is incorporated herein by reference). High copy number vectors are generally, however, not preferred for expression of large genes or multiple genes. For non-replicating and integrating vectors and generally for any vector, it is useful to include at least an E. coli origin of replication, such as from pUC, plP, plI, and pBR. For phage based vectors, the phage phiC31 and its derivative KC515 can be employed (see Hopwood et al., supra).
Also, plasmid pSET152, plasmid pSAM, plasmids pSE101 and pSE211, all of which integrate site-specifically in the chromosomal DNA of S. lividans, can be employed.
Typically, the expression vector will comprise one or more marker genes by which host cells containing the vector can be identified and/or selected. Useful antibiotic resistance conferring genes for use in Streptomyces host cells include the ermE (confers resistance to erythromycin and lincomycin), tsr (confers resistance to thiostrepton), aadA
(confers resistance to spectinomycin and streptomycin), aacC4 (confers resistance to apramycin, kanamycin, gentamicin, geneticin (G418), and neomycin), hyg (confers resistance to hygromycin), and vph (confers resistance to viomycin) resistance confernng genes.
The recombinant PKS gene on the vector will be under the control of a promoter, typically with an attendant ribosome binding site sequence. A preferred promoter is the actI promoter and its attendant activator gene actII-ORF4, which is provided in the pRMI
and pRMS expression vectors, supra. This promoter is activated in the stationary phase of WO 00/31247 _ 55 _ PCT/US99/27438 growth when secondary metabolites are normally synthesized. Other useful Streptomyces promoters include without limitation those from the ermE gene and the melC 1 gene, which act constitutively, and the tipA gene and the merA gene, which can be induced at any growth stage. In addition, the T7 RNA polymerase system has been transferred to Streptomyces and can be employed in the vectors and host cells of the invention. In this system, the coding sequence for the T7 RNA polymerase is inserted into a neutral site of the chromosome or in a vector under the control of the inducible merA
promoter, and the gene of interest is placed under the control of the T7 promoter. As noted above, one or more activator genes can also be employed to enhance the activity of a promoter.
Activator genes in addition to the actII-ORF4 gene discussed above include dnrI, redD, and ptpA genes (see U.S. patent application Serial No. 09/181,833, supra), which can be employed with their cognate promoters to drive expression of a recombinant gene of the invention.
The present invention also provides recombinant expression vectors that drive I 5 expression of the epothilone PKS and PKS enzymes that produce epothilone or epothilone derivatives in plant cells. Such vectors are constructed in accordance with the teachings in U.S. patent application Serial No. 09/114,083, filed 10 July 1998, and PCT
patent publication No. 99/02669, each of which is incorporated herein by reference.
Plants and plant cells expressing epothilone are disease resistant and able to resist fungal infection.
For improved production of an epothilone or epothilone derivative in any heterologous host cells, including plant, Myxococcus, Pseudomonas, and Streptomyces host cells, one can also transform the cell to express a heterologous phosphopantetheinyl transferase. See U.S. patent application Serial No. 08/728,742, filed 11 Oct. 1996, and PCT
patent publication No. 97/13845, both of which are incorporated herein by reference.
In addition to providing recombinant expression vectors that encode the epothilone or an epothilone derivative PKS, the present invention also provides, as discussed above, DNA compounds that encode epothilone modification enzyme genes. As discussed above, these gene products convert epothilones C and D to epothilones A and B, and convert epothilones A and B to epothilones E and F. The present invention also provides recombinant expression vectors and host cells transformed with those vectors that express any one or more of those genes and so produce the corresponding epothilone or epothilone derivative. In one aspect, the present invention provides the epoK gene in recombinant WO 00/31247 _ 56 - PCT/US99/27438 form and host cells that express the gene product thereof, which converts epothilones C
and D to epothilones A and B, respectively.
In another important embodiment, and as noted above, the present invention provides vectors for disrupting the function of any one or more of the epoL, epoK, and any of the ORFs associated with the epothilone PKS gene cluster in Sorangium cells. The invention also provides recombinant Sorangium host cells lacking (or containing inactivated forms of? any one or more of these genes. These cells can be used to produce the corresponding epothilones and epothilone derivatives that result from the absence of any one or more of these genes.
The invention also provides non-Sorangium host cells that contain a recombinant epothilone PKS or a PKS for an epothilone derivative but do not contain (or contain non-functional forms of) any epothilone modification enzyme genes. These host cells of the invention are expected produce epothilones G and H in the absence of a dehydratase activity capable of forming the C-12-C-13 alkene of epothilones C and D. This dehydration reaction is believed to take place in the absence of the epoL gene product in Streptomyces host cells. The host cells produce epothilones C and D (or the corresponding epothilone C and D derivative) when the dehydratase activity is present and the P450 epoxidase and hydroxylase (that converts epothilones A and B to epothilones E
and F, respectively) genes are absent. The host cells also produce epothilones A and B (or the corresponding epothilone A and B derivatives) when the hydroxylase gene only is absent.
Preferred for expression in these host cells is the recombinant epothilone PKS
enzymes of the invention that contain the hybrid module 4 with an AT specific for methylmalonlyl CoA only, optionally in combination with one or more additional hybrid modules. Also preferred for expression in these host cells is the recombinant epothilone PKS
enzymes of the invention that contain the hybrid module 4 with an AT specific for malonyl CoA only, optionally in combination with one or more additional hybrid modules.
The recombinant host cells of the invention can also include other genes and corresponding gene products that enhance production of a desired epothiIone or epothilone derivative. As but one non-limiting example, the epothilone PKS proteins require phosphopantetheinylation of the ACP domains of the loading domain and modules through 9 as well as of the PCP domain of the NRPS. Phosphopantethein-ylation is mediated by enzymes that are called phosphopantetheinyl transferases (PPTases). To produce functional PKS enzyme in host cells that do not naturally express a PPTase able WO 00/31247 _ 57 _ PCT/US99/27438 to act on the desired PKS enzyme or to increase amounts of functional PKS
enzyme in host cells in which the PPTase is rate-limiting, one can introduce a heterologous PPTase, including but not limited to Sfp, as described in PCT Pat. Pub. Nos. 97/13845 and 98/27203, and U.S. patent application Serial Nos. 08/728,742, filed 11 Oct.
1996, and 08/989,332, each of which is incorporated herein by reference.
The host cells of the invention can be grown and fermented under conditions known in the art for other purposes to produce the compounds of the invention.
The compounds of the invention can be isolated from the fermentation broths of these cultured cells and purified by standard procedures. Fermentation conditions for producing the compounds of the invention from Sorangium host cells can be based on the protocols described in PCT patent publication Nos. 93/10121, 97/19086, 98/22461, and 99/42602, each of which is incorporated herein by reference. The novel epothilone analogs of the present invention, as well as the epothilones produced by the host cells of the invention, can be derivatized and formulated as described in PCT patent publication Nos.
93/10121, IS 97/19086, 98/08849, 98/22461, 98/25929, 99/01124, 99/02514, 99/07692, 99/27890, 99/39694, 99/40047, 99/42602, 99/43653, 99/43320, 99/54319, 99/54319, and 99/54330, and U.S. Patent No. 5,969,145, each of which is incorporated herein by reference.
Invention Compounds Preferred compounds of the invention include the 14-methyl epothilone derivatives (made by utilization of the hybrid module 3 of the invention that has an AT
that binds methylmalonyl CoA instead of malonyl CoA); the 8,9-dehydro epothilone derivatives (made by utilization of the hybrid module 6 of the invention that has a DH and KR instead of an ER, DH, and KR); the 10-methyl epothilone derivatives (made by utilization of the hybrid module 5 of the invention that has an AT that binds methylmalonyl CoA
instead of malonyl CoA); the 9-hydroxy epothilone derivatives (made by utilization of the hybrid module 6 of the invention that has a KR instead of an ER, DH, and KR); the 8-desmethyl-14-methyl epothilone derivatives (made by utilization of the hybrid module 3 of the invention that has an AT that binds methylmalonyl CoA instead of malonyl CoA
and a hybrid module 6 that binds malonyl CoA instead of methylmalonyl CoA ); and the desmethyl-8,9-dehydro epothilone derivatives (made by utilization of the hybrid module 6 of the invention that has a DH and KR instead of an ER, DH, and KR and an AT
that specifies malonyl CoA instead of methylmalonyl CoA).

WO 00/31247 _ Sg _ PCT/US99/27438 More generally, preferred epothilone derivative compounds of the invention are those that can be produced by altering the epothilone PKS genes as described herein and optionally by action of epothilone modification enzymes and/or by chemically modifying the resulting epothilones produced when those genes are expressed. Thus, the present invention provides compounds of the formula:
1z ~10 X13 Xs 1a X11 a 7 /16 R14 Ra ,.X
1s R2 _ i ~ ' OH
(1) including the glycosylated forms thereof and stereoisomeric forms where the stereochemistry is not shown, wherein A is a substituted or unsubstituted straight, branched chain or cyclic alkyl, alkenyl or alkynyl residue optionally containing 1-3 heteroatoms selected from O, S and N; or wherein A comprises a substituted or unsubstituted aromatic residue;
R2 represents H,H, or H,lower alkyl, or lower alkyl,lower alkyl;
XS represents =O or a derivative thereof, or H,OH or H,NR2 wherein R is H, or alkyl, or acyl or H,OCOR or H,OCONR2 wherein R is H or alkyl, or is H,H;
R6 represents H or lower alkyl, and the remaining substituent on the corresponding carbon is H;
X' represents OR, NR2, wherein R is H, or alkyl or acyl or is OCOR, or OCONR2 wherein R is H or alkyl or X' taken together with X9 forms a carbonate or carbamate cycle, and wherein the remaining substituent on the corresponding carbon is H;
R8 represents H or lower alkyl and the remaining substituent on the carbon is H;
X9 represents =O or a derivative thereof, or is H,OR or H,NR2, wherein R is H, or alkyl or acyl or is H,OCOR or H,OCONR2 wherein R is H or alkyl, or represents H,H or wherein X9 together with X' or with X' 1 can form a cyclic carbonate or carbamate;
R~° is H,H or H,lower alkyl, or lower alkyl,lower alkyl;

