Introduction

Danaus chrysippus is a widespread and abundant butterfly which is distributed throughout the Old World tropics in open habitats. In Uganda, Kenya, Tanzania and neighbouring countries, but not elsewhere, an extensive range of polymorphic forms is sympatric and they interbreed freely (Owen & Chanter, 1968; Smith, 1975a, 1980; Gordon, 1984). Unusual sex ratios have been described (Owen & Chanter, 1968; Smith, 1975b; Gordon, 1984), both in the field and within broods; less commonly, non-Mendelian segregations for colour genes have been reported, some of which are associated with all-female broods (Smith, 1976; Gordon, 1984) and others not (Owen & Chanter, 1968; Gordon, 1984). Moreover, there are several reports of morph ratios in the field which differ between the sexes (Smith, 1980; Smith et al., 1993, 1997). Here, I describe a type of non-Mendelian segregation not hitherto reported in D. chrysippus or, to my knowledge, in any other butterfly, together with evidence for variable penetrance of two recessive alleles contingent upon, in one case sex, and in the other, epistatic interaction with two other colour gene loci. The basic colour genetics of D. chrysippus is now fairly well known (Clarke et al., 1973; Smith, 1975a, 1980) (Fig. 1).

Fig. 1
figure 1

The four major phenotypes of Danaus chrysippus in Tanzania: (a) chrysippus, (b) alcippus, (c) dorippus, (d) albinus. The stippled areas are either tawny orange or nutbrown; the black and white areas are as shown. Hind-wing colour and pattern are controlled by the A locus: (a) and (c) have the genotype AA, (b) and (d) are aa; Aa heterozygotes are variable, being either indistinguishable from AA or showing varied amounts of white (weak alcippus and albinus), especially lining the veins, but always less than in aa. Ground colour is controlled by the B locus with two alleles, brown (B) being variably dominant over orange (b); Bb heterozygotes are either brown or, more usually, show a variable extent of brown on the costal and basal areas of the forewing and the basal area of the hindwing, the remaining areas being orange. The pattern of the forewing apex is governed by the C locus; (c) is CC whereas (a) and (b) are cc; (d) has a transiens (Cc) fore-wing. The B and C loci are closely linked with a cross-over value of 1.9% in males only (Smith, 1975a) whereas the A chromosome segregates independently; none of the colour genes is sex-linked (Clarke et al., 1973; Smith, 1975a).

Methods

The founder pair for this study was taken in copula in the field at Dar es Salaam, Tanzania. Eggs were obtained in a sleeve on a branch of the foodplant Calotropis gigantea. All the females used are direct descendants of the founder female but wild males were sometimes introduced to obtain required matings. Pairing took place in an outdoor insectary from which mated females were removed and placed in sleeves. All broods were reared in separate cages and fed on C. gigantea.

Results

Non-Mendelian segregation (broods 82–85)

The broods used in this study are shown in Table 1. The founder brood 82 (Table 2) shows some unusual features which stimulated the remainder of the breeding programme. Segregation is 16:11 (1:1) for orange (bb):brown (Bb) (omitted from Table 2 for reasons of clarity), 20:7 (3:1) for dorippus (A–C–): albinus (aaC–) and 10:17 for sex. The most obvious anomaly is the absence of cc forewing phenotypes (chrysippus or alcippus) from the progeny; because both parents are visibly Cc (transiens), a 3:1 segregation, including 1/4×27=6.75 cc, is expected. Moreover, the eight Cc heterozygotes with penetrant c (transiens) are distributed nonrandomly with respect to the A locus, as seven of them have the visually determinable genotypes aa Bc/bC or aa bC/ bc, whereas the aa bC/ bC and aa Bc bc genotypes are both missing. Assuming random assortment for the A and C loci, H0 for a 9:3:4 segregation (with the A–cc and aacc classes amalgamated) is rejected (χ22=9.02; 0.02> P0.01). Analysed in four classes without pooling, the combined segregation for the A and C loci has an exact probability (two-tailed) of 1.8×10−5. Association of the A and bC chromosomes (n=12) and the a and bc chromosomes (n=4), combined with the absence of the alternative arrangements, a with bC and A with bc, has a highly significant exact probability (two-tailed) of 1.1×10−5.

