Differences in the electroantennal responses of apple- and hawthorninfesting races of Rhagoletis pomonella to host fruit volatile compounds

Transcription

1 Chemoecology 8: (1998) /98/ $ Birkhäuser Verlag, Basel, 1998 Differences in the electroantennal responses of apple- and hawthorninfesting races of Rhagoletis pomonella to host fruit volatile compounds Jürg E. Frey 1, Jeffrey L. Feder 2, Joanne Palma 3 and Guy L. Bush 3 1 Department of Zoology and Weed Science, Swiss Federal Research Station, CH-8820 Wädenswil, Switzerland 2 Department of Biological Sciences, University of Notre Dame, Notre Dame, IN 46556, USA, jeffrey.1.feder.2@nd.edu 3 Department of Zoology, Michigan State University, East Lansing, MI 48824, USA Summary. Domestic apple (Malus pumila)- and hawthorn (Crataegus sp.)-infesting races of Rhagoletis pomonella, Walsh (Diptera: Tephritidae) provide an excellent model to examine the role that host plant specificity plays during sympatric speciation (i.e., divergence in the absence of geographic isolation). Previous work has shown that these races differ in their propensities to accept apple and hawthorn fruits in behavioral choice assays, and that this discrimination translates into host fidelity in the field (i.e., apple flies tend to mate on and oviposit into apples and hawthorn flies on hawthorns). We present the results of a study examining possible physiological factors contributing to host choice differences in R. pomonella. We tested whether apple and hawthorn flies differ in their electroantennogram (EAG) responses to biologically relevant volatile compounds emitted from apples and hawthorns. Significant differences were found in the relative EAG responses of apple and hawthorn flies to host fruit compounds at five of six paired study sites across the eastern United States. The geographic pattern of EAG variation was complex, however, with local populations of apple and hawthorn flies tending to be more similar to one another than to flies of the same race at distant sites. This pattern was largely due to EAG responses for several compounds showing longitudinal or latitudinal clines, the latitudinal clines being similar to those observed for allozyme loci in the host races. We also found evidence for sex-related differences, as males tended to have higher mean EAG responses to compounds than females. Host-associated differences were therefore nested within geographic and sex-related differentiation in R. pomonella. Further behavioral studies are needed to distinguish whether the EAG differences are responsible for, as opposed to being a consequence of, host-plant fidelity and adaptation. Crosses are also required to establish a genetic basis for the EAG responses, although we did find significant correlations between EAG scores for several compounds and the allozymes NADH-Diaphorase-2 and Hydroxyacid dehydrogenase at one of Correspondence to: J. L. Feder the study sites. Questions therefore remain concerning the evolutionary significance of the EAG response differences between apple and hawthorn fly races. Nevertheless, these differences raise the possibility that antennal responses to fruit-related volatile compounds contribute to host plant discrimination in R. pomonella. Regardless, the EAG responses represent another set of traits, in addition to diapause/eclosion time phenotypes and allozyme frequencies, differing between apple and hawthorn host races of R. pomonella. Key words. Rhagoletis pomonella (Diptera: Tephritidae) apple maggot fly sympatric speciation olfaction host fruit odor electroantennogram Introduction Models of sympatric speciation proposed for certain phytophagous insects posit a central role for host plant associated mating as a pre-zygotic isolating mechanism in lieu of geographic barriers to gene flow (Bush 1966, 1969, 1975). These models hypothesize that gene flow is initially reduced following a host shift due to a direct relationship between host plant discrimination and mate choice. The essence of the argument is that adults tend to mate and oviposit on the same species of plant that they fed on or within as juveniles. Such host fidelity establishes a system of semi-autonomous insect demes assortatively mating on different plants, facilitating the evolution of additional host-associated adaptations involved in survivorship and/or performance that can eventually lead to speciation (Bush 1969, 1975). Domestic apple (Malus pumila)- and hawthorn (Crataegus sp.)-infesting populations of the apple maggot fly, Rhagoletis pomonella provide an excellent model to examine the issue of host discrimination and its relevance to population divergence. Rhagoletis pomonella has been controversial ever since Benjamin Walsh (1867) proposed that the fly s shift from hawthorn, its native host, to apple represented an example of an incipient sympatric speciation event. Subsequent studies have confirmed the status of apple and

2 176 J. E. Frey et al. CHEMOECOLOGY hawthorn flies as genetically differentiated and partially reproductively isolated host races (Feder et al. 1988; McPheron et al. 1988), the hypothesized first step in sympatric speciation (Bush 1969, 1975). Six allozyme loci display significant allele frequency differences between apple and hawthorn-fly populations across eastern North America (Feder & Bush 1989; Feder et al. 1990a; see materials and methods section for details). In addition, mark-release-recapture experiments have demonstrated that differences in host preference and in the timing of adult eclosion combine to reduce genetic exchange between the fly races to 6% per generation (Feder et al. 1994). Behavioral work has further underscored that apple and hawthorn flies differ in their propensities to accept apple and hawthorn fruits (Prokopy et al. 1988; Luna & Prokopy 1994; Feder et al. 1994). Host discrimination therefore