Nucleotide Sequence and Deduced Functions of a Set of Cotranscribed Genes of Streptomyces coelicolor A3(2) Including the Polyketide Synthase for the Antibiotic Actinorhodin”

A 5.3-kb region of the Streptomyces coelicolor acti- norhodin gene cluster, including the genes for polyketide biosynthesis, was sequenced. Six identified open reading frames (ORF1-6) were related to genetically characterized mutations of classes actI, VII, IV, and VB by complementation analysis. ORF1-6 run diver-gently from the adjacent act111 gene, which encodes the polyketide synthase (PKS) ketoreductase, and appear to form an operon. The deduced gene products of ORF1-3 are similar to fatty acid synthases (FAS) of different organisms and PKS genes from other polyketide producers. The predicted ORF5 gene product is similar to type I1 &lactamases of Bacillus cereus and Bacteroides fragilis. The ORF6 product does not re- semble other known proteins. combining the genetical, biochemical, and similarity data, the potential activi- ties of the products of the six genes can be postulated as: + ORF2); 2) acyl carrier protein (ORF3); 3) putative cyclase/dehydrase (ORF4); 4) dehydrase (ORF5); and 5) “dimerase” (ORF6). The data show that the actinor- hodin PKS consists of discrete monofunctional components, like that of the Escherichia coli (Type 11) FAS, rather than the multifunctional polypeptides for the macrolide PKSs and vertebrate FASs (Type I). Polyketides are a vast class by living

publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "aduertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ Supported by a fellowship from the Spanish Ministerio de Educacibn.
11 Supported by a postdoctoral fellowship from the Spanish Consejo Superior de Investigaciones Cientificas.
The nucleotide sequence(s) reported in this paper has been submitted X63449.
to the GenBankTM/EMBL Data Bank with accession numbeds) **To whom correspondence should be addressed: Facultad de Medicina de la Universidad Autbnoma de Madrid, Arzobispo Morcillo 4, 28029 Madrid, Spain. Fax: 34-1-5854015;Tel.: 34-1-3975452. some of them have a possible role in the physiology of the producers, while others are commercially valuable antibiotics or pharmacologically active materials, pigments, or flavoring agents (3). Despite their chemical heterogeneity, all of them share a common pattern of biosynthesis via the successive condensation of simple carboxylic acid metabolites. Their synthesis is conceptually similar to that of long-chain fatty acids, as first suggested by Collie (4). However, while fatty acid and polyketide biosynthesis resemble each other in enzymology, there are substantial differences in detail: the lack of some reductive steps in polyketide chain assembly, the heterogeneity of choices of starter and chain extender units among the different pathways, and the need for aromatization in some of them. These differences account for the great variety of chemical structures found among the polyketides.
Recently, several sets of genes for fatty acid synthases (FAS)' of animals (5)(6)(7), fungi (8)(9)(10)(11)(12), and Escherichia coli (13) and polyketide synthases (PKS) from actinomycetes, fungi, and plants (reviewed in Ref. 3) have been cloned and sequenced. Two structural types of complex enzymes (type I : multifunctional proteins, and type 11: multienzyme complexes) can be found among the different PKSs (see Ref. 3 for review). This basic knowledge of the physical structure, organization, and functions of genes encoding polyketide synthases will constitute a powerful tool to understand the mechanisms by which the primary polyketide carbon chain is assembled and channelled toward the final product in the different pathways.
Actinorhodin is a polyketide antibiotic produced by Streptomyces coelicolor A3 (2), which is genetically the most characterized actinomycete (14). Blocked mutants were isolated and grouped into seven phenotypic classes (actI-VII) (15); the ability of different classes to convert intermediates secreted by other mutants, blocked later in the pathway, into actinorhodin, placed the mutants in a biosynthetic sequence. By genetic complementation of one of the late-blocked mutants, the whole pathway was cloned in a single DNA fragment that conferred the ability to produce actinorhodin on a nonproducer, Streptomyces paruulus, when it was introduced into it by transformation (16).
