A relA / spoT Homologous Gene from Streptomyces coelicolor A3(2) Controls Antibiotic Biosynthetic Genes*

A 0.972-kilobase pair DNA fragment from Streptomy- ces lividans that induces the production of the blue-pigmented antibiotic actinorhodine in S. lividans when cloned on a multicopy plasmid has led to the isolation of a 4-kilobase pair DNA fragment from Streptomyces coeli- color containing homologous sequence. Computer-as-sisted analysis of the DNA sequence revealed three pu- tative open reading frames (ORFs), ORF1, ORF2, and ORF3. ORF2 extends beyond the sequenced DNA frag- ment, and its deduced product shares no similarities with any other known proteins in the data bases. ORF3 is also truncated, and its 41-amino acid C-terminal product is identical to the S. coelicolor adenine phosphori- bosyltransferase. The 847-amino acid ORF1 protein, with a predicted molecular mass of 94.2 kDa, strongly resembled the relA and spoT gene products from Escherichia coli and the homologs from Vibrio sp. strain S14, Haemophilus influenzae , Streptococcus equisimilis H46A, and Mycoplasma genitalium . Unlike these proteins, the ORF1 amino acid sequence analysis revealed the presence of a putative ATP/GTP-binding domain. A mutant was generated by deleting most of the ORF1 gene that showed an actinorhodine-nonproducing

A 0.972-kilobase pair DNA fragment from Streptomyces lividans that induces the production of the bluepigmented antibiotic actinorhodine in S. lividans when cloned on a multicopy plasmid has led to the isolation of a 4-kilobase pair DNA fragment from Streptomyces coelicolor containing homologous sequence. Computer-assisted analysis of the DNA sequence revealed three putative open reading frames (ORFs), ORF1, ORF2, and ORF3. ORF2 extends beyond the sequenced DNA fragment, and its deduced product shares no similarities with any other known proteins in the data bases. ORF3 is also truncated, and its 41-amino acid C-terminal product is identical to the S. coelicolor adenine phosphoribosyltransferase. The 847-amino acid ORF1 protein, with a predicted molecular mass of 94.2 kDa, strongly resembled the relA and spoT gene products from Escherichia coli and the homologs from Vibrio sp. strain S14, Haemophilus influenzae, Streptococcus equisimilis H46A, and Mycoplasma genitalium. Unlike these proteins, the ORF1 amino acid sequence analysis revealed the presence of a putative ATP/GTP-binding domain. A mutant was generated by deleting most of the ORF1 gene that showed an actinorhodine-nonproducing phenotype, while undecylprodigiosin and the calcium-dependent antibiotic were unaffected. The mutant strain grew at a much lower rate than the wild-type strain, and spore formation was delayed. When the gene was propagated on a low copy number vector, not only was actinorhodine production restored, but actinorhodine and undecylprodigiosin production was enhanced in both the mutant and wild-type strains and morphological differentiation returned to wild-type characteristics. (p)ppGpp synthetase activity was not detected in purified ribosomes from the ORF1-deleted mutant, while it was restored by complementation of this strain.
Streptomyces species have a complex cell life cycle that involves morphological and biochemical differentiation. Antibiotic production is usually initiated at the transition between vegetative growth and the development of the spore-bearing aerial mycelium (1,2), suggesting that there may be a close relationship between both processes. During this developmental regulation, antibiotic biosynthesis is controlled by a series of metabolites and regulatory gene products operating at different levels, with their final target being the structural genes of the antibiotic pathway.
In S. coelicolor, several so-called pleiotropic genes outside the biosynthetic clusters have been implicated in the regulation of the multiple antibiotic pathways; mutations in absA (14) and absB (15) completely abolish the biosynthesis of all four antibiotics, from which the production of both pigmented antibiotics is restored by the afsQ1-afsQ2 gene pair in absA but not absB mutants (16). Neither actinorhodine nor undecylprodigiosin (as well as reduced amounts of methylenomycin and CDA) could be detected in afsB mutants (17), which were suppressed by the afsR gene (18,19). In abaA-ORFB (20) mutants, actinorhodine and undecylprodigiosin production is blocked, while the level of CDA is reduced and methylenomycin remains unaffected. The bld (bldA-D and bldF-G) (21,22) genes have been described as being required for both antibiotic production and aerial mycelium formation. The bldA gene encodes a leucyl-tRNA (23) that recognizes the UUA codon (extremely rare in Streptomyces mRNA because of the high G ϩ C content of their DNA), and the suggestion has been made that this gene might constitute a translational regulatory mechanism controlling sporulation genes and some antibiotic pathways (9,24).
