Purification of thymidine-diphospho-D-glucose 4,6-dehydratase from an erythromycin-producing strain of Saccharopolyspora erythraea by high resolution liquid chromatography.

TDP-D-glucose 4,6-dehydratase was purified from Saccharopolyspora erythraea, the producer of the macrolide antibiotic erythromycin A, by a high resolution chromatographic method that exploited the difference in the behavior of the protein on anionic exchange chromatography in Tris/HCl or phosphate buffers. By this method, the enzyme was purified approximately 900-fold by two anionic exchange steps to more than 90% homogeneity. It was further purified to apparent homogeneity by hydrophobic interaction chromatography. The enzyme is a homodimer of Mr 36,000 subunits, is highly specific for TDP-D-glucose, requires NAD+ as cofactor, and shows a K'm of 34 microM and V'max of 26 mumol h-1 mg-1 of protein for TDP-D-glucose. TDP and TTP strongly inhibit the enzyme at 2 mM. The maximal TDP-D-glucose 4,6-dehydratase activity coincides with the time of erythromycin production, suggesting that this enzyme is involved in antibiotic biosynthesis.

Macrolide antibiotics are made in bacteria largely from simple fatty acids and glucose (1). For erythromycin A, the macrolide antibiotic produced by Succharopolyspora erythruea (formerly Streptomyces erythreus (2)), it has been established that its two deoxy sugars, D-desosamine and L-mycarose, are made from D-glucose (3,4). Although the specific steps used in the conversion of glucose to these two deoxy sugars are not known, the biosynthesis of other deoxy sugars like rhamnose and fucose is well-understood and involves nucleotidyl diphosphate esters of glucose (5) or mannose (6). The first step in these pathways is the oxidation of a nucleotidyl diphosphoglucose to its 4-keto-6-deoxy derivative. Enzymes that catalyze this oxidation have been purified from Escherichia coli (7), Pseudomonas ueruginosa (8), and Pasteurella pseudotuberculosis (9) as well as from plant (10) and mammalian (11) sources. This type of enzyme has also been studied in other antibiotic-producing bacteria (12, 13), and one report describes the partial purification of thymidine-diphospho-Dglucose 4,6-dehydratase from Streptomyces rimosus, a producer of the macrolide antibiotic tylosin (13).
Deoxy (amino) sugar formation has a key role in the biosynthesis of the macrolides and several other types of antibi-otics (14). To further our investigation of the genetics and biochemistry of erythromycin production, we purified and characterized TDP-D-glucose 4,6-dehydratase from S. erythruea. The purification was accomplished with fast protein liquid chromatography by exploiting the different behavior of the enzyme on anionic exchange chromatography in the presence of Tris/HCI and phosphate buffers. The s. erythraea enzyme requires the presence of NAD+ for oxidation of TDP-D-glucose in vitro and has the strictest substrate specificity and lowest turnover rate among the known bacterial enzymes.

DISCUSSION
We observed an exceptional dependence of protein resolution by fast protein liquid chromatography on the eluting buffer during the purification of TDP-D-glucose 4,6-dehydratase from S. erythruea. Because of the different behavior of this enzyme on anionic exchange chromatography in Tris/ HCl or phosphate buffers, we were able to purify it using only three high resolution chromatographic steps. This buffer effect appears to be general based on the behavior of another protein fraction from S. erythruea (Fig. l ) , where an essentially single peak from the Mono Q-Tris/HCl column could be resolved by rechromatography into many peaks simply by changing to a phosphate buffer. Furthermore, we have purified an L-valine dehydrogenase from Streptomyces coelicolor using the same approach.' Whereas the exact reason for our observations is unknown, Scopes (18) has noted that phosphate buffers have been used successfully in anionic exchange chromatography.
Some of the properties of TDP-D-glucose 4,6-dehydratase from S. erythruea differ significantly from the enzymes isolated from other bacteria. Although it also is a homodimer of M, 36,000 subunits (the other bacterial enzymes are homodimers of M, 38,000-43,000 subunits (5-9)), the enzyme does not contain a tightly bound nucleotide cofactor and thus requires NAD+ for activity. In this respect, the S. erythruea enzyme is like the enzymes from P. ueruginosa (8) and P.
pseudotuberculosis (9). The cofactor is not required by the enzyme from S. rimosus, although its activity was stimulated by addition of NAD+ (13). Several other properties of the two  (7)). Comparison of the respective Vmax/K,,, values for these two enzymes, 0.012 versus 0.10 mol" s-l mg-' of protein, reveals that the S. erythraea enzyme is 10 times less efficient than the E. coli enzyme.
One function of the TDP-D-glucose 4,6-dehydratases from the two Streptomyces could be to provide the first intermediate of the biosynthetic pathway to deoxy sugars like mycarose that are present in the macrolide antibiotics produced by each species. In fact, the formation of TDP-L-mycarose from TDP-D-glucose has been demonstrated in a cell-free system from the tylosin-producing strain of s. rimosus (19). Our observation (see "Results") that the level of the S. erythraea TDP-Dglucose 4,6-dehydratase is correlated with antibiotic production (Fig. 6) supports a similar role for this enzyme in erythromycin biosynthesis, although the presence of the enzyme before the onset of the latter process suggests that it could have other biological functions. Moreover, we have detected TDP-D-glucose 4,6-dehydratase activity in Streptomyces liuiwhich is not known to produce a macrolide antibiotic. Thus, as in E. coli (7), the enzyme might also be involved in the formation of cell wall constituents like rhamnose, even though this 6-deoxy sugar is not widely distributed among Streptomyces species (20). Proof for the involvement of this J. A. Vara and C. R. Hutchinson, unpublished results.