Journal of Chemical Ecology, Vol. 11, No. 12, 1985 ISOFLAVONOID FEEDING DETERRENTS FOR Costelytra zeatandica Structure-Activity Relationships G E O F F R E Y A. L A N E , 1 D A V I D R. B I G G S , l G R A E M E B. R U S S E L L , 1 O L I V E R R. W. S U T H E R L A N D , 2 E. M A R E E W I L L I A M S , 2 J O H N H. M A I N D O N A L D , 3 and D E B O R A H J. D O N N E L L 3 t Applied Biochemistry Division, DS1R Palmerston North, New Zealand 2Entomology Division 3Applied Mathematics Division, DSIR Auckland, New Zealand (Received January 22, 1985; accepted May 2, 1985) Abstraet--A number of naturally occurring isoflavonoids of differing substitution patterns and oxidation states have been tested for feeding deterrent activity in a bioassay with larvae of Costelytra zealandica White. The most active deterrents, which reduced feeding significantly at 0.2-1.0 #g/g, arc those compounds containing a ring B-fused cycloprenoid moiety. The least active compounds were highly oxidized coumestans and isoflavones. The ring B-fused cyclic isoprenoid moiety and the presence of a 2'-oxy function appear to be structural features important for high activity. It is suggested that the feeding deterrent activity of isoflavonoids relates to their stereochemistry and that the most active compounds have or can adopt a similar nonplanar molecular shape with a similar arrangement of polar and lipophilic groups. Key Words--Isoflavonoids, structure-activity relationships, insect feedingdeterrent activity, Costelytra zealandica, Coleoptera, Scarabaeidae, stereochemistry. INTRODUCTION R e c e n t w o r k has shown that the s u b t e r r a n e a n , r o o t - f e e d i n g larvae of the beetle C o s t e l y t r a z e a l a n d i c a W h i t e (Coleoptera: Scarabaeidae), which inflict m a j o r d a m a g e on N e w Z e a l a n d pastures, are d e t e r r e d f r o m f e e d i n g by several isoflavonoid constituents o f l e g u m e roots (Russell et a1.,1978; Sutherland et. al., 1980; L a n e et. al., 1985). Several o f the m o s t active c o m p o u n d s have also been r e c o g n i z e d as phytoalexins (antifungal m e t a b o l i t e s elicited in response to stress 17t3 098-0331/85/1200-1713504.50/0 cQ 1985 Plenum Publishing Corporation 1714 LANE ET AL. or infection) in the aerial parts of legumes. This observation led to the proposal that in the defense strategy of these plants, the isoflavonoids may serve two different and perhaps independent roles, as phytoalexins and also as insect feeding deterrents (Sutherland et al., 1980). Isoflavonoids are only one of several classes of compounds implicated in legume resistance to grass grub (Sutherland et al., 1975, 1982a, b; Hutchins et al., 1984). The isoflavonoids present in different legumes differ in structural type, yield, and level of feeding deterrent activity. In order to understand and predict the effect of the isoflavonoids of a particular legume on the feeding of C. zealandica larvae, it is important to determine whether deterrent activity can be related quantitatively to defined structural features of the constitutive isoflavonoid molecules. It is generally accepted that molecular size and shape are important determinants for semiochemical activity (Silverstein, 1979), and studies with insect feeding deterrents have shown that redox potential, intramolecular hydrogen bonding capability (Norris, 1977), and functional group stereochemistry (Kubo and Ganjian, 1981 ; Kojima and Kato, 1981) can be important factors in relating molecular structure to insect gustatory response. In this study we have examined the effect of a range of naturally occurring isoflavonoids, some synthetic analogs, and some simple chromenes on the feeding behavior of C. zealandica larvae in order to determine those molecular features that are important for high feeding-deterrent activity. We have drawn together data from our earlier studies on isoflavonoids (Sutherland et al., 1980; Lane et al., 1985) and new data acquired for this study, and analyzed them with a uniform statistical treatment. METHODS AND MATERIALS Feeding Deterrent Assay. Field-collected third-instar larvae of Costelytra zealandica, which had been starved for 24 hr, were enclosed individually in 5.2cm glass Petri dishes with a 4% agar-4% cellulose powder disk (1.5 cm). Feeding-deterrent activity was assessed by comp~xing 24-hr fecal pellet counts of larvae offered disks containing both standard feeding stimulant (0.1 M sucrose + 0.01 M ascorbic acid) (Sutherland, 1971) and test material with similar counts of larvae offered disks containing feeding stimulants only (control) and of a third group offered disks prepared with neither (blank). Twenty larvae were tested with each medium. Blank and control treatments were run on all feeding occasions, and the difference between them was taken as the standard stimulated response (standard). Most isoflavonoids were tested at four concentrations in the disk, usually 2, 20, 100, and 200/~g/g. Lack of material limited the range of some compounds to three concentrations, while the activity of a few others required a wider test range. The required amount of each compound was dissolved in solvent, added 1715 FEEDING DETERRENTS to a weighed sample of cellulose powder, and the solvent evaporated. Test media were then prepared by adding an appropriate amount of hot 4 % agar solution to each cellulose powder sample, stirring vigorously, and pouring a measured amount into a glass Petri dish lid containing either sufficient concentrated sucrose and ascorbic acid solution to make up the desired final concentration (test and control) or an equivalent volume of distilled water (blank). The combined ingredients were then stirred thoroughly and set aside to cool. This procedure was adopted to minimize oxidation of the ascorbic acid. Test disks were then cut from the media with a cork borer. Statistical Treatment of Data. Although the fecal pellet count is the most accurate, reproducible, and convenient method of measuring ingestion by C. zealandica larvae (Sutherland, 1971), the feeding behavior of individual insects presented with the same medium at the same time often differs widely (for example, see Figure 1). Precise quantification of the behavior and comparison of response to often quite similar test chemicals was therefore difficult. Instead of the mean fecal pellet count, a Hodges-Lehmann estimate (Hollander mad Wolfe, 1973) was used as the measure of the overall level of response of all insects tested at a particular concentration of a test compound. As used here, it is the median of all possible pairwise averages of individual fecal pellet counts. Its advantage is that it is unaffected by a few very large or very small counts, which were a feature of some sets of the results. o o e-- ~ z o 1 30- -@ [] o c 8 D g s E ~-~ 20~ 8 [] ~ o [] [] 3 o -~ 1 0 -"1 Z Control 0.001 0,01 01 1 10 Blank Phaseollin Conc.