[PDF] Polymorphism and evolution in the butterfly

/-/eredity7l (1993) 242 - 251Genetical Society of Great BritainReceived 3 December 1992Polymorphism and evolution in the butterflyDanaus chrysippus (L.) (Lepidoptera:Danainae)DAVID A. S. SMITH, DENIS F. OWEN*, IAN J. GORDON & AGOROACHAI M. OWINYDepartment of Biology, Eton College, Windsor SL4 6EW, tSchoo/ of Biological and Mo/ecu/ar Sciences, Oxford BrookesUniversity, Headington, Oxford 0X3 OBP, U.K., Department of Zoology, University of Nairobi, P0 Box 30197, Nafrobi,Kenya and §Department of Zoology, Makerere University, P0 Box 7062, Kampala, UgandaAnalysisof the genetic structure of a sample of the polymorphic butterfly Danaus chiysippus fromKampala, Uganda shows that the population is undergoing substantial evolutionary change.Comparison with samples from the same area going back to before 1900, indicate that thefrequency of form alcippus has increased from 16 per cent to 71 per cent (1909 - 91) while f. dorip-pus has decreased from 14 per cent to 2 per cent, f. aegyptius from 66 per cent to 24 per cent and f.albinus from 4 per cent to 3 per cent. Genotype frequency differences between the sexes at two ofthe three loci examined suggest that a balanced polymorphism is maintained by opposing selectiveforces acting on males and females. Non-gametic (genotypic) disequilibrium between two pairs ofunlinked loci indicates that natural selection is involved, again with sex differences. It is suggestedthat the polymorphism originated after hybridization of allopatric races which evolved during thePleistocene but are now maintained sympatrically. The selective agents have not been identified butmimetic relationships, both Batesian and Müllerian, are almost certainly involved.Keywords:allopatricevolution, D. chiysippus, mimicry, non-gametic disequilibrium, polymorph-ism, sympatric evolution.IntroductionInmuch of Africa the butterfly Danaus chrysippus (L.)(Nymphalidae: Danainae) is polymorphic for colourand pattern. As it is both a Batesian model for severalunrelated species, belonging to four different butterflyfamilies (or subfamilies depending on the authority)and also a Müllerian mimic sharing in rings with severalother unrelated unpalatable species, it is a remarkableexception to the rule that both Batesian models andMüllerian mimics are monomorphic (Owen, 1971;Smith, 1980; Gordon, 1984). The polymorphism isbest developed in East and Central Africa, especially inUganda; there are also large areas, both within andoutside Africa, where the butterfly is monomorphicand this is the general rule, with only local exceptions,throughout its extensive Palaearctic, Oriental andAustralasian range.In East Africa there are four main colour forms, stillreferred to by their traditional Latin names: aegyptius(incorrectly called chiysippus by many previous investi-*Correspondence242

gators in the East African region), alcippus, dorippusand albinus (Fig. 1). Colour plates can be found inRothschild etal. (1975) and Smith(1980).Many alcippus have only a small white patch in thehindwing or simply white scaling lining the veins; thesewe refer to as weak alcippus (also known as alcip-poides); similarly there are weak albinus (also known assemialbinus). Many dorippus have a variable number of'aegyptius' subapical spots on the underside of the fore-wing; these are also called transiens. In all four formsthe orange ground colour may be completely orpartially replaced by brown.Forms aegyptius, alcippus, dorippus and albinus areinvolved in a Müllerian mimicry complex with Acraeaencedana Pierre and A. encedon (L.) (Acraeinae) andare also models for many presumed Batesian mimics,particularly Hypolimnas misippus (L.) (Nymphalidae).Maps showing the distribution of the forms of D.chrysippus in Africa are given in Owen & Chanter(1968), Owen (1971), Rothschild et al. (1975) andPierre (1973, 1980). The switch between orange andbrown ground colour phenotypes (which have notreceived Latin names) probably has little impact on

POLYMORPHISM AND EVOLUTION IN D. CHRYSIPPUS 243

Fig. 1 Colour forms (f.) of Danaus ch,ysippus. (a)1. aegyptius. (b) f. alcippus. (c) 1. dorippus. (d) f. albinus. Theblack and white areas are as shown; the stippled areas areeither orange or brown.generalized mimetic resemblance (Gordon, 1984),especially as many of the mimics are similarly variable;Smith (1980) has suggested that the orange - brownpolymorphism relates to climatic adaptation, possiblythrough thermoregulation, influencing activity levelsand, in particular, courtship behaviour.The genetics of the polymorphism in D. ch,ysippusinvolves three loci, A, B and C (Smith, 1975a). The Alocus controls the pattern of the hindwing, the A allelegiving brown or orange aegyptius and aa a large whitepatch (alcippus). Aa heterozygotes vary betweenhaving a reduced white patch, white scaling lining theveins (both scored as weak alcippus) or no trace ofwhite (scored as aegyptius); hence not all heterozygotesare phenotypically distinguishable from AA homo-zygotes. Estimates for the penetrance of a in hetero-zygotes, based on breeding at Dar es Salaam, Tanzaniaare 74.4 per cent (N= 47) (Smith, 1975a) and 65.6 percent (N=66) (Smith, 1980) with an average of 70 percent. An estimate of 69.7 per cent penetrance for theKampala sample (N =211)suggests good agreementwith Dar es Salaam but assumes the alleles are in

