Inactivation of chloramphenicol by O-phosphorylation. A novel resistance mechanism in Streptomyces venezuelae ISP5230, a chloramphenicol producer.

Plasmid pJV4, containing a 2.4-kilobase pair insert of genomic DNA from the chloramphenicol (Cm) producer Streptomyces venezuelae ISP5230, confers resistance when introduced by transformation into the Cm-sensitive host Streptomyces lividans M252 (Mosher, R. H. Ranade, N. P., Schrempf, H., and Vining, L. C.(1990) J. Gen. Microbiol. 136, 293-301). Transformants rapidly metabolized Cm to one major product, which was isolated and purified by reversed phase chromatography. The metabolite was identified by nuclear magnetic resonance spectroscopy and mass spectrometry as 3′-O-phospho-Cm, and was shown to have negligible inhibitory activity against Cm-sensitive Micrococcus luteus. The nucleotide sequence of the S. venezuelae DNA insert in pJV4 contains an open reading frame (ORF) that encodes a polypeptide (19 kDa) with a consensus motif at its NH terminus corresponding to a nucleotide-binding amino acid sequence (motif A or P-loop; Walker, J. E., Saraste, M., Runswick, M. J., and Gay, N. J.(1982) EMBO J. 1, 945-951). When a recombinant vector containing this ORF as a 1.6-kilobase pair SmaI-SmaI fragment was used to transform S. lividans M252, uniformly Cm-resistant transformants were obtained. A strain of S. lividans transformed by a vector in which the ORF had been disrupted by an internal deletion yielded clones that were unable to phosphorylate Cm, and exhibited normal susceptibility to the antibiotic. The results implicate the product of the ORF from S. venezuelae as an enzymic effector of Cm resistance in the producing organism by 3′-O-phosphorylation. We suggest the trivial name chloramphenicol 3′-O-phosphotransferase for the enzyme.

that inhibits protein biosynthesis by binding reversibly to the peptidyl transferase center of 50 S ribosomal subunits (Pongs, 1979), is produced by Streptomyces venezuelae and some related species (Vining and Westlake, 1984). During growth under conditions where Cm is not produced, S. venezuelae is relatively sensitive to the antibiotic, but high-level resistance is induced by exposure to Cm; in cultures grown under Cmproducing conditions, resistance increases concurrently with Cm synthesis. However, ribosomes extracted from producing or nonproducing mycelium are equally sensitive to the antibiotic Vining, 1970, 1972). Resistance to Cm is mediated in most eubacteria by chloramphenicol acetyltransferase (CAT; EC 2.3.1.28: reviewed by Shaw (1983) and Shaw and Leslie (1991)); this enzyme modifies Cm by acetylation, yielding 3Ј-Oacetyl-Cm, which is only very weakly bound by ribosomes and thus is not an antibiotic. Since S. venezuelae lacks CAT activity (Shaw and Hopwood, 1976;Nakano et al., 1977), alternative mechanisms have been sought to explain the ability of the producing organism to avoid inhibition by one of its own products (Vining and Westlake, 1984).
By cloning genomic S. venezuelae DNA in the streptomycete vector pIJ702 and transforming the Cm-hypersensitive Streptomyces lividans M252 to resistance, Mosher and co-workers (1990) obtained a transformant (RM3) harboring a recombinant plasmid (pJV3) with a 6.5-kbp insert. Deletion from pJV3 of a 5.2-kbp segment (encompassing 4.1-kbp from the insert and 1.1-kbp from the pIJ702 vector) yielded pJV4, from which a region implicated in Cm resistance was subcloned as a 2.4kbp KpnI-SstI fragment. Cultures of S. lividans RM3 rapidly metabolized [U-14 C]chloramphenicol to unidentified labeled products (Mosher et al., 1990).
We report here the isolation and characterization of the major product of Cm metabolism by S. lividans RM3 and RM4 (M252 transformed with pJV4). Identification of this metabolite as the 3Ј-phospho ester of Cm implicates a mechanism of Cm resistance for the producing organism that has not hitherto been encountered in streptomycetes or other microbial systems.

EXPERIMENTAL PROCEDURES
Materials-Restriction endonucleases and DNA modifying enzymes were obtained from Life Technologies, Inc., Promega, New England Biolabs, and Pharmacia Biotech Inc. Cm and D-threo-p-nitrophenylserinol (Cm-base) were from Sigma.
