The biosynthetic pathway of the aminonucleoside antibiotic puromycin, as deduced from the molecular analysis of the pur cluster of Streptomyces alboniger.

The pur cluster which encodes the puromycin biosynthetic pathway from Streptomyces alboniger was subcloned as a 13-kilobase fragment in plasmid pIJ702 and expressed in an apparently regulated manner in the heterologous host Streptomyces lividans. The sequencing of a 9.1-kilobase DNA fragment completed the sequence of pur. This permitted identification of seven new open reading frames in the order: napH, pur7, pur10, pur6, pur4, pur5, and pur3. The latter is followed by the known pac, dmpM, and pur8 genes. Nine open reading frames are transcribed rightward as a unit in opposite direction to that of the pur8 gene which is expressed as a monocistronic transcript from the rightmost end. napH encodes the known N-acetylpuromycin N-acetylhydrolase. The deduced products from other open reading frames present similarities to: NTP pyrophosphohydrolases (pur7), several oxidoreductases (pur10), the putative LmbC protein of the lincomycin biosynthetic pathway from Streptomyces lincolnensis (pur6), S-adenosylmethionine-dependent methyltransferases (pur5), a variety of presumed aminotransferases (pur4), and several monophosphatases (pur3). According to these similarities and to previous biochemical work, a puromycin biosynthetic pathway has been deduced. No cluster-associated regulatory gene was found. However, both pur10 and pur6 genes contain a TTA codon, which suggests that they are translationally controlled by the bldA gene product, a specific tRNA.

The pur cluster which encodes the puromycin biosynthetic pathway from Streptomyces alboniger was subcloned as a 13-kilobase fragment in plasmid pIJ702 and expressed in an apparently regulated manner in the heterologous host Streptomyces lividans. The sequencing of a 9.1-kilobase DNA fragment completed the sequence of pur. This permitted identification of seven new open reading frames in the order: napH, pur7, pur10, pur6, pur4, pur5, and pur3. The latter is followed by the known pac, dmpM, and pur8 genes. Nine open reading frames are transcribed rightward as a unit in opposite direction to that of the pur8 gene which is expressed as a monocistronic transcript from the rightmost end. napH encodes the known N-acetylpuromycin N-acetylhydrolase. The deduced products from other open reading frames present similarities to: NTP pyrophosphohydrolases (pur7), several oxidoreductases (pur10), the putative LmbC protein of the lincomycin biosynthetic pathway from Streptomyces lincolnensis (pur6), S-adenosylmethionine-dependent methyltransferases (pur5), a variety of presumed aminotransferases (pur4), and several monophosphatases (pur3). According to these similarities and to previous biochemical work, a puromycin biosynthetic pathway has been deduced. No cluster-associated regulatory gene was found. However, both pur10 and pur6 genes contain a TTA codon, which suggests that they are translationally controlled by the bldA gene product, a specific tRNA Leu .
Nucleoside antibiotics constitute an important group of microbial secondary metabolites some of which are effective agents against plant and human diseases. Examples are ribavarin (antiviral), polyoxins (herbicides), and mildiomycin (plant antifungal antibiotic). Given the key role accomplished by nucleosides and nucleotides in biochemical processes, the nucleoside antibiotics have found a fundamental application as specific inhibitors of a high variety of biochemical reactions (for a review, see Ref. 1). These features rise up the question of how the producing organisms defend themselves against the toxic effects of their products and, in most cases, of the relevant biosynthetic precursors.
Puromycin is an aminoacyl nucleoside antibiotic produced by Streptomyces alboniger. It is a broad spectrum secondary metabolite active against Gram-positive bacteria, protozoans, and mammalian cells, including tumor cells. It has been a key compound in various cell-free systems directed to elucidate the mechanism of protein synthesis and the mode of action of other inhibitors of this process (2,3). Concerning the biosynthetic pathway of puromycin, some data are available. Adenosine is known to be a direct precursor for the 3Ј-amino-3Ј-deoxyadenosine moiety of puromycin (4). Moreover, a commercial sample of puromycin was found to be contaminated by small amounts of N 6 ,N 6 ,O-tridemethylpuromycin, N 6 ,O-didemethylpuromycin, and O-demethylpuromycin. This led to the proposal that puromycin biosynthesis probably proceeds through these intermediates and in that order (5). Three enzymes of the pathway, an O-demethylpuromycin O-methyltransferase (DmpM), 1 a puromycin N-acetyltransferase (Pac), and an N-acetylpuromycin N-acetylhydrolase (NapH) have been characterized (6 -8). Biochemical studies suggest that Pac inactivates the intermediate N 6 ,N 6 ,O-tridemethylpuromycin by acetylation, and that DmpM methylates N-acetyl-O-demethylpuromycin to form Nacetylpuromycin (7). The latter would be excluded from the cells and then hydrolyzed by the extracellular NapH activity (8) yielding the biologically active puromycin antibiotic. In addition to Pac, the putative transmembrane protein Pur8 was shown to confer resistance to puromycin in Streptomyces lividans and, consequently, in S. alboniger, possibly by promoting an active efflux energized by a proton-dependent electrochemical gradient (9). The complete set of genes (15 kb) encoding the puromycin biosynthetic pathway (pur cluster) from S. alboniger has been cloned in low-copy number cosmids and expressed in a regulated pattern in heterologous hosts, S. lividans and Streptomyces griseofuscus (10). Three genes of pur have been sequenced, pac, dmpM, and pur8, which encode Pac, DmpM, and Pur8, respectively (9,11,12). The three genes are contiguously located at the right end of the cluster and in this order. In addition, the napH gene, located at the left end of pur, was isolated and its product (NapH) characterized (8). Here, we report the completion of the nucleotide sequence of pur, an analysis of its transcriptional organization, and a novel proposal for the puromycin biosynthetic pathway.