exists in R. pomonella and is restricting gene flow between apple and hawthorn-fly races, as predicted by models of sympatric divergence. But what is the physiological basis for host acceptance differences between the races? Olfaction is a logical choice for a trait involved in host discrimination. Previous studies have shown that adult R. pomonella flies are strongly attracted to specific odors emanating from host fruit (Prokopy et al. 1973). Volatile esters isolated from whole fruit extracts have been shown by electroantennogram (EAG) assays and behavioral observations to be important in eliciting the response to fruits (Fein et al. 1982; Carlé et al. 1987; Averill et al. 1988; Frey et al. 1992). For instance, baiting fly traps with volatiles isolated from apples resulted in significantly higher catches of flies in the field than unbaited traps (Prokopy et al. 1976; Reissig et al. 1982). Also, tests using either the active chemical fraction or synthetic host fruit volatiles present in apples resulted in the positive orientation of flies in olfactometer tubes, a wind tunnel and the field (Fein et al. 1982; Aluja & Prokopy, 1992). It is therefore clear that R. pomonella flies use chemical cues to find host fruits and that olfactory information transmitted from antennae play a significant role in this process. But just because R. pomonella flies use olfaction to locate fruits does not mean that the host races also use their sense of smell to distinguish between apples and hawthorns. In fact, within the volatile region found to be attractive to R. pomonella by Fein et al. (1982), the chemical profiles of apples and hawthorns are fairly similar, which could have facilitated the original shift of the fly to apple (Carlé et al. 1987). Nevertheless, certain chemical differences do exist between the two fruits (Carlé et al. 1987), affording the possibility that the fly races are exploiting these differences when deciding between fruits. As an initial test of the chemoreception hypothesis, Frey & Bush (1990) analyzed the EAG responses of apple and hawthorn flies to 9 compounds at 6 different concentrations each. These 9 compounds are known to contribute to apple and hawthorn fruit odors (Fein et al. 1982, Dimick & Hoskin 1983, Carlé et al. 1987). They found that apple and hawthorn flies differed in their EAG responses to several of these biologically relevant fruit volatiles at certain concentrations. It is therefore possible that the antennae of apple and hawthorn flies are tuned to chemical differences that exist between their respective host fruits. However, the sample sizes analyzed in Frey and Bush s (1990) study (n=15 apple flies and 6 hawthorn flies) were small for population level comparisons. Consequently, the findings should be viewed as cursory in nature. In addition, the apple flies that were tested came from a non-diapausing laboratory line originating in Geneva, N.Y., a line that had been in culture for at least 70 generations prior to the experiment. In contrast, the hawthorn flies came from a natural population near East Lansing, MI. Caution is therefore required in drawing any firm conclusions from this work as either the EAG responses in the apple race could have been altered by laboratory rearing or the reported differences between the host races could have been due to geographic, rather than host-plant associated variation. Here, we rectify the deficiencies of the previous EAG study by conducting a detailed survey of 6 paired apple and hawthorn fly populations collected throughout the eastern United States. We report finding quantitatively significant differences in the relative EAG responses of apple and hawthorn flies to chemical stimuli at 5 of the 6 paired study sites. Sympatric populations of apple and hawthorn flies therefore do significantly vary in their EAG responses to the volatile components of host odor. However, the geographic pattern of variation we observed was complex, with local populations of apple and hawthorn flies tending to be more similar to one another than to flies of the same race at distant sites. This pattern was due in large part to several of the compounds displaying significant longitudinal or latitudinal variation (i.e., clines). We explore the evolutionary implications of our findings to host recognition and sympatric race formation in the discussion section. Material and methods Collecting sites The flies analyzed in the EAG study were collected from 6 paired apple and hawthorn populations across the eastern United States (Fig. 1). Five of these sites were essentially sympatric, with host race populations separated by a distance of 1 km or less (Fig. 1). In fact, at the Grant, MI., Carlsville, WI. and Urbana, IL sites, apple and hawthorn trees were interspersed with one another in old fields. However, at Gas City, IN., infested fruits were collected from apple and hawthorn trees separated by a distance of 6 km, and should not be considered strictly sympatric due to potentially limited gene flow. At 5 of the 6 study sites (Amherst, MA., Gas City, IN., Okemos, MI., Carlsville, WI. and Urbana, IL.), flies were collected as larvae in infested host fruits in either the summer of 1986 or 1987 and were subsequently reared to adulthood in the laboratory (Fig. 1). These flies were allowed to reach sexual maturity in the lab, but were not exposed to host fruits during their lifetimes, and so were tested as naive adults. Naive flies were also used from the Grant, MI. site, but were sampled in a different manner than those from the other five sites. Here, flies were collected as newly eclosing adults in mosquitonet traps (tents) that we constructed beneath host trees in the summer of The tent-collected adults from Grant, MI. were transported back to the laboratory, where they were also allowed to reach sexual maturity before being tested.