Bacteria, Plasmids, and Culture Conditions-The properties of S. lividans M252, and those of transformants RM3 (containing pJV3) and RM4 (containing pJV4) have been described (Mosher et al., 1990). Escherichia coli was grown in 2 ϫ YT broth (10 g of yeast extract, 16 g * This work was supported by grants from the Natural Sciences and Engineering Research Council of Canada (to L. C. V.) and the Science and Engineering Research Council of the United Kingdom through its Biotransformations Programme (to W. V. S.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence ( of Bacto-tryptone and 5 g of NaCl per liter). Conditions for E. coli growth, transformation, transfection, and plasmid isolation were essentially as described by Sambrook et al. (1989). Batch cultures of S. lividans were grown in GNY medium (20 ml of glycerol, 8 g of nutrient broth, 3 g of yeast extract, and 5 g of K 2 HPO 4 per liter) for 48 h at 30°C (Malik and Vining, 1970). Conditions for transforming Streptomyces and for plasmid isolation were as described (Hopwood et al., 1985). Table I lists the strains and plasmids used in this work.
High Performance Liquid Chromatography (HPLC)-Culture broths and extracts were analyzed for Cm and its derivatives by HPLC on a 4.6 ϫ 50-mm Phenomenex Ultracarb ODS-30 C 18 column (5-m particle size). The column was initially equilibrated with 30 mM potassium phosphate buffer, pH 3.3. At the sample injection, a programmed methanol gradient was started: the methanol concentration increased linearly to 25, 50, and 100% at 1, 6, and 7 min, respectively, remained at 100% for 3 min, and then reduced to zero over 1 min. The column was re-equilibrated for 5 min before the next injection. Nitrophenyl derivatives were detected by their absorbance at 273 nm.
Isolation of 3Ј-Phospho-Cm from Suspensions of RM4 Mycelium-To investigate Cm metabolism by S. lividans RM3 and RM4, mycelium was grown in GNY medium supplemented with Cm at a final concentration of 12.5 g/ml (38.7 M). Cultures were routinely inoculated with spore suspensions and incubated at 30°C on a rotary shaker (3.7-cm eccentricity; 250 rpm). Mycelial suspensions for converting Cm to 3Јphospho-Cm were prepared as follows; 250 ml of RM4 cells grown for 48 h were centrifuged (35,000 ϫ g for 10 min at 4°C), and washed aseptically with water. They were resuspended in 250 ml of water containing Cm (200 g/ml), and incubated in a 1-liter Erlenmeyer flask for 23 h under the conditions used for growing cultures. The cells were then removed by centrifugation, and the supernatant was decanted and stored at Ϫ20°C. The supernatant was concentrated to 1 ml at 40°C in vacuo, and acidified with 100 l of 1 M ammonium formate buffer, pH 2.5. The resulting solution, pH 3.4, was clarified by filtration and applied to a C 18 reversed-phase silica column (1 ϫ 23 cm, 40-m particle size, Bakerbond) packed in water. After 1 ml of 0.1 M ammonium formate buffer, pH 2.5, had been added, the products were eluted with water. Fractions were monitored by HPLC; those that contained Cm or 3Ј-phospho-Cm were pooled appropriately and concentrated in vacuo. By this procedure 3Ј-phospho-Cm greater than 99% pure by HPLC and NMR analysis, and free from contamination with Cm, was isolated in milligram quantities.
The 31 P NMR spectrum of 3Ј-phospho-Cm was recorded at 101.2 MHz in D 2 O with 85% phosphoric acid as an external reference. Data were accumulated using 90°pulses (6 s) with delays of 1 or 2 s. The data size was 32 K.
Mass Spectrometry-Mass spectra were obtained on an API/III triple quadrupole mass spectrometer (Perkin-Elmer SCIEX Instruments, Thornhill, Ontario) using nebulizer-assisted electrospray ionization (ion spray). The mass spectrum of 3Ј-phospho-Cm was acquired in the negative ion mode; 2 l of a 7.5 g/l solution was injected into a stream of 50% acetonitrile (flow rate 10 l/min). The mass spectrometer scanned over a mass range of 100 -500 in steps of 0.1 Da, with a dwell time of 2 ms/step. The spectrometer was calibrated with a cesium nitrate standard; mass assignments were judged to be reliable to Ϯ0.5 Da.