Hardy - Weinberg equilibrium, which they are unlikelyto be (see below). At Kampala, penetrance of a isnearly equal in the two sexes whereas at Dar es Salaamit is 24 per cent higher in males than in females. The Blocus controls ground colour, B giving brown and bborange. Many heterozygotes are detectable but they arevariable: Bb genotypes always have a brown forewingand a little brown on the costal margin of the hindwingbut the hindwing may be entirely brown as in BBhomozygotes. The B locus was first identified byClarke et al. (1973) and confirmed by Smith (1975a).The C locus controls the forewing pattern, C givingdorippus (or albinus when combined with a whitehindwing), and cc giving aegyptius (or alcippus whencombined with a white hindwing). Many Cc hetero-zygotes can be recognized phenotypically by thepresence of pale subapical spots on the underside ofthe forewing (transiens). At Dar es Salaam, Tanzania,the expressivity of c in Cc genotypes is influenced bythe B locus, 45 per cent (N= 129) of bbCc hetero-zygotes being transiens compared with 74.6 per cent(N= 173)of B - Cc (Smith, 1980).The B and C loci are closely linked with a recombin-ation value of 1.9 per cent (3.8 per cent in males only asmeiosis is probably achiasmatic in females). TheA locus assorts independently of B and C at Dar esSalaam (Smith, 1975a, 1980) and also did so in crossesbetween alcippus (aa B - )fromSierra Leone andsubspecies petilia (AA bb) from Australia (Clarke et al.,1973).Here we report on genotype frequency in a sampleof D. chrysippus obtained at Kampala, Uganda, inDecember 1991 and present evidence indicating thatpowerful natural selection is operating at the presenttime and that a considerable evolutionary change hasoccurred in the past 91 years (approximately 1092breeding generations).Materials and methodsSource of materialD. chrysippus is an abundant butterfly and largesamples are easily obtained by using an ordinarybutterfly net. In December 1991 we collected a samplein the Kampala area at sites where previous sampleshad been obtained by us (1964 - 66, Owen & Chanter(1968); 1972, A.M.O., unpublished data). We alsoexamined two samples in the Hope EntomologicalCollections, Oxford, which had been randomlyobtained on the instructions of Professor E. B. Poulton.One of these, obtained by C. A. Wiggins in 1909 - 12, isfrom the same area as the more recent samples,whereas the other comes from Kome Island (just off

244 D. A. S. SMITH ETAL.Table 1 Frequency (percentage) of phenotypes and genotypes in a sample ofDanaus chrysippus collected in the Kampala region, Uganda, in December 1991Form and colourGenotypeMaleFemalePopulationalcippus, orangeaa bb cc46.724.035.5alcippus,brownaaB - cc3.77.75.7weak alcippus (= alcippoides),orangeAa bb cc26.224.025.1weak alcippus (= alcippoides),brownAa B - cc0.98.74.7aegyptiust orangeA - bbcc15.924.019.9aegyptiust, brownA - B - cc1.95.83.8albinus, orangeaa bb C - 0.90.00.5albinus-transiens, orangeaa bb Cc3.70.01.9weak albinus( =semialbinus),orangeAa bb C - 0.01.00.5dorippus, orangeA - bbC - 0.01.90.9dorippus-transiens, orangeA - bbCc0.02.91.4Sample sizeA - bbCc107104211tScored as chiysippus by some earlier workers, this name should be restricted to thesimilar but distinct Oriental and Palaearctic subspecies.Entebbe in Lake Victoria), was obtained in 1946, andis included because D. chrysippus ranges widely andthe colour forms (at a given time) are stabilized overrelatively large geographical areas (Owen, 1971).Three earlier large collections, all mainly pre-1900,were examined: two in the Natural History Museum,London and one in the Hope Entomological Collec-tions, Oxford. Although museum samples are notor-iously non-random with rare phenotypes generallyover-represented, in this case the phenotype frequen-cies in the three collections are in fact, surprisingly,formally homogeneous and this increases our confid-ence that they provide a realistic early baseline fromwhich subsequent evolutionary changes are assessed.ResultsGenetic structure of the Kampala population in 1991Table1 lists the frequencies of phenotypes and geno-types in the Kampala sample obtained in 1991; Table 2gives the frequencies of dominant and recessive pheno-types observed at each of the three polymorphic loci:A, B and C. At all three loci only a proportion of heter-ozygotes is distinguishable from dominant homo-zygotes and their scores are therefore combined. Asthe Uganda population of D. chrysippus is unlikely tobe in Hardy - Weinberg equilibrium, it follows that theestimation of allele frequencies using the observedfrequency of recessive phenotypes, which is the onlymethod available to us, must be illegitimate.There are several factors, known to apply either tothe sampled population or to others in East Africa,which are expected to be inimical to the maintenance