The stringent response and (p)ppGpp formation have been extensively studied in Escherichia coli (25,26). These polyphosphorylated nucleotides are synthesized by at least two possible routes. The main one is attributed to the (p)ppGpp synthetase I activity, which is encoded by the relA gene and operates on ribosomes under amino acid deprivation when codon-specified uncharged tRNAs are bound to the ribosomal acceptor site (27). The reaction involves a pyrophosphoryl transfer from ATP to GTP or GDP. The transient (p)ppGpp accumulation leads to complex regulatory adjustments such as a reduction in stable RNA transcription rate (28 -32) and an increase in expression of certain amino acid operons (33,34). Defective ribosomal (p)ppGpp synthesis was observed in relC mutants, which have an altered L11 protein in the 50 S ribosomal unit, the same subunit implicated in the binding of the RelA protein (35,36). A putative relC mutant of S. coelicolor has been isolated (37). The strain is deficient in the production of actinorhodine and undecylprodigiosin as well as in its ability to form aerial mycelium. In this relaxed mutant, there is a 10-fold reduction of (p)ppGpp upon amino acid starvation when compared with the parental strain. Based on this observation and the isolation and characterization of relaxed mutants from other Streptomyces species (38 -41) and Bacillus subtilis (42), a correlation was suggested between the stringent response at either the onset of secondary metabolism because of (p)ppGpp formation or morphological differentiation due to the reduction of the intracellular GTP level (40). The second route for (p)ppGpp synthesis in E. coli, mediated by a ribosome-independent enzyme, (p)ppGpp synthetase II activity (spoT gene product) (43), is deduced to occur because during carbon and energy source deprivation, (p)ppGpp does accumulate in relA-deleted strains (44), while it is no longer detectable in strains carrying deletions of both the relA and spoT genes (45).
Recently, a (p)ppGpp synthetase from Streptomyces antibioticus has been purified and characterized (46,47) and shown to possess differential catalytic properties, which raises the possibility that the reported enzyme would not represent the analog of the RelA (or SpoT) protein from E. coli; nevertheless, the presence in S. antibioticus of a pathway similar to that of relA in E. coli for (p)ppGpp synthesis could not be excluded. The isolation of relC mutants in S. antibioticus (41,48) supports this hypothesis.
We report here the isolation and characterization of a new gene that strongly resembles relA and spoT and that is implicated in the regulation of antibiotic production in S. coelicolor.
Media, Culture Conditions, and Microbiological Procedures-E. coli strains were grown on L agar or in L broth, supplemented with 0.2% maltose and 10 mM MgSO 4 when necessary (60). Streptomyces manipulations were as described previously (61). Thiostrepton (Sigma, catalog No. T-8902) was used at a concentration of 50 g/ml in agar medium and 10 g/ml in broth cultures. Hygromycin B (Sigma, catalog No. H-2638) was used at 200 and 50 g/ml in solid and liquid media, respectively.
DNA Sequencing-DNA sequencing was done by the dideoxy chain termination method (62); DNA sequence was determined from both strands, routinely using the 7-deaza-dGTP reagent kit from U. S. Biochemical Corp. (catalog No. 70750) following the manufacturer's recommendations. Convenient DNA fragments were previously cloned in either M13mp18 or M13mp19 vectors using suitable restriction fragments.
Computer Analysis of Sequences-The DNA sequence was analyzed using the software programs of the University of Wisconsin Genetics Computer Group (Version 8.0-AXP) (63). Analysis of open reading frames was done using CODONPREFERENCE with a codon usage table made from 100 Streptomyces genes (64); comparisons of sequences were made against the EMBL nucleic acid data base (daily updated) and the SwissProt data base (weekly updated), using FASTA, TFASTA, and BESTFIT. Protein alignments were made using PILEUP from the same package and displayed using PRETTYBOX from the Extended Genetics Computer Group package (65). Gene Disruption and Deletion-For insertional inactivation, an internal ORF fragment was cloned into the C31 derivative PM1 vector, and the resulting recombinant phage was used to lysogenize the S. coelicolor J1501 strain by insert-directed recombination (66). To obtain the deleted mutant, fragments flanking the region to be deleted were cloned into the PM1 vector, and the recombinant phage was used to lysogenize the wild-type strain by insert-directed recombination. The resulting lysogens were first isolated as a single colony and later allowed to grow without selection in order to obtain the double recombinants. The deletion was confirmed by Southern blot analysis (67).