By a combination of genetic complementation of the blocked act mutants and insertional inactivation of the cloned genes, the physical localization and organization of the act genes was determined on a 25-kb region of DNA within the cloned fragment (17). The act genes were grouped into three  (17). Below the general restriction map of the act cluster, the region analyzed on this paper is magnified, showing relevant restriction sites referred to in the text (and others). The sequenced region is shown as a shaded bar; the arrangement and direction of the different ORFs is given below the restriction map of the enlarged region. act DNA fragments (thick lines) cloned on different vectors are placed under the corresponding regions. Dotted lines with arrowheads (only for the relevant plasmids) show the orientation of the cloned fragment in relation to the vector. different regions: a central regulatory region (actII), flanked by "early" and "intermediate" genes (actI, 111, IV, VII, and VB) on one side and by intermediate and late genes (actVA and VI) on the other (Fig. 1). Nearly 6 kb of the central region of the cluster (the act11 region) has recently been characterized and its gene products associated with antibiotic export and with the regulatory mechanism for the biosynthetic genes (18). To the left of the act11 region, genes corresponding to the actVA and actVI mutant classes have been characterized (19).' The act111 gene was sequenced (20) and its product identified as the ketoreductase, one of the catalytic functions found in all FASs and presumed to be a component of the PKS complex for actinorhodin biosynthesis. This paper describes the molecular characterization of the region adjacent to and to the right of actIII, which comprises the act early and some intermediate genes.

MATERIALS AND METHODS
Bacterial Strains, Plasmids, and Bacteriophages-The E. coli strains were JMlOl(21) and XL1-Blue (22). The Streptomyces strains and E. coli and Streptomyces vectors are summarized in Tables I and  11, respectively. Media, Culture Conditions, and Microbiological Procedures-Streptomyces manipulations were as in Hopwood et al. (23). Thiostrepton (a gift of S. J. Lucania, Bristol-Myers-Squibb Research Institute, Princeton, NJ) was used a t a concentration of 50 Gg/ml for solid media and 10 rg/ml for liquid media. Hygromycin (Sigma, Cat. number H2638) was used at 200 and 50 pg/ml in solid and liquid E. Martinez, M. A. Fernandez-Moreno, D. A. Hopwood, and F. Malpartida, manuscript in preparation. media, respectively. E. coli strains were grown on L agar or in L broth (24).
DNA Sequencing-DNA sequencing was carried out by the dideoxy-chain termination method (25); we used the 7-deaza-dGTP reagent kit from United States Biochemical (Cat. number 70750), following the manufacturer's recommendations. Convenient DNA fragments were previously cloned on either M13 mp18 or M13 mp19 vectors from suitable restriction fragments or generated by Ex0111 digestion (26).
Computer Analysis of Sequences-The DNA sequence was analyzed for open reading frames using CODONPREFERENCE (from UWGCG package (27)). Amino acid sequences were analyzed using various programs from the UWGCG package (version 7.0, April 1991); comparisons of sequences were made against the EMBL gene data base (daily updated, October 1991), and Swissprot Data Base, (Release 19.0, updated August 1991), using FASTA, TFASTA, BESTFIT, COMPARE, and DOTPLOT. Protein alignments were made using LINEUP (from the UWGCG package).
Gene Disruption and Mutant Assignments-For gene disruption and phage "complementation" experiments, we used insert-directed recombination as described previously (28), using S. coelicolor strain 51501 or act mutants as hosts. The trans-complementation tests were carried out by transformation as described elsewhere (23) with either low or high copy number Streptomyces plasmids.
DNA and RNA Manipulations-For isolation, cloning, and manipulation of nucleic acids, the methods used were those in Hopwood et al. (23) for Streptomyces and Maniatis et al. (24) for E. coli.
DNA sequencing began at BamHI site 14 (Fig. l ) , adjacent to the act111 gene (20), and extended rightward 5.3 kb to include the actVB region (to a BglI site close to position 19.2). Computer analysis of the DNA sequence, using CODONPRE-FERENCE, revealed a set of six ORFs (Fig. 3) which were named (from left to right) ORF1-6, respectively. All of them are oriented from left to right.
