Abstract

Mitochondrial and nuclear DNA information was analysed among four subspecies of the African Queen butterfly, Danaus (Anosia) chrysippus sensu lato (s.l.), along with four other Danaus species drawn from all three subgenera (D. (Danaus) plexippus, D. (Salatura) genutia, D. (A.) gilippus, D. (A.) eresimus) and two outgroup species from the same tribe, Tirumala septentrionis and Amauris niavius. A mitochondrial phylogeny derived from the 12S rRNA (347 bp) and COI (537 bp) loci indicates two very distinct haplotypes for subspecies D. (A.) c. dorippus, dorippus-1 and dorippus-2. Interestingly, dorippus-1, on the one hand, and all other D. (A.) chrysippus haplotypes, on the other, are the most distantly related clades within the genus and have different most recent ancestors from different subgenera, though sharing the common ancestor of the monophyletic genus. A phylogeny based on the EF1-α nuclear locus (400 bp) shows that the two well-separated mitochondrial lineages of dorippus are identical for this gene and reciprocally monophyletic to the other D. (A.) chrysippus lineages. Thus, nuclear and cytoplasmic phylogenies are not only discordant, but also suggest that both D. (A.) chrysippus s.l. and subspecies dorippus are polyphyletic. Paradoxically, four African subspecies, chrysippus-orange, chrysippus-brown, alcippus and dorippus, though substantially vicariant, hybridize extensively in East Africa wherever the ranges of two or more of them overlap. Linkage disequilibrium, and hence sexual isolation, in sympatry between colour (nuclear) genes and unlinked mitochondrial (cytoplasmic) loci is consistent across populations and therefore indicates the operation of positive natural selection. Together with data from previous experimental and field work, our results suggest that extensive hybridization occurs among once allopatric or parapatric lineages, that are now nascent species. We deduce that hybridism among lineages in sympatry is currently enforced, in the face of assortative mate choice, by a bacterial symbiont, Spiroplasma, a male-killer that forces females in female-biased populations to pair with heterotypic males. In discussion we emphasize that neither D. (A.) chrysippus s.l. as presently circumscribed, nor its component clades, conform to any established concept of species.

INTRODUCTION

The genus Danaus (monarchs and queens), subtribe Danaina, tribe Danaini (milkweed butterflies), subfamily Danainae, family Nymphalidae, is currently divided into three non-coordinate subgenera: Danaus, Salatura and Anosia (Ackery & Vane-Wright, 1984). This paper describes the polyphyly found in the African Danaus (Anosia) chrysippus complex. Hereinafter, we omit the subgeneric notation.

Throughout the paper the epithet ‘chrysippus’ is necessarily used in three senses:

  1. chrysippus sensu lato (s.l.)’ is the specific epithet of the D. chrysippus species complex.

  2. In areas where D. chrysippus s.l. is polymorphic, i.e. within the hybrid zone, ‘chrysippus sensu stricto (s.s.)’ refers to a morph identified by colour pattern (see later) that is sympatric with other named morphs (e.g. dorippus and alcippus) of D. chrysippus s.l.

  3. In parts of Africa and elsewhere, where D. chrysippus s.l. is monomorphic, four largely allopatric phenotypes (genotypes) −chrysippus-orange (chrysippus-o) (AAbbcc), chrysippus-brown (chrysippus-b) (AABBcc), dorippus (AAbbCC) and alcippus (aabbcc or aaB-cc) − are subspecies (Fig. 1).

Figure 1.

Geographical distribution of the colour forms of Danaus chrysippus s.l. in the Afrotropical region, showing the approximate boundaries of the hybrid zone and locations of the main sampling sites. Distributions of the four major African phenotypes (genotypes) of D. chrysippus s.l.: chrysippus-orange (AAbbcc), chrysippus-brown (AABBcc), alcippus (aabbcc) and dorippus (AAbbCC), are shown in the areas where the respective forms occur alone. Additional forms found commonly in the hybrid zone, but rarely outside it, are alcippoides (Aabbcc), transiens (A-BbCc), albinus (aabbC-) and semialbinus (AabbC-).

Figure 1.

Geographical distribution of the colour forms of Danaus chrysippus s.l. in the Afrotropical region, showing the approximate boundaries of the hybrid zone and locations of the main sampling sites. Distributions of the four major African phenotypes (genotypes) of D. chrysippus s.l.: chrysippus-orange (AAbbcc), chrysippus-brown (AABBcc), alcippus (aabbcc) and dorippus (AAbbCC), are shown in the areas where the respective forms occur alone. Additional forms found commonly in the hybrid zone, but rarely outside it, are alcippoides (Aabbcc), transiens (A-BbCc), albinus (aabbC-) and semialbinus (AabbC-).

Almost 40 years ago, Owen & Chanter (1968), investigating D. chrysippus s.l. at Kampala, Uganda, established that three colour forms, alcippus, dorippus and chrysippus s.s., treated as subspecies by Talbot (1943), were sympatric throughout the year. [Neither Talbot nor Owen and Chanter distinguished chrysippus-o from chrysippus-b, though both occur at Kampala (Smith et al., 1993) where the latter pair worked.] Owen and Chanter reared broods from females that had mated in the wild and found hybridism to be commonplace. Most hybrids were phenotypically identifiable and occurred at high frequency in the field. From limited observations, they supposed that mate selection was random and hybrid forms at little or no disadvantage. In the same paper, Owen and Chanter presented the first maps, based on study of museum material, to show that the three forms are largely allopatric, but with distributions that overlap extensively in Central and East Africa in what we term the ‘hybrid zone’ (Fig. 1). The geographical distributions have subsequently been refined but essentially confirmed (Rothschild et al., 1975; Smith et al., 1998; Lushai et al., 2003a, 2003b, c)

Genetic control of colour pattern in D. chrysippus s.l. is well known (Owen & Chanter, 1968; Clarke, Sheppard & Smith, 1973; Smith, 1975a, 1980, 1998; Gordon, 1984; Smith et al., 1998; Lushai et al., 2003a; see colour comparisons in Fig. 1). Ground colour is controlled by the B locus with two alleles, brown (B) being variably dominant over orange (b); Bb heterozygotes may be brown, but are usually intermediate between the BB and bb genotypes (Smith, 1975a). Forewing pattern is governed by the C locus: dorippus and albinus are CC; chrysippus s.s. and alcippus are cc. Most Cc heterozygotes are the visually identifiable form transiens (Smith, 1998). The B and C loci are closely linked with a cross-over value, in males only, of 3.8% (Smith, 1975a) and, at Dar es Salaam, Tanzania, strong linkage disequilibrium for the repulsion phase (Smith, 1980). transiens in Kenya, where chrysippus s.s. is almost invariably bc/bc, is generally orange (bC/bc) whereas at Dar es Salaam, where most chrysippus s.s. are brown (Bc/Bc), transiens (Bc/bC) is predominantly brown (Smith, 1980). Hindwing colour is governed by the autosomal A locus that is unlinked to the B/C loci (Clarke et al., 1973): chrysippus s.s. and dorippus with orange or brown hindwings are AA; alcippus and albinus with white hindwings are aa; Aa heterozygotes such as alcippoides and semialbinus are intermediate between AA and aa.

