Cloning and characterization of a gene cluster for geldanamycin production in Streptomyces hygroscopicus NRRL 3602


*Corresponding author, E-mail address:


We illustrate the use of a PCR-based method by which the genomic DNA of a microorganism can be rapidly queried for the presence of type I modular polyketide synthase genes to clone and characterize, by sequence analysis and gene disruption, a major portion of the geldanamycin production gene cluster from Streptomyces hygroscopicus var. geldanus NRRL 3602.


Geldanamycin [1–4] (Fig. 1) and several other benzoquinone microbial products classified as ansamycin antibiotics[5], herbimycin A [6,7], macbecins[8], ansatrienins[9] and reblastatin [10,11] were discovered between 1970 and 2000 in screens for antibacterial, antifungal and antiviral compounds. Interest in the benzoquinone ansamycins increased greatly upon the discovery of the antitumor properties of geldanamycin[12] and herbimycin A[13], and the remarkable cytotoxic properties of the ansamitocins[14].

Figure 1.

Main steps in the proposed biosynthesis of geldanamycin. The enzyme-bound product shown is assembled from AHBA and the carbon chain extender substrates, malonyl-CoA, 2-methoxymalonate and 2-methylmalonyl-CoA, by the PKS modules indicated beneath the product structure. The dashed line at position 4 signifies that the 4,5-C=C may be introduced by the PKS or afterwards as in A. Abbreviations: ACP, acyl carrier protein; LDD, loading didomain; mod, module. The order of the four tailoring steps that convert progeldanamycin into geldanamycin is unknown.

It was initially believed that geldanamycin and herbimycin A interfered with signal transduction pathways in tumor cells by direct inhibition of oncogenic protein tyrosine kinases. However, Neckers and co-workers[15] showed in 1994 that the principal cellular target is not a tyrosine kinase but Hsp90, a ubiquitous and abundant protein chaperone of mammalian cells[16]. Geldanamycin competes with ATP for the ATP-binding site of Hsp90 and, when bound, inhibits the ATP-dependent functions of Hsp90. A particular function is its ability to chaperone nascent protein kinases that are critical components of signal transduction pathways, especially those in certain cancer cells[16]. In the presence of geldanamycin, herbimycin A or macbecin, the immature kinases undergo rapid degradation, as a consequence of ubiquitination and subsequent catabolism by the proteosome, and the levels of the mature kinases become depleted. This can result in a cytostatic effect on a cancer cell or in some cases apoptosis and cell death.

The discovery that Hsp90 and one or more of its protein kinase cohorts are overproduced in several types of human cancers has led to considerable interest in geldanamycin and its analogs as potential anti-cancer drugs[17]. Many geldanamycin analogs have been produced by replacement of the C17 O-methoxy group with substituted amines. One such drug, 17-allylamino-17-demethoxygeldanamycin, is currently undergoing Phase I clinical trials[18].

We are interested in engineering the geldanamycin polyketide synthase (PKS) genes to make novel geldanamycin analogs. To obtain the genes from the producer, Streptomyces hygroscopicus NRRL 3602, we faced a set of challenges common to the isolation of any PKS gene cluster from a streptomycete: bacteria commonly produce more than one polyketide metabolite (e.g. Streptomyces coelicolor has 10[19] and Streptomyces avermitilis at least 13[20] PKS gene clusters). We therefore developed a reliable method for querying the genomic DNA of the NRRL 3602 strain for the number of modular PKS genes, then used this method to help clone most of the geldanamycin production genes.