The translation start point for each ORF was tentatively located (Fig. 2) using the following criteria: (a) distribution of GC content in the third position of the codons (30); ( b ) codon usage (for this purpose a table of codon usage from 64 different Streptomyces sequenced genes was used3); and (c) location of a potential ribosome binding site (RBS), based on reasonable complementarity to the 3'-end of the 16 S ribosomal RNA sequence, immediately upstream of the region where the two previous criteria were satisfactory. In nearly all cases the choice of start codon was unambiguous, and four ORFs (ORF1, ORF3, ORF4, and ORF6) are preceded by a good RBS. The choice of start codon for ORFl was given special consideration because a TTG (nucleotides 107-109) appeared to be the most likely candidate. Although unusual as a start codon, TTG is used in a low proportion of genes in some microorganisms (31), including one putative example in Streptomyces (32) in spite of its high GC content. The translation product starting at this TTG (preceded by a good RBS) would be a polypeptide of about the same size as, and aligning very closely with, two highly similar PKS gene products in the gra (33) and tcm (34) clusters (see below). The nearest inframe potential start codon upstream of this TTG, a GTG, would lie within the adjacent BamHI fragment (sites 13-14 in Fig. 1) coding for the act111 gene (nucleotides 8-6 of the complementary strand) (20) and would give a polypeptide 36 amino acids longer than the corresponding gra and tcm gene products. Moreover, it would lack a good RBS and have a long N-terminal run of unusual codons. Even further upstream (nucleotides 27-29), another GTG is preceded by a relatively good RBS but also suffers from the other objections. Both these potential GTG starts seem highly improbable. The lengths of the six ORFs in amino acid residues, and the corresponding molecular weights, would be: ORF1,424 amino acids (Mr 45,034); ORF2,407 amino acids (M, 42,523); ORF3, 86 amino acids (Mr 9,242); ORF4, 316 amino acids (Mr 34,643); ORF5, 297 ( M , 31,967); ORF6, 177 amino acids (Mr 18,381). Three pairs of ORFs (ORF1/2, ORF3/4, and ORF4/ 5) appear to be "translationally coupled" (35,36), overlapping by 4 bp in each case. There is a 25-bp untranslated segment between ORF2 and ORF3, and an 11-bp gap between ORF5 and ORF6.
An interesting polymorphism, at nucleotide 4665, was found in ORF5. It was previously shown (17) that the S. coelicolor strain from which the act DNA was derived, M145 (16), lacks BglII site 19, which is present in another S. coelicolor A3 (2) derivative, 51501 (as well as in Streptomyces liuidans). Site 19 was picked up by homologous recombination on various subclones of the act DNA which had been passaged through 51501. In strain 51501 the hexanucleotide starting at this position is AGATCT, giving BglII site 19, whereas in strain M145 the sequence is CGATCT. The ORF5 product in 51501 would have a glutamic acid residue in this region of the protein, while M145 would have alanine instead.
The 3'-end of ORF6 seems to represent the limit of the act gene cluster; 21 nucleotides further downstream of ORF6 is the end of another ORF, as deduced from computer analygis, which appears to run in the opposite direction to that of ORF6.'

Characterization of the act Early and Intermediate Region
Definition of the Limits of the act Early and Intermediate Region: An Operon Carrying Six act Genes-In order to find out if DNA beyond ORF6 was involved in actinorhodin biosynthesis, pMMl was constructed ( Fig. 1). No blue colonies (indicating restitution of actinorhodin production) were obtained when strains B135 and B185 (actVB mutants), carrying the rightmost of the known act mutations, were transformed M. J. Bibb, personal communication. vector as follows, pIJ2925, e pSU20, pUC19, then rescued and ligated to Streptomyces vectors as indicated.
with pMM1. To confirm that this result was really due to p M M l not including the actVB gene, rather than a low level of gene products, the same fragment was cloned on the high copy number vector pIJ702 to yield pMM4, which gave the same result. We therefore conclude that the region to the right of SphI site 19.2 does not code for any act early and intermediate genes. In addition, using a set of overlapping DNA fragments, which covered the whole sequenced region from BamHI site 14 to BglII site 21, as probe in S1 nuclease protection experiments, a transcript of nearly 5.5 kb was identified, which covers the region coding for the actI, VII, IV, and VB genes (data not shown). The limits of the S1

BglII 19
P w I I * * * * * * ~" " "~~+~~" " "~+~~~~~~~~~+ " " " " ----+ -----" " +~~"~~~~~+ "~" " " + " " " " -+ mRNA. Insertional inactivation with a BglII fragment (sites 16 to 19), which covers ORFs 2-5, was previously reported (17), in good agreement with our conclusions. (Note also that there are untranslated nucleotides only between ORFs 2 and 3 and ORFs 5 and 6: see above.) Identification of the act Early and Intermediate Loci-To try to assign act mutant classes to specific ORFs, we cloned a set of overlapping fragments ( Fig. 1 and Table 11), either in att-0C31 derivatives or in pIJ702, pIJ486, and pIJ941 plasmids and introduced them by transduction or transformation into several act mutants. For the phages, restoration of actinorhodin production (blue color) needed a single crossover to introduce the wild-type allele carried on the phage vector into the chromosome; this occurs when either the end or the beginning of the transcript, including the promoter, is present in the cloned fragment and when the crossover takes place between the 5'-end of the mRNA and the mutation point. Thus, the percentage of wild type (blue) lysogens would reflect the distance between the mutation and the 5'-end of the cloned DNA. (Recombination downstream of the mutation, between it and the 3'-end of mRNA, would restore functional genes only if an efficient promoter from the phage can transcribe stably the genes located downstream of it; an orientation-dependent phenotype would indicate this situation (17).