Table 1 Eight broods of Danaus chrysippus reared at Dar es Salaam, Tanzania
Table 2 Segregation for the A and C loci in brood 82 of Danaus chrysippus (expected numbers in parentheses): the parental genotypes are Aa Bc/bC (female) and Aa bC/bc (male)

By deduction, the array of gamete pairings found in brood 82 (Table 3) suggests that half of those predicted, assuming random assortment, are absent (four cells are in doubt because penetrance of the a and c alleles is impossible to estimate), whereas all observed pairings exceed expectation. And yet, it is clear that all gametes, with the possible exception of A/bc in males, are produced and functional.

Table 3 Observed numbers of gamete pairings and their deduced compatibilities in brood 82 of Danaus chrysippus

Three F2 broods (Table 4) were bred from the brood 82 progeny to generate better data for the interpretation of the non-Mendelian segregations observed and, as all selected parents were aa, to reduce genetic noise. Parents were selected for their visually identifiable B- and C-locus genotypes. Furthermore, brood size was maximized and careful egg–adult mortality records kept. Not only were all the female parents sibs, but the same male sired two broods.

Table 4 F2 broods of Danaus chrysippus from which two expected phenotypic classes are missing in both sexes

The segregation expected with independent assortment is 1:1:1:1 for Bc/ bC, bC/ bC, bC/ bc and Bc/ bc: however, the large progenies probably comprise only two of the expected four genotypes. Yet it is clear in all these broods that both parents contribute both their BC chromosomes equally to offspring of both sexes. How can I be sure that all the orange albinus should be scored as aa bC/ bc rather than aa bC/ bC? Two facts support the interpretation that the latter genotype is missing. First, the penetrance of c in these broods is slightly above average and homogeneous, 47% in bC/ bc and 81.9% in Bc/ bC, compared to 45% (n=129) and 74.4% (n=173), respectively, in known backcross progenies from Tanzania (Smith, 1980); secondly, all three brood segregations fit a 1:1 expectation; there is no surplus of orange (bb) individuals which would be expected if both bC/ bc and bC/ bC genotypes were recovered. A parsimonious interpretation is that the bC/ bC genotype is missing, as is the easily recognized Bc/ bc genotype. Note that both these genotypes are also absent from the parental brood 82.

Penetrance of the c allele in Cc heterozygotes

Penetrance of the a and c alleles clearly varies with genetic background in broods 82–85. Broods 87–95 were reared to provide larger samples for estimating penetrance against a wider variety of genetic backgrounds. The results for the C locus (Table 5) show that penetrance of c in Cc genotypes is higher in a Bb (72.3%) than a bb (44.9%) background and in aa (67.6%) compared to A- (49.0%) genotypes. Because the χ2-value for interaction is not significant, the two effects are probably additive, the aaBbCc genotype showing penetrance of 83.3%, compared to 0% for A–bbCc (this sample is small (n=12) and the estimate atypically low). The highly significant interaction between the A- and B-locus segregations is caused by the non-Mendelian effects described above for broods 82–85. Penetrance is probably not generally affected by sex (χ21=3.44, 0.1> P0.05) although the influence of sex is significantly heterogeneous (χ23=10.99, 0.02> P0.01) and requires further investigation.

Table 5 Penetrance of the c allele in Cc genotypes of Danaus chrysippus against different genetic backgrounds (broods 82–95) (expected numbers in parentheses)

Penetrance of the a allele in Aa heterozygotes

In contrast to the C locus, penetrance of a in Aa heterozygotes (Table 6) is strongly influenced by sex, as it is significantly higher (62.5%) in males than females (25.5%). It is not, however, affected by C-locus genotype; a possible B-locus effect cannot be tested with these data as broods 87–88 are entirely brown.

Table 6 Penetrance of the a allele in Aa genotypes of Danaus chrysippus against different genetic backgrounds (broods 87–88) (expected numbers in parentheses)

Discussion

Non-Mendelian segregation

The segregations in broods 82–85, and especially the last three, suggest at first glance the operation of a balanced recessive lethal system. This interpretation is, however, effectively ruled out by the mortality data for broods 83–85 Table 4 which show that the mean egg–adult mortality of 26.4% is well below the minimum of 50% expected if two classes from a 1:1:1:1 segregation were to suffer total mortality. Brood mortality is, moreover, well below the average of 39.8% (n=52 broods measured) for bisexual broods reared by me in Tanzania and is homogeneous across the three broods. Preoviposition mortality or reabsorption of eggs probably never occurs in butterflies because eggs are fertilized singly, immediately before laying, as they pass the opening from the bursa copulatrix. Usually females carry only one fertilized egg at a time.