2Materials and methods

2.1Bacterial strains and culture conditions

The geldanamycin-producing strain, first described by DeBoer et al. [1,2] as S. hygroscopicus var. geldanus var. nova UC-5208, was obtained from the Northern Regional Research Laboratory of the Agricultural Research Service as S. hygroscopicus NRRL 3602. To confirm production of geldanamycin, spores were used to inoculate 5 ml of R2YE liquid media[38]. The culture was incubated at 28°C for 36 h, transferred into 100 ml geldanamycin production medium[1] and the final culture was incubated at 28°C in 500-ml Erlenmeyer flasks at 300 rpm for another 5 days. Following low-speed centrifugation, the cell pellet from the culture was extracted with methanol by stirring for 10 min. The methanol broth was clarified by centrifugation (17 500×g) and the supernatant was analyzed for the presence of geldanamycin using high-performance liquid chromatography (HPLC) under the following conditions: column Inertsil C18 (4.6×150 mm, Ansys Technologies, Inc.), mobile phase 60% acetonitrile (isocratic), flow rate (2 ml min−1), temperature (40°C), detection (UV 315 nm), injection volume (10–20 μl). Geldanamycin (Sigma-Aldrich) was quantified by comparing the peak area at 315 nm with that measured for a standard solution. The titer of geldanamycin was approximately 250 mg l−1.

2.2Manipulation of DNA and organisms

For genomic DNA extraction, a spore stock was used to prepare a seed culture as described above. The entire seed culture was transferred into 50 ml of the same growth medium in a 250-ml baffled Erlenmeyer flask and incubated for 48 h at 28°C. A 20-ml portion of the cell suspension was centrifuged (10 000×g) and the resulting pellet was washed with 10 ml buffer 1 (Tris, 50 mM, pH7.5; 20 mM EDTA). The pellet was pulverized with mortar and pestle under liquid nitrogen and transferred into 3.5 ml of buffer containing 150 μg ml−1 RNase (Sigma-Aldrich). After incubation of the mixture at 30°C for 20 min, the salt concentration was adjusted by adding 850 μl 5 M NaCl solution, then the mixture was extracted multiple times with phenol:chloroform:isoamylaclohol (25:24:1, vol/vol) with gentle agitation followed by centrifugation for 10 min at 3500×g. After precipitation with 1 vol of isopropanol, the genomic DNA knot was spooled on a glass rod and redissolved in water (200 μl). This method yielded about 1 mg DNA with a protein factor of about 2, as determined by the ratio of the UV absorbances at 260 and 280 nm. Standard agarose gel electrophoresis using 0.7% Seakem® LE-Agarose (BioWhiaker Molecular Applications, Rockland, ME) at a voltage of 50 mV overnight revealed that the sample contained mainly high-molecular-mass DNA fragments of about 60 kb.

2.3Genomic analysis of S. hygroscopicus NRRL 3602

The following degenerate keto synthase (KS) primers were used to scan the genomic DNA of S. hygroscopicus for PKS genes: degKS1F (5′-TTCGAYSCSGVSTTCTTCGSAT-3′), degKS2F (5′-GCSATGGAYCCSCARCARCGSVT-3′), degKS3F (5′-SSCTSGTSGCSMTSCAYCWSGC-3′), degKS5R (5′-GTSCCSGTSCCRTGSSCYTCSAC-3′), degKS6R (5′-TGSGYRTGSCCSAKGTTSSWCTT-3′) and degKS7R (5′-ASRTGSGCRTTSGTSCCSSWSA-3′). Forward (F) and reverse (R) primers were tested in all possible combinations in standard PCR reactions with annealing temperatures between 50 and 60°C. A typical 50 μl PCR reaction consisted of 200 ng genomic DNA, 200 pmol of each primer, 0.2 mM dNTP (containing 7-deaza-dGTP), 10% DMSO and 2.5 U Taq DNA polymerase (Roche Applied Science).

Carbamoyl transferase gene fragments were amplified with degenerate forward primer degCT2F (5′-AARGTSATGGGSYTSGCSCCSTA-3′) and reverse primers degCT3R (5′-CCSARSGCSCKSGGSCCRAAYTC-3′) or degCT4R (5′-CKSGCSSWSCCRTCSACRTGSGT-3′) using an annealing temperature of 55°C.

3-Amino-5-hydroxybenzoate (AHBA) synthase gene fragments were amplified with a set of two degenerate forward primers, degAH-F1 (5′-GTSATCGTSCCSGCSTTCACSTTC-3′) and degAH-F2 (5′-ATCATGCCSGTSCAYATGGCSGG-3′), and two reverse primers, degAH-R1 (5′-GGSTBSGKGAACATSGCCATGTA-3′) and degAH-R2 (5′-CKRTGRTGSARCCASTKRCARTC-3′). Again forward and reverse primers were tested in all combinations in the described standard PCR reactions with annealing temperatures between 50 and 65°C.