Antibiotic Assay-Assay disks (1.3-cm diameter) impregnated with aqueous solutions of Cm or 3Ј-phospho-Cm at known concentrations were placed on Difco nutrient agar seeded (2% v/v) with a 48-h nutrient broth culture of Micrococcus luteus. The relative antibiotic activities of the test samples were determined by measuring the diameters of inhibition zones around the disks after incubation overnight at 30°C.
Construction and Use of Shuttle Vector pSV1.6 -Standard procedures were used for manipulating DNA from Streptomyces spp. (Hopwood et al., 1985) and E. coli (Sambrook et al., 1989). DNA fragments from restriction digests separated by agarose gel electrophoresis were purified with a Geneclean kit (BIO-101); E. coli plasmids were purified with a Wizard miniprep kit (Promega). A 1.6-kbp fragment of the 2.4-kbp insert in pJV7 was produced by digesting the plasmid with SmaI. Cloning the purified fragment in pUC19 gave pDC1.6. To construct an E. coli/Streptomyces shuttle vector from pDC1.6, the plasmid was digested with HindIII and ligated to pIJ6017, also digested with HindIII. The ligation mixture was used to transform S. lividans M252; selection for transformants resistant to kanamycin and Cm (20 and 12.5 g/ml, respectively) yielded pSV1.6.
Construction and Use of the ORF2/ORF3 Deletion Vector pDC-ORF1-To test whether ORF1 alone could confer Cm resistance, a 1-kbp fragment containing ORF2 and ORF3 was excised from pJV7. The plasmid was first linearized with BstEII, and the 3Ј overhang was filled in using the Klenow fragment of DNA polymerase I; digestion of the product with Acc65I furnished a 1.4-kbp fragment of insert DNA. Plasmid pJV4 was then digested with Acc65I and EcoICRI to recover the pIJ702-derived 4.7-kbp vector fragment free of the 2.4-kbp insert. The 1.4-and 4.7-kbp fragments were ligated, and the reaction product was used to transform S. lividans M252. From a transformant resistant to thiostrepton (20 g/ml), plasmid pDC-ORF1 was isolated. Analysis of culture filtrates by HPLC showed that cells containing pDC-ORF1 were not able to convert Cm to 3Ј-phospho-Cm. No colonies were obtained when cells containing pDC-ORF1 were plated on solid media containing Cm (12.5 g/ml).
Construction and Use of the Disrupted ORF2 Vector pJV11-⌬Sal-The SalI site in the polylinker region of pJV7 was removed by digesting the plasmid with HindIII and XbaI, followed by end-filling and religation. Digestion of the resulting plasmid pJV7a with SalI and religation yielded pJV7b in which the 363-bp SalI-SalI segment internal to ORF2 in the plasmid insert had been deleted. The modified insert was excised from pJV7b as a 2.0-kbp BamHI-SacI fragment, and recloned in pHJL400 linearized by digestion with BamHI and SacI, to give pJV11-⌬Sal. As a control, the intact insert from pJV7a was recloned in pHJL400 in the same way to give pJV11. Colonies obtained by transforming S. lividans M252 with pJV11 grew within 2 days of transfer to MYM agar containing 8 g/ml Cm, whereas transformants containing pJV11-⌬Sal failed to grow within 5 days on this medium. Mycelium from the transformants grown for 48 h in GNY medium containing 5 g/ml thiostrepton for plasmid maintenance was washed and resus- pended in aqueous Cm (50 g/ml). Analysis of the suspension at intervals by HPLC showed no conversion to 3Ј-phospho-Cm by mycelium from pJV11-⌬Sal transformants, whereas conversion occurred in control suspensions of mycelium carrying pJV11.
DNA Sequence and Sequence Analysis-For sequencing, the 2.4-kbp KpnI-SstI fragment of pJV4 was subcloned as smaller segments in the vectors M13 mp18, pTZ18R, and pBluescriptII SKϩ. Sets of overlapping nested deletions were prepared with exonuclease III and mung bean nuclease (Mosher, 1993). Single-stranded DNA, prepared either directly or by infection of a phagemid transformant with helper phage (VCSM13), was sequenced by the dideoxy chain termination method (Sanger et al., 1977) using a Sequenase version 2.0 kit (U. S. Biochemical Corp.). Both strands were sequenced, and ambiguous regions were resolved by using synthetic oligodeoxynucleotides (15-21-mers) derived from the generated sequence as sequencing primers, as well as by using deaza-dGTP or dITP in place of dGTP in the sequencing reactions.