WO 00/31247 _ 59 _ PCT/US99/27438 X' ~ is =O or a derivative thereof, or is H,OR, or H,NR2 wherein R is H, or alkyl or acyl or is H,OCOR or H,OCONR2 wherein R is H or alkyl, or is H,H or wherein X11 in combination with X9 may form a cyclic carbonate or carbamate;
R'2 is H,H, or H,lower alkyl, or lower alkyl,lower alkyl;
X'3 is =O or a derivative thereof, or H,OR or H,NR2 wherein R is H, alkyl or acyl or is H,OCOR or H,OCONR2 wherein R is H or alkyl;
R'4 is H,H, or H,lower alkyl, or lower alkyl,lower alkyl;
R~6 is H or lower alkyl; and wherein optionally H or another substituent may be removed from positions 12 and 13 and/or 8 and 9 to form a double bond, wherein said double bond may optionally be converted to an epoxide.
Particularly preferred are compounds of the formulas R12 R1o " ,o 13 ~ 9 A ~ ,a 14 X11 ,s R R
,s r.,p , 2 3 a,, s 0 pH Xs 1 (a), A
1(b) and WO 00/31247 - (Q _ PCT/US99/2'1438 R~2 Rio O, ,z ~ ,o R~s ,3 ~ 9 A
/,s ~4 R~4 X R8 a ,.X' ,5 R2 , 2 3 4,, 5 6 'R
O OH Xs 1(c) wherein the noted substituents are as defined above.
Especially preferred are compounds of the formulas R~2 Rto Z O
~z " ,o R~s ,3 ~ s ,4 ,a X~~
/,s R~4 R8 . Z
,s R2 O , 2 3 4,' 5 T ~R
O OH Xs 1(d) and R~s ,a .~ ~~ s A ,a Z
/,s R~4 Re s ~ X7 ~s R2 , z .
1 2 3 4, 5 Rs 1 (e) wherein both Z are O or one Z is N and the other Z is O, and the remaining substitueaE
are as defined above.
As used herein, a substituent which "comprises an aromatic moiety" contains!
least one aromatic ring, such as phenyl, pyridyl, pyrimidyl, thiophenyl, or thiazoiyl.1e substituent may also include fused aromatic residues such as naphthyl, indolyl, benzothiazolyl, and the like. The aromatic moiety may also be fused to a nonaromati~ng WO 00/31247 _ 61 - PCTNS99/27438 and/or may be coupled to the remainder of the compound in which it is a substituent through a nonaromatic, for example, alkylene residue. The aromatic moiety may be substituted or unsubstituted as may the remainder of the substituent.
Preferred embodiments of A include the "R" groups shown in Figure 2.
As used herein, the term alkyl refers to a C~-C8 saturated, straight or branched chain hydrocarbon radical derived from a hydrocarbon moiety by removal of a single hydrogen atom. Alkenyl and alkynyl refer to the corresponding unsaturated forms.
Examples of alkyl include but are not limited to methyl, ethyl, propyl, isopropyl, n-butyl, tert-butyl, neopentyl, i-hexyl, n-heptyl, n-octyl. Lower alkyl (or alkenyl or alkynyl) refers to a 1-4C radical. Methyl is preferred. Acyl refers to alkylCO, alkenylCO or alkynylCO.
The terms halo and halogen as used herein refer to an atom selected from fluorine, chlorine, bromine, and iodine. The term haloalkyl as used herein denotes an alkyl group to which one, two, or three halogen atoms are attached to any one carbon and includes without limitation chloromethyl, bromoethyl, trifluoromethyl, and the like.
The term heteroaryl as used herein refers to a cyclic aromatic radical having from five to ten ring atoms of which one ring atom is selected from S, O, and N;
zero, one, or two ring atoms are additional heteroatoms independently selected from S, O, and N; and the remaining ring atoms are carbon, the radical being joined to the rest of the molecule via any of the ring atoms, such as, for example, pyridyl, pyrazinyl, pyrimidinyl, pyrrolyl, pyrazolyl, imidazolyl, thiazolyl, oxazolyl, isoxazolyl, thiadiazolyl, oxadiazolyl, thiophenyl, furanyl, quinolinyl, isoquinolinyl, and the like.
The term heterocyle includes but is not limited to pyrrolidinyl, pyrazolinyl, pyrazolidinyl, imidazolinyl, imidazolidinyl, piperidinyl, piperazinyl, oxazolidinyl, isoxazolidinyl, morpholinyl, thiazolidinyl, isothiazolidinyl, and tetrahydrofuryl.
The term "substituted" as used herein refers to a group substituted by independent replacement of any of the hydrogen atoms thereon with, for example, Cl, Br, F, I, OH, CN, alkyl, alkoxy, alkoxy substituted with aryl, haloalkyl, alkylthio, amino, alkylamino, dialkylamino, mercapto, nitro, carboxaldehyde, carboxy, alkoxycarbonyl, or carboxamide.
Any one substituent may be an aryl, heteroaryl, or heterocycloalkyl group.
It will apparent that the nature of the substituents at positions 2, 4, 6, 8, 10, 12, 14 and 16 in formula ( 1 ) is determined at least initially by the specificity of the AT catalytic domain of modules 9, 8, 7, 6, 5, 4, 3 and 2, respectively. Because AT domains that accept malonyl CoA, methylmalonyl CoA, ethylmalonyl CoA (and in general, lower alkyl malonyl CoA), as well as hydroxymalonyl CoA, are available, one of the substituents at these positions may be H, and the other may be H, lower alkyl, especially methyl and ethyl, or OH. Further reaction at these positions, e.g., a methyl transferase reaction such as that catalyzed by module 8 of the epothilone PKS, may be used to replace H at these positions as well. Further, an H,OH embodiment may be oxidized to =O or, with the adjacent ring C, be dehydrated to form a ~-bond. Both OH and =O are readily derivatized as further described below.
Thus, a wide variety of embodiments of R2, R6, Rg, R'°, R'Z, R'4 and R'6 is synthetically available. The restrictions set forth with regard to embodiments of these substituents set forth in the definitions with respect to Formula ( 1 ) above reflect the information described in the SAR description in Example 8 below.
Similarly, (i-carbonyl modifications (or absence of modification) can readily be controlled by modifying the epothilone PKS gene cluster to include the appropriate sequences in the corresponding positions of the epothilone gene cluster which will or will not contain active KR, DH and/or ER domains. Thus, the embodiments of X5, X', X9, X"
and X13 synthetically available are numerous, including the formation of ~-bonds with the adjacent ring positions.
Positions occupied by OH are readily converted to ethers or esters by means well known in the art; protection of OH at positions not to be derivatized may be required.
Further, a hydroxyl may be converted to a leaving group, such as a tosylate, and replaced by an amino or halo substituent. A wide variety of "hydroxyl derivatives" such as those discussed above is known in the art.
Similarly, ring positions which contain oxo groups may be converted to "carbonyl derivatives" such as oximes, ketals, and the like. Initial reaction products with the oxo moieties may be further reacted to obtain more complex derivatives. As described in Example 8, such derivatives may ultimately result in a cyclic substituent linking two ring positions.
The enzymes useful in modification of the polyketide initially synthesized, such as transmethylases, dehydratases, oxidases, glycosylation enzymes and the like, can be supplied endogenously by a host cell when the polyketide is synthesized intracellularly, by modifying a host to contain the recombinant materials for the production of these modifying enzymes, or can be supplied in a cell-free system, either in purified forms or as relatively crude extracts. Thus, for example, the epoxidation of the ~c-bond at position 12-13 may be effected using the protein product of the epoK gene directly in vitro.
The nature of A is most conveniently controlled by employing an epothilone PKS
which comprises an inactivated module I NRPS (using a module 2 substrate) or a knockout (using a module 3 substrate) as described in Example 6, hereinbelow.
Limited variation can be obtained by altering the AT catalytic specificity of the loading module;
further variation is accomplished by replacing the NRPS of module 1 with an NRPS of different specificity or with a conventional PKS module. However, at present, variants are more readily prepared by feeding the synthetic module 2 substrate precursors and module 3 substrate precursors to the appropriately altered epothilone PKS as described in Example 6.
Pharmaceutical Compositions The compounds can be readily formulated to provide the pharmaceutical compositions of the invention. The pharmaceutical compositions of the invention can be used in the form of a pharmaceutical preparation, for example, in solid, semisolid, or liquid form. This preparation will contain one or more of the compounds of the invention as an active ingredient in admixture with an organic or inorganic Garner or excipient suitable for external, enteral, or parenteral application. The active ingredient may be compounded, for example, with the usual non-toxic, pharmaceutically acceptable Garners for tablets, pellets, capsules, suppositories, pessaries, solutions, emulsions, suspensions, and any other form suitable for use.
The carriers which can be used include water, glucose, lactose, gum acacia, gelatin, mannitol, starch paste, magnesium trisilicate, talc, corn starch, keratin, colloidal silica, potato starch, urea, and other carriers suitable for use in manufacturing preparations, in solid, semi-solid, or liquified form. In addition, auxiliary stabilizing, thickening, and coloring agents and perfumes may be used. For example, the compounds of the invention may be utilized with hydroxypropyl methylcellulose essentially as described in U.S. Patent No. 4,916,138, incorporated herein by reference, or with a surfactant essentially as described in EPO patent publication No. 428,169, incorporated herein by reference.
Oral dosage forms may be prepared essentially as described by Hondo et al., 1987, Transplantation Proceedings XIX, Supp. 6: 17-22, incorporated herein by reference.
Dosage forms for external application may be prepared essentially as described in EPO

patent publication No. 423,714, incorporated herein by reference. The active compound is included in the pharmaceutical composition in an amount su~cient to produce the desired effect upon the disease process or condition.
For the treatment of conditions and diseases caused by infection, immune system S disorder (or to suppress immune function), or cancer, a compound of the invention may be administered orally, topically, parenteraIly, by inhalation spray, or rectally in dosage unit formulations containing conventional non-toxic pharmaceutically acceptable carriers, adjuvant, and vehicles. The term parenteral, as used herein, includes subcutaneous injections, and intravenous, intrathecal, intramuscular, and intrasternal injection or infusion techniques.
Dosage levels of the compounds of the present invention are of the order from about 0.01 mg to about 100 mg per kilogram of body weight per day, preferably from about 0.1 mg to about 50 mg per kilogram of body weight per day. The dosage levels are useful in the treatment of the above-indicated conditions (from about 0.7 mg to about 3.5 mg per patient per day, assuming a 70 kg patient). In addition, the compounds of the present invention may be administered on an intermittent basis, i.e., at semi-weekly, weekly, semi-monthly, or monthly intervals.
The amount of active ingredient that may be combined with the Garner materials to produce a single dosage form will vary depending upon the host treated and the particular mode of administration. For example, a formulation intended for oral administration to humans may contain from 0.5 mg to 5 gm of active agent compounded with an appropriate and convenient amount of Garner material, which may vary from about 5 percent to about 95 percent of the total composition. Dosage unit forms will generally contain from about 0.5 mg to about 500 mg of active ingredient. For external administration, the compounds of the invention may be formulated within the range of, for example, 0.00001 %
to 60% by weight, preferably from 0.001 % to 10% by weight, and most preferably from about 0.005% to 0.8% by weight.
It will be understood, however, that the specific dose level for any particular patient will depend on a vaxiety of factors. These factors include the activity of the specific compound employed; the age, body weight, general health, sex, and diet of the subject; the time and route of administration and the rate of excretion of the drug;
whether a drug combination is employed in the treatment; and the severity of the particular disease or condition for which therapy is sought.

WO 00/31247 _ 65 _ PCT/US99/27438 A detailed description of the invention having been provided above, the following examples are given for the purpose of illustrating the present invention and shall not be construed as being a limitation on the scope of the invention or claims.
Example 1 DNA Sequencing of Cosmid Clones and Subclones Thereof The epothilone producing strain, Sorangium cellulosum SMP44, was grown on a cellulose-containing medium, see Bollag et al., 1995, Cancer Research 55: 2325-2333, incorporated herein by reference, and epothilone production was confirmed by LC/MS
analysis of the culture supernatant. Total DNA was prepared from this strain using the procedure described by Jaoua et al., 1992, Plasmid 28: 157-165, incorporated herein by reference. To prepare a cosmid library, S. cellulosum genomic DNA was partially digested with Sau3AI and ligated with BamHI-digested pSupercos (Stratagene). The DNA
was packaged in lambda phage as recommended by the manufacturer and the mixture then used to infect E. coli XLl-Blue MR cells. This procedure yielded approximately 3,000 isolated colonies on LB-ampicillin plates. Because the size of the S.
cellulosum genome is estimated to be circa 10' nucleotides, the DNA inserts present among 3000 colonies would correspond to circa 10 S. cellulosum genomes.
To screen the library, two segments of KS domains were used to design oligonucleotide primers for a PCR with Sorangium cellulosum genomic DNA as template.
The fragment generated was then used as a probe to screen the library. This approach was chosen, because it was found, from the examination of aver a dozen PKS genes, that KS
domains are the most highly conserved (at the amino acid level) of all the PKS
domains examined. Therefore, it was expected that the probes produced would detect not only the epothilone PKS genes but also other PKS gene clusters represented in the library. The two degenerate oligonucleotides synthesized using conserved regions within the ketosynthase (KS) domains compiled from the DEBS and soraphen PKS gene sequences were (standard nomenclature for degenerate positions is used): CTSGTSKCSSTBCACCTSGCSTGC and TGAYRTGSGCGTTSGTSCCGSWGA. A single band of 750 bp, corresponding to the predicted size, was seen in an agarose gel after PCR employing the oligos as primers and S. cellulosum SMP44 genomic DNA as template. The fragment was removed from the gel and cloned in the HincII site of pUC 118 (which is a derivative of pUC 18 with an insert sequence for making single stranded DNA). After transformation of E. coli, plasmid DNA

