Diazaborines are synthetic compounds, having one boron and two nitrogen atoms in a six-membered ring1"3. Most show moderate to good antibacterial activity. Analysis of the antibacterial spectrum of diazaborines synthesized in our laboratories showed that only Gram-negative organisms were affected, suggesting that these antibacterials might interfere with the biosynthesis of a specific component of such bacteria. To investigate this we used 84474 (see formula below) as a model compound, assuming that inhibition of LPS biosynthesis by this substance is a general property of diazaborines. The compound was tested using Escherichia coll K-12 and Salmonella typhimurium mutants. The MIC (minimum inhibitory concentration) values, determined by the tube dilution method, were 1.25 and 2.5 jxg mP1 for E. coll and Salmonella, respectively. E. coli K-12 has a rough colony morphology, that is, it produces LPS molecules that lack the O-antigenic side chains. The LPS molecule contains one galactose residue in a 1,6 branch of the core oligosaccharide chain4'5. As LPS is the only macromolecular structure in these bacteria that contains galactose, we used the incorporation of galactose into whole E. coli cells as an index of LPS biosynthesis. A galactose-epimerase-negative mutant of E. coli K-12, PL2, was chosen for this experiment to avoid the spread of the radioactive label, introduced via galactose, into the general carbon metabolism of the bacteria.As shown in Fig. 1, the incorporation of galactose into macromolecular material was significantly inhibited by very low concentrations of the diazaborine. Other antibacterials with different modes of action, such as chloramphenicol, epicillin, mecillinam and bicyclomycin, showed no such effects (Table 1). Interestingly, all /3-lactam antibiotics tested enhanced galactose incorporation; the reason for this is unknown. However, transport of galactose into bacterial cells was not inhibited by the diazaborine. This was shown by centrifuging the bacteria, after incubation with 3H-galactose, through a layer of silicone oil and by measuring the radioactivity in the cell pellet6. Both with and without diazaborine, the bacterial cells accumulated similar amounts of galactose (data not shown). The step which is sensitive to the inhibitor must therefore be the incorporation of labelled galactose into macromolecular material. The incorporation of galactose into LPS was demonstrated by phenol extraction of E. coli cells after incubation in radiolabelled galactose-containing medium by the method of Westphal and Jann7, and subsequent SDS-polyacrylamide gel electrophoresis of the aqueous phase. About 75% of the radioactivity migrated on the gel with the mobility of a purified, marker LPS. Some radioactivity was associated with an unidentified, low-molecular weight compound, which migrated with the bromophenol marker. The inhibition of LPS biosynthesis by diazaborine was also demonstrated by electron microscopy. We prepared antibodies against S. typhimurium SF1135, a wild-type strain which contains a smooth LPS having complete O-antigenic poly-saccharide chains. The binding of these antibodies to the O-antigenic chains of the wild-type cells was visualized by labelling with ferritin as described elsewhere8. No binding occurred if the corresponding galactose-epimerase-negative mutant, SF1195, was mixed with the ferritin-labelled antibodies. This was expected, as the Salmonella strain used had been grown in the absence of galactose and was therefore unable to incorporate this sugar into the core oligosaccharide of LPS; thus it produced an incomplete, 'rough' LPS molecule. Miihlradt et a/.8 showed that addition of galactose to the growth medium of these mutants leads, within minutes, to the appearance of the complete, 'smooth' LPS on the cell surface and it can be detected by binding of the ferritin-labelled antibodies. By this method, we were able to determine the kinetics of LPS biosynthesis and monitor its appearance on the outer cell wall of the bacteria. The synthesis of complete LPS molecules was strongly inhibited when Salmonella galE" cells were treated with diazaborine for a short time before the addition of galactose to the medium. The electron microscopic images of control and 