Isolation of a Novel Cm Metabolite
Analysis by HPLC of filtrates from cultures grown in GNY medium with Cm at the highest concentration tolerated (12.5 g/ml) revealed that S. lividans RM3 and RM4 rapidly converted the antibiotic to a single major extracellular product ( Fig. 2) with an absorption maximum (273 nm) identical to that of Cm. In direct comparisons, the retention time of the Cm product (5.9 min) differed not only from that of Cm (6.3 min) but also from other related reference compounds, including 1Ј,3Ј-diacetyl-Cm (6.8 min), Cm-base (the free amine of Cm; 4.9 min), and N-acetyl-p-nitrophenylserinol (5.5 min).
Mycelium grown in a variety of media was active in converting Cm to its major metabolite; activity persisted in mycelium harvested after the end of the growth phase, or resuspended in water. Complete conversion was achieved by adding Cm at the concentration tolerated (12.5 g/ml) to cultures of S. lividans RM4 growing in GNY medium. Furthermore, complete conversion was obtained by incubating washed mycelium with 200 g/ml of Cm in water for 48 h. The use of aqueous cell suspensions simplified isolation of the metabolite, and facilitated its purification by reversed phase chromatography. The major chromatographic product was collected as a single sharp peak, well separated from other substances absorbing at 273 nm, and was evaporated to dryness at 40°C in vacuo.

Identification of the Cm Metabolite
The 1 H NMR spectrum of the Cm metabolite was similar to that of Cm in that it contained signals consistent with the presence of a dichloroacetyl group and a 1,4-disubstituted phenylpropanoid derivative. However, it differed in the spin-spin coupling pattern of protons associated with the propanediol moiety. Whereas the H-3aЈ and H-3bЈ resonances for Cm were present as quartets due to coupling with each other and with H-2Ј, the resonances in the metabolite showed complex spinspin coupling, and were displaced downfield with respect to those given by Cm. The data suggested that the propanediol protons of the metabolite were deshielded by an electron-withdrawing functional group probably attached at C-3Ј.
The 13 C NMR spectrum of the metabolite was similar to that of Cm, with most of the signals slightly downfield of corre-  sponding Cm signals. However, the signal assigned to C-3Ј showed a larger shift, from 62.2 to 66.6 ppm, consistent with the presence of an electron-withdrawing substituent on this carbon. That the latter is an orthophosphate group is supported by the 31 P NMR spectrum of the product. The chemical shift ( 31 P, ␦ ϭ 6.474 ppm) of the single strong signal was in the region of the spectrum predicted for an organophosphate ester (data not shown). Examination of the sample by low resolution, negative ion-spray mass spectrometry gave a group of molecular ions in the relative proportions expected for a substance containing two chlorine atoms (Fig. 3). The deduced M r of the compound is 401.97, the value predicted for a monophosphate ester of chloramphenicol. From the combined evidence, the most probable structure of the metabolite is, therefore, 3Ј-phospho-Cm.
To determine whether 3Ј-phospho-Cm exhibited antibiotic activity, the purified compound was compared with Cm in a disk-diffusion assay using M. luteus as the test organism. Inhibition zones were observed only with relatively large samples of 3Ј-phospho-Cm (Ͼ350 g/disk), whereas 0.15 g of Cm gave a measurable zone; in a standard assay, the apparent specific activity of 3Ј-phospho-Cm was 0.04% that of Cm. We conclude from this result that phosphorylation of the 3Ј-OH position of Cm reduces its antibiotic activity significantly.

Sequence and Analysis of the pJV4 Insert
The 2.4-kbp KpnI-SstI insert of S. venezuelae DNA cloned in pJV4 contained a 2355-bp nucleotide sequence (Fig. 4) with 73.6% G ϩ C content. A sequence analysis (Fig. 5) of both strands for third position G ϩ C bias, and for codon usage (Wright and Bibb, 1992), indicated one incomplete and three complete ORFs, the codon usage in which was typical of streptomycetes (Bibb et al., 1984).

