During the course of our screening program for new antibiotics active against Valsa ceratosperma, the causative fungus of the apple canker disease, we found that Streptomyces exfoliatus BT-38 produced a new phenoxazine antibiotic, exfoliazone (I). In this paper, we report the isolation andstructure of I. S. exfoliatus BT-38, which was isolated from Matsumoto city, Nagano Prefecture, Japan, was cultivated at 27°C in a 60-liter jar fermenter containing 30 liters of a medium with agitation rate of 400rpm and air flow of 30 liters/minute. The medium consisted of glucose 2.5%, soybean meal 1.5% dry yeast 0.2% and CaCO3 0.4%; pH was adjusted to 6.2. After fermentation for 90 hours, the culture broth was separated into filtrate and mycelium by centrifugation. The flow diagram for the isolation of I is shown in Fig. 1. The culture filtrate was adsorbed on Diaion HP-20 (10 liters, batch treatment), which was washed with water and then eluted with MeOH.The active eluate was concentrated under reduced pressure to 5 liters and the aqueous residue was extracted with CHC13. Further purification was made by silica gel column chromatography, Toyopearl HW-40column chromatography and preparative HPLC. Crystallization from hot CHC13gave pure I (30.8mg) as orange needles.The physico-chemical properties of I were as follows: MP 294~296°C; IR vmax (KBr)cm"1 3310, 1700, 1620; UV /lmaxnm (e) 238 (75,300), 400 (42,300). The HREI-MS of I showed a molecular ion peak at m/z 284.0825, indicating its molecular formula to be C15H12O4N2 (Calcd 284.0853); XH NMR (500MHz, DMSO-d6) S2.22 (3H, s), 4.59 (2H, s), 6.46 (1H, s), 7.51 (1H, d, /=8.5Hz), 7.56 (1H, dd, /=2.0and 8.5Hz), 7.75 (1H, d, /=2.0Hz),8.25 (1H, s) and 9.68 (1H, s, exchangeable). These spectral data were very similar to those of N-acetylquestiomycin A1}, suggesting that I belongs to the group ofphenoxazine antibiotics (Fig. 2). The 1 3C NMRspectral data of I and iV-acetylquestiomy-cin A are summarized in Table 1. Comparison of 13C NMRspectral data of these two compoundsenabled us to assign 13 out of the 15 signals of I. The spectral differences between them were that the aromatic methine observed at Sc 130.0 in TV-acetylquestiomycin A was replaced by a quaternary aromatic carbon at Sc 140.0 in I with appearance of a newoxymethylene signal at dc 63.4 in the latter. From the molecular formula of I, this new functional group was ascribed to a hydroxymethyl group.These results suggested that I was a hydroxymethyl derivative of TV-acetylquestiomycin A. Comparison of *H NMRspectral data of the two antibiotics revealed that the two aromatic singlet protons on ring C remained unchanged in I (<5H 8.25 and 6.46. cf. SH 8.37 and 6.41 in 7V-acetylquestiomycin A). Therefore, the hydroxymethyl group must be located on ring A. The aromatic ring protons on ring A were observed at c5H7.51 (7=8.5Hz), 7.56 (/= 8.5 and 2.0Hz) and 7.75 (/=2.0Hz). This ABMtype pattern suggested that the position of the hydroxymethyl group must be either C-7 or C-8. The C-8 position was favored, because it allowed the lowest field ring A hydrogen to be located at C-9, peri to the nitrogen at position 102). The proposed structure of I was supported by the close similarity of the chemical shifts of the ring A protons between I and an analogous metabolite, acetyl- michigazone3) (Fig. 3).I showed antifungal activity only against V. ceratosperma. The dosage for 50% inhibition of mycelial growth (ED50) was 70 ^g/ml. Tested so far, it was inactive against Gram-positive and Gramnegative bacteria, yeasts and other fungi. Detailed biological activities of I will be reported elsewhere.