A 5.2 kb region from the oleandomycin gene cluster in Streptomyces antibioticus located between the oleandomycin polyketide synthase gene and sugar biosynthetic genes was cloned. Sequence analysis revealed the presence of three open reading frames (designated oleI, oleN2 and oleR). The oleI gene product resembled glycosyltransferases involved in macrolide inactivation including the oleD product, a previously described glycosyltransferase from S. antibioticus. The oleN2 gene product showed similarities with different aminotransferases involved in the biosynthesis of 6-deoxyhexoses. The oleR gene product was similar to several glucosidases from different origins. The oleI, oleR and oleD genes were expressed in Streptomyces lividans. OleI and OleD intracellular proteins were partially purified by affinity chromatography in an UDP-glucuronic acid agarose column and OleR was detected as a major band from the culture supernatant. OleI and OleD showed oleandomycin glycosylating activity but they differ in the pattern of substrate specificity: OleI being much more specific for oleandomycin. OleR showed glycosidase activity converting glycosylated oleandomycin into active oleandomycin. A model is proposed integrating these and previously reported results for intracellular inactivation, secretion and extracellular reactivation of oleandomycin.
Macrolide antibiotics constitute a chemical class of antibiotics that have clinical applications mainly against bacterial infections caused by Gram-positive bacteria (Nakayama, 1984). Most macrolides are produced by streptomycete species, and the biosynthesis of a potentially lethal antibiotic by these microorganisms requires the existence of a self-resistance mechanism in the producing organism to avoid suicide. To date, three different mechanisms of self-resistance in macrolide-producing organisms have been described. The first resistance mechanism involves modification of the antibiotic target site (the ribosome) by mono- or dimethylation of a single adenine residue in the 23S RNA and has been reported in several macrolide producers. This mechanism has been found to confer self-resistance to the producers of erythromycin (Skinner and Cundliffe, 1982), tylosin (Zalacain and Cundliffe, 1989; 1991) and carbomycin (Zalacain and Cundliffe, 1990). In contrast, in the oleandomycin producer, ribosomes have been found to be sensitive to oleandomycin throughout the cell cycle, even during the production phase (Fierro et al., 1987). The second mechanism involves cell permeability. Several genes encoding ABC (ATP-binding cassette) transporters and conferring resistance to different macrolides have been reported from several macrolide producers (Rosteck et al., 1991; Schöner et al., 1992; Rodriguez et al., 1993; Olano et al., 1996). These ABC transporters participate in the export of macrolides through ATP-dependent processes. The existence of antibiotic-modifying enzymes as a third resistance mechanism has also been described. The oleandomycin producer possesses an intracellular glycosyltransferase activity that inactivates oleandomycin by glycosylation of a hydroxyl group present in one of the sugars (desosamine) attached to the macrolactone ring (Vilches et al., 1992; Quirós and Salas, 1995) (Fig. 1). This producer strain also contains a second activity that converts inactive glycosylated oleandomycin into active antibiotic (Vilches et al., 1992), the responsible enzyme being purified from the culture supernatant of this organism (Quirós et al., 1993). Inactivation of macrolides by glycosylation has also been reported in the non-macrolide producer Streptomyces lividans (Jenkins and Cundliffe, 1991), and recently such activity has been described to be present in 15 out of 32 streptomycete strains tested as producers of different polyketides (Sasaki et al., 1996).
Here we report the cloning, sequencing and expression of two genes from the oleandomycin producer Streptomyces antibioticus ATCC 11891, which code for a glycosyltransferase and a glycosidase, and biochemical evidence supporting their role in oleandomycin inactivation and reactivation.
