Hepatoenteritis in humans caused by toxic cyanobacterial blooms in domestic water supplies that have become eutrophic is a growing concern. Microcystis aeruginosa is the most frequently implicated blue-green alga in these poisonings, and the hepatotoxins associated with this cyanophyte are cyclic heptapeptides known as microcystins. Circumstantial evidence is slowly emerging linking toxic Microcystis blooms with a higher incidence of liver cancer among populations in Third World countries such as China that depend on surface drinking water. In 1979, however, an outbreak of hepatoenteritis on Palm Island in northern Queensland, Australia, was traced to a different cyanobacterium, Cylindrospermopsis raciborskii (Woloszynska) Seenaya and Subba Raju, a species that had not been previously found to be toxic. We report here the isolation and gross structure determination of an unusual alkaloid, cylindrospermopsin, which is hepatotoxic with symptoms indistinguishable from those originally described for the cyanobacterial extract. C. raciborskii was grown in culture as previously described. An aqueous extract (0.9% NaCl) of the ultrasonicated, freeze-dried alga (0.7 g) was fractionated (bioassay-guided) by successive gel filtration on Toyopearl HW40F with 1:1 MeOH/H2O and reversed-phase HPLC on C18 with 5% MeOH in H2O to give white microcrystals of cylindrospermopsin (1, C15H21N5O7S; positive ion HRFABMS, glycerol matrix: MH+ m/z 416.1236, Δ = 0.4 mmu), in 0.5% yield, [α]D -31° (H2O, c 0.1), as the only detectable hepatotoxin. The intense negative ion FABMS (M - H m/z 414) and UV spectrum in H2O [λmax 262 (ε 5500), sh 290 nm (2100)] was consistent with 1 being a substituted uracil. Comparison of the 13C chemical shifts in both D2O and H2O (Figure 1) and 1JCH for C-5 (175 Hz) with values reported for uracil indicated that the substitution was on C-6. The toxin appeared to be a sulfate ester since the air CIDMS of the MH+ ion (FAB mode) showed fragment ions at m/z 336.1688 (C15H22N5O4, Δ = -1.6 mmu), 318, 274 [MH - (hydroxymethyl)uracil]+, 194, and 176 for the loss of SO3 and H2SO4 from the MH+ and [MH - (hydroxymethyl)uracil]+ ions. Detailed analysis of the 500-MHz 1H and 125-MHz 13C NMR spectra in D2O, aided by two-dimensional COSY, HMQC, and HMBC experiments, enabled us to assign all of the 1H and 13C signals and to propose the structure shown in Figure 1. Chemical shifts suggested that nitrogen was attached to the carbons resonating at 45.0 (C-10), 48.3 (C-15), 53.6 (C-8), and 57.9 (C-14) ppm whereas oxygen was present on the carbons absorbing at 70.7 (C-7) and 78.2 (C-12) ppm. Isotope shifts for the C-7, C-8, and C-15 signals in H2O (see ΔδC values in Figure 1) established that NH's were on C-8 and C-15 and an OH group was on C-7. The sulfate group was therefore attached to C-12, and its placement here was supported by the CIDMS data. The coupling constants (Jtrans = 11.1-11.8 Hz, Jcis = 2.0-3.9 Hz, and Jgem = -13.9 ± 0.5 Hz) associated with the signals for the protons on C-8, C-9 (28.5 ppm), C-10, C-11 (36.3 ppm), C-12, C-13 (39.8 ppm), and C-14 showed that these nuclei were located in six-membered rings which required that (1) the same nitrogen be connected to C-10 and C-14, (2) the sulfate ester group on C-12 be oriented axially, and (3) the methyl substituent on C-13 be equatorially disposed. The proton on C-8 (3.87 ppm) and one of the protons on C-15 (3.84 ppm) were coupled (HMBC cross peaks) to a guanidino carbon resonating at 156.5 ppm, and this meant that the toxin was a tricyclic guanidine as depicted in 1. The uracil ring had to be attached to C-7 since HMBC cross peaks were clearly seen between H-5 and C-4, C-6, and C-7 (1a). The 13C-13C COSY spectrum of uniformly 80 + % 13C enriched 1, isolated from alga grown on NaH13CO3 (99%), confirmed the gross structure (see supplementary material). The 13C signals for C-2 and C-6 were broad, and their chemical shifts, unlike the ones for uracil, were very sensitive to small changes in pH around 7, hinting that N-1/C-2 might have an enol rather than an amide structure. Favoring the enol tautomer could be a consequence of the uracil and guanidine units being coplanar with 18(N)-H hydrogen-bonded to N-1. If so, then the NOESY spectrum (NOE correlations: 5-H ↔ 7-H; 7-H ↔ 8-H; 9-Heq ↔ 11-Hov; 10-H ↔ 14-H; 13-H ↔ 11-Hax, 12-H, and 15-H (cis); and 13-CH3 ↔ 12-H, 14-H, and 15-H2) suggests that the toxin has the relative stereochemistry shown in 1a.