84474-treated Salmonella cells, visualized by binding of the ferritin-labelled antibodies, are shown in Fig. 2. A distinct reduction in the number of newly formed 'smooth' LPS chains was detected when the dose of inhibitor was >10jxgmr1. None of the standard antibiotics investigated in this system (chloramphen-icol, cephaloridine, epicillin and bicyclomycin), each applied at 50 ixgmr1, gave any detectable alteration in the amount of antibody bound per cell. A modification (N. v. Jeney, personal communication) of the ELISA method9 gave quantitative results that corroborated the electron microscopic observations. With this test, inhibition could be observed at concentrations of diazaborine as low as 2 jxg ml'1 (data not shown).Table 1Incorporation of 3H-galactose in E. coli PL2 Concentration(iJigmr1) c.p.m.Control (no antibiotic) 73,398 84474 20 3,304Chloramphenicol 20 71,750 Epicillin 10 126,374Mecillinam 20 121,410 Bicyclomycin 20 91,255The basic procedure was as described in Fig. 1 legend except that time of incubation with antibiotic was 30 min.Mean of two independent experiments. The O-antigenic polysaccharide portion, which is linked to the LPS 'core', can be lost due to mutation, without any apparent deleterious effect on the cell10. Cells having LPS in which the oligosaccharide chain is even less extended as a result of a mutation are still viable, the most drastic changes being in the 'Re' mutants. These produce only a truncated LPS without any sugars attached to the 2-keto-3-deoxy-octonic acid (KDO) moieties. Some of these 'deep rough' mutants show limitations in their growth behaviour in certain media10. The existence of such viable mutants suggests that inhibition by a drug of the extensionof the polysaccharide part of LPS would not be a lethal event. However, the mutationally induced blockage in the synthesis of KDO, which is incorporated as a trisaccharide into the lipopolysaccharide chain, seems to result in non-viable variants11. We therefore tested for interference with KDO metabolism or utilization by 84474. Metabolism of KDO was monitored by allowing the cells to use radioactive D-arabinose 5-phosphate, a precursor in the synthesis of KDO which can be actively transported into bacterial cells via the hexose mono-phosphate transport system11. This uptake in S. typhimurium AG 701 (uhpc) was efficient and was not disturbed by 10 |jiM diazaborine. This was shown by washing the bacteria with physiological saline at room temperature after incubation with labelled arabinose 5-phosphate and 84474. Both control and drug-treated bacteria retained an equal amount of radioactivity at various time intervals (data not shown). However, incorporation into material that could not be extracted from the cells by hot water treatment, for example, LPS and other macromolecu-lar compounds, was strongly inhibited (Fig. 3;tl&z). Fifty per cent inhibition occurred in the rnicromolar concentration range of the inhibitor. Fig, 1 Incorporation of 3H-galactose in E. coli PL2. E. coli PL2 (galE), a mutant originally described by Buttin16 (obtained from the E. coli Stock Center, New Haven), was grown at 37 C in a synthetic M9 medium17 containing 0.5% glycerol as a carbon source for the early logarithmic phase (A60o = 0.3), at which point IjxCimr1 3H-galactose (9.3 or 22.3 Ci mmo!"1) was added. Growth was continued for 10 min and 0.5-ml portions were filtered on to Whatman GF/C filter disks. The filters were washed with 100 ml boiling water, dried and counted after addition of a toluene-based scintillation fluid.Fig. 2 The bacterial strains used here were obtained from G. Schmidt, Max-Planck-Institut fur Immunologie, FRG. S. typhi-murium SF 1195 (galE) was grown in several flasks under aeration in M9 medium with glycerol to the late logarithmic phase 0^600 = 1). At this point various quantities of 84 474 were added to give 2, 10, 50 and lOOjxgml"1 final concentrations of the diazaborine. One sample contained no inhibitor. After further incubation for 5 min galactose was added to a final concentration of 50 mM. The cultures were incubated for another 5 min then the bacteria were collected by centrifugation and washed once with saline. After centrifugation the cell pellet was suspended in 200 jjj of a ferritin-labelled anti-SF1135 