Table 2 Frequencies of dominant and recessive phenotypesin a sample of 211 D. chiysippus from Kampala, UgandaLocusMaleD(%)R(%)FemaleD(%)R(%)Z1)A44.955.168.331.710.820**B6.593.522.177.99.249**C4.795.35.894.20.002D, dominant; R, recessive. **fl >P>0.00 1 (with Yates'correction).of a Hardy - Weinberg equilibrium. (i) Selective elimina-tion of Aa heterozygotes was observed at Kampala in1964 - 66 from a comparison of the proportions ofweak alcippus among adults reared from wild larvae inthe laboratory and in the wild population (Owen &Chanter, 1968). (ii) Sexual selection based on maleheterozygote advantage is recorded in Tanzania (Smith,1975c, 1981). (iii) Assortative mating for both B and Cloci is known from Tanzania (Smith, 1984). (iv)Linkage disequilibrium for B and C loci is also presentin Tanzania (Smith, 1980). (v) Both primary andsecondary sex ratios generally deviate from 1:1throughout East Africa (Smith, 1975b; I. J. Gordon,unpublished data) and both were low (female biased) inUganda during 1964 - 66 (Owen & Chanter, 1968). (vi)Migration has frequently been reported. Thus there areat least six possible factors which either singly or incombination are incompatible with the maintenance ofa Hardy - Weinberg equilibrium.

POLYMORPHISM AND EVOLUTION IN 12 CHRYSIPPUS 245The data (Table 2) show highly significant sex differ-ences for phenotype frequency at the A and B loci: inparticular, aa (white hindwing) is more frequent inmales, and conversely, A - (non-whitehindwing) infemales (P P>0.02x2 values are calculated using Brandt and Snedecor'sformula. A - C - and aa C - genotypes are amalgamated toavoid low expected values.

heterozygotes or linkage phases in double hetero-zygotes, thus ruling out exact tests for linkage disequili-brium. However, chi-squared tests applied to the sixdata subsets in Table 3 (A-B and A-C for each of males,females and the population), on the null hypothesis thatassociation of genotypes in A-B and A-C combinationsis random, are all non-significant. This result suggeststhere is no linkage disequilibrium between the A andBC chromosomes, either within sexes or in the popula-tion as a whole.On the other hand, some form of disequilibrium isclear from the comparison of phenotype frequencydistributions between sexes for both A-B and A-Ccombinations (Table 3). Among A-B phenotypes, theAB, Ab and aB 'gametes' are all under-represented inmales but exceed expectation in females; the oppositeapplies to the ab 'gamete'. The data are not sufficient todetermine if the effects at the A and B loci might beadditive or interactive but suggest the former. A similarpattern, although involving mainly the A locus, is foundin the A-C genotypes; AC and Ac 'gametes' are morefrequent than expected in females whereas aC and acare over-represented in males. The A-B pattern sug-gests some kind of second-order, genotypic (non-gametic or trigenic) disequilibrium within each sex. Thesex differences could be caused by abnormal segrega-tion of the type previously recorded at Dar es Salaam(Smith, 1976b) and Nairobi (I. J. Gordon, unpublisheddata) in which there is some form of association atsegregation between autosomes and sex chromosomes,possibly resulting from Y-autosome translocations. Asthe B and C loci are closely linked (Smith, 1975a),some similarity between the A-B and A-C locus inter-actions is not unexpected.Changesin phenotype frequencyTable4 shows changes in the frequency of alcippus,aegyptius,albinus and dorippus in the Kampala areaover 91 years, or slightly more, which is approximately1092 breeding generations, the generation time (egg toegg) of D. chiysippus in Uganda being about 1 month.The heterozygote weak alcippus (= alcippoides)(Aa cc)is included under alcippus (aa cc) and the heterozygoteweak albinus (=semialbinus)(Aa C - ) is includedunder albinus (aa C - );this unfortunate loss of geneticinformation is unavoidable for reasons given below.Although the three mainly pre-1900 samples areunlikely to be strictly random, and we do not have theexact dates for many of the specimens, they provide auseful baseline against which to measure subsequentchanges in phenotype frequency. Moreover, despite thecaveats, we are surprised to find that they are formally