A set of four gene specific primers, AH-B-spF (5′-AGGACAGTGGCGCGGCAAGAA-3′), AH-B-spR (5′-GGTCGACGATCTTCGCGCGGCG-3′), AH-N-spF (-5′-TCGACGTGGCTGCCGCGGCTT-3′) and AH-N-spR (5′-TGTCGACGAGGGCGTTGCGGG-3′), was used to distinguish between AHBA-B and AHBA-N synthase genes.

PCR amplimers were gel-purified and cloned into pCR2.1-TOPO using TA cloning (Invitrogen). For each primer pair, a representative set of cloned amplimers (600–800 bp) was sequenced using a Beckman CEQ2000 with M13 forward and reverse primers.

2.4Library construction and gene isolation

A genomic library of the NRRL3602 strain was constructed at TIGR using the proprietary single-copy BAC vector pHOS3. Sheared genomic DNA was made blunt-ended using T4 DNA polymerase and Bst XI adaptors were ligated to the blunt ends. Adaptors were removed and inserts were size-selected by three rounds of field inversion gel electrophoresis. The inserts were ligated to the vector and electroporated into Escherichia coli. A total of 4896 BAC clones were arrayed into 384-well microtiter plates and spotted in high density onto nylon filters (Amplicon Express). Probes were labeled using [α-32P]dCTP and a random prime-labeling system (redi prime II, Amersham Pharmacia Biotech). Filters were hybridized at 68°C for 12 h using ExpressHyb hybridization solution (Clontech). After removal of the probe and hybridization solution, the filter was washed twice for 30 min each time with 100 ml of buffer I [2×SSC: 300 mM NaCl, 30 mM sodium citrate pH 7.0, 0.05% sodium dodecyl sulfate (SDS)] at room temperature and then three times for 60 min each time at 50°C with 100 ml of buffer II (0.1×SSC, 0.1% SDS) with continuous shaking. Finally, the filter was rinsed several times with 0.05×SSC and analyzed by autoradiography.

BAC DNA was prepared by alkaline lysis, starting with 10 ml culture volume. The resulting DNA was first treated with RNase (Sigma-Aldrich) at 30°C for 3 h and then with plasmid safe DNase (Epicentre Technologies) at 37°C o/n. After heat inactivation at 70°C for 10 min the DNA was precipitated with 1 volume isopropanol for 30 min on ice and recovered by centrifugation at 1880×g for 45 min to separate the remaining smaller fragments from the large, intact BAC plasmids. The final pellet was washed with 70% EtOH and redissolved in 80 μl water. This method typically yielded about 100 μg of BAC DNA.

2.5DNA sequence analysis

BAC DNA was sequenced to completion using the standard shotgun procedures[39]. The DNA and deduced protein sequences were analyzed with Sequencer 4.1 (Gene Codes Corporation) and MacVector 6.5.3 (Accelrys) software, and compared with sequences in the public databases using the CLUSTAL W[40] and BLAST[41] programs. The sequence data for the genes displayed in Fig. 2 have been deposited with GenBank (accession number AY179507).

Figure 2.

Geldanamycin gene cluster. The colored broad arrows indicate the genes and their direction of transcription. PKS and gdmF genes are red, 2-methoxymalonate synthesis genes are orange, the AHBA synthesis gene is green and tailoring enzyme genes are yellow. The right end of the cluster continues beyond ORF22. The domains in each of the seven modules of the gdm PKS are indicated by the abbreviations above the gdmA arrows (see Table 1 for definitions). Scale beneath the genes indicates the distance in kb from the beginning of the 85-kb region sequenced.

2.6Disruption of the gdmH gene

The gdmH gene disruption vector was made as follows. The aphII neomycin/kanamycin resistance gene from Tn5 was excised as a Stu I–Sma I fragment from SuperCos-1 (Stratagene), then inserted into the Msc I site within gdmH carried in a 4-kb Bst XI fragment, containing the gdmN, gdmH and gdmI genes, and cloned in pOJ260[42] to give pKOS246-33. The Xba I–Eco RI fragment from pKOS246-33 was excised and cloned into the Xba I–Eco RI sites of pKC1139[42] to give pKOS279-37.