The partial vicariance of alcippus, chrysippus-o, chrysippus-b and dorippus probably reflects past geographical isolation (Lushai et al., 2003a). At the present time, however, two or more of these forms are found in permanent sympatry across Uganda, southern Kenya, Tanzania and parts of several neighbouring countries, where they form polymorphic, interbreeding populations. The year-round status of D. chrysippus s.l. in Dhofar (southern Oman) around Salalah (Fig. 1) is not clear. Whereas chrysippus-o, dorippus and hybrids may be locally abundant in the aftermath of the variable monsoon rains (September−October; Lushai et al., 2003a), by November they are scarce, which suggests their presence may be seasonal. Studying interactions in Africa between the orange/brown hind-winged forms dorippus, chrysippus-o and chrysippus-b, on the one hand, and the white hind-winged alcippus on the other, Lushai et al. (2003a) showed that, despite widespread hybridism, the forms are partially isolated even in sympatry, thus suggesting a recent history of allopatric evolution. Furthermore, both hybridism and heterozygote excess were inferred to be, at least in part, pathological symptoms of differences between chrysippus s.s. and alcippus in susceptibility to maternally inherited, male-killing, bacterial (Spiroplasma) infection (Jiggins et al., 2000), as this affects the sex ratio and, hence, might enforce heterotypic mating. Here, we return to the same theme (see Lushai et al., 2003a), but focus instead on forms dorippus and chrysippus s.s.

Because D. chrysippus s.l. is an aposematic species that is chemically defended (Brower, Edmunds & Moffitt, 1975, 1978; Rothschild et al., 1975) and dominates several African mimicry cycles (Smith, 1973a, 1976, 1979; Owen & Smith, 1993; Owen et al., 1994), both colour polymorphism and the ubiquity of non-mimetic hybrid phenotypes, sometimes at high frequencies, are paradoxical. Dominance has not evolved at any of the three colour gene loci, as expected for switch genes in populations of a visually polymorphic and mimetic species (Sheppard, 1967). Since most heterozygotes are identifiable by a phenotype that is intermediate between the two homozygotes, they more likely result from hybridism among once isolated, nascent species (Lushai et al., 2003a), rather than from crosses between morphs of a polymorphic species. The extensive area of so-called ‘polymorphism’, which we designate and delineate as the ‘hybrid zone’ (Fig. 1), is unique among the 157 species of aposematic and distasteful milkweed butterflies (Ackery & Vane-Wright, 1984).

chrysippus-o, chrysippus-b and dorippus are not only substantially vicariant, but also mate assortatively where they fly together (Smith, 1984). Moreover, they differ in migration behaviour into and away from zones of sympatry (Smith & Owen, 1997; Lushai et al., 2003b) and, consequently, in seasonal abundance within such areas (Smith et al., 1997). Sex ratios and, by deduction, levels of Spiroplasma infection, also differ among sympatric colour morphs (Smith, 1975b; Gordon, 1984; Smith et al., 1997; Lushai et al., 2003a). Thus, there is evidence to suggest that, despite widespread hybridism in East Africa, the colour morphs are to some degree sexually isolated, even in sympatry. If isolation is maintained among colour forms in sympatry, it should be possible to find evidence for linkage disequilibrium between unlinked genes in these polymorphic and interbreeding populations.

To investigate the possibility of linkage disequilibrium between nuclear and cytoplasmic genomes, we present new DNA sequence data for the mitochondrial (mtDNA) 12S rRNA (12S), cytochrome-c oxidase subunit I (COI) and the nuclear (nDNA) elongation factor 1-alpha (EF1-α) genes of chrysippus-o, chrysippus-b and dorippus, for which the colour inheritance is well known, from both monomorphic and polymorphic populations of D. chrysippus s.l. in Africa, the Middle East and India.

METHODS

Collection of butterflies

Butterflies comprising samples from which mtDNA and nDNA sequence information was extracted were collected randomly in the field in 1998–2001, boxed and later killed in ethyl ethanoate vapour, immediately before storage in 95% ethanol at −20 °C. We aimed to collect each race of D. chrysippus from at least one area where it is monomorphic, as well as from polymorphic sites; this aim was achieved in every case except for D. c. dorippus, which is monomorphic only in Somalia where it was unsafe to collect. Voucher specimens (wings) of all butterflies that provided DNA were retained in papers and deposited at the Hope Department of Entomology, Oxford University Museum of Natural History (OUMNH). Dar es Salaam, Kampala and Athi River material was collected from the same sites from which we have substantial field and breeding data (Owen & Chanter, 1968; Smith, 1975b, 1980, 1998; Gordon, 1984; Smith et al., 1993, 1997, 1998). Fresh specimens of dorippus, chrysippus-o and chrysippus-b were collected to check morphology on criteria given by Ackery & Vane-Wright (1984).

Extraction, amplification and sequencing ofDNA

The new data consist of partial mtDNA (12S, 347 bp and COI, 537 bp), and nDNA (EF1-α, 400 bp) sequences. MtDNA was extracted from samples of Tirumala septentrionis and nine Danaus taxa, and nDNA from Amauris niavius and seven Danaus taxa (Table 1). Methods were as described for the 12S gene by Lushai et al. (2003a). Additional to details provided elsewhere, primers for the COI gene were 5′−TTGATTTTTTGGTCATCCAGAAGT−3′ (forward) and 5′−ATACTTCTCTAGCATATAAAG−3′ (reverse): this produced a fragment of ∼700 bp. To nullify the possibility that PCR had amplified nuclear copies of COI mitochondrial genes (pseudogenes), we carried out a second run on selected individuals (see Table 1) with different primers, 5′−GGAGGATTTGGAAATTGATTA GTTCC−3′ (forward) and 5′−TCCAATGCACTAATCT GCCATATTA−3′ (reverse), which amplified a shorter fragment of ∼550 bp. In every case, the COI sequences obtained from the same individuals with the alternative primer (N = 10) were in agreement with those using the initial primer set.

Table 1.

Provenance, sample size (N) and GenBank references† for individual danaine butterflies sequenced for both 12S rRNA and COI mitochondrial loci and the EF1-α nuclear gene. Sample sizes in parentheses indicate individuals sequenced twice for COI using alternative primers

Species/form Provenance GenBank numbers and sample sizes
 
12S rRNA COI N EF1-α N 
Amauris niavius Tanzania – – – AY296132  1 
Tirumala septentrionis Malaysia AF389888 AF394182 – – 
Danaus plexippus Australia, USA AF389889 AF394183 9 (1) AY296133  1 
D. genutia Thailand, Malaysia AF389892 AY256344 2♯ AY296134  1 
D. eresimus Cayman Is. AF389895 AF394185 AY296135  1 
D. gilippus Cayman Is. AF389896 AF394186 – – 
D. c. chrysippus-o Oman*, India, Kenya AF389902 AF394192 17 (2) AY296139  2 
D. c. chrysippus-b Tanzania, Zambia, Uganda AF389903 AF394193 23 (2) AY296143  2 
D. c. alcippus Ghana, Uganda, Oman AF389901 AF394191 20 (1) – – 
D. c. dorippus-1 Kenya AF389890 AF394184 12 (2) AY296136  2 
D. c. dorippus-2‡ Kenya*, Uganda, Oman, Tanzania AF389900 AF394190 19♯ (2) AY296137  2♯ 
Total    126 (10)  12 
Species/form Provenance GenBank numbers and sample sizes
 