Sequencing of the oleI, oleN2 and oleR genes
From a cosmid library of chromosomal DNA from the oleandomycin producer S. antibioticus (Rodriguez et al., 1993), several overlapping clones were isolated after screening by in situ colony hybridization using as DNA probes the Streptomyces griseus strD, strE and strM genes from the streptomycin biosynthetic pathway. These genes encode three enzymes catalysing early steps in the biosynthesis of 6-deoxyhexoses (Distler et al., 1987; Pissowotzki et al., 1991) and are conserved in many different streptomycetes producing 6-deoxyhexoses containing secondary metabolites (Stockmann and Piepersberg, 1992). One of these cosmids, cosAB63 (Fig. 2), also showed hybridization signals against a DNA fragment derived from the orf3 gene of the oleandomycin polyketide synthase (Swan et al., 1993). Sequencing of the region homologous to the strDEM genes showed the presence of several open reading frames (ORFs) encoding different enzymes involved in the biosynthesis of the 6-deoxyhexoses present in oleandomycin, L-oleandrose and D-desosamine (C. Olano, I. Aguirrezabalaga, C. Méndez and J. A. Salas, unpublished). A region comprising the strDEM and the polyketide synthase homologous regions was sequenced (5250 nucleotides; GenBank accession number AF055579) and the sequence was analysed for coding regions (Devereux et al., 1984; Wright and Bibb, 1992). This analysis predicted the existence of three likely Streptomyces ORFs: two of them being transcribed in the same direction and the third one convergent to these two (Fig. 2). The first ORF (from left to right), designated oleI, starts in an ATG codon and comprises 1274 nucleotides ending in a TGA codon; it would code for a polypeptide of 424 amino acids with a Mr of 42 380. oleI is preceded by a typical ribosomal-binding site (RBS) (GGAG) (Bibb and Cohen, 1982). Downstream (135 nucleotides) of oleI, there is a second ORF (designated oleN2) convergently transcribed to oleI. It starts in an ATG codon that is preceded by a putative RBS (GGAGG). It comprises 1050 nucleotides and ends in a TGA codon; it codes for a polypeptide of 369 amino acids with a Mr of 39 580 Da. The third ORF, designated oleR, is separated from the oleN2 gene by 20 nucleotides. It starts in an ATG codon it comprises 2304 nucleotides ending in a TGA codon and codes for a polypeptide of 768 amino acids with a Mr of 80 681. No potential RBS sequence is recognized in the vicinity of the starting codon. The three genes showed the characteristic high GC content and codon bias of Streptomyces genes.
Identification of the functions of the different genes
Potential functions of the gene products were assigned after comparison with proteins in databases. The OleI protein showed similarities with glycosyltransferases and glucuronosyltransferases. The highest score was with the products of the mgt gene from S. lividans, 41.6% identical amino acids (Jenkins and Cundliffe, 1991), and of the oleD gene from S. antibioticus, 41.9% (Hernández et al., 1993), which encode glycosyltransferases responsible for inactivation of different macrolides. Close to the C-terminus of the OleI protein sequence there is an amino acid sequence that is well conserved in different glycosyltransferases and is considered a glycosyltransferase motif (Fig. 3). The OleN2 protein showed similarity with several aminotransferases involved in the biosynthesis of 6-deoxysugars in different antibiotic biosynthetic pathways: EryCI and EryCIV from the erythromycin pathway in Saccharopolyspora erythraea, 63.2% and 55.3% respectively (Dhillon et al., 1989; Gaisser et al., 1997; Summers et al., 1997), TylB from the tylosin pathway in S. fradiae, 61.8% (Merson-Davies and Cundliffe, 1994), DnrJ from the daunorubicin in S. peucetius, 44.3% (Stutzman-Engwall et al., 1992) and LmbS from the lincomycin pathway in S. lincolnensis, 39.5% (Peschke et al., 1995). All these proteins are similar to the mechanistically well-characterized AscC dehydratase from the ascarylose pathway in Yersinia pseudotuberculosis (Thorson et al., 1994). The above-mentioned proteins, including OleN2, are possibly involved in sugar biosynthesis catalysing the introduction of an amino group into C-3 of dTDP-activated deoxyhexose intermediates in the biosynthesis of the different drugs. At least two enzyme activities would be required for this amination. These activities would be codified in the case of erythromycin by the EryCI and EryCIV proteins (Gaisser et al., 1997; Summers et al., 1997) and in the case of oleandomycin by the homologous oleN1 (C. Olano, unpublished) and oleN2 (this paper).