Transcription Analysis-To perform S1 mapping experiments, S. alboniger was grown in S medium. RNA was isolated as described (24). RNA was annealed to DNA probes (labeled with [␥-32 P]ATP or [␣-32 P]dCTP to determine transcription initiation and termination, respectively) and then digested with S1 nuclease as described elsewhere (12). The resulting products were developed on either 1% alkaline agarose gels (low resolution experiments) or on 4, 5, or 6% polyacrylamide gels containing 7 M urea for high-resolution determination of transcription initiation/termination sites. The sizes of the protected fragments were evaluated by comparing their mobilities with those of DNA markers or DNA fragments obtained by sequencing reactions.
Computer Analysis-DNA and protein sequences were analyzed with the University of Wisconsin Genetics Computer Group package version 8 (25). Comparisons of the amino acid sequences of Pur proteins with the sequence data bases were also performed using the BLAST program (26) at the NCBI server. The DNA Strider group of programs (27) was occasionally used for sequence analysis.

RESULTS
Boundaries of the pur Cluster-Prior to sequencing, we attempted to define, as precisely as possible, the ends of the pur cluster. By comparing the restriction maps of puromycin-producing and non-producing pKC505-derivative plasmids, we previously showed that pur was encoded by a single DNA fragment of approximately 15 kb (10; Fig. 1). Moreover, the napH and pur8 genes could mark the ends of the cluster ( Fig.  1; Refs. 8 and 9). Therefore, the 13-kb ClaI-EcoRI (labeled E*, Fig. 1 and below) fragment from pCXS ( Fig. 1) was subcloned in the high-copy number plasmid pIJ702. The resulting construct (pRCP11; Fig. 1) was introduced into S. lividans. Puromycin was present in the fermentation broth of a relevant transformant, as determined by TLC and Pac assays (not shown), which indicates that all the structural genes for puromycin production are present in the 13-kb ClaI-E * fragment. Moreover, the amount of puromycin produced in S. lividans(pRCP11) was similar to that in a control S. lividans(pPB5.13) ( Fig. 1 and data not shown). In transformants carrying pRCP11, puromycin production starts at the middle of the log-phase, similarly to S. lividans(pPB5.13) and S. alboniger (10). In the particular case of S. lividans(pRCP11) this cannot be attributed to a variation of plasmid copy number/ structure, since these apparently remained unchanged during all stages of the growth curve (not shown). Therefore, it appears that in S. lividans expression of the pur cluster from pRCP11 is regulated in a similar manner than in S. alboniger or S. lividans(pPB5.13) (10).
