Heredity 64 (1990) 371-375 The Genetical Society of Great Britain Received 11 October 1989

Host acceptance and life-history traits in busckii: tests of the hierarchy-threshold model

Steven P. Courtney and Department of Biology, University of Oregon, Jeffrey J. Hard Eugene, Oregon 97403, U.S.A. The hierarchy-threshold model for host-choice in predicts positive genetic correlations between use of different hosts, and between host use and the number of eggs carried by females. Life history theory predicts negative covariance between reproductive capacity and lifespan; hence host use and lifespan should be negatively correlated. These predictions were tested in half-sib breeding designs with the cosmopolitan Drosophila busckii. The predictions of the hietarchy-threshold model were met, supporting previous findings with other Drosophila. Predictions from life-history theory were not met for the population studied.

INTRODUCTION expected to maintain genetic variability. Courtney et a!. argue that insects rank potential hosts in Anumber of studies have recently documented the order of decreasing acceptability; high ranking presence of genetic variance affecting host choice hosts are always acceptable, while low-ranking behaviour within natural populations of plant- hosts may become acceptable, dependent upon feeding insects (Tavormina, 1982; Jaenike and environmental conditions. An accepting a Grimaldi, 1983; Via, 1984; Lofdahl, 1987; Jaenike, low-ranking host must therefore also accept all 1987; Thompson, 1988a; Courtney and Chen, higher-ranked hosts. The model finds some 1988). In reviewing these and other data, Futuyma empirical support in studies by other workers, not- and Petersen (1985) have drawn attention to the ably Wiklund (1981, 1982), Singer (1982, 1986), scarcity of information regarding the factors that Lofdahl (1987) and Thompson (1988a). Factors maintain such variation within populations. In par. influencing the rate at which hosts become accep- ticular, Futuyma and Petersen point out the need ted include current eggload borne by females (Fitt, for studies of genetic covariance of other charac- 1986) and adult experience (Jaenike, 1982, 1983). ters, and of the nature of genetic correlations Specific predictions made by Courtney et a!. (1989) between characters. Genetic correlations due to include: (1) Significant levels of additive genetic pleiotropy may impose constraints on the evol- variance for use of low-ranking hosts will be ution of host choice behaviour. Similarly, corre- maintained in populations, since such variance is lated responses to natural selection may explain rarely subject to direct selection (this holds true both observed patterns of host use, and the main- even if the host is never experienced in the popula- tenance of additive variance for host selection. tion in question, i.e., it is a novel host); (2) less Here we test a model which predicts pleiotropic variance is expected for use of higher-ranked hosts interactions between host use characters, using a (unless they are novel) due to direct selection on population of the cosmopolitan Drosophila busckii. such traits; (3) significant positive genetic correla- Courtney et a!. (1989) have proposed acompre- tions are expected between use of different low- hensive theory for the mechanistic control of host ranking, novel hosts; (4) there will be genetic choice behaviour. The "hierarchy-threshold" correlations between use of low-ranking hosts and model which they advance makes a number of factors influencing acceptance (high eggload predictions concerning genetic variance for host increases acceptance—therefore we expect a posi- use, and covariance with other characters. The tive genetic correlation between these traits); (5) model shows that pleiotropy between traits is no such correlations are expected for characters 372 S. P. COURTNEY AND J. J. HARD measured against use of high-ranking hosts. These tive prediction of life-history theory and the predictions have been met by several studies. hierarchy-threshold model is thus that we expect Lofdahl (1987) reports additive genetic variance negative covariance between lifespan and accept- for use of a novel cactus species by D. mojavensis; ance of low-ranking hosts, due to pleiotropic similar results obtain for D. suboccidentalis, a effects through reproductive allocation. mushroom feeder, when presented with two novel hosts, commercial mushrom (Agaricus campestris) and cucumber (Cucumis sativa) (Courtney and METHODS Chen, 1988; Chen, 1987). Courtney et a!. (1989) Wecarried out a standard half-sib breeding design report a significant positive genetic correlation using tenth generation descendants of D. busckii between use of low-ranked Cucumis and eggload caught at Eugene, Oregon on mushroom baits. in 6-day-old D. suboccidentalis, but no correlation Stocks were maintained at high population num- between use of higher-ranked