DNA and RNA Manipulations-Isolation, cloning, and manipulation of nucleic acids were as described previously for Streptomyces (61) and E. coli (60). For constructing the S. lividans DNA library, the chromosomal DNA was totally digested with BamHI, and the DNA fragments were cloned into the BamHI site of pIJ486. S. coelicolor A3(2) J802 total DNA was partially digested with Sau3AI. The resulting fragments were fractionated by centrifugation in a 10 -40% sucrose gradient as described (60), and 15-20-kb DNA fragments were pooled and cloned into the BamHI site of EMBL4 phage; the recombinant phages were packaged in vitro using the EMBL4 system (Stratagene, catalog No. 242201) according to the manufacturer's recommendations. The library was probed using a DNA fragment labeled by the polymerase chain reaction (68) with Thermostase (Linus, catalog No. MB014) following the manufacturer's recommendations.
For high resolution S1 mapping, the method of Murray (69) was used. For actI-ORF1 (70), a 798-base pair SphI-SacI fragment (from positions 13.4 to 14.1) containing the actI-ORF1 promoter region uniquely labeled at the 5Ј-end of the SacI site within the actI-ORF1 coding region was used as probe. For actVI-ORF1 (71), a 847-base pair KpnI-BssHII fragment (nucleotides 1406 -2252) that contained the actVI-ORF1 promoter region labeled at the 5Ј-end of the BssHII site within the internal actVI-ORF1 coding region was used. For actII-ORF4 (9), the actII-ORF4 promoter region included in a 635-base pair fragment (nucleotides 4824 -5458) was uniquely labeled at the 5Ј-end of the XhoI site (nucleotide 5458) within the actII-ORF4 coding region and was used as probe. RNA was extracted as described (61) from 3-day-old mycelium grown on the surface of cellophane discs on R5 agar plates as described (72).
Antibiotic Production-Actinorhodine and undecylprodigiosin were isolated from 6-day-old S. coelicolor mycelia grown on R5 solid medium over cellophane discs. For the blue pigment extraction, both mycelium and agar medium were separately processed; 10 and 40 ml of water were added to the fragmented agar and the scraped mycelium, respectively. The final pH was then increased to 10 with 1 N NaOH, and diffusion of the blue pigments was allowed for at least 2 h at 4°C. After 20 min of centrifugation at 25,000 ϫ g at 4°C, the absorption of both supernatants was measured at 610 nm. Results are presented as total actinorhodine produced. Undecylprodigiosin was extracted from the scraped mycelium with 10 ml of methanol acidified with 1 N HCl and estimated spectrophotometrically at 530 nm. Blank correction was made using S. coelicolor MAFM0195 for actinorhodine and S. coelicolor CM01 for undecylprodigiosin. Antibiotic valorations are expressed as A units/g of obtained mycelia, wet weight. CDA production was tested as described (73). At least three independent colonies were analyzed, and determinations were repeated twice.
Ribosome Purification-Ribosomes from E. coli and S. coelicolor were isolated essentially as describe previously (74). Either Streptomyces mycelia or E. coli cells (2-5 g, wet weight) were used. The clear ribosomal pellet was slowly resuspended in buffer A (50 mM Tris acetate, pH 8, 15 mM magnesium acetate, 60 mM potassium acetate, 30 mM NH 4 acetate, 1 mM dithiothreitol, 0.2 mM EDTA, 0.5 mM phenylmethylsulfonyl fluoride) at 4°C. Ribosomal concentration was estimated spectrophotometrically at 260 nm and is expressed as A units/ml. Protein was measured as described (75) using bovine serum albumin as standard.