(ppm) FIG. 1. Result of one assay of the feeding response of 20 third-instar Costelytra zealandica larvae to each of five concentrations of phaseollin and to control and blank treatments. [3 = individual fecal pellet counts; O = Hodges-Lehmann estimate; line shows estimated dose-response curve. 1716 LANE ET AL. The data were analyzed by a procedure analogous to probit analysis. The Hodges-Lehmann estimates were plotted against log concentration. A straight line was fitted to the points obtained and the FD50 was calculated as the concentration corresponding to the midpoint between the blank and the control fecal pellet counts, i.e., that concentration at which stimulated feeding was reduced to 50% of the standard. The slope of each line was also calculated. Figure 1 shows individual pellet counts, Hodges-Lehmann estimates, and the fitted line for one of the trials with phaseollin (I). Providing a suitable concentration range was used, the assumption of linear change in feeding response with log-dose appeared adequate; the data did not allow the fitting of more complex models such as a sigmoid curve. In a few cases, all concentrations tested depressed feeding to a level close to or even below that for the blank. It was then clear that the FD~o was less than the lowest concentration tested. These experiments were repeated with a more suitable range of concentrations. In order to calculate standard errors for the FDso values and the slopes, pooled standard errors of the relevant Hodges-Lehmann estimates were required. For each Hodges-Lehmann estimate, the standard error was estimated at the 100 (1 - cr % confidence level by the method of Hollander and Wolfe (1973). The term cr was set at 0.32 to give a confidence interval of twice the standard error; smaller values of c~ make the interval (and hence the standard error) sensitive to extreme data values. For each experiment, the root mean square of the standard errors of the Hodges-Lehmann estimates of responses at all concentrations of the test compound was calculated as an overall (pooled) estimate of the error. This allowed calculation of standard errors for the slope and intercept of the regression line. Following Finney (1971, p. 78) these lead to standard error limits log FDs0 d, log FDs0 + d, where standard errors of blank and control are also incorporated in d. For the FDs0, the limits are (FDs0 + c, FDso x c), where c = antilog (d) (Tables 1-4). In order to test whether the slope of the regression line differed from zero, it was divided by its standard error and the absolute value taken, to give a statistic z. A value of z greater than 1.65 indicated statistical significance, corresponding to the one-sided 5 % point for the normal distribution. Where a compound was tested twice, the requirement was at least one of the two z values should be more than 1.95; this gives an overall test with a nominal significance level of 5 %. For three tests, the requirement was similarly increased. Chemicals. The following compounds were obtained commercially.: VIII (Aldrich) XV and XVII (ICN Pharmaceuticals Inc.), XXVII (Eastman, Kodak Ltd.), XXXIV and XXXV (Calbiochem). L. Jurd, USDA, Albany, CA, provided compounds XXVIII-XXXI and O. Gottlieb, Instituto de Quimica, Universidade de Sao Paulo, Brazil, provided compounds XI and XXXII. The source 1717 FEE~,NO DETERRENTS HO H IoL A )1 c ~L,,,,H o. , R=. ~"YI [I. R=Me Ill. RO V. R=Me , [..,I H Me VH,. VII. gMe ~ e IX. R'= H x. R'=Me XI. of compounds I, IV, V, IX, X, XII-XIV and XVIII has been previously described (Sutherland et al., 1980). The isolation of VI, VII, XVI, and XX-XXIII as insect feeding deterrents and antifungal components from the roots of Lupinus angustifolius is the subject of a separate report (Lane et al., 1985). The following compounds were isolated or synthesized as described. 7-O-Methylphaseollin (I1). Phaseollin (1) was treated with ethereal diazomethane in MeOH. Crystallization from aq. MeOH gave II, mp 120-122~ Comparison of the UV, mass spectral, and [IH]NMR data with those reported for phaseollin (I) (Perrin et al., 1972) indicated the structure II. UV (EtOH) 1718 LANE ET AL. Xm,~207 (e 47,000), 230.5 (e 37,000), 279.5 (e 13,000), 285 sh (e 11,000), 315.5 nm (e 3,000); EIMS (probe, 70 eV) m/z (relative intensity) 336 (M +, 63), calcd 336.13615, found 336.13699, 322 (24), 321 (100), 293 (40), 160 (26); [IH]NMR (CDC13, 80 MHz) ~ 7.43 (1 H, d, J = 8.5 Hz, H-5), 6.93 (1 H, d, J = 8.1 Hz, H-6'), 6.62 (1 H, dd, J = 2.5, 8.5 Hz, H-6), 6.49 (1 H, d, J = 9.9 Hz, H-c0, 6.45 (1 H, d, J = 2.6 Hz, H-8), 6.32 (1 H, dd, J = 0.7, 8.0 Hz, H-5,), 5.55 (1 H, d, J = 9.9 Hz, H-r 5.48 (1 H, br d, J = 6.3 Hz, H-4), 4.2 (1 H, m, H-2), 3.77 (3 H, s, OCH3), 3.6-3.4 (2 H, m, H-2, H-3), 1.41 (3 H, s, CH3), 1.38 (3 H, s, CH3). (Numbering as for structure I). Phaseollidin (II1) and Kievitone (XXIV). Etiolated hypocotyls of Phaseolus vulgaris cv. Tendergreen were treated with conidial suspensions of Sclerotinia fructicola (Wint.) Rehm, Saccardo in 0.5 ppm cycloheximide and incubated 24 hr at 25 ~ Ethanolic extracts of freeze-dried hypocotyls were chromatographed on silica gel with CHC13-EtOH and petroleum ether-Et20. Phaseollidin (III), purified by repeated chromatography on Sephadex LH20 with CHC13 and CHCI3-EtOH (19 : 1) gave UV, EIMS, and [IH]NMR spectra in agreement with reported data (Perrin et al., 1972). Kievitone (XXIV) was purified by repeated chromatography on Sephadex LH-20 with CHC13-petroleum ether-EtOH (10 : 10 : 5). The UV, ELMS, and [~HINMR were in agreement with data reported by Smith et al. (1973). 2 '-Hydroxyformononetin (XIX). This was synthesized from 2,4-dihydroxyacetophenone and 2-hydroxy-4-methyoxybenzaldehyde as described by Farkas et al. (1974). The mp 220-222~ (lit. 220-221~ Farkas et al., 1974), UV, and [~H]NMR data were in agreement with reported values (Dewick, 1977; Braz Filho et al., 1977). ( + )-Medicarpin (Racemic XII) and ( + )-Vestitol (Racemic IX). Catalytic hydrogenation of XIX in HOAc at room temperature and atmospheric pressure over 10% Pd/carbon and column chromatography on silica gel with petroleum ether-Et20 afforded (+)-medicarpin (racemic XII), mp 198-200~ (lit. 194195~ Cocker et al., 1965), and (+)-vestitol (racemic IX), mp 173-175~ (lit. 173-175~ Farkas et al., 1974). The UV, MS, and [1H]NMR spectra were identical to those for (-)-medicarpin (XII) and (-)-vestitol (IX) respectively. (+_)-Methylequol (XXIO. Catalytic hydrogenation of XVIII in HOAc over 10% Pd/carbon (Nottle and Beck, 1974) gave XXV, mp 160-161 ~ (lit. 160~ Nottle and Beck, 1974). The MS and [IH]NMR data were in accord with those reported by Breytenbach et al. (1981). Dihydroformononetin (XXV1). Catalytic hydrogenation of the acetate of XVIII in acetone at room temperature and atmospheric pressure over 10% Pd/ carbon and mild hydrolysis with NaHCO3-MeOH gave XXVI, nap 197-198~ (lit. 197-199~ Breytenbach et al., 1981). The MS and [IH]NMR data were in agreement with those reported by Breytenbach et al. (1981). Epirotenone (XXXIll). Rotenone (VIII) was converted to the crystalline 1719 FEEDING DETERRENTS Xtl. XIII. R'= R"=H XIV. R'=Me, R"=OH XV. R'=R"=H XVI. R'=OH, R"=H XVII. R'=H, R"=Me xx.:. XXI. R=OH XXIII. XVlll. R=H XIX. R=OH xx XXIV. epimeric mixture "mutarotenone" by treatment with NaOAc in EtOH as described by Crombie et al. (1961). The mixture was separated by preparative recycle HPLC on a silica gel column with petroleum ether-EtOAc (5 : 2) using a Waters Prep-500A chromatograph, with refractive index detection. The separation was monitored by analytical HPLC on a 25-cm, 10#m silica gel column with hexane-EtOAc (5 : 1) with UV detection at 280 nm, and rotenone could not be detected in the purified fraction ( < 1%). Crystallization from MeOH gave XXXIII mp 88-91~ [cq D + 60 ~ (c 1.1, benzene) [lit. mp 90~ [c~]D + 75.6 ~ (c 0.6), Crombie et al., 1961]. The [~3C]NMR (CDCI~, 20 MHz) data for 1720 L A N E ET AL. HO"~/j....OMe xxv. R=H~ XXVI. FI=O XXVII. R'=OH, R'~ XXVIII. R'=H, R"=OH XXXI. XXIX. R=H XXX. R=OH HO.