Agaricus and bers (>1000) over Carolina instant Drosophila eggload. medium at 25°C, and under constant light. In all, D. suboccidentalis might be seen as an atypical 23 males were successfully mated to an average of species. It is an ecologically monophagous species 5•5 females, producing 126 full-sib families. (sensu Smiley, 1978), which uses only fungi of Families were raised on instant medium as before. genus Ramaria in the study population of the Half-sib families were distributed randomly among Cascade mountains in western Oregon (Courtney vials and with respect to position to the rearing el a!., 1989). Here, other fungi are rare and chamber and time of onset. This procedure ensures unpredictable, and Ramaria is the only relatively that common environment (larval rearing media) abundant resource. Monophagous populations effects did not influence the estimates of heritabiliy are, by definition, subject to little direct selection or genetic correlations, derived from half-sib on host choice. Results for D. suboccidentalis might designs (Falconer, 1981). were sexed at emer- therefore reflect the special circumstances of a gence, and the sexes separated. Females were then monophagous species. The present study was car- divided among four treatments. One group was ried out to determine whether similar results may dissected at age 6 days, to determine the eggload be seen in polyphagous species, which live under carried. Eggload is measured as an index of repro- more variable selection regimes. The cosmopolitan ductive allocation, independent of mating history. D. busckii was chosen because of its extreme Eggload was measured in virgins, because mated breadth of diet in the wild (many rotting substrates, females would begin to lay eggs. This seems a including mushrooms, are used), and because it is reasonable way to study host choice, since females distantly related to D. suboccidentalis. D. busckii mate at the oviposition site, and must therefore is in subgenus Drosophila; D. subocciden tails is a sometimes approach hosts as virgins. A second member of the quinaria complex of subgenus Dor- group was maintained (virgin) over medium, and silopha. If the predictions of the hierarchy- the lifespan of each individual recorded. The threshold model are met for such ecologically and remaining flies were mated at day 6 and exposed taxonomically diverse species, the theory may be for 24 hours to approximately 1 g of fresh material robust. of one of two hosts: A. campestris or C. sativa. An additional interest was to incorporate into Oviposition substrates were then examined for the our experimental design consideration of another presence of eggs; if one egg or more was found, life-history character, adult lifespan. Considerable then the host was scored as accepted. The total attention has been given to theories of life-history number of eggs laid on each host was also recor- evolution which rely on antagonistic pleiotrophy ded. Only one host was offered to each ; This between life-history components. For instance, procedure is necessary because exposure to a host Rose and Charlesworth (1981) show that, in some changes the behaviour of females on other hosts populations of D. meianogaster, early fecundity (Courtney and Gardner unpublished) and may and lifespan are negatively genetically correlated. alter both fecundity and survivorship. Some other studies support this result (e.g., Roach, Heritabilities of the numbers of eggs laid on 1986; Luckinbill et ai., 1987; Schemer et ai., 1989) either host, of female lifespan and of eggload were but others find equivocal (Hughes and Clark, 1988) estimated using standard methods employing or even opposite results (Giesel et aL, 1982). What- nested ANOVA (Falconer, 1981). Heritabilities of ever the generality of the results, antagonistic host acceptance were calculated using methods pleiotropy between fecundity and lifespan could applicable to threshold traits (Robertson and have important ramifications given that fecundity Lerner, 1949); standard errors for unbalanced itself is predicted to covary with host use. A deriva- designs are not calculable, and significance of these HOST ACCEPTANCE IN DROSOPHILA BUSCKII 373 threshold characters were estimated from x2analy- bers of eggs laid on the two hosts were not however sis. Genetic correlations were estimated from correlated, although each did weakly correlate with correlations across family means, following the eggload. We interpret these results as supporting suggestions of Via (1984) for situations where only thehierarchy-thresholdmodel. Acceptance one character can be measured per individual. This (whether or not at least one egg is laid) is not is a conservative procedure which, relative to other closely associated with the numbers of eggs laid measures of rA,tendsto underestimate correlations by accepting females. This argues that acceptance, of characters; it does have