In order to associate specific ORFs with the act mutations, phages 0ME15,0ME17,0AB19, and 0G7 were made ( Fig. 1 and Table 11). @ME15 carries 65% of ORF2, complete ORF3, ORF4, and ORF5, and an almost complete ORF6 (lacking the three C-terminal codons). @ME17 contains 16% of ORF4 and complete ORF5 and ORF6, plus 500 bp to the right. @AB19 harbors less than 10% of ORF3, the whole of ORF4, ORF5, and ORF6, and 1.5 kb to the right. 0G7 carries less than 6% of the C terminus of ORF5, a complete ORF6, and approximately 2.5 kb to its right. To locate unambiguously the actVB mutations plasmids pIJ2350, pMM4, and pMM5 were constructed carrying the 3kb BglII fragment (sites 19-21 in Fig. 1) in both orientations: in ~152350, ORF6 is oriented in the same direction as the me1 gene, while in pMM5 ORF6 and me1 are divergent. pMM4 is the same as pIJ2350 but lacks the 600-bp SphI fragment which harbors the whole of ORF6.
The above recombinant plasmids and phages were introduced by transformation or transduction into different act mutants. The results (summarized in Table 111) enabled us to locate act mutations representing each of the phenotypic classes as follows.
The actIV mutations (act-112 and act-131) seem to lie in ORF5, in accord with the frequency of complementation observed with 0ME15,@ME17,@AB19, and 0G7: the closer the mutation is to the 5'-end of the cloned fragment the lower is the frequency of blue lysogens. Thus, the failure to detect complementation of B31 (act-131) with @ME17 placed the mutation either close to the N terminus of the ORF5 protein (or conceivably in the C terminus of ORF4), while in TK16 (whose complementation frequency is higher), the act mutation (act-11 7) must be located nearer the C terminus of ORF5.
The actVB mutations were localized within ORF6 because of trans-complementation using pIJ2350, but not by pMM1, pMM4, and pMM5, and by the frequencies of blue colonies obtained using the set of four phages. Complementation of in S. coelicolor A3 (2) (+) and no complementation by (-). For actIV and actVB mutants "complemented" by phages, the results are given as the percentage of positive (blue) lysogens out of the total of -2000. Roman numbers in brackets refer to the act mutant classes. ' NT, not tested.
actVB mutants with pIJ2350 but not with pMM5 suggests that ORF6 might be transcribed from a vector promoter, while neither pMM4 nor pMMl are able to restore the wild type phenotype because they lack ORF6. Moreover, the only overlapping region between 0G7 and @ME15 is precisely ORF6, confirming that ORF6 is responsible of actVB complementation. The frequencies of blue lysogens observed with B135 and B185 (actVB mutants) when lysogenized with @ME17 are the same as for TK16, suggesting that act-235 and act-285 might be located close to the 5'-end of the ORF6 protein.
The experiments leading to the identification of ORF4 as actVII were described elsewhere (38).
