Biochemical Characterization of VlmL, a Seryl-tRNA Synthetase Encoded by the Valanimycin Biosynthetic Gene Cluster*

Previous studies have shown that the valanimycin producer Streptomyces viridifaciens contains two genes encoding proteins that are similar to seryl-tRNA synthetases (SerRSs). One of these proteins (SvsR) is presumed to function in protein biosynthesis, because it exhibits a high degree of similarity to the single SerRS of Streptomyces coelicolor. The second protein (VlmL), which exhibits a low similarity to the S. coelicolor SerRS, is hypothesized to play a role in valanimycin biosynthesis, because the vlmL gene resides within the valanimycin biosynthetic gene cluster. To investigate the role of VlmL in valanimycin biosynthesis, VlmL and SvsR have been overproduced in soluble form in Escherichia coli, and the biochemical properties of both proteins have been analyzed and compared. Both proteins were found to catalyze a serine-dependent exchange of 32P-labeled pyrophosphate into ATP and to aminoacylate total E. coli tRNA with l-serine. Kinetic parameters for the two enzymes show that SvsR is catalytically more efficient than VlmL. The results of these experiments suggest that the role of VlmL in valanimycin biosynthesis is to produce seryl-tRNA, which is then utilized for a subsequent step in the biosynthetic pathway. Orthologs of VlmL were identified in two other actinomycetes species that also contain orthologs of the S. coelicolor SerRS. The significance of these findings is herein discussed.

The aminoacyl tRNA synthetases (AARSs) 2 are a well studied family of enzymes that ligate a specific amino acid to the cognate tRNA for subsequent incorporation into protein (1). The ligation process involves a two-step reaction sequence. The first step proceeds by reaction of the amino acid with ATP to generate an aminoacyl adenylate AA-AMP. This step is followed by transfer of the amino acid to the 2Ј-or 3Ј-hydroxyl group of the 3Ј-terminal ribose of the tRNA.
The AARSs are divided into two unrelated classes. The enzymes of each class share a characteristic catalytic folding pattern, identifiable sequence motifs, and distinctive mechanistic features (2,3). Structural studies have shown that the active site of class I synthetases contains a Rossmann dinucleotide binding fold, whereas members of the class II family of enzymes exhibit a seven-stranded antiparallel ␤-sheet flanked by three ␣-helices in a barrel-like structure (4). All class I enzymes attach their amino acid substrate to the 2Ј-hydroxyl group of the tRNA-terminal adenosine, whereas class II AARSs (except PheRS) attach the amino acid substrate to the 3Ј-hydroxyl group (2).
Investigations over the last two decades have shown that some AARSs participate in biochemical processes besides direct involvement with protein biosynthesis (5,6). These alternate functions are of two types, those that involve the synthesis of an aminoacylated tRNA and those that do not. An example of the former is the conversion of Glu-tRNA Gln and Asp-tRNA Asn into their glutaminylated and asparaginylated versions by amidotransferases (1). Examples of the latter type of function include AARSs that participate in translational regulation, promote the splicing of mitochondrial cytochrome genes, and serve as precursors of species with cytokine-like activity (4,6).
The antibiotic valanimycin (see Fig. 1) is a naturally occurring azoxy compound isolated from the fermentation broth of Streptomyces viridifaciens MG456-hF10 by Yamato et al. (7). In addition to antibacterial activity, valanimycin exhibits potent antitumor activity against in vitro cell cultures of mouse leukemia L1210, P388/S (doxorubicin-sensitive), and P388/ADR (doxorubicin-resistant) (7). As a naturally occurring azoxy compound, valanimycin is a member of a growing family of natural products that now includes the cycad toxins macrozamin and cycasin (7-10), the carcinogen elaiomycin (11)(12)(13), the antifungal agents maniwamycins A and B (14), the nematocidal compounds jietacins A and B (15), and the antifungal agent azoxybacilin (16). Precursor incorporation experiments have established that valanimycin is derived from Lvaline and L-serine via the intermediacy of isobutylamine and isobutylhydroxylamine ( Fig. 1) (17). Furthermore, the azoxy nitrogen atoms in valanimycin have been shown to originate from the nitrogen atoms of isobutylamine and serine (17). Enzymatic and genetic investigations (18 -20) led to the cloning and sequencing of the valanimycin biosynthetic gene cluster, which was found to contain 14 genes (21). The functions of five proteins encoded by these genes have been elucidated. These functions include a valine decarboxylase (VlmD) (21), a twocomponent flavin monooxygenase that converts isobutylamine to isobutylhydroxylamine (VlmH, VlmR) (18 -20), a valanimycin resistance gene (VlmF) (22), and a Streptomyces antibiotic regulatory protein (VlmI) that regulates valanimycin production. 3 The identity of the gene(s) responsible for the formation of the novel N-N bond that is present in valanimycin has thus far remained elusive.