~O XXXII. L H HO/ ~ / --" 0....//~-.....~ 'i "OMe XXXIII. "~/'"OMe 6Me Me XXXIV. R=H XXXV. R=OMe MeC XXXVl. XXXVll, XXXIII are closely similar to those for rotenone (VIII) recorded under identical conditions: epirotenone (XXXIII), 6 188.9 (C-4), 167.3 (C-7), 157.9 (C-9), 149.5 (C-4'), t47.4 (C-2'), 143.9 (C-5'), 143.0 (-C=), 129.9 (C-5), 113.4 (C-10), 112.9 (C-8), 112.4 (=CH2), 110.5 (C-6'), 104.9 (C-6), 100.9 (C-3'), 87.9 (furanoid CH-O), 72.2 (C-2), 66.3 (CH2-O), 56.3 (OCH3), 55.8 (OCH3), 44.5 (C-3), 31.1 (Ar-CH2), 17.1 (-CH3); rotenone (VIII), 6 188.8, 167.3, 157.8, 149.5, 147.3, 143.8, 143.0, 129.8, 113.2, 112.9, 112.4, 110.5, 104.8, 100.9, 87.7, 72.2, 66.2, 56.3, 55.7, 44.5, 31.2, 17.0. The assignments for XXXIII are based on those reported by Crombie et al. (1975) for VIII, and the numbering used is that shown for I and VIII. The assignments are supported by multiplicities determined using a DEPT pulse sequence (Doddrell et al., 1982). The [IH]NMR (CDCI3, 80 MHz) data for XXXIII are also very similar to those reported for VIII (Carlson et al., 1973). 6 7.81 (1 H, d, J = 8.5 Hz, 5H), 6.77 (1 H, d, J = 0.6 Hz, 6'-H), 6.47 (1 H, d, J = 8.6 Hz, 6-H), 6.42 (1 H, s, 3'-H), 5.26 (1 H, br t, J = 8.9 Hz, furanoid CH-O), 5.01 (1 H, br s, FEEDING DETERRENTS 172! C = C H ) , 4.9 (2 H, m, C = C H , 2-H), 4.59 (1 H, dd, J 3.0, 11.9 Hz, 2a-H), 11.6 Hz, 2a-H), 3.81 (1 H, br d, J 4.1 Hz, 3-H), 3.77 4.14 (1 H, br d, J (3 H, s, OCH3), 3.74 (3 H, s, OCH3), 3.29 (1 H, dd, J = 15.8, 9.6 Hz, Ar-CH), 2.92 (1 H, dd, J = 15.9, 8.3 Hz, Ar-CH), 1.70 (3 H, s, CH3). The observed couplings are almost identical to those reported for VIII by Carlson et al. (1973), and any chemical shift differences are slight. The coupling of 4.1 Hz for 3-H, obscured by the OCH 3 signals in CDC13, was determined from a spectrum in pyridine-d 5 (Carlson et al., 1973). Homonuclear shift-correlated two-dimensional NMR experiments (COSY; Bax and Freeman, 1981) showed an identical pattern of connectivities for XXXIII and VIII, and closely similar resolved subspectra for the protons in the furanoid ring and those about the C/D ring junction. These data indicate the heterocyclic ring conformations of XXXIII and VIII are closely similar, despite the different relative configurations in the two compounds. Dihydroprecocene II (XXXV1). Catalytic hydrogenation of XXXV in HOAc at room temperature and atmospheric pressure over 10% Pd/carbon gave XXXVI, mp 62~ [lit 61-62~ Ahluwalia et al. (1982)] with [tH]NMR data in agreement with that of Ahluwalia et al. (1982). 6-Acetyl-5-hydroxy-2,2-dimethylchromene (XXXVII). This was prepared from 2,4-dihydroxyacetophenone and 3-hydroxy-3-methylbutanal dimethyl acetal in pyridine as described by Bandaranayake et al. (1971). The mp and [~H]NMR data were in agreement with those reported by Bandaranayake et al. (1971). = = ----- RESULTS Table 1 gives FDso values and slopes of dose-response lines, together with the standard errors, for compounds giving a significant correlation between dose and feeding response. The criteria for inclusion in Table 1 are that on at least one occasion: (1) the dose-response gave an FDs0 not more than 1000 ~g/g (five times the maximum concentration tested), and (2) the slope of the regression line was negative and significantly different from zero at the 5 % level. Different sets of values for a compound correspond to measurements on different days. The active feeding deterrents in Table 1 are listed in decreasing order of activity (as measured by the FDso), although the wide error limits of the FDso values imply that the ranking is approximate. No account has been taken of differences in the slopes of the dose-response lines. While these isoflavonoids, as a group, all show feeding-deterrent activity, the table spans a very wide range of activities. With FDso values of 0.02 #g/g and 0.03 #g/g, phaseollinisoflavan (IV) and phaseollin (I) are among the most active insect feeding deterrents yet recorded. The variation between the FDso values of compounds tested on dif- 1722 LANE ET AL, TABLE 1. EFFECT OF ISOFLAVONOIDS ON FEEDING OF Costelytra zealandica LARVAE: SIGNIFICANT DOSE-REsPONSE CORRELATION Feeding response Compound ( -)-Phaseollinisoflavan (IV) (-)-Phaseollin (I) (-)-Rotenone (VIII) (+)-2'-O-Methylphaseollin isoflavan (V) (-)-7-O-Methylphaseollin (II) LicoisoflavoneB (VI) (-)-Phaseollidin (III) (-)-Clausequinone (XI) (-)-Vestitol (IX) 2'-Hydroxyformononetin(XIX) 2'-Hydroxygenistein(XVI) (-)-Medicarpin (XII) (-)-Maackiain (XIII) (+)-Pisatin (XIV) Kievitone(XXIV) Luteone (XXI) FDs0 (t~g/g) c (• +)4 Slope of doseresponse line 0.02 0.03 0.06 4.7 2.2 3.3 -3.8 -6.3 -4.3 0.21 0.35 5.7 4.1 -3.1 -2.1 1.2 2.9 1.6 6.5 16 9.3 3.4 14 20 27 42 43 49 50 55 120 260 2.8 2.2 2.2 2.7 4.7 10 7.1 3.6 1.8 9.6 2.2 1.8 2.1 3.6 8.1 -3.7 -3.9 -5.8 -6.5 -6.0 NSb -4,7 -1.1 -3.8 NS -2.7 -10.2 -3.4 NS - 10 - 12 -7.8 -3.6 - 3.9 "The error limits are (FDso + c, FDs0 x c). bNS: slope not significantlydifferentfrom zero at 5 % level. All other slopes are significant. ferent occasions (compare Tables 1 and 3) reflects the inherent scatter of data in such a behavioral bioassay. The slopes of the regression lines given in Table 1 are given as fecal pellet count per log concentration. On average, the standard stimulated response was approximately 15 pellets. The slope values listed correspond to a reduction in feeding of between 4 and 20% of this average when the concentration of the compound is doubled. Isoflavonoids which failed the FDso or the slope criteria for feeding-deterrent activity over the concentration range are listed in Table 2 together with their feeding responses at the highest concentration (200 /~g/g). Compounds VII, XXII, and XXXI showed significant reduction in feeding at this concentration by the Wilcoxon's Rank sum test, and XXIII and X showed significant reduction in feeding at a lower concentration (Lane et al., 1985). These compounds, while FEEDING DETERRENTS 1723 TABLE 2. EFFECT OF [SOFLAVONOIDS ON FEEDING OF Costelytra