Sequence of pur-Starting at the ClaI site (Fig. 1), a total of 9.12 kb were sequenced (Fig. 2). The sequence covered the FIG. 1. Restriction maps of the DNA inserts from several constructs, puromycin production phenotype of S. lividans transformants and organization of the ORFs of pur. A, the restriction maps of cosmids pPB5.13, pPB4.6, and pPB11.40 and plasmid pCXS were modified from (10). Plasmid pRCP11 was obtained by subcloning the 13-kb ClaI-EcoRI (labeled E* in other restriction maps as well) fragment from pCXS into Bluescript. From the resulting plasmid, this fragment was isolated as a KpnI-PstI piece, which was finally inserted in the KpnI-PstI replicon fragment of pIJ702. ϩ and Ϫ indicate production or no production of puromycin by the relevant S. lividans transformants, respectively. Enzymes: C, ClaI; E, EcoRI; K, KpnI; N, NcoI; P, PstI; S, SpeI; X, XhoI; Xb, XbaI. B, ORFs and transcriptional organization of the pur cluster. The incomplete orfA and orf1 are interrupted by the ClaI and EcoRI sites, respectively. The gray region indicates the sequence obtained in this work. Only relevant restriction sites are indicated. Size of the ORFs is indicated by small arrows drawn immediately below. Lines a, b, c, and d indicate the DNA fragments used as probes in low resolution S1-protection assays. Above them, continuous arrows indicate the size and direction of clearly identified transcripts, whereas dotted arrows indicate possible transcripts.  Fig. 1). It contained seven complete and one incomplete ORFs (Figs. 1 and 2). All shared a codon usage and a G ϩ C content at the third position typical of Streptomyces (28), with an exception made of the 5Ј-end of the napH coding sequence (see below). From left to right these ORFs were named/identified as: orfA, napH, pur7, pur10, pur6, pur4 (previously named prg1; Ref. 10), pur5 and pur3. The incomplete orfA probably did not pertain to the cluster, because its incompleteness did not prevent puromycin production from pRCP11 ( Fig. 1). Therefore, the left end of the cluster should reside between orfA and napH. Similarly, in plasmids pPB11.40 and pRCP11 ( Fig. 1), 5Ј to pur8 there is another incomplete ORF (orf1) (Ref. 9; Fig. 1), which indicates that the right end of pur lies between these two ORFs. Several characteristics of the intact ORFs are indicated in Table I. Previously, it was shown that the napH gene is part of a 2.5-kb SphI-NcoI fragment from the left end of pur (8). The single complete ORF of 1458 nucleotides found in this fragment ( Fig. 2) was, therefore, attributed to napH. Its deduced product has 485 amino acids (Table I). Both upstream and downstream of the chosen initiator codon there are other putative initiator codons (Fig. 2). They were rejected as probable initiators because, for those located downstream, destruction of a BamHI site at position 730 (Figs. 1 and 2) prevented expression of NapH activity (8) and, for those located upstream, only the selected initiator codon was preceded by a good ribosomal binding site (Table I; Fig. 2). NapH is an extracellular enzyme (8) and, in agreement with this, its N-terminal region presents the characteristics of a signal peptide (29). Thus, it contains a positively charged leading region (MLHRIQRKR), followed by a hydrophobic central region (AMTAGAVGVLFLAQLVI) and a polar region (SSSSAAA). The putative signal peptide contains 33 residues, a value close to the mean size of 35 for the putative signal peptides from Streptomyces (30). The presence of this signal peptide also helps to select the corresponding DNA se-FIG. 2. Nucleotide sequence of a 9.12-kb DNA fragment from the pur cluster. The deduced gene products are indicated in the one-letter code under the DNA sequence. Possible ribosome binding sites (rbs) are indicated by dotted lines. Putative translation initiation codons are in bold letters. The start and direction of each of the ORFs are indicated by up arrows and named accordingly. A presumptive signal peptide of NapH is underlined and its putative cleavage site is indicated by a vertical arrow. A proposed motif for Pur10 and similar oxidoreductases is underlined. An inverted repeat located 3Ј to pur5, which could form a stem-loop for transcription termination is indicated by horizontal arrows. Putative Ϫ10 and Ϫ35 regions of napH are indicated. Possible transcription initiation and termination sites are indicated by bold triangles and an open triangle, respectively. Restriction sites with an asterisk are not unique in the sequence DNA; they are referred to in the text. The 5Ј region of the previously reported pac sequence (11), which has been revised in this work, is indicated by small letters.
quence as a coding region, despite the fact that its codon usage deviates from that typical of Streptomyces. Comparisons of the NapH sequence with those in data banks disclosed significant similarities (28.9% identity; 49.3% similarity) with the aminopeptidase Y from Saccharomyces cerevisiae (31).
The ORF corresponding to pur7 comprises 459 nucleotides, of which the first eight overlap with the 3Ј end of the napH coding sequence. A similar situation was previously found for other Streptomyces genes (32). pur7 encodes a hydrophilic protein (Pur7) of 152 residues (Table I; Fig. 2) which contains a highly conserved domain present in proteins from a wide variety of organisms. Besides this, no generally significant similarities are present in this group of proteins. This domain was defined by Koonin (33) as a region of approximately 40 residues, generally located at the N terminus. Its consensus sequence is GX 5 EX 4 -(STAGC)-(LIVMA)-XRE-(LIVMF)-XEE, where X represents any amino acid and residues in parentheses indicate the most frequent ones in those positions (Fig. 3). It has been proposed that this domain is the catalytic center of NTP-pyrophosphohydrolases which produce PP i and the relevant NMP. Indeed, such activity has been found in several of these proteins of which the most thoroughly studied are MutT from E. coli, which names the family (34 -37), MutX from Streptococcus pneumoniae (38), MutT from Proteus vulgaris (39), and MutT from humans (40). MutT hydrolyses 8-oxo dGTP and, less efficiently, other nucleotides (41). Curiously, dATP is a better substrate for MutX, which presents the highest similarity to Pur7, than for other members of the MutT family (42). Moreover, Orf17 from E. coli (42), which does not complement the mutT mutation in E. coli, hydrolyses all dNTPs. These data suggest a nucleotide-pyrophosphohydrolase activity for Pur7.