from ten independent clones was isolated and sequenced. The analysis revealed nine unique sequences that each corresponded to a common segment of KS domains in PKS
genes. Of the nine, three were identical to a polyketide synthase gene cluster previously isolated from this organism and determined not to belong to the epothilone gene cluster from the analysis of the modules. The remaining six KS fragments were excised from the vector, pooled, end-labeled with 32P and used as probe in hybridizations with the colonies containing the cosmid library under high stringency conditions.
The screen identified 15 cosmids that hybridized to the pooled KS probes. DNA
was prepared from each cosmid, digested with NotI, separated on an agarose gel, and transferred to a nitrocellulose membrane for Southern hybridization using the pooled KS
fragments as probe. The results revealed that two of the cosmids did not contain KS-hybridizing inserts, leaving 13 cosmids to analyze further. The blot was stripped of the label and re-probed, under less stringent conditions, with labeled DNA
containing the sequence corresponding to the enoylreductase domain from module four of the DEBS
gene cluster. Because it was anticipated that the epothilone PKS gene cluster would encode two consecutive modules that contain an ER domain, and because not all PKS
gene clusters have ER domain-containing modules, hybridization with the ER
probe was predicted to identify cosmids containing insert DNA from the epothilone PKS
gene cluster. Two cosmids were found to hybridize strongly to the ER probe, one hybridized moderately, and a final cosmid hybridized weakly. Analysis of the restriction pattern of the NotI fragments indicated that the two cosmids that hybridized strongly with the ER
probe overlapped one another. The nucleotide sequence was also obtained from the ends of each of the 13 cosmids using the T7 and T3 primer binding sites. All contained sequences that showed homology to PKS genes. Sequence from one of the cosmids that hybridized strongly to the ER probe showed homology to NRPSs and, in particular, to the adenylation domain of an NRPS. Because it was anticipated that the thiazole moiety of epothilone might be derived from the formation of an amide bond between an acetate and cysteine molecule (with a subsequent cyclization step), the presence of an NRPS domain in a cosmid that also contained ER domains) supported the prediction that this cosmid might contain all or part of the epothilone PKS gene cluster.
Preliminary restriction analysis of the 12 remaining cosmids suggested that three might overlap with the cosmid of interest. To verify this, oligonucleotides were synthesized for each end of the four cosmids (determined from the end sequencing WO 00/31247 - (7 _ PCT/US99/27438 described above) and used as primer sets in PCRs with each of the four cosmid DNAs.
Overlap would be indicated by the appearance of a band from a non-cognate primer-template reaction. The results of this experiment verified that two of the cosmids overlapped with the cosmid containing the NRPS. Restriction mapping of the three S cosmids revealed that the cosmids did, in fact, overlap. Furthermore, because PKS
sequences extended to the end of the insert in the last overlapping fragment, based on the assumption that the NRPS would map to the 5'-end of the cluster, the results also indicated that the 3' end of the gene cluster had not been isolated among the clones identified.
To isolate the remaining segment of the epathilone biosynthesis genes, a PCR
fragment was generated from the cosmid containing the most 3'-terminal region of the putative gene cluster. This fragment was used as a probe to screen a newly prepared cosmid library of Sorangium cellulosum genomic DNA of again approximately 3000 colonies. Several hybridizing clones were identified; DNA was made from six of them.
Analysis of NotI-digested fragments indicated that all contained overlapping regions. The cosmid containing the largest insert DNA that also had the shortest overlap with the cosmid used to make the probe was selected for further analysis.
Restriction maps were created for the four cosmids, as shown in Figure 1.
Sequence obtained from one of the ends of cosmid pKOS35-70.8A3 showed no homology to PKS sequences or any associated modifying enzymes. Similarly, sequence from one end of cosmid pKOS3S-79.85 also did not contain sequences corresponding to a PKS
region. These findings supported the observation that the epothilone cluster was contained within the ~70 kb region encompassed by the four cosmid inserts.
To sequence the inserts in the cosmids, each of the NotI restriction fragments from the four cosmids was cloned into the NotI site of the commercially available pBluescript 2S plasmid. Initial sequencing was performed on the ends of each of the clones. Analysis of the sequences allowed the prediction, before having the complete sequence, that there would be 10 modules in this PKS gene cluster, a loading domain plus 9 modules.
Sequence was obtained for the complete PKS as follows. Each of the 13 non-overlapping NotI fragments was isolated and subjected to partial HinPI
digestion.
Fragments of ~2 to 4 kb in length were removed from an agarose gel and cloned in the AccI site of pUC 118. Sufficient clones from each library of the NotI
fragments were sequenced to provide at least 4 -fold coverage of each. To sequence across each of the NotI sites, a set of oIigos, one S' and the other 3' to each NotI site, was made and used as WO 00/31247 _ 6g _ PCT/US99/27438 primers in PCR amplification of a fragment that contained each NotI site. Each fragment produced in this manner was cloned and sequenced.
The nucleotide sequence was determined for a linear segment corresponding to ~72 kb. Analysis revealed a PKS gene cluster with a loading domain and nine modules.
Downstream of the PKS sequence is an ORF, designated epoK, that shows strong homology to cytochrome P450 oxidase genes and encodes the epothilone epoxidase. The nucleotide sequence of 15 kb downstream of epoK has also been determined: a number of additional ORFs have been identified but an ORF that shows homology to any known dehydratase has not been identified. The epoL gene may encode a dehydratase activity, but this activity may instead be resident within the epothilone PKS or encoded by another gene.
The PKS genes are organized in 6 open reading frames. At the polypeptide level, the loading domain and modules l, 2, and 9 appear on individual polypeptides;
their corresponding genes are designated epoA, epoB, epoC and epoF respectively.
Modules 3, 4, S, and 6 are contained on a single polypeptide whose gene is designated epoD, and modules 7 and 8 are on another polypeptide whose gene is designated epoE . It is clear from the spacing between ORFs that epoC, epoD, epoE and epoF constitute an operon.
The epoA, epoB, and epoK gene may be also part of the large operon, but there are spaces of approximately 100 by between epoB and epoC and 115 by between epoF and epoK
which could contain a promoter. The present invention provides the intergenic sequences in recombinant form. At Ieast one, but potentially more than one, promoter is used to express all of the epothilone genes. The epothilone PKS gene cluster is shown schematically below.
PKS
epn.1 epo8 epoC epnD epaE epoF' epoK
Load Mod I Mod 2 Mod 3,4,5,&6 Mod 7& 8 Mod 9 (NRPS) P450 A detailed examination of the modules shows an organization and composition that is consistent with one able to be used for the biosynthesis of epothilone. The description that follows is at the polypeptide level. The sequence of the AT domain in the loading module and in modules 3, 4, 5, and 9 shows similarity to the consensus sequence for malonyl loading domains, consistent with the presence of an H side chain at C-14, C-12 WO 00131247 _ 69 _ PCT/US99/27438 (epothilones A and C), C-10, and C-2, respectively, as well as the loading region. The AT
domains in modules 2, 6, 7, and 8 resemble the consensus sequence for methylmalonyl specifying AT domains, again consistent with the presence of methyl side chains at C-16, C-8, C-6, and C-4 respectively.
The loading module contains a KS domain in which the cysteine residue usually present at the active site is instead a tyrosine. This domain is designated as KS'' and serves as a decarboxylase, which is part of its normal function, but cannot function as a condensing enzyme. Thus, the loading domain is expected to load malonyl CoA, move it to the ACP, and decarboxylate it to yield the acetyl residue required for condensation with cysteine.
Module 1 is the non-ribosomal peptide synthetase that activates cysteine and catalyzes the condensation with acetate on the loading module. The sequence contains segments highly similar to ATP-binding and ATPase domains, required for activation of amino acids, a phosphopantotheinylation site, and an elongation domain. In database searches, module 1 shows very high similarity to a number of previously identified peptide synthetases.
Module 2 determines the structure of epothilone at C-15 - C-17. The presence of the DH domain in module 2 yields the C-16-17 dehydro moiety in the molecule.
The domains in module 3 are consistent with the structure of epothilone at C-14 and C-15; the OH that comes from the action of the KR is employed in the lactonization of the molecule.
Module 4 controls the structure at C-12 and C-13 where a double bond is found in epothilones C and D, consistent with the presence of a DH domain. Although the sequence of the AT domain appears to resemble those that specify malonate loading, it can also load methylmalonate, thereby accounting in part for the mixture of epothilones found in the fermentation broths of the naturally producing organisms.
A significant departure from the expected array of functions was found in module 4. This module was expected to contain a DH domain, thereby directing the synthesis of epothilones C and D as the products of the PKS. Rigorous analysis revealed that the space between the AT and KR domains of module 4 was not large enough to accommodate a functional DH domain. Thus, the extent of reduction at module 4 does not proceed beyond the ketoreduction of the beta-keto formed after the condensation directed by module 4.
Because the C-12,13 unsaturation has been demonstrated (epothilones C and D), there must be an additional dehydratase function that introduces the double bond, and this WO 00/31247 _ ~0 _ PCT/US99/27438 function is believed to be in the PKS itself or resident in an ORF in the epothilone biosynthetic gene cluster.
Thus, the action of the dehydratase could occur either during the synthesis of the polyketide or after cyclization has taken place. In the former case, the compounds S produced at the end of acyl chain growth would be epothilones C and D. If the C-12,13 dehydration were a post-polyketide event, the completed acyl chain would have a hydroxyl group at C-13, as shown below. The names epothilones G and H have been assigned to the 13-hydroxy compounds produced in the absence of or prior to the action of the dehydratase.
S
E
)H
Epothilones G (R=H) and H (R=CH3).
Modules 5 and 6 each have the full set of reduction domains (KR, DH and ER) to yield the methylene functions at C-1 l and C-9. Modules 7 and 9 have KR
domains to yield the hydroxyls at C-7 and C-3, and module 8 does not have a functional KR
domain, consistent with the presence of the keto group at C-S. Module 8 also contains a methyltransferase (MT) domain that results in the presence of the geminal dimethyl function at C-4. Module 9 has a thioesterase domain that terminates polyketide synthesis and catalyzes ring closure. The genes, proteins, modules, and domains of the epothilone PKS are summarized in the Table hereinabove.
Inspection of the sequence has revealed translational coupling between epoA
and epoB (loading domain and module 1 ) and between epoC and epoD. Very small gaps are seen between epoD and epoE and epoE and epoF but gaps exceeding 100 by are found between epoB and epoC and epoF and epoK. These intergenic regions may contain promoters. Sequencing efforts have not revealed the presence of regulatory genes, and it is possible that epothilone synthesis is not regulated by operon specific regulation in Sorangium cellulosum.

WO00/31247 _ ~1 _ PCT/US99/27438 The sequence of the epothilone PKS and flanking regions has been compiled into a single contig, as shown below.

lO 421 AGGTCGAGGTCGGAAAGACCCCCTACGCGCGCTTCGATCTCAACGACTACTCGGTCCCCC

WO _ ~2 _ PCT/US99/27438 TGCTGGTGCT ACCGCGGCAG ACACGCGGCG
CGGCTGCGCG

CGGCTCGCGG GGGCTGCGGG

CAGGGACAGA

GGGGGCGCAG

WO _ ~3 _ PCT/US99/27438 10201 GGACGCCTTGGAGGAGATCCTCGTTGCGGT.CGTACGGGAGGTGCTCGGGCTGGAGGTGGT

WO _ ~4 _ PC'T/US99/27438 WO _ ~S _ PCT/US99/27438 WO _ ~6 _ PCT/US99/27438 WO - ~'7 PCT/US99/27438 00/31247 _ WO _ ~g _ PCT/US99/27438 lO 25191 TCGATGTACTGAAGCTGGAGGATCGCAGCGACACCCAGCATGTTCGGTCGTTGGCGTCAG

WO _ ~9 _ PCTIUS99/27438 WO _ gO _ PCT/US99/Z7438 ZS 33121 GAGCGATTCAGGCCGAAGCCGGCTGGTCGCTGCTCGCCGAGCTCGCGGCC,GATGAGACCA

WO00/31247 _ gl _ PCT/US99/27438 WO _ g2 _ PCT/US99/Z7438 WO _ g3 _ PCTNS99/27438 95?81 GCATGTTTTCAAGTGCTGACCGCGCTGCTCACCACGCCGGAATCCATCGAGATTCGGAGG

WO _ g4 _ PCT/US99/27438 45$41 CGGCTGACGGATCTCCACGAACCGGATCTCCCGCGGTCCAGGGCTCCGGTGAATCAAGCG

WO 00/31247 . gS _ PCTNS99/27438 CGAAAAAGAA
CGCGTCTTTG
GAGCAAGAGC

GCTGCCGCTT GCGGACACTC

CTCGACCGGC

WO 00/31247 _ g6 _ PCT/US99/27438 WO 00/31247 _ g~ _ PCTNS99/27438 TCAGGATCTA
CAATCTCGCA
CCACGGAGCG
CTTCTATATG

CGAGTTTCTC
GTCGATCGAG
GGCGCGAGAT
CATGCACATC

GCTGCTCGCC
GCGAGGACGA
CGTCGTCAGG
CCCCGCGTTC

TCCCTCGGGC GCGTGGGTGC
GCGCGAGATG

CGGAGGCTCA CAGCATAGGA
TGAGCCTTCC

AGCGAATCAG
AGTGAGACGA

TTTCCCGCGA

TGAAGGCCGC CCACGACGTG
TCCTGGGTCC

GTCAGTCGAG ATCGAGCGCG

ACATCGAGAT ACCAGTTTCC

ATTCTCCGCT

CACGAGCAGA ACCAGAACAG

CCGCATGCGG ACATCCTTGC CCTCCGCTCG

WO _ $g _ PCT/US99/27438 WO _ g9 _ PCTNS99/27438 WO 00/31247 _ 9O _ PCT1US99/27438 6?861 GCCCCTGGTTCGCCTACTGGATCTTCCCGTACCCCGCGACGATGCAGGTGTGGCCGCAGT

lO 71161 ACCGGAGCATATGATGTTCCGTTTCCGTCACAGCGAGGTTCGCCAGGAGGAGGACACGCT

2S Example 2 Construction of a Myxococcus xanthus Expression Vector The DNA providing the integration and attachment function of phage Mx8 was inserted into commercially available pACYC184 (New England Biolabs). An 2360 by MfeI-SmaI from plasmid pPLH343, described in Salmi et al., Feb. 1998, J. Bact.
180(3):
30 614-62I, was isolated and ligated to the large EcoRI-XmnI restriction fragment of plasmid pACYC 184. The circular DNA thus formed was ~6 kb in size and called plasmid pKOS3S-77.
Plasmid pKOS3S-77 serves as a convenient plasmid for expressing recombinant PKS genes of the invention under the control of the epothilone PKS gene promoter. In one 3S illustrative embodiment, the entire epothilone PKS gene with its homologous promoter is inserted in one or more fragments into the plasmid to yield an expression vector of the invention.
The present invention also provides expression vectors in which the recombinant PKS genes of the invention are under the control of a Myxococcus xanthus promoter. To 40 construct an illustrative vector, the promoter of the pilA gene of M.
xanthus was isolated as a PCR amplification product. Plasmid pSWU3S7, which comprises the pilA gene promoter and is described in Wu and Kaiser, Dec. 1997, J. Bact. 179(24):7748-7758, was mixed with PCR primers Seql and Mxpill primers:

WO 00/31247 - 92 - PCT/US99/2743$
Seql : S'-AGCGGATAACAATTTCACACAGGAAACAGC-3 ; and Mxpill: S'-TTAATTAAGAGAAGGTTGCAACGGGGGGC-3', and amplified using standard PCR conditions to yield an 800 by fragment. This fragment was cleaved with restriction enzyme KpnI and ligated to the large KpnI-EcoRV
restriction fragment of commercially available plasmid pLitmus 28 (New England Biolabs).
The resulting circular DNA was designated plasmid pKOS3S-71B.
The promoter of the pilA gene from plasmid pKOS3S-71 B was isolated as an 800 by EcoRV-SnaBI restriction fragment and ligated with the large MscI
restriction fragment of plasmid pKOS3S-77 to yield a circular DNA ~6.8 kb in size. Because the 800 by fragment could be inserted in either one of two orientations, the ligation produced two plasmids of the same size, which were designated as plasmids pKOS3S-82.1 and pKOS3S-82.2. Restriction site and function maps of these plasmids are presented in Figure 3.
Plasmids pKOS3S-82.1 and pKOS35-82.2 serve as convenient starting materials for the vectors of the invention in which a recombinant PKS gene is placed under the 1 S control of the Myxococcus xanthus pilA gene promoter. These plasmids comprise a single PacI restriction enzyme recognition sequence placed immediately downstream of the transcription start site of the promoter. In one illustrative embodiment, the entire epothilone PKS gene without its homologous promoter is inserted in one or more fragments into the plasmids at the PacI site to yield expression vectors of the invention.
The sequence of the pilA promoter in these plasmids is shown below.
CGACGCAGGTGAAGCTGCTTCGTGTGCTCCAGGAGCGGAAGGTGAAGCCGGTCGGCAGCGCCGCGGAGATTC
CCTTCCAGGCGCGTGTCATCGCGGCAACGAACCGGCGGCTCGAAGCCGAAGTAAAGGCCGGACGCTTTCGTG
AGGACCTCTTCTACCGGCTCAACGTCATCACGTTGGAGCTGCCTCCACTGCGCGAGCGTTCCGGCGACGTGT
CGTTGCTGGCGAACTACTTCCTGTCCAGACTGTCGGAGGAGTTGGGGCGACCCGGTCTGCGTTTCTCCCCCG
2S AGACACTGGGGCTATTGGAGCGCTATCCC.TTCCCAGGCAACGTGCGGCAGCTGCAGAACATGGTGGAGCGGG
CCGCGACCCTGTCGGATTCAGACCTCCTGGGGCCCTCCACGCTTCCACCCGCAGTGCGGGGCGATACAGACC
CCGCCGTGCGTCCCGTGGAGGGCAGTGAGCCAGGGCTGGTGGCGGGCTTCAACCTGGAGCGGCATCTCGACG
ACAGCGAGCGGCGCTATCTCGTCGCGGCGATGAAGCAGGCCGGGGGCGTGAAGACCCGTGCTGCGGAGTTGC
TGGGCCTTTCGTTCCGTTCATTCCGCTACCGGTTGGCCAAGCATGGGCTGACGGATGACTTGGAGCCCGGGA