12S rRNA COI N EF1-α N 
Amauris niavius Tanzania – – – AY296132  1 
Tirumala septentrionis Malaysia AF389888 AF394182 – – 
Danaus plexippus Australia, USA AF389889 AF394183 9 (1) AY296133  1 
D. genutia Thailand, Malaysia AF389892 AY256344 2♯ AY296134  1 
D. eresimus Cayman Is. AF389895 AF394185 AY296135  1 
D. gilippus Cayman Is. AF389896 AF394186 – – 
D. c. chrysippus-o Oman*, India, Kenya AF389902 AF394192 17 (2) AY296139  2 
D. c. chrysippus-b Tanzania, Zambia, Uganda AF389903 AF394193 23 (2) AY296143  2 
D. c. alcippus Ghana, Uganda, Oman AF389901 AF394191 20 (1) – – 
D. c. dorippus-1 Kenya AF389890 AF394184 12 (2) AY296136  2 
D. c. dorippus-2‡ Kenya*, Uganda, Oman, Tanzania AF389900 AF394190 19♯ (2) AY296137  2♯ 
Total    126 (10)  12 
*

Samples were taken from a single site in each country (see Table 3 for D. chrysippus s.l. sites) with the exception of dorippus-2 from Kenya, which was collected at six locations, Athi River (near Nairobi), Masai Mara, Amboseli, Mombasa, Malindi and Galana River; chrysippus-o from Oman came from two locations, near Nazwa in the north (N = 5) and near Salalah in the south (Dhofar) (N = 9).

Because of hybridism among D. chrysippus s.l. forms in East Africa and Oman, the GenBank numbers shown are for those sequences that are diagnostic for their respective colour forms (Table 2) in areas where the species is monomorphic or, in the case of dorippus, nearly so. There was minor variation at non-diagnostic sites (Table 2) and some mtDNA introgression at hybrid zone sites (Table 3). ♯D. genutia and six D. c. dorippus-2 individuals were sequenced using only the alternative COI primer.

Six dorippus-2 individuals have chrysippus-o, chrysippus-b or alcippus cytoplasm (Table 3) and one has the chrysippus-oEF1-α allele.

Additionally, as there are many populations containing only one haplotype and no evidence for heteroplasmy in those with several, confusion with nuclear copies of cytoplasmic genes (pseudogenes) is improbable. Furthermore, the dorippus butterflies with divergent and controversial COI sequences have equally aberrant 12S sequences and nuclear copies of mitochondrial rRNA genes are unknown. Some butterflies, e.g. D. genutia (N = 2) and six D. chrysippus s.l. from Kenya (Table 1), were only amplified using the second set of COI primers; therefore, to include all results in a phylogenetic tree, only 537 bp of sample sequences were used in tree models and the calculation of genetic distances among D. chrysippus s.l. forms. Primers 5′−CGTGAACGTGGTATCACTATTGATATTGC−3′ (forward) and 5′−CTTCAGGGAGGGCTTCGTGATGC−3′ (reverse) were used to amplify the nDNA EF1-α gene.

These primers were used in PCR amplifications (see Lushai et al., 2003a) carried out in an Omnigene (Hybaid Ltd), using a hot start at 94 °C/2 min (1-cycle), denaturation at 92 °C/1 min, annealing at 49–55 °C/1 min (dependent on primer specificity), extension at 72 °C/2 min (30 cycles) and a final extension step of 72 °C/7 min (one cycle). Sequencing for the 12S gene is detailed in Lushai et al. (2003a) and the same procedure was used for the COI and EF1-α genes. In all cases, both sense and antisense fragments were sequenced, screened by eye with CHROMAS 1.45 or ALIGN-IR 1.2 and exported as text files. They were formatted as interleaved consensus sequences by sample, prior to multiple alignment using CLUSTAL X (1.5b) (Thompson, Higgins & Gibson, 1994). Sequences obtained were also screened against the National Center for Biotechnology Information (NCBI) GENBANK to compare for homology in the BLAST-NR database. For example, 89% homology was shown for COI sites with the Nymphalid butterfly Phycoides vesta (GenBank ref. AY156686) (Wahlberg, Oliveira & Scott, 2003a) and 94% for EF1-α sites based on a sequence from a close outgroup Amauris ellioti (AY218253) (Wahlberg, Weingartner & Nylin, 2003b). All COI and EF1-α sequences described here were submitted to GenBank using the NCBI submission program SEQUIN 3.7 (for other accession numbers see Table 1).

Phylogenetic analysis

The phylogenetic trees for 12S + COI (Fig. 2) and EF1-α (Fig. 3) are rooted by Tirumala septentrionis and Amauris niavius, respectively, as outgroups. A null hypothesis that the 12S and COI data-sets for Danaus genetic distances are consistent was tested by a Mantel Test, which gave r = 0.83; P < 0.005; this significant result supports the null hypothesis and justifies our combining sequence data for the two loci (Lushai et al., 2003c). The tree topologies for terminal clades, analysed using PAUP 4.0b (Swofford, 1998), are strongly supported for both mitochondrial loci by Maximum Parsimony (MP) and Maximum Likelihood (ML) algorithms. Bootstrap support values (1000 replicates) are indicated along with Bremer support values prior to each node where the support is significant (Decay indices were calculated using Autodecay 4.0; Eriksson, 1999). Branch lengths are drawn proportional to nucleotide changes, with branch significance (P < 0.01) indicated, calculated using ML analysis (PHYLIP 3.573c; Felsenstein, 1993). Otherwise, genetic distances (%) described in the text are based on the uncorrected-p model.

Figure 2.

One of two most parsimonious trees, based on amalgamated mtDNA 12S rRNA (347 bp†) and COI (537 bp†) genes for the genus Danaus rooted by outgroup Tirumala septentrionis (†values before alignment). Tree statistics: characters in matrix 887, variable uninformative characters 56, parsimony informative characters 71, heuristic search length 168, consistency index = 0.85, homoplasy index = 0.15, retention index = 0.80, rescaled consistency index = 0.68. 1000 Bootstrap replicates/Bremer Support values are shown for each node where significant. Branch lengths are drawn proportional to nucleotide changes and indicated in parentheses. All branch lengths have Maximum Likelihood (ML) branch significance (P < 0.01) unless marked *. The Parsimony topology depicted here is congruent with Distance (NJ) and ML algorithms not shown. See Table 1 for provenances and sample sizes.

Figure 2.

One of two most parsimonious trees, based on amalgamated mtDNA 12S rRNA (347 bp†) and COI (537 bp†) genes for the genus Danaus rooted by outgroup Tirumala septentrionis (†values before alignment). Tree statistics: characters in matrix 887, variable uninformative characters 56, parsimony informative characters 71, heuristic search length 168, consistency index = 0.85, homoplasy index = 0.15, retention index = 0.80, rescaled consistency index = 0.68. 1000 Bootstrap replicates/Bremer Support values are shown for each node where significant. Branch lengths are drawn proportional to nucleotide changes and indicated in parentheses. All branch lengths have Maximum Likelihood (ML) branch significance (P < 0.01) unless marked *. The Parsimony topology depicted here is congruent with Distance (NJ) and ML algorithms not shown. See Table 1 for provenances and sample sizes.

Figure 3.