246 D. A. S. SMITH ETAL.Table 4 Comparison of phenotype frequencies of Danaus ch,ysippus around Kampala, Uganda, over 100 years (approximately1092 generations)SourceDatealcippusaegyptiusalbinusdorippusNNatural History Museum (NHM), LondonMainly pre-190038.227.69.225.076Rothschild Collection (NHM)Mainly pre-190028.540.78.722.1172HopeDepartment(HD),OxfordMainlypre-190029.034.28.528.3272Wiggins Collection (HD)1909 - 1215.866.34.013.9101Komelsland,LakeVictoria(HD)194646.838.01.313.979OwenandChanter(1968)1964 - 6637.449.63.69.4530Owiny1971 - 7256.534.80.08.723Owen, Gordon and Smith199171.123.72.82.4211Values are percentages.homogeneous (x)= 6.295;0.5 > P> 0.3) and all differfrom the more recent ones in similar ways.As shown in Table 4, the frequency of dorippus hasdeclined steadily and consistently over the wholeperiod; albinus also declined up to 1946 but has beenfairly stable at low frequency since then. The alcippusphenotype shows the most dramatic change in fre-quency. It reached a low point of 16 per cent during1 909 - 12 and then increased steadily to reach 71 percent in 1991. The apparent drop in frequency during1964 - 66 may not be real as the standard errors for1946, 1964 - 66 and 1971 - 72 overlap.Figure 2 shows changes in frequency of the aa(+ some Aa) and cc phenotypes for the period1909 - 91. We start the graphs in 1900 by averaging thethree pre-1900 collections (Table 4). The graph for thecc phenotypes (aegyptius + alcippus) is based on directobservation; the data for the aa phenotypes(alcippus + albinus) are substantially inflated in Fig. 2(by 30.3 per cent for the 1991 sample) because of theinclusion with the recessive aa phenotype of some Aabutterflies (weak alcippus + weak albinus). This error isunavoidable, if the large random sample for 1964 - 66and one of the museum samples were included (whichis essential), to ensure comparable treatment of allsamples: the 1964 - 66 data (Owen & Chanter, 1968)are for butterflies marked and released and thereforenot available for examination. The study was carriedout before the genetics of D. chrysippus had been fullyworked out and all individuals with some white in thehindwing were scored as alcippus (or albinus).Regression coefficients are required to quantify withprecision the changes of phenotype frequency over theperiod 1910 - 91; unfortunately, the small number ofdata points combined with large variation in samplesizes precludes such treatment. Further analysis hastherefore to be limited to comparing phenotype fre-quencies at the start and end of this period and toexamining the regularity of the changes by rank corre-

1900 1910 1920 1930 1940 1950 1960 1970 1980YearFig.2Changesin frequency of the aa and cc phenotypes inDanaus chrysippus at Kampala, Uganda, during 1900 - 91.lation. For the forewing phenotype, aegyptius +alcippus (the cc phenotype) increased from 81.6 percent to 94.9 per cent; a z-test on these data (arcsinetransformed) shows that the change is very significant(z 2.697, 0.01 > P> 0.002). Moreover, the increase isprobably regular (Kendall's r= 1, P= 0.0 17). Makingdue allowance for imperfections in the data, the ccphenotype increased by an average of 0.16 per cent ayear or 0.0 13 per cent per generation. The white hind-winged phenotype (alcippus + albinus, including weak(heterozygous) forms) increased even more signifi-cantly by 54.1 per cent, from 19.8 per cent in 1910 to73.9 per cent in 1991 (z=7.043, P<0.001) althoughin this case the increase cannot be accepted as regular(v=0.8, P=0.08). The phenotype increase averages0.67 per cent a year and 0.056 per cent per generation.If the Hardy - Weinberg equilibrium were assumed (seeabove), these observations would suggest frequencyincreases over 81 years of approximately 7 per cent forthe c allele and, more tentatively, in the order of 40 percent for the a allele.

0a,a0Ca,0'a0>.0Ca,20

95a,0.0Ca,a000>0Ca,C.a,U-

POLYMORPHISM AND EVOLUTION IN D. CHRYSIPPUS 247Sustained changes in allele frequency of the magni-tude described could result from either introgressionfrom populations to the west of Uganda, or powerfulnatural selection, or possibly a combination of both.We are unable to estimate changes at the B locus, whichwas only recognized in 1973 (Clarke et al., 1973) butwe expect substantial change to have occurred in viewof the interactions between the A and B loci describedabove and between the closely linked B and C locidemonstrated in Tanzania.DiscussionSubstantialnegative linkage disequilibrium (repulsionphase chromosomes in excess) between the B and Cloci in D. chiysippus, based on breeding experiments,was described at Dar es Salaam (Smith, 1980); thisresult could not be tested or replicated in our smallfield sample from Kampala, where the frequency ofboth dominant alleles (B and C) is too low. Therefore,although the Dar es Salaam and Kampala populationsmay yet prove similar in having both linkage and non-gametic disequilibria, on current evidence they differfundamentally: the former shows strong negativelinkage disequilibrium for the linked B and C loci inboth sexes whereas the latter shows some other kind ofdisequilibrium, which we call genotypic, for both A andB phenotypes considered alone (Table 2) and for A-Band A-C combinations of unlinked loci (Table 3). Thegenotypic disequilibrium takes the form of marked sexdifferences for genotype frequency. The effect does notresult from orthodox sex linkage (Clarke et a!., 1973;Smith, 1 975a). For all three loci, both genotype fre-quencies and additive or epistatic interactions witheach of the other loci differ significantly between thesexes either at Dar es Salaam, Kampala or both.As the a and b alleles are subject to adverse selec-tion in females and positive selection in males, theopposing selection pressures, favouring alternativealleles in the two sexes, must result in balanced poly-morphisms at both A and B loci. The genotypicdisequilibria show that the A, B and C locuspolymorphisms in D. chiysippus, at Kampala as at Dares Salaam, are interlocked in interactions with eachother and with sex, possibly involving multiple feed-back effects. Allele frequencies must be maintained bya balance of selective forces, acting on a coadaptedgene complex (Lewontin, 1974), which operates differ-ently in the two sexes. We predict that disequilibria ofthe types described here will in time be established as ageneral feature of all polymorphic East African popu-lations of D. chrysippus but we are aware of the import-ance of replicating the Kampala and Dar es Salaamresults in a large random sample from at least one otherarea.