A novel feature of the valanimycin biosynthetic gene cluster is the presence of a gene (vlmL) that appears to encode a class II seryl-tRNA synthetase (SerRS) (21). The translation product of vlmL exhibits a relatively low identity (38%) to the single SerRS that is present in Streptomyces coelicolor. This finding suggests that an additional SerRS gene might be present in S. viridifaciens. The presence of a second SerRS gene (svsR) in S. viridifaciens has been subsequently confirmed by PCR experiments (21). As expected, the translation product of svsR exhibited a high degree of identity (91%) to the SerRS of S. coelicolor but a relatively low level of identity to VlmL (39%). The presence of a gene in the valanimycin gene cluster that appears to encode a SerRS is very intriguing, because the role that this protein might play in valanimycin biosynthesis is quite unclear. In this paper, we have described a phylogenetic analysis of VlmL as well as the overexpression and biochemical characterization of VlmL and the "housekeeping" seryl-tRNA synthetase of S. viridifaciens, SvsR.

EXPERIMENTAL PROCEDURES
General-Unless otherwise indicated, all reagents used in this study were purchased from Sigma, Roche Applied Science, Bio-Rad, or G. E. Healthcare. Oligonucleotides were obtained from the Sigma Genosys. Restriction enzymes were obtained from either New England Biolabs (Beverly, MA) or Promega (Madison, WI). (29.5 Ci/mmol) and [ 32 P]Na 4 P 2 O 7 (102.3 Ci/ mmol) were procured from PerkinElmer Life Sciences. Competent cells of Escherichia coli DH10B and BL21 (DE3) were purchased from Invitrogen and EMD Biosciences, Inc. (Madison, WI), respectively, and were used according to the manufacturer's recommendations. The purity and molecular weight of overproduced proteins were evaluated by SDS-PAGE using broad range protein molecular weight markers from Bio-Rad. Prestained protein molecular weight markers from New England Biolabs were used for Western blotting. Protein concentrations were determined with the Advanced Protein Assay Reagent from Cytoskeleton Inc. (Denver, CO) or by use of the BCA reagent from Pierce. Bovine serum albumin was used as a standard.
DNA Manipulations-Genomic DNA was prepared from S. viridifaciens MG456-hF10 using DNAZOL reagent (MRC Inc., Cincinnati, OH) after pulverization of mycelium frozen with liquid nitrogen. Plasmid DNA was purified with a QIAprep spin plasmid kit (Qiagen, Valencia, CA). DNA fragments were isolated from agarose gels with a QIAquick gel extraction kit (Qiagen). PCR products were separated on agarose gels and purified from the gels. Digestion with restriction endonucleases and ligation experiments were carried out by standard procedures under conditions recommended by the manufacturers. Automated DNA sequencing was performed at Lone Star Sequencing Laboratories (Houston, TX) using universal and synthetic oligonucleotide primers.
Computational Analyses-Primary sequence alignments were performed using The Gene Inspector, version 1.5, software (Textco, Inc., West Lebanon, NH) run on a Macintosh computer. BLAST (basic local alignment search tool) analyses were performed at the National Center for Biotechnology Information web site. GraphPad Prism, version 4.0 (GraphPad Software Inc., San Diego, CA), was used for the nonlinear regression analysis of all enzyme assay data. Sequencher, version 4.1 (Gene Codes Corporation, Ann Arbor, MI), was used to compile DNA sequence data. Phylogenetic analysis of SerRSs was carried out by the creation of a multiple sequence alignment with ClustalW (24). A tree was then generated based on a Kimura distance matrix and the neighbor-joining method using the PHYLIP software package (25).