Measurement of (p)ppGpp Synthesis-Standard ATP:GTP 5Ј-pyrophosphotransferase assays were done in a final volume of 50 l containing 2 mM ATP, 1.3 mM GTP, 10 Ci/ml [␣-32 P]GTP (3000 Ci/mmol; Amersham Corp.), and 60 A units of ribosomes/ml in buffer A. The reactions were allowed to proceed for 45 min at 30°C and stopped by the addition of 2 l of 88% formic acid. After the removal of precipitated 2 M. A. Ferná ndez-Moreno, personal communication. (p)ppGpp Synthetase from S. coelicolor protein by centrifugation for 2 min at 9000 ϫ g, 2.5 l of the resulting supernatants was spotted onto polyethyleneimine cellulose thin-layer plates. Plates were developed as described by Cashel (76), and reaction products were identified by autoradiography. The migration position of ppGpp was confirmed by comparison with the authentic compound (generously supplied by K. Ochi) and with the reaction products, pppGpp and ppGpp, of isolated ribosomes from E. coli JM101 either alone or carrying extra copies of the relA gene (pGG21) (27). For quantification of the reaction products, portions of the plates corresponding to the equivalent migrating position of the (p)ppGpp compounds were cut and subjected to liquid scintillation counting. Corrections were made by subtracting the counts obtained from pppGpp and ppGpp regions following chromatography of a reaction mixture containing only GTP as substrate. The possible synthesis of polyphosphorylated nucleotides by a nucleotide 3Ј-pyrophosphokinase was tested by incubating the ribosomal fraction with either 1.3 mM [␣-32 P]GTP (7.6 mCi/mmol) or 2 mM [␥-32 P]ATP (2.5 mCi/mmol) alone.

RESULTS
Cloning and Sequencing of the DNA That Activates Actinorhodine Production in S. lividans-S. lividans, a streptomycete closely related to S. coelicolor, has the whole genetic information for actinorhodine biosynthesis; however, this antibiotic is not produced under routine laboratory conditions. By introducing extra copies of the actII-ORF4 gene (or other pleiotropically acting genes) into S. lividans, actinorhodine is synthesized (9,19,20,77). Since these trans-acting elements will give some understanding of the mechanisms that are involved in antibiotic production, we are interested in their isolation and characterization. Thus, BamHI fragments of S. lividans strain TK21 chromosomal DNA were ligated into the BamHI site of the high copy number plasmid pIJ486. S. lividans TK21 was transformed with the ligation mixture, and recombinant clones were selected for thiostrepton resistance. Among the transformants, a blue colony was isolated, and its phenotype was confirmed by retransformation. The plasmid DNA (named pSCNB079) obtained from this colony revealed a single (unique) 1-kb BamHI insert in the cloning site of the vector. This 1-kb DNA fragment was subcloned in E. coli and sequenced. Computer-assisted analysis of ORFs using the program CODONPREFERENCE showed a continuous ORF; other putative ORFs were discarded as coding regions because their codon usage did not fit the Streptomyces pattern. The cloned fragment was presumably part of the internal coding region of a gene.
Southern blotting of BamHI-digested chromosomal DNA from S. lividans TK21 and S. coelicolor J1501, probed with pSCNB079, showed a single hybridizing band of the same intensity in both strains of 1 kb in S. lividans and of 3.3 kb in S. coelicolor. This confirmed that the sequenced DNA was not the result of rearrangement during the cloning experiments and indicated the presence of a similar gene in S. coelicolor.
Cloning and Sequencing of the S. lividans Homologous Gene from S. coelicolor-From a genomic library of S. coelicolor, three different recombinant clones were isolated when probed with pSCNB079. All the recombinant phages carried a 3.3-kb BamHI hybridizing fragment. From one of these ( Fig. 1,  lambda 16.4), the 3.3-kb fragment and some of its adjacent region were subcloned and sequenced.

(p)ppGpp Synthetase from S. coelicolor
Computer-assisted analysis of the resulting 4-kb DNA sequence revealed three possible ORFs (Fig. 1), which were named ORF1, ORF2, and ORF3. No other putative ORF could be deduced by consideration of Streptomyces codon usage.