In order to localize the positions of the actI mutations, the available actI mutants were transformed with recombinant plasmids and phages carrying different combinations of ORFs 1-3. p M M l l (containing only ORFl complete) and pMM13 (ORFl,ORF2, and ORF3) were constructed on the high copy number vector pIJ486. None of the actI mutants transformed with them yielded an Act+ phenotype. Moreover, when the wild type S. coelicolor 51501 was transformed with p M M l l or pMM13 an Act-phenotype was observed. These results strongly suggest that the Act-phenotype observed with those constructions might well be due to a titration by the cloned fragments of some trans-acting element needed for expression of the act genes and thus presumably preventing expression of other act transcripts; a good candidate for such a trunsacting element is the actII-ORF4 gene product whose correct translation is essential for transcription of the actI gene (18). Furthermore, when p M M l l was introduced into S. coelicolor 51501, previously lysogenized with a hygromycin-resistant/ thiostrepton-sensitive phage carrying actIIORF4, and therefore with one extra copy of the activator gene, the colonies selected (tsr/hyg) overcame the mutagenic effect. In addition, the failure of complementation with high copy number plasmids contrasted with the success in using the low copy plasmid pIJ941 or "complementation" using att-phages. The low copy number recombinant plasmids were pMM14 (carrying ORFl complete), pMM15 (ORF1 and 2 complete), and pMM16 (ORF1, 2, and 3 complete); and the phages were 0L22 (containing complete ORFs 1, 2, and 3) and @AB23 (carrying complete ORFs 1 and 2, without any ORF3 sequences), both of them starting at SphI site 13.4 (Fig. 1).
Table I11 (a) shows the results obtained with the different constructions in the complementation of the actI mutants. Four of the 12 actI mutations were complemented by ORFl alone, and therefore lie in this ORF. The other eight were not complemented by ORFl and are therefore candidates for ORF2 mutations (although we cannot exclude the possibility that some of them might be deletions involving ORFl and ORF2 or polar mutations in ORF1). The absence of any within ORF3 may be due to the fact that ORF3 represented a small DNA target in relation to the whole actI region (less than 9% of ORFs 1, 2, and 3) when the wild type S. coelicolor was mutagenized (15).
In view of the data reported here, based on DNA sequencing and complementation of the act mutants, the conclusion previously reported by Bartel et al. (39) must be corrected as follows: the minimal act DNA fragment which caused alloesaponarin I1 production by Streptomyces galilueus strains (a XhoI fragment extending from a site to the left of the act region to XhoI site 17.1, cloned in pANT43) would have contained only ORFl and ORF2 complete; it would have lacked the 3'-end of ORF3 (a component of the act1 locus) and the whole of ORF4 (the &VI1 locus). Consequently the observed phenotype must have been the result of heterologous interaction between the ORFl and/or ORF2 gene products from the act DNA and some endogenous components of S. galilaeus in the recombinant cultures.
Deduced Biochemical Functions of act Early and Intermediate Genes-To better understand the functions of each of the ORFs revealed by DNA sequencing, we searched the available data bases with their translated products. ORF1, ORF2, and ORF3 show strong similarities with other genes whose products are known to be components of either fatty acid or polyketide synthases of different organisms. The resemblances are strongest (and are end-to-end) with the gra and tcm ORF1, 2, and 3 gene products (33,34) (between 55 and 80% identity) and less extensive, although significant, with other gene products like 6-methylsalicylic acid synthase from Penicillium patulum f40), Fatty acid synthase from rat ( 5 ) , and 6-deoxyerythronolide-B synthases from Smcharopolyspom erythraea (41, 42) (between 25 and 29% identity and 46 and 51% similarity). From these similarities, two main catalytic domains can be postulated within ORF1. These are: a ~-ketoacyl-synthase (condensing enzyme) (Fig. 4) with significant alignments around the Cys residue, the presumptive active site of the synthase, with those in other Type I or Type I1 fatty acid and polyketide synthases. A second feature is around a Ser residue in a GHS motif which could represent a putative acyl transferase domain. This would imply that ORFl codes for a bifunctional ketosynthase/acyl transferase and could reflect the lack of any identifiable separate acyltransferase(s) within the cluster. As in thegra and tern clusters (33, 34), actI-ORF1 and actI-ORF2 strongly resemble each other (29% identity and 49% similarity), and their stop and start codons overlap as in many other bacterial operons (35,36). This suggests a translational coupling which probably ensures equimolar production of the two gene products. There is no sign of the corresponding active site domains in ORF2. Therefore, as previously suggested for the corresponding genes in the granaticin and ~tracenomycin PKS (33,34), we suggest that the ORF2 product may perhaps function, together with the ORFl product, as a heterodimeric protein.
The ORF3 product is a small polypeptide which, from its resemblance to other known proteins, would function as an acyl-carrier protein (ACP); a typical phosphopantetheine binding domain can be identified centered on the so-called "active-site" Ser, which is well conserved in other FAS and P-keto synthase motif.