Cloning of vlmL and svsR-vlmL was amplified from cosmid pVal35 (19) by PCR using an N-terminal primer containing an NdeI site (bold) (5Ј-TATTACATATGCATGACCCTCAC-GAGT-3Ј) and a C-terminal primer with an EcoRI site (bold) (5Ј-TTATGAATTCGAGAGTGGTTCCGTC-3Ј). The vlmL PCR product was cloned into a PCR3.2 TOPO TA vector (Invitrogen) and then subcloned into the NdeI and EcoRI sites of pFLAG-CTC to produce pFLAG-CTC-VlmL. The svsR gene was amplified from the genomic DNA of S. viridifaciens using primers designed from the nucleotide sequence of the S. coelicolor seryl-tRNA synthetase. The N-terminal primer (5Ј-TAT-TACATATGATTGACCTTCGCCTGCTT-3Ј) contained an NdeI site, whereas the C-terminal primer (5Ј-TTATGAATTC-CTTGGCCACCGGCTCCAG-3Ј) carried an EcoRI site. A mixture of Vent and TaqDNA polymerase was used for amplification in standard PCR buffer with annealing at 66°C for 30 s. The 1.3-kb PCR product was purified after separation on an agarose gel, cloned into pGEM-T (Promega), and sequenced. The gene was then subcloned into the NdeI and EcoRI sites of pFLAG-CTC to produce pFLAG-CTC-SvsR. The N-terminal PCR primers for vlmL and svsR were both designed to replace the native GTG start codon of each gene with an ATG codon to improve expression in E. coli.
Overexpression, Purification, and Characterization of Proteins-Overproduction and purification of VlmL and SvsR from pFLAG-CTC-VlmL and pFLAG-CTC-SvsR were carried out according to the protocols outlined in the FLAG vector instruc-tion manual (Sigma). The conditions for optimal expression of VlmL and SvsR were standardized by Western blotting using FLAG monoclonal antibodies (Sigma). E. coli BL21(DE3) cells harboring the desired plasmid were grown overnight in Luria-Bertani medium supplemented with 100 mg/ml ampicillin and then diluted 100-fold into fresh 2ϫ Luria-Bertani medium plus 0.4% glucose and ampicillin. The cultures were grown at 37°C until an A 600 of ϳ0.6 was reached, whereupon isopropyl-1thio-␤-D-galactopyranoside was added to a final concentration of 0.5 mM. After 2 h, the cells were harvested and stored at Ϫ20°C until utilized for isolation of the desired protein. The cells from a 1-liter culture were suspended in 25 ml of TBS buffer (50 mM Tris, pH 7.5, 150 mM NaCl), which contained 0.3% Triton X-100, 2 mg/ml lysozyme, 50 g each of RNase and DNase, and one tablet of EDTA-free protease inhibitor mixture (Roche Applied Science). After 40 min of incubation at 37°C, a cell-free extract was prepared by centrifugation at 18,700 ϫ g for 30 min. The cell-free extract was filtered through a 0.2micron syringe filter and passed over a 2-ml column of FLAG antibody resin. Unbound proteins were removed by washing with 20 bed volumes of TBS buffer. FLAG-tagged proteins were then eluted with a 100 M solution of FLAG peptide (Sigma) in TBS buffer. The protein was concentrated, and the FLAG peptide was removed by repeated concentration and dilution using an Amicon Ultra-4 centrifugal filter with a M r 10,000 cutoff (Millipore Corp., Bedford, MA). The resulting protein was stored at Ϫ80°C in TBS buffer containing 15% glycerol. Western blotting with GroEL antibodies (Sigma) was carried out with each protein preparation to eliminate the possibility of GroEL contamination (26). Native molecular weights for VlmL and SvsR were determined by gel filtration using a 10 ϫ 300mm Superose 6 column and TBS buffer with a flow rate of 0.3 ml/min. The void volume of the column was determined with dextran blue 2000. The molecular weight calibration standards (G. E. Healthcare) used to calibrate the column consisted of thyroglobin (669,000), ferritin (440,000), catalase (232,000), aldolase (158,000), albumin (67,000), and chymotrypsinogen A (25,000).
ATP-[ 32 P]PP i Exchange Assays-The ATP-pyrophosphate (PP i ) exchange reactions were carried out at 30°C in a 100-l reaction volume containing 75 mM Tris-HCl, pH 7.5, 10 mM MgCl 2 , 1 mM tris(2-carboxyethyl)phosphine (United States Biochemical, Cleveland, OH), 1 mM tetrasodium pyrophosphate (0.5 Ci of [ 32 P]PP i ), and 5 mM ATP. The reactions were initiated by the addition of enzyme, and incubations were carried out for variable periods of time. The amino acid concentrations ranged from 20 to 800 M, whereas the concentrations of enzyme ranged from 40 to 400 nM. The reactions were quenched by the addition of 0.5 ml of an activated charcoal: tetrasodium pyrophosphate:perchloric acid mixture (1.6:4.5:5% in water). Each quenched reaction mixture was filtered with a vacuum manifold containing a 24-mm GF/C filter (Whatman Inc., Florham Park, NJ). The charcoal on each filter was washed twice with 1 ml of PP buffer (2% tetrasodium pyrophosphate, 20% aqueous HClO 4 ) and then washed twice with water. The filter paper disc bearing the charcoal was transferred to 10 ml of SX18 -4 scintillation mixture (BD Biosciences), and the level of [ 32 P]PP i incorporation into ATP was determined by liquid scintillation counting.