The breeding data of Owen & Chanter (1968) atKampala show some unusual segregation patterns forboth A and BC autosomes and sex chromosomes, inparticular the occurrence of all-female broods at highfrequency. Similar and more extensive results havebeen obtained at Dar es Salaam (Smith, 1975a,b,1976b, 1980) and at Nairobi (I. J. Gordon, unpub-lished data). Features common to all three areasinclude distorted sex and phenotype ratios togetherwith interaction between them. The causal geneticmechanisms are still under investigation and remain farfrom fully understood but we believe that sex - autosome translocations and either nuclear or cyto-plasmic lethal genes, at least one of which ismale-specific, are involved.A remarkable feature of our Kampala sample is thesymmetry of genotype frequency differences betweenthe sexes, the shortage of particular genotypes in onesex, when tested against the null hypothesis that auto-somal gene frequencies are expected to be the same inboth sexes, being almost exactly balanced by a surplusin the other (Table 3). We have evidence from Dar esSalaam suggesting that a Y-BC autosome translocationmay be carried out by some females (the female isheterogametic in Lepidoptera), resulting in transmis-sion of the normally autosomal B and C genes exclu-sively down the female line in all-female broods (Smith,1976b). (This explanation was originally rejected infavour of meiotic drive but we now have evidence(unpublished data) to suggest that the all-female broodsresult from the death of male embryos.) These inter-pretations are supported by unpublished data fromNairobi (I. J. Gordon) and Mombasa (D. Schneider).Thus we have evidence for the occurrence of all-femalebroods over a large area in East Africa but fromnowhere else.There is, however, no evidence from Kenya orTanzania for a sex - autosome translocation involvingthe A chromosome, which could provide an explana-tion, on similar grounds to that for the B locus above,for the considerable sex differences in A - andaagenotype frequencies at Kampala. Therefore we specu-late that a Y-A autosome translocation, inheritedexclusively by females, mainly involving the A allele,could explain the predominance of non-white hindwingin females compared with males at Kampala. It is prob-able that a combination of Y-autosome translocationpolymorphisms and sex-specific mortality in ovoaccounts for at least some of the differing genotypefrequencies between sexes that we have described inthis paper. Why such sex differences, whatever theirorigins and mode of inheritance, should have beensubject to positive selection over much of East Africais an open question that might in time be answered interms of the comparative ecology or behaviour of the

248 D. A. S. SMITH ETAL.sexes. A clear understanding of the genetic mechan-isms involved is obviously essential to further progressand will only be achieved through a breeding pro-gramme using Kampala material.The changes in genotype frequency in D. chrysippusat Kampala over the past 90 years (Fig. 2), in particularthe increase in frequency of alcippus, are similarin magnitude and time-span to the spread of thecarbonaria allele in the moth Biston betularia (L.)(Geometridae) in the U.K. in the nineteenth century(Kettlewell, 1973). The main difference is that themoth has oniy one generation a year so that the evolu-tionary change in the butterfly involved approximately12 times the number of generations, although within acomparable time-scale. The considerable geneticchanges described here could result either fromimmigration and gene flow or entirely from naturalselection in situ, or from a combination of the two. Ourdata provide no means of distinguishing all thesepossibilities, although two of our findings indicateemphatically that natural selection is involved. Firstly,the significant differences in genotype frequencybetween males and females for two of the three lociinvestigated indicate that, at individual loci, differentgenotypes have a selective advantage in the two sexes, asituation that must result in balanced polymorphisms,with equilibria determined by the relative strengths ofthe opposing selective forces. Secondly, the genotypicdisequilibrium within sexes found for two pairs ofunlinked autosomal loci constitutes prima facie evi-dence for the operation of natural selection at a higherlevel than the single gene locus or even the chromo-some but rather at the level of the integrated genecomplex or the genome (Lewontin, 1974; Gordon,1987).Although there is an impressive body of intuitiveevidence suggesting that mimicry has been an import-ant element in the evolution of polymorphism in D.chiysippus (Trimen, 1887; Poulton, 1912; Owen &Chanter, 1968; Owen, 1971; Gordon & Smith,1989), it is probable that sexual selection, as found atDar es Salaam (Smith, 1975b, 1981, 1984), and otherfactors which we are unable to identify, are alsoinvolved, especially in the evolution of sex differences.There are 157 species of danaine butterflies in theworld (Ackery & Vane-Wright, 1984), D. chrysippusbeing the only one with a conspicuous colour andpattern polymorphism covering a substantial geo-graphical area (some other species are polymorphic ona local scale where contact between geographical racescreates hybrid zones). The question thus arises as tohow the polymorphism originated and how it is main-tained in an unpalatable species (Rothschild et al.,1975) that in theory should be monomorphic. Con-

