Physiology end Genetkv ol Tme-Phptqdiage/n/ereclions Oulnn (Fr~nco),August 31 September 6, 1997 Ed. INRA, Paris, 1999 (LssCdloques, n.90) - I I . e 'IbMLlN RS., BORDWJ.H., 1997. Multicmponent index for evaluating mistma by Sitka spruce lo LCwhite NC-4501302 VMPR RBAIS: 1 . 7 1 3 No R e p r i n t s A v a i l a b l e pine weevil (Coleoptera: Curculionidac). J. Econ E n r o d . 90.704-714. ACHOWLEDGEMENTS The authors thank Pamela Cheers for editing the text and Diane Paqud for her word Resistance to galling adelgids varies among families of Engelmann spruce (Picea engelmani P.) processing work. WILLIAM J. MATTSON', ALVIN YANCHUK', GYULA KISS', BRUCE BIRR' , ' Forestry Sciences Laboratory, North Central Forest Research Sfatlon , 5985 Nlghway K, Rhlnelander, WI 5450 1 'Reseamh Branch, British Columbla Ministry of Forests , 1450 Government Stmet, Vlctorfa, B.C., Canada VBW 3E7 'Kalamalke Research StatEon, Brftlsh Columbia Minlstty of Fomsts 340 1 Resemlr Road, Vernon, B.C., V1B 2C7 RESUME Cootey gall adelgids, Adelges cooleyi, and round gall adelgids, Adelges abietis, differentially infested 110 half-sib families of Engelmann spruce, Plcea engelmannil at 9 study sites in British Columbia. There was a negative genetic correlation (-0.53) between the infestations of the two gall-fonlng specles. Cooley gall abundance exhibited a negative genetic correlation (-0.66) with tree growth, whereas round gall abundance exhibited a positive genetic correlation (0.79) with growth. Heritability (h') of resistance against the adelgids averaged about 0.60 for Cooley galls and 0.20 for round galls INTRODUCTION World-wide, there are 30 species of spruce (Picea spp: Pinaceae) which sewe as the primary host for the majority of the 49 or so specles of primitive aphids in the famlly Adelgidae (Carter 1971, Ghosh 1983). Each adelgid species typically causes a uniquely shaped gall to form on the developing shoots of susceptible spruces. Such galls can persist on the affected trees for many years depending on the species of gall- substantive family level variation in Engelmann spruce resistance to two common, maker and the species and environment of the host spruce, thereby leaving a useful index to a tree's general resistance. For example, in undisturbed plantations in northern shoot-galling adelgids: the nalive, Cooley spruce gall aphid, A. cooleyi, and the introduced (from Eurasia), round spruce gall aphid, A. abietis, both of which produce Minnesota, we have found galls that were formed as long as 30 years earlier (Mattson very distinctive galls (Rose and Lindquist 1977). We also investigated the heritability et al. 1994). (h') of such resistance and tested for genetic correlations between plant resistance to the two species of adelgids and between tree growth and resistance. We predicted that Usually, the impact of adelgids on tree growth and survival appears (a) tree resistance to one adelgid species would be positively correlatedwith resistance inconsequential. But, to be sure, no one has ever measured their long-termimpacts on to the second adelgid owing to trees employing the same or very similar resistance bud demography. Cooley galls, caused by Adelges cooleyl, for example, typically kill mechanisms against both species, and (b) tree resistance and growth would be their subjugated shoots. Round galls caused by A. abiells and A. lariciafus, on the other positively correlated because the most likely mechanisms of resistance would be rapid hand, do not kill the shoot but allow it to grow in an apparently substandard manner. inducible defenses, such as strong and swift hypersensitive reactions, that are Ragged spruce galls caused by Pineus sirnilis can cause hormonal imbalances in dependent on vigorous growth (Herms and Mallson 1992, 1997). young seedling stems, triggering a cork-screw growth-habit that predisposes them to snow damage. In rare cases, some highly susceptible lndivldual trees have nearly all of their shoots attacked by adelgids and consequently are rendered uncompetitive and METHODS eventually die (Mattson, pers. obs). In western North America where there are 15 species of adelglds, it is not uncommon to find galls from as many as 4 species on the This study was overlaid on a genetic trial utilizing 110 half-sib Engelmann spruce P. engelmannii, Sitka spruce, families originating from the east Kootenay region in S.E. British Columbia. Families P. siichensis, while spruce, P. glauca, and the vast hybrld swarm called interior spruce were planted in linear, 10 tree plots, replicated in 2-3 blocks at each of 9 different, widely (mostly P. engelmsnniix glauca) in interior British Columbia. separated geographic sites. Trees were about 15 years of age, and ranged in height same host plant, as is h e case for Engelmann spruce, from 2-5 m in September 1994 when measurements were made. The first 5 living trees Numerous field studies in Europe and North America have reported substantive, of all families in each replicate were scored for numbers of Cooley and round galls by consistent differences among individual trees in their resistance to the galling adelgids a pair of forestry technicians. Each tree was examined (one person on each side) for (e.g. see Mattson et al. 1994, Mattson et al. 1998). thereby implying that resistance has two, 30-second intervals (using a timer) during which Cooley galls were counted, and an important genetic component to it. For example, Mattson el al. (1998) estimated that then round galls counted. Such a brief inspection period was sufficient to provide a broad sense heritability (H') of resistance against Adelges abiefis was about 0.86 (the reliable difference among differently infested trees. For example, there was a strong theoretical maximum'being 1.0), using one clonal orchard of Picea abies in the southem linear regression relationship (p < 0.01, 1 = 50-75% between such 30 second counts Paris basin. Of course, such limited studies (i.e. one -population. of trees, one by forestry technicians and one minute counts (virtually 100% censuses) done on the environment, one population of aphids) give but only a glimpse of the possible genetic same trees by two highly experienced gall workers (wjm 8 bab). Usually, the 30 second bases of spruce resislance to adelgids. counts allowed the measurement of nearly all of the galls on lightly infested trees, and at least one-third to one half of them on the heavily infested trees. Tree heights (cm) This study was undertaken to further basic understanding about the genetics of and tree diameters (mm) were likewise measured at the same time. The round galls spruce resistance to insects. Specifically, we investigatedthe hypothesis that there is were initiated primarily by A. abiefis, but there may also have been a small fraction of 64 * RESULTS and DISCUSSION superficially similar roung galls caused by A. ladciatus, though its distribution in S.E.. British Columbia is not well substantiated. Because over 11,000 trees were to be examined, it was not economically feasible to attempt to definitively segregate these two species of round gall makers, if Indeed they occurred together. Therefore, h e round gall counts may represent cumulative infestations of two species of adelgids. Due to mlsundemtandings, the whole tree gall C O U ~ Were ~ S converted to a tlumefical index, and recorded as infestation classes: 0=0, 1=I-10, 2=11-20, 3=21-30, 4=31-40, 5=4l -99 Variation among sites and families Cooley galls were common at all 9 sites, but as expected, the\[ abundance differed substantially and significantly among siles (Tables 1.2). For example, (hey averaged about 7 galldtree at the least infested site, and 31 gallsftree at the most infested site. Analyses of variance confirmed that there were also substantive. significant (pe.001) differences among half-sib families (F) in their resistance to A. - 1 galls. coolevi at all 9 sites. According lo a variance components analysis on !he overall model For statistical analyses and tests, we used the numerical indices directly, but furthermore we converted them back into a semblance of the original data by using a random number generator to pick a gall number within the appropriate numerical range coveredby the index assigned to a tree. All such generated numbers (x) were then testing all of the sites together, lamily effecls accounted for 8.6 % of the total variation (Table 2). At the most heavily infested sites, the majority (>75 percent) of families were in middle to high range infestation classes, i.e. medium to light-heavy infestalion classes. Only very few families were in the extremes, i.e. in very light and very heavy converted to their log, (x+l) equivalent for meeting the assumptions of ANOVA. We infestation categories. This implies that nearly all families are moderately to highly used two ANOVA models. One was for analyzing separately each geographic locale susceptible to A. cooleyj. or site: X,= p + F,+ R, + FP, + e,, where F is family, R is replicate, and FR are family by replicate interaction effects, e is residual error among indlvlduals. The second was Round galls were much less common than Cooley galls at 6 of the 9 sltes, but + R, (S,) + F, + FISk + F, at 3 sites they were equally abundant (Tables 1.2). At the least infested location, round Because of computatlonal overload problems with an galls averaged about 2/tree, whereas at the most infested location they averaged about unbalanced design arising from missing observations, the pooled data set was 34ltree. Analyses of variance revealed that there were highly signilicant family effects manipulatedto create a perfectly balanced data set conslstlng of 82 families and three (p < .001) on round gall numbersltree at all 9 sites. According to a variance components individual trees per replicate (2) at all study sites. Hence, those statistics for the pooled analysis on the overall model, family effects accounted for 2.4% of the total variation study refer to this reduced data set, and consist o l the class counts for galls. Statistics (Table 2). At the most heavily infested siles, the majority of spruce half-sib families fell shown for the individual sites are based on the full set of 110 families, all replicates, and into the medium and heavy infestation classes. As was the case for Cooley gall all individual trees, but using the random counts. In every ANOVA, all effects were susceplibility, only few families fell into the very light and very heavy infestationclasses. considered random effects. To estimate the magnludes of genetic correlations (rJ, and Thus, almost all families are moderately to highly susceptible to !he round gall adelgids. for analyzing the pooled data set across all sites: X, = p + S, R, (S.) + em, where S is site. narrow-sense, individual heritabilities (h'), we computed the variance components for the main effects, and their interactions, and then followed standard estimation protocols Tree size effects on galling (Becker 1984, Falconer 1989, Kiss and Yanchuck 1991, Stonecypher 1992). Because larger trees have larger canopies and offer more potential growing points for adelgid infestations, and families varled significantly in growth rates, we plotted mean galldlreelfamily against family diameter (d.b.h.) and height to search for Trblr 1. Mean number of galls (log, (x + 1)) per tree at nine different research slles, and the any such potential relationships. Universally, there was a vety poor to negligible pmbablllty of a larger F value In randomized, complete black deslgn tesNng Me null hypolheses relationship between mean Cooley gall counts/treelfamily and mean family d.b.h. and that there are no family, repllcale, and family x replicate effects on gall counts per tree. Study sltes are ranked In ascending order according to the mean abundance of Cooley galls per tree. Cooky gallr: height (Table 2.). On the other hand, quite the opposite was true for round galls. Round gall numbers per family clearly increased linearly with mean family height (and diameter) at all study areas, ranging from about 1-3 gallslm of height at the most lightly infested Roundgalls: areas, to about 7-12 galldm of height at the most severely infested areas (table 4.). Study Sltes: mean Fam Rep FxR mean Fam Rep FxR E.WhlteREvsr 1,768 e.001 0.24 0.07 0.959 c.001 0.06 e.010 Lodge Creek 1,818 e.001 e.001 0.07 1.554 c.001 c.001 e.001 To remove the effects of family growth rate which is potentially confounded with Roche Creek 2.162 e.001 0.07 e.01 2.547 e.001 0.1 <.001 the inherent resistance/susceptibility of families, we divided galldtree by tree height (m) JumboCreek 2.613 e.001 r1.001 401 0.539 c.001 e.010 e.010 and then transformed the ratio (r) using a log, r +.01) transformation before running PerryCreek 2.733 e.001 c.001 e.001 2.689 e.001 0.02 e.001 analyses of variance. We did not employ height as a true covariate in ANOVA because the regression relationship of galls on height was not perfectly collinear among all 110 families. Hone Creek 3.173 e.001 0.41 0.12 1.338 <.001 0.47 0.04 WdwmereCk 3.233 e.001 c.01 e.001 1.755 4.001 0.247 e.001 Converting raw gall counts to galls per unit height did not remove family effects; they were still vety highly significant (p < 0.001) at all study sites (Table 3.). Although, standardizing galling by height did not eliminate famlly effects from the model, il did Table 2. Analysls of variance of data pooled from all sibs, uslng gal class counts, showing rearrange the individual infestation rankings of families, but largely within their original degrees of freedom (dl). Mean Squares (MS) and components of variance for the main effects quartiles. For example, most of those families ranking in the highest and lowest lor both Cooley and round gal analyses. infestationquartiles