The pur6 gene would encode a hydrophilic protein of 772 amino acids (Table I; Fig. 2). Similarly to pur10, it possesses a TTA codon which encodes leucine 3 (Fig. 2). Sequence compar-isons only showed certain similarities between the N-terminal halves of Pur6 and the LmbC deduced protein of the lincomycin biosynthetic pathway from S. lincolnensis. Thus, the first 300 amino acids of these two proteins share 24.6% identity and 48.1% similarity. The precise function of LmbC is still unknown. However, it presents significant similarities with members of the peptide synthetase family of enzymes as well as to a subfamily of aminoacyl adenylate-forming domains of aromatic amino acid-activating enzymes (47). LmbC has been proposed as an L-tyrosine-activating enzyme, besides other putative activating functions. All these proteins contain an AMPbinding domain, which could not be detected in Pur6. Given the low similarity of Pur6 with peptide synthetases, i.e. with TycA (tyrocidine synthetase from Bacillus subtilis, accession number P09095) the values are 19.6% identity and 40.2% similarity, it is difficult to assign an amino acid (i.e. tyrosine, a moiety of puromycin) activating activity for Pur6, although there are some hints to suggest that it might be implicated in the transfer of tyrosine to 3Ј-amino-3Ј-deoxyadenosine (see "Discussion").
The pur4 gene would determine a protein (Pur4) of 429 amino acids (Table I; Fig. 2). Although this structure is chiefly hydrophilic, it contains some hydrophobic stretches in its Nterminal half. A partial sequence of Pur4 was previously reported (10). It presented significant similarities with the deduced amino acid sequences of eryC1, from the erythromycin biosynthetic pathway of Saccharopolyspora erythraea (48), orf10.4 from the rfb cluster of Salmonella typhimurium (49), and the pleiotropic, putative regulatory gene degT of B. subtilis (50). Also, significant similarities of Pur4 were found with the products of the dnrJ, strS, tylB, and lmbS genes, which are implicated in the biosynthetic pathways of several antibiotics from Streptomyces (47,(51)(52)(53). Based on the clear similarities of these genes with the degT gene, a regulatory role in the relevant biosynthetic pathways was proposed for some of them (10,51,52). More recently, however, it was suggested that these genes including pur4 may encode aminotransferases (54 -56). Indeed, all of them have a conserved lysine residue (lysine 204 in Pur4; Fig. 2) which could be used to bind the presumptive aminotransfer coenzyme pyridoxal phosphate by forming a Schiff base. Moreover, this conserved lysine is located in a motif (GX 3 DX 7 AX 8 EDX 14 GX 13 KX 4 -5 geGGX 19 G; residues 153-229 in Pur4; Fig. 2) according to Piepersberg (56). Therefore, Pur4 may display an aminotransferase activity.
The pur5 gene encodes a protein (Pur5) of 228 amino acids (Table I; Fig. 2). This product is mainly hydrophilic with a hydrophobic stretch between residues 40 and 80. Sequence comparison of Pur5 with proteins from data banks detected a small glycine-rich region between amino acids 64 and 80 which is highly conserved, usually at the N-terminal region, in a variety of methyltransferases and other enzymes which use AdoMet as a cosubstrate (Fig. 4). This domain corresponds to region I as defined by Ingrosso et al. (57). Its consensus sequence was later slightly modified (58) as hhD/EhGXGXGX-hXXXhh⌬, where h is a hydrophobic amino acid, X is any amino acid, and ⌬ is generally a charged amino acid. The presence of this charged amino acid in the last position of the domain could be questioned, according to the alignment presented in Fig. 4. Two other regions (II and III), which were proposed elsewhere (57,59) are not rigorously detected in several AdoMet-dependent methyltransferases including Pur5 (58; not shown). The presence of region I in Pur5 suggests that it may have a AdoMet-dependent methyltransferase activity. Finally, the pur3 gene would encode a protein (Pur3) of 273 amino acids (Table I; Fig. 2). Although this product is mainly hydrophilic it contains two significant hydrophobic regions between residues 90 -120 and 190 -230. Pur3 has significant similarity with a family of proteins (60), which includes a variety of inositol monophosphatases and other proteins of different origins (Fig. 5), including the bovine IMP (28.3% identity and 44.8% similarity), a key enzyme of the inositol phosphate second messenger pathway (61). This enzyme hydrolyses D-inositol phosphates producing inositol. It is also significant the similarity of Pur3 with the Hal2 protein from S. cerevisiae (27.9% identity; 47.9% similarity), which is involved in salt tolerance and methionine biosynthesis (62), and CysQ from E. coli (27.7% identity; 48.3% similarity), which is required for cysteine biosynthesis and could help control the pool of 3Јphosphoadenosine 5Ј-phosphosulfate (63). All these proteins share two highly conserved domains (60), as well as certain identical residues mainly at the N-terminal half (Fig. 5). The tridimensional structure of IMP (64) indicates that these two domains participate in the binding of phosphate and Mg 2ϩ , which are essential for enzymic activity. The two domains are also present in the bovine inositol polyphosphate phosphatase (65), which has scarce similarity in other parts of the sequence with other proteins of the family. As indicated by Neuwald et al. (60), most proteins carrying these two domains could be phosphatases. This activity could, therefore, be attributed to Pur3.