the advantage of and the numbers of eggs laid following acceptance employing standard tests of significance. Percent- are largely independent characters in D. busckii. ages were arcsine square-root transformed before Acceptance of the low-ranking host is positively analysis. correlated with eggload as predicted. Use of the high-ranking host is not correlated with eggload, again as predicted. Correlations between eggload RESULTS and the number of eggs laid on either host probably reflect rather simple effects of the number of eggs Heritabilitiesof the characters studied are shown available for laying. in table 1. As predicted, acceptance of the low- Predictionsbased on expectationsof ranking host, cucumber, showed significant addi- antagonistic pleiotropy between lifespan and egg- tive genetic variance, even though this host must load were not met (table 2). No significant relation- be very rarely encountered and used in the field. ship between these two characters was found. Heritability of acceptance of the high-ranking host, Neither did acceptance of cucumber correlate A. campestris, was not significantly different from negatively with lifespan. Eggload may not be the zero, again as predicted from the hierarchy- most appropriate measure of reproductive alloca- threshold model. Heritabilities of the numbers of tion; nevertheless, significant additive genetic vari- eggs laid on either host and of life-history charac- ance was detected for this trait, indicating the teristics were low, and only that of eggload was possibility of detecting significant correlations. An significantly different from zero. unexpected result was the significant positive corre- A significant positive genetic correlation was lation between lifespan and acceptance of the high- found between acceptance of the two hosts ranking host. Possible interpretations include: (1) (table 2). Acceptance of the low-ranking host was the widespread presence of deleterious mutations also positively correlated with eggload. The num- in inbred lines, leading to positive correlations between life-history components (Rose, 1984). We Table 1 Heritabilities of traits in the D. busckii population reject this interpretation because our stocks were recently derived from natural populations, and h2 SE. x2 P were kept at high numbers. Also, no correlation — — between lifespan and eggload was seen, as predic- E. Eggload 024011 ted from this explanation. (2) Interaction between L. Longevity 016018— — AcAcceptance of Cucumis 025 425001 behaviours favouring survival and those promoting AaAcceptance of Agaricus 008 293ns acceptance of a normal host. This might for instance occur where adults are attracted to feed Acceptance traits were treated as threshold characters, for on the oviposition substrate. We do not favour this which no standard errors are calculable for unbalanced data sets; significance levels are determined by x2statisticfor these hypothesis, and currently lack a satisfactory traits. explanation for the observed correlation.

Table 2 Genetic correlations between traits, as described in the text. n =23in all cases

L. Ac Aa Nc Na

E. Eggload —0038 0.628*** 0234 0.466*Q.437* L.Longevity —0036 0.527***—0132 0330 AcAcceptance of Cucumis 0.552** 0125 AaAcceptance of Agaricus 0290 NcNumber of eggs on Cucumis 0151 NaNumber of eggs on Agaricus Significance levels P<0.05*, oor ooos. Full-sib correlations did not yield additional information, except that the number of eggs laid by sibs on the two hosts were significantly correlated (r=0.452,P<001) 374 S. P. COURTNEY AND J. J. HARD

DISCUSSION lower-ranking hosts following host deprivation (e.g., after periods of inclement weather, when Thepredictions of the hiearchy-threshold model dispersal is prevented, or when host abundance is (Courtney et aL, 1989) have been met forD. busckii. low). Longer-lived flies may well then be the part Acceptance of hosts does appear to depend on of the population most likely to survive inter-host physiological status of the female. Observed pat- periods, and hence to be subject to selection for terns of genetic correlations imply pleiotropy, since acceptance of lower-ranked hosts. A priori then, it is unlikely that selection in the uniform lab we may expect positive genetic correlations environment would maintain sufficient levels of between longevity and acceptance of low-ranked linkage disequilibrium to explain such high genetic hosts, due to linkage disequilibrium. We should correlations. All observations therefore support the not however detect this pattern (expected under proposition of Courtney et a!., that host choice is field conditions) in our lab-reared populations strongly affected by internal factors, and that unless linkage is very strong. pleiotropy between characters may underlie and Our results indicate the presence within popu- help to explain the observed high levels of genetic lations of D. busckii of