3 Vol. 8, 1998 EAG responses in Rhagoletis 177 Fig. 1 Map of collecting sites for apple and hawthorn flies used in the EAG study. Also given are the year that samples were collected, distance in kilometers separating apple and hawthorn trees at sites (sym=sympatric, with trees interspersed in old fields and separated by distances 0.01 km), sample sizes of flies measured for EAG responses (af=apple race females, am=apple race males, hf=hawthorn race females, hm= hawthorn race males), and latitude and longitude of sites EAG measurements Antennal sensitivity was measured by electroantennogram (EAG) methods as described in Roelofs (1977) and modified by Frey & Bush (1990). In short, we used head preparations of flies fixed with the inner side of the right uncut antenna exposed to a split air stream flowing at a constant rate of 750 ml/min over the preparation. Volatiles were introduced to a fly s head space by opening different channels of an electronic valve (General Valve Corporation) connected to one of two streams of the split air-flow system. Small (0.5 2 cm), rectangular pieces of 1 Whatman filter paper were used as the carrier medium for compounds. A 25 l sample of a given concentration of test compound (diluted vol./vol. in light paraffin oil [Fluka]) was allowed to absorb into the filter paper and the filter paper was then placed inside of a disposable Pasteur pipette henceforth referred to as an application unit. Application units were connected into the air stream containing the electronic valve and the valve opened for 0.5 second to expose a fly to a brief pulse of the compound. Application units were used for a maximum of 15 times to avoid concentration loss. We tested the EAG responses of flies to the following compounds at the indicated vol./vol. concentrations relative to the paraffin oil: butyl-2-methyl butanoate at 10 5,10 4 (abbreviated B25 & B24); ethyl pentanoate at 10 3 (EP3); hexyl butanoate at 10 6,10 5,10 4 (HB6, HB5, HB4); nonanol at 10 6, 10 5 (NL6, NL5); nonanone at 10 6 (NN6). These five compounds are known to contribute to the odors of host fruits (Fein et al. 1982; Dimick & Hoskin 1983; Carlé et al. 1987) and the nine compound/ concentration combinations that we used were the same as those tentatively shown by Frey & Bush (1990) to elicit differential EAG responses between the host races. Hexyl butanoate, is abundant in both apples and hawthorns, while butyl-2-methyl butanoate, ethyl pentanoate and 2-nonanol are known to occur in apple odors. The last compound, 2-nonanone, is found in blueberries but does not occur in apples or hawthorn. It was included as an outgroup compound to test for EAG differences in chemicals not found in hosts normally used by R. pomonella. Every fly was tested for the entire series of nine stimuli three times. As there were generally only minor differences among these three runs, we included for analysis the one replicate having the highest signal-to-noise ratio. To correct for changes in the antennal responses during the experimental period and for possible differences in the quality of head preparations among flies, a n-propyl hexanoate standard was applied at a concentration of 10 4 at intervals of 90 seconds. The usual sequence for an EAG run was standard, HB6, NL6, NN6, B25, HB5, standard, NL5, B24, HB4, EP3, standard. The time between stimuli presentation was staggered for a long enough period to allow depolarization to return to baseline, thus limiting the potential for sensory adaptation between compounds. EAG responses were quantified using the ratio between the amplitude of the response to the compound and the amplitude of the response to the standard applied before and after that compound (Guerin & Visser 1980). Data analysis of EAG responses Statistical analysis of the EAG data is complicated by the fact that responses to different compounds and concentrations of compounds are not necessarily independent. [Odors are, in general, a complex blend of chemical compounds and most higher animals have a range of differentially specialized odor receptors (Bernays and Chapman, 1994). With the exception of pheromone receptors, odor receptors are generally not absolutely specific and each receptor can be stimulated by two or more related chemical compounds. As a result, the response spectra of different receptors may overlap considerably (Mustaparta 1984). The EAG measures the summed total of all the receptor cells stimulated by a compound, and cannot distinguish the composition of the population of receptors responding (Boeckh et al. 1965). Hence, the responses of a fly to two compounds are usually not completely independent but due, to varying degrees, to the two compounds stimulating similar underlying sets of receptors.] Treating the compound/concentration combinations as separate variables or tests therefore introduces the possibility of redundancy in the analysis (i.e., multicollinearity; Glantz & Slinker 1990). We therefore took 2 different approaches to analyzing the EAG data. One approach involved calculating Spearman rank correlation coefficients to test for significant differences between the host races in their EAG responses to single compound/concentration treatments

4 178 J. E. Frey et al. CHEMOECOLOGY Table 1 Compounds displaying significant EAG differences between sexes within host race populations at sites, as determined by Mann- Whitney U tests. Sample sizes (number of females and males scored) are given in Figure 1. Bolded asterisks indicate table-wide significance, as determined by sequential Bonferroni procedure. No compound showed a significant EAG difference between the sexes in either host race at the Carlsville, WI. or Gas City, IN. sites. Negative ( ) sign indicates that females had significantly lower EAG scores than males Site Race Compounds showing sex effect Amherst, MA. A HB4*, NL6**, NN6* H Grant, MI. A H B24 Okemos, MI. A EP3**, HB4*, NL5 H B25*, NL5**, NN6* Urbana, IL. A H NN6* *P 0.05, ** P 0.01, *** P (Apple flies were assigned a value of 0 for these correlations and hawthorn flies 1). Because of the possible confounding effects of sex, correlations were performed separately for females and males, and then combined into a common correlation coefficient between hosts by