TCCTCGCCGAGGGGATTGTTCCAAGCCTTGAGAATTGGGGGGCTTGGAGTGCGCACCTGGGTTGGCATGCGT
AGTGCTAATCCCATCCGCGGGCGCAGTGCCCCCCGTTGCAACCTTCTCTTAATTAA
To make the recombinant Myxococcus xanthus host cells of the invention, M. xanthus cells are grown in CYE media (Campos and Zusman, 1975, Regulation of 3S development in Myxococcus xanthus: effect of 3': S'-cyclic AMP, ADP, and nutrition, Proc. Natl. Acad. Sci. USA 72: S 18-S22) to a Klett of 100 at 30°C at 300 rpm. The remainder of the protocol is conducted at 2S°C unless otherwise indicated. The cells are then pelleted by centrifugation (8000 rpm for 10 min. in an SS34 or SA600 rotor) and WO 00/31247 - 93 " PCT/US99/27438 resuspended in deionized water. The cells are again pelleted and resuspended in 1/100th of the original volume.
DNA (one to two ~,L) is electroporated into the cells in a 0.1 cm cuvette at room temperature at 400 ohm, 25 ~,FD, 0.65 V with a time constant in the range of 8.8 - 9.4. The DNA should be free of salts and so should be resuspended in distilled and deionized water or dialyzed on a 0.025 pm Type VS membrane (Millipore). For low efficiency electroporadons, spot dialyze the DNA, and allow outgrowth in CYE. Immediately after electroporation, add 1 mL of CYE, and pool the cells in the cuvette with an additional 1.5 mL of CYE previously added to a 50 mL Erlenmeyer flask (total volume 2.5 ml).
Allow the cells to grow for four to eight hours (or oven~ight) at 30 to 32°C
at 300 rpm to allow for expression of the selectable marker. Then, plate the cells in CYE soft agar on plates with selection. If kanamycin is the selectable marker, then typical yields are 103 to 105 per pg of DNA. If streptomycin is the selectable marker, then it must be included in the top agar, because it binds agar.
With this procedure, the recombinant DNA expression vectors of the invention are electroporated into Myxococcus host cells that express recombinant PKSs of the invention and produce the epothilone, epothilone derivatives, and other novel polyketides encoded thereby.
Example 3 Construction of a Bacterial Artificial Chromosome (BAC) for Expression of Epothilone in Myxococcus xanthus To express the epothilone PKS and modification enzyme genes in a heterologous host to produce epothilones by fermentation, Myxococcus xanthus, which is closely related to Sorangium cellulosum and for which a number of cloning vectors are available, can also be employed in accordance with the methods of the invention. Because both M.
xanthus and S. cellulosum are myxobacteria, it is expected that they share common elements of gene expression, translational control, and post translational modification (if any), thereby enhancing the likelihood that the epo genes from S cellulosum can be expressed to produce epothilone in M. xanthus. Secondly, M. xanthus has been developed for gene cloning and expression. DNA can be introduced by electroporation, and a number of vectors and genetic markers are available for the introduction of foreign DNA, including those that permit its stable insertion into the chromosome. Finally, M.
xanthus can be WO 00/31247 - 94 _ PCT/US99/27438 grown with relative ease in complex media in fermentors and can be subjected to manipulations to increase gene expression, if required.
To introduce the epothilone gene cluster into Myxococcus xanthus, one can build the epothilone cluster into the chromosome by using cosmids of the invention and homologous recombination to assemble the complete gene cluster. Alternatively, the complete epothilone gene cluster can be cloned on a bacterial artificial chromosome (BAC) and then moved into M. xanthus for integration into the chromosome.
To assemble the gene cluster from cosmids pKOS35-70.1A2, and pKOS35-79.85, small regions of homology from these cosmids have to be introduced into Myxococcus xanthus to provide recombination sites for larger pieces of the gene cluster.
As shown in Figure 4, plasmids pKOS35-154 and pKOS90-22 are created to introduce these recombination sites. The strategy for assembling the epothilone gene cluster in the M. xanthus chromosome is shown in Figure 5. Initially, a neutral site in the bacterial chromosome is chosen that does not disrupt any genes or transcriptional units.
One such 1 S region is downstream of the devS gene, which has been shown not to affect the growth or development of M. xanthus. The first plasmid, pKOS35-154, is linearized with DraI and electroporated into M. xanthus. This plasmid contains two regions of the dev locus flanking two fragments of the epothilone gene cluster. Inserted in between the epo gene regions are the kanamycin resistance marker and the galK gene. Kanamycin resistance arises in colonies if the DNA recombines into the dev region by a double recombination using the dev sequence as regions of homology. This strain, K35-159, contains small regions of the epothilone gene cluster that will allow for recombination of pKOS35-79.85.
Because the resistance markers on pKOS35-79.85 are the same as that for K35-159, a tetracycline transposon was transposed into the cosmid, and cosmids that contain the transposon inserted into the kanamycin marker were selected. This cosmid, pKOS90-23, was electroporated into K35-159, and oxytetracycline resistant colonies were selected to create strain K35-174. To remove the unwanted regions from the cosmid and leave only the epothilone genes, cells were plated on CYE plates containing 1 %
galactose. The presence of the galK gene makes the cells sensitive to 1 % galactose.
Galactose resistant colonies of K35-174 represent cells that have lost the galK marker by recombination or by a mutation in the galK gene. If the recombination event occurs, then the galactose resistant strain is sensitive to kanamycin and oxytetracycline. Strains sensitive to both antibiotics are verified by Southern blot analysis. The correct strain is identified and designated K35-WO 00/31247 _ 95 - PCT/US99/2743$
175 and contains the epothilone gene cluster from module 7 through two open reading frames past the epoL gene.
To introduce modules 1 through module 7, the above process is repeated once more. The plasmid pKOS90-22 is linearized with DraI and electroporated into K35-175 to create K35-180. This strain is electroporated with the tetracycline resistant version of pKOS35-70.1A2, pKOS90-38, and colonies resistant to oxytetracycline are selected. This creates strain K35-185. Recombinants that now have the whole epothilone gene cluster are selected by resistance to 1% galactose. This results in strain K35-188. This strain contains all the epothilone genes as well as all potential promoters. This strain is fermented and tested for the production of epothilones A and B.
To clone the whole gene cluster as one fragment, a bacterial artificial chromosome (BAC) library is constructed. First, SMP44 cells are embedded in agarose and lysed according to the BIO-RAD genomic DNA plug kit. DNA plugs are partially digested with restriction enzyme, such as Sau3AI or HindIII, and electrophoresed on a FIGE
or CHEF
gel. DNA fragments are isolated by electroeluting the DNA from the agarose or using gelase to degrade the agarose. The method of choice to isolate the fragments is electroelution, as described in Strong et al., 1997, Nucleic Acids Res. 19:
3959-3961, incorporated herein by reference. The DNA is ligated into the BAC (pBeIoBACII) cleaved with the appropriate enzyme. A map of pBeIoBACII is shown below.

WO 00/31247 _ 96 _ PGT/US99/27438 ri~ -~ SPs Ssl 1 Notl $ H Notl I
I
psN Ipxp lac Z
~ao R1 par C
pnr B r CtJI
s"~ ~' pBeIoBAC 11 7.4 kb xho I
par A
on S
Xba t r Eco RV
Eco FtV
The DNA is electroporated into DH10B cells by the method of Sheng et al., 1995, Nucleic Acids Res. 23: I990-1996, incorporated herein by reference, to create an S. cellulosum genomic library. Colonies are screened using a probe from the NRPS region of the epothilone cluster. Positive clones are picked and DNA is isolated for restriction analysis to confirm the presence of the complete gene cluster. This positive clone is designated pKOS35-178.
To create a strain that can be used to introduce pKOS35-178, a plasmid, pKOS35-164, is constructed that contains regions of homology that are upstream and downstream of the epothilone gene cluster flanked by the dev locus and containing the kanamycin resistance galK cassette, analogous to plasmids pKOS90-22 and pKOS35-154. This plasmid is linearized with DraI and electroporated into M. xanthus, in accordance with the method of Kafeshi et al., 1995, Mol. Microbiol. 15: 483-494, to create K35-183. The plasmid pKOS35-178 can be introduced into K35-183 by electroporation or by transduction with bacteriophage P1 and chloramphenicol resistant colonies are selected.
Alternatively, a version of pKOS35-178 that contains the origin of conjugative transfer from pRP4 can be constructed for transfer of DNA from E coli to K35-183. This plasmid WO 00/31247 _ 97 - PCT/US99/27438 is made by first constructing a transposon containing the oriT region from RP4 and the tetracycline resistance maker from pACYC 184 and then transposing the transposon in vitro or in vivo onto pKOS35-178. This plasmid is transformed into S17-l and conjugated into M. xanthus. This strain, K35-190, is grown in the presence of 1%
galactose to select for the second recombination event. This strain contains all the epothilone genes as well as all potential promoters. This strain will be fermented and tested for the production of epothilones A and B.
Besides integrating pKOS35-178 into the dev locus, it can also be integrated into a phage attachment site using integration functions from myxophages Mx8 or Mx9.
A
transposon is constructed that contains the integration genes and att site from either Mx8 or Mx9 along with the tetracycline gene from pACYC 184. Alternative versions of this transposon may have only the attachment site. In this version, the integration genes are then supplied in trans by coelectroporation of a plasmid containing the integrase gene or having the integrase protein expressed in the electroporated strain from any constitutive promoter, such as the mgl promoter (see Magrini et al., Jul. 1999, J. Bact.
181(13): 4062-4070, incorporated herein by reference). Once the transposon is constructed, it is transposed onto pKOS35-178 to create pKOS35-191. This plasmid is introduced into Myxococcus xanthus as described above. This strain contains all the epothilone genes as well as all potential promoters. This strain is fermented and tested for the production of epothilones A and B.
Once the epothilone genes have been established in a strain of Myxococcus xanthus, manipulation of any part of the gene cluster, such as changing promoters or swapping modules, can be performed using the kanamycin resistance and galK
cassette.
Cultures of Myxococcus xanthus containing the epo genes are grown in a number of media and examined for production of epothilones. If the levels of production of epothilones (in particular B or D) are too low to permit large scale fermentation, the M. xanthus-producing clones are subjected to media development and strain improvement, as described below for enhancing production in Streptomyces.
Example 4 Construction of a Streptomyces Expression Vector The present invention provides recombinant expression vectors for the heterologous expression of modular polyketide synthase genes in Streptomyces hosts.

WO 00/31247 _ 9g _ PCT/US99/Z7438 These vectors include expression vectors that employ the actI promoter that is regulated by the gene actII ORF4 to allow regulated expression at high levels when growing cells enter stationary phase. Among the vectors available are plasmids pRMI and pRMS, and derivatives thereof such as pCK7, which are stable, low copy plasmids that carry the S marker for thiostrepton resistance in actinomycetes. Such plasmids can accommodate large inserts of cloned DNA and have been used for the expression of the DEBS
PKS in S. coelicolor and S. lividans, the picromycin PKS genes in S. lividans, and the oleandomycin PKS genes in S. lividans. See U.S. Patent No. 5,712,146. Those of skill in the art recognize that S. lividans does not make the tRNA that recognizes the TTA codon for leucine until late-stage growth and that if production of a protein is desired earlier, then appropriate codon modifications can be made.
Pactl fd teem eryA!
actl!

on SCP2"
pCKT
on ColEl eryAtl bia eryAlll Plasmid pCK7 Another vector is a derivative of plasmid pSET152 and comprises the actII ORF4-PactI expression system but carnes the selectable marker for apramycin resistance. These vectors contain the attP site and integrase gene of the actinophage phiC31 and do not replicate autonomously in Streptomyces hosts but integrate by site specific recombination into the chromosome at the attachment site for phiC31 after introduction into the cell.
Derivatives of pCK7 and pSET152 have been used together for the heterologous production of a polyketide, with different PKS genes expressed from each plasmid. See U.S. patent application Serial No. 60/129,731, filed 16 Apr. 1999, incorporated herein by reference.

WO 00/31247 _ 99 _ PCT/US99/27438 Pacl Bglll cr term ORF
nt ~C31 eryAlll ttP
pKOS010-153 apramycir~
pUC18 on Plasmid pKOS010-153, a pSET152 Derivative The need to develop expression vectors for the epothilone PKS that function in Streptomyces is significant. The epothilone compounds are currently produced in the slow growing, genetically intractable host Sorangium cellulosum or are made synthetically. The streptomycetes, bacteria that produce more than 70% of all known antibiotics and important complex polyketides, are excellent hosts for production of epothilones and epothilone derivatives. S. lividans and S. coelicolor have been developed for the expression of heterologous PKS systems. These organisms can stably maintain cloned heterologous PKS genes, express them at high levels under controlled conditions, and modify the corresponding PKS proteins (e.g. phosphopantetheinylation) so that they are capable of production of the polyketide they encode. Furthermore, these hosts contain the necessary pathways to produce the substrates required for polyketide synthesis, e.g.
malonyl CoA and methylmalonyl CoA. A wide variety of cloning and expression vectors are available for these hosts, as are methods for the introduction and stable maintenance of large segments of foreign DNA. Relative to the slow growing Sorangium host, S
lividans and S. coelicolor grow well on a number of media and have been adapted for high level production of polyketides in fermentors. A number of approaches are available for yield improvements, including rational approaches to increase expression rates, increase precursor supply, etc. Empirical methods to increase the titers of the polyketides, long since proven effective for numerous other polyketides produced in streptomycetes, can also be employed for the epothilone and epothilone derivative producing host cells of the invention.