Expression of the OleI, OleD and OleR proteins in S. lividans
To confirm that the oleI and oleR genes code for proteins involved in glycosylation and deglycosylation of oleandomycin, both genes were expressed in S. lividans under the control of the erythromycin resistance promoter from Sacc. erythraea (Bibb et al., 1985). The oleD gene from S. antibioticus, which encodes a previously identified oleandomycin glycosyltransferase (Hernández et al., 1993), was also expressed after subcloning in the same expression vector. Cell-free extracts of S. lividans expressing OleI (strain LQI) and OleD (strain LQD) showed oleandomycin glycosylating activity as determined by the transfer of tritiated glucose from UDP-[3H]-glucose. However, the LQI extracts exhibited were much more active than LQD extracts (Fig. 4A). Oleandomycin glycosylating activity was also detected in cell-free extracts of the oleandomycin producer, S. antibioticus (used as a positive control), but extracts from clones LQI and LQD showed higher levels of activity (Fig. 4A). Cell-free extracts of S. lividans EM4 (only containing the vector pEM4 without insertion) did not show any activity. The possible overexpression of the OleI and OleD proteins was analysed by SDS–PAGE. No apparent differences among the tracks corresponding to S. lividans EM4, LQI and LQD were observed (data not shown). To verify that expression of both glycosyltransferases was present, a partial purification of the putative expressed proteins was attempted. Cell-free extracts were applied to UDP-glucuronic acid agarose affinity columns and eluted using high salt concentration. This affinity step has been shown to facilitate purification of a glycosyltransferase from S. antibioticus (Quirós and Salas, 1995). Fractions eluted at high salt concentration using extracts from the control (strain EM4) showed negligible activity levels and no apparent major band was observed in SDS–PAGE (Fig. 5A, lane 3). In contrast, elution from extracts of LQI and LQD strains showed high levels of activity (data not shown) and a major protein band was found in SDS–PAGE (Fig. 5A, lanes 5 and 7). Estimation of the molecular mass of these two proteins gave values close to those deduced from gene sequencing.
To evaluate possible differences between both glycosyltransferase activities, enzymatic assays were performed using different macrolide antibiotics as substrates (Fig. 6). Interestingly, OleI and OleD clearly differed in their pattern of substrate specificity. OleI showed a high activity against oleandomycin but its activity against other macrolides (carbomycin, tylosin and erythromycin) was very low (Fig. 6A). In contrast, OleD was able to glycosylate all macrolides tested, being more active against carbomycin and oleandomycin (Fig. 6B).
Glycosidase activity in strain LQR was detected in the culture supernatant and no activity was observed in control strain EM4 (Fig. 4B). A protein of a molecular mass similar to that deduced from the gene sequence was detected by SDS–PAGE analysis from the culture supernatant of strain LQR (Fig. 5B, lane 3).
Effect of glycosylated oleandomycin on in vitro protein synthesis
To test the potential inhibitory activity of glycosylated oleandomycin on protein synthesis, the effect of different concentrations of glycosylated oleandomycin in a DNA-directed in vitro coupled transcription/translation system was tested. Oleandomycin caused between 30% and 60% inhibition of in vitro protein synthesis at concentrations between 0.5 and 1 μM, whereas 100 μM glycosylated oleandomycin only produced 17% decreased activity.