meta-analysis using the weighted estimation method of Hedges & Olkin (1985; also see Britten 1996). This common correlation was tested for significance using z-tests (Hedges & Olkin 1985), with table-wide levels set using the sequential Bonferroni method (Holm 1979; Rice 1989). The second approach involved stepwise logistic regression analysis performed on square root transformed EAG scores. Our philosophy was to use the regression analysis to discern whether EAG responses to subsets of compounds significantly distinguished (discriminated) apple- from hawthorn-fly populations at paired sites. Stepwise regression involves the sequential selection of candidate variables in the model based on how much independent information each one contains about the dependent variable, after allowing for the information contained in the other variables already included in the regression equation (Glantz & Slinker 1990). As a consequence, stepwise logistic regression usually avoids producing a regression model with serious multicollinearity, allowing for latitude in interpreting the meaning of the included variables (Glantz & Slinker 1990). This analysis also discounted any redundancy in our results stemming from testing the same compound at different concentrations. Differences in EAG responses to compounds among the six study sites were tested for significance by Kruskal-Wallis non-parametric analysis of variance by ranks. These tests were conducted separately for females and males within each host race. The geographic pattern of EAG variation was also investigated by Cluster analysis. This was done using the JMP 3.1 computer program (SAS Institute Inc., Cary, NC, USA) according to the hierarchical clustering method of Ward (1963), based on the averages (midpoints) of the mean relative EAG responses of females and males of each host race at sites. Patterns of spatial differentiation were further dissected by calculating Spearman rank order correlation coefficients between EAG scores and the latitude and longitude of sites (see Figure 1 for a listing of geographic information concerning the sites). Because of the possible confounding effects of sex and host plant, these correlations were derived separately for each sex within each race and then combined by meta-analysis into a single common correlation using the weighted estimation method of Hedges & Olkin (1985). The common correlations were tested for significance using z-tests. Allozyme analysis of EAG responses Flies from the Grant, MI. study site were also genetically scored for the allozymes NADH-Diaphorase-2 (DIA-2), Aconitase-2 (ACON-2), and Hydroxyacid dehydrogenase (HAD) using standard horizontal starch gel electrophoresis techniques (Berlocher & Smith 1983; Feder et al. 1989). DIA-2, ACON-2 and HAD represent 3 of the 6 allozymes that display significant allele frequency variation between apple and hawthorn flies (Feder et al. 1988, 1990a; McPheron et al. 1988; Feder & Bush 1989). These 3 allozymes map to 3 different regions of the R. pomonella genome; DIA-2 is located on linkage group I, ACON-2 on group II, and HAD on linkage group III (Berlocher & Smith 1983; Feder et al. 1989, Roethele et al. 1997). Significant linkage disequilibrium has been found between DIA-2, ACON-2 and HAD and other genetic markers (including the 3 other allozymes displaying host-associated differentiation) within, but not among, these 3 linkage groups in natural fly populations (Feder et al. 1990a, unpubl. data). DIA-2, ACON-2 and HAD are therefore good genetic indicators of variation within the 3 genomic regions currently known to differ between the host races. Spearman rank order correlations were calculated between the number of DIA-2 100, ACON-2 95 or HAD 100 alleles that flies possessed and their relative EAG responses to the 9 volatile stimuli (Note: Homozygotes for the indicated allele were scored as 2, heterozygotes as 1 and homozygotes for alternate alleles as 0). ACON-2 95, DIA and HAD 100 are the common alleles at these loci and are found at higher frequencies in the hawthorn than the apple population at the Grant, MI. site (Feder et al. 1988, 1990b, 1993). Because of the possible confounding effects of sex and allele frequency differences between the host races, rank order correlations were derived separately for each sex within each race and then combined by meta-analysis into a single common correlation using the weighted estimation method of Hedges & Olkin (1985). The common correlations were tested for significance by z-tests. Results Sex-related differences in EAG responses to olatile compounds Females and males differed in their EAG responses to single compound/concentration treatments. There was a general tendency within both apple and hawthorn fly races at sites for males to have higher relative EAG scores to compounds than females (Table 1). In 33 of 54 comparisons involving the apple race and 35 of 54 comparisons involving the hawthorn race, males had higher mean EAG scores than females (data not shown). The combined result (68/108) was significant at the P 0.01 level, as determined by a sign test. In 11 of these 68 instances (6 in the apple race and 5 in the hawthorn race), the EAG responses of males were significantly higher than those of females, as indicated by Mann-Whitney U tests (Table 1). In contrast, females did not have significantly higher EAG scores than males for any of the 9 stimuli in either host race at any of the 6 paired study sites (Table 1). The tendency for males to have higher EAG scores was also reflected in the mean EAG responses to compounds combined for all sites (Table 2). Males had higher mean EAG scores across sites for 7 of the 9 stimuli in the apple race (the only exceptions being HB6 & NL5) and 8 of 9 compounds in the hawthorn race (the only exception being EP3; Table 2). The combined result (15/18) was significant at the P 0.01 level, as determined by a sign test. The effects of sex therefore had to be considered when testing for significant EAG response differences between the host races. We accomplished this by either considering host-related differences separately for females and males or independently testing the sexes and