The gdmH gene was disrupted by introducing pKOS279-37 into the S. hygroscopicus NRRL3602 strain by conjugation from E. coli ET12567/pUB307 according to a published method[43]. Exconjugants resistant to apramycin (pKC1139 carries the accIV(3) gene) and kanamycin were isolated and one of them was grown at 30°C in 6 ml of R5 liquid medium[38] supplemented with 100 μg ml−1 of kanamycin for 2 days in 50-ml culture tubes at 200 rpm. Approximately 5% of this culture was transferred into 6 ml of fresh R5/apramycin liquid medium and the culture was grown at 37°C for 3 days in order to force chromosomal integration of pKOS279-37. After recovery of the mycelia by centrifugation, cells were plated on tomato paste medium containing 100 μg ml−1 kanamycin and grown at 30°C for sporulation. Spores collected from these plates were diluted and replated on the same medium for single colonies. Among 100 colonies screened, 20 of them were apramycin sensitive and kanamycin resistant when assayed on plates containing apramycin or kanamycin, using 60 or 50 μg ml−1 of antibiotic, respectively. Genomic DNA was isolated from 11 of these 20 colonies by an established method[38] and probed by Southern-blot hybridization[38] with the aphII gene to determine that all kanamycin resistant recombinant strains had the restriction fragment pattern upon digestion with Pst I–Eco RV expected for integration of the aphII gene into the gdmH locus by a double crossover recombination (hybridizing bands at 2.9 and 3.2 kb that were absent in the NRRL3602 strain).

To determine geldanamycin production, each of the 11 strains was individually cultured in 35 ml of the geldanamycin production medium[1] as described above. After 4 days, 500 μl of broth from each flask was mixed with 500 μl of methanol, the mixture was centrifuged at 12,000 rpm in a desktop microcentrifuge for 5 min to remove mycelia and other insoluble ingredients, then the supernatant fraction was analyzed by HPLC/MS. The results showed that geldanamycin was present (retention time and low-resolution MS data were identical to the reference standard) and that two new compounds were present with molecular masses and formulas of 518.2759 (C28H40NO8[M-H]) and 520.2916 (C28H42NO8[M-H]), calculated on the basis of high-resolution MS data. These data are consistent with 4,5-dihydro-7-descarbamoyl-7-hydroxygeldanamycin and its hydroquinone form.

3Results and discussion

3.1Survey of modular PKS genes in S. hygroscopicus NRRL 3602

Geldanamycin biosynthesis can be depicted as shown in Fig. 1 on the basis of isotope-labeling studies by the Rinehart laboratory [4,21]. That work forecast the involvement of a PKS that would use AHBA [22–25] as the chain initiation substrate for a seven-module type I PKS, and malonate, 2-methylmalonate and 2-methoxymalonate, as chain extender substrates to produce the putative progeldanamycin (Fig. 1). Geldanamycin can be formed from this polyketide by four tailoring steps, as shown.

For the initial approach we used standard bioinformatics methods to choose six highly conserved yet recognizably different regions of the amino acid sequence of KS domains among 20 known modular PKSs. Three primer combinations (see Section 2) consistently gave correctly sized PCR products with all organisms tested (data not shown) and were chosen to survey the complexity of PKS genes in the genome of S. hygroscopicus NRRL 3602. Of the 90 KS amplimers cloned and sequenced, 63 represented unique type I PKS KS gene fragments, but when they were analyzed further by comparison with the KS sequences of the ansamitocin[23] and rifamycin[24] PKSs, the evidence that they were part of the geldanamycin PKS genes was not conclusive.

Homologs of the genes for formation of AHBA and for the C7 carbamoylation step were then chosen to design a new set of PCR primers that were used to screen the genomic DNA for AHBA synthase [23–25] and carbamoyltransferase (CT) [23,26,27] homologs. We analyzed 56 AHBA amplimers and identified two different DNA sequences that would encode AHBA synthases that are 75% identical, AHBA-B and AHBA-N. The data, when analyzed by standard bioinformatics methods (Section 2), strongly suggest that one AHBA synthase homolog belongs to the family associated with the biosynthesis of benzoquinone ansamycins (AHBA-B)[23] and the other with naphthoquinone ansamycins (AHBA-N) [24,25]. All 20 CT amplimers analyzed were identical, which let us conclude that there is only one CT gene in this strain.