zealandica LARVAE: NO SIGNIFICANT DOSE-RESPONSE CORRELATION Compound Feeding response" (%), 200 ~zg/g Licoisoflavone A (VII) (-)-Sativan (X) Genistein (XV) Biochanin A (XVII) Formononetin (XVIII) Wighteone (XX) Angustone A (XXII) Angustone B (XXIII) (+)-Methylequol (XXV) Dihydroformonorletin (XXVI) Coumestrol (XXVII) 7-Deoxy-5'-hydroxycoumestrol(XXVIII) Osajin (XXIX) Pomiferin (XXX) 4-Hydroxy-3-aryleoumarylquinone (XXXI) 67b 84 80 76 79 69 31b 74 98 175 86 77 76 110 48h ~Based on Hodges-Lehmann estimates, standard = 100. bSignificant reduction in feeding: P < 0.05 Wilcoxon's rank sum test. probably active, do not show the steady reduction in feeding with increasing dose over the range tested that is characteristic of the compounds in Table 1. Licoisoflavone A (VII) has consistently shown a mild depression in feeding at all concentrations, while for XXII and XXXI the highest dose may be at the beginning of the response range. In order to compare directly the effect of certain structural differences on the feeding response of C. zealandica larvae, sets of two or three homologous isoflavonoids were tested on the same day under the same conditions. The FDso values and slopes of dose-response lines for these comparative tests are given in Table 3. For comparison of active compounds in these sets, a parallel line analysis was used to check whether there were significantly different feeding responses and to provide a numerical measure of the difference. Parallel response lines were fitted to the Hodges-Lehmann estimates of the pellet counts for the compounds to be compared. The common slope was then checked for statistical significance. The Y-axis intercepts were compared using Fisher's LSD test (5% level), and the ratio of FDso values was calculated to give a measure of the relative activity. This analysis was not appropriate for comparisons involving compounds with an FDso > 1000 and/or a positive or nonsignificant slope of the dose-response line. All the compounds tested on this occasion (Table 3) gave similar results, TABLE 3. COMPARATIVE EFFECT OF SELECTED ISOFLAVONOIDS ON FEEDING OF Costelytra zealandica LARVAE Feeding response Compound FDs0 (/~g/g) ( - ) - M e d i c a r p i n (XII) ( - )-Phaseollidin ( l i d ( - ) - P h a s e o l l i n (1) 76 5.3 0.050 2'-Hydroxygenistein (XV1) Licoisoflavone A (VII) Lieoisoflavone B (VI) Genistein (XV) 2'-Hydroxygenistein (XVI) Formononetin (XVIII) 2'-Hydroxyformononetin (XIX) (• (XXV) (+_)-Vestitol (rac. IX) (_,• (rac. XII) ( + ) - V e s t i t o l (XXXII) (+)-Vestitol (rac. IX) ( - ) - V e s t i t o l (IX) C (• +) Relative activity" C ( x , +) 4.7 4.0 2.3 - 4 . 4 NS' - 5.1 - 8.1 -1 75 2.0 4.9 44 1.9 -5.1 - 3 . 4 NS -7.8 1 -68 2.2 -2.3 - 2 . 5 NS -6.0 --- -390 + 2 . 1 NS - 1.2 NS --- > 1000 370 NO -17 -- - 6.2 - 2 . 8 NS + 1.2 NS ---- 460 35 43 23 3.1 2.5 - 3 . 2 NS -6.4 -7,6 -1.6 NS 1 - 5,4 -8.5 380 320 2.7 > 1000 97 NO b 18 Epirotenone (XXXIII) Rotenone (VIII) 56 0,44 5.0 2.0 Epirotenone (XXXIII) Rotenone (VIII) 2,1 0.42 2.3 1.8 4.6 3.6 Epirotenone (XXXIII) Rotenone (VIII) Slope of d o s e response line 120 5.0 2.2 1 86 2.9 - 7.8 - 11 1 8.4 1.5 -3,8 -3.1 1 34 3.1 ~Inverse ratio of FDs0 values for pairs of active compounds as determined by the parallel line analysis, expressed as i : n. bNO: not obtainable, FD50 not possible with compounds of positive slope. ' N S : not significant, slope not significantly different from zero at 5% tevel; activity difference nol significant at 5% level, z FEEDING DETERRENTS 1725 TABLE 4. EFFECT OF SIMPLE CHROMENES ON FEEDING OF Costelytra zealandica LARVAE Feeding response Compound FDs0 (#g/g) C ( x , +) PrecoceneII (XXXV) Precocene I (XXXIV) 6-Acetyl-5-hydroxy-2'-2'-dimethylchromene(XXXVII) DihydroprecoceneII (XXXVI) 2.4 6.9 8.4 47 1.7 3.8 2.1 2.1 within the limits of error, to those obtained previously (Table 1 and 2). Thus, for example, phaseollidin (III) and phaseollin (I) gave comparable FDs0 and slope values, and (-)-medicarpin (XII) again gave a nonsignificant slope. Table 4 records the FDso values and slopes of three simple chromenes and a chroman. In all cases, the data showed these compounds to have feedingdeterrent activity. To simplify the discussion of each of the isoflavonoids, the isoflavone numbering system (see structure I) has been used throughout. Unless specifically indicated in the diagrams or in the text, the isoflavans and pterocarpans (reduced ring C compounds) have the same stereochemistry as phaseollin (I). The exceptions are (+)-pisatin (XIV) which has the opposite configuration to phaseollin at C-3 and C-4 and (+)-methylequol (XXV) which is a racemate. The isoflavanones, dihydroformononetin (XXVI) and kievitone (XXIV) Were also obtained as racemates (Wong, 1975). DISCUSSION The data included in this paper are the quantitative measures of a behavioral response of field-collected insects. In interpreting them as a whole, we have sought to identify trends. Some have emerged which are clearly supported by the statistical tests. Other apparent trends are less strongly supported by the analysis but still suggest a particular interpretation. We have attempted to find a common molecular basis for this pattern of results and consider a number of hypotheses in turn. Structural Correlations. Of 36 isoflavonoids, including optical isomers, tested for feeding-deterrent activity towards Costelytra zealandica larvae, 18 were active. Phaseollin (I) and related compounds with a cyclic isoprenoid unit fused to ring B, and rotenone (VIII) are particularly active with a significant effect on feeding at concentrations below 1/zg/g. The feeding deterrent activity of rotenone (VIII) is of interest as it is well known as an insect toxin. Feeding 1726 LANE ET AL. deterrent activity is not restricted to a particular isoflavonoid class. While all the pterocarpans tested (I-III and XII-XIV) showed feeding deterrent activity and neither of the coumestans tested (XXVII, XXVIII) showed activity, isoftavans, isoflavones, and isoflavanones occurred in both the "active" (Table 1) and "inactive," (Table 2) categories. Many of the isoflavonoids showing insect feeding deterrent activity are also antifungal (Sutherland et al., 1980, Hart et al., 1983). However, a study of lupin isoflavones (Lane et al., 1985) has shown that the two activities do not always coincide and appear to have different structural requirements. Thus, 2'-hydroxygenistein (XVI) shows insect feeding-deterrent activity but is not antifungal, while the antifungal prenylisoflavone, wighteone (XX), does not show insect feeding-deterrent activity. Perrin and Cruickshank (1969) suggested a stereochemical basis for the antifungal activity of pterocarpans, but this was later rejected by Van Etten (1976), who found the antifungal activity of 17 isoflavonoids tested could not be correlated with a common three-dimensional shape. Whatever the structural requirements for the antifungal activity of isoflavonoids, further evidence that they are not identical