Comparison of Derived Amino Acid (aa) Sequences with
Data Base Sequences ORF1-There are two possible ATG start sites for ORF1 (nt 28 -30 and nt 100 -102; see Fig. 4); the first gives a polypeptide 25 amino acids longer than the second. A FASTA sequence alignment (Pearson and Lipman, 1988) comparing ORF1 with sequences in the GenBank and EMBL data bases showed ORF1 to be 42.2% identical over 386 amino acids to CMR, the Cm resistance protein of Rhodococcus fasciens (Desomer et al., 1992), and 39.8% identical over 372 amino acids to CmlG of S. lividans (Dittrich et al., 1991). The amino termini of CMR and CmlG align optimally with Met-25 of ORF1 rather than with Met-1, and since AraJ (Reeder and Schleif, 1991) and NorA (Yoshida et al., 1990), two other hydrophobic proteins with sequence similarity to ORF1, also align at their amino termini with Met-25, the translational start codon for ORF1 is likely to be the ATG at nt 100 -102. This being so, then ORF1 is a polypeptide of 412 amino acids (M r 41,479). ORF1 shows greater than 60% sequence similarity over 300 amino acids to a number of antibiotic resistance proteins presumed to be integral membrane components and capable of promoting active and specific antibiotic efflux (Levy, 1992). Examples include the quinolone resistance determinant (NorA) of Staphylococcus aureus noted above, the methylenomycin resistance protein (Mmr) of the methylenomycin producer Streptomyces coelicolor (Neal and Chater, 1987), and the tetracycline resistance protein (TetL) of Bacillus stearothermophilus (McMurry et al., 1987). All such proteins probably contain 12-14 membrane-spanning ␣-helical segments that form a transmembrane channel (Paulsen and Skurray, 1993).
ORF2-Translation of ORF2 begins at either of two possible start codons: GTG at nt 1412-1414 or ATG at nt 1424 -1426. The sequence 5Ј-GGTGA-3Ј, showing complementarity to the 3Ј end of S. lividans 16 S rRNA (⌬G ϭ Ϫ11.6 kcal: Tinoco et al., 1973;Bibb and Cohen, 1982) is present 4 bp (nt 1403-1407) upstream of the GTG codon, and also 8 bp (nt 1411-1415) upstream of the ATG codon. The 4-bp distance is below the range (5-12 bp) commonly observed for the separation between a ribosome-binding site and the translational start point in streptomycetes (Strohl, 1992). Beginning 26 bp upstream of the GTG codon, the sequence 5Ј-CACCGT-3Ј (nt 1381-1386) represents a possible Ϫ10 hexamer recognized by E 70 -like RNA polymerase (Strohl, 1992). However, the region upstream of the GTG codon contains a second such sequence (5Ј-TACGGT-3Ј; nt 1400 -1405). Transcription from this hexamer should initiate at the first nucleotide of the GTG codon. Such a concurrence of translational and transcriptional start sites is not uncommon in streptomycetes (Strohl, 1992).
Initiation of translation at the GTG codon would yield a polypeptide of 178 amino acids (M r 18,804), whereas a polypep- tide initiated at the ATG codon would contain only 174 amino acids (M r 18,315). Although a FASTA comparison of the amino acid sequence with the GenBank and EMBL data bases revealed no significant similarities, use of the alignment program MPsrch (Sturrock and Collins, 1993) showed resemblance between the NH 2 -terminal region of the ORF2 product and predicted nucleotide-binding sites in such ATP-requiring proteins as pantothenate kinase of E. coli (Song and Jackowski, 1992) and GlnQ of B. stearothermophilus (Wu and Welker, 1991). Examination of ORF2 using the Motifs program (Genetics Computer Group Inc., version 7.3), showed the presence of the "P-loop" phosphate-binding motif found in many classes of adenine and guanine nucleotide-binding proteins (Saraste et al., 1990). The ORF2 sequence GGSSAGKS (aa 10 -17; see Fig. 4) fits the P-loop motif consensus (A/G)XXXXGK(S/T) for an ATP/ GTP binding site (Walker et al., 1982;Saraste et al., 1990), and implicates ORF2 in a process such as phosphoryl transfer. Also present in the ORF2 polypeptide is the sequence DADG (aa 57-60; see Fig. 4), which corresponds to a proposed consensus element (DXXG) for GTP/GDP-binding sites (Dever et al., 1987). The linear separation between the aspartic acid of DXXG, which interacts with nucleotide-bound Mg 2ϩ , and the P-loop lysine, thought to interact with the ␤and ␥-phosphates of the bound nucleotide, conforms to the observed separation (40 -80 residues) between these amino acids (Saraste et al., 1990;Mimura et al., 1991).
