We have compared the 13C NMR chemical shifts of the trihalomethyl cations to those of the respective trihalomethanes and found a consistent trend (see Table I). The Δδ13C chemical shift differences12 are 158.6, 194.7, and 234.7 for the trichloro-, tribromo-, and triiodomethyl derivatives, respectively. This trend is in agreement with the positive charge stabilization (increased back-bonding) order of Cl > Br > I. The 13C shifts of trihalomethyl cations correlate well with the electronegativities of the halogen atoms.I4The 13C NMR chemical shifts of the trihalomethyl cations and related reported methyl substituted mono- and dihalocarbenium ions3 also correlate well with the 11B NMR chemical shifts of the corresponding isostructural, isoelectronic boron halidesI3 (see Figure 1). An excellent linear relationship is obtained with a correlation coefficient of unity.15 From this plot, with the known 11B NMR chemical shift of BF3 (10.4 ppm) we can estimate the corresponding 13C NMR shift of the yet unknown trifluoromethyl cation as to be 140 ppm which is only 31.2 ppm more deshielded than trifluoromethane at δ13C 118.8. This indicates substantial stabilization of the trifluoromethyl cation by fluorine back-bonding. Experimental verification, however, must await preparation of the still elusive CF3+ ion under long-lived conditions. Anatoxin-a(s) is a neurotoxic alkaloid associated with the blue-green alga Anabaena flos-aquae. It potent toxicity (LD50 20-40 bg/kg mice) is attributed to exceptional anticholinesterase activity.2 We report here the isolation of anatoxin-a(s) from a cultured strain NRC 525-17 and a field-collected bloom implicated in animal poisonings3 and the determination of its structure as 1. Freeze-dried alga was extracted with 0.05 N AcOH/EtOH. The filtered extract was partitioned between water and CH2Cl2, the aqueous layer was washed with n-BuOH and evaporated in vacuo, and the residue was extracted successively with small portions of 0.05 N AcOH/MeOH and 0.05 N AcOH/EtOH to give a toxic concentrate. Gel filtration on Toyopearl HW40F followed by HPLC on CN and ODS columns gave pure anatoxin-a(s) as a colorless solid in 0.05% yield. Toxin isolation was followed by assaying fractions for anticholinesterase activity.5 Anatoxin-a(s) decomposed rapidly in basic solution but was relatively stable in neutral or acidic (pH 3-5) media6. Anatoxin-a(s) from cultured and field-collected A. flos-aquae exhibited identical chemical and spectral properties, including optical (CD in H2O: [θ] ... -3300, +3900). Mass spectral analysis of anatoxin-a(s) [positive FABMS (m/z 253.1067, MH+), negative FABMS (m/z 251, M-H-), FDMS (m/z 253, MH+)] indicated the molecular formula C7H17N4O4P. The 1H and 13C NMR spectra7b revealed the presence of dimethylamino and P-OMe (JH-P 11.0 Hz; Jc-P 6.7 Hz) groups, a 1,2,3-trisubstituted propane unit, and an sp2 carbon that was fully substituted by heteroatoms (δ 163.7). The methoxyl protons and carbon were the only ones showing distinct coupling to phosphorus. Only one signal was seen in the 31P NMR spectrum and its chemical shift (δ 6.16) agreed well for either a phosphate ester or phosphoramide.8 The 3JBCP (-10.1 Hz) and 1JH,C values for the protons in one of the methylenes of the propane unit suggested that this CH2 was in a five-membered ring, along with the adjacent CH. More information was obtained from NMR analysis of anatoxin-a(s) that had been uniformly enriched to 50% 13C and 90+% 15N10,11 (See also Supplementary Material). The following conclusions could be made: (1) The sp2 carbon at 163.7 ppm was connected to three nitrogens of a guanidine group and that two of these nitrogens were attached to the CH and CH2 in the five-membered ring. (2) The NMe2 group was connected to the side-chain CH2 on the resulting imidazoline.12 (3) No nitrogens were connected to the phosphorus; a methyl phosphate group was therefore present in the toxin. (4) The methyl phosphate group was attached to one of the nitrogens (2Jp-N 4 Hz); the toxin was therefore zwitterionic. Anatoxin-a(s) slowly decomposed during storage at -20 °C into a mixture of 2, 3 (sometimes), and monomethyl phosphate, separable by Toyopearl HW40F chromatography. Compound 2 (FABMS, MH+ m/z 159.1245; CD in H2O, +2400), which differed from compound 3 by an oxygen, could be converted into 3 (FABMS, MH+ m/z 143.1298; CD in H2O, [θ]198 +11000) by catalytic hydrogenation (Pd-C/MeOH). Hydrolytic removal of the monomethyl phosphate group caused a diamagnetic shift of the H-5 signal from 4.71 ppm in 1 to 4.48 ppm in 2; the methylene 1H chemical shifts, however, were essentially identical for the two compounds. Although the 1H chemical shifts for 3 were similar to those for 2, except for one of the H-6 signals which was shifted upfield appreciably (-0.37 ppm), the 13C chemical shifts were significantly different, i.e., upfield for C-5 (-7.8 ppm) and C-2 (-1.9 ppm) and downfield for C-4 (+2.7 ppm) and C-6 (+2.4 ppm).13,14 These chemical shift differences were consistent with placements of the hydroxyl group on N-1 in 215 and the methyl phosphate group on N-1 in 1. Anatoxin-a(s) therefore had to have structure 1. To elucidate the absolute configuration at C-5, R- and S-3 were prepared from D- and L-asparagine, respectively (Scheme I). p-(Benzyloxycarbonyl)-N3-(tert-butoxycarbonyl)-L-2,3-diaminopropionic acid (4),16 for example, was converted to dimethylamide 5 via the N-hydroxysuccinimide ester.17 After removal of the amino-protecting groups (trifluoroacetic acid; H2/Pd-C), the resulting diamine was reduced with BH3-Me2S complex18 to give the triamine 6, which was then treated with S,S'-dimethyl-N-tosyliminodithiocarbonimidate19 to furnish the tosylguanidine 7. Removal of the N-tosyl group was accomplished by refluxing 7 in 48% HBr.20 Synthetic 3 showed identical chromatographic properties and 1H and 13C NMR spectra with the degradation product. The CD spectrum of 3 derived from anatoxin-a(s) was identical with that of synthetic 3 from L-Asn ([θ] 198 +13 000), which meant that C-5 was S. Anatoxin-a(s) is a unique phosphate ester of a cyclic N-hydroxyguanidine. The structure and reactivity is reminiscent of an ester of N-hydrosuccinimide or 1-hydroxybenzotriazole. Cholinesterase inactivation may proceed by nucleophilic attack of Ser at the esteratic site of the enzyme on the phosphate group of 1 with concomitant elimination of 2.21 The natural hydroxamate siderophores contain bidentate hydroxamate donor groups in acyclic, exocyclic, and endocyclic arrangements.1 Of these, the sexadentate endocyclic ligands, such as desferriferrioxamine E, have the highest affinities for iron(III), because of the involvement of the macrocyclic rings in coordination of the metal ion. The only synthetic endocyclic hydroxamate ligand previously reported is a diamino bishydroxamate macrocycle containing pendant carboxylate donors.2 Up to the present time, no synthetic or natural endocyclic trishydroxamate cryptand has been reported, although the potential of such ligands for effective binding of trivalent metal ions has been pointed out.3 The orientation of oxygen donor groups in the hydroxamates places stringent demands on the polyatomic chains which link them together in a manner that places three pairs of oxygen donors symmetrically around a six-coordinate (octahedral) metal ion. Molecular models indicate that eight or more connecting atoms are needed to accomplish this effectively. It is the purpose of this paper to report the synthesis and properties of the first trishydroxamate cryptand. In this case the ligand has nine and 11 atoms between