versely, why is D. chrysippus monomorphic overthe greater part of its very extensive geographicalrange, for example in West Africa, tropical Asia andAustralia?Form alcippus (aa cc) is the only form present intropical Africa west of Cameroun and the simplesthypothesis is that western alleles have been and still areinfiltrating central and eastern populations. One of us(D.A.S.S.) has proposed that the present sympatriccolour forms are derived from previously allopatricsubspecies which were isolated by forest barriersbefore or during the Pleistocene (Smith, 1980; Smith etal., 1988). D. chrysippus is a savanna butterfly and itspresent wide distribution in Africa must be a relativelyrecent phenomenon associated with a decrease in thearea covered by forest as a result of a drier climate,accelerated by an unprecedented impoverishment offorest vegetation from human activities, especially inthe last few hundred years. On this hypothesis there isnow hybridization between three until recently allo-patric subspecies: alcippus (aa B - ccand aa bb cc) tothe west of the formerly vast Chad Basin (Lake Mega-Chad), dorippus (AA bb CC) in the north-east(Somalia), and aegyptius (AA BB cc) in the south. Theabsence of full dominance at all three diallelic locisupports our belief that hybridization is a recent eventand unusual segregation for sex and other charactersmay indicate hybrid dysgenesis.The hypothesis of an allopatric origin for the Eastand Central African polymorphism is supported bysome persuasive evidence, the most pertinent being thepresent ranges of alcippus, dorippus and aegyptius,centred respectively in the west, north-east and southof Africa, where the populations remain monomor-phic. In the precise form stated by Smith (1980), thishypothesis is undoubtedly a simplification of the truestate of affairs. For example, aegyptius is by far the mostfrequent phenotype in Africa north of the Sahara,including the Nile Valley in Egypt and the CanaryIslands, whereas alcippus is occasionally encountered,usually as weak alcippus, in localities as far apart as theCanary Islands (Owen & Wiemers, 1992), Algeria (B.Samraoui, unpublished data) and Egypt (El-Aziz,1991). The allopatric hypothesis clearly requiresrefinement as follows. Firstly, as D. chrysippus rangeswithout a break across North Africa and the MiddleEast to the Indian subcontinent and beyond, the deri-vation of the North African populations may be fromthe Asian subspecies chiysippus, which differs forseveral characters from the rather similar form fromsouthern Africa. In recognizing these differences,Talbot (1943) accepted the name liboria Hulstaert(1931) for the distinctive race occupying the southernthird of the continent. Secondly, the alcippus gene

POLYMORPHISM AND EVOLUTION IN D. CHRYSIPPUS 249might have infiltrated to North Africa and the CanaryIslands from across the Sahara at some time in therecent past when much of the area was dry savanna.Another of us (D.F.O.) has proposed an hypothesisto account for the maintenance, as distinct from theorigin, of the East African polymorphism in terms ofBatesian mimicry theory (Owen, 1970). In Batesianmimicry the mimics, unless the models are exceedinglydistasteful or toxic (Brower, 1960), must generally beless frequent than their models otherwise predatorswould learn that butterflies of particular colourpatterns are usually palatable and the advantage gainedby mimicry would soon be lost. Although Batesianmimicry confers advantages to mimics through decep-tion, it is in general a threat to models as predatorsmust occasionally gain experience on palatable speciesand thence proceed to attack unpalatable ones(Edmunds, 1974). In East Africa in particular, D.chrysippus is a model for a considerable variety ofBatesian mimics, several of which (Hypolimnasmisippus L., Pseudacraea poggei Dewitz, Mimacraeamarshalli Talbot, Papilio dardanus Brown) are them-selves polymorphic, with alternative forms mimickingdifferent phenotypes of D. chrysippus (and in the lattercase, other model species). However, in West Africathere are few mimics; here D. chrysippus is mono-morphic and the alcippus phenotype has only oneBatesian mimic, the alcippoides phenotype of thefemale H. misippus which is always relatively infre-quent (Edmunds, 1969; Smith, 1976a). Thus, on thishypothesis, polymorphism in D. chiysippus was envis-aged to have arisen sympatrically in areas where thespecies had become threatened by an overload ofmimics; in contrast, where there were few mimics,polymorphism did not evolve.These two hypotheses can now be dovetailed toprovide an explanation both for the origin of the poly-morphism and for its persistence over such a largearea. Our data indicate a substantial increase in thefrequency of alcippus in Uganda over the past 100years, a period during which agricultural activities andforest destruction have provided vastly more suitablehabitat for D. chiysippus. Indeed, our entire study areawould have been under forest only a few hundred yearsago. We therefore envisage that alcippus might be arelatively recent (post-Pleistocene) invader to Ugandafrom the west. During this time, say 20,000 years, the aallele has reached the east coast of Kenya and Tanzaniaand spread through the Sudan to the Arabian penin-sular and beyond (where its frequency is low), creatinga west - east morph-ratio dine. The phenotype recurs,possibly as an independent mutant and sometimes athigh frequency, in West Malaysia (Corbet & Pendle-bury, 1978) and Sumatra where it forms Müllerian co-