genetic variance for use of variance for host acceptance behaviour. a novel host Cucumis. If ecological conditions ever Our results for D. busckii confirm earlier, arose where use of this host was favoured, this similar findings in D. suboccidentalis. These two genetic variance would already be present in the species are at opposite ends of the resource- population, and available for the operation of breadth continuum, since D. suboccidentalis selection. As pointed out by Lofdahl (1987) and utilises a narrow range of hosts (so far we have Courtney et al. (1989), the presence of genetic found it only on Ramaria) while D. busckii uses variance for use of novel hosts implies that host many substrates including mushrooms, skunk cab- shifts may occur rapidly. What maintains this vari- bage, rotting wood and garbage. These two species ance? Our results clearly implicate correlated are also taxonomically distant, occurring in selection on reproductive allocation as one plaus- different subgenera of Drosophila. That the under- ible explanation. If eggload early in life is subject lying genetic structure of the two species is so to fluctuating selection, we may expect genetic similar argues that the hierarchy-threshold model variance for this character to be maintained. Given is not sensitive to taxonomic and ecological con- the variable thermal environment and patchy host siderations, and may be generally applicable. distribution of both D. busckii and D. subocciden- The exact values for h2 and r in any lab study talis, we feel that further attention should be given must always be treated with caution. Nevertheless to effects on female eggload. our results are in broad agreement with earlier Our results also suggest that selection on host results for D. mojavensis and D. suboccidentalis use may result in responses in other characters. In where heritability of acceptance of novel hosts was particular, selection against use of suboptimal low even if significant (12—22 per cent) (Lofdahl, hosts may initially result in lowered early fecun- 1987; Courtney and Chen, 1988; Courtney et aL, dity. However we believe that such selection 1989). Correlations between characters were rela- against a fitness component would result in a tively high compared to other behavioural studies change in the correlation between the traits; in this (Roff and Mousseau, 1987) but comparable to our case the internal threshold for host acceptance earlier results for D. suboccidentalis. This again would rise, and the correlation would weaken. supports our mechanistic explanation that correla- Selection on acceptance of one host will also have tions result from the constraint implied by physio- correlated responses on acceptance of other hosts. logical processes affecting both egg maturation and If our population of D. busckii had experienced host choice. strong directional selection for use of Cucumis we An additional life-history component, female would also expect the population to readily accept longevity, was not correlated with host choice in Agaricus and other hosts ranked higher than the manner predicted. This suggests that selection Cucumis. This effect has potential importance in could operate on variation affecting longevity insect pests, where colonists onto one novel host independently from selection acting on variation (crop) are also likely to be colonists of other hosts. affecting host acceptance. Nevertheless these Our studies have not addressed the relationship characters may show some association. Long-lived of adult and larval characters. In particular we flies will tend to have more opportunities for have not considered here the suitability of different acceptance or rejection of hosts. The hierarchy- hosts for larval development. Both Cucumis and threshold model also suggests that accept Agaricus are well able to support larval growth of HOST ACCEPTANCE IN DROSOPHILA BUSCKI1 375

D. busckii. Generally, adult oviposition range of JAENIKE, J. AND GRIMALDI, D. 1983. Genetic variance for host insects is less broad than larval dietary tolerance preference within and among populations of Drosophila tripunctata. Evolution, 37, 1023-1033. (Wiklund, 1975). We agree with Futuyma (1983), JERMY, T. 1984. Evolution of insect/hostplant relationships. Jermy (1984) and others that understanding ovipo- Am. Nat., 124, 609-630. sition choice is therefore critical to our knowledge LOFDAHL, K. 1987. A genetic analysis of habitat selection in of the evolution of host use. The hierarchy- the cactophilic species, Drosophila mojavensis. In Huettel, threshold model, which concerns the mechanism M. D. (ed.) Evolutionary Genetics of Invertebrate Behavior, Plenum Press, New York, pp. 153-162. of oviposition choice, implies that hosts are ranked LUCKINBILL, L. S., CLARE, M. J., KRELL, W. L., CIROCCO, in order of acceptability. The evolutionary W. C. AND RICHARDS, P. A. 1987. Estimating the number