Several macrolide producers posses specific methylases capable of modifying the ribosome and render it resistant to the drug (Skinner and Cundliffe, 1982; Zalacain and Cundliffe, 1989; 1990; 1991). This is a very efficient resistance mechanism that enables the producer strain to survive the biosynthesis of a potentially lethal drug. In contrast, the oleandomycin producer organism has been shown to contain ribosomes (the antibiotic target site) susceptible to oleandomycin all along the cell cycle, even during antibiotic biosynthesis (Fierro et al., 1987). However, cell-free extracts of this producer strain contain glycosyltransferase activity that inactivates oleandomycin by glycosylation (Vilches et al., 1992), and a glycosyltransferase was purified and enzymatically characterized (Quirós and Salas, 1995). Furthermore, a gene (oleD) encoding a glycosyltransferase from S. antibioticus showing high similarity (both at the amino acid and the DNA levels) to the macrolide glycosyltransferase gene (mgt ) of S. lividans was also cloned and sequenced (Hernández et al., 1993) and its product was presumably considered responsible for oleandomycin glycosylation. It is commonly observed that antibiotic resistance and biosynthetic genes are closely linked in the chromosome, but the oleD gene was found in a chromosomal region where no oleandomycin biosynthetic genes were present. Now we have found a gene (oleI ) encoding another glycosyltransferase but within the oleandomycin gene cluster. Activity assays using OleI and OleD proteins against different macrolides showed a very different pattern of substrate specificity. OleI was nearly specific for oleandomycin showing a very high level of activity against this macrolide and negligible against the other macrolides. These data, together with the close linkage between the oleI gene and other oleandomycin biosynthetic genes, suggest that the oleI product is the enzyme responsible for oleandomycin glycosylation during its biosynthesis. On the other hand, OleD was not specific for oleandomycin and glycosylated all other macrolides assayed. This fact, together with the identification of the oleD gene in a chromosomal region far apart from the oleandomycin gene cluster (Hernández et al., 1993), makes it more plausible to attribute to the oleD gene a more general role not directly involved in the biosynthesis of oleandomycin. This hypothesis is also supported by the fact that macrolide glycosylating activities similar to that of OleD have been found in at least 15 different streptomycete species (Sasaki et al., 1996). Substrate specificity profile and molecular weight concordance also suggest that OleI probably corresponds to the glycosyltransferase previously purified and characterized (Quirós and Salas, 1995).
In the vicinity of the oleI gene, another gene (oleR) was found. Database comparison of the OleR protein suggested that this protein was a glycosidase. Expression of this gene in S. lividans confirmed that this protein present in the culture supernatant of LQR strain was a glycosidase able to convert glycosylated oleandomycin into oleandomycin and possibly the OleR protein corresponds to the glycosidase previously purified and characterized from the culture supernatant of S. antibioticus (Quirós et al., 1993). Interestingly, OleR showed great similarity to EryBI protein from Sacc. erythraea, an erythromycin producer. The eryBI locus was initially proposed to be involved in the biosynthesis of mycarose in the erythromycin pathway (Weber et al., 1990), but recent experimental evidence suggests that the eryBI gene is not involved in erythromycin biosynthesis because its insertional inactivation produces a wild-type phenotype (Gaisser et al., 1998). The great similarity between EryBI and OleR proteins (56.7% identical amino acids) suggests that EryBI also encodes a glycosidase. By assuming a parallelism in the biosynthetic pathways of oleandomycin and erythromycin, one might assume that EryBI could have played a role in reactivation of glycosylated erythromycin during its biosynthesis. However, although most (or probably all) of the erythromycin gene cluster has been fully sequenced, no erythromycin glycosylating gene has been reported to date. Possibly the eryBI gene in Sacc. erythraea did play a role in the past but lost its function through the evolution at the time when the producer organism acquired the more precise survival mechanism, i.e. target site modification by a constitutive ribosomal methylase.