To produce epothilones by fermentation in a heterologous Streptomyces host, the epothilone PKS (including the NRPS module) genes are cloned in two segments in derivatives of pCK7 (loading domain through module 6) and pKOS010-153 (modules through 9). The two plasmids are introduced into S lividans employing selection for thiostrepton and apramycin resistance. In this arrangement, the pCK7 derivative replicates autonomously whereas the pKOS010-153 derivative is integrated in the chromosome. In both vectors, expression of the epothilone genes is from the actI promoter resident within the plasmid.
To facilitate the cloning, the two epothilone PKS encoding segments (one for the loading domain through module six and one for modules seven through nine) were cloned as translational fusions with the N-terminal segment of the KS domain of module 5 of the ery PKS. High level expression has been demonstrated from this promoter employing KSS
as the first translated sequence, see Jacobsen et al., 1998, Biochemistry 37:
4928-4934, incorporated herein by reference. A convenient BsaBI site is contained within the DNA
segment encoding the amino acid sequence EPIAV that is highly conserved in many KS
domains including the KS-encoding regions of epoA and of module 7 in epoE.
The expression vector for the loading domain and modules one through six of the epothilone PKS was designated pKOS039-124, and the expression vector for modules seven through nine was designated pKOS039-126. Those of skill in the art will recognize that other vectors and vector components can be used to make equivalent vectors. Because preferred expression vectors of the invention, described below and derived from pKOS039-124 and pKOS039-126, have been deposited under the terms of the Budapest Treaty, only a summary of the construction of plasmids pKOS039-124 and pKOS039-is provided below.
The eryKSS linker coding sequences were cloned as an ~0.4 kb PacI-BgIII
restriction fragment from plasmid pKOSlO-153 into pKOS039-98 to construct plasmid pKOS039-117. The coding sequences for the eryKSS linker were linked to those for the epothilone loading domain by inserting the ~8.7 kb EcoRI-XbaI restriction fragment from cosmid pKOS35-70.1A2 into EcoRI-XbaI digested plasmid pLItmus28. The ~3.4 kb of BsaBI-NotI and ~3.7 kb NotI-HindIII restriction fragments from the resulting plasmid were inserted into BsaBI-HindIII digested plasmid pKOS039-117 to construct plasmid pKOS039-120. The ~7 kb PacI-XbaI restriction fragment of plasmid pKOS039-120 was inserted into plasmid pKA018' to construct plasmid pKOS039-123. The final pKOS039-124 expression vector was constructed by ligating the ~34 kb XbaI-AvrII
restriction fragment of cosmid pKOS35-70.1A2 with the 21.1 kb AvrII-XbaI restriction fragment of pKOS039-123.
The plasmid pKOS039-126 expression vector was constructed as follows. First the S coding sequences for module 7 were linked from cosmids pKOS35-70.4 and pKOS35-79.85 by cloning the ~6.9 kb BgIII-NotI restriction fragment of pKOS35-70.4 and the ~5.9 kb NotI-HindIII restriction fragment of pKOS35-79.85 into BgIII-HindIII
digested plasmid pLitmus28 to construct plasmid pKOS039-119. The ~12 kb NdeI-NheI
restriction fragment of cosmid pKOS35-79.85 was cloned into NdeI-XbaI digested plasmid pKOS039-119 to construct plasmid pKOS039-122.
To fuse the eryKSS linker coding sequences with the coding sequences for module 7, the ~1 kb BsaBI-BgIII restriction fragment derived from cosmid pKOS35-70.4 was cloned into BsaBI-BcII digested plasmid pKOS039-117 to construct plasmid pKOS039-121. The 21.5 kb AvrII restriction fragment from plasmid pKOS039-122 was cloned into AvrII-XbaI digested plasmid pKOS039-121 to construct plasmid pKOS039-125. The 21.8 kb PacI-EcoRI restriction fragment of plasmid pKOS039-125 was ligated with the ~9 kb PacI-EcoRI restriction fragment of plasmid pKOS039-44 to construct pKOS039-126.
Plasmids pKOS039-124 and pKOS126 were introduced into S. lividans K4-114 sequentially employing selection for the corresponding drug resistance marker.
Because plasmid pKOS039-126 does not replicate autonomously in streptomycetes, the selection is for cells in which the plasmid has integrated in the chromosome by site-specific recombination at the attB site of phiC31. Because the plasmid stably integrates, continued selection for apramycin resistance is not required. Selection can be maintained if desired.
The presence of thiostrepton in the medium is maintained to ensure continued selection for plasmid pKOS039-124. Plasmids pKOS039-124 and pKOS039-126 were transformed into Streptomyces lividans K4-114, and transformants containing the plasmids were cultured and tested for production of epothilones. Initial tests did not indicate the presence of an epothilone.
To improve production of epothilones from these vectors, the eryKSS linker sequences were replaced by epothilone PKS gene coding sequences, and the vectors were introduced into Streptomyces coelicolor CH999. To amplify by PCR coding sequences from the epoA gene coding sequence, two oligonucleotides primers were used:

N39-73, 5'-GCTTAATTAAGGAGGACACATATGCCCGTCGTGGCGGATCGTCC-3'; and N39-74, 5'-GCGGATCCTCGAATCACCGCCAATATC-3'.
The template DNA was derived from cosmid pKOS35-70.8A3. The ~0.8 kb PCR
product was digested with restriction enzymes PacI and BamHI and then ligated with the ~2.4 kb BamHI-NotI and the ~6.4 kb PacI-NotI restriction fragments of plasmid pKOS039-120 to construct plasmid pKOS039-136. To make the expression vector for the epoA, epoB, epoC, and epoD genes, the ~5 kb PacI-AvrII restriction fragment of plasmid pKOS039-136 was ligated with the ~50 kb PacI-AvrII restriction fragment of plasmid pKOS039-124 to construct the expression plasmid pKOS039-1248. Plasmid pKOS039-1248 has been deposited with the ATCC under the terms of the Budapest Treaty and is available under accession number To amplify by PCR sequences from the epoE gene coding sequence, two oligonucleotide primers were used:
N39-67A, 5'-GCTTAATTAAGGAGGACACATATGACCGACCGAGAAGGCCAGCTC-CTGGA-3', and IS N39-68, 5'-GGACCTAGGCGGGATGCCGGCGTCT-3'.
The template DNA was derived from cosmid pKOS35-70.1A2. The ~0.4 kb amplification product was digested with restriction enzymes PacI and AvrII and ligated with either the 29.5 kb PacI-AvrII restriction fragment of plasmid pKOS039-126 or the 23.8 kb PacI-AvrII restriction fragment of plasmid pKOS039-125 to construct plasmid pKOS039-1268 or plasmid pKOS039-1258, respectively. Plasmid pKOS039-1268 was deposited with the ATCC under the terms of the Budapest Treaty and is available under accession number The plasmid pair pKOS039-1248 and pKOS039-1268 (as well as the plasmid pair pKOS039-124 and pKOS039-126) contain the full complement of epoA, epoB, epoC, epoD, epoE, epoF, epoK, and epoL genes. The latter two genes are present on plasmid pKOS039-1268 (as well as plasmid pKOS039-126); however, to ensure that these genes were expressed at high levels, another expression vector of the invention, plasmid pKOS039-141 (Figure 8), was constructed in which the epoK and epoL genes were placed under the control of the ermE* promoter.
The epoK gene sequences were amplified by PCR using the oligonucleotide primers:
N39-69, 5'-AGGCATGCATATGACCCAGGAGCAAGCGAATCAGAGTG-3'; and N39-70, 5'-CCAAGCTTTATCCAGCTTTGGAGGGCTTCAAG-3'.

The epoL gene sequences were amplified by PCR using the oligonucleotide primers:
N39-71A, 5'-GTAAGCTTAGGAGGACACATATGATGCAACTCGCGCGCGGGTG-3'; and N39-72, 5'-GCCTGCAGGCTCAGGCTTGCGCAGAGCGT-3'.
S The template DNA for the amplifications was derived from cosmid pKOS35-79.85. The PCR products were subcloned into PCR-script for sequence analysis.
Then, the epoK and epoL genes were isolated from the clones as NdeI-HindIII and HindIII-EcoRI
restriction fragments, respectively, and ligated with the ~6 kb NdeI-EcoRI
restriction fragment of plasmid pKOS039-134B, which contains the ermE* promoter, to construct plasmid pKOS039-140. The ~2.4 kb NheI-PstI restriction fragment of plasmid pKOS039-140 was cloned into XbaI-PstI digested plasmid pSAM-Hyg, a plasmid pSAM2 derivative containing a hygromycin resistance conferring gene, to construct plasmid pKOS039-141.
Another variant of plasmid pKOS039-1268 was constructed to provide the epoE
and epoF genes on an expression vector without the epoK and epoL genes. This plasmid, pKOS045-12 (Figure 9), was constructed as follows. Plasmid pXH106 (described in J.
Bact., 1991, 173: 5573-5577, incorporated herein by reference) was digested with restriction enzymes StuI and BamHI, and the ~2.8 kb restriction fragment containing the xylE and hygromycin resistance conferring genes was isolated and cloned into EcoRV-BgIII digested plasmid pLitmus28. The ~2.8 kb NcoI-AvrII restriction fragment of the resulting plasmid was ligated to the ~18 kb PacI-BspHI restriction fragment ofplasmid pKOS039-1258 and the ~9 kb SpeI-PacI restriction fragment of plasmid pKOS039-42 to construct plasmid pKOS045-12.
To construct an expression vector that comprised only the epoL gene, plasmid pKOS039-141 was partially digested with restriction enzyme NdeI, the ~9 kb NdeI
restriction fragment was isolated, and the fragment then circularized by ligation to yield plasmid pKOS039-150.
The various expression vectors described above were then transformed into Streptomyces coelicolor CH999 and S. lividans K4-114 in a variety of combinations, the transformed host cells fermented on plates and in liquid culture (RS medium, which is identical to R2YE medium without agar). Typical fermentation conditions follow. First, a seed culture of about 5 mL containing SO ug/L thiostrepton was inoculated and grown at 30°C for two days. Then, about 1 to 2 mL of the seed culture was used to inoculate a production culture of about 50 mL containing 50 ~g/L thiostrepton and 1 mM
cysteine, WO00/31247 _ 1~4 _ PCT/US99/27438 and the production culture was grown at 30°C for 5 days. Also, the seed culture was used to prepare plates of cells {the plates contained the same media as the production culture with 10 mM propionate), which were grown at 30°C for nine days.
Certain of the Streptomyces coedicolor cultures and culture broths were analyzed for production of epothilones. The liquid cultures were extracted with three times with equal volumes of ethyl acetate, the organic extracts combined and evaporated, and the residue dissolved in acetonitrile for LC/MS analysis. The agar plate media was chopped and extracted twice with equal volumes of acetone, and the acetone extracts were combined and evaporated to an aqueous slurry, which was extracted three times with equal volumes of ethyl acetate. The organic extracts were combined and evaporated, and the residue dissolved in acetonitrile for LC/MS analysis.
Production of epothilones was assessed using LC-mass spectrometry. The output flow from the UV detector of an analytical HPLC was split equally between a Perkin-Elmer/Sciex API 100LC mass spectrometer and an Alltech S00 evaporative light scattering 1 S detector. Samples were injected onto a 4.6 x 150 mm reversed phase HPLC
column (MetaChem 5 m ODS-3 Inertsil) equilibrated in water with a flow rate of 1.0 mL/min. UV
detection was set at 250 nm. Sample components were separated using H20 for 1 minute, then a linear gradient from 0 to 100% acetonitrile over 10 minutes. Under these conditions, epothilone A elutes at 10.2 minutes and epothilone B elutes at 10.5 minutes.
The identity of these compounds was confirmed by the mass spectra obtained using an atmospheric chemical ionization source with orifice and ring voltages set at 75 V and 300 V, respectively, and a mass resolution of 0.1 amu. Under these conditions, epothilone A
shows [M+HJ at 494.4 amu, with observed fragments at 476.4, 318.3, and 306.4 amu.
Epothilone B shows [M+HJ at 508.4 amu, with observed fragments at 490.4, 320.3, and 302.4 amu.
Transformants containing the vector pairs pKOS039-I24R and pKOS039-1268 or pKOS039-124 and pKOS039-1268 produced detectable amounts of epothilones A and B.
Transformants containing these plasmid pairs and the additional plasmid pKOS039-141 produced similar amounts of epothilones A and B, indicating that the additional copies of the epoK and epoL genes were not required for production under the test conditions employed. Thus, these transformants produced epothilones A and B when recombinant epoA, epoB, epoC, epoD, epoE, epoF, epoK, and epoL genes were present. In some WO 00/31247 - 105 - PCT/US99/2743$
cultures, it was observed that the absence of propionate increased the proportion of epothilone B to epothilone A.
Transformants containing the plasmid pair pKOS039-1248 and pKOS045-12 produced epothilones C and D, as did transformants containing this plasmid pair and the additional plasmid pKOS039-150. These results showed that the epoL gene was not required under the test conditions employed to form the C-12-C-13 double bond.
These results indicate that either the epothilone PKS gene alone is able to form the double bond or that Streptomyces coelicolor expresses a gene product able to convert epothilones G and H to epothilones C and D. Thus, these transformants produced epothilones C and D when I 0 recombinant epoA, epoB, epoC, epoD, epoE, and epoF genes were present.
The heterologous expression of the epothilone PKS described herein is believed to represent the recombinant expression of the largest proteins and active enzyme complex that have ever been expressed in a recombinant host cell. The epothilone producing Streptomyces coelicolor transformants exhibited growth characteristics indicating that either the epothilone PKS genes, or their products, or the epothilones inhibited cell growth or were somewhat toxic to the cells. Any such inhibition or toxicity could be due to accumulation of the epothilones in the cell, and it is believed that the native Sorangium producer cells may contain transporter proteins that in effect pump epothilones out of the cell. Such transporter genes are believed to be included among the ORFs located downstream of the epoK gene and described above. Thus, the present invention provides Streptomyces and other host cells that include recombinant genes that encode the products of one or more, including all, of the ORFs in this region.
For example, each ORF can be cloned behind the ermE* promoter, see Stassi et al., 1998, Appl. Microbiol. Biotechnol. 49: 725-731, incorporated herein by reference, in a pSAM2-based plasmid that can integrate into the chromosome of Streptomyces coelicolor and S. lividans at a site distinct from attB of phage phiC3l, see Smokvina et al., 1990, Gene 94: 53-59, incorporated herein by reference. A pSAM2-based vector carrying the gene for hygromycin resistance is modified to carry the ermE* promoter along with additional cloning sites. Each ORF downstream is PCR cloned into the vector which is then introduced into the host cell (also containing pKOS039-1248 and pKOS039-1268 or other expression vectors of the invention) employing hygromycin selection.
Clones carrying each individual gene downstream from epoK are analyzed for increased production of epothilones.