Aminoacylation Assays-Aminoacylation assays for VlmL and SvsR were carried out with total E. coli tRNA (Roche Applied Science). Typically, a 50-l reaction mix contained 100 mM HEPES, pH 7.2, 5 mM ATP, 5 mM dithiothreitol, 30 mM KCl, 10 mM MgCl 2 , 40 M L-serine, 1.5 Ci of L-[ 3 H(G)]serine, 3.2 mM tRNA Ser (200 g of total E. coli tRNA), and concentrations of enzyme at 0.8 -5.0 nM. The concentration of tRNA Ser in the total E. coli tRNA was calculated from data provided by the manufacturer. Negative controls were carried out by omission of enzyme, tRNA, or ATP from the reaction mixtures. The reactions were incubated at 35°C for variable time intervals or with varied substrate concentrations. Aliquots of the incubation mixtures were spotted onto 24-mm papers disks (Whatman grade 3) that were presoaked in 10% trichloroacetic acid and dried. The paper discs were washed twice with cold 10% trichloroacetic acid containing 5 mM serine and then washed twice with 5% trichloroacetic acid. Each washed paper disc was sequentially submerged for 5 min in 75% ethanol and 100% ether and then dried completely at 60°C. The quantity of [ 3 H]seryl-tRNA present on each disc was determined by liquid scintillation counting using SX18 -4 scintillation mixture (BD Biosciences).
Acid Urea Gel Electrophoresis to Detect Aminoacylated tRNA-The aminoacylation reactions were carried out as described above using L-[ 3 H(G)]serine and E. coli tRNA and the aminoacylated tRNA separated from unacylated tRNAs by a modification of the method of Varshney et al. (27). Aliquots (10 l) of the reaction mixtures were diluted with an equal amount of quenching buffer (8 M urea, 50 mM EDTA, pH 8.0, and 100 mM sodium acetate buffer, pH 5.0), and the aminoacylated tRNA was separated by acidic, denaturing 8 M urea PAGE (14%) using a Mini-Protean cell (Bio-Rad). The electrophoretic separation was run for ϳ2 h at 100 V, and the gel was then submerged in 15 ml of amplifying solution (G. E. Healthcare) after fixing. The treated gel was then dried and autoradiography carried out for 1-6 days depending on the assay.

Cloning and Sequence Analysis of VlmL and SvsR-Previous
work identified a housekeeping seryl-tRNA synthetase gene in S. viridifaciens in addition to vlmL (21). In the current studies, the S. viridifaciens housekeeping SerRS (svsR) and vlmL were amplified using primers that contained restrictions sites to facilitate the cloning of each gene into the protein expression vector pFLAG-CTC. The amino acid translation product of each gene was analyzed by BLAST. SvsR and VlmL both display a strong degree of similarity to other seryl-tRNA synthetases, and both proteins display the three motifs that are associated with class II AARSs (2) (see supplemental Fig. 6). SvsR exhibits Ͼ90% identity to the housekeeping SerRSs of other Streptomyces strains, but only 38% identity to VlmL. BLAST analysis reveals that two other actinomycetes, Streptomyces avermitilis and Frankia sp. EAN1pec, each contain two genes that encode SerRSs. One of the two SerRS genes (GenBank TM BAC70005, ZP_00567562) present in each of these organisms encodes a protein that exhibits a high degree of identity to VlmL. Fig. 2 displays a phylogenetic tree created with PHYLIP software (25) relating these and other bacterial SerRSs to one another. The phylogenetic tree suggests that VlmL (GenBank TM AAN10249), BAC70005, and ZP_00567562 are more closely related to the SerRSs of Gram-negative bacteria than to the SerRSs of Gram-positive bacteria. An alignment between the amino acid sequences of VlmL and the SerRSs of S. coelicolor, S. avermitilis, and Frankia sp. EAN1pec is shown in supplemental Fig. 6.
