Methanol seeds extracts were partitioned between chloroform and water. Then, chloroform was evaporated, extracts cromatographed on a silica gel column (chloroform-ethyl acetate-methanol gradient) and column fractions processed on a Beckman C18 HPLC column (25 cm × 1 cm i.d., 5 mm particle size) to yield pure ACG. Structural characterization was achieved by IR, ¹H-NMR, ¹³C-NMR, and EIMS.

Acetylated analogs (enzymatic acetylation) were obtained by dissolving the ACG in mixture of dichloro-methane (5 ml) and vinyl acetate (1.2 mol per OH group to be acetylated) in a screw cap vial. Then lipase (Candida antarctica B) was added (10% - 30% of ACG weight) and vial placed on an orbital shaker (37˚C, 150 - 200 rpm) until completion of reaction as shown by TLC. Finally, lipase was filtered and washed with dichloro-methane_. Solvent was removed from the liquid fraction in a rotary vacuum evaporator and aceylated compounds purified by flash column chromatography.

Methoxy methylated ACG derivatives were obtained by reaction with N,N-diisopropylethylamine and methoxymethyl chloride in dichloromethane under a nitrogen atmosphere. At completion of the reaction (shown by TLC) solvent was removed, residue chromatographed on flash column and chemical structure of products assessed by ¹H-NMR and ¹³C-NMR by comparison with ACG precursors.

Spodoptera frugiperda larvae were obtained from our laboratory population. The larval diet was prepared as follows: yeast, 3 g; milled and boiled bean, 250 g; wheat germ, 12.5 g; agar-agar, 12.5 g; ascorbic acid, 1.5 g; methyl p-hydroxybenzoate, 1.5 g; formaldehyde 38% water solution, 4 ml; water, 500 ml. Acetone solutions of natural ACG and derivatives were prepared (100 μg∙ml⁻¹).

A portion of larval diet was thoroughly mixed with an acetone ACG solution (or ACG derivative) to leave 100 µg of compound per g of diet (treated) [24] and solvent left to evaporate at room temperature. Another portion was impregnated with acetone and used as control diet. Traces of solvent were removed in a desiccator. Cotton plugged test tubes containing 2^nd instar larvae (1 per tube) were added with either control or treated diet and kept at 27˚C and 60% ± 15% relative humidity. Upon consumption of 50% of control diet, remaining diets were removed and weighed accurately and feeding ratio was calculated as FR₅₀ = T/C [28] (T and C represent weights of diets consumed in the treated and control experiments, respectively).

Control and treated diets were placed in cotton plugged test tubes (10 replicates each) with 2^nd instar larvae (1 per tube) and kept as mentioned before until the 1^st generation of adults emerged. Larval developmental periods as well as mortality rates were recorded for both treatment and control experiments [29] .

Homogeneous size 2^nd instar larvae were placed in a test tube (1 per tube, 10 replicates) and larval weights measured. Then, accurately weighed portions of either treated or control diet were offered to larvae and kept at 27˚C. Larval weight and diet additions were recorded. Average diet consumption (CI), growth rate (GR) and efficiency in the consumption index (ECI) were calculated, rates expressed as treatment-control ratio and values expressed as ( GR _T / GR _C ) 100% ( CI _T /CI_C) 100% and (ECI_T/ECI_C) 100% in the tables [29] . Control is considered as 100%.

Results are reported as mean ± SD. The differences in the mean values were evaluated by ANOVA test. The Tukey test was used for pairwise multiple comparisons of groups. In all analyses, p < 0.05 indicates significant difference of means [30] .