Transcript Mapping-Previous work showed that the pac and dmpM genes were transcribed in a single RNA which terminated 3Ј to the second gene (12). The sequence reported here and elsewhere (11,12) shows that nine ORFs of pur, closely linked to each other, share the same direction of transcription (Fig. 1). Moreover, an inverted repeat, which could form a strong stem-loop of ⌬G ϭ Ϫ35.2 kcal/mol (66), is located 3Ј to pur5 (Fig. 2). It could act as a transcription termination structure similar to that found for other Streptomyces genes (67,68). Other putative transcription termination structures were previously found 3Ј to dmpM and pur8 (9,12). To determine the transcriptional organization of the pur cluster, total RNA was isolated from a late log phase culture of S. alboniger, when puromycin was actively synthesized. Initially, a broad localization of possible transcription initiation sites was carried out by low resolution S1 protection assays with probes indicated in Fig. 1. The results suggested that a transcript, which started 5Ј to napH and covered all nine ORFs with the same direction of transcription was present. In addition, transcription initiation between pur5 and pur3 was also detected (not shown). The clear lack of transcription initiation between pur3 and pac is in contrast to previous data, which indicated both promoter activity and transcription initiation 5Ј to pac (11). However, these data were obtained from S. lividans transformants which contained the pac gene inserted in plasmid vectors and were grown in a different culture medium than the one used here. It is possible, therefore, that we have been unable to detect those events in S. alboniger.
A higher resolution study was also carried out using S1 protection experiments to confirm the existence of the transcription start points detected by low resolution nuclease-S1 assays. To locate the transcription initiation site 5Ј to napH, we used a probe (Fig. 6) covering its 5Ј coding region and the noncoding region between this gene and orfA. The nuclease assay showed a protected fragment of about 500 bp (Fig. 6), which indicated that napH transcription started around position 679, 41 bp upstream of the translational initiator GTG (Fig. 2). Possible Ϫ10 and Ϫ35 regions are indicated in Fig. 2.
To identify a possible transcription initiation site 5Ј of pur3, the probe used (Fig. 6), covered the 5Ј end of this ORF and the region between pur5 and pur3. A protected fragment of approximately 290 bp (Fig. 6) suggested the presence of a transcription initiation signal around position 8210, 43-bp upstream of the pur3 initiator ATG (Fig. 2). No clear Ϫ10 and Ϫ35 regions were found. Transcription initiation upstream of pur8 was examined using a probe that covered the 5Ј end of pur8 and all the noncoding region up to orf1 (9). A protected fragment of 124 The stem-loop 3Ј to pur5 referred to above suggested that transcription termination could take place at this region. To examine this possibility, a high resolution S1 protection experiment was performed using a probe (Fig. 6) that comprised all the DNA between pur5 and pur3, including the stem-loop (Fig.  2), plus a Bluescript tail. The results indicated the presence of major protected fragments of approximately 840 and 512 bp, and minor protected ones of 535-545 and 1080 bp sizes (Fig. 6). The 1080-bp fragment corresponds to probe/probe reannealing. The 840-bp fragment corresponds to full protection of the probe minus the non-homologous plasmid sequence, thus confirming the existence of a transcript that extends through the stem-loop into pur3. The protected band of approximately 512 bp may correspond to transcription termination close to the 5Ј end of the stem-loop. The minor protected bands suggest that transcription termination occasionally occurs at the region close to or covering the loop located within the inverted repeat 3Ј to pur5. However, since these protected fragments have not been found 3Ј to the putative transcription terminator, it is questionable that these terminations really take place in vivo, and other possible roles for this stem-loop must be considered (see "Discussion"). Therefore, these studies and those reported elsewhere (12) show that the pur cluster is transcribed into at least two mRNA species: one of them is a polycistronic messenger that covers nine ORFs, and the other one is a monocistronic messenger transcribed in the opposite direction that corresponds to pur8. In addition, there may be two other transcripts that span napH through pur5 and pur3 through dmpM, respectively (Fig. 1). DISCUSSION Gene Organization of the pur Cluster-This paper shows that the genes of the pur cluster, which determines the puromycin biosynthetic pathway of S. alboniger, are located in a single DNA fragment of approximately 13 kb. It comprises 10 ORFs, nine of which are contiguous and have the same direction of transcription and only one (pur8), located at the right end of the cluster, is transcribed in the opposite direction. The different enzymic steps of this pathway should be assigned to these ORFs. In fact, the biochemical studies of some of the expressed proteins (5,7,8), together with the similarities of the deduced gene products with known proteins, have allowed the proteins to attribute specific functions to all the genes of the cluster. As suggested by others, the possibility that some unlinked structural gene(s) could play a role in the biosynthesis of this antibiotic should not be excluded (47,69). If so, such gene(s) should also be present in S. lividans, since puromycin production was achieved in this organism carrying the 13-kb fragment. The analysis of the biosynthetic gene cluster for puromycin is of special interest since it is the first one to be isolated for a nucleoside antibiotic. It should serve as a model system for related clusters.