One of two most parsimonious trees based on the EF1-α gene for the genus Danaus rooted by outgroup Amauris niavius. Tree statistics: total characters in matrix 400 (two excluded), variable uninformative characters 23, parsimony informative characters 5 of 8, heuristic tree length 29, consistency index = 0.96, homoplasy index = 0.03, retention index = 0.88, rescaled consistency index = 0.84. Other details are as in Fig. 2. D. c. alcippus and D. gilippus are missing from this analysis as the PCR failed.

Figure 3.

One of two most parsimonious trees based on the EF1-α gene for the genus Danaus rooted by outgroup Amauris niavius. Tree statistics: total characters in matrix 400 (two excluded), variable uninformative characters 23, parsimony informative characters 5 of 8, heuristic tree length 29, consistency index = 0.96, homoplasy index = 0.03, retention index = 0.88, rescaled consistency index = 0.84. Other details are as in Fig. 2. D. c. alcippus and D. gilippus are missing from this analysis as the PCR failed.

Results of mtDNA sequences

Twelve dorippus (dorippus-1) butterflies from Athi River, near Nairobi, Kenya, have distinctive mtDNA sequences that were not encountered elsewhere or by Lushai et al. (2003a). The new dorippus haplotype, designated NB, comprises a majority sequence (N = 6, Table 2) used for phylogenetic analysis (Fig. 2), and four rarer variants (N = 2, 2, 1, 1) that differ from it by 1–2 base pair changes at non-diagnostic sites; however, all 12 specimens differ invariably from all other D. chrysippus s.l. haplotypes in at least 46/884 (5.2%) of nucleotide sites (Table 2). In contrast, 13 dorippus butterflies, collected from seven sites that span the whole sampled range of dorippus, from Dar es Salaam (Tanzania) in the south to Salalah (Oman) in the north have a haplotype, designated DP (dorippus-2), that has two diagnostic sites in COI (C at 288 and 366), and one (T at 529) shared with NB. Five of these dorippus-2 butterflies, sequenced for a larger COI 676 bp fragment, displayed a total of five diagnostic sites, compared to all other haplotypes, and a further three shared with dorippus-1 (Smith, Lushai & Allen, 2005).

Table 2.

Variable sites, sites that are diagnostic for haplotypes (underlined) and genetic distances for the 12S rRNA (347 bp) and COI (537 bp) mitochondrial genes and the EF1-α (400 bp) nuclear gene among subspecies of Danaus chrysippus s.l.

Haplotype and OUMNH voucher number[…] N 12S001111111222223043556669024791752681299008092 COI000011111111122222333333333333344444444444455555504581222335573567801123355667780133456789991222773821106923167422189287091416281215810843345358913 
ST1a[261] 15 CT-ATTTATATCACA AACACTTCTATTTATTGTAATTACTCCTACTATTTATCGCAGTAAACTT 
ST1b[101] A.............. ................................................. 
ST1c[156] .......T....... ................................................. 
ST2a[257] 21 ............... ......................G.......................... 
ST2b[299] ..........A.... ......................G.......................... 
ST2c[277] ...........TT.. ......................G.......................... 
GH[163] 20 ........––..... ......................G.......................... 
NBa[57] .ATTCA-.––...TTCTTAACACGACGTCC..TTAA.AATT.TTCTCGCTCATT.CAGTTT.. 
NBb[52] .GTTCA-.––...TTCTTAACACGACGTCC..TTAA.AATT.TTCTCGCTCATT.CAGTTT.. 
NBc[65] .ATTCA-......TTCTTAACACGACGTCC..TTAA.AATT.TTCTCGCTCATT.CAGTTT.. 
NBd[53] .ATTCA-.––...TG TCTTAACACGACGTCC..TTAA.AATT.TTCTCGCTCATT.CAGTTT.. 
NBe[68] ..TTCA-.––...TTCTTAACACGACGTCC..TTAA.AATT.TTCTCGCTCATT.CAGTTT.. 
DPa[55] ............... .................C.........C..................T.. 
DPb[647] ............... .................C....G....C..........A.......T.. 
DPc[654] ............... .................C....G....C..................T.. 
DPd[651] ............... ................AC....G....C..........A.......T.. 
DPe[648] ............... .................C.........C..................TAA 
DPf[649] ............... .................C....G..T.C..................T.. 
DPg[661] ............... .................C....G....C.T................T.. 
Haplotype and OUMNH voucher number[…] N 12S001111111222223043556669024791752681299008092 COI000011111111122222333333333333344444444444455555504581222335573567801123355667780133456789991222773821106923167422189287091416281215810843345358913 
ST1a[261] 15 CT-ATTTATATCACA AACACTTCTATTTATTGTAATTACTCCTACTATTTATCGCAGTAAACTT 
ST1b[101] A.............. ................................................. 
ST1c[156] .......T....... ................................................. 
ST2a[257] 21 ............... ......................G.......................... 
ST2b[299] ..........A.... ......................G.......................... 
ST2c[277] ...........TT.. ......................G.......................... 
GH[163] 20 ........––..... ......................G.......................... 
NBa[57] .ATTCA-.––...TTCTTAACACGACGTCC..TTAA.AATT.TTCTCGCTCATT.CAGTTT.. 
NBb[52] .GTTCA-.––...TTCTTAACACGACGTCC..TTAA.AATT.TTCTCGCTCATT.CAGTTT.. 
NBc[65] .ATTCA-......TTCTTAACACGACGTCC..TTAA.AATT.TTCTCGCTCATT.CAGTTT.. 
NBd[53] .ATTCA-.––...TG TCTTAACACGACGTCC..TTAA.AATT.TTCTCGCTCATT.CAGTTT.. 
NBe[68] ..TTCA-.––...TTCTTAACACGACGTCC..TTAA.AATT.TTCTCGCTCATT.CAGTTT.. 
DPa[55] ............... .................C.........C..................T.. 
DPb[647] ............... .................C....G....C..........A.......T.. 
DPc[654] ............... .................C....G....C..................T.. 
DPd[651] ............... ................AC....G....C..........A.......T.. 
DPe[648] ............... .................C.........C..................TAA 
DPf[649] ............... .................C....G..T.C..................T.. 
DPg[661] ............... .................C....G....C.T................T.. 
Haplotype and OUMNH voucher number[…] N EF1-α011112233370027691340067558677 Genetic distances (uncorrected-p), mtDNA below diagonal, nDNA above diagonal voucher
 
  
ST1[262] CCCCTGGCCC ST1 – 1.5 – 2.3 2.3 
ST2[257] ..TTG.ATT. ST2 0.1 – – 1.3 1.3 
GH[163] xxxxxxxxxx GH 0.3 0.2 – – – 
NB[53] TTTTGAAT.T NB 5.7 5.8 5.5 – 0.0 
DP[55] TTTTGAAT.T DP 0.3 0.3 0.6 5.7 – 
Haplotype and OUMNH voucher number[…] N EF1-α011112233370027691340067558677 Genetic distances (uncorrected-p), mtDNA below diagonal, nDNA above diagonal voucher
 
  
ST1[262] CCCCTGGCCC ST1 – 1.5 – 2.3 2.3 
ST2[257] ..TTG.ATT. ST2 0.1 – – 1.3 1.3 
GH[163] xxxxxxxxxx GH 0.3 0.2 – – – 
NB[53] TTTTGAAT.T NB 5.7 5.8 5.5 – 0.0 
DP[55] TTTTGAAT.T DP 0.3 0.3 0.6 5.7 – 

Notation: x, no data. Equivalences between haplotype codes and D. chrysippus s.l. taxa are: ST1, chrysippus-o; ST2, chrysippus-b; GH, alcippus; NB, dorippus-1; DP, dorippus-2. EF1-α sites 175 and 316 are excluded from the phylogenetic analysis (Fig. 3) for reasons explained in the text.