Overlapping the TGA stop codon of ORF4 is an ATG (nucleotides 3830-3832); this start codon is not preceded by a typical RBS but is the first one after a typical change in the distribution of rare codons between reading frames (Fig. 3). We postulate that the beginning of ORF5 is this ATG, rather than the more upstream GTG (nucleotides 3419-3421) within ORF4. Searching the available data bases with the ORF5translated product revealed strong similarities with the products of cphA (45) (55% similarity and 27% identity over 232 overlapping amino acids), cfiA (46) (49% similarity and 26% identity over 239 overlapping amino acids) and ccrA (47) (49% similarity and 26% identity over 239 overlapping amino acids) (see Fig. 6 for Dotplots and alignments). These three proteins are Type I1 /3-lactamases. There are less extensive similarities with other p-lactamases belonging to the same metallothioprotein group such as those from Bacillus cereus 569/H (48) and B. cereus 5/B/6 (49). The corresponding regions are in a domain thought to be important for Zn2' binding: 2 closely located His residues, 1 Cys, and a further His near the C terminus of the proteins (48. 49). The two regions around the His residues can be aligned together, while the region next to the Cys residue of the Type I1 /3-lactamases, although very similar to the region in the ORF5 product, lies immediately upstream of the Cys rather than being centered on it. Nevertheless, a Cys residue is located within this domain of ORF5 (amino acids 143, Fig. 6), whose counterpart in the Type I1 p-lactamases is a Tyr residue.
It is not at all obvious why the actIV gene product, which has been postulated (39) to act as a dehydrase that would remove a hydroxyl group from C-5 after the two carbon rings of the isochromanequinone are established (Fig. 5), should resemble 8-lactamases, which are believed to be specific for hydrolysis of the #?-lactam ring system; such a reaction cannot rationally be implicated in a biosynthetic scheme for actinorhodin. However, perhaps it is significant that the segments of the p-lactamases that resemble the ActIV protein are in regions thought to be important for Zn2+ binding, which include 2 closely located His residues (only 1 in CphA), a Cys, and a further His near the C terminus of the proteins (48). Corresponding residues are seen in the ActIV protein, except that the Cys is replaced by an aspartic acid residue. In some other Zn2+ proteins, 1 acidic residue acts as one of the three required ligands for the metal (50-52). Thus, perhaps the significance of the resemblance with the p-lactamases is a common evolutionary origin as Zn2+ enzyme, rather than a relationship between the present day enzymatic activities of the proteins, with the ActIV protein having acquired a role as a dehydrase. Interestingly, the homologies between the ORF5 product and the 8-lactamases lie immediately downstream of the processing site of the pre-@-lactamases (46,48,49); this is consistent with the presumed intracellular location of the ActIV protein (as a biosynthetic enzyme), in contrast to the extracellular location of the p-lactamases.
Searching the available data bases with the sequence of the putative ORF6 product failed to show any significant similarity with other proteins. Cole et al. (29) argued that the I Acelate biosynthetic block caused by the actVB mutations, here shown to lie in ORF6, i s probably in the dimerization reaction which would join two molecules of a late precursor of actinorhodin to produce the final antibiotic structure. Little is known about such phenolic oxidative coupling reactions in biological systems and so the further study of the ORF6 gene product may provide important insights into this class of bioorganic reactions. The sequencing strategy is outlined in Fig. 7.

CONCLUSION
This paper has described a set of five cotranscribed genes. Four of them (ad-ORF1, 2, 3, and actVII) code for components of the polyketide synthase that assembles and cyclizes the carbon backbone of the actinorhodin half-molecule. A fifth gene, act111 (20), which encodes the ketoreductase of the PKS, is adjacent to the actI/VII genes but, perhaps surprisingly, is transcribed independently from them, in the opposite direction, and from a promoter that appears to represent a different class from the promoter for the actI/VII transcript. This transcript also carries the gene (actIV) that appears to catalyze the next step in biosynthesis after the PKS but, again unexpectedly, continues on to include the actVB gene whose putative role, as a dimerase for the immediate precursors of the actinorhodin half molecule, acts considerable later in the pathway. Clearly, many intriguing features remain to be elucidated about the org~ization of the act cluster, which may be the first actinomycete antibiotic gene cluster to have been completely sequenced.