3.2Cloning and sequence analysis of geldanamycin production genes

The AHBA-B synthase and CT amplimers were used as probes for primary screening of a genomic library made in a single-copy BAC vector, pHOS3, a derivative of pBeloBAC11[28] modified for Bst XI cloning. We screened 4896 BACs with average insert sizes of 45 kb, which was equivalent to ca. 20× coverage of this genome, and identified 36 AHBA synthase and 16 CT positive BACs. AHBA-B and AHBA-N synthase containing BACs were distinguished by PCR analysis, which established that about half of these BACs belonged to each of the AHBA-B and AHBA-N types. Interestingly, when these were analyzed for the presence of PKS genes by PCR analysis, we found that none of the 20 AHBA-B synthase containing BACs contained PKS genes, whereas 14 out of 17 AHBA-N synthase containing BACs also had PKS genes. We thus focused on the CT containing BACS to identify the geldanamycin PKS genes. Nine out of 15 CT BACs also contained PKS genes and from these CT+PKS BACs, we subcloned four unique KS amplimers using the degenerate PCR primers. These putative geldanamycin KS sequences (one of which was contained in the 63 KS amplimers described above) were then used as probes in a secondary screen of the BAC library at high stringency to identify seven additional overlapping BACs.

Two overlapping clones, pKOS256-107–3 and pKOS256-154-1, spanning a ∼85-kb region, were fully sequenced by the shotgun method, and the resident genes were assigned to geldanamycin production on the basis of database comparisons (Fig. 2). Twenty-three open reading frames (ORF) and the deduced functions of their products are listed in Table 1. The CT gene probe matched the sequence of gdmN, and only one homolog of the AHBA synthase genes was found (gdmO). An additional 15 ORFs downstream of ORF16 were sequenced and analyzed (not shown) but, on the basis of their deduced functions, are not believed to be involved in geldanamycin production. (There are incomplete reports about cloning the geldanamycin genes from other streptomycetes isolated independently from the 3602 strain [29,30].)

Table 1.  Gene designations and functions
  1. Abbreviations for PKS domain functions: ACP, acyl carrier protein; AL, acyl-CoA ligase; AT, acyltransferase; CT, carbamoyltransferase; DH, dehydratase; ER, enoyl reductase; KR, keto reductase; KS, keto synthase; X, unknown.

Predicted polypeptideAmino acidsHomologous gene(s)Putative function
ORF16227RhtB family transportersEfflux
ORF17300SC3C8.01Secreted protein
ORF19294AraC familyTranscriptional regulation
ORF20253TetR familyTranscriptional regulation
GdmX156JadX; MmyYUnknown
GdmA16842 PKS (modules 0–3)
Loading  AL X ACP
Module 1  KS AT DH ER KR ACP
Module 2  KS AT DH ER KR ACP
Module 3  KS AT KR ACP
GdmA23435 PKS (modules 4–5)
Module 4  KS AT DH KR ACP
Module 5  KS AT KR ACP
GdmA33896 PKS (modules 6–7)
Module 6  KS AT DH ER KR ACP
Module 7  KS AT DH KR ACP
GdmF257RifFAmide synthase
GdmH370FkbHMethoxymalonyl-ACP biosynthesis
GdmI370FkbIMethoxymalonyl-ACP biosynthesis
GdmJ91FkbJMethoxymalonyl-ACP biosynthesis
GdmK288FkbKMethoxymalonyl-ACP biosynthesis
GdmG218FkbGMethoxymalonyl-ACP biosynthesis
GdmR2926LuxR familyTranscriptional regulation
GdmO354Asm47AminoDHQ synthase
GdmR1962LuxR familyTranscriptional regulation

3.3Key features of the geldanamycin biosynthesis genes and their deduced products

