Heredity 65 (1990) 163-168 The Genetical Society of Great Britain Received 6 February 1990

Selection for changes in the incidence of wing dimorphism in firmus

D. A. Roff Department of Biology, McGill University, 1205 Dr Penfield Avenue, Montreal, Quebec, Canada, H3A IB1.

Previous studies suggested that in the the heritability of wing dimorphism is approximately 0-65. This estimate was based on a full sib analysis and hence may be confounded by non additive effects. To confirm this estimate a selection experiment was undertaken, lines being selected for increased and decreased incidence of macroptery. The response to selection was asymmetric, response for decreased percentage macroptery being faster than that for increased proportion macroptery. The realized heritability in the "up" line was approximately 0-25, and in the "down" line approximately 0-95, the mean across all lines being 0-6: this agrees very well with the estimate of 0-67 obtained from a full sib analysis. Two possible reasons for the asymmetric response are suggested. In the related cricket species, , the genes controlling wing morph appear to be on the sex chromosome while in C. firmus no sex linkage is evident. A new method of estimating realised heritability for threshold traits is presented.

INTRODUCTION of reproduction and a reduction in fecundity (Roff, 1984, 1986a; Denno et a!., 1989; Roff and Fair- Migrationis an important component of the lives bairn, 1990). Thus there is a trade-off between flight of many permitting them to colonize and capability and reproduction. persist in temporary habitats (Southwood, 1962; To understand how the incidence of a par- Johnson, 1969; Harrison, 1980; Dingle, 1985). But ticular wing morph will evolve under changing though there are long term benefits to migration environmental conditions we must understand the there are also costs such as increased mortality and genetic basis of the traits involved in this trade-off decreased reproduction (Southwood, 1962; Roff, and the genetic correlations between them. Herita- 1977), and hence it is not surprising that migratory bility estimates of wing dimorphism for two polymorphisms are common (Harrison, 1980; crickets, Gryllusfirmus and Allonemobiusfasciatus, Dingle, 1985). Many species are wing suggest that a significant fraction of the variation dimorphic, some individuals within the population (>50 per cent) can be attributed to additive genetic or a having wings and being capable of effects (Roff, 1986b; Mousseau and Roff, 1989; flight and others having either no wings or reduced maternal effects were insignificant in the study of wings and being incapable of flight. Within these G.firmus but could not be examined in the analysis species there is a clear dichotomy between the of A. fasciatus). However, in both cases heritability morph capable of migration and that which is not. estimates were obtained by full sib analysis and But an individual with wings does not have to thus contain an unknown fraction of non additive and hence the existence of wing dimorphic species effects. To overcome this deficiency a selection suggests that there is a cost not only to migration experiment was undertaken using Gryllus firmus, itself but also to possessing the capability of migra- lines being selected for increased and decreased tion, viz wings, wing muscles and associated struc- incidence of macroptery. The results of this experi- tures. This hypothesis has been confirmed in a wide ment and a new method for estimating realized variety of insects, in which it has been found that heritability of a threshold trait are presented in the flight-capable morphs show a delay in the onset this paper. 164 D. A. ROFF