relationship between acceptability and suitability of genetic elements that defer senescence in Drosophila. Evol. EcoL, 1, 37—46. for larval development is largely unexplored RAUSHER, M. D. 1984. Tradeoffs in performance on different (Thompson, 1988b, but see Rausher, 1984; hosts: Evidence from within- and between-site variation in Rausher and Englander, 1988). Given that selec- the beetle Deloyala guttata. Evolution, 38, 582—595. tion against use of low-ranking hosts will usually RAUSHER, M. D. AND ENGLANDER, R. 1987. The evolution of be both weak and sporadic, the exact relationship habitat preference. II. Evolutionary genetic stability under soft selection. Theor. Pop. Biol., 31, 116—139. of adult preference and performance requires ROACH, D. 1986. Life-history variation in Geranium much further investigation. carolinianum. 1. Covariation between characters at different stages of the life cycle. Am. Nat., 128, 47—57. Acknowledgements We thank Ian Fisher and Alex Gardner ROBERTSON, A. AND LERNER, M. 1949. The heritability of for help with laboratory work. W. Bradshaw and L. Heisler all-or-none traits: viability of poultry. Genetics, 34, 395— have made useful comments on our ideas. 411. ROFF, D. AND MOUSSEAU, T. A. 1987 Quantitative genetics and fitness: lessons from Drosophila. Heredity, 58, 103—118. ROSE, M. 1984. Genetic covariation in Drosophila life history: untangling the data. Am. Nat., 123, 565-569. REFERENCES ROSE. M. R. AND CHARLESWORTH, B. 1981. Genetics of life history in . I. sib analysis of adult CHEN, G. K. 1987.Oviposition behavior of Drosophila sub- females. Genetics, 97, 173—186. occidentalis. M.Sc. Thesis, University of Oregon. SCHEINER, S. M., CAPLAN, R. L., LYMAN, R. F. 1989. A search COURTNEY. S. P. AND CHEN, G. K. 1988. Genetic and environ- for trade-offs among life history traits in Drosophila mental variation in oviposition behaviour in the melanogaster. Evol. EcoL, 3, 51—63. mycophagous Drosophila suboccidentalis Spcr. Func. Ecol., SINGER, M. C. 1982. Determinants of multiple host use by a 2,521—528. phytophagous insect population. Evolution, 37, 389-403. COURTNEY, S. P., GARDNER, A. AND CHEN, G. K. 1989. A SINGER. M. C. 1986 The definition and measurement of ovi- general model for individual host selection. Oikos, 55, position preference in plant-feeding insects. In Miller, J. 5 5-65. and Miller, T. A. (eds) Methods for Studying Mechanistic FALCONER, D. S. 1981. An Introduction to Quantitative Genetics, Interactions Between Insects and Plants. Springer-Verlag, 2nd edn, Longman, New York. Berlin, 1987. FITF,G.P. 1986. The influence of a shortage of hosts on the SMILEY, J. T. 1978. Plant chemistry and the evolution of host specificity of oviposition behavior in species of Dacus specificity: New evidence from Heliconius and Passiflora. (Diptera: Tephritidae). Physiol. EnlomoL, 11, 133—143. Science, 201, 745—747. FUTUYMA, D. j.1983.Selective factors in the evolution of TAVORMINA, S. i. 1982. Sympatric genetic divergence in the host choice by phytophagous insects. In Ahmad, S. (ed.) leaf-mininginsect Liriomyza brassicae (Diptera: Herbivorous Insects: Host-Seeking Behavior and Mechan- Agromyzidae) Evolution, 36, 523-534. isms, Academic Press, New York, pp. 227-244. THOMPSON, j.N.1988a. Variation in preference and specificity FUTUYMA, D. J. AND PETERSON, S. C. 1985. Genetic variation in monophagous and oligophagous swallowtail butterflies. in the use of resources by insects. Ann. Rev. Entomol., 30, Evolution, 42, 118-128. 217—238. THOMPSON, J. N. 1988b. Evolutionary ecology of the relation- GIESEL, J. T., MURPHY, P. A. AND MANLOVE, M. N. 1982. The ship between oviposition preference and performance of influence of temperature on genetic interrelationships of offspring in phytophagous insects. Entomol. Exp. AppI., 47, life history traits in a population of Drosophila 3-14. melanogaster: what tangled data sets we weave. Am. Nat., vIA, 5. 1984. Quantitative genetics of polyphagy in an insect 119, 464-479. herbivore, 1. Genotype-environment interaction in larval HUGHES. 0. M. AND CLARK, A. G. 1988. Analysis of the genetic performance on different hostplant species. Evolution, 38, structure of life history of Drosophila melanogaster using 88 1—895. recombinant extracted lines. Evolution, 47, 1309-1320. WIKLUND, C. 1981. Generalist vs. specialist oviposition .JAENIKE, i. 1982. Environmental modification of oviposition behavior in Papilio machaon (Lepidoptera) and functional behaviour in Drosophila. Am. Nat., 119, 784-802. aspects of the hierarchy of oviposition preferences. Qikos, JAENIKE, i. 1983. Induction of host preference in Drosophila 36, 163—170. melanogaster. Oecologia, 58, 320-325. wIKLUND, C. 1982. Generalist versus specialist utilization of .JAENIKE, j.1987.Genetics of oviposition-site preference in host plants among butterflies. Proc. mt. Symp. Insect-Plant Drosophila tripunctata. Heredity, 59, 363-369. Relationships, 5, 181—191.