Additional fermentation and strain improvement efforts can be conducted as illustrated by the following. The levels of expression of the PKS genes in the various constructs can be measured by assaying the levels of the corresponding mRNAs (by quantitative RT PCR) relative to the levels of another heterologous PKS mRNA
(e.g.
picromycin) produced from genes cloned in similar expression vectors in the same host. If one of the epothilone transcripts is underproduced, experiments to enhance its production by cloning the corresponding DNA segment in a different expression vector are conducted. for example, multiple copies of any one or more of the epothilone PKS genes can be introduced into a cell if one or more gene products are rate limiting for biosynthesis. If the basis for low level production is not related to low level PKS gene expression (at the RNA level), an empirical mutagenesis and screening approach that is the backbone of yield improvement of every commercially important fermentation product is undertaken. Spores are subjected to UV, X-ray or chemical mutagens, and individual survivors are plated and picked and tested for the level of compound produced in small scale fermentations. Although this process can be automated, one can examine several thousand isolates for quantifiable epothilone production using the susceptible fungus Mucor hiemalis as a test organism.
Another method to increase the yield of epothilones produced is to change the KSY
domain of the loading domain of the epothilone PKS to a KSQ domain. Such altered loading domains can be constructed in any of a variety of ways, but one illustrative method follows. Plasmid pKOS39-1248 of the invention can be conveniently used as a starting material. To amplify DNA fragments useful in the construction, four oligonucleotide primers are employed:
N39-83: 5'-CCGGTATCCACCGCGACACACGGC-3', N39-84: 5'-GCCAGTCGTCCTCGCTCGTGGCCGTTC-3', and N39-73 and N39-74, which have been described above. The PCR fragment generated with N37-73 and N39-83 and the PCR fragment generated with N39-74 and N39-84 are treated with restriction enzymes PacI and BamHI, respectively, and ligated with the ~3.1 kb PacI-BamHI fragment of plasmid pKOS39-120 to construct plasmid pKOS039-148.
The -r0.8 kb PacI-BamHI restriction fragment of plasmid pKOS039-148 (comprising the two PCR amplification products) is ligated with the ~2.4 kb BamHI-NotI
restriction fragment and the ~6.4 kb PacI-NotI restriction fragment of plasmid pKOS39-120 to construct pKOS39-136Q. The ~S kb PacI-AvrII restriction fragment of plasmid pKOS039-WO 00/31247 _ 10~ - PCT/US99/27438 136Q is ligated to the ~50 kb PacI-AvrII restriction fragment of plasmid pKOS039-124 to construct plasmid pKOS39-124Q. Plasmids pKOS039-124Q and pKOS039-1268 are then transformed into Streptomyces coelicolor CH999 for epothilone production.
The epoA through epoF, optionally with epoK or with epoK plus epoL, genes cloned and expressed are su~cient for the synthesis of epothilone compounds, and the distribution of the C-12 H to C-12 methyl congeners appears to be similar to that seen in the natural host (A:B::2:1 ). This ratio reflects that the AT domain of module 4 more closely resembles that of the malonyl rather than methylmalonyl specifying AT
consensus domains. Thus, epothilones D and B are produced at lower quantities than their unmethylated counterparts C and A. The invention provides PKS genes that produce epothilone D and/or B exclusively. Specifically, methylmalonyl CoA specifying AT
domains from a number of sources (e.g. the narbonolide PKS, the rapamycin PKS, and others listed above) can be used to replace the naturally occurring at domain in module 4.
The exchange is performed by direct cloning of the incoming DNA into the appropriate 1 S site in the epothilone PKS encoding DNA segment or by gene replacement through homologous recombination.
For gene replacement through homologous recombination, the donor sequence to be exchanged is placed in a delivery vector between segments of at least 1 kb in length that flank the AT domain of epo module 4 encoding DNA. Crossovers in the homologous regions result in the exchange of the epo AT4 domain with that on the delivery vector.
Because pKOS039-124 and pKOS039-1248 contain AT4 coding sequences, they can be used as the host DNA for replacement. The adjacent DNA segments are cloned in one of a number of E. coli plasmids that are temperature sensitive for replication. The heterologous AT domains can be cloned in these plasmids in the correct orientation between the homologous regions as cassettes enabling the ability to perform several AT
exchanges simultaneously. The reconstructed plasmid (pKOS039-124* or pKOS039-1248*) is tested for ability to direct the synthesis of epothilone B and/or by introducing it along with pKOS039-126 or pKOS039-1268 in Streptomyces coelicolor and/or S. lividans.
Because the titers of the polyketide can vary from strain to strain carrying the different gene replacements, the invention provides a number of heterologous methylmalonyl CoA specifying AT domains to ensure that production of epothilone D at titers equivalent to that of the C and D mixture produced in the Streptomyces coelicolor host described above. In addition, larger segments of the donor genes can be used for the WO00/31247 _ 1~g- PCT/US99/27438 replacements, including, in addition to the AT domain, adjacent upstream and downstream sequences that correspond to an entire module. If an entire module is used for the replacement, the KS, methylmalonyl AT, DH, KR, ACP - encoding DNA segment can be obtained from for example and without limitation the DNA encoding the tenth module of the rapamycin PKS, or the first or fifth modules of the FK-520 PKS.
Example 5 Heterologous Expression of EpoK and Conversion of Epothilone D to Epothilone_B
This Example describes the construction of E. coli expression vectors for epoK.
The epoK gene product was expressed in E. coli as a fusion protein with a polyhistidine tag (his tag). The fusion protein was purified and used to convert epothilone D to epothilone B.
Plasmids were constructed to encode fusion proteins composed of six histidine residues fused to either the amino or carboxy terminus of EpoK. The following oligos were used to construct the plasmids:
55-lOl.a-I:
5'-AAAAACATATGCACCACCACCACCACCACATGACACAGGAGCAAGCGAAT-CAGAGTGAG-3' 55-lOl.b:
5'-AAAAAGGATCCTTAATCCAGCTTTGGAGGGCTT-3', ss-lol.c:
5'-AAAAACATATGACACAGGAGCAAGCGAAT-3', and 55-lOl.d:
5'-AAAAAGGATCCTTAGTGGTGGTGGTGGTGGTGTCCAGCTTTGGAGGGCTTC-AAGATGAC-3'.
The plasmid encoding the amino terminal his tag fusion protein, pKOS55-121, was constructed using primers 55-lOl.a-land 55-lOl.b, and the one encoding the carboxy terminal his tag, pKOS55-129, was constructed using primers 55-lOl.c and 55-lOl.d in PCR reactions containing pKOS35-83.5 as the template DNA. Plasmid pKOS35-83.5 contains the ~S kb NotI fragment comprising the epoK gene ligated into pBluescriptSKII+
(Stratagene). The PCR products were cleaved with restriction enzymes BamHI and NdeI
and ligated into the BamHI and NdeI sites of pET22b (Invitrogen). Both plasmids were sequenced to verify that no mutations were introduced during the PCR
amplification.
Protein gels were run as known in the art.
Purification of EpoK was performed as follows. Plasmids pKOS55-121 and pKOS55-129 were transformed into BL21(DE3) containing the groELS expressing plasmid pREP4-groELS (Caspers et al., 1994, Cellular and Molecular Biology 40(5):
635-644). The strains were inoculated into 250 mL of M9 medium supplemented with 2 mM MgS04, 1% glucose, 20 mg thiamin, 5 mg FeCl2, 4 mg CaCl2 and 50 mg levulinic acid. The cultures were grown to an OD6~ between 0.4 and 0.6, at which point IPTG was added to 1 mM, and the cultures were allowed to grow for an additional two hours. The cells were harvested and frozen at -80°C. The frozen cells were resuspended in 10 ml of buffer 1 (5 mM imidazole, 500 mM NaCI, and 45 mM Tris pH 7.6) and were lysed by sonicating three times for 15 seconds each on setting 8. The cellular debris was pelleted by spinning in an SS-34 rotor at 16,000 rpm for 30 minutes. The supernatant was removed and spun again at 16,000 rpm for 30 minutes. The supernatant was loaded onto a 5 mL
nickel column (Novagen), after which the column was washed with 50 mL of buffer 1 (Novagen). EpoK was eluted with a gradient from 5 mM to 1 M imidazole.
Fractions containing EpoK were pooled and dialyzed twice against 1 L of dialysis buffer ( 45 mM
Tris pH7.6, 0.2 mM DTT, 0.1 mM EDTA, and 20% glycerol). Aliquots were frozen in liquid nitrogen and stored at -80°C. The protein preparations were greater than 90% pure.
The EpoK assay was performed as follows (See Betlach et al., Biochem (1998) 37:14937, incorporated herein by reference). Briefly, reactions consisted of 50 mM Tris (pH7.5), 21 pM spinach ferredoxin, 0.132 units of spinach ferredoxin: NADP+
oxidoreductase, 0.8 units of glucose-6-phosphate dehydrogenase, 1.4 mM NADP, and 7.1 mM glucose-6-phosphate, 100 ~M or 200 ~M epothilone D (a generous gift of S.
Danishefsky), and 1.7 pM amino terminal his tagged EpoK or 1.6 pM carboxy terminal his tagged EpoK in a 100 ~.L volume. The reactions were incubated at 30°C for 67 minutes and stopped by heating at 90°C for 2 minutes. The insoluble material was removed by centrifugation, and 50 ~L of the supernatant were analyzed by LC/MS. HPLC
conditions: Metachem 5 ~, ODS-3 Inertsil (4.6 X 150 mm); 80% H20 for 1 min, then to 100% MeCN over 10 min at 1 mL/min, with UV (~,~"~ 250 nm), ELSD, and MS
detection. Under these conditions, epothilone D eluted at 11.6 min and epothilone B at 9.3 min. the LC/MS spectra were obtained using an atmosphere pressure chemical ionization source with orifice and ring voltages set at 20 V and 250 V, respectively, at a mass resolution of 1 amu. Under these conditions, epothilone E shows an [M+H~ at m1z 493, with observed fragments at 405 and 304. Epothilone B shows an [M+H~ at mlz 509, with observed fragments at 491 and 320.

WO 00/31247 PCT/t1S99/27438 The reactions containing EpoK and epothilone D contained a compound absent in the control that displayed the same retention time, molecular weight, and mass fragmentation pattern as pure epothilane B. With an epothilone D concentration of 100 p.M, the amino and the carboxy terminal his tagged EpoK was able to convert 82% and S 58% to epothilone B, respectively. In the presence of 200 pM, conversion was 44% and 21 %, respectively. These results demonstrate that EpoK can convert epothilone D to epothilone B.
Example 6 Modified Epothilones from ChemobincvnrhP~;~

This Example describes a series of thioesters provided by the invention for production of epothilone derivatives via chemobiosynthesis. The DNA sequence of the biosynthetic gene cluster for epothilone from Sorangium cellulosum indicates that priming of the PKS involves a mixture of polyketide and amino acid components. Priming involves loading of the PKS-like portion of the loading domain with malonyl CoA
followed by decarboxylation and loading of the module one NRPS with cysteine, then condensation to form enzyme-bound N-acetylcysteine. Cyclization to form a thiazoline is followed by oxidation to form enzyme bound 2-methylthiazole-4-carboxylate, the product of the loading domain and NRPS. Subsequent condensation with methylmalonyl CoA
by the ketosynthase of module 2 provides the substrate for module, as shown in the following diagram.

~S-CoA
ATP S H
O ATP S
SH ~ ~N S-Enz ~ ~~~S-Enz H3+N O_ /~H
oxidation S
~~S-E nz modu/1e~2~''s' ubstrate MM-CoA
S
epothilone ~- ~--~S-E~
module 3 substrate The present invention provides methods and reagents for chemobiosynthesis to produce epothilone derivatives in a manner similar to that described to make 6-dEB and erythromycin analogs in PCT Pat. Pub. Nos. 99/03986 and 97/02358. Two types of feeding substrates are provided: analogs of the NRPS product, and analogs of the module 3 substrate. The module 2 substrates are used with PKS enzymes with a mutated NRPS-like domain, and the module 3 substrates are used with PKS enzymes with a mutated KS
domain in module 2.

The following illustrate module 2 substrates (as N-acetyl cysteamine thioesters) for use as substrates for epothilone PKS with modified inactivated NRPS:

N ~a N

Y V a ~N
O O
O
9~
The module 2 substrates are prepared by activation of the corresponding carboxylic acid and treatment with N-acetylcysteamine. Activation methods include formation of the acid chloride, formation of a mixed anhydride, or reaction with a condensing reagent such as a carbodiimide.

Exemplary module 3 substrates, also as NAc thioesters for use as substrates for epothilone PKS with KS2 knockout are:
O O
R / \ s~\/ s~\/