the bidentate hydroxamate units. Cyanoethylation at 25-35 °C in p-dioxane of the 1,1,1-tris(hydroxymethyl)ethane gives the tricyano ether 1 in 83% yield. Hydrolysis of 1 with hydrogen chloride gas in methanol leads to the tricarboxylic acid trimethyl ester 2 (yield = 64%) (Calcd for C17H30O9: C, 53.97; H, 7.94. Found: C, 53.63; H, 7.93. FAB MS (M + H)+ = 379). Treatment of 2 with lithium aluminum hydride affords the triol 5 in 83% yield (Calcd for C14H30O6: C, 57.12; H, 10.27. Found: C, 57.24; H, 10.20. FAB MS (M + H)+ = 295). The tosylate 6, prepared from tosyl chloride and 5, was treated with O-benzylhydroxylamine in 1,2-dimethoxyethane to give the tris(O-benzylhydroxylamine) 7 (Calcd for C35H51N3O6: C, 68.96; H, 8.37; N, 6.90. Found: C, 68.68; H, 8.45; N, 6.69. FAB MS (M + H)+ = 610, yield = 37%). The triacid chloride 4 was obtained by allowing 3, which was prepared from its methyl ester, to react with oxalyl chloride in benzene, yield >95%. High dilution acylation of 7 with 4 in benzene gives the protected macrobicyclic trishydroxamate 8 (Calcd for C49H69N3O12·1H2O: C, 61.06; N, 4.36; H, 8.00. Found: C, 60.95; N, 4.18; H, 7.10. FAB MS (M + H)+ = 893) in 30% yield. The macrobicyclic trishydroxamate cryptand 9 was obtained from 8 by catalytic hydrogenation. This cryptand (1,13-dimethyl-3,11,15,23,26,34-hexaoxa-6,20,29-trioxo-7,19,30-tris(hydroxyaza)bicyclo[11.11.11]pentatricontane, H3THX) was characterized by 1H NMR, 13C NMR, elemental analysis, and FAB MS (Calcd for C28H51N3O12·1/2H2O: C, 53.27; H, 8.40; N, 6.59. Found: C, 53.16; H, 8.19; N, 6.38. FAB MS (M + H)+ = 622). This cryptand is very soluble in methanol and sparingly soluble in water. Its sodium salt is very water soluble. The 1:1 Fe(III) complex of the cryptand was prepared by combining Fe(III) chloride with a slight excess of the neutral (acid) form of the ligand to form a 2.0 × 10-4 molar solution in 2:8 v/v methanol-water and gradually increasing the pH to the desired neutral value. The absorbance spectra of the cryptate (Figure 1) compare well with those of the Fe(III) chelate of desferriferrioxamine B, DFB, reported by Anderegg et al.4 At pH 4.0 the Fe(III) cryptate has a molar absorbance of 2700 at λmax 423 nm (see Figure 1), while that of FeIII-DFB is 2640 (λmax 440 nm) at pH 4.4 The absorbance shifts with pH are also similar for the cryptate and DFB complexes, with an isosbestic point at 480 nm for the cryptand and 481 for the DFB complex. The close similarity in the magnitudes of the molar absorbances of the iron(III) complexes of DFB and the cryptand provide assurance that all three bidentate donor groups of the latter are coordinated to the Fe(III) center. The observed isosbestic point indicates the conversion of one pure Fe(III) complex to another as the pH increases. It is suggested that the reaction corresponds to a monoprotonated complex FeHTHX+, having two coordinated hydroxamate groups, and one protonated, non-coordinated hydroxamate group, which is converted at higher pH to the octahedral Fe(III) cryptate FeTHX, with three coordinated hydroxamate groups arranged in an octahedral fashion around the metal ion. At pH 4.4 and above there is little further increase in absorbance, and one therefore concludes the cryptate to be fully formed, with a molar absorbance of 2750 at λmax = 430 nm. Because the Ga(III) ionic radius is only slightly smaller than that of Fe(III), the new cryptand would also be expected to complex Ga(III) strongly in an octahedral fashion. The 1:1 Ga(III) complex was prepared by the reaction of molar equivalents of Ga(OH)4- and the ligand in aqueous solution at pH 8.9. This is above the pH at which Ga(III) precipitates as Ga(OH)3. The white solid which separated was characterized by elemental analysis and mass spectra (Calcd for C28H48N3O12Ga·2H2O: C,...).