ORF3-Translation of ORF3 from the possible GTG start codon (nt 2266 -2264; Fig. 4) would generate a polypeptide of 89 amino acids with a deduced M r of 9,767. A comparison of the amino acid sequence with those in GenBank and EMBL data bases revealed no significant similarities.

Identification of the Cm Resistance Determinant in the 2.4-kbp Insert
The striking sequence similarity between ORF1 and gene products associated with antibiotic resistance in other bacteria suggested a role for ORF1 in the Cm resistance of S. lividans RM4. However, the genes for CMR and CmlG confer resistance to concentrations of Cm greater than 200 g/ml on their hosts (Dittrich et al., 1991;Desomer et al., 1992), whereas the maximal resistance to Cm conferred upon S. lividans M252 by pJV4 is 12.5 g/ml. To determine if Cm resistance is affected by the absence of ORF2 and ORF3, a 1.4-kbp Acc651-BstEII fragment containing only ORF1 was excised from pJV7 (E. coli vector pTZ18R carrying the 2.4-kbp KpnI-SstI insert from pJV4) and subcloned in the residual pJV4 vector segment obtained by digesting pJV4 with Acc65I and EcoICRI to remove the 2.4-kbp insert. The presence of the 1.4-kbp insert in the resulting plasmid (pDC-ORF1; see Fig. 6) was verified by restriction enzyme analysis. Transformation of S. lividans M252 with pDC-ORF1 gave Cm-sensitive colonies, a result indicating that ORF1 by itself did not confer the Cm resistance phenotype of pJV4 transformants.
Evidence that resistance depended on the presence of an intact ORF2 in pJV4 was obtained by deleting a segment of DNA between two SalI sites within ORF2. The deletion plasmid was constructed in the Streptomyces-E. coli shuttle vector pHJL400. Transformation of S. lividans M252 with the vector carrying the 2.4-kbp S. venezuelae DNA insert from which the SalI-SalI fragment had been deleted yielded Cm-sensitive colonies, in contrast to the Cm-resistant transformants obtained with a control plasmid in which ORF2 was intact. Transformants with the deletion in ORF2 also differed from those with ORF2 intact in that they failed to convert Cm to 3Ј-phospho-CM.

Subcloning and Expression of ORF2 in S. lividans
To determine whether ORF2 alone conferred Cm resistance in S. lividans RM4, the 1.67-kbp SmaI-SmaI segment, lacking large segments of both ORF1 and ORF3 (see Fig. 6), was excised from pJV7 and cloned in the E. coli vector pUC19. The resultant plasmid (pDC1.6) was linearized with HindIII, and ligated to the Streptomyces vector pIJ6017 carrying the thiostrepton-inducible promoter tip (Murakami et al., 1989). 2 The E. coli-Streptomyces shuttle vector (pSV1.6) so formed was used to transform S. lividans M252, and colonies were selected for Cm resistance. In bioassays these resistant transformants were inhibited by Cm concentrations above 10 g/ml, compared with 12.5 g/ml for RM4. When cultures were grown in GNY medium containing Cm, HPLC analysis of the culture supernatant showed that the antibiotic was converted to 3Ј-phospho-Cm; however, concentrations in the supernatant were only 20% of those in RM4 cultures. We were unable to measure 3Јphospho-Cm concentrations in the cells from these cultures. Since Cm resistance was observed without induction of the tip promoter, the presence of a functioning promoter upstream of ORF2 seems likely. Overall, the results indicate that phospho-rylation by the ORF2 product is mainly responsible for the Cm resistance phenotype of S. lividans RM4, but the ORF1 product may also have a role.

DISCUSSION
In a previous report (Mosher et al., 1990), the presence of small quantities of metabolites derived from Cm-base in cultures of the Cm-resistant S. lividans strain RM3 supplemented with Cm was offered as evidence that a chloramphenicol amide hydrolase mediated Cm resistance in S. venezuelae. In the present study, a more exhaustive analysis of the culture broths of S. lividans RM3 and RM4 revealed a hitherto unidentified polar metabolite as the major product of Cm metabolism. Enough of the metabolite was obtained by incubating Cm with resting RM4 cells to allow its isolation, purification by C 18 reverse-phase column chromatography, and identification as the 3Ј-O-phosphoryl ester of Cm. That ORF2 has a role in the Cm resistance conferred by pJV4 was established by the concurrent loss of Cm kinase activity and resistance when an internal segment was deleted. We conclude that ORF2 encodes the enzyme that catalyses the phosphorylation of Cm, for which we have proposed the trivial name chloramphenicol 3Ј-O-phosphotransferase.