Paradoxically, DP differs from NB by 50/884 (5.7%), the greatest genetic distance based on the mitochondrial genes we have recorded between any lineages in Danaus (Smith et al., 2005). The 13 DP butterflies have a majority sequence (N = 5, Table 2) and six minor variants (N = 2, 2, 1, 1, 1, 1) that differ at 1–3 non-diagnostic nucleotide sites. Six butterflies of dorippus phenotype, from five different sites, have haplotypes characteristic of chrysippus s.s. (ST1, ST2) or alcippus (GH) matrilines (Table 3). The only site where NB (N = 12) and DP (N = 3) dorippus butterflies were collected together was a field at Athi River, near Nairobi, where dorippus of GH haplotype (N = 4) also occurred: all dorippus (N = 19) were identical morphologically and for an EF1-α nuclear gene sequence (N = 3, see below).

Table 3.

Frequency distribution of haplotypes of Danaus chrysippus s.l. from Africa and Asia

Phenotype and colour genotype Provenance Haplotype
 
ST1 ST2 GH DP NB N 
dorippus Kenya*  0  0  4  7 12 23 
AAbC/bC Kampala, Uganda  1  0  0  2  0  3 
Dar es Salaam, Tanzania  0  1  0  3  0  4 
Salalah, Oman  0  0  0  1  0  1 
Total   1  1  4 13 12 31 
chrysippus-o Masai Mara, Kenya  3  0  0  0  0  3 
AAbc/bc Kampala, Uganda  1  0  0  0  0  1 
Oman*  9  0  4  1  0 14 
Petiala, India  4  0  0  0  0  4 
Total  17  0  4  1  0 22 
chrysippus-b Kampala, Uganda  0  2  2  0  0  4 
AABc/Bc Dar es Salaam, Tanzania  0 13  0  0  0 13 
Lusaka, Zambia  0  8  0  0  0  8 
Total   0 23  2  0  0 25 
alcippus Kampala, Uganda  0  1 10  0  0 11 
aabc/bc or aaBc/–c Cape Coast, Ghana  0  0  9  0  0  9 
Salalah, Oman  0  0  1  0  0  1 
Total   0  1 20  0  0 21 
N  18 25 30 14 12 99 
Phenotype and colour genotype Provenance Haplotype
 
ST1 ST2 GH DP NB N 
dorippus Kenya*  0  0  4  7 12 23 
AAbC/bC Kampala, Uganda  1  0  0  2  0  3 
Dar es Salaam, Tanzania  0  1  0  3  0  4 
Salalah, Oman  0  0  0  1  0  1 
Total   1  1  4 13 12 31 
chrysippus-o Masai Mara, Kenya  3  0  0  0  0  3 
AAbc/bc Kampala, Uganda  1  0  0  0  0  1 
Oman*  9  0  4  1  0 14 
Petiala, India  4  0  0  0  0  4 
Total  17  0  4  1  0 22 
chrysippus-b Kampala, Uganda  0  2  2  0  0  4 
AABc/Bc Dar es Salaam, Tanzania  0 13  0  0  0 13 
Lusaka, Zambia  0  8  0  0  0  8 
Total   0 23  2  0  0 25 
alcippus Kampala, Uganda  0  1 10  0  0 11 
aabc/bc or aaBc/–c Cape Coast, Ghana  0  0  9  0  0  9 
Salalah, Oman  0  0  1  0  0  1 
Total   0  1 20  0  0 21 
N  18 25 30 14 12 99 
*

All samples were collected at single sites as indicated except for dorippus from Kenya, six sites, and chrysippus-o from Oman, two sites (Table 1).

The chrysippus-o (ST1), chrysippus-b (ST2), alcippus (GH) and dorippus-2 (DP) haplotypes are similar. However, ST1 differs invariably from GH at three sites (0.3%) in large samples [TA/-- indels at 199–200 in 12S (Lushai et al., 2003a)] and an A/G transition at 330 in COI); from DP at three sites (0.3%) (T/C transitions at 288 and 366, and a C/T transition at 529); from ST2 at only one site (0.1%) (A/G at 330 in COI). ST2 differs from GH by 0.2% (TA/-- at 199–200 in 12S) and from DP by 0.3% (T/C at 288 and 366, and C/T at 529). Finally, GH and DP are separated by 0.6% (TA/-- at 199–200 in 12S, T/C at 288 and 366, and C/T at 529 in COI (Table 2).

Though DNA site differences among these forms are few, they are, individually or in combination, diagnostic of each colour form in samples (Table 3) taken from populations, both polymorphic and monomorphic, over a considerable geographical range χ29 = 163.7; P < 0.001; DP and NB haplotype numbers amalgamated to avoid low expected numbers). Furthermore, if polymorphic populations are analysed separately, the linkage between C locus and cytoplasmic genome is significantly maintained (Table 3), as for the A locus and mitochondrial 12S gene (Lushai et al., 2003a). chrysippus-o (ST1) has a majority (N = 15) haplotype (Table 2) and two minor variants at 1–2 non-diagnostic sites (N = 1, 1). Similarly, chrysippus-b (ST2) has one predominant haplotype (N = 21) (Table 2) and two rare ones (N = 1, 1). We found no variation in alcippus (GH), either in Ghana (N = 9), where D. chrysippus s.l. is monomorphic, in Uganda (N = 10), or in Oman (N = 1), where alcippus is sympatric with chrysippus-o and dorippus-2. In contrast to the small genetic distances that separate chrysippus-o, chrysippus-b, alcippus and dorippus-2, dorippus-1 (NB) is relatively distant from the other four, from which it differs by 5.7%, 5.8%, 5.5% and 5.7%, respectively.

Nucleotide sequences for the EF1-αnuclear gene

Eleven Danaus butterflies, with A. niavius as outgroup, were sequenced for the EF1-α nuclear gene. There are ten variable sites in the 400 bp fragment sequenced for four D. chrysippus s.l. clades (Table 2). Of four dorippus butterflies sequenced, two [53, 59 (Kenya)] of dorippus-1 haplotype and one dorippus-2[55 (Kenya)] are identical for the nuclear gene although they differ from chrysippus-o[102 (Kenya), 262 (Oman)] at nine sites (2.3%) and chrysippus-b[129 (Uganda), 257 (Zambia)] at five (1.3%). One dorippus-2 butterfly [73 (Kenya)] has essentially the chrysippus-b nuclear sequence but differs in one non-diagnostic A/G transition; chrysippus-o and chrysippus-b differ at six sites (1.5%). The interesting finding is that the two dorippus clades, which are highly divergent for mtDNA, are identical for the nDNA sequence, as well as for colour genes and morphology (see below). They appear to be one species. Moreover, it is noteworthy that genetic distances for nDNA (EF1-α) among chrysippus-o, chrysippus-b and dorippus-2 range from 1.3 to 2.3%, whereas for mtDNA, which is expected to be less conserved by an order of magnitude, distances are a mere 0.1–0.3%.