MATERIALS AND METHODS maintained to provide further estimates of herita- bility. One "within family" macropterous and two Experimental protocol micropterous lines were so established, designated Detailsof the species and rearing methods are as WL1, WS1 and WS2, respectively. given in Roll (1986b), and only the salient For each of the lines Li, Si and Cl, six cages points are presented here. The stock of G. firmus containing 60 newly hatched nymphs per cage, and used in the present study was derived from for the within family lines (WL1, WS1 and WS2), approximately 40 individuals (approximate sex two cages of 60 nymphs/cage were established. ratio 1: 1) from a single location in northern Males and females of a desired morph (macrop- Florida in 1981. They are maintained in diapause terous for Li, micropterous for Si and both for averting conditions (25-30°C, no set photoperiod Cl), were selected upon eclosion into adults until but the laboratory lights ensure a relatively long approximately 100 (and never less than 50) of each light period), with a breeding stock of between sex were obtained. In the case of WL1, WSi and 100-300 individuals. For the selection experiments WS2, 20 females and ten males were selected. The individuals were raised in batches of 60 individuals relatively large number of parents was used to per disposable mouse cage, as described in Roff prevent inbreeding depression. The mass selection (1986b). Food was provided, ad libitum and com- procedure outlined above was followed on all gen- prised Purina© rabbit chow and fresh lettuce erations subsequent to the first. To provide a repli- leaves. cate, a second series of lines was set up some Previous studies were conducted at 30°C and months after the first, the protocol followed being a photoperiod of 17 h L; 7 h D: under these condi- identical except that a mass selection design was tions the proportion of macropterous males and followed from the outset and no initial group of females is about 76 and 64 per cent, respectively single pair matings constructed. These lines will (Roll, 1986b). To reduce this percentage to around be designated L2, S2 and C2. 50 per cent in the females, in the present experi- In G. rubens wing morph is controlled to a meni crickets were reared at 28°C, 15 h L: 9 h D. large extent by sex-linked genes (Walker, 1987; An e;timate of the heritability of wing dimorphism Gryllus have an XO system with the female under these conditions was obtained in the first being XX); to test for this possibility in G. selection experiment as follows: eggs were firmus, the following crosses were made in obtained from the stock culture and the nymphs generation 10 of the first experiment, Li x SI, raised under the experimental conditions. From LIxWS1, LixWS2, WL1xS1, WL1xWS2. In this group 20 pairs were extracted comprising 10 each cross two cages, each comprising ten adult pairs LW x LW (macropterous x macropterous) females from one line and five adult males from and 10 pairs SWxSW (micropterousx another, were set up and 180 nymphs collected micropterous). Estimates of the heritability of wing from these for rearing. All crosses were reciprocal dimDrphism were made from these crosses for the with respect to sex. founding population. Since these matings do not include mixed crosses, estimates of heritability have to be corrected for assortative mating (dis- Statistical cussed below). analysis To establish a macropterous line (hereafter Theestimation of heritability from full sib data referred to as the Li line) 200 adults (100 males, for a threshold trait is summarised in Roff (i986b, 100 females) from the 10 LW x LW matings were typographical corrections to formulae given in this mixed together, with approximately equal rep- paper are presented in Mousseau and Roff, 1989). resentation from each family. Similarly, a In the present analysis the situation is complicated micropterous line (Si) was started by mixing by the fact that the matings are assortative. To together the offspring from the SW x SW crosses. calculate the uncorrected estimate of heritability, A Control line (Cl) was formed by mixing the H, we require an estimate of the mean proportion offspring from all crosses. Based on initial esti- of micropterous individuals per family, p. The pro- mates of percentage macroptery in each family, 3 portion p is normally estimated by (i /n where "extreme" families were selected and matings nisthe number of families and p, is the proportion made between siblings within each family, 20 of micropterous individuals in family i (Roff, females and ten males per family: final data on i986b). Since in the present case mating is assorta- these families indicated that they were not tive, p was estimated from the parental population. "extreme" (see Results) but the lines were The heritability estimate can be corrected using SELECTION AND WING DIMORPHISM 165 the formula given by Falconer (1981, P. 164), The probability that an individual in generation i + 1 will be micropterous is, /12 ={—1+../[1 +4rH]}/(2r) (1) z÷1)2)dx =P±1.(3) where 112isthe narrow sense heritability, H is the (1/)J exp (—05(x — uncorrected estimate of heritability, and r is the phenotypic correlation between parents measured The likelihood of obtaining the observed series on the underlying continuous scale. This cannot be of micropterous individuals in the selection experi- directlyestimated from the dichotomous ment is, phenotypes of the parents (macropterous or N micropterous). However, given the proportions of H CP(1-P' (4) macropterous males and females in the population from which the parents are derived, and assuming where n, is the number of offspring in generation that the dichotomous trait is a consequence of an i, r- is the number of micropterous offspring in interaction between a threshold and a continuously generation i, and N is the number of generations. varying character (Falconer, 1981), r can be esti- Taking logs, we obtain mated by Monte Carlo simulation. The base popu- lation comprised 51 per cent macropterous females LLcC {r, log (P)+(n —ri) log (1 —P)}. (5) and 29 per cent macropterous males. The value of r, estimated from 100 runs of the simulation was Thus the problem of finding the best estimate 063. To examine the robustness of this value to of h2 reduces to finding the value of h2 that maxim- variation in percentage macroptery I ran the simu- izes LL. Apart from h2, parameters to be estimated lation with percentages of macropterous males and are the frequency of micropterous individuals in females ranging from 20-50 per cent, using both the initial population, and the mean phenotypic equal and unequal values for the parents. The mean value of the parents at each generation. Although value of r, estimated from 100 runs per parameter the frequency of macroptery differs between sexes, combination, was very insensitive to variation the two heritabilities do not differ (Roff, 1986b), within this range, the maximum and minimum and hence in the present estimation procedure both values being 066 and O62, respectively. sexes are considered jointly. The total number of Realized heritability is usually estimated by a parameters to be estimated is 2(N+1)+1, where regression of cumulative response on cumulative N is the total number of generations. The best selection differential (Falconer, 1981). The same estimate of the initial frequency is r0/ n0 (two esti- procedure can be used with a threshold trait in the mates since there are two sexes), while the mean following manner (Method 1): the mean value, on value of the parents can be computed according the underlying continuous scale, of the offspring to the first method described, leaving only h2 to in generation i is estimated as z,,theordinate on be estimated. The best estimate of h2, given the the standardised normal curve corresponding to p estimates of the other parameters as described is estimated as z1,theproportion of micropterous above, can be found by numerical methods (see, offspring. The estimated value of the micropterous for example, Press et al., 1986). Confidence inter- parents selected to produce the i + 1 generation is vals were estimated as the values of h2 at which given by For the LL equals X=o.o5 with 2(N+1)+1 degrees of macropterous line the relevant estimates are —z1, freedom (Draper and Smith, 1981). Confidence and—;—exp (—05z)/('J[i[1 —pJ). intervals are asymmetric, but the differences were A potential problem with the above method is small and in the present paper the estimated stan- that it cannot be used if in one or more generations dard error of the estimate was computed as only one morph is observed, since z-cannotthen 05{(C—h2)/2+(h—C1)/2), where C and C1 be estimated. This problem is likely to arise when are, respectively, the upper and lower 95 per cent samples are small and/or p is close to 1 or 0. A confidence bounds on h2. method that does not depend upon this constraint is as follows (Method 2): let G, be the mean phenotypic value of the selected micropterous RESULTS parents in the ith generation. The expected value Heritability estimates and selection response of the offspring, z1,isgiven by Heritability estimates derived from the 20 full sib =z•(1—h2)+ h2G,. (2) families are, after correction for assortative mating, 166 D. A. ROFF 066+ 02l (±S.E.) and 069 022 for females and The two methods of estimating h2 from males, respectively. These estimates compare very response to selection gave similar results and indi- favorably to the values of 068±0085 and 062± cate that the realised h2 in the macropterous lines 0075 for females and males, respectively, obtained was less than in the micropterous lines (table 1). under rearing conditions of 30°C and 17L:7D In all three "within family" lines (WL1, WSi and (Roff, 1986b). The relatively high value of h2 sug- WS2) the estimated values of h2 are similar in gests that the population should respond rapidly magnitude to the mass selected lines. Plots of to selection. This prediction was verified by all cumulative response on cumulative selection selected lines, the lines selected for macroptery differential were examined visually and statisti- increasing to approximately 90 per cent macrop- cally for evidence of curvature indicating a limit terous females and those selected for microptery on the underlying scale to response to selection; decreasing to approximately 5 per cent macrop- there was no evidence of non linearity. The terous females by generation 5 (fig. 1). Based on averaged estimate of heritability in the lines selec- ascending rank by percentage macropterous, the ted for increased percentage macroptery is 025, family selected as WL1 was 5th rank of the ten and that obtained for the lines selected for LWx LW crosses, while WS1 and WS2 were 3rd decreased incidence of macroptery is 095 (0.90 if and 5th rank among the ten SWx SW crosses. The estimates greater than 1 are set to 1). The heritabil- response of SW1 and WS2 were both more rapid ity estimate averaged across morphs and lines is than either Si or S2, but the response of WL1 was 060; this agrees well with the value of 067 comparable to Li and L2 (fig. 1). obtained from full sib analysis in the present experiment and that of 065 previously reported (Roff, 1984).