Biosynthesis of an antibiotic by a producer organism as an inactive compound that is reactivated either during its export or extracellularly obviously confers an evolutionary advantage to the producer strain to survive during antibiotic biosynthesis. In this sense, it has been proposed that biosynthesis of streptomycin proceeds through inactive phosphorylated intermediates and conversion into active streptomycin by a phosphatase as the final step in biosynthesis (Miller and Walker, 1970; Walker and Walker, 1971). S. antibioticus has not evolved to posses oleandomycin-resistant ribosomes but rather has developed a combination of modifying enzymes (intracellular glycosyltransferase and extracellular glycosidase) and efficient secretion system. Two oleandomycin resistance genes from S. antibioticus have been characterized, both encoding ABC transporters (Rodriguez et al., 1993; Olano et al., 1996). One of them, oleB, can interact and undergoes conformational changes upon interaction with glycosylated oleandomycin (Buche et al., 1997) and has been reported to secrete glycosylated oleandomycin through the cell membrane (Olano et al., 1996). We propose a model (Fig. 7) for intracellular inactivation, secretion and extracellular reactivation of oleandomycin by S. antibioticus. The OleI glycosyltransferase would prevent the appearance of free intracellular oleandomycin either by acting on an intracellular oleandomycin intermediate or by direct glycosylation of the antibiotic. In the former, the glycosylation would take place on an intermediate close to the final product as indicated by the high specificity of the enzyme. In the latter, oleandomycin biosynthetic enzymes could be organized as a multienzymatic complex that would channel the intermediates through the different enzymatic activities. In this way, oleandomycin would be finally channelled to the OleI glycosyltransferase that would cause its inactivation. In any case, the final intracellular product of the pathway would be glycosylated oleandomycin, which is inactive as an antibiotic. The OleB (and also perhaps OleC) transporters would be responsible for secretion through the membrane of this inactive molecule that would be extracellularly reactivated by the product of the OleR gene. The way back of the antibiotic into the cells would be prevented by the OleB transporter acting as a one way in–out pump.
Bacterial strains, culture conditions and vectors
S. antibioticus ATCC 11891, an oleandomycin producer, was used as the donor of chromosomal DNA. For sporulation it was grown for 7 days at 30°C on GAE plates (Hardisson et al., 1978). Growth in liquid medium was carried out at 30°C in TSB medium (trypticase soya broth, Oxoid). S. lividans TK21 was used as the Streptomyces host for gene expression. Escherichia coli XL1-Blue (Bullock et al., 1987) was used as the host for subcloning. E. coli BL21 (DE3) was used for the in vitro protein synthesis assays (Studier and Moffat, 1986). pUC18, pSL1180 (Pharmacia) and pIJ2925 (Janssen and Bibb, 1993) were used as vectors for subcloning in E. coli, and M13mp18 and M13mp19 phage vectors were used for DNA sequencing. pWHM4 was used as subcloning vector in Streptomyces (Vara et al., 1989). pUR12 (Tang et al., 1996) was obtained from Dick Hutchinson.
DNA manipulation and sequencing
DNA manipulations were carried out according to standard procedures for E. coli (Sambrook et al., 1989) and for Streptomyces (Hopwood et al., 1985). Southern hybridization was according to standard procedures (Hopwood et al., 1985). DNA sequencing was carried out by use of the dideoxynucleotide chain termination method (Sanger et al., 1977) using an ALF express automatic DNA sequencer (Pharmacia). To overcome band compression artefacts, 7-deaza-dGTP was routinely used instead of dGTP (Mizusawa et al., 1986). Both DNA strands were sequenced with universal primers or with internal oligoprimers (17 mer). Computer-aided database searching and sequence analyses were carried out using the University of Wisconsin Genetics Computer Group programs package (UWGCG; Devereux et al., 1984) and the BLASTP program (Altschul et al., 1990).
Expression in S. lividans
For expression in S. lividans TK21, the different genes were cloned under the control of the promoter of the erythromycin resistance (ermE*p) gene from Sacc. erythraea (Bibb et al., 1985). The ermE*p was subcloned as a 0.3 kb PstI fragment from pUR12 into the PstI site of pWHM4 generating pEM4.
For the expression of the oleI gene, a 2.2 kb SacI–HindIII fragment from pUOS004 (a pUC vector containing a 2.6 kb BamHI fragment from cosAB63 and containing the entire oleI gene and 618 bp of the 3′ end of oleN2 gene; C. Olano, unpublished) was subcloned into the same restriction sites of pSL1180 generating pLQ8-1. Then the oleI gene (and the 3′ end of the oleN2 gene) was rescued as a 2.2 kb XbaI–EcoRI fragment using these sites from the pSL1180 polylinker and subcloned into the same sites of expression vector pEM4. In the final construction, pLQI, the oleI gene is located immediately downstream of ermE*p.