Three EF1-α genotypes are identified as ef1 (chrysippus-o), ef2 (chrysippus-b) and ef3 (dorippus-1+dorippus-2) (Table 4). As all sequences were easily read, we assume these butterflies were homozygous, though the species would benefit from a comprehensive allelic marker analysis, i.e. by specific microsatellites. Whereas chrysippus-o and chrysippus-b show no amino acid substitutions compared to one another, the (dorippus-1+dorippus-2) lineage differs from (chrysippus-o+chrysippus-b) by 1/133 amino acid substitutions (0.01%).

Table 4.

Congruence for haplotype and nuclear genotype in subspecies of Danaus chrysippus s.l. in Africa

Forms Genotype
 
Haplotype 
Colour genes EF1-α 
chrysippus-o AAbc/bc ef1 ST1 
chrysippus-b AABc/Bc ef2 ST2 
alcippus aa-c/–c – GH 
dorippus-1 AAbC/bC ef3 NB 
dorippus-2 AAbC/bC ef3 DP 
Forms Genotype
 
Haplotype 
Colour genes EF1-α 
chrysippus-o AAbc/bc ef1 ST1 
chrysippus-b AABc/Bc ef2 ST2 
alcippus aa-c/–c – GH 
dorippus-1 AAbC/bC ef3 NB 
dorippus-2 AAbC/bC ef3 DP 

Peptide sequences in theCOI gene

The COI amino acid sequence is invariable among the alcippus (GH), chrysippus-o (ST1), chrysippus-b (ST2) and dorippus-2 (DP) haplotypes, but is divergent in dorippus-1 (NB) by 4/179 amino acids (2.2%). Paired comparisons of dorippus-1/dorippus-2 and dorippus-1/chrysippus-o were tested for their Ka/Ks ratio, as set out in MEGA 2.1 (Kumar et al., 2001) and described by Hurst (2002), to investigate any abnormal selection for the divergent COI sequence data. Results indicate that synonymous differences per synonymous site (Ks = 0.24 ± 0.03 and 0.23 ± 0.03, respectively), exceed the number of non-synonymous differences per non-synonymous site (Ka = 0.03 ± 0.01 and 0.02 ± 0.01, respectively), with a Fisher's Exact Test of P = 1 for both comparisons, i.e. as expected for purifying selection, rather than positive selection (Zhang, Kumar & Nei, 1997).

Linkage disequilibrium for cytoplasmic and nuclear genomes

Although samples sequenced for mtDNA at polymorphic sites are small, if the [NB + DP = dorippus (CC)] and [(ST1 + ST2 + GH) = (chrysippus s.s.+alcippus) (cc)] haplotypes are pooled, the Kampala and Dar es Salaam populations show significant linkage disequilibrium for C locus genotype and haplotype, as do the results for the combined Kenya sites (Table 5). There is, moreover, association of (NB + DP) with C- and (ST1 + ST2 + GH) with cc in all samples; the highly significant combined probability indicates strong linkage disequilibrium (D′ (Lewontin, 1964) = 0.93) in polymorphic populations between the cytoplasmic and nuclear genomes of C- (dorippus) genotypes, on the one hand, and cc (chrysippus s.s.+alcippus) on the other. Linkage disequilibrium for unlinked genes that is consistent between populations is prima facie evidence, not only for sexual isolation, but also for positive natural selection (Lewontin, 1974).

Table 5.

Linkage disequilibrium for haplotype and C locus nuclear genotype in Danaus chrysippus s.l. from sites in the Afrotropical region that are polymorphic for the C locus

Haplotype C locus genotype NB + DP
 
ST1 + ST2 + GH
 
Exact P 
C- cc C- cc 
Kenya* 19  3 26 0.013 
Kampala, Uganda  2 16 19 0.018 
Dar es Salaam, Tanzania  3 13 17 0.006 
Salalah, Oman†  1  9 11 0.182 
Totals 25 41 73 2.56 × 10−7 
Haplotype C locus genotype NB + DP
 
ST1 + ST2 + GH
 
Exact P 
C- cc C- cc 
Kenya* 19  3 26 0.013 
Kampala, Uganda  2 16 19 0.018 
Dar es Salaam, Tanzania  3 13 17 0.006 
Salalah, Oman†  1  9 11 0.182 
Totals 25 41 73 2.56 × 10−7 

Note: The C- genotype has the dorippus phenotype, cc is chrysippus-o, chrysippus-b or alcippus (Table 3). Visually identifiable Cc heterozygotes (transiens) are omitted from the table but a few C- phenotypes are expected to be Cc for genotype.

*

The Kenya sample consists of individuals pooled from six different polymorphic sites (Table 1). If the pooled Kenyan sites are excluded P = 1.97 × 10−5.

The monomorphic sample (N = 5) from Nazwa, north Oman, is omitted.

Phylogenetic analysis

The MP tree for the 12S (347 bp) and COI (537 bp) sequences (Fig. 2) shows that all but one clade of D. chrysippus s.l. (forms chrysippus-o, chrysippus-b, alcippus and dorippus-2) comprise a monophyletic group that is sister to the New World clade (D. eresimus+D. gilippus), with [D. chrysippus s.l.+ (D. eresimus+D. gilippus)] sister to the Indo-Australian D. genutia (Fig. 2). With the exception of three subspecies of D. chrysippus s.l. (chrysippus-o, chrysippus-b and alcippus) all clades are well supported by bootstrap (87+) and Bremer support and, moreover, backed by large sample sizes (Table 1). The enigmatic feature of the tree, the topology of which is fully supported by ML analyses, is the exclusion from the [D. chrysippus s.l.+ (D. eresimus+D. gilippus)] clade of D. chrysippus s.l. form dorippus-1, which is instead sister to the American species D. plexippus. The relationship of dorippus-1 to the other D. chrysippus s.l. clades (including dorippus-2) is polyphyletic, as they have different most recent ancestors. It was chiefly with the possibility of trying to resolve the dorippus paradox that we sequenced a small number of samples for the nuclear EF1-α gene (Tables 1, 2).

In the EF1-α gene, two C−T transitions at tertiary positions (DNA sites, 175 and 316, respectively), have taken place in the basal taxa (D. plexippus, D. c. dorippus-1) of Danaus in comparison to the outgroup A. niavius. Reversions at these same sites in more terminal taxa (D. eresimus, D. genutia, other D. chrysippus s.l. clades) have reduced the resolution of the parsimony tree in comparison to Distance (Neighbour-joining) and ML algorithms. However, exclusion of these two nucleotides (which explains apparent sequence differences between Table 2 and Fig. 3) significantly improves the Parsimony tree score: tree number falls from 38 to 2; consistency index rises from 0.69 to 0.96, etc.