3.3.1The PKS

A seven-module PKS encoded by gdmA1-A3 and immediately followed by the gdmF gene encoding an amide synthase, as found for the rifamycin and ansamitocin PKSs [23,24], is entirely consistent with the biosynthetic hypothesis shown in Fig. 1. The N-terminal region of module 1 of the GdmA1 protein contains two domains strongly resembling their counterparts in the rifamycin and ansamitocin PKSs. Both of the latter enzymes use AHBA as the starter unit, and the RifA loading module recently has been shown to load the AHBA onto the acyl carrier protein (ACP0) domain through formation of an intermediate acyl-adenylate catalyzed by the acyl ligase domain[31]. The starter unit is transferred intramolecularly from ACP0 to the KS1 of the following module for reaction with the 2-methylmalonate chain extender unit bound to ACP1[32]. Modules 2 and 5 of the GdmA1 and GdmA2 proteins, respectively, have acyltransferase (AT) domains with the specific motifs similar to other ATs that utilize 2-methoxymalonate (unpublished data). The substrate for AT2 and AT5 is presumed to be provided by the products of the gdmG-gdmK genes, given the recent genetic and functional characterization of their homologs from the FK520[33] and ansamitocin [34,35] gene clusters. In contrast, the AT domain of module 6 of GdmA3 clearly falls within the class that uses malonyl-CoA, and those of GdmA1 module 1 and module 3, GdmA2 module 4 and GdmA3 module 7 strongly resemble those that use 2-methylmalonyl-CoA [36,37]. Finally, the presence of enoyl reductase (ER) domains in modules 1, 2 and 3, dehydratase (DH) domains in modules 1, 2, 4, 6 and 7, and keto reductase (KR) domains in all the modules fit the requirements for the predicted functionality of progeldanamycin (Fig. 1).

The GdmA PKS has one notable feature. The ER domain in module 6 is apparently functional yet geldanamycin has a C4 cis-double bond. Consequently, it is plausible that the PKS reduces this double bond and that the C4 cis-double bond is introduced oxidatively either before or after formation of progeldanamycin (Fig. 1A).

3.3.2Tailoring genes

The predicted functions for most of these genes are consistent with the requirements for conversion of progeldanamycin (or its 4,5-dihydro form) into geldanamycin (Fig. 1). Disruption of the gdmH gene, carried out by insertion of a kanamycin resistance gene (Section 2), resulted in strain K279-37 that produced geldanamycin together with a compound whose high-resolution mass spectral data are consistent with 4,5-dihydro-7-descarbamoyl-7-hydroxygeldanamycin. This result confirms the function assigned to gdmN (the partial inactivation of this gene is believed to be the consequence of read-through expression) and shows that gdmH is dispensible or that its mutation is compensated in trans by a paralog. Distinction among the three genes, gdmL, gdmM and gdmP, likely to govern C17 hydroxylation, C21 oxidation and C4,5 unsaturation, if that actually occurs, will have to await studies of mutants or the expressed and purified enzymes. The putative C17 O-methyltransferase gene presumably lies outside the currently sequenced region.

3.3.3Other genes

Possible regulatory genes (gdmR1, gdmR2, ORF19 and ORF20) and a gene for the amino dehydroquinate synthase needed for AHBA biosynthesis (gdmO) are listed in Table 1 as part of the gdm gene cluster because their homologs have been found in other clusters of bacterial secondary metabolism genes. ORF16 and ORF17 are candidates for putative geldanamycin resistance and/or export genes. All but one of the remaining genes for AHBA biosynthesis in the geldanamycin pathway, as deduced from the sequences of homologs involved in the biosynthesis of other benzoquinone ansamycins [22–24], are clustered elsewhere in the genome, more than 20 kb from the end of the BAC that contains ORF22 (data not shown). This is consistent with the observation that three of the AHBA biosynthesis genes for ansamitocin production and the remaining asm genes are in a subcluster separated from all the other genes for AHBA biosynthesis by 30 kb[23].


We thank David Hopwood, Robert McDaniel, Chris Reeves and Peter Revill for helpful comments during preparation of the manuscript. This research was supported in part by a Small Business Innovative Research grant from the National Institute of Health (R43 CA96262 and AL38947).