For the expression of the oleR gene, a 6 kb SacI–SphI fragment (Fig. 2) from cosAB63 containing the oleR gene and the 5′ end of orf1 of the oleandomycin polyketide synthase was subcloned into the same sites of pIJ2925 originating pLQ9-1. Then, the fragment was rescued as a 6 kb BglII fragment (these sites were from the pIJ2925 polylinker) and subcloned into the BamHI site of pUC18 generating pLQ9-2. Finally, a 6 kb XbaI–EcoRI fragment was subcloned downstream of the ermE*p into the same sites of the expression vector pEM4, generating pLQR.
For expression of the oleD gene, a 2.3 kb PvuII fragment from cos25G8 containing the oleD gene (Hernández et al., 1993) was subcloned into the SmaI site of pUC18 originating pLQD1. From this construction, a 2.3 kb XbaI–EcoRI fragment was subcloned into the same sites of the expression vector pEM4, generating pLQD.
Preparation of cell-free extracts
S. lividans clones containing the different plasmid constructions were grown on TSB medium at 30°C for 72 h in the presence of 5 μg ml−1 thiostrepton. The mycelia were collected by centrifugation and washed in 50 mM Tris-HCl (pH 8.0) containing 1 mM EDTA and 1 mM dithiothreitol (DTT). Mycelia were broken by ultrasound (10 pulses 10 s each with intermittent cooling on ice water) in a MSE ultrasonic disintegrator at 150 W and 20 kHz. Cell debris was removed by centrifugation at 30 000 × g for 15 min and the supernatant was used for assay of the glycosyltransferase activity. Extracellular proteins were obtained from similar cultures. The mycelia were removed by centrifugation and 25 mM DTT was added to the culture supernatant. Then proteins in the supernatant were precipitated using ammonium sulphate (95% saturation) and the precipitates were collected by centrifugation. After extensive dialysis against the buffer mentioned above, the sample was used for the glycosidase assay.
Macrolide glycosyltransferase activity was assayed by measuring the transfer of [3H]-glucose from UDP-D-[6-3H]-glucose (sp. act. 529 GBq mmol−1) into different macrolides (Quirós and Salas, 1995). Conversion of glycosylated oleandomycin into oleandomycin was measured by determining the release of [3H]-glucose from [3H]-glycosylated oleandomycin (Quirós et al., 1993). Tritiated glycosylated oleandomycin used as substrate was prepared as described elsewhere (Quirós et al., 1993).
Partial purification of glycosyltranferases was achieved by affinity chromatography on UDP-glucuronic acid agarose columns (Sigma Chemicals). Cell-free extracts were applied to the column at a flow rate of 0.1 ml min−1. The column was then washed with three column volumes of the buffer mentioned above and then eluted using the same buffer including 1 M NaCl. Samples were collected, analysed by 12% PAGE containing 0.1% SDS and used for enzymatic assays of oleandomycin glycosylation.
In vitro coupled trancription–translation
The effect of oleandomycin and glycosylated oleandomycin on in vitro protein synthesis was determined by using an in vitro coupled transcription–translation system (Köhrer et al., 1996). In this system, the incorporation of L-[35S]-methionine into trichloroacetic acid-insoluble material as a consequence of the expression of genes harboured in plasmid pUC18 was measured in the presence and the absence of different concentrations of both drugs.
Protein determinations and polyacrylamide gel electrophoresis
Protein concentration in the different samples was determined by the protein dye-binding assay (Bradford, 1976). Expression of the different genes in S. lividans TK21 was followed by PAGE in the presence of sodium dodecyl sulphate as described (Laemmli, 1971). Staining was carried out with Coomassie blue.
This work was supported by grants of the European Union (BIO4-CT96-0080) and from the Spanish Ministry of Education (PB94-1319). We wish to thank C. Köhrer for collaboration on the in vitro protein synthesis assays. We also thank Wolfgang Piepersberg for the str genes, Dick Hutchinson for plasmids pWHM4 and pUR12 and Mervin Bibb for pIJ2925. We also thank Mr S. J. Lucania (Squibb) for samples of thiostrepton.