IgG solution in 0.89% NaCl/30 mM sodium barbital buffer, pH 7.2, and left in an ice bath for 1 h. The ferritin-labelled antibody was prepared according to Miihlradt etal.8. After incubation, the bacterial cells were pelleted by centrifugation, washed with the NaCl-containing barbital buffer, resuspended in the same medium and applied to formvar-carbon-layered copper grids, a, Salmonella SF1195 without antibiotic treatment and b, after pretreatment with 50 |xg mF1 84 474. Scale bars, 200 nm. KDO is formed from D-arabinose 5-phosphate and phos-phoenol pyruvic acid via an 8-phosphorylated intermediate by the action of KDO 8-phosphate synthetase, and is subsequently dephosphorylated by the enzyme KDO 8-phosphate phos-phatase12'13. KDO is then 'activated' by the CMP-KDO-synthetase, with CTP as second substrate, to give CMP-KDO. From this intermediate, KDO is linked to the 'precursor molecule' by the membrane-bound CMP-KDO-transferase14. It is expected that inhibition of one of the enzymes involved in KDO metabolism by diazaborine would result in an accumulation of the substrate of this particular enzyme. We demonstrated this by culturing S. typhimurium AG 701 with 3H-D-arabinose 5-phosphate for several hours to avoid pool effects and by adding diazaborine in the last 30 min of the incubation period; there was an elevated KDO content. This was shown by treating the bacteria with hot water and then analysing the filtrates by paper electrophoresis (Fig. 3&). The accumulation of KDO can be explained either by an inhibition of CMP-KDO-synthetase or by an accumulation of CMP-KDO, the last of the substrates located in the cytoplasm. Extraction of the highly unstable CMP-KDO14 would invariably lead to the generation of KDO. When tested in our institute, there was no observed effect of the diazaborine on the activity of partially purified CMP-KDO synthetase (F. M. Unger and H. Grasmuk, personal communication). Thus, it is reasonable to assume that 84474 interferes with the correct transfer of CMP-KDO to the lipid A-precursor molecule. The conclusion that 84 474 affects KDO metabolism is consistent with the observed intrinsic resistance of Bacteroides species to diazaborines; the LPS of Bacteroides fragilis is thought to lack KDO15. Fig. 3 Incorporation of 3H-D-arabinose 5 -phosphate, a, S. typhimurium AG 701 (from P. Ray, Burroughs Wellcome) was grown in 50 ml of M9 medium with glucose. When the bacterial growth was in the early log phase, the culture was subdivided into 2-ml fractions, 0.5 jxCi 3H-D-arabinose 5-phosphate (42.5 mCi mmoF1)18 per ml was added, and the incubation continued for 15 min. The inhibitor was added at various concentrations together with D-arabinose 5-phosphate. The samples were filtered through nitrocellulose filters and washed repeatedly with 3 ml of boiling water. The filters were dried and counted in a toluene-based scintillation fluid, b, When the intracellular radioactivity of low-molecular weight material was analysed, the time of labelling of the cells was different. The bacteria were grown continuously in the presence of 3H-D-arabinose 5-phosphate (49 mCi mmol"1), then collected on nitrocellulose filters, washed twice with cold water and several times with boiling water. The hot water fractions were collected separately, evaporated to small volumes under reduced pressure, applied to Whatman 3MM paper and separated by high-voltage electrophoresis in 0.05 M ammonium formate, pH 2.9, at 50 V cm"1 for 60-90 min. Picric acid and authentic KDO served as reference compounds. The radioactivity on the paper was determined by cutting the paper into 1 cm-wide strips and counting the water eluate in a scintillation spectrometer. The reference KDO was detected by heating the paper to 140 C for 2 min - this produces a brightly fluorescing spot at the position of KDO in UV light (F. M. Unger, personal communication). The prominent radioactive peaks show the same mobility as authentic KDO. A, Control, no additions; o, 2 M-gml"1 84474; o, 10 ^gmF1 84474.We thank W. Brunowsky, V. Mongold and E. Foglar for technical assistance and Dr P. Ray for a gift of labelled arabinose 5-phosphate. We also thank Drs M. Brown and F. M. Unger for their help with improving the manuscript and H. Oath for typing it.