In broods 83–85 (Table 4), all gametes are viable and apparently produced by both sexes in equal numbers. As both types of sperm are equally recovered in the progeny, sperm competition is ruled out. Compatible pairings are Bc with bC and bC with bc and incompatible pairings bC with bC and Bc with bc. As these four pairings behave, in terms of compatibility, identically among the three F2 broods and the parental brood 82, the mechanism must be inherited. Moreover, because broods 83 and 85, on the one hand, and 84 on the other, are reciprocal crosses, compatibility must be irrespective of whether the haploid genomes involved are carried by sperm or eggs. These results suggest the operation of a prezygotic isolating mechanism which operates at the point of fertilization.

It may be relevant that the parents of brood 82 are heterozygous for three genes with alleles which are at fixation in the allopatric subspecies chrysippus (AAbbcc) (North Africa and Asia), liboria (AABBcc) (southern Africa), alcippus (aabbcc) (West Africa) and dorippus (AAbbCC) (north-east Africa), all of which overlap and interbreed in Tanzania. Recombination through independent A/ BC assortment in both sexes and B/ C cross-over in males has resulted in many phenotypes (transiens (–––– Cc), brown dorippus (A–B–C–) and albinus (aa––C–) which are confined to the postulated hybrid zone. It may be that prezygotic isolation, together with disturbed sex ratios, indicates incipient speciation involving two or more of these subspecies.

Variable penetrance of the a and c alleles

Penetrance of the a allele is apparently unaffected by genotype at the C locus and a possible effect from the B locus remains to be investigated. The partial sex-limitation of penetrance is possibly a stage in the evolution of full sex-limited inheritance which is rather common in butterflies (though not in Batesian models such as D. chrysippus), e.g. Papilio dardanus (Clarke & Sheppard, 1963) and P. memnon (Clarke et al., 1968), Hypolimnas misippus (Smith & Gordon, 1987; Gordon & Smith, 1989) and H. bolina (Clarke & Sheppard, 1975), Colias and Catopsilia species, although in all these cases it is the male in which gene penetrance is suppressed. However, in Pseudacraea eurytus (Carpenter, 1949; Owen & Chanter, 1972) sex-limited phenotypes occur in both sexes. Expressivity of a in the Aa genotype is very variable; at the extremes, individuals with, on the one hand, a few white scales lining the hindwing veins and, on the other, a white patch scarcely smaller than found in aa individuals, suggests that evolution of dominance has not occurred. It is my impression that, not only penetrance but also expressivity (which has not been measured), is higher in males. Owen & Chanter (1968), working in Uganda, found that the frequency of weak alcippus (Aa) was significantly lower in the field than in a sample reared from wild-collected larvae. They suggested that selection for distinctiveness between the polymorphic forms, which improves mimetic resemblance to H. misippus and Acraea encedon (now known to include A. encedana (Owen et al., 1994)), might be responsible. If this is the case, selection for improved mimicry is stronger in the female than the male.

Interaction of the C and c alleles in C-locus heterozygotes differs from the A-locus interaction in that the phenotype of the transiens (Cc) form (with penetrant c) is much closer to dorippus than to chrysippus, indeed so much so that it is unlikely that any predator would discriminate between transiens and dorippus. In this case dominance may have evolved, an hypothesis which is testable by crossing distantly allopatric races. On the other hand, the near dominance of the C allele may be a simple consequence of the biochemistry of gene action.

The penetrance data at both A and C loci, showing as they do multiple epistatic interactions with sex and other loci, which in turn have variable penetrance, serve to emphasize the practical impossibility on present knowledge of measuring allele frequencies at any of the A, B or C loci directly from field samples. Moreover, the occurrence of non-Mendelian segregations for phenotype, not only of the type described here, but also associated with all-female broods (Smith, 1975b, 1976; Gordon, 1984) and bisexual broods (Gordon, 1984) strongly suggests that intergeneration changes of colour gene frequencies are at present hard to predict. The highly complex set of interactions in the polymorphic population at Dar es Salaam has probably arisen as a result of the recent admixture of several well-differentiated subspecies to form an extensive hybrid zone.