5 Vol. 8, 1998 Apple race Hawthorn race Compound Female (136) Male (63) Female (187) Male (120) B B EP HB HB HB NL NL NN EAG responses in Rhagoletis 179 Table 2 Mean relative EAG responses to volatile compounds standard error combined for all sites for apple and hawthorn race females and males. Sample sizes (number of flies scored) are given in parentheses following sex designation combining the results into a single common test statistic by meta-analysis. Host-associated differences in EAG responses to single olatile compounds Apple and hawthorn flies did not qualitatively differ in their EAG responses to any single compound. Both host races responded to all 9 stimuli that we tested. There was no instance of one race responding to a compound and the other race displaying no antennal sensitivity to the same compound. In addition, no compound diagnostically distinguished apple from hawthorn fly populations to the extent that flies from one race always had higher (or lower) EAG scores than those of the other race at a paired site. The distributions of EAG responses for all 9 stimuli overlapped between the races at all 6 sites. We illustrate the extent of this overlap for NN6, HB4 and B25 for females and males at the Amherst, MA. site in Figs. 2a f. As we discuss below, the EAG responses to all 3 of these compounds differed significantly between the host races at Amherst (Table 3). Nevertheless, the distributions of raw EAG scores still overlapped for these 3 compounds from a low of 36.3% for NN6 in females up to 83.8% for NN6 in males (Figs. 2a f). Significant quantitative differences were found in the EAG responses of apple and hawthorn flies to individual compounds. Twelve out of a total of 54 common Spearman rank correlation coefficients calculated by meta-analysis across the sexes indicated significant rank order differences in the EAG responses of apple and hawthorn populations at paired sites to host fruit volatiles (Table 3). These 12 significant differences were distributed among 5 different study sites (Amherst, MA., Carlsville, WI., Grant, MI., Okemos, MI., and Urbana, IL.) and involved every compound/ concentration combination with the exception of HB5 (Table 3). Four of these 12 tests (NN6, B25 & HB4 at Amherst, MA. and NL5 at Okemos, MI.) were significant on a table-wide basis, as determined by a sequential Bonferroni procedure (Table 3). There was a general tendency within both sexes at sites for the hawthorn race to have higher relative EAG scores than the apple race. Hawthorn flies had higher mean EAG responses than apple flies for 33 out of 54 comparisons in both of the sexes (data not shown). The combined result (66/108) was significant at the P 0.05 level, as determined by a sign test. Furthermore, hawthorn flies had higher mean EAG scores than apple flies for 9 of the 12 single compound tests displaying significant host-associated variation (Table 3). The trend for hawthorn flies to have higher responses is also graphically illustrated in Figures 2a-f, which present the distributions of relative EAG scores of flies to NN6, HB4 and B25 at the Amherst, MA site. Multicollinearity in EAG responses to different olatile compounds Spearman rank correlation coefficients calculated between pairs of different compounds at sites suggested the presence of substantial multicollinearity (non-independence) in the EAG data set. Table 4 presents the common correlation coefficients between pairs of compounds across the six study sites for females and males, as determined by meta-analysis. Twenty-nine and 21 out of a total of 36 tests were significant (P 0.05) for females and males, respectively (Table 4). Twenty-five of the correlations in females and 12 in males were significant on a table-wide basis (Table 4). Caution is therefore urged against over-interpreting the pattern of single compound tests displaying significance (or lack of significance) between the host races. These tests are not necessarily independent and may reflect a common underlying cause, such as different compounds triggering similar sets of chemoreceptors. Evidence supporting a functional connection among EAG responses to compounds could be found in the observation that matrices of pairwise rank order correlations were significantly correlated across sites and between the sexes (common rank order correlation between sites for females= s.e., P as determined by z- test; males= , P ; between sexes=0.881, P ). In other words, EAG responses to certain sets of compounds that were highly correlated (or uncorrelated) at one site or within one sex, displayed the same relative relationship across other sites and for the other sex. In particular, EAG responses to the compounds B25, HB5, HB6, NL6 and NN6 showed a high degree of association, as did EP3 with HB4 (Table 4). NL5 tended to display stronger

6 180 J. E. Frey et al. CHEMOECOLOGY Fig. 2 Percentages of female and male flies of apple and hawthorn-origin at the Amherst, MA. site having relative EAG responses to the compounds NN6, HB4 & B25 within the indicated ranges of scores. The mean relative EAG responses are given for each sex/host race class s.e., along with the percent overlap of the distributions between the apple and hawthorn races. % overlap was calculated as 100 i (A i H i )/ ( i A 2 i )( i H 2 i ), where A i =the frequency of apple flies having EAG responses within range i, and H i =the frequency of hawthorn flies having EAG responses within range i relationships with EP3 and HB4 than to the other compounds, while B24 was more closely tied to B25, HB5, HB6, NL6 and NN6 (Table 4). Stepwise logistic regression analysis of EAG responses Stepwise logistic regression (Table 5) substantiated the implications of the single compound tests that apple and hawthorn flies can be distinguished based on their EAG responses to host fruit volatiles. At 5 of the 6 study sites, EAG responses significantly discriminated the host origin of flies when females and males were separately tested (Table 5). The only exception was the Carlsville, WI. site, where the host affiliation of both females and males was indeterminate based on their EAG scores. The stepwise regression analyses allowed more leeway in interpreting the pattern of compounds differentiating the host races than the single compound tests, because the regressions