mimicry rings with white hindwinged forms of Danausaffinis (Fabricius) and D. genutia (Cramer) (Ackery &Vane-Wright, 1984). These authors suggest the possi-bility that the 'alcippus' pattern is primitive in D.chrysippus on the grounds that the a allele is bothrecessive (it is actually only partially so) and is wide-spread at low frequency in the Oriental Region. Similarwhite hindwinged forms, however, also occur in racesof the closely related D. gilippus in the New World andin D. plexippus in Hawaii. It follows that recurrentmutation for white hindwing in the genus Danaus is anequally feasible explanation.We envisage that dorippus moved into Uganda fromits heartland in the north-east at some unspecified butprobably recent period and combined with alcippus toproduce albinus (which exists only where alcippus anddorippus are sympatric) whereas aegyptius moved upfrom the south (and possibly down from the north aswell). Similar events probably occurred and are occur-ring in other parts of the range of D. chiysippus whereit is now polymorphic. The picture is further compli-cated by the fact that the species is seasonally migra-tory, thus promoting gene flow in some parts of itsrange but apparently not in others.The polymorphism is probably now maintainedsympatrically as suggested by Owen (1970). It is neces-sary, however, to incorporate the historical hypothesisof allopatric origins with that of sympatric maintenanceof the polymorphism to overcome the problem pointedout by Turner (1977) that a novel aposematic mutant ina model species must be at a severe initial disadvantageas predators would fail to recognize it as distasteful.Even a gradual diversification of forms, through racialmingling and hybridization, could threaten the safety ofall of them as predators experience increasing difficultyin relating several different images with distastefulness.In particular, there is evidence that the rarer forms ofD. chiysippus are disproportionately at risk frompredation (Smith, 1979). The answer to this paradoxmay lie in the existence in Uganda of up to fourMüllerian co-mimicry rings comprising three sympatricsavanna species, D. chrysippus, Acraea encedon and A.encedana, with many of the phenotypes of each specieshaving matching forms in one or both of the others (A.encedon also has several non-mimetic forms). In thissituation it is possible for a new form, whether mutantor novel immigrant, to join an existing Müllerianmimicry ring and obtain immediate protection. More-over, if the various matching phenotypes of the threespecies had evolved by slow convergence in isolatedallopatric refuges and all expanded their ranges tooverlap and hybridize, more or less in unison and inresponse to the same habitat changes, this couldexplain the unique polymorphic Müllerian assemblages

250 0. A. S. SMITH ETAL.in East Africa. (It is possibly not a coincidence that allthree species share a feature, rare in butterflies,unrelated to mimicry and highly suggestive of hybriddysgenesis, namely female-biased populations with ahigh frequency of all-female brooding.) UnlikeBatesian mimicry, these associations would not beexpected to be number-dependent nor would theranges of the various phenotypes necessarily show aperfect match because all are protected. We arecurrently investigating the two Acraea species to assessthe extent of their polymorphic resemblances to differ-ent populations of D. chrysippus.AcknowledgementsOurresearch in Uganda in 1991 was financed by agrant from the Royal Society of London. We are alsograteful to the Nuffield Foundation for financial helpthrough a grant made to Sir Cyril Clarke, Departmentof Genetics and Microbiology, University of Liverpool.Derek Whiteley drew the figures. The authors wish tothank an anonymous referee whose comments greatlyimproved the manuscript.ReferencesACKERY,P. R. AND VANE-WRIGHT, R, .1984.Milkweed Butterflies:Their Cladistics and Biology, British Museum (NaturalHistory), London.BROWER, J. VAN Z. 1960. Experimental studies in mimicry. PartIV. The reactions of starlings to different proportions ofmodels and mimics. Am. Nat,, 94,271 - 282.CLARKE, C. A., SHEPPARD, P. M. AND SMITH, A. G. 1973. Thegenetics of fore and hindwing color in crosses betweenDanaus chrysippus from Australia and from Sierra Leone.J. Lep. Soc., 27, 73 - 77.CORBET, A. 5. AND PENDLEBURY, H. M. 1978. The Butterflies of theMalay Peninsular, 3rd edn, rev. Eliot, J. N. Kuala Lumpur,Malaysia.EDMUNDS, M. 1969. Polymorphism in the mimetic butterflyHypolimnas misippus L. in Ghana. Heredity, 24, 28 1 - 302.EDMUNDS, M. 1974. Defence in Animals.' a Survey of Anti-Predator Defences. Longman, Harlow.EL-AzIz, S. A. 1991. Biological and morphological studies onDanaus chrysippus and Pieris rapae (Lepidoptera:Rhopalocera) in Assiut. Unpublished M.Sc Thesis, AssiutUniversity, Egypt.GORDON, i. .i. 1984. Polymorphism of the tropical butterflyDanaus chtysippus L. in Africa. Heredity, 53, 583 - 593.GORDON, I. .j.1987.Natural selection for rare and mimeticcolour pattern combinations in wild populations of thediadem butterfly, Hypolimnas misippus L. Biol. J. Linn.Soc., 31, 1 - 23.GORDON, I. J. AND SMITH, D. A. S. 1989. Genetics of the mimeticAfrican butterfly Hypolimnas misippus: hindwing poly-morphism. Heredity, 63, 409 - 425.