The most likely translation initiation codon for ORF1 is at position 311 (a TTG codon), as deduced by its overall distribution of GC content in the third position, the codon usage within ORF, and the presence of a good putative ribosome-binding site (GAG-GAG, nucleotides 300 -305) at an appropriate distance (82). The similarities observed between the putative ORF1 product and other known proteins (see below) were used as additional criteria. The stop codon (TAG) is located at position 2852.  p)ppGpp Synthetase from S. coelicolor ORF1 encodes a protein of 847 amino acids with a predicted molecular mass of 94.2 kDa. Comparison of the ORF1 product showed a strong resemblance to the following proteins: RelA from E. coli (38% identity, 61% similarity) (27), RelA from Vibrio sp. strain S14 (38% identity, 60% similarity) (78), SpoT from E. coli (43% identity, 62% similarity) (43), RelA from H. influenzae (37% identity, 60% similarity) (79), SpoT from H. influenzae (38% identity, 60% similarity) (79), the Rel-like proteins from S. equisimilis H46A (40% identity, 63% similarity) (80) and from M. genitalium (25% identity, 50% similarity) (81), and a putative SpoT from M. leprae cosmid B1177 (62% identity, 77% similarity). 3 Additionally, the 166-amino acid N-terminal ORF1 product was shown to be almost identical to the translated DNA extreme region near the secD and secF genes from S. coelicolor A3(2). 4 There is a conserved mismatch in amino acid sequence at position 24, alanine instead of proline, due to a change of a guanine to a cytosine nucleotide in the DNA sequence. This difference could be attributed to the different strains used. The N-terminal region of the ORF1 protein is ϳ90 amino acid residues longer than the homologous ones. Six nucleotides were different between S. lividans and S. coelicolor within the original fragment, while the corresponding products were almost identical with only a conserved change (leucine instead of valine at position 197).
The amino acid sequence of the ORF1 protein reveals a particularly well conserved ATP/GTP-binding domain (amino acids 458 -465). This sequence motif, (A/G)XXXXGK(S/T), generally referred to as the "A" consensus sequence (83) or the "P-loop" (84), is not present in RelA or SpoT proteins (27, 43, 78 -81) and represents a gap in these proteins when aligned with the ORF1 product (Fig. 2).
The second ORF (ORF2, nucleotides 2933-4066) extends beyond the sequenced DNA. Comparison of its 378-amino acid C-terminal product with protein sequences contained in data bases gave no similarities to other known proteins and therefore no clue as to its possible function.
The third ORF (ORF3, nucleotides 2-127) is incomplete, and translation of this short DNA sequence was shown to be identical to that of the 41-amino acid C terminus of the adenine phosphoribosyltransferase from S. coelicolor reported in the data base. 4 Based on the observed similarities, we infer that the DNA sequence reported here is adjacent to the secD and secF region.
Implication of the ORF1 Gene in Antibiotic Production-To explore the possible role of the ORF1 product in antibiotic production in S. coelicolor, mutants were generated by either insertional inactivation or chromosomal deletion of ORF1. To disrupt ORF1, the original 1-kb BamHI fragment from S. lividans was cloned in the BglII site of PM1 (in both orientations) and used to lysogenize S. coelicolor J1501. That ORF1 had indeed been interrupted was confirmed by Southern blot analysis of appropriately digested total DNA from four lysogens carrying the recombinant prophage inserted in one orientation and four in the opposite. No obvious phenotypic differences were observed between any of these and the parental strain.
To delete ORF1, a clone was constructed by sequentially ligating the 0.805-kb Sau3AI-BamHI fragment (nucleotides 1-805) and the 1.429-kb XhoI fragment (nucleotides 2567-3995) in the same relative orientation as in the chromosome into the BamHI and SalI sites of E. coli vector pIJ2925, respectively; the resulting fragment, carrying the intended deletion, was rescued by digestion with BglII and ligated to the PM1 vector previously digested with BglII and BamHI, which re-places the thiostrepton resistance marker with the recombinant fragment. Insertion of the phage through one of the flanking fragments was confirmed by Southern blotting. One of the lysogens was spread on agar plates without selection, and spores from this first unselected round were analyzed on plates for hygromycin sensitivity. The hygromycin-sensitive colonies are expected either to carry the internal deletion, after double crossover with the prophage fragment, or to have simply lost the prophage from the chromosome. Six different hygromycinsensitive colonies were analyzed by Southern blotting in order to determine their chromosomal structure. Three of them were shown to carry the expected physical deletion; two still contained the prophage (their sensitivity may be the result of the generation of a mutation on the hygromycin resistance gene); and the last gave the same pattern as the wild type. These deleted mutants (named 18J strain) were shown to grow slower than and not to sporulate as well as the wild type. Actinorhodine production was almost abolished, while undecylprodigiosin and CDA production was little affected (Table I). Normal sporulation rate and actinorhodine biosynthesis were restored by introducing plasmid pSCNB080, which contained the complete region (Fig. 1). However, the amount of both pigmented antibiotics seemed to be higher in both S. coelicolor strains 18J and J1501 when transformed with pSCNB080 (Table I). Interestingly, transformation of the 18J strain with actII-ORF4 in plasmid pPAS4 led to actinorhodine production (see below), without affecting the low growth rate and the deficiency in sporulation (data not shown).