discounted redundant information in the data sets. There were both similarities and differences in the compounds included in the stepwise logistic regressions of square root transformed EAG responses at different sites (Table 5). For example, at Amherst, HB4 and NN6 were important in both sexes in discriminating between the host races, at Gas City HB4, at Grant HB5, HB6, B24 and B25, and at Okemos HB6 (Table 5). There was therefore a tendency for EAG responses to hexyl butanoate, a compound found in both host fruits, to distinguish apple from hawthorn flies at sites. This statement should be qualified, however, by saying that this compound was also tested at different concentrations more often than any of the other compounds. When both sexes were considered together, nonanol (a compound more prevalent in apples) was included in the stepwise regressions at 3 of 4 sites displaying significant host-associated differentiation (Table 5). Other than hexyl butanoate and

7 Vol. 8, 1998 Site Compound Amherst Carlsvil. Gas City Grant Okemos Urbana n (108) (83) (42) (143) (47) (83) B B EP HB HB HB NL NL NN =P 0.05, 2=P 0.01, 4=P ; as determined by z-test of common correlation coefficient. Tests with P were significant on a table-wide basis EAG responses in Rhagoletis 181 Table 3 Compounds displaying significant EAG response differences between apple and hawthorn populations at paired sites, as determined by z-tests of common Spearman rank order correlation coefficients derived by meta-analysis from separate tests performed for each host race and sex. Positive (+) sign of correlation coefficient indicates that hawthorn flies had greater relative EAG responses to a compound than apple flies at a site, while negative sign ( ) indicates that they had lower responses. Total sample sizes (n=number of flies scored) at each site are given in parentheses in the first row of the table nonanol, however, there appeared to be no general pattern in the compounds differentiating the races across sites. Geographic ariation Significant differences existed in relative EAG responses to compounds among the six study sites, as indicated by Kruskal-Wallis non-parametric ANOVA conducted separately on females and males within each host race (Table 6). Twenty-four out of 36 tests indicated significant inter-site variation in EAG responses to individual volatile compounds at the P 0.05 level, and 14 of these tests were significant on a table-wide basis (Table 6). Only EP3 failed to show significant among site variation. Male apple flies also tended to display less geographic differentiation in EAG responses than other flies, as only NL5 and NL6 differed significantly for male apple flies (Table 6). Cluster analysis revealed that mean EAG scores for apple and hawthorn flies at paired sites tended to be more similar to each other than to geographically distant populations of the same host race (Fig. 3a). The degree of differentiation between the host races at paired sites was therefore generally lower than that among geographic populations of either race, the only major exception being the Amherst, MA site (Fig. 3a). EAG responses to several compound/concentration treatments significantly varied with latitude or longitude, as determined by common Spearman rank correlation coefficients derived by meta-analysis from individual correlations calculated separately for each race and sex (Table 7). Specifically, NL5 and NL6 showed significant relationships with latitude, while HB5, HB6 and NN6 significantly varied with longitude (Table 7). The results for NL5, NL6 and HB5 were significant on a table-wide basis. The clinal patterns for NL6 and HB5 are illustrated in Figs. 4a, b, respectively. Inspection of these graphs suggested that latitudinal and longitudinal variation in EAG responses were the major factors responsible for the clustering of apple and hawthorn populations at paired sites. Notice, for example, how the ranges of most EAG responses to NL6 and HB5 within sites either do not overlap or only partially overlap with those from other sites (Figs. 4a, b). To test this hypothesis, we performed separate cluster analyses for the compounds displaying (HB5, HB6 NL5, NL6 and NN6) and not displaying (B24, B25, HB4 and EP3) clinal variation. The results indicated that the tendency for paired apple and hawthorn populations to cluster together disappeared when the analysis was based on compounds not showing clinal geographic variation (compare Figs. 3a and b). Hostassociated differences in EAG responses to fruit-related volatile compounds are therefore nested within a complex pattern of latitudinal, longitudinal and sex-related differentiation within R. pomonella. Genetic analysis of EAG responses at Grant, MI The allozymes DIA-2 and HAD displayed significant relationships to the EAG responses of flies at the Grant site to certain stimuli, but ACON-2 did not (Table 8). DIA was negatively correlated to the magnitude of relative EAG responses of flies to B24, HB5, HB6 and NN6, while HAD 100 was negatively correlated to HB4 and NN6 (Table 8). The magnitude of the correlations were not high, ranging from to 0.317, which are considered moderate in meta-analysis (Britten 1996). In addition, only the result for DIA-2/NN6 was significant on a table-wide basis. Nevertheless, the genetic results suggested an association between the EAG responses to certain compounds and allozyme markers in two of the three genomic regions displaying host-associated differentiation in R. pomonella. Discussion Two major findings emerged from our study. First, apple and hawthorn flies displayed significant differences in their relative EAG responses to biologically relevant fruit odor volatiles. At 5 of 6 paired populations, the EAG responses of apple and hawthorn flies