Transcription of the pur Cluster-The transcription analyses of the pur cluster described here and elsewhere (12) indicate the existence of a polycistronic transcript which starts 5Ј to napH and terminates 3Ј to dmpM and comprises all ORFs of pur, except pur8, which is transcribed in the opposite direction as a monocistronic transcript. The expression of most ORFs of pur as a polycistronic transcript is in agreement with several other clusters for antibiotic biosynthesis (69 -71). As indicated elsewhere (71), these polycistronic transcripts should facilitate the regulation of antibiotic biosynthesis improving its efficiency. In addition, two other transcripts, comprising napH through pur5 and pur3 through dmpM, respectively, may be synthesized in the pur cluster. Nevertheless, as indicated above, the transcription terminator function of the stem-loop located between pur5 and pur3, which would explain the exist- FIG. 5. Alignment of Pur3 with several proteins. The bovine brain inositol monophosphatase (IMP) was introduced as an example of proteins displaying this enzymic activity. QutG from Aspergillus nidulans (28.3% identity; 49.4% similarity; 88) and Qa-X from Neurospora crassa (27% identity; 49.6% similarity; 89) are implicated in the utilization of quinic acid in these organisms. SuhB from E. coli (30.6 identity; 50.6 similarity) is implicated in the regulation of a transcriptional -factor of genes encoding heat-shock proteins (90). Yhr046C is an ORF from S. cerevisiae which pertains to this family of proteins (25.5% identity; 47.4% similarity) (91). The values of identity and similarity refer to comparisons of the different proteins with Pur3. Alignments and features are as in Fig. 3. The final amino acids of several proteins are not presented. ence of these two transcripts, may be questioned. Therefore, transcription termination 3Ј to pur5 and transcription initiation 5Ј to pur3 may not occur in vivo. An alternative role for this stem-loop might be to confer variable stability to different segments of the polycistronic transcript (71). Finally, additional promoters active under certain physiological conditions and that could contribute to the regulation of pur expression should not be discarded.
The Puromycin Biosynthetic Pathway-Sequencing of a gene cluster provides an analysis of its molecular organization and an insight, by sequence comparison with data banks, into the functions of many of its putative gene products. Moreover, if additional biochemical data are available, this insight can be achieved with a high degree of certainty. The rationale to propose the initial steps of the puromycin biosynthetic pathway from S. alboniger takes into account the finding that [U-14 C]adenosine is a direct precursor of the 3Ј-amino-3Ј-deoxyadenosine moiety of puromycin (4). The 3Ј addition of an amino group to the ribose moiety of adenosine should take place, similarly to other deoxysugars, through a 3Ј-keto intermediate (for a review, see Ref. 55). Therefore, in the puromycin pathway this intermediate should be 3Ј-keto-3Ј-deoxyadenosine, which by means of an aminotransferase would be converted to 3Јamino-3Ј-deoxyadenosine. However, in both Gram-positive and Gram-negative bacteria and in Ehrlich ascites tumor cells, 3Ј-amino-3Ј-deoxyadenosine is triphosphorylated by adenosine kinase, producing 3Ј-amino-3Ј-deoxy-ATP, a strong inhibitor of DNA-dependent RNA polymerase (72)(73)(74). Therefore, it seems likely that, if produced, 3Ј-keto-3Ј-deoxyadenosine would also be 5Ј-triphosphorylated. The resulting product, 3Ј-keto-3Ј-deoxy-ATP, could also be highly toxic. If so, S. alboniger should have a means to prevent the harmful effects of these putative intermediates. This could be either a resistant RNA polymerase or an adenosine kinase which does not recognize the 3Јderivatives. Moreover, it seems safe to assume that these mechanisms of autodefense are not present in S. lividans, where pur determines puromycin production. Therefore, the resistance system should most likely be encoded by this cluster. We propose Pur7 as the responsible enzyme conferring this resistance. Pur7 is a member of the MutT family of NTPpyrophosphohydrolases, which