Much early work on the relationship between the structure of Cm and its antibiotic activity (summarized by Pongs, 1979) emphasized the importance of an unmodified primary alcohol at C-3Ј. Consistent with this, bacterial resistance to Cm is commonly mediated by enzymic acetylation of the C-3Ј hydroxyl group (Shaw and Leslie, 1991). The discovery that a Cm-producing streptomycete modifies the same functional group by an alternative mechanism suggests that this transformation is responsible for self-resistance. Phosphorylation is widely used to confer resistance to antibiotics, both in producing and in susceptible nonproducing organisms (Cundliffe, 1992). Nonetheless, in the more than four decades during which Cm has been used to treat infections in humans and animals, phosphorylation has not hitherto been reported as a mechanism for inactivating this antibiotic. Its presence in a Cm-producing streptomycete implies that, here at least, the origin of a resistance mechanism cannot easily be attributed to horizontal transfer of a gene conferring protection on an antibiotic producer.
Although the deduced primary structure of chloramphenicol 3Ј-O-phosphotransferase does not show end-to-end similarity 2 M. J. Bibb, personal communication.

FIG. 5.
A, analysis of %G ϩ C in the 2355-bp sequence of S. venezuelae DNA using the FRAME programme of Bibb et al. (1984) adapted for the Macintosh Plus microcomputer (Uchiyama and Weisblum, 1985;Doran et al., 1990). B, restriction enzyme sites on the cloned S. venezuelae DNA fragment (thick bar, same scale as A); the arrows immediately below represent the ORFs deduced from the FRAME analysis, and the location of translational start and stop codons. The short arrows show the sequencing strategy using both the exonuclease derived deletions and primers deduced from sequenced DNA. to proteins in current data bases, it does contain localized regions of sequence similarity to a number of proteins requiring nucleotide cofactors. The presence near the NH 2 terminus of a consensus P-loop sequence similar to known ATP/GTP-binding sites is compatible with the role of chloramphenicol 3Ј-O-phosphotransferase in phosphoryl transfer. The P-loop motif occurs in a wide variety of proteins that, although diverse in biochemical function, have both a common nucleotide binding sequence and similar structures (Saraste et al., 1990;Schulz, 1992). Among these are the human Ha-ras p21 protein, E. coli adenylate kinase, EF-Tu and dethiobiotin synthetase (Jurnak, 1985;Pai et al., 1989Pai et al., , 1990Muller and Schulz, 1992;Huang et al., 1994;Alexeev et al., 1994). The ATP-binding site is also present in the transport protein superfamily that includes the cystic fibrosis gene product (Hyde et al., 1990;Mimura et al., 1991). In many proteins of this class the polypeptide is folded into a barrel of parallel ␤-sheets surrounded by ␣-helices to give a "core structure" (Milner-White et al., 1991). The glycine-rich region of the loop contributes to forming a giant anion hole accommodating the nucleotide triphosphate oxygen atoms.
Although three intact ORFs are present in the 2.4-kbp KpnI-SstI DNA fragment cloned in pJV4 from S. venezuelae, large segments of ORF1 and ORF3 can be deleted with only a 10 -20% loss of the Cm resistance phenotype conferred on S. lividans M252 transformants by the plasmid. Cm resistance and the ability to phosphorylate Cm are retained only when ORF2 remains intact; selective disruption of ORF2 eliminates both capabilities. The conclusion that cpt is involved in the resistance conferred by pJV4 is consistent with chemical characterization of the Cm inactivation product as 3Ј-phospho-Cm, and sequence analysis of the deduced gene product indicated that the primary structure contains a nucleotide-binding motif. Although ORF1 contributes to the Cm resistance of S. lividans RM4, it does not have a predominant role. This is unexpected given its similarity to proteins associated with resistance to greater than 200 g/ml Cm in other bacteria. Its lack of activity in pJV4 could be attributed to the absence of upstream regulatory sequences, but the comparable Cm resistance conferred by pJV3, in which 4.1 kbp of upstream sequence from S. venezuelae remains intact, renders this explanation unsatisfactory. A possible alternative is that ORF1 encodes an efflux protein for 3-phospho-Cm, rather than for Cm itself, and that it functions in the presence of chloramphenicol 3Ј-O-phosphotransferase to facilitate the export of inactivated antibiotic.