8 182 J. E. Frey et al. CHEMOECOLOGY Table 4 Common Spearman rank order correlation coefficients between EAG responses of flies to pairs of compounds across the six study sites, as determined by meta-analysis. Correlations were calculated separately for females (above diagonal) and males (below diagonal). The Spearman rank order correlation between the female and male matrices was 0.881, P , as determined by a z-test Compounds B24 B25 EP3 HB4 HB5 HB6 NL5 NL6 NN6 B B EP HB HB HB NL NL NN =P 0.05, 2=P 0.01, 3=P 0.001, 4=P ; as determined by z-test of common correlation coefficient. Tests with P were significant on a table-wide basis within each sex were statistically distinguishable from each other (Table 5). EAG responses therefore represent another trait, in addition to diapause/eclosion time phenotypes (Feder et al. 1993, 1997a,b) and allozyme frequencies (Feder et al. 1988, 1990a; McPheron et al. 1988), that differ between the host races. Second, the pattern of EAG variation was complex, with host-associated differences nested within sex-related and geographic variation (Tables 1, 2, 6 and 7, Fig. 3a). Paired apple and hawthorn fly populations tended to have more similar EAG responses than geographically distant populations, as revealed by cluster analysis (Fig. 3a). This pattern was due to significant longitudinal and latitudinal correlations in EAG responses (Table 7), corresponding with allozymes showing latitudinal allele frequency clines (Feder & Bush 1989). There are 3 possible interpretations of the EAG results. The first is that EAG responses differ between apple and hawthorn fly races because they directly influence host plant choice. The second is that the EAG responses do not influence host choice, but show host-related differences due to pleiotropy or linkage with other traits under selection. The third possibility is that EAG responses are induced by the host fruit itself or some other environmental factor correlated with apples and hawthorns. If the first hypothesis is true and antennal sensitivities influence host plant choice, then it suggests that olfactory responses to the derived host apple evolved multiple, independent times in R. pomonella. The similarity of paired apple and hawthorn populations in the cluster analysis would be due to local apple-fly populations having originated from local hawthorn populations, resulting in them sharing common characteristics in their EAG profiles. The latitudinal and longitudinal clines are difficult to explain, however, as one would suspect that if the EAG responses affect host choice, then the pattern of these responses should be consistent across sites. But such geographic clines are possible if the chemical compositions of apple and hawthorn fruits change across the eastern United States. The problem with this scenario is that certain features of apple and hawthorn fruit chemistry would have to change in parallel with latitude, while others would have to co-vary with longitude. Furthermore, the chemical profiles of the two fruits would still have to remain different enough at sites for selection for olfactory differentiation to counteract gene flow occurring between local apple and hawthorn populations (Feder et al. 1994). If the second hypothesis is true and antennal sensitivities have little to do with host choice but are indirectly under host-plant associated selection, then geographic variation in EAG responses would again reflect a balance between selection and gene flow. Only in this case, the EAG response differences would be due to either pleiotropy (selection on the chemoreception system for a function other than host choice) or linkage between receptor loci and unrelated gene(s) experiencing host-dependent selection. Several lines of evidence are suggestive of pleiotropy or linkage. As we discussed above, the EAG responses to several compounds displayed significant latitudinal or longitudinal variation (Table 7). The EAG responses for two of the compounds (HB6 and NN6) showing clinal variation were also significantly correlated with the allozyme DIA-2 at the Grant, MI. site, while NN6 showed a significant relationship to HAD (Table 8). The allozymes DIA-2 and HAD display significant allele frequency differentiation between the host races (Feder et al. 1988, 1990a; McPheron et al. 1988) and have been associated with diapause-related traits in R. pomonella important in coordinating the life-cycle of flies with the fruiting phenologies (seasonalities) of host plants (Feder et al. 1993, 1997). DIA-2 and HAD also display a complex pattern of clinal geographic variation in R. pomonella (Feder & Bush 1989; Feder et al. 1990a, Berlocher & McPheron 1996). A linkage may therefore exist between EAG responses for certain compounds and developmental traits under host-associated and geographic selection. The third hypothesis is that environmental conditions are responsible for the pattern of EAG phenotypic variation. Under this scenario, the host races differ because environmental conditions faced by flies infesting apples and hawthorns are different. These differences affect the development of the chemorecep-

9 Vol. 8, 1998 Site Sex X 2 P df Compounds in regression Amherst T HB4 3 NN6 2 HB5 B25 EP3 MA. F HB4 2 NN6 2 HB5 M HB4 1 NN6 NL6 EP3 Carlsvil T B24 WI. F B24 HB4 M B24 Gas City T NL5 NL6 IN. F NL5 1 HB4 NL6 B25 M HB4 1 Grant T NL6 4 B24 3 HB4 2 NN6 2 B25 1 HB5 1 MI. F NL6 4 B24 4 NN6 3 HB4 1 B25 1 HB5 M HB6 1 B25 HB5 B24 EP3 Okemos T NL5 3 EP3 2 HB4 1 MI. F NL5 1 EP3 HB6 M HB6 1 Urbana T HB6 2 NL5 1 NL6 1 IL. F HB5 1 NL6 M NL5 1 NN6 1 EAG responses in Rhagoletis 183 Table 5 Results of stepwise logistic regression for square root transformed EAG responses. Models shown are those that gave the greatest discrimination between the host races at paired sites based on a chisquare (X 2 ) test statistic. Sex refers to whether the regression was performed on both sexes combined (T), or separately on females (F) and males (M). Probability level (P) and degrees freedom (df) are given along with the abbreviations for the compounds (Note: The order of compounds corresponds to their significance levels in the regressions) 1=P 0.05, 2=P 0.01,3=P 0.001, 4=P ; probability level for inclusion of compound in the regression tion system, resulting in host-associated variation. Paired populations tend to cluster in their EAG responses because environmental conditions experienced by apple and hawthorn flies at local sites are more similar than those faced by geographic populations of the same race. Gene flow does not diminish the extent of these EAG differences because the differences are induced by the host fruit environment itself. Clinal patterns of EAG variation for certain compounds could be