Transcriptional Analysis of the act Cluster-As almost no actinorhodine was detected in the ORF1-deleted strain, it was of interest to analyze the transcription of some of the act genes, such as actI-ORF1 (70), actVI-ORF1 (71), and the transcriptional regulator gene actII-ORF4 (9), by S1 mapping. No protected fragment was detected for either actI-ORF1 or actVI-ORF1 when RNA from the 18J mutant carrying the pPAS3 control plasmid was probed (Fig. 3, A and B); only a basal transcription of the actII-ORF4 gene was found (Fig. 3C), showing that transcription of the actinorhodine pathway-specific regulator gene is clearly affected. In contrast, S1-resistant fragments of the expected size ( Fig. 3C; see wild type) were present when RNA from the same strain containing actII-ORF4 in plasmid pPAS4 was used in the experiments, suggesting that the nonproducing phenotype was indeed due to a limitation in the transcription of the positive regulatory gene. As expected, the increase in ActII-ORF4 protein cause by the enhancement of its specific transcript level led to actI-ORF1 and actVI-ORF1 transcription (Fig. 3, A and B). (p)ppGpp Synthetase Activity Measurements-Due to the strong similarities between the ORF1 and RelA/SpoT proteins, experiments were set up to measure the (p)ppGpp synthetase activity in both the wild-type (J1501) and ORF1-deleted (18J) strains. Ribosomes from both strains harboring either the (p)ppGpp Synthetase from S. coelicolor pPAS3 or pSCNB080 plasmid and from E. coli JM101 with or without the pGG21 plasmid were obtained as described under "Experimental Procedures." The (p)ppGpp compounds formed in the reaction mixtures are shown in Fig. 4. Interestingly, no (p)ppGpp synthesis was observed with purified ribosomes from the mutant strain carrying the control plasmid, pPAS3 (Fig. 4, lane 4), while (p)ppGpp formation was almost undetectable when ribosomes were preincubated with 2 M thiostrepton, being unaffected by the addition of an equivalent amount of dimethyl sulfoxide (Fig. 4, lanes 9 and 8, respectively). No formation of polyphosphorylated nucleotides was detected when assayed in the presence of either GTP or ATP alone (data not shown). The (p)ppGpp formation in S. coelicolor was ϳ25-fold lower than that in E. coli JM101 (Table II), with a correlation between the copy number of the ORF1 gene and (p)ppGpp synthetic activity in the former strain (Table II). Interestingly, a ppGpp/pppGpp ratio of 3.22 was found in the reaction with ribosomes from the J1501 strain with the pPAS3 control plasmid, while values of 0.75 and 1.18 were observed with the 18J and J1501 strains harboring extra copies of the ORF1 gene, respectively. This difference was also observed with ribosomes from E. coli when compared with the ribosomes from the same strain carrying extra copies of the relA gene (pGG21) (ppGpp/ pppGpp ratios of 3.90 and 1.35, respectively).

DISCUSSION
By selecting a clone from S. lividans that stimulated actinorhodine production in S. lividans, we have isolated a gene in S. coelicolor that seems to be involved in the control of antibiotic biosynthesis in this bacterium. The ORF1-encoded 847amino acid protein strongly resembles a group of enzymes involved in the biosynthesis of (p)ppGpp compounds, being produced under stringent conditions and generally referred to as (p)ppGpp synthetases.