on the basis of gallsllree still occurred within the highest and lowest infestation quartiles, respectively, after slandardization by tree heighl. However, the Cooley OaHs Saum of Vertetion dl Sltss 8 MS 315.5'" Round GaIlr internal rankings of almost all these families changed relative to one another. For Varfsn~, COW. MS Vadance Comp. 0.62 S02.4' 0.99 example, at Bloom Creek. the mean change in round gall family infestation rank (old rank-new rank) was +2.8 for the highest infestationquartile. In other words, the average family increased in mean gall loading by about 3 ranks. For the lowest infestation quartile. the mean change in family rank was -3.5, meaning lhat rank dropped by more than 3 (i.e. lower average gall loading). Only 5 new families moved into the highest infestation quartile, the other 22 just changed their rank orders. The same was true for FxS the lowest infestation quartile. Therefore, we conclude that there are real and substantive family effects on galling that are confounded to only a small degree with the Error 2952 0.9 0.87 0.7 ' ' slgnlfknnt (pc0.001)effecls, * percentage of total variance 0.74 effects of family growth rates. Table 3. Statistics lor linear regressions of mean Cooley and round pall nrunt~ree/famlly Genetlc correlations between galling and tree growth versus mean lamlly tree helght (cm) at each of 8 sludy areas. If zero occurs In the slope columns, slopes are not slgnlflmn~fmm different from zero. Othemise, elopes are slgnlficantly (p < .05) larger than zero. Each regression Is based o n 2 216 obsew%tlons. To investigate possible genetic relationships between gall inlesletlons and tree growth rates, we calculated the genetic correlations (r9)between Cooley and round gall Cooley gals: Study Sltes BbomC& Roue gals: r" 54.0443 fitopa . 0.05 0,0954 infestations and height. at each of the nine study sites, and on the entire, pooled data r' set (Table 5). In the case of Cooley galls, all correlatlons were negative, ranging In value 0.17 from -0.02 to -.0.79. The pooled data set gave rv= -0.21. The consistent assoclatlon implies that the tree's traits for fast growth and resistance lo Cooley gall aphids are positively linked, as might be the case if resistance depended on rapid inducible Table 5. Genetic mnslatkms (rJ between adelgld gall hfeslalbm and tree gm* (helghtJ, and between Cwley (C) and round (R) galls levels per tree, and estimates of the heritability (h) of resistance against adelgldr using two measures of infeslallon. gall nohree and gall no.fm at each of 8 study altes in British Columbia. Each estimate In the table Is based upon measurements on approximately 1000-2000 trees, except lor the pooled sites. Study Sites Tttble 4. Mean number ot log, (galYmMer of tree helphl* 0.0 1) at dne dWennt research sltes. vs helght and the probablllv of a larger F value In randomized, complete block deslgn testing the null hypotheses lhal there are no lamlly, replicate, and family x repllcate effects on gall per unit of C,geNs Gemtk correlallons: Heritablllty (h') estimates: R.qsls RsellaYsC.aalls C.galls R.galls C.galls R.gslls no. Per no. per vs blght nohree noJm lree tree noJm nolm Bbom Creek -0.79 0.36 -0.51 -0.31 1.0+ 0.41 1.0+ 0.17 E. While River -0.49 1.0+ -0.54 -0.71 0.34 0.13 0.44 0.09 Horse Creek -0.28 1.0+ 0.03 -0.59 0.35 0.26 0.44 0.1 1 helght per tree. Study sltes are ranked in ascending order according to the mean abundance ol Cooley galls per tree. Cooley galls: Study Round galls: Sites: mean Fam Rep FxR mean Fam Rep FxR Jumbo Creek -0.42 0.84 -0.49 -0.61 0.32 0.26 0.39 0.21 E.WhlteRlver -0.018 e.OO1 0.09 0.03 -2.002 e.OO1 0.04 e.OO1 Lodge Creek -1.0+ 0.62 -1.0+ -1.0+ 0.28 0.05 0.32 0.00 Lodge Creek 0.199 e.00f e.OO1 0.018 4.592 e.001 e.001 c.001 tussler River 5.37 0.84 -0.48 -0.69 0.95 0.51 1.0+ 0.26 Perryereek -0.61 1.0+ -0.33 -0.69 0.83 0.12 0.85 0.00 JumboCreek 1.003 e.001 c.001 c.010 -3.217 c.001 e.001 c.010 RocheCreek -0.02 0.84 0.19 0.38 0.78 0.17 0.84 0.09 Peny Creek 1.006 e.001 e.001 c.001 0,954 401 0.03 e.001 Wdvmre -0.55 0.74 -0.93 -0.85 0.69 0.18 1.0+ 0.11 Creek Lussier Rivet 1.887 e.001 0.06 e.010 0.857 401 c.002 c.030 HoneCreek 1.839 e.001 0.31 0.49 -1,132 e.001 0.55 0.010 1 n.a. not available, because analysis was not done for the particular variable of concern. defenses (such as hypersensitive reactions) which are swiftest and strongest in t vigorously growing tlssues. On the other hand, for round galls quite the opposite was found. All correlations were positive, ranging from 0.36 to 0.84. The pooled r, estimate high-high I was 0.92. E3ecause