explained by latitudinal and longitudinal differences in environmental conditions. Environmental effects could also account for the tendency of apple flies to Table 6 Geographic variation in EAG responses to volatile compounds among collecting sites, as determined by Kruskal-Wallis non-parametric analysis of variance by rank. Tests were conducted separately for each host race and sex. H values corrected for tied ranks and significance level for each test are given. Bolded asterisks indicate table-wide significance, as determined by sequential Bonferroni procedure Apple race Hawthorn race Compound Female Male Female Male B ** 11.2* B * ** EP HB HB * 19.4** HB6 12.6* ** 15.3** NL NL6 18.3** * NN *P 0.05, ** P 0.01, *** P 0.001, **** P , as approximated by chi-square distribution with 5 df have lower EAG responses to compounds than hawthorn flies at paired sites (Table 2). It would be particularly interesting if the lower antennal sensitivity of the apple race to volatile compounds translates into a reduced ability to detect host fruit odors. If true, then this could help explain why apple flies readily accept both apples and hawthorns as potential hosts in fruit choice experiments, while hawthorn flies generally shun apples and show strong preferences for hawthorns (Prokopy et al. 1988); apple fruits may possess a deterrent volatile compound in higher concentration than hawthorns which apple flies might not detect as efficiently as hawthorn flies due to their reduced chemoreceptive abilities. We must stress that we have no confirmatory evidence for this scenario, but offer it as an hypothesis. Nevertheless, this hypothesis could be tested in R. pomonella through reciprocal larval transplant experiments and, if true, would represent a unique example of the Hopkins host selection principle involving a combination of hypotheses 1 and 3 discussed above. Instead of the larval feeding environment imprinting a preference for a novel host, it would lower the discrimination ability of the adult, resulting in the broadening of the host range and the inclusion of a new plant previously avoided. We currently are unable to distinguish which of the 3 above hypotheses concerning the EAG responses, either alone or in combination, is correct. There are reasons to question whether the observed differences in antennal sensitivities between apple and hawthorn flies affect host fruit choice. EAG analysis records only the peripheral neural responses of insects to volatile compounds, revealing little about how this information is processed in the central nervous system. If host acceptance decisions are primarily made in the central nervous system, then our results may have little bearing on

10 184 J. E. Frey et al. CHEMOECOLOGY Fig. 3 Dendogram of the relationships among apple- (A) and hawthorn-fly (H) fly populations at sites, as determined by a Cluster Analysis based on mean relative EAG responses to volatile compounds. (a) Dendogram derived from the complete 9 compound/concentration data set. (b) Dendogram derived from B24, B25, HB4 and EP3, the 4 compounds not displaying geographic clines in EAG response among sites. Dendograms are drawn to scale according to the relative EAG distances between populations indicated at the branch nodes Table 7 Common Spearman rank order correlation coefficients between mean EAG scores for compounds at sites and latitude or longitude, as determined by meta-analysis from separate tests performed for each host race and sex. Bolded asterisks indicate tablewide significance, as determined by sequential Bonferroni procedure Compound Latitude Longitude B B EP HB HB HB ** NL NL NN * *P 0.05, ** P 0.01, *** P 0.001, as determined by z-tests the issue of host fidelity in Rhagoletis. In addition, fruit odors are complex blends of compounds present at different concentrations. We tested only single compounds at fixed concentrations in the current study, which may or may not reflect the chemical odor cues that flies are behaviorally responding to in nature. Nevertheless, EAG responses did differ between the host races, and so we cannot dismiss the possibility that they contribute to host discrimination. Further experiments are required to test for a direct relationship between EAG responses and host fruit acceptance behaviors. One way that this could be accomplished would be to determine whether EAG responses correspond to the readiness of flies to orient to apple versus hawthorn whole fruit extracts and volatile compounds in a wind tunnel assay. If antennal sensitivity is important in determining host fruit acceptance, then a relationship should exist between EAG scores and the preferences of flies for apple compared to hawthorn fruit extracts. These studies can then be augmented by solid-phase microextraction (SPME) and gas chromatography-electroantennogram detection (GC-EAD) techniques, as well as single sensillum recordings, to pin-point the exact chemical and physiological factors underlying any detected behavioral differences. Test crosses and reciprocal egg transplant experiments between apple and hawthorn fruits are also needed to show that the EAG responses are not developmentally labile traits affected by larval and pupal rearing conditions. We can rule out the scenario that the EAG differences we found were due to conditioning in adults, as we used naive flies in all of our tests. Imprinting on the central nervous system during the larval stage is also unlikely (Prokopy et al. 1988). But it is still possible that differences in environmental conditions that Rhagoletis larvae and pupae face influence the number and/or development of key components of the chemoreception system responsible for the EAG differences between the races. In conclusion, while the present results are intriguing, additional work is needed to clarify their importance to host plant specific mating in R. pomonella. Mark-release-recapture studies indicate that host fidelity exists for apple and hawthorn flies and reduces inter-host gene flow to 6% per generation (Feder et al. 1994). This and other studies suggest that inherent differences in host preference contribute to host fidelity (Prokopy et al. 1988; Luna & Prokopy 1994). Here, we found that differences in peripheral EAG responses to