Protein Overexpression and Characterization-VlmL and SvsR were overexpressed in E. coli as C-terminal FLAG-tagged proteins using the pFLAG-CTC vector in which expression is driven from the inducible trc promoter. Both proteins were overproduced largely in soluble form with yields of ϳ5 mg/liter culture broth. The maximum amounts of protein were observed after 2 h of induction at 37°C. Longer incubation at 37°C or the use of lower temperatures resulted in proteolysis and reduced yields. On SDS-PAGE analysis, both VlmL and SvsR exhibited the molecular weights expected for the denatured proteins and appeared to be Ͼ95% pure (data not shown). The native molecular weights of VlmL and SvsR were determined by gel filtration in TBS buffer on a Superose 6 column calibrated with standard protein molecular weight markers. The elution volume of VlmL corresponded to a molecular weight of ϳ90,000, indicating that VlmL exists as a dimer. This observation is consistent with the behavior exhibited by other SerRSs (3,28). SvsR displayed two major fractions on gel filtration. The elution volume for one fraction was similar to that of VlmL, indicating a dimeric form for the enzyme, whereas a second fraction amounting to ϳ50% of the total protein eluted at an early time point, similar to the 669,000 thyroglobin standard. This high molecular weight fraction appears to be an aggregated form of SvsR, because SDS-PAGE analysis indicated that the protein sample used for analysis was of high purity. Potential contamination of the overproduced SvsR by GroEL (native molecular weight of 14-mer ϳ800,000) was ruled out by Western blotting with GroEL antibodies.
PP i Exchange Assays-The adenylation activity exhibited by VlmL and SvsR with L-serine as a substrate was assayed using the amino acid-dependent exchange of radiolabel from [ 32 P]PP i into ATP. This assay measures the reversible formation of the aminoacyl-AMP derivative of the amino acid (Equation 1).
VlmL and SvsR were each assayed in a buffered medium containing ATP and [ 32 P]PP i . The progress of the exchange reaction was monitored versus time under conditions of varied substrate concentration by adsorption of the radiolabeled ATP onto activated charcoal and quantification of the adsorbed radioactivity by liquid scintillation counting. Nonlinear regression analyses of reaction velocity versus substrate concentration were transformed into Lineweaver-Burk plots to calculate K m and V max values, the latter of which were then used to calculate k cat . Fig. 3 shows the substrate versus velocity and Lineweaver-Burk plots (inset) for VlmL and SvsR, respectively, and the measured kinetic parameters are summarized in Table 1. The data show that VlmL displays a somewhat tighter binding affinity for L-serine than does SvsR, whereas the k cat for L-serine turnover by VlmL is only ϳ7% of that exhibited by SvsR. The k cat /K m values for the two enzymes show that SvsR is the more efficient of the two enzymes at catalyzing the adenylation of L-serine.
Aminoacylation Activities of VlmL and SvsR-The ability of VlmL and SvsR to catalyze the aminoacylation of tRNA by L-serine was evaluated. Each of the overproduced proteins was incubated with total E. coli tRNA and L-[ 3 H(G)]serine and the [ 3 H]seryl-tRNA produced from the pool of tRNA was measured by trichloroacetic acid precipitation on Whatman filter paper. Intriguingly, both VlmL and SvsR catalyzed the aminoacylation of tRNA by L-serine. The rates of tRNA aminoacylation catalyzed by VlmL and SvsR were linear with time (Fig. 4), and omission of enzyme, ATP, or tRNA resulted in the complete lack of precipitated radioactivity (data not shown). The aminoacylation reactions catalyzed by VlmL and SvsR followed Michaelis-Menten kinetics under conditions in which the concentration of L-serine or tRNA (Fig. 5) was varied. The maximal rates for the conversion of radiolabeled serine into radiolabeled seryl-tRNA by VlmL and SvsR were calculated to provide a comparison of the efficiency of the aminoacylation reaction catalyzed by each enzyme. The calculations show that these two enzymes catalyze the aminoacylation reaction at comparable rates.
Acid Urea Gel Analysis of Aminoacylation of tRNA-In order to confirm that [ 3 H]seryl-tRNA was produced in the aminoacylation reactions catalyzed by both VlmL and SvsR, the labeled seryl-tRNA produced in the enzymatic incubations was separated by PAGE using acidic, urea-containing gels, and the labeled seryl-tRNA was visualized by autoradiography. It was observed that the amount of labeled seryl-tRNA produced by both enzymes increased with time in a manner that was consistent with the increase in labeled seryl-tRNA observed by trichloroacetic acid precipitation (Fig. 4, insets). Acidic urea PAGE analysis of control incubations carried out in the absence of tRNA confirmed that the radioactive product observed by PAGE in the complete incubations was tRNA-dependent (data not shown). Additional control experiments verified that the formation of radiolabeled seryl-tRNA was also dependent on the presence of ATP and the presence of VlmL or SvsR (data not shown).