The RelA and SpoT proteins of E. coli have been studied in some detail (26), and RelA from Vibrio sp. strain S14 (78) and the (p)ppGpp synthetases from Bacillus sp. (85,86) and from S. antibioticus (46,47) have also been characterized. The predicted ORF1 product has a molecular mass of 94.2 kDa, which is close to that described for E. coli RelA (84,000 Da) and SpoT (79,000 Da) (27,43), for Vibrio RelA (84,500 Da) (78), for the ribosome-dependent (p)ppGpp synthetase from Bacillus stearothermophilus (86,000 Da) (86), and for the ribosome-independent (p)ppGpp synthetase I from S. antibioticus (88,000 Da) (46). Nevertheless, the ORF1 protein is ϳ90 amino acids longer at its N terminus than the homologs so far sequenced.
That the cloned ORF1 gene codes for a (p)ppGpp synthetase is also supported by the measurements of this activity in purified ribosomes from S. coelicolor. In this context, it should be emphasized that no (p)ppGpp formation was detected with ribosomes from the S. coelicolor ORF1-deleted mutant, while the activity was restored by complementation of the 18J strain.

(p)ppGpp Synthetase from S. coelicolor
Like the ribosome-dependent (p)ppGpp synthetase I (RelA) from E. coli (26), the enzyme from S. coelicolor was inhibited by thiostrepton (Table II), its activity was detected in the presence of 18% methanol, and Mg 2ϩ ions were absolutely required, not being replaced by Mn 2ϩ or Zn ϩ2 (data not shown). The increase in the relative proportion of ppGpp with respect to pppGpp in ribosomes from the J1501 strain harboring the control plasmid (pPAS3) ( Table II) when compared with either the 18J or J1501 strain carrying extra copies of the ORF1 gene (pSCNB080) might be consistent with the observation that ppGpp is synthesized from pppGpp in E. coli (87). Nevertheless, the direct pyrophosphoryl transfer from ATP to the GDP formed from GTP by ribosomal nucleotidases cannot yet be excluded and might also account for the observed differences.
An interesting feature of the ORF1-deduced product is the presence of a putative ATP/GTP-binding motif (83), (A/ G)XXXXGK(S/T), which has not been described in any other known protein related to (p)ppGpp metabolism. The presence of this conserved motif in the ORF1 protein is in agreement with its biochemical function because ATP and GTP are both substrates of the reaction catalyzed by the (p)ppGpp synthetase. Additionally, a consensus GTP-binding domain has been proposed by Dever et al. (88) to be composed of three conserved elements: (A/G)XXXXGK, DXX(A/G), and NKXD, with a spacing of either 40 -80 or 130 -170 amino acid residues between the first and second elements and of 40 -80 residues between the second and third elements. The first two elements are involved in interactions with the phosphate portion of the GTP molecule, and the last element is involved in nucleotide specificity (89). Although with a mismatch, this conserved GTPbinding fingerprint is observed in the ORF1 protein (amino acids 458 -464, APKSSGK; amino acids 513-516, DVIA; and amino acids 587-590, NKIR, with spacings of 48 and 70 amino acids, respectively), suggesting that it could be specifically involved in GTP binding. The deviation in the consensus sequence of the last element in the ORF1 protein (the aspartic acid of NKXD is replaced by arginine, NKIR) might be related to a lower affinity for GTP, as has been demonstrated by Feig et al. (90) for the p21 ras protein, in which the mutation of this residue to asparagine resulted in a reduction in affinity for GTP by a factor of 100 (K D ϭ 10 Ϫ8 to 10 Ϫ6 M). Thus, it is still reasonable to suggest that the ORF1 protein could have the capacity to bind GTP. Further biochemical studies of this (p)ppGpp synthetase as well as the characterization of mutations within the putative nucleotide-binding domain will be of particular interest for understanding its biochemical function and are currently in progress.