Although bootstrap and Bremer support for the four major Danaus clades, plexippus (dorippus-1+dorippus-2+eresimus), genutia and (chrysippus-o+chrysippus-b) is not strong, the EF1-α topology (Fig. 3) supports the mtDNA phylogeny (Fig. 2) in showing dorippus-1 and (chrysippus-o+chrysippus-b) to be polyphyletic and distantly related. There are, however, several differences between the nuclear and mitochondrial phylogenies:

  1. The EF1-α sequences of dorippus-1 and dorippus-2 are identical, whereas their haplotypes are widely divergent.

  2. (dorippus-1+dorippus-2 = dorippus) is also identical with D. eresimus for EF1-α;

  3. The EF1-α gene tree (Fig. 3) suggests that (chrysippus-o+chrysippus-b=chrysippus s.s.) and (dorippus-1+dorippus-2=dorippus) are separate species that are reciprocally monophyletic, whereas the haplotype tree (Fig. 2) shows dorippus-2 to be sister to the chrysippus s.s. haplotypes and distant from dorippus-1. chrysippus s.s. has two largely vicariant, monophyletic subspecies, chrysippus-o (bc/bcef1ef1), north of the hybrid zone and chrysippus-b (Bc/Bcef2ef2) ranging from southern Africa to the southern part of the hybrid zone (Fig. 1), whereas dorippus (bC/bCef3ef3) is a single nuclear lineage with two highly diverged, polyphyletic haplotypes, one of which (dorippus-1) was found only in the polymorphic population at Athi River, whereas the other (dorippus-2) occurred throughout the sampled range.

Comparative morphology ofdorippusandchrysippus

Examination of new specimens of dorippus (N = 23) and chrysippus s.s. (N = 20), including both orange and brown forms of the latter, shows that they share the five pattern characters diagnostic of all D. chrysippus s.l. (Ackery & Vane-Wright, 1984): they are also inseparable for five structural characters; male genitalia, including the disposition and number of bristles on the aedaeagus (Smith et al., 2002), short spines on the 5th tarsal segment of mid and hind legs, the 2-5-11 segmental formula for paired processes in larvae, the sharply angled hindwing cross-vein m1-m2 and the 1 : 2 ratio of hindwing discal cell length: hindwing length. The morphological identity of dorippus and chrysippus s.s. by no means proves their conspecifity since D. chrysippus s.l. and D. gilippus are also structurally indistinguishable (Ackery & Vane-Wright, 1984), whereas their F1 hybrids are sterile (Smith et al., 2002).

DISCUSSION

DNA sequence divergence amongdorippus, chrysippus-oandchrysippus-b

dorippus-1, dorippus-2, chrysippus-o and chrysippus-b lack morphological apomorphies with respect to one another. The mtDNA sequences of the two latter have diverged from dorippus-1 by 5.7% and 5.8%, respectively, but from dorippus-2 by a mere 0.3%. Most dorippus samples contain only the DP (dorippus-2) haplotype, which has a wide geographical distribution and sits conformably, within a monophyletic domain, as sister to the other D. chrysippus s.l. clades. On the other hand, the NB (dorippus-1) haplotype, found only at Athi River, is sister to D. plexippus and closer to the Indo-Australian D. genutia than to any other D. chrysippus s.l. clades (Fig. 2). Note that the two clades (chrysippus-o+chrysippus-b+alcippus+dorippus-2) and dorippus-1 have similar polyphyletic relationships for both 12S and COI genes when analysed independently (Lushai et al., 2003c).

The weakly supported EF1-α gene tree (Fig. 3) suggests a sister relationship between dorippus (both haplotypes) and D. plexippus. Compared to both forms of dorippus, chrysippus-o is separated by 2.3% and chrysippus-b by 1.3% nucleotide base pair differences (Table 2). Hence, the clades (chrysippus-o+chrysippus-b) and (dorippus-1+dorippus-2) are reciprocally monophyletic (Fig. 3). In summary, the two dorippus clades have relationships with the remaining D. chrysippus s.l. clades that are discordant for mitochondrial and nuclear sequences: in the case of dorippus-1, its relationship to all the other clades is polyphyletic for cytoplasmic DNA and reciprocally monophyletic for nuclear DNA. In contrast, for cytoplasmic DNA, dorippus-2 is sister to all D. chrysippus s.l. clades other than dorippus-1 and, for nuclear DNA, is reciprocally monophyletic to these same clades. This extraordinary divergence within D. chrysippus s.l., and within dorippus in particular, suggests the species group has a complex history.

We believe that the hypothesis to explain extreme intraspecific divergence of mtDNA in the hermaphrodite snail Cepaea nemoralis (Thomaz, Guiller & Clarke, 1996), namely that it results from subdivision of populations into nearly isolated demes, must be inappropriate for the highly vagile, migratory and r-selected D. chrysippus s.l. (Smith & Owen, 1997). The pattern of variation in D. chrysippus s.l. corresponds to phylogeographical category II of Avise (2000), in which it is supposed that extremely divergent haplotypes within a species originate either (1) from hitherto unidentified sympatric sibling species, or (2) from previously isolated lineages. As widespread reticulation (multidirectional hybridism) occurs among all D. chrysippus s.l. clades today, the latter alternative more likely explains extreme haplotype divergence within dorippus. One hypothesis is that hybridization occurred in Africa between an ancestral D. chrysippus s.l. and a more basal species, now presumably extinct, that was closely related to D. plexippus. An alternative hypothesis, which we favour, is that dorippus-1 may itself be the descendent by anagenesis (evolution without branching) of this ancient species that has more recently hybridized with other D. chrysippus lineages. The relatively basal position of dorippus-1 in both mitochondrial and nuclear gene trees (Figs 2, 3), and of dorippus-2 only in the latter (Fig. 3), supports the second hypothesis.

The sister relationship between the basal clades dorippus-1 (Africa) and D. plexippus (America), which is independently supported by the history of three genes, one nuclear and two mitochondrial, suggests that the ancestral Danaus, or its immediate descendants, colonized both Old and New Worlds early in the history of the genus. If a molecular clock conversion factor for mtDNA evolution at the COI locus in prawns of 2.4% per Myr (Knowlton et al., 1993) is applied to the 7.8% COI sequence divergence of dorippus-1 from dorippus-2, the result suggests a coalescence time of ∼3.3 million years BP in the Pliocene. A similar calculation, based on substitutions at the 12S locus of 0.6% per Myr in the mollusc Littorina (Rumbak et al., 1994), may be applied to the 2.3% 12S divergence for dorippus-1 and dorippus-2, giving a similar coalescence time of ∼3.8 Myr. Lushai et al. (2003c) show that COI and 12S molecular evolution in Danaus is clock-like; the two clocks are significantly correlated, the former evolving approximately four times faster than the latter.

Evidence for assortative mating and linkage disequilibrium

Records of mated pairs (N = 718) and unmated individuals among morphs of D. chrysippus s.l. in the field at Dar es Salaam were kept from 1972 to 1975 without interruption (Smith, 1973b, 1975c, 1984). With respect to mate choice, the main conclusions were all statistically significant, as follows:

  1. When seasonal fluctuations of sex and morph ratios were controlled, mate selection at both B and C loci was independently assortative and the effects on each locus were additive.

  2. Males mated at random but females exercised a preference for homotypic males.

  3. Paradoxically, tested against a null hypothesis that sex ratios are 1 : 1 in both dorippus and (chrysippus-o + chrysippus-b), pairings in the wild assessed from their progenies (N = 61) were asymmetrically heterotypic, with the dorippus (C-)♂ × chrysippus (cc) ♀ pairing above expectation and its reciprocal below χ23 = 9.4: P < 0.05). These old data suggest that introgression of chrysippus s.s. haplotypes into the dorippus phenotype is expected to occur at high frequency in the East African hybrid zone, whereas chrysippus phenotypes with dorippus cytoplasm should be rare. Although our DNA data (Table 4) support this expectation, sample sizes are too small to be conclusive. However, the mating data suggest a mechanism by which a chrysippus s.s. haplotype could be captured by a lineage of dorippus morphology in a possibly recent event, thus explaining the origin of dorippus-2 as a form of hybrid origin.