Females Table 1 Heritability estimates from selection experiments. See text for a description of the methods of estimation

Heritability estimate (SE.) Line Method 1 Method 2 C,) 0 Li 033 (005) 025 (0.04) L2 022 (0.05) 033 (0.04) a) WL1 Oil (0.10) 024 (0.07) Si 078 (016) 1•27 (020) I—0 C) S2 086(018) 0.74(0.11) C', WSI — 072 (0.19) WS2 — 132 (0.32) 0)a) Li+S1* 056 (0i2) 076 (0.14) C', L2+S2 0.54(0.13) 0.54(0.08) C ci *Mean of the two heritabilities. S.E. estimated as the square C) root of the mean variance. a) Crossesbetween lines Therewas no significant heterogeneity between cages within crosses, and therefore, cages were combined for the following analyses. If the genes controlling wing morph are sex linked, the 6 8 10 frequency of macropterous offspring in highly selected lines should depend upon which parent Generation is macropterous, as indicated in table 2. For G. Figure 1 Response to selection for increasing and decreasing rubens, crosses after five generations of selection incidence of macroptery in Gryllusfirmus. Line1 (LI, indicated that a large proportion of the genes con- Si, Ci); —— — "Withinfamily" lines (WL1, WS1, WS2); trolling wing morph are sex linked in this species Line 2 (L2, S2, C2). (table 2; Walker, 1987). No such variation is SELECTION AND WING DIMORPHISM 167