One out of three antibiotics produced by S. coelicolor, actinorhodine, was severally affected in the 18J strain. This dramatic reduction is due to a decrease in the specific mRNA level of the transcriptional activator gene of the act cluster (actII-ORF4). Surprisingly, undecylprodigiosin production is only slightly reduced in the 18J mutant, although both pathways are controlled by their respective positive regulators (8,9), which are very similar to each other. Nevertheless, a possible role for the ORF1 gene in the regulation of both pathways is suggested since extra copies of this gene resulted in an increase in both actinorhodine and undecylprodigiosin in both the J1501 and 18J strains. It is well known that several metabolites and regulatory genes operate at different points and in a particular mode on antibiotic biosynthesis, giving rise to a signaling within an intricate regulatory network. An alternatively acting signal or any other factor independent of the ORF1 mechanism could be sufficient to trigger undecylprodigiosin biosynthesis, in contrast to the ORF1 requirement for actinorhodine production. The fact that actinorhodine and undecylprodigiosin are not equally reduced in the ORF1-deleted mutant, while both of them are enhanced by extra copies of this gene, is an interesting observation, and the mechanism of these differences needs to be studied in more detail.
The production of actinorhodine in strain 18J could be restored by extra copies of either ORF1 or actII-ORF4. An effect similar to that observed with the actII-ORF4 gene has been described previously in absA and absB mutants (91,92). These data also support the suggestion that the ActII-ORF4 protein is by itself sufficient to activate transcription of the biosynthetic genes, and if any other additional factor is required, either it does not constitute a limitation or its action might be overtaken by the overproduced ActII-ORF4 protein (93). As expected, deficiencies in morphological differentiation of strain 18J carrying extra copies of actII-ORF4 cannot be complemented, and they are only completely restored in trans by the ORF1 gene.
We do not yet know why the disruptants in the ORF1 gene showed an apparently normal phenotype. A residual (p)ppGpp synthetase activity cannot be excluded in these lysogens. This question is of interest not only for defining putative functional peptides, but also for understanding the activation of actinorhodine production in S. lividans by the original BamHI fragment (internal region of the ORF1 gene), which allowed us to isolate the gene. In E. coli, the relA1 gene products (␣-and ␤-fragments) have been shown to complement each other in trans to yield some (p)ppGpp synthetic activity, while overexpressed RelA1 ␤-fragment abolishes (p)ppGpp formation of a relA ϩ strain, probably due to its competition with the wild-type gene product for ribosomal binding (44). Furthermore, a Cterminally truncated RelA protein is still active (94,95), although it then becomes relC-independent, unlike the wild-type protein (94). The differences in phenotype observed between the lysogens and the ORF1-deleted mutant might be interpreted in this context.
The involvement of (p)ppGpp in the stringent response has been studied in E. coli in some detail (25,26). Depletion of amino acids leads to the synthesis of these polyphosphorylated nucleotides by (p)ppGpp synthetase I activity, which apparently mediates several complex changes in gene expression, due to the inability of the cell to maintain sufficient aminoacylated tRNAs for the demands of protein synthesis. There are several reports that (p)ppGpp formation takes place during stringent response in several Streptomyces species (37-41, 96 -99). A relaxed (presumptively relC) mutant in S. coelicolor has been isolated (37) in which the onset of aerial mycelium formation was delayed and the production of actinorhodine and undecylprodigiosin was abnormal. The 18J strain reported here was shown to grow slower than the wild type and to have a reduced formation of spores (in agreement with the observations of Ochi (37)), but only actinorhodine was severally affected in our mutant. In addition, the relaxed mutant isolated by Ochi (37) did produce actinorhodine after 10 days on agar plates, while no such effect could be detected in our ORF1deleted mutant under the same conditions. Thus, we cannot yet exclude that the observed differences could reflect either the existence of alternative effects on the (p)ppGpp biosynthetic pathway or the presence of more than one pathway for (p)ppGpp formation in S. coelicolor.
Differentiation and production of secondary metabolites start concomitantly in response to nutrient limitation, and although a possible role of (p)ppGpp in initiating antibiotic biosynthesis has been suggested (37-41, 96, 98), no direct link was established by others (99 -101). Furthermore, (p)ppGpp accumulation due to nutritional shiftdown or serine hydroxamate treatment does not seem to be sufficient to trigger antibiotic production (99 -101), suggesting that sensing of growth rate or (p)ppGpp Synthetase from S. coelicolor growth cessation may be of critical importance (101). Further biochemical and genetic characterization of the ORF1 protein and the deleted mutant will provide some insight into the role played by (p)ppGpp levels in the onset of antibiotic biosynthesis as well as in other regulatory events in the cellular physiology of Streptomyces strains.