play an important role in the detoxification of certain mutagenic/carcinogenic keto deriva-FIG. 6. Transcription analysis. Total RNA used was isolated (14) from S. alboniger samples taken at 5.0 A 660 nm (10), when puromycin was actively produced as shown by a Pac assay. RNA samples (50 g) were hybridized to the 32 P-labeled probes described below. High resolution S1 mapping was performed as described (11). Lanes G, A, T, and C contain standards from sequencing reactions of known sequences from pur (B, C, and D), and a fragment from M13 mp18 (A). For D, 29 digested with HindIII and labeled with 32 P was also used as size standard. Numbers on the left margins indicate the size (in nucleotides) of the standards and those on the right the size (in nucleotides) of the DNA fragments protected from the nuclease S1 digestion. Lines 1 and 2 represent S1 assays performed with S. alboniger RNA and yeast tRNA, respectively. Line 3 represents DNA probes. A, transcription initiation of napH. The 2.07-kb BglII-XhoI fragment from pur (nucleotides 363-2429; Fig. 2) was cloned in Bluescript. From the resulting construct (pPS1.8), an 855-bp SplI (internal site to the napH coding sequence, Fig. 2)-SacI (from the polylinker of the vector) fragment was isolated, labeled at the SplI end and then used as a probe. B, transcription initiation of pur3. From plasmid pPS5.2 (5.5-kb XhoI-EcoRI fragment from pur, cloned in pUC19; XhoI, nucleotide 7523, Fig. 2. EcoRI, right end, Fig. 1), a 984-bp SalI (internal site to the pur3 coding sequence; Fig. 2)-PstI (from the polylinker of the vector) fragment was isolated, labeled at the SalI end and then used as a probe. C, transcription initiation of pur8. The 2-kb SalI-EcoRI rightmost fragment of pur was cloned in Bluescript. From the resulting plasmid (pPS2.0; 9), a 743-bp BamHI (internal site to the pur8 sequence)-PstI (from the polylinker of the vector) fragment was isolated, labeled at the BamHI end, and then used as a probe. D, transcription termination of pur5. From plasmid pPS1.0 (984-bp SalI-PstI fragment described above, cloned in Bluescript), a 1.08-kb StyI (internal site to pur5 sequence; Fig. 2)-SspI (from the vector polylinker) fragment, was isolated, labeled at StyI and used as a probe.
tives of the NTPs. Thus, the MutT protein from E. coli and humans hydrolyses 8-oxo-dGTP producing PP i and an inactive 8-oxo-dGMP (41,75). Therefore, 3Ј-keto-3Ј-deoxy-ATP could be inactivated by Pur7 producing a supposedly nontoxic 3Ј-keto-3Ј-deoxy-AMP and PP i . This proposal is also based on the finding that the mono-and diphosphorylated derivatives of 3Ј-amino-3Ј-deoxy-ATP lack biological activity (76,77). Therefore, 3Ј-keto-3Ј-deoxy-AMP would then be the substrate for an aminotransferase to produce a non-toxic 3Ј-amino-3Ј-deoxy-AMP. In this way, the initial steps of the pathway could proceed through inactive intermediates. Because puromycin biosynthesis initiates when growth is still maintained (7), adenosine would preferentially be converted into ATP rather than into 3Ј-keto-3Ј-deoxyadenosine. Therefore, it seems plausible to propose that adenosine would enter the pathway via ATP (compound I; Fig. 7). If so, the latter would be converted by the putative oxidoreductase Pur10 (NAD ϩ could be its cofactor; Ref. 55) into 3Ј-keto-3Ј-deoxy-ATP (compound II; Fig. 7), which would be hydrolyzed by Pur7 to yield 3Ј-keto-3Ј-deoxy-AMP (compound III; Fig. 7). This intermediate would then be modified by the presumptive aminotransferase Pur4 to produce 3Ј-amino-3Ј-deoxy-AMP (compound IV; Fig. 7). As described for aminotransferases which are implicated in amino sugar biosynthesis in Gram-negative bacteria (78), Pur4 would be a pyridoxal phosphate/L-glutamate(L-glutamine)-dependent enzyme (54,56). Assuming that the pathway proceeds in this manner, the 5Ј-phosphate group should be eliminated at some step. This hydrolysis could be carried out by Pur3, which presents significant similarities to a variety of monophosphatases (Fig. 5). However, it should not be at this stage of the pathway because the resulting intermediate would be phosphorylated back to 3Ј-amino-3Ј-deoxy-ATP. Therefore, Pur3 should act on a subsequent intermediate.