~H

O
H H
H
O
O
N
H
O
H
\ ' These compounds are prepared in a three-step process. First, the appropriate aldehyde is treated with a Wittig reagent or equivalent to form the substituted acrylic ester.
The ester is saponified to the acid, which is then activated and treated with N-acetylcysteamine.
Tllustrative reaction schemes for making module 2 and module 3 substrates follow.
Additional compounds suitable for making starting materials for polyketide synthesis by the epothilone PKS are shown in Figure 2 as carboxylic acids (or aldehydes that can be converted to carboxylic acids) that are converted to the N-acylcysteamides for supplying to the host cells of the invention.
A. Thiophene-3-carboxylate N-acetylcysteamine thioester A solution of thiophene-3-carboxylic acid (128 mg) in 2 mL of dry tetrahydrofuran under inert atmosphere was treated with triethylamine (0.25 mL) and diphenylphosphoryl azide (0.50 mL). After 1 hour, N-acetylcysteamine (0.25 mL) was added, and the reaction was allowed to proceed for 12 hours. The mixture was poured into water and extracted three times with equal volumes of ethyl acetate. The organic extracts were combined, washed sequentially with water, 1 N HCI, sat. CuS04, and brine, then dried over MgS04, filtered, and concentrated under vacuum. Chromatography on Si02 using ether followed by ethyl acetate provided pure product, which crystallized upon standing.
B. Furan-3-carboxylate N-acetylcysteamine thioester A solution of furan-3-carboxylic acid (112 mg) in 2 mL of dry tetrahydrofuran under inert atmosphere was treated with triethylamine (0.25 mL) and diphenylphosphoryl azide (0.50 mL). After 1 hour, N-acetylcysteamine (0.25 mL) was added and the reaction was allowed to proceed for 12 hours. The mixture was poured into water and extracted three times with equal volumes of ethyl acetate. The organic extracts were combined, washed sequentially with water, 1 N HCI, sat. CuS04, and brine, then dried over MgS04, filtered, and concentrated under vacuum. Chromatography on Si02 using ether followed by ethyl acetate provided pure product, which crystallized upon standing.
C. Pyrrole-2-carboxylate N-acetylcysteamine thioester A solution of pyrrole-2-carboxylic acid ( 112 mg) in 2 mL of dry tetrahydrofuran under inert atmosphere was treated with triethylamine {0.25 mL) and diphenylphosphoryl azide (0.50 mL). After 1 hour, N-acetylcysteamine (0.25 mL) was added and the reaction was allowed to proceed for 12 hours. The mixture was poured into water and extracted three times with equal volumes of ethyl acetate. The organic extracts were combined, washed sequentially with water, 1 N HCI, sat. CuS04, and brine, then dried over MgS04,filtered, and concentrated under vacuum. Chromatography on Si02 using ether followed by ethyl acetate provided pure product, which crystallized upon standing.
D. 2-Methyl-3-(3-thienyl)acrylate N-acetylcysteamine thioester (1) Ethyl 2-methyl-3-(3-thienyl)acrylate: A mixture of thiophene-3-carboxaldehyde (1.12 g) and (carbethoxyethylidene)triphenylphosphorane (4.3 g) in dry tetrahydrofuran (20 mL) was heated at reflux for 16 hours. The mixture was cooled to ambient temperature and concentrated to dryness under vacuum. The solid residue was suspended in 1:I ether/hexane and filtered to remove triphenylphosphine oxide.
The filtrate was filtered through a pad of Si02 using 1:1 ether/hexane to provide the product (1.78 g, 91%) as a pale yellow oil.
(2) 2-Methyl-3-(3-thienyl)acrylic acid: The ester from (1) was dissolved in a mixture of methanol (5 mL) and 8 N KOH (5 mL) and heated at reflux for 30 minutes. The mixture was cooled to ambient temperature, diluted with water, and washed twice with ether. The aqueous phase was acidified using 1 N HCl then extracted 3 times with equal volumes of ether. The organic extracts were combined, dried with MgS04, filtered, and concentrated to dryness under vacuum. Crystallization from 2:1 hexane/ether provided the product as colorless needles.
(3) 2-Methyl-3-(3-thienyl)acrylate N-acetylcysteamine thioester: A solution of 2-Methyl-3-(3-thienyl)acrylic acid (168 mg) in 2 mL of dry tetrahydrofuran under inert atmosphere was treated with triethylamine (0.56 mL) and diphenylphosphoryl azide (0.45 mL). After 15 minutes, N-acetylcysteamine (0.15 mL) is added and the reaction is allowed to proceed for 4 hours. The mixture is poured into water and extracted three times with equal volumes of ethyl acetate. The organic extracts are combined, washed sequentially with water, 1 N HCI, sat. CuS04, and brine, then dried over MgS04,filtered, and concentrated under vacuum. Chromatography on Si02 using ethyl acetate provided pure product, which crystallized upon standing.
The above compounds are supplied to cultures of host cells containing a recombinant epothilone PKS of the invention in which either the NRPS or the KS
domain of module 2 as appropriate has been inactivated by mutation to prepare the corresponding epothilone derivative of the invention.
Example 7 Producing Epothilones and Epothilone Derivatives in Sorangium cellulosum SMP44 The present invention provides a variety of recombinant Sorangium cellulosum host cells that produce less complex mixtures of epothilones than the naturally occurnng epothilone producers as well as host cells that produce epothilone derivatives. This Example illustrates the construction of such strains by describing how to make a strain that produce only epothilones C and D without epothilones A and B. To construct this strain, an inactivating mutation is made in epoK. Using plasmid pKOS35-83.5, which contains a NotI fragment harboring the epoK gene, the kanamycin and bleomycin resistance markers from Tn5 are ligated into the ScaI site of the epoK gene to construct pKOS90-55. The orientation of the resistance markers is such that transcription initiated at the kanamycin promoter drives expression of genes immediately downstream of epoK. In other words, the mutation should be nonpolar. Next, the origin of conjugative transfer, oriT, from RP4 is ligated into pKOS90-55 to create pKOS90-63. This plasmid can be introduced into S17-1 and conjugated into SMP44. The transconjugants are selected on phleomycin plates as previously described. Alternatively, electroporation of the plasmid can be achieved using conditions described above for Myxococcus xanthus.
Because there are three generalized transducing phages for Myxococcus xanthus, one can transfer DNA from M. xanthus to SMP44. First, the epoK mutation is constructed in M. xanthus by linearizing plasmid pKOS90-55 and electroporating into M.
xanthus.
Kanamycin resistant colonies are selected and have a gene replacement of epoK.
This strain is infected with Mx9, MxB, Mx4 is 18 hft hrm phages to make phage lysates. These lysates are then individually infected into SMP44 and phleomycin resistant colonies are selected. Once the strain is constructed, standard fermentation procedures, as described 1 S below, are employed to produce epothilones C and D.
Prepare a fresh plate of Sorangium host cells (dispersed) on S42 medium. S42 medium contains tryptone, 0.5 g/L; MgS04, 1.5 g/L; HEPES, 12 g/L; agar, 12 g/L, with deionized water. The pH of S42 medium is set to 7.4 with KOH. To prepare S42 medium, after autoclaving at 121 °C for at least 30 minutes, add the following ingredients (per liter):
CaCl2, 1 g; K2HP04, 0.06 g; Fe Citrate, 0.008 g; Glucose, 3.5 g; Ammonium sulfate, 0.5 g;
Spent liquid medium, 35 mL; and 200 micrograms/mL of kanamycin is added to prevent contamination. Incubate the culture at 32°C for 4-7 days, or until orange sorangia appear on the surface.
To prepare a seed culture for inoculating agar plates/bioreactor, the following protocol is followed. Scrape off a patch of orange Sorangium cells from the agar (about S
mm2) and transfer to a 250 ml baffle flask with 38 mm silicone foam closures containing 50 ml of Soymeal Medium containing potato starch, 8 g; defatted soybean meal, 2 g; yeast extract, 2 g; Iron (III) sodium salt EDTA, 0.008 g; MgS04.7H20, 1 g;
CaC12.2H20, 1 g;
glucose, 2 g; HEPES buffer, 11.5 g. Use deionized water, and adjust pH to 7.4 with 10%
KOH. Add 2-3 drops of antifoam B to prevent foaming. Incubate in a coffin shaker for 4-5 days at 30°C and 250 RPM. The culture should appear an orange color.
This seed culture can be subcultured repeatedly for scale-up to inoculate in the desired volume of production medium.

WO 00/31247 - 117 _ PCT/US99/27438 The same preparation can be used with Medium 1 containing (per liter) CaC12.2H20, 1 g; yeast extract, 2 g; Soytone, 2 g; FeEDTA, 0.008 g; Mg S04.7H20, 1 g;
HEPES, 11.5 g. Adjust pH to 7.4 with 10% KOH, and autoclave at 121°C
for 30 minutes.
Add 8 ml of 40% glucose after sterilization. Instead of a baffle flask, use a 250 ml coiled spring flask with a foil cover. Include 2-3 drops of antifoam B, and incubate in a coffin shaker for 7 days at 37°C and 250 RPM. Subculture the entire 50 mL into 500 mL of fresh medium in a baffled narrow necked Fernbach flask with a 38 mm silicone foam closure.
Include 0.5 ml of antifoam to the culture. Incubate under the same conditions for 2-3 days.
Use at least a 10% inoculum for a bioreactor fermentation.
To culture on solid media, the following protocol is used. Prepare agar plates containing (per liter of CNS medium) KN03, 0.5 g; Na2HP04, 0.25 g; MgS04.7H20, 1 g;
FeCl2, 0.01 g; HEPES, 2.4 g; Agar, 15 g; and sterile Whatman filter paper.
While the agar is not completely solidified, place a sterile disk of filter paper on the surface. When the plate is dry, add just enough of the seed culture to coat the surface evenly (about 1 mL).
Spread evenly with a sterile loop or an applicator, and place in a 32°C
incubator for 7 days. Harvest plates.
For production in a 5 L bioreactor, the following protocol is used. The fermentation can be conducted in a B. Braun Biostat MD-1 SL bioreactor.
Prepare 4 L of production medium (same as the soymeal medium for the seed culture without HEPES
buffer). Add 2% (volume to volume) XAD-16 absorption resin, unwashed and untreated, e.g. add 1 mL of XAD per SO mL of production medium. Use 2.5 N H2S04 for the acid bottle, 10% KOH for the base bottle, and SO% antifoam B for the antifoam bottle. For the sample port, be sure that the tubing that will come into contact with the culture broth has a small opening to allow the XAD to pass through into the vial for collecting daily samples.
Stir the mixture completely before autoclaving to evenly distribute the components.
Calibrate the pH probe and test dissolved oxygen probe to ensure proper functioning. Use a small antifoam probe, ~3 inches in length. For the bottles, use tubing that can be sterile welded, but use silicone tubing for the sample port. Make sure all fittings are secure and the tubings are clamped off, not too tightly, with C-clamps. Do not clamp the tubing to the exhaust condenser. Attach 0.2 pm filter disks to any open tubing that is in contact with the air. Use larger ACRD 50 filter disks for larger tubing, such as the exhaust condenser and the air inlet tubing. Prepare a sterile empty bottle for the inoculum.
Autoclave at 121 °C
with a sterilization time of 90 minutes. Once the reactor has been taken out of the WO 00/31247 _ 11 g _ PCT/US99/27438 autoclave, connect the tubing to the acid, base, and antifoam bottles through their respective pump heads. Release the clamps to these bottles, making sure the tubing has not been welded shut. Attach the temperature probe to the control unit. Allow the reactor to cool, while sparging with air through the air inlet at a low air flow rate.
After ensuring the pumps are working and there is no problem with flow rate or clogging, connect the hoses from the water bath to the water jacket and to the exhaust condenser. Make sure the water jacket is nearly full. Set the temperature to 32°C. Connect pH, D.O., and antifoam probes to the main control unit. Test the antifoam probe for proper functioning. Adjust the set point of the culture to 7.4. Set the agitation to 400 RPM.
Calibrate the D.O. probe using air and nitrogen gas. Adjust the airflow using the rate at which the fermentation will operate, e.g. 1 LPM (liter per minute). To control the dissolved oxygen level, adjust the parameters under the cascade setting so that agitation will compensate for lower levels of air to maintain a D.O. value of SO%. Set the minimum and maximum agitation to 400 and 1000 RPM respectively, based on the settings of the control unit. Adjust the settings, if necessary.
Check the seed culture for any contamination before inoculating the fermenter.
The Sorangium cellulosum cells are rod shaped like a pill, with 2 large distinct circular vacuoles at opposite ends of the cell. Length is approximately 5 times that of the width of the cell. Use a 10% inoculum (minimum) volume, e.g. 400 mL into 4 L of production medium. Take an initial sample from the vessel and check against the bench pH.
If the difference between the fermenter pH and the bench pH is off by > 0. I units, do a 1 point recalibration. Adjust the deadband to 0.1. Take daily 25 mL samples noting fermenter pH, bench pH, temperature, D.O., airflow, agitation, acid, base, and antifoam levels. Adjust pH
if necessary. Allow the fermenter to run for seven days before harvesting.
Extraction and analysis of compounds is performed substantially as described above in Example 4. In brief, fermentation culture is extracted twice with ethyl acetate, and the ethyl acetate extract is concentrated to dryness and dissolved/suspended in 500 p.L of MeCN-HZO (1:1). The sample is loaded onto a 0.5 mL Bakerbond ODS SPE
cartridge pre-equilibrated with MeCN-H20 ( 1: I ). The cartridge is washed with 1 mL of the same solvent, followed by 2 mL of MeCN. The MeCN eluent is concentrated to dryness, and the residue is dissolved in 200 p,L of MeCN. Samples (50 pL) are analyzed by HPLC/MS on a system comprised of a Beckman System Gold HPLC and PE Sciex APIl00LC single quadrapole MS-based detector equipped with an atmospheric pressure chemical ionization source. Ring and orifice voltages are set to 75V and 300V, respectively, and a dual range mass scan from m/z 290-330 and 450-S50 is used.
HPLC
conditions: Metachem Sp, ODS-3 Inertsil (4.6 X 150 mm); 100% H20 for 1 min, then to 100% MeCN over 10 min a 1 mL/min. Epothilone A elutes at 0.2 min under these conditions and gives characteristic ions at m/z 494 (M+H), 476 (M+H-H20), 318, and 306.
Example 8 Epothilone Derivatives as Anti-Cancer Agents The novel epothilone derivatives shown below by Formula ( 1 ) set forth above are potent anti-cancer agents and can be used for the treatment of patients with various forms of cancer, including but not limited to breast, ovarian, and lung cancers.
The epothilone structure-activity relationships based on tubulin binding assay are (see Nicolaou et al., 1997, Angew. Chem. Int. Ed. Engl. 36: 2097-2103, incorporated herein by reference) are illustrated by the diagram below.
H G E
D
S ~ ~'v a I
,,, ,,~~OH
,.,,y' C
H
A) (3S) configuration important; B) 4,4-ethano group not tolerated; C) (6R, 7S
) configuration crucial; D) (8S) configuration important, 8,8-dimethyl group not tolerated;
E) epoxide not essential for tubulin polymerization activity, but may be important for cytotoxicity; epoxide configuration may be important; R group important; both olefin geometries tolerated; F) (15S) configuration important; G) bulkier group reduces activity;
H) oxygen substitution tolerated; I) substitution important; ~ heterocycle important.
Thus, this SAR indicates that modification of the C1-C8 segment of the molecule can have strong effects on activity, whereas the remainder of the molecule is relatively tolerant to change. Variation of substituent stereochemistry with the C1-C8 segment, or removal of the functionality, can lead to significant loss of activity.
Epothilone derivative compounds A-H differ from epothilone by modifications in the less sensitive portion of the molecule and so possess good biological activity and offer better pharmacokinetic characteristics, having improved lipophilic and steric profiles.
These novel derivatives can be prepared by altering the genes involved in the biosynthesis of epothilone optionally followed by chemical modification. The 9-hydroxy-epothilone derivatives prepared by genetic engineering can be used to generate the carbonate derivatives (compound D) by treatment with triphosgene or 1,1' carbonyldiimidazole in the presence of a base. In a similar manner, the 9,11-dihydroxy-epothilone derivative, upon proper protection of the C-7 hydroxyl group if it is present, yields the carbonate derivatives (compound F). Selective oximation of the 9 oxo-epothilone derivatives with hydroxylamine followed by reduction (Raney nickel in the presence of hydrogen or sodium cyanoborohydride) yield the 9-amino analogs.
Reacting these 9-amino derivatives with p-nitrophenyl chloroformate in the presence of base and subsequently reacting with sodium hydride will produce the carbamate derivatives (compound E). Similarly, the carbamate compound G, upon proper protection of the C7 hydroxyl group if it is present, can be prepared form the 9-amino-11 hydroxy-epothilone derivatives.
Illustrative syntheses are provided below.
Part A. Epothilone D -7, 9-cyclic carbonate To a round bottom flask, a solution of 254 mg epothilone D in 5 mL of methylene chloride is added. It is cooled by an ice bath, and 0.3 mL of triethyl amine is then added.
To this solution, 104 mg of triphosgene is added. The ice bath is removed, and the mixture is stirred under nitrogen for 5 hours. The solution is diluted with 20 mL of methylene chloride and washed with dilute sodium bicarbonate solution. The organic solution is dried over magnesium sulfate and filtered. Upon evaporation to dryness, the epothilone D-7, 9 -cyclic carbonate is isolated.
Part B. Epothilone D-7,9-cyclic carbamate (i) 9-amino-epothilone D
To a rounded bottom flask, a solution of 252 mg 9-oxo-epothilone D in S mL of methanol is added. Upon the addition of 0.5 mL 50% hydroxylamine in water and 0.1 mL

acetic acid, the mixture is stirred at room temperature overnight. The solvent is then removed under reduced pressure to yield the 9-oxime-epothilone D. To a solution of this 9 oxime compound in 5 mL of tetrahydrofuran (THF) at ice bath is added 0.25 mL 1 M
solution of cyanoborohydride in THF. After the mixture is allowed to react for 1 hour, the ice bath is removed, and the solution is allowed to warm slowly to room temperature. One mL of acetic acid is added, and the solvent is then removed under reduced pressure. The residue is dissolved in 30 mL of methylene chloride and washed with saturated sodium chloride solution. The organic layer is separated and dried over magnesium sulfate and filtered. Upon evaporation of the solvent yields the 9-amino-epothilone D.
(ii) Epothilone D-7,9-cyclic carbamate To a solution of 250 mg of 9-amino-epothilone D in 5 mL of methylene is added 110 mg of 4-nitrophenyl chloroformate followed by the addition of 1 mL of triethylamine.
The solution is stirred at room temperature for 16 hours. It is diluted with 25 mL of methylene chloride. The solution is washed with saturated sodium chloride and the organic layer is separated and dried over magnesium sulfate. After filtration, the solution is evaporated to dryness at reduced pressure. The residue is dissolved in 10 mL
of dry THF.
Sodium hydride, 40 mg (60% dispersion in mineral oil), is added to the solution in an ice bath. The ice bath is removed, and the mixture is stirred for 16 hours. One-half mL of acetic acid is added, and the solution is evaporated to dryness under reduced pressure. The residue is re-dissolved in 50 mL methylene chloride and washed with saturated sodium chloride solution. The organic layer is dried over magnesium sulfate and the solution is filtered and the organic solvent is evaporated to dryness under reduced pressure. Upon purification on silica gel column, the epothilone D-7,9-carbamate is isolated.
The invention having now been described by way of written description and examples, those of skill in the art will recognize that the invention can be practiced in a variety of embodiments and that the foregoing description and examples are for purposes of illustration and not limitation of the following claims.