Table 2Percentage of macropterous offspring from crosses LW under "optimal" conditions; Roff, unpub- between lines selected for increased and decreased lished data). incidence of macroptery (LW). Sample sizes, within sexes, Although the crickets in the selection experi- for the between line crosses ranged from 41 to 102 with a mean of 63 (SD=17) ments are reared under good conditions, as eviden- ced by high survival (>80 per cent), some Female Male Female Male individuals are likely to be stressed due to small, parent parent offspring offspring random variations between and within cages, and Macropterous* Micropterous 500 (51) 1000 (99) possibly harassment by larger nymphs. Suppose Micropterous Macropterous 50.0 (47) 00(11) that, if exposed to some stress, S, prior to the Li Li 947 879 developmental stage at which wing morph is deter- WL1 WL1 938 822 mined, a cricket will always develop into a SW Si Si 16 31 WS1 WS1 00 00 morph. Further, suppose that, on average, 10 per WS2 WS2 0.0 0•0 cent of nymphs are exposed to S in each generation. Ll Si 60•7 637 Under these conditions lines selected for increas- SI Li 667 483 ing macroptery will asymptote at 90 per cent rather WL1 Si 47.2 167 than 100 per cent, while lines selected for decreased SI WLI 719 518 incidence of macroptery can achieve zero per cent. Li WS1 75•6 729 WS1 Li 683 507 If, as is likely, genetic variation exists for the WL1 WS2 576 34.9 response to S, selection for an increasing percent- WS2 WL1 826 475 age of macroptery will also select against Li WS2 796 70•0 individuals that show a response to S. Because of the very low selection intensity that can be achieved * Theproportion expected if all genes controlling wing morph are sex linked and the lines homozygous. Values in brackets when the frequency of LW individuals exceeds 90 show data for G. rubens(fromtable 4 of Walker, 1987). per cent (with respect to both the response to S and the incidence of macroptery in the absence of S) the response to selection will be slow and the evident in any of the present crosses (table 2), realized heritability relatively low, as found in the indicating that in this species most of the genes present experiments (fig. 1). However, the selected controlling wing morph are autosomal. group in the line selected for increased microptery will contain some genetically macropterous individuals and hence the selection differential will DISCUSSION be reduced. A slower response would therefore be expected in the S lines compared to the L lines Althoughthe estimates of heritability derived from although the former is capable of evolving further full sib analysis and from selection are very similar than the latter. The effect of S on the response to when morphs are combined (approximately 06, selection and the realized heritability is easily table 1), there is a clear asymmetry in the response calculated using equations 2 and 3 and adjusting to selection. Lines selected for increasing for the effect of S. To demonstrate these effects I macroptery diverged from the control lines and used S 01 and a heritability of 08 in the absence approached their limiting value (100 per cent) of S (i.e., S =0).In the case of selection for much more slowly than lines selected for a increased incidence of microptery neither the decreased incidence of macroptery. Two factors response to selection (fig. 2), nor the estimated may be responsible for this asymmetric response. realized heritability are significantly affected (the First, two lines of evidence suggest that this may estimated value of h2 declines from 08 to 0.76). be a consequence, in part, of stress causing crickets But the response to selection when selection is for to switch, developmentally, to the micropterous increased incidence of macroptery is greatly form; reduced (fig. 2), and the realized heritability only (a)if crickets from the LW lines are weighed daily 018. These results match those observed (fig. 1) during development they all develop into the quite closely, but I have not attempted to estimate micropterous morph (Shannon and Roff, values of S or h2 which give a "best" fit to the data. unpublished data). The second possible reason for the difference (b) when reared under conditions that cause high in response is that micropterous females begin mortalities (>50 per cent) inordinately high reproduction earlier and produce more eggs than frequencies of micropterous forms are the macropterous morph (RoIl, 1984). It is possible obtained (50-100 per cent LW vs. 95 per cent that macropterous females with genotypes that lie 168 D. A. ROFF

table 3 of Walker [1987]). The mechanism preserv- 0 9-f ing such high genetic variation in these species has not been resolved though the trade-off between wing morph and the age schedules of reproduction I (Roff, 1984) may play an important role. Acknowledgements I am most grateful for the constructive criticisms of Drs D. J. Fairbairn, H. Dingle, J. Endler, T. Prout t— and M. Turelli. Catherine Helick and Sharon David provided indispensable technical assistence. This work was supported by an operating grant from National Science and Engineering I Council of Canada. 00 ) 2 4 6 8 10 12 14