round and Cooley gall resistance have different relationships to tree growlh, it implies that the resistance mechanisms against them are different. (.Family 76 Correlstlone Between Cooley and round gall adelgld8 To obtain a synoptic view of famlly resistance to both Cooley and round gall adelgids, we ranked each spruce famlly according lo its percent departure from the grand mean resistance level of all 110 families at each of the 9 study sites. For Coolsy gall--devlatlon from grand mean (%) example, at each locale, every family was rescaled for the two gall types using the following formula: (,m, gall, mean-local gall grand mean)/local gall, grand mean x 100, where means are from log, (x+l) data. These new variables, percent deviations, were Flgure 1. Plot showlng the mean round and Cooley gall Infestalbnranks of each of 1 10 spwce lamilies. using a larnity's average percent deviation from Vle grand mean round and Cooley gal Infestationlevel of all Ireos, at each of 9 study sites, In other words, each point is Ihe mean ol then used in a randomized block ANOVA, with study areas treated as blocks, to test 9 individual devlallons measurements for a single famlly, showing its tendency lo be higher, or again for family differences, lower than others with respecl to infestationsby round and Cooley gall adelgids. For example. Family 126 Is Identified In the bwer kfl(low-low)quadrant, exhibitingbebw average levels of both gelling adelglds. As before, there were hlghly slgnlficant (p <0.001) family effects on the rescaled measures of resistance to galling. To demonstrate the fammes' resistance relationships Genetlc correlations between Cooley and round gall lnfeslations to both Cooley and round gall adelgids, we plotted mean round gall rank against mean Cooley gall rank for all 110 families (Figure 1). The result was an obvious left to right Using the log, (x+l) gall counts, we found that there were negative genetic sloping scatlergram. Such a pattern implies a negative genetk correlation between the correlations (-0.33 to -0.93) between Cooley and round gall infestalions per tree at 7 of resistance traits for these two adelgids. Sixty-eight percent of the families fell into the 9 sites (Table 5). At 2 sites {he correlalions were near zero or positive (0.03 and 0.19). upper left (27 percent) and lower right (41 percent) quadrants of the graph. Only a meager 11 percent of the families occurred in the lower left quadrant where Using the entire data (all sites) set in the pooled model gave a substantial, negative genetic correlation of -0.53. susceptibility to both species of adelgids is below the grand mean. Therefore, typically. when one resistance trait Is high, then the olher is low, and vice versa. Contrary to this Estimating heritabilities of reslstance to edefgfds trend, only one family, 1-126, had substantially low levels of both adelgids: its infestations were on average 23 percent below the grand mean for Cooley galls, and 32 Heritability estimates depend on the environment in which they are measured. percent below the grand mean for round galls. At the other contrary extreme, family f- Ihe particular genetic structure of the populations, and a bevy of other factors. 80, had \he highest combined level of both adelgids: its infestations were on average Therefore, we computed herilabilily eslimates for each of the 9 study locales, realizing about 20, and 23 percent higher than the respective grand means for Cooley and round that insect population size and ils genetic structure may vary substanlially among them. galls. For Cooley adelglds, h' values ranged from 0.28-0.95 wilh one estimate exceeding 1.0, would make breeding for resistance against both insects more difficult as very few owing to estimation errors (Yable 5). The pooled estlmate and i s standard error over all genotypes would be segregating for positive attributes for all three traits (1.e.. growth and sites was 0.61 5 0.1 1. Cooley and round gall adelgid resistance ). , For round gall adelglds, hevalues ranged from 0.05-0.51 (Table 5). The pooled estlmate and its standard error over all sites was 0.20 & 0.05. * Adjustments to the data lo accounl for differenlial attack among large and small trees, suggested that tree size per se does not largely change the probability that a tree will be attacked and that family pedigree Is more important. Ideally one should know the Because each heritability estimate was usually based on about 1000 individual impact of galling aphids on tree growth which would then permit an assessment of the tree measurements, derived from 