Claims (66)

Claims

1. An isolated recombinant nucleic acid compound that comprises a nucleotide sequence encoding at least a domain of an epothilone polyketide synthase (PKS) protein and/or encoding a functional region of an epothilone modification enzyme.

2. The nucleic acid of claim 1, wherein said domain is selected from the group consisting of a loading domain, a thioesterase domain, a non-ribosomal peptide synthase (NRPS), an acyltransfase (AT) domain, a ketosynthase (KS) domain, an acyl carrier protein (ACP) domain, a ketoreductase (KR) domain, a dehydratase (DH) domain, and an enolyreductase (ER) domain, a methyl transferase domain and a functional oxidase domain.

3. The nucleic acid of claim 1 or 2 that comprises the coding sequence of an epoA gene, and/or the coding sequence of an epoB gene, and/or the coding sequence of an epoC gene, and/or the coding sequence of an epoD gene, and/or the coding sequence of an epoE gene, and/or the coding sequence of an epoF gene, and/or the coding sequence of an epoK gene, and/or the coding sequence of an epoL gene.

4. The nucleic acid of claim 1 that further comprises a promoter positioned to transcribe said encoding nucleotide sequence in host cells in which said promoter is operable.

5. The nucleic acid of claim 4, wherein said promoter is a promoter from a Sorangium gene, or from a Myxococcus gene, or from a Streptomyces gene, or from an epothilone PKS gene, or from a pilA gene, or from an actinorhodin PKS gene.

6. The nucleic acid of any of claims 1-5 that is a recombinant DNA
expression vector.

7. Host cells which contain the nucleic acid of any of claims 4-6.

8. The cells of claim 7 which are Sorangium cells, or Myxococcus cells, or Pseudomonas cells, or Streptomyces cells.

9. A method to produce a polyketide which method comprises culturing the cells of claim 7 under conditions wherein the encoding nucleotide sequence is expressed to obtain a functional PKS.

10. A recombinant Sorangium cellulosum host cell that contains a mutated gene for an epothilone PKS protein or epothilone modification enzyme, wherein said mutated gene was inserted in whole or in part into genomic DNA of said cell by homologous recombination with a recombinant vector comprising all or a part of an epothilone PKS gene or epothilone modification gene.

11. The recombinant host cell of claim 10 that makes epothilone C or D but not A or B due to a mutation inactivating or deleting the epoK gene, or makes epothilone A or C but not B or D due to a mutation in epoD altering module 4 AT domain specificity, or makes epothilone B or D but not A or C due to a mutation in epoD altering module 4 AT domain specificity, or makes epothilone C but not epothilone A, B or D due to a mutation in epoD
altering module 4 AT domain specificity and a mutation in epoK, or makes epothilone D but not epothilone A, B or C due to a mutation in epoD
altering module 4 AT domain specificity and a mutation in epoK.

12. Recombinant Streptomyces or Myxococcus host cells that express an epothilone PKS gene or an epothilone modification enzyme gene, optionally comprising one or more of said epothilone PKS or modification enzyme genes integrated into their chromosomal DNA and/or one or more of said epothilone PKS or modification enzyme genes on an extrachromosomal expression vector.

13. The host cells of claim 12 that are S. coelicolor CH999.

14. A method to produce an epothilone or epothilone derivative which comprises culturing the cells of claims 12 or 13.

15. A modified functional epothilone PKS wherein said modification comprises at least one of:
replacement of at least one AT domain with an AT domain of different specificity;
inactivation of the NRPS-like module 1 or of the KS2 catalytic domain;
inactivation of at least one activity in at least one .beta.-carbonyl modification domain;
addition of at least one of KR, DH and ER activity in at least one .beta.-carbonyl modification domain; and replacement of the NRPS module 1 with an NRPS of different specificity.

16. The modified PKS of claim 15 contained in a cell or contained in a cell-free system, wherein said cell or system contains additional enzymes for modification of the product of said epothilone PKS.

17. The modified PKS of claim 16 wherein said modifying enzymes comprise at least one of a methyltransferase, an oxidase or a glycosylation enzyme.

18. A method to prepare an epothilone derivative which method comprises providing substrates including extender units to the modified PKS of any of claims 15-17.

19. A modified functional epothilone PKS wherein said modification comprises inactivation of the NRPS of module 1 or the KS2 of module 2 thereof.

20. A method to make an epothilone derivative which method comprises contacting the modified PKS of claim 19 with a module 2 substrate or a module substrate and extender units.

21. Recombinant host cells which comprise the modified PKS of any of claims 15-17 or 19.

22. The cells of claim 21 that produce an epothilone derivative selected from the group consisting of 16-desmethyl epothilones, 14-methyl epothilones, hydroxyl epothilones, 10-methyl epothilones, 8,9-anhydro epothilones, 9-hydroxyl epothilones, 9-keto epothilones, 8-desmethyl epothilones, and 6-desmethyl epothilones.

23. A compound selected from the group consisting of 16-desmethyl epothilones, 14-methyl epothilones, 11-hydroxyl epothilones, 10-methyl epothilones, 8,9-anhydro epothilones, 9-hydroxyl epothilones, 9-keto epothilones, 8-desmethyl epothilones, and 6-desmethyl epothilones.

24. A recombinant PKS enzyme that comprises one or more domains, modules, or proteins of a non-epothilone PKS and one or more domains, modules, or proteins of an epothilone PKS, and/or contains a loading domain that comprises a KS Q domain.

25. The PKS enzyme of claim 24, wherein said PKS comprises a deoxyerythronolide B synthase (DEBS) loading domain and 5 modules of DEBS and an NRPS of the epothilone PKS, wherein said PKS comprises all of a non-epothilone PKS with an MT domain of the epothilone PKS.

26. A compound of the formula:

including the glycosylated forms thereof and stereoisomeric forms where the stereochemistry is not shown, wherein A is a substituted or unsubstituted straight, branched chain or cyclic alkyl, alkenyl or alkynyl residue optionally containing 1-3 heteroatoms selected from O, S and N; or wherein A comprises a substituted or unsubstituted aromatic residue;
R2 represents H,H, or H,lower alkyl, or lower alkyl,lower alkyl;
X5 represents =O or a derivative thereof, or H,OH or H,NR2 wherein R is H, alkyl or acyl, or H,OCOR2, H,OCONR2 wherein R is H or alkyl, or is H,H;
R6 represents H or lower alkyl, and the remaining substituent on the corresponding carbon is H;
X7 represents OR, or NR2, wherein R is H, alkyl or acyl or is OCOR, or OCONR2 wherein R is H or alkyl or X7 taken together with X9 forms a carbonate or carbamate cycle, and wherein the remaining substituent on the corresponding carbon is H;
R8 represents H or lower alkyl and the remaining substituent on the carbon is H;
X9 represents =O or a derivative thereof, or H,OR or H,NR2 wherein R is H, alkyl or acyl, or is H,OCOR or H,OCONR2, wherein R is H or alkyl, or represents H,H or wherein X9 together with X7 or with X11 can form a cyclic carbonate or carbamate;
R10 is H,H or H,lower alkyl, or lower alkyl,lower alkyl;
X11 is =O or a derivative thereof, or H,OR, or H,NR2 wherein R is H, alkyl or acyl or H,OCOR or H,OCONR2 wherein R is H or alkyl, or is H,H or wherein X11 in combination with X9 may form a cyclic carbonate or carbamate;
R12 is H,H, or H,lower alkyl, or lower alkyl,lower alkyl;

X13 is =O or a derivative thereof, or H,OR or H,NR2 wherein R is H, alkyl or acyl or is H,OCOR or H,OCONR2 wherein R is H or alkyl;
R14 is H,H, or H,lower alkyl, or lower alkyl,lower alkyl;
R16 is H or lower alkyl; and wherein optionally H or another substituent may be removed from positions 12 and 13 and/or 8 and 9 to form a double bond, wherein said double bond may optionally be converted to an epoxide.

27. A compound of the formula wherein both Z are O or one Z is N and the other Z is O and the remaining substituents are defined as in claim 26.

28. A recombinant vector selected from the group consisting of pKOS35-70.8A3, pKOS35-70.1A2, pKOS35-70.4, pKOS35-79.85, pKOS039-124R, and pKOS039-126R.

29. Isolated or recombinant host cells that contain a nucleic acid that codes for expression of Sorangium cellulosum epothilone polyketide synthase ("PKS") but does not contain a functionally active product of a Sorangium cellulosum epoK
gene whereby the host cells produce more epothilone C than epothilone A or more epothilone D than epothilone B.

30. The cells of claim 29 which are Sorangium cells, or Myxococcus cells, or Pseudomonas cells, or Streptomyces cells.

31. A method to produce epothilone C or epothilone D which method comprises culturing the cells of claim 29 under conditions wherein the nucleic acid is expressed to obtain a functional PKS.

32. A recombinant Sorangium cellulosum host cell that contains a mutated epoK gene and produces more epothilone C than epothilone A or more epothilone D
than epothilone B.

33. The recombinant host cell of claim 32 that makes epothilone C but not epothilone A, B or D due to a mutation in epoD
altering extender module 4 acyltransferase ("AT") domain specificity or makes epothilone D but not epothilone A, B or C due to a mutation in epoD
altering extender module 4 AT domain specificity.

34. Recombinant Streptomyces or Myxococcus host cells that express Sorangium cellulosum epothilone polyketide synthase ("PKS") genes and do not express a fully functional Sorangium cellulosum epoK gene product, optionally comprising one or more of said epothilone PKS genes integrated into their chromosomal DNA or in an extrachromosomal expression vector, wherein said recombinant cells produce more epothilone C than epothilone A or more epothilone D
than epothilone B.

35. A method to produce epothilone C or D which comprises culturing the cells of claims 33 or 34.

36. The cells of claim 29, wherein said cells do not contain a Sorangium cellulosum epoK gene.

37. The cells of claim 29, wherein said cells contain an altered Sorangium cellulosum epoK gene and said altered epoK gene produces a gene product that does not convert epothilone C to epothilone A or epothilone D to epothilone B.

38. The cells of claim 29, wherein said cells produce both epothilone C
and epothilone D.

39. The cells of claim 29, wherein said cells do not produce epothilone A
or epothilone B.

40. The cells of claim 29, wherein said nucleic acid comprises epoA, epoB, epoC, epoD, epoE and epoF genes.

41. The cells of claim 29, wherein said nucleic acid comprises an altered epoD extender module 4 acyltransferase ("AT") domain coding sequence.

42. The cells of claim 41, wherein said altered AT domain coding sequence results in increased production of epothilone C and decreased production of epothilone D as compared to cells containing an unaltered AT domain coding sequence.

43. The cells of claim 41, wherein said altered AT domain coding sequence results in increased production of epothilone D and decreased production of epothilone C as compared to cells containing an unaltered AT domain coding sequence.

44. The cells of claim 30 that are Streptomyces cells.

45. The cells of claim 36 that are Streptomyces cells.

46. The cells of claim 38 that are Streptomyces cells.

47. The cells of claim 39 that are Streptomyces cells.

48. The cells of claim 30 that are Myxococcus cells.

49. The cells of claim 36 that are Myxococcus cells.

50. The cells of claim 38 that are Myxococcus cells.

51. The cells of claim 39 that are Myxococcus cells.

52. The cell of claim 32, wherein said mutated epoK gene comprises an insertion or deletion relative to an unmutated epoK gene.

53. The cell of claim 32, wherein said mutated epoK gene is produced by random mutagenesis.

54. The cell of claim 32, wherein said mutated epoK gene is produced by homologous recombination.

55. The cell of claim 32 that does not produce epothilone A or epothilone B.

56. The cells of claim 34, wherein said cells do not contain a Sorangium cellulosum epoK gene.

57. The cells of claim 34, wherein said cells contain an altered Sorangium cellulosum epoK gene and said altered epoK gene produces a gene product that does not convert epothilone C to epothilone A or epothilone D to epothilone B.

58. The cells of claim 34, wherein said cells do not produce epothilone A
or epothilone B.

59. The cells of claim 34 that contain epoA, epoB, epoC, epoD, epoE and epoF PKS genes.

60. The cells of claim 59, wherein said epoD gene comprises an altered extender module 4 acyltransferase ("AT") domain coding sequence.

61. The cells of claim 60, wherein said altered AT domain coding sequence results in increased production of epothilone C and decreased production of epothilone D as compared to cells containing an unaltered AT domain coding sequence.

62. The cells of claim 60, wherein said altered AT domain coding sequence results in increased production of epothilone D and decreased production of epothilone C as compared to cells containing an unaltered AT domain coding sequence.

63. The method of claim 31, wherein both epothilone C and epothilone D
are produced.

64. The method of claim 31, wherein epothilone D is produced.

65. The method of claim 35, wherein epothilones C and D are produced.

66. The method of claim 35, wherein epothilone D is produced.