DISCUSSION
Previous investigations have shown that the valanimycin producer S. viridifaciens contains two genes that appear to code for SerRSs (21). Sequence analysis suggests that one of these   genes (svsR) is likely to play a primary metabolic role, whereas the second gene (vlmL), which is embedded in the valanimycin gene cluster, appears to be involved in valanimycin biosynthesis (21). A phylogenetic analysis (Fig. 2) of the VlmL protein reveals that it is not closely related to the SerRSs of Gram-positive and Gram-negative bacteria. However, orthologs of VlmL exist in both S. avermitilis MA-4680 and Frankia sp. EAN1pec. The sequence differences between the VlmL subfamily and housekeeping SerRSs are greatest in the N-terminal region (supplemental Fig. 6), which encodes a coiled-coil domain (28). This domain has been shown to interact with the cognate tRNA (28), but it might also interact with other proteins involved in valanimycin biosynthesis. Analysis of the genomic regions surrounding the VlmL orthologs of S. avermitilis and Frankia sp. reveals the presence of several genes that exhibit strong similarities to genes found within the valanimycin biosynthetic gene cluster.
Consequently, it appears that a locus for the biosynthesis of a secondary metabolite related to valanimycin is likely to be present in both of these actinomycete species.
To clarify the function of vlmL in valanimycin biosynthesis, the biochemical characterization of both the VlmL and SvsR proteins was undertaken. VlmL and SvsR were overproduced in E. coli in soluble form as C-terminal FLAG-tagged fusion proteins and purified by affinity chromatography. The purified proteins were first assayed for the L-serine-dependent exchange of [ 32 P]PP i into ATP, an assay that measures the enzyme-dependent reversible formation of the seryl-AMP derivative (see Equation 1). Both proteins displayed activity in this assay, with VlmL exhibiting a somewhat slower turnover rate than SvsR (Table 1). VlmL and SvsR were then assayed for their ability to catalyze the aminoacylation of total E. coli tRNA with serine (Equation 2).
AARS ⅐ AA Ϫ AMP ϩ tRNA º AARS ϩ AA Ϫ tRNA ϩ AMP (Eq. 2) Michaelis-Menten-like behavior was observed for both proteins in this assay (Fig. 5). The data (Table 1) show that both proteins catalyze the aminoacylation reaction with comparable efficiency. From these observations, we hypothesize that the function of VlmL in valanimycin biosynthesis is to produce seryl-tRNA, which then serves as the substrate for a subsequent step in the biosynthetic pathway. Precedent for involvement of an aminoacyl tRNA in a biochemical pathway unrelated to protein biosynthesis is provided by the formation of 5-aminolevulinic acid from L-glutamyl-tRNA in chloroplasts and some bacteria (29). Although the details of the later stages in the valanimycin biosynthetic pathway are presently unknown, several steps will clearly be required to produce valanimycin from seryl-tRNA and isobutylhydroxylamine. These include N-N bond formation, a formal four-electron oxidation of the N-N bond to the level of an azoxy group, and dehydration of the serine-derived hydroxyl group. The most novel of these processes is clearly N-N bond formation. It is possible that the ester moiety of the seryl-tRNA is somehow required for this reaction. Prior investigations have shown that the oxygen of isobutylhydroxylamine is not the source of the valanimycin azoxy oxygen atom (23). This observation limits the number of mechanisms that can be envisioned for formation of an N-N bond. An interesting issue raised by our hypothesis is the degree of cross-talk that occurs between the seryl-tRNA produced by SvsR for use in protein biosynthesis and the seryl-tRNA generated by VlmL for valanimycin biosynthesis. Previous studies have shown (21) that disruption of the vlmL gene strongly reduces, but does not eliminate, valanimycin production. The phenotype of the vlmL mutant suggests that seryl-tRNA produced by SvsR can be used for valanimycin production, although the process appears to be inefficient. This inefficiency might result from compartmentalization of the valanimycin pathway. Alternatively, VlmL may utilize only a single tRNA Ser isoacceptor as a substrate in contrast with SvsR, which would be expected to utilize all of the isoaccepting tRNA Ser species. Future studies will examine this question as well as the role played by seryl-tRNA in valanimycin biosynthesis.