Generation REFERENCES Figure 2 Predicted response to selection for increasing and S = decreasing incidence of macroptery when h2 =0'8, 0 DENNO,R. F., OLMSTEAD, K. L. AND McCLOUD, E. S. 1989. (————)andS=01 (—).Inthe latter case realized h2 Reproductive cost of flight capability: a comparison of life is reduced to 076 in the "down" line and 018 in the "up" history traits in wing dimorphic planthoppers. EcoL Ent., line. 14, 31-44. DINGLE, H. 1985. Migration. In Kerkut, G. A. and Gilbert, close to the threshold at which wing morph is L. 1. (eds) Comprehensive Insect Physiology, Biochemistry and Pharmacology. Vol. 9, Behaviour, Pergamon Press, determined also reproduce sooner and produce New York. more eggs than macropterous females with DRAPER, N. R. AND SMITH R1981.Applied Regression Analy- genotypes that are far removed from the threshold. sis, 2nd ed. John Wiley, New York. Since there is no way of distinguishing these FALCONER, D. S. 1981. Introduction to Quantitative Genetics. 2nd edn. Longman, London and New York. genotypes morphologically both types would be HARRISON, R. G. 1980. Dispersal polymorphism in insects. Ann. selected but the former would contribute more Rev. EcoL Syst., 11, 95—118. offspring to the next generation. Likewise, in the JOHNSON, C. G. 1969. Migration and Dispersal of Insects by micropterous females the age of first reproduction Flight. Methuen, London. and total fecundity may depend upon how close KEVAN, D. McE. 1980. The taxonomic status of the Bermuda is beach cricket (Orthopteroida: ). Syst. Ent., 5, the genotype to the threshold value, 83—95. micropterous females with genotypes far removed LIM, H-C., VICKERY, V. R. AND McKEVAN, D. K. 1973. from the threshold reproducing the earliest and Cytogenetic studies in relation to within the producing the most eggs. This would mean that family Gryllidae (). 1. Subfamily . natural selection would act against artificial selec- Can. J. Zool. 51, 179-186. MOUSSEAU, T. A. AND ROFF, D. A. 1989. Geographic variability tion when selecting for the macropterous morph in the incidence and heritability of wing dimorphism in but would act with artificial selection when select- the striped ground cricket, Allonemobius fasciatus. Hered- ing for the micropterous morph. ity, 62, 315—318. The finding that in two closely related species PRESS, W. H., FLANNERY, B. P., TEAKOLSKY, S. A. AND the genes controlling wing morph are in one case VETFERLING, w. T. 1986. Numerical Recipes. Cambridge University Press, Cambridge. on the sex chromosomes (G. rubens) and in the ROFF, D. A. 1977. Dispersal in dipterans: its costs and con- other on the autosomal chromosomes (G. firmus) sequences. J. Anim. Ecol., 46, 443-456. was unexpected. Both species have the same num- ROFF, D. A. 1984. The cost of being able to fly: a study of wing ber of chromosomes (2n =29in males, Lim et a!., polymorphism in two species of crickets. Oecologia, 63, 30—37. 1973; although G. firmus has not been examined, ROFF, D. A. 1986a. The evolution of wing dimorphism in insects. the species G. bermudensis, is now considered only Evolution, 40, 1009—1020. a subspecies of G. firmus [Kevan, 1980]). Further ROFF, D. A. 1986b. The genetic basis of wing dimorphism in study on other species within the Gryllus the sand cricket, Gryllus firmus and its relevance to the evolution of wing dimorphisms in insects. Heredity, 57, are needed to establish the general location of the 221—231. genes determining wing morph. ROFF, D. A. AND FAIRBAIRN, D. j.1990.Wing dimorphisms Walker (1987) selected for changed incidence and the evolution of migratory polymorphisms among the in the proportion of the macropterous morph of Insecta. Am. Zool. (In press). G. rubens, and as with G. firmus, obtained a rapid SOUTH WOOD, T. R. E. 1962. Migration of terrestrial in relation to habitat. Biol. Rev., 37,171-214. response, the realized heritability being 098 016 WALKER, T. .1.1987.Wing dimorphism in Gryllus rubens (estimated by Method 1 using data presented in (Orthoptera: Cryllidae). Ann. ent. Soc. Am., 80,547-560.