with those responsible for insect feeding-deterrent activity is provided by the results reported here for 7-O-methylphaseollin (II) and rotenone (VIII). Both of these compounds show very high feeding-deterrent activity (Table 1), but both lack antifungal activity (R.A. Skipp, personal communication; Georgopoulos, 1977). Two structural features that are characteristic of most of the highly active feeding deterrents are: (1) the presence of a 2'-oxy function, and (2) the occurrence of a 2.2-dimethyl-l,H-pyran (cyclic isoprenoid group) fused to ring B (I. II, IV, V). All the active isoflavonoids in Table 1 contain an oxy function at the 2'-position of ring B. Conversely, none of the 2'-deoxy compounds tested was active (Table 2). When pairs of homologous compounds were compared, the 2'deoxy compounds, formononetin (XVIII), genistem (XV), methylequol (XXV), and wighteone (XX) were inacnve (Table 2), whereas their 2'-hydroxy analogs, XIX, XVI, IX, and XXI, were active (Table 1). Direct comparative assays of several of these pairs proved less clear-cut (Table 3). The 2'-deoxy compounds genistein (XV), formononetin (XVIII), and methylequol (XXV) were again clearly inactive, and the difference between genistein (XVt and its 2'--hydroxy homolog (XVI) was confirmed. However, a difference between formononetin (XVIII) and 2'-hydroxyformononetin (XIX) and between (_+)-methylequol (XXV) and (_+)-vestitol (racemic IX) could not be unequivocally established as the 2'-hydroxy compounds on this occasion did not show a statistically significant dose-response slope. The cyclic ether homolog of (:t:_)-methylequol (XXV), (+)-medicarpin (racernic XII) was also inactive, although (-)-medicarpin (XII) is a marginally active feeding deterrent (Table 1). These results highlight the difficulty of comparing inactive compounds with those of marginal activity, but the balance of evidence supports the view that the difference in activity between 2'-deoxy compounds and their 2'-oxy counterparts is real. Several reactive or FEEDING DETERRENTS 1727 marginally active compounds do contain a 2'-oxy moiety (X, XXVII, XXVIII, XXII, XXIII, XXX1, VII), and the contrast between the inactivity of the coumestans (XXVII, XXVIII) and the active pterocarpans has already been noted. Thus, while 2'-oxygenation is a feature of all the active feeding deterrents, ,its occurrence does not always correlate with activity. The second important structural feature is the presence of a ring B-fused cyclic isoprenoid moiety (2,2-dimethyl-l,H-pyran). The significance of this feature is highlighted by the high absolute activity of phaseollin (I) and its homologs and by the high relative activity of compounds I and VI compared to compounds III, XII and VII, XVI respectively (Table 3). Phaseollin with a cyclic isoprenoid group fused to ring B is much more active than (-)-medicarpin (XII) and 75 times more active than phaseollidin (III) with a noncyclic isoprenyl group. A similar tendency is seen when 2'-hydroxygenistein (XVI) is compared with the complex isoflavones containing isoprenyl (VII) and cyclic isoprenoid (VI) groups. Licoisoflavone B (VI) containing a ring B cyclic isoprenoid moiety is 68 times more active than 2'-hydroxygenistein (XVI) and the analog with a noncyclic isoprenyl group (VII) gave a nonsignificant dose response. The occurrence of a cyclic isoprenoid moiety on the isoflavonoid molecule is not always associated with feeding-deterrent activity. Thus, angustone B (XXIII) shows only marginal signs of activity, and the ring A isoprenoids, osajin (XXIX) and pomiferin (XXX) are both inactive. On the other hand, of the highly active compounds, only rotenone (VIII) does not have a cyclic isoprenoid moiety fused to ring B, but rather, it contains an isoprenoid-derived furano group attached to ring A. The activity of this complex isoflavonoid is discussed below. The high activity of the complex isoflavonoids prompted us to investigate the feeding deterrent activity of some simple 2,2-dimethylchromenes (Table 4). These compounds can be considered as partial structures to the ring B cyclic isoprenoid isoflavonoids. Precocene I (XXXIV) and precocene II (XXXV) are of particular interest as they have been found to act as antijuvenile hormones against a range of insects by cytotoxic action on the corpora allata (Bowel"s, 1976, 1981). The chromenes XXXIV, XXXV, and XXXVII all showed appreciable feeding deterrent activity, while the chroman XXXVI was less active. The chromene encecalin (analog of XXXIV, XXXV with R = COCH3) has been reported as a feeding deterrent to Heliothis zea (Proksch and Rodriguez, 1983). The feeding-deterrent activity of these compounds may reflect structural similarities to the cyclic isoprenoid isoflavonoids or may have an independent basis. Modes o f Action: Reactive Centers. In considering and interpreting the pattern of results discussed thus far, we have not found any direct parallels with previous studies of insect feeding deterrents. Oxygen functionality and oxidation state have been found to be significant factors in studies of the insect feeding-deterrent activity of phenolics. Elliger et al. (1980) found the inhibition of growth of Heliothis zea larvae caused by the presence of a range of flavonoids 1728 LANEET AL. in their food was associated with ring B ortho-dihydroxylation. Norris (1977) found that the presence of a carbonyl and an adjacent hydroxyl in compounds such as 5-hydroxynaphthoquinone, quercetin, and kaempferol gave high feeding inhibition toward Scolytus multistriatus. He concluded that there was a strong positive correlation of antifeeding activity with the degree of oxidation state of ring C of the flavonoids and also with the intramolecular hydrogen-bonding capability of the molecule. Any correlation between the oxidation state of the isoflavonoids and the feeding response of Costelytra zealandica does not follow the same pattern, since it is the reduced isoflavans and pterocarpans which are more active than the highly oxidized isoflavones, coumestans, and quinones. Sites for intramolecular hydrogen bonding are available around the carbonyl region in isoflavones, and it could be argued that this factor accounts for the higher activity of the 2'-hydroxyisofiavones (VI, XVI, XIX, XXI). However, such a mechanism cannot account for the activity of the 2'-hydroxyisoflavans (IV, IX) and pterocarpans (I-III, XII, XIII). Moreover, the presence of a strongly intramolecularly hydrogen-bonded 5-hydroxy group in genistein (XV) and its homologs (VI, VII, XVI, XVII, XX-XXIII) does not seem to confer any special increase in feeding-deterrent activity. Therefore, we conclude that our results with C. zealandica feeding are not explained by the relative location of oxygen functionality alone. Reactive centers of the electron donor-acceptor type in close proximity, as typified by the intramolecular hydrogen-bonding sites discussed above, have been identified as the key structural feature for the