It has been proposed that tyrosine would be attached by its -COOH group to the 3Ј -NH 2 group of 3Ј-amino-3Ј-deoxyadenosine to produce N 6 ,N 6 ,O-tridemethylpuromycin (5,7). According to our proposal, formation of this bond should take place on the 5Ј-phosphoderivative (Fig. 7), thus removing it from the nucleotide pool to prevent any additional phosphorylation. This reaction would be equivalent to a peptide bond formation step, which would previously require an activation of tyrosine. Amino acid activation is performed by aminoacyl-tRNA synthetases, antibiotic peptide synthetases, the bacterial peptidoglycan precursor synthetases, and condensing enzymes of the ␤-lactams antibiotics (79 -82). Assuming that the relevant gene(s) is(are) not on a separate locus(loci), the only putative enzyme, by elimination, of the puromycin biosynthetic pathway to which this function could be attributed is Pur6. Although it presents low level similarities to these enzymes and apparently lacks any AMP binding domain, this might not be an impeding difficulty. Thus, LeuS, a leucyl-tRNA synthetase, has low similarities to members of the peptide antibiotic synthetases and lacks an apparent AMP binding site (83). The only ORF from data banks which presents a similarity to Pur6 is LmbC from the lincomycin biosynthetic gene cluster. This ORF appears to belong to the peptide synthetase family of proteins. It has been proposed to be an activating enzyme of either tyrosine, before conversion to L-dihydroxyphenylalanine, or propylproline (a tyrosine-derivative intermediate), prior to condensation at the -NH 2 group of the sugar moiety of lincomycin (47). In our case, the activation of tyrosine could be performed by a different enzyme (i.e. a tyrosinyl-tRNA synthetase) and Pur6 might only catalyze its linkage to 3Ј-amino-3Јdeoxy-AMP to produce N 6 ,N 6 ,O-tridemethylpuromycin-5Јphosphate (compound V; Fig. 7). Given the rarity of the linkage attributed to Pur6, a sequence deviation from peptide bond forming enzymes should not be surprising. Indeed, Pur6 could be a member of a variety of enzymes which are implicated in the biosynthetic pathways of certain nucleoside antibiotics of bacterial and fungal origin, like chriscandin, A201A, and the agricultural fungicides polyoxins, where a variety of polycarbon chains are attached to the amino-ribofuranosyl moieties by forming a peptide bond-like linkage (1).
It has been proposed that N 6 ,N 6 ,O-tridemethylpuromycin is dimethylated at N 6 (5). However, the resulting intermediate inhibits protein synthesis (84), and to prevent it, a Pac-dependent inactivation by N-acetylation of the -NH 2 group of the tyrosinyl moiety of tridemethylpuromycin has been suggested (7). According to our proposal, this acetylation should take place on compound V (Fig. 7). The resulting intermediate (compound VI; Fig. 7) would be N 6 -dimethylated by Pur5, which presents a typical motif of AdoMet-dependent N-methylases (Fig. 4). Curiously, AdoMet-dependent N 6 -methylation of either N 6 ,N 6 ,O-tridemethylpuromycin or N 6 ,O-didemethylpuromycin with cell extracts from S. alboniger has not been detected. This result was attributed to the possible requirement of a precursor derivative such as a phosphate ester (5). Indeed, our proposal suggests that this modified precursor is compound VI (Fig. 7). The resulting N-acetyl-O-demethylpuromycin-5Ј-phosphate (compound VIII, via compound VII; Fig. 7) would be O-methylated by DmpM (5,7). However, the affinity of DmpM for N-acetyl-O-demethylpuromycin (compound IX; Fig. 7) is very high (K m ϭ 2.3 M; 7), which suggests that this, instead of VIII, is the real substrate for this enzyme. Therefore, the 5Ј-phosphate group could be removed by Pur3 prior to O-methylation. If so, this removal could take place even at an earlier step (i.e. from compound V; Fig. 7). This sequence of reactions would produce N-acetylpuromycin (compound X; Fig.  7), which would be secreted to be N-hydrolyzed by NapH, the puromycin (compound XI; Fig. 7) activating enzyme (Ref. 8; Fig. 7).
Regulation of the pur Cluster-It has been surprising not to find evidence for a pathway-specific regulatory gene associated to the pur cluster. While this is also the case for the tcm biosynthetic pathway isolated from the tetracenomycin C producer, Streptomyces glaucescens (69), many antibiotic biosynthetic gene clusters have a transcriptional activator associated with them (85). Given that puromycin production is clearly temporally controlled, it is possible that some pleiotropic regulator is responsible for activating the pur cluster. A number of pleiotropic regulatory genes have been isolated from Streptomyces coelicolor which affect multiple antibiotic pathways, but are not pathway-associated (86). The observation that pur expression appears to be regulated in S. lividans(pRCP11) implies that such a putative pleiotropic regulator(s) is not specific to S. alboniger and, indeed may even be widespread among actinomycetes. The isolation and characterization of these genes may uncover novel regulatory mechanisms in secondary metabolite biosynthetic pathways within this important group of bacteria. In this context, the existence of TTA codons in the 5Ј ends of the coding regions of both pur10 and pur6 may be of interest, since it suggests that both genes are controlled by the product of the bldA gene (43). Interestingly, according to our proposal for puromycin biosynthesis, Pur10 would catalyze the first step of the pathway, and it is tempting to speculate on the role of translational regulation in expression of the pur cluster.