Progenies from wild mated pairs (Smith et al., 1998) show that chrysippus s.s. (cc)♂ × dorippus (CC)♀ pairs are rare at both Athi River (N = 0/18) and Dar es Salaam (N = 1/69). Although mate selection in the field has not been studied at the former location, the predominance of heterotypic pairings is suggested by the progenies of wild mated pairs, of which only 3/18 were homogametic. At Athi River, the year-round excess of Cc (transiens) and cc (chrysippus s.s.) females (Smith et al., 1997), and the relative scarcity of their respective males, suggests that CC (dorippus)♂ × cc (chrysippus)♀ and CC (dorippus)♂ × Cc (transiens)♀ pairings prevail at all times.

The significance of these old observations is enhanced by the EF1-α sequence data, which suggest that dorippus (both haplotypes) and chrysippus s.s. are reciprocally monophyletic and, in the context of a small genus that has radiated only within the last 5 million or so years (Lushai et al., 2003c), relatively distant clades that differ concordantly for nuclear EF1-α and colour genes, despite sharing structural morphology. Furthermore, the population at Dar es Salaam, which is polymorphic for the dorippus (bC/bC) and chrysippus-b (Bc/Bc) linked colour genes (Smith, 1980), shows strong linkage disequilibrium (D′ = −0.78) at the tightly linked B and C loci for the wild-type linkage arrangements (Smith, 1980). Here, we present further evidence for strong linkage disequilibrium involving the nuclear C alleles and unlinked cytoplasmic genes, between dorippus (C-) and (chrysippus s.s.+alcippus) (cc). Therefore, prezygotic isolation between dorippus and chrysippus s.s. is no surprise. Indeed, the question now to be answered is why, despite behavioural isolation, dorippus and chrysippus s.s. hybridize at such high frequency at Athi River, Dar es Salaam, Kampala (Owen & Chanter, 1968; Smith et al., 1993) and elsewhere (Smith et al., 1997).

Heterozygote excess

The data on mate choice at Dar es Salaam appear paradoxical. On the one hand, recalculating from the data of Smith (1984), taking account of the overall difference between sex ratios in dorippus and chrysippus s.s., mate selection, measured by the Phi Coefficient (r), was significantly assortative: r = 0.14, P < 0.001; χ23 = 83.0, P < 0.001, N = 718. In dry seasons when chrysippus s.s. was rare (< 20%), r = 0.25, P < 0.001; χ23 = 211.7, P < 0.001, N = 341. On the other hand, if observed frequencies are compared with an expectation that sex ratios in both forms were 1 : 1, the dorippus♂ × chrysippus s.s.♀ pairing significantly exceeded this and the reciprocal was below expectation (Smith, 1973b, 1984). However, recalculation shows that, on the same hypothesis, asymmetric heterotypic mating was confined to wet seasons χ23 = 9.0, 0.05 > P > 0.01, N = 377), when the frequency of chrysippus s.s., and especially female chrysippus, was relatively high (> 40%).

The apparent conflict of evidence is resolved if the extraordinary sex ratios and associated sex differences in genotype frequencies (Smith et al., 1997) are considered. The likely explanation for asymmetric heterotypic mating is that chrysippus s.s. females are obliged to mate unlike males becausechrysippus males are scarce, sometimes exceedingly so, either throughout the year at Athi River (Table 6), or mainly in wet seasons at Dar es Salaam (Smith, 1975b). On the other hand, female dorippus are able to find homotypic males without difficulty and generally choose them as mates in preference to chrysippus males. In practice, therefore, whereas female preference is homotypic, realized pairings may be predominantly heterotypic. Thus, heterozygote excess is the product of the dorippus♂ × chrysippus♀ heterotypic pairing (Smith et al., 1997), and is a maladaptive consequence of denying preference to chrysippus s.s. females. As the excess of heterozygotes is itself strongly female biased at Dar es Salaam, the Nairobi region (Smith et al., 1997) and Kampala (Smith et al., 1993), it must result from infection by Spiroplasma, a maternally inherited, male-killing, bacterial symbiont (Jiggins et al., 2000; I. J. Gordon, pers. comm.), transmitted predominantly by chrysippus s.s. mothers. We must even entertain the idea that the hybrid zone, involving as it does two distantly related and reciprocally monophyletic (nuclear genome) or polyphyletic (cytoplasmic genome) taxa that are behaviourally isolated by mate choice (Smith, 1984), allochronic migration (Smith & Owen, 1997) and substantial vicariance (Fig. 1), is a pathological outcome of Spiroplasma infection (Lushai et al., 2003a).

Table 6.

Comparisons of sex ratios among C locus genotypes of Danaus chrysippus s.l. at Dar es Salaam (1974) and Athi River (1986−95) (from Smith et al., 1997); genotype frequencies are corrected for penetrance of the c allele in the Cc genotype (Smith, 1998)

Genotype Phenotype Dar es Salaam
 
Athi River
 
♂ ♀ ♂/♀ ♂ ♀ ♂/♀ 
CC dorippus  390 152 2.57  98  59 1.66 
Cc transiens  722 533 1.36 167 432 0.39 
cc chrysippus s.s.  271 238 1.14  15  86 0.17 
Totals  1383 923 1.50 280 577 0.49 
χ22 (P   45.2 (< 0.001)    84.0 (< 0.001)   
Genotype Phenotype Dar es Salaam
 
Athi River
 
♂ ♀ ♂/♀ ♂ ♀ ♂/♀ 
CC dorippus  390 152 2.57  98  59 1.66 
Cc transiens  722 533 1.36 167 432 0.39 
cc chrysippus s.s.  271 238 1.14  15  86 0.17 
Totals  1383 923 1.50 280 577 0.49 
χ22 (P   45.2 (< 0.001)    84.0 (< 0.001)   

Notes: (1) The Dar es Salaam statistics are for butterflies netted and marked in the field throughout the year, while the Athi River data are based on eggs collected intermittently in the wild and reared to adult in the laboratory. Males are almost invariably over-represented in field-caught samples, whereas collections of eggs are not expected to be biased for sex ratio. (2) The chrysippus s.s. forms are chrysippus-b at Dar es Salaam and chrysippus-o at Athi River.

The possible involvement of a male-killer in the bizarre population dynamics of D. chrysippus s.l. is supported by the marked contrast, for both sex ratio at hybrid zone sites (Table 6) and mtDNA diversity, between dorippus on the one hand and chrysippus-o and chrysippus-b on the other. Haplotype diversity at hybrid zone sites, measured by the information index H′, is 2.224 for dorippus (N = 25 with 12 variants), 0.329 for chrysippus-o (N = 17 with three variants), and 0.223 for chrysippus-b (N = 23 with three variants). Significance tests for differences (Magurran, 1988) give, for dorippus vs. chrysippus-o, t26 = 6.4, P < 0.001, and for dorippus vs. chrysippus-b, t28 = 7.7, P < 0.001: both results are significant. The contrasted sex ratios for dorippus, chrysippus-o (Kenya), chrysippus-b