activity of quinoid and flavonoid (Norris, 1977) and isodon diterpenoid (Kubo and Genjian, 1981) feeding deterrents. The mode of action of such compounds as juglone, polygodial, and warburganal has been attributed to the interaction of the reactive electrophilic center with the sulfhydryl site of a receptor macromolecule (Norris, 1976; Singer et al., 1975, Ma, 1977). If such binding to sulfhydryl sites were important in the gustatory response of C. zeaIandica, then the quinone isoflavonoids (XI, XXXI) would be expected to be especially deterrent, but this was not observed. Further, such structural features are not evident in the most highly active isoflavonoids (I, II, IV-VI, VIII). In Situ Activation. If isoflavonoid feeding deterrents are not characterized by the presence of reactive electrophilic centers, they may be precursors of reactive species generated in situ which could react with a receptor macrotmolecule. A number of possibilities are suggested by the literature. Recently, !Bakker et al. (1983) have reported the photoactivation of several pterocarpan ~pbytoalexins to free-radical species which deactivated the enzyme glucose-6phosphate dehydrogenase in an in vitro assay system. Other photoactive compounds such as furocoumarins have been reported as insect feeding deterrents (Yajima et al., 1977). The pertinence of a photoactivation mechanism to resistance to root-feeding larvae is not clear. In any case, the pattern of photoactiv- FEEDING DETERRENTS 1729 ities reported by Bakker et al. (1983) [pisatin (XIV) > phaseollin (I); medicarpin (XII) inactive] differs from the pattern of feeding deterrent activity for these compounds with C. zealandica (Table 1), and there was no exposure of materials to direct sunlight in the feeding assay procedure. While a photoactivation mechanism seems unlikely, the involvement of free radical species in the insect response to isoflavonoids cannot be excluded. The reported oxidation of the cyclic isoprenoid moiety of precocene (XXXIV) to a reactive epoxide in insect corpora allata (Pratt et al., 1980) indicates another mode of bioactivation of these molecules. The reduced activity of dihydroprecocene II (XXXVI) compared to precocene II (XXXV) (Table 4) suggests that the presence of the reactive double bond favors feeding-deterrent activity, but this feature is clearly not essential for activity. The oxidative formation of reactive ortho-quinone-methide species has been invoked by Jurd et al. (1979) to explain the pattern of insect toxicity and chemosterilant activity of a range of benzylphenols. A similar process has been suggested to account for the anti-juvenile hormone activity of a prenyl phenol analogous to precocene II (XXXV) (Bowers et al., 1982), and valence isomerization to the corresponding quinone-methide has been suggested as a mode of activation of the precocenes (XXXIV, XXXV) (Bowers, 1981). Chemical evidence has recently been obtained for the formation of para-quinone-methides from 7-hydroxyflavans (Attwood et al., 1984). Of the isoflavonoids studied here, those with a benzylic proton ortho or para to a phenolic hydroxyl, or with a fused 2,2-dimethyl-l,4-pyranyl ring, can be considered as quinone-methide precursors. However, the presence or absence of these structural features alone does not correlate with the observed pattern of feeding-deterrent activity (e.g., compare XVI, Table 1; and XX, Table 2). Stereochemistry. From this discussion it is clear that reactivity factors such as intramolecular hydrogen bonding, redox potentials, and the formation of reactive species could be envisaged as accounting for the feeding-deterrent activity of some of the active isoflavonoids but not in a consistent way. If we make the assumption that there is a common basis to the observed pattern of structure-activity relationships, it cannot be found in such reactivity considerations alone. This has led us to consider the stereochemical similarities of the compounds, in particular phaseollin (I) and rotenone (VIII). These are both highly active deterrents, have few reactive centers, and have rather rigid polycyclic structures with relatively few stable conformations. However, rotenone (VIII) does not contain a ring B cyclic isoprenoid moiety characteristic of the other highly active feeding deterrents, and the alignment of the 2'-oxygen is quite different from that in the pterocarpans. Further, the absolute stereochemistry at C-3 is opposite to that for (-)-pterocarpans such as phaseollin (I) (Wong, 1975). In a study of the bifunctional binding of rotenone with mitochondrial NADH 1730 LANE ET AL. dehydrogenase, Yaguzhinskii and Kolesova (1975) concluded that the polar ring B and the carbonyl group determine the interaction of the molecule with the binding site. That this binding is stereospecific is indicated by the fact that epimeric compounds are inactive, which the authors attributed to the inability of the carbonyl to lie in the proper region at the phase boundary. The nature of the furanoid moiety on ring A seemed to be less specific. Rotenone can thus be considered as a nonplanar molecule with lipophilic and polar ends and containing a central dipole (Figure 2). This stereospecific enzyme binding model raises the possibility that there might be a stereochemical basis for the observed pattern of feeding deterrent activity, albeit less specific than for NADH dehydrogenase inhibition. We suggest that the common feature for the active feeding deterrents is that they have, or can adopt, a similar nonplanar molecular shape with a similar arrangement of polar and lipophilic groups. The molecular structures of phaseollin (I) and rotenone (VIII) have been the subject of both X-ray crystallographic (DeMartinis et al., 1978; Arora et al., 1975) and NMR studies (Pachler and Underwood, 1967; Perrin et al., 1972; Carlson et al., 1973) which indicated a consistent shape in both the solid and solution phases. Both Dreiding models and computer generated plots (PLUTO; Motherwell, 1976) (Figure 2) suggest that when the molecules are appropriately aligned their stereochemistry is cornRotenone FIG. 2. Projections of molecular structures of rotenone and phaseollin from X-ray crystal coordinates. FEEDING DETERRENTS 1731 parable. The ring A of phaseollin can be aligned with the ring B of rotenone so that the dihydrofuran oxygen (2'-oxygen) occurs within the same region as the important carbonyl of rotenone. The cyclic isoprenoid group is then in a similar region to the dihydrofuran ring of rotenone and the phenolic hydroxyl of phaseollin in a similar region to the two methoxyl groups of rotenone. This comparison suggests a basis for the importance of the isoflavonoid 2'-oxy function, which could serve a similar role to the rotenone carbonyl in binding to a macromolecule, and the importance of the cyclic isoprenoid group on ring B in defining a lipophilic region of the molecule. Of the active chiral isoflavonoids initially tested (Table 1), all the isoflavans and pterocarpans, with the exception of ( +)-pisatin (XIV) are of the same absolute stereochemistry as ( -)-phaseollin and are able to adopt a similar relative arrangement of 2'-oxygen and aromatic rings. (+)-Pisatin (XIV) can be aligned similarly to rotenone with the angular hydroxyl (on C-3) in the vicinity of the carbonyl group. The inactive 2'-deoxyisoflavones (XV, XVII, XVIII, XX) and isoflavan (XXV) cannot be aligned with a similar arrangement of groups, nor can the planar coumestans (XXVII, XXVIII). The 2'-hydroxyisoflavones can be aligned to some extent by rotation of ring B out of the plane of rings A and C, and most of these compounds are active. The high activity of licoisoflavone B (VI) may be associated with the possibility of aligning both the 2'-hydroxy and the cyclic isoprenoid group similarly to phaseollin. In the inactive prenyl homolog (XXIII), the molecule no longer has defined polar and lipophilic ends. The contrast between the activity of vestitol (IX) and the inactivity of sativan (X) may be associated with the steric effect of the 2'-O-methyl, as 2'-O-methyl phaseollin isoflavan (V) also appears to be appreciably less active than its 2'-hydroxy counterpart (IV) (Table t). The activity of the simple chromenes (XXXIV, XXXV, XXXVII) can be rationalized if they are considered to be partial phaseollin-like structures corresponding to the ring B and cyclic isoprenoid regions. Thus, the stereochemieal binding model does appear to be consistent with the observed pattern of results discussed so far. Some feeding assays were carried out, using available stereoisomers, to investigate the importance of stereochemistry to feeding-deterrent activity of C. zealandica larvae. The activity of rotenone was compared on several occasions with that of epirotenone (XXXIII) which has the opposite configuration at the junction of tings C-D. The NMR data (Methods) indicate an essentially enantiomeric relationship between rings C and D of rotenone and epirotenone and hence an opposite sense of folding. The results (Table 3) show that epirotenone is significantly less active than rotenone, although the difference is considerably less than for enzyme inhibition (epirotenone 1000-fold less active, Yaguzhinskii and Kolesova, 1975). Further, (+)-vestitol (XXXII) appeared to be less active in a comparative assay with (-~-vestitol (IX) and synthetic racemic (+)-vestitol, although this difference was statistically not significant. Thus stereochemical considerations appear to be a factor in isoflavonoid feeding deterrent activ- 1732 LANE ET AL. ity, and the relative activity of these stereoisomers can be related to how closely they approximate a phaseollin-like shape. CONCLUSIONS Kojima and Kato (1981) concluded in their study of clerodin feeding deterrents that a definite steric environment around an active binding center is required for high activity. Similarly, the active isoflavonoids (including rotenone) can be considered as defining, in terms of stereochemistry, polarity, and lipophilicity, the environment of the key oxygen. Further elucidation of the stereochemical and possible reactivity factors involved in the feeding-deterrent activity of isoflavonoids awaits the availability of suitable synthetic analogues. The sensitivity of C. zealandica larvae to phaseollin, its analogs, and rotenone is remarkable. While our results suggest a common stereochemicat basis for this phenomenon, the biochemical mechanism of this behavioral response remains to be established. The ability of these root-feeding larvae to detect and avoid low concentrations of the insect toxin, rotenone, normally found in roots of several tropical legumes (e.g., Dalbergia and Tephrosia spp.) is intriguing for its evolutionary implications. C. zealandica is indigneous to New Zealand flora, although pterocarpans and isoflavans are present (Briggs et al., 1975; Ingham, 1983). The wide differences in activity we have found for isoflavonoid feeding deterrents of the same structural class emphasize the need for quantitative methods and detailed analysis in determining the ecological role of plant secondary metabolites. Thus, phaseollin is likely to play a significant role in defense against insect attack at much lower plant concentrations than the simple pterocarpan, medicarpin, and the insect resistance of legumes may relate to the structures of the endogenous isoflavonoids. This raises the possibility of modifying the isoflavonoid biosynthetic pathway to enhance the resistance of a susceptible legume toward the feeding of C. zealandica larvae. Acknowledgments--We would like to thank Dr. L. Jurd and Professor O. Gottlieb for the gift of samples, Mrs. C. Ford for the collection of grass grubs, and H. Wright, D. Gardiner, R. M. P, Beeler, P. Dittmer, W. J. Greenfield, and A. Ngan for technical help. REFERENCES AHLUWALIA,V.K., JOLLY,R.S., and BALA,S. 1982. Synthesis of precocenes. Chem. Ind. (London) 36%370. ARORA,S.K., BATES,R.B., GRADY,R.A., and DELFEL,N.E. 1975. Crystal and molecular structure of the one to one complex of rotenone and carbon tetrachloride. J. Am, Chem. Soc. 97:57525755. ATTWOOD,M.R., BROWN,B.R., LISSETER,S.G., TORRERO,C.L., and WEAVER,P.M, 1984. Spec- FEEDING DETERRENTS 1733 tral evidence for the formation of quinone methide intermediates from 5- and 7-hydroxyflavonoids. J. Chem. Soc. Chem. Commun. 177-179. BAKKER,J., GOMMERS,F.J., SMITS, L., FUCHS, A., and DE VRIES, F.W. 1983. Photoactivation of isoflavonoid phytoalexins: Involvement of free radicals. Photochem. PhotobioL 38:323-329. BANDARANAYAKE,W.M., CROMBIE,L., and WHITING, D.A. 1971. 3-Hydroxy-3-methyM,l-dimethoxybutane, a new reagent for dimethyl chromenylation: Synthesis of lonchocarpin, jacareubin, evodionol methyl ether and other chromens. J. Chem. Soc. (C) 811-816. BAX, A., and FREEMAN,R. 1981. Investigation of complex networks of spin-spin coupling by two dimensional NMR. J. Magn. Reson. 44:542-561. BOWERS, W.S. 1976. Anti-juvenile hormones from plants: Chemistry and biological activity, pp. 129-142, in G.B. Marini-Bettolo (ed.). Natural Products and the Protection of Plants. Elsevier, Amsterdam. 846 pp. BOWERS, W.S. 1981. How anti-juvenile hormones work. Am. Zool. 21:737-742. BOWERS,W.S., EVANS,RH., MARSELLA,RA., SODERLUND,D.M., and BETTARINI,F. 1982. Natural and synthetic allatotoxins: Suicide substrates for juvenile hormone biosynthesis. Science 217:647-648. BRAZFILHO, R., GOTTLIEB, O.R., DE MORAES,A.A., PEDREmA,G., and PINHO, S.L.V. 1977. The chemistry of Brazilian Myristicaceae IX. Isoftavonoids from Amazonian species. Lloydia 40:236-238. BREYTENBACH,J.C., VAN ZYL, J.J., VAN"DER MERWE, P.J., RALL, G.J.H., and Roux, D.G. 1981. Some photochemical and oxidative conversions of pterocarpans and isoflavans: Functional requirements for cyclization of isoflavans to pterocarpans. J. Chem. Soc. Perkin 1 2684-2691. BRIGGS, L.H., CAMBIE, R.C., and MONTGOMERY,R.K. 1975. New Zealand phytochemical survey. 13. Constituents of the wood and bark of Sophora microphylla and S. tetraptera. N.Z.J. Science 18:555-558. CARLSON, D.G., WEISLEDER, D., and TALLENT, W.H. 1973. NMR investigations of rotenoids. Tetrahedron 29:2731-2741. COCKER, W., McMURRY, T.B.H. and STANILAND,RA. 1965. A synthesis of demethylhomopterocarpin. J. Chem. Soc. 1034-1037. CROMBIE, L., GODIN,EJ., WHITING, D:A., and SIDDALINGAIAH,K.S. 1961. Some chemistry of the B/C-ring system of rotenoids. J. Chem. Soc. 2876-2889. CROMBIE,L., KILBEE,G.W., and WHITING,D.A. 1975. Carbon-13 magnetic resonance spectra of natural rotenoids and their relatives. J.. Chem. Soc. Perkin 1 1497-1499. DEMARTINIS, C., MACKAY,M.F., and POPPLETON, B.J. 1978. The crystal and molecular structure of (-)-phaseollin. Tetrahedron 34:1849-1852. DEWICK, RM. 1977. Biosynthesis of pterocarpan phytoalexins in Trifolium pratense. Phytochemistry 16:93-97. DODDRELL, D.M., PEGG, D.T., and BENDALL, M.R. 1982. Distortionless enhancement of NMR signals by polarization transfer. J. Magn. Reson. 48:323-327. ELLIGER, C.A., CHAN, B.C., and WAISS, A.C. 1980. Flavonoids as larval growth inhibitors. Naturwissenschaften 67:358-360. FARKAS,L., GOTTSEGEN,A., NOGRADI,M., and ANTUS, S. 1974. Synthesis of sopborol, violanone, lonchocarpin, claussequinone, philenopteran, leiocalycin and some other natural isoflavonoids by the oxidative rearrangement of chalcones with thallium (III) nitrate, a~ Chem. Soc. Perkin 1 305-312, FINNEY, D.J. 1971. Probit Analysis, 3rd ed. Cambridge University Press, Cambridge. 333 pp. GEORGOPOULOS,S.G. 1977. Development of fungal resistance to fungicides, pp. 439-495, in M.R. Siegel and H.D. Sisler (eds.). Antifungal Compounds, Vol. 2. Marcel Dekker, New York. 674 Pp. HART, S.W., KOGAN,M., and PAXTON,J.D. 1983. Effect of soybean phytoalexins on the herbivorous insects Mexican bean beetle and soybean looper. J. Chem. Ecol. 9:657-672. 1734 LANE ET AL. HOLLANDER, M., and WOLFE, E.A. 1973. Nonparametric Statistical Analysis. Wiley, New York. 503 pp. HUTCttlNS, R.F.N., SUTFIERLAND,O.R.W., GNANASUNDERAM,C., GREENFIELD, W:J., WILLIAMS, E.M., and WRIGHT, H.J. 1984. Toxicity of nitro compounds from Lotuspedunculatus to grass grub (Costelytra zealandica) (Coleoptera: Scarabaeidae). J. Chem. Ecol. 10:81-93. INGHAM, J.L. 1983. Naturally occurring isoflavonoids (1855-1981). Fortsch. Chem. Org. Naturst. 43:1-363. JURD, L., FYE, R.L., and MORGAN, J. 1979. New types of insect chemosterilants. Benzylphenols and benzyl-l,3-benzodioxole derivatives as additives to housefly diet. J. Agric. Food Chem. 27:1007-1016. KOJIMA, Y., and KATO, N. 1981. Stereocontrolled synthesis of clerodin homologs--a synthetic approach to structure activity relationships. Tetrahedron 37:2527-2538. KUBO, I., and GANJIAN, I. 1981. Insect antifeedant terpenes, hot tasting to humans. Experientia 37:1063-1064. LANE, G.A., SUTnERLAND,O.R.W., and SKIPP, R.A. 1985. Isoflavonoids as insect feeding deterrents and antifungal components from roots of Lupinus angustifolius. Unpublished. MA, WEI CHUN. 1977. Alterations of chemoreceptor function in army worm larvae (Spodoptera exempta) by a plant derived sesquiterpene and by sulfhydryl reagents. Physiol. Entomol. 2:19% 207. MOTHERWELL, S. 1976. PLUTO. A programme for plotting molecular and crystal structures. University Chemical Laboratory, Cambridge, England. NORRIS, D.M. 1976. Physicochemical aspects of the effects of certain phytochemicals on insect gustation. Symp. Biol. Hung. 16:197-201. NORRIS, D.M. 1977. Role of repellents and deterrents in feeding of Scolytus multistriatus, pp. 215230, in P.A. Hedin (ed.). Host Plant Resistance to Pests. American Chemical Society, Washington, D.C. 286 pp. NOTTLE, M.C., and BECK, A.B. 1974. Urinary sediments in sheep feeding on oestrogenic clover. Ill. The identification of 4'-O-methyl equol as a major component of some sediments. Aust. J. Agric. Res. 25:509-514. PACHLER, K.G.R., and UNDERWOOD, W.G.E. 1967. A proton magnetic resonance study of some pterocarpan derivatives. Tetrahedron 23:1817-1826. PERRIN, D.R., and CRUICKSHANK, I.A.M. 1969. The antifungal activity of pterocarpans towards Monilinia fructicola. Phytochemistry 8:971-978. PERRIN, D.R., WHITTLE, C.P., and BATTERHAM,T.J. 1972. The structure of phaseollidin. Tetrahedron Lett. 1673-1676. PRATT, G.E., JENNINGS, R.C., HAMNETT, A.F., and BROOKS, G.T. 1980. Lethal metabolism of precocene I to a reactive epoxide by locust corpara allata. Nature 284: 320-323. PROKSCIf, P., and RODRIGUEZ, E. 1983. Chromenes and benzofurans of the Asteraeeae, their chemistry and biological significance. Phytochemistry 22:2335-2348. RUSSELL, G.B., SUTHERLAND,O.R.W., HUTCHINS, R.F.N., and Christmas, RE. 1978. Vestitol: A phytoalexin with insect feeding-deterrent activity. J. Chem. Ecol. 4:571-579. SILVERSTEIN, R.M. 1979. Enantiomeric composition and bioactivity of chiral semiochemicals in insects, pp. 133-146, in F.J. Ritter (ed.). Chemical Ecology: Odour Communication in Animals, Elsevier, North Holland, Amsterdam. SINGER, G., ROZENTAE,J.M., and NORRIS, D.M. 1975. Sulfhydryl groups and the quinone receptor in insect olfaction and gustation. Nature 256:222-223. SMITh, D.A,, VAN ETTEN, H.D., SERUM, J.W., JONES, T.M. B~FEMAN, D.F., WILLIAMS, T.H., and COFFEN~D.L. 1973. Confirmation of the structure of kievitone, an antifungal isoflavanone isolated from Rhizoctonia-infected bean tissues. Physiol. Plant Pathol. 3:293-297. SUTHERLAND, O.R.W. 1971. Feeding behaviour of the grass grub Costelytra zealandica White (.Coleoptera: Melolonthinae) 1. The influence of carbohydrates. N.Z.J. Sci, 14:18-24. FEEDING DETERRENTS 1735 SUTHERLAND,O.R.W., HOOD, N.D., and HILLIER, J.R. 1975. Lucerne root saponins a feeding deterrent for the grass grub, Costelytra zealandica (Coleoptera: Scarabaeidae). N.Z.J. Zool. 2:93-100. SUTHERLAND, O.R.W., RUSSELL, G.B., BraGS, D.R., and LANE, G.A. 1980. Insect feeding-deterrent activity of phytoalexin isoflavonoids. Biochem. Syst. Ecol. 8:73-75. SUTHERLAND, O.R.W., HUTCH~NS, R.F.N., and GREENFIELD, W.J. 1982a. Effect of lucerne saponins and Lotus condensed tannins on survival of grass grub Costelytra zealandica. N.Z.J. Zool. 9:511-514. SUTHERLAND,O.R.W., HUTCHINS,R.F.N., RUSSELL, G.B., LANE, G.A., and BIGGS, D.R. 1982b Biochemical plant resistance mechanisms: An evaluation of basic research, pp. 245-254, in K. E. Lee (ed.). Proceedings of the 3rd Australasian Conference on Grasslands Invertebrate Ecob ogy. S. A. Government Printer. Adelaide. 402 pp. VAN ETTEN, H.D. 1976. Antifungal activity of pterocarpans and other selected isoflavonoids. Phytochemistry 15:655-659. WONG,E. 1975. Isoflavonoids, pp 744-800, in J.B. HARBORNE,T.J. MABRY, and H. MABRY(eds.). The Flavonoids. Chapman and Hall, London, 1204 pp. YAGUZmNSKII, L.S., and KOLESOVA,G.M. 1975. A mode of the bifunctional binding of rotenone and piericidin with mitochondrial NADH dehydrogenase. Biokhimiya 40:456-460. YAnMA, T., KATO, N., and MUNAKATA,K. 1977. Isolation of insect antifeeding principles in Orixa japonica Thunb. Agric. Biol. Chem. 41:1263-1268.