The isolation of ACG from the chloroform sub-extract was guided by the Kedde’s reagent (TLC). Repeated chromatographic processes permitted the separation of known ACG. As shown in Figure 1, the following compounds were isolated. Mono-THF ACG: annonacin (1), cis-annonacin (2), cis-annonacin-10-one (3); adjacent bis-THF ACG: asimicin (4), rolliniastatin-2 (5); non

adjacent bis THF ACG: cherimolin-1 (6), cherimolin-2 (7), almuñequin (8); bis-THF β-OH ACG: laherradurin (9) and itrabin (10). Derivatives obtained are shown in Figure 2: annonacin (3 OAc) (11) [27] , annonacin (4 OAc) (12) [27] , asimicin (3 OAc) (13), rolliniastatin-2 (3 OAc) (14) [27] , rolliniastatin-2 (MOM) (15), laherradurin (3 OAc) (16) and itrabin (3 OAc) (17) [27] .

The identification of chemically acetylated ACGs (13 and 16), both bis-THF, was achieved by comparing spectroscopic data of ¹H-NMR and ¹³C-NMR with those of original ACGs samples. In ¹H-NMR it was observed that the signals of the protons corresponding to the hydroxyl-based carbons were displaced to low fields (δ 5.0 - 4.0 ppm) and the signals corresponding to the protons of the hydroxyl groups present in the original ACGs disappeared. Likewise, new singlet signals appeared at high fields corresponding to the methyl protons of the acetate groups (δ 2.2 - 2.0 ppm).

In ¹³C-NMR it was observed that the signals corresponding to the hydroxyl-based carbons move to low fields (δ 75.0 - 71.0 ppm) and new signals appeared to lower fields corresponding to the quaternary carbons of the acetate group (δ 171.0 - 170.0 ppm).

The identification of methoxy methylated bis-THF ACG, rolliniastatin-2 (MOM) (15), was achieved by comparing ¹H-NMR and ¹³C-NMR spectroscopic data with that of the original ACG. In ¹H-NMR it was observed that the proton signals corresponding to the hydroxyl base carbons that were at δ 4.2 - 4.5 ppm were displaced at δ 3.9 ppm and the signals corresponding to the protons of the hydroxyl groups present in the original ACG disappeared. Likewise, new singlet signals appeared between δ 3.3 - 3.5 ppm and δ 4.6 - 4.9 ppm, corresponding to the methylic and methylenic protons of the methoxy methyl group respectively.

In ¹³C-NMR it was observed that the signals corresponding to the hydroxyl base carbons that were between δ 71.3 - 74.2 ppm, moved to lower fields towards δ 79.0 - 81.8 ppm and new signals appeared at δ 55.0 - 56.0 and δ 96.0 - 97.0 ppm, corresponding to the methylic and methylenic carbons of the methoxy methyl group respectively.

All ACGs evaluated previously had toxic effects on S. frugiperda at the dose of 100 μg∙g⁻¹ or less [24] . As shown in Table 1, the incorporation of 15 (100 μg∙g⁻¹) into the artificial diet of S. frugiperda displayed the strongest antifeedant effects (FR₅₀) followed by 14 causing a marked decrease in larval growth. Antifeedant action of remaining ACG and derivatives was negligible under experimental conditions (Table 1).

As shown in Table 1, natural ACG containing adjacent bis-THF groups were more active than those with non-adjacent groups. Moreover, compounds with a single THF group were also less active, except for compound 2. The structural

characteristics of the lactonic group also seem to be of importance. Adjacent bis-THF ACG with β-hydroxylated lactone groups had less activity than those with unsaturated lactone groups.

The natural acetogenin rolliniastatin-2 (100 μg∙g⁻¹ of diet) produced the most important toxic action causing 100% mortality of early instar larvae.

Acetylation and methoxy-methylation of ACG reduced their toxicity (Table 1) and caused malformations and size decrease in surviving adults (see Figure 3(A)-(C)) that led them to death before laying eggs.

^†100 μg∙g⁻¹ of diet. ^††Mean ± SD.

Nutritional effects were assessed in order to infer the mechanism of action leading to mortality. The decreases in ECI values for natural adjacent bis-THF ACG (4, 9 and 10) and acetylated bis-THF ACG (13, 14, 16 and 17) are consistent with chronic intoxication leading to death, thus preventing the production of offspring (Table 2). The addition of the methoxy-methylated derivative of rolliniastatin-2 (15) caused less nutritional effects than its acetylated counterpart (14).