110 families, the precision of the estimates are reasonably good. For example, the sampling variances of h2 estimates are benefit of developing lines of trees resistant to gall forming adelgids. Unfortunately, nothing is really known about the impact of such gall formers on tree survival and approximately 32 x ht/lOOO (Falconer 1989), and thus its standard errors are growth. approximately (sqrt (32 x h')) 1 1000. Although we only calculated standard errors for the overall h' estimates (after Becker 1984), it is evident that heritabllities were significantly larger than zero. Understanding of the genetics of tree resistance lo phytophagous insects is in its infancy. More research is clearly needed in order to be able to effectively employ natural mechanisms of plant resistance against potential tree pests. CONCLUSIONS REFERENCES Heritability estimates varied widely across individual sites, but the overall estimates confirmed that h' was substantially larger ( -0.6) for Cooley galls, than for round galls (-0.2). This may suggest lhat there is more genetic variation present for Cooley gall resistance, but the lower level of attack by round gall adelgids may have caused the heritability estimates to be less than those for Cooley gall resistance. Not surprisingly, for both gall adelglds there was a tendency for the highest heritability estimates to be linked to the sites with highest insect densities, as has been observed spp elsewhere by Strong el el. (1993) for a cecidomyiid leaf-galling fly on willow, In Sweden. Nevedheless, these values Indicate there are good levels of genetic variation in reslstance to both gall-forming Insects on Interior spruce, and breeding for resistance would be possible. However, relationships between height growth and resistance to both gall forming insects were quite different. Cooley gall abundance was negatively correlated wilh height growth which suggests faster growing families are more resistant. Whereas, for round gall adelgids, faster growing famllies are more susceptible. This BECKER W.A., 1984. Manualof quanIitalive genetics, Wash, State Unlv, Press, Pullman. CARTERC.I., 1971. Conifer woolly aphids (Adelgldae) In Britaln. Foresfv Comm. BUN.42: 1-4 1 FALCONER D.S., 1989. Infmdmfbn to quanfilative genelks. John Wiley, N.Y., 438 pp. Gtiosti A.K., 1983. A review of the family Adelgidae from the Indian subregion ( Homoplera: Aphidoidea). Orienfal Insects 17: 2-29. HEUMS D.A. and MATTSONW.J., 1992. The dilemma of plants: to grow or defond. Quad. Rev. Biot 67: 283-335. HERMS D.A., and M AW.J.. 1997. ~ Trees, stress, end pests, pp. 13-25. In: Plan! health cam for wody arnrrmenlals,.Unlv. Ill.Coop. Ext. Sawlce, U&ana-Champaign, 223 p. KISS Q.K. and YANCW A.D., 1991. Preliminary evaluation of genetk control of weevil resistance of Interfor spruce in British Colombia. Can J. For. Res. 21: 230-234. MATTSON W.J., BIRR B.A.. and IAWUENCE R.K.. 1994. Variation In the susceptibilityof North Amerkan white spruce populallons to the gall-krming adelgid, Adelges abielis (Homoplera: Adelgidae) pp. 135-147, In Price, P., Mattson, W.J., and Baranchlkov, Y, (eds.) The ecology and evolution of gall-fomlng Insects. USDA Forest Sew. GTR NC-174, St. Paul, Mn., 222 p. Mrcnsm W.J. LEVIEUX J., and PIOUD., 1998. Genetk and environmental contributions to variation In the resistance of Pkea abhs to the gall-tormlngadelgid, Adejges abietis (Homoplera: Adelgidae). pp. 304-315, In Csoka, G.. Mattson. W.J.. Stone. G.N.. and Price. P.W. (eds.) The biology of gallforming arfhmpods, USDA For. Sew. GTR NC-199, St. Paul, Mn 303 p. ROSEA.H. and. CIND~VIST, O.H., 1977. Insects of eastem spruces, flr, and hemlock. For. Tech. Rept. 23, Ministry of Supply and Servkes, Ottawa, 167 pp. S~floFIaD.R., LARSSON S. and Q U L L ~ EU., R ~1993. Heritability of host plant resistance to herbivory changes with gallmfdge density during an outbreak on willow.. Evolution 47: 291-300. S~mEcvrneflR.W., 1992, Computational methods, pp, 195-228, in Fins, L., Friedman, S.T., and Brotschol, J.V..(eds.) Hendbook of quanfiteflve forest genefks. Kiuwer Academlc Publishers, London. 403 p. Physiology and Genetics of Tree-Phytophage Interactions International Symposium Gujan (France) August 31 - September 5, 1997 Physiology and Genetics of Tree-Phytophage Interactions International Symposium Gujan (France), August 31 - September 5, 1997 organized by lnstitut National de la Recherche Agronomique (INRA) International Union of Forestry Research Organizations (IUFRO) INSTITUT NATIONAL DE LA RECHERCHE AGRONOMIQUE 147, rue de I'Universite - 75338 Paris Cedex 07 \.