HULSTAERT, R. r'. G. 1931. Lepidoptera Rhopalocera: Fam.Danaidae.' Subfam. Danainae and Tellervinae. GeneraInsectorum, Bruxelles, p. 193.KETFLEWELL, B. 1973. The Evolution of Melanism. Clarendon,Oxford.LEWONTIN, R. C. 1964. The interaction of selection and linkage.I. General considerations; heterotic models. Genetics, 49,49 - 67.LEWONTIN, R. c. 1974. The Genetic Basis of EvolutionaryChange. Columbia University Press, New York.OWEN, n. F. 1970, Mimetic polymorphism and the palatabilityspectrum. Oikos, 21, 333 - 336.OWEN, D. F. 1971. Tropical Butterflies. Clarendon, Oxford.OWEN, D. F. AND CHANTER, D. o. 1968. Population biology oftropical African butterflies. 2. Sex ratio and polymorph-ism in Danaus chrysippus L. Rev. Zoo!. Bot. Afr., 78,8 1 - 97.OWEN, D. F. AND WIEMERS, M. 1992. The butterflies of Fuerte-ventura. Entom. Gaz., 43, 87 - 92.PIERRE, i. 1973. Etude d'un cycle mimétique ayant Danauschrysippus (Linné) pour modèle et Hypolimnas misippus(Linné) (Nymphalidae) pour mime. Arch. Zool. Exp. Gen.,114,73 - 96.PIERRE, .i. 1980. Variation geographique du polymorphisme etdu mimétisme de Danaus chrysippus et d'Hypolimnasmisippus en Afrique et en Asie. C. r. somm. Seanc. Soc.Biogeogr., 486, 179 - 187.POULTON, E. B. 1912. Comment on Col. Mander's paper. Proc.Roy. Ent. Soc. Lond., 1912, vii - ix.ROTHSCHILD, M., VON EUW, 3., REICHSTEIN, J., SMITH, D. A. S. ANDPIERRE, j.1975.Cardenolide storage in Danaus chrysippus(L.) with additional notes on D. plexippus (L.). Proc. Roy.Soc. Lond., B, 90, 1-3 1.SMITH, D. A. S. 1 975a. Genetics of some polymorphic forms ofthe African butterfly Danaus chrysippus L. (Lepidoptera:Danaidae). Entom. Scand., 6, 134 - 144.SMITH, D. A. s. 1 975b. All-female broods in the polymorphicbutterfly Danaus chrysippus L. and their ecological signifi-cance. Heredity, 34,363 - 371.SMITH, D. A. 5. 1 975c. Sexual selection in a wild population ofthe butterfly Danaus chrysippus L. Science, 187, 664 - 665.SMITH, D. A. S. 1976a. Phenotypic diversity, mimicry andnatural selection in the African butterfly Hypolininasmisippus L. (Lepidoptera: Nymphalidae). Biol. J. Linn.Soc., 8, 183 - 204.SMITH, ID. A. s. 1976b. Evidence for autosomal meiotic drive inthe butterfly Danaus chiysippus L. Heredity, 36, 139 - 142.SMITH, D. A. 5. 1979. The significance of beak marks on thewings of an aposematic, distasteful and polymorphicbutterfly. Nature, 281, 215 - 216.SMITH, ID. A. 5. 1980. Heterosis, epistasis and linkage dis-equilibrium in a wild population of the polymorphicbutterfly Danaus ch.ysippus (L.). Zoo!. J. Linn. Soc., 69,87 - 109.SMITH, D. A. S. 1981. Heterozygous advantage expressedthrough sexual selection in an African butterfly. Nature,289, 174 - 175.SMITH, D. A. s. 1984. Mate selection in butterflies: competition,coyness, choice and chauvinism. In: Vane-Wright, R. I. and

POLYMORPHISM AND EVOLUTION IN D. CHRYSIPPUS 251Ackery, P. R. (eds) The Biology of Butterflies, AcademicPress, London, Symp. R. Ent. Soc. Lond., 11,225 - 244.SMITH, D. A. S., SHOESMITH, E. A. AND SMITH, A. G. 1988. Pupalpolymorphism in the butterfly Danaus chrysippus (L.):environmental, seasonal and genetic influences. Biol. .1.Linn. Soc., 33, 17 - 50.TALBOT, o. 1943. Rev. notes on the genus Danaus Kiuk (Lep.Rop. Danaidae). Trans. R. Ent. Soc. Lond., 93, 115 - 148.TRIMEN, R. 1887. South African Butterflies: a Monograph of theExtra-Tropical Species. 1. London.TURNER, S. R. u. 1977. Butterfly mimicry: the genetical evolu-tion of an adaptation. In: Hecht, M. K., Steere, M. C. andWallace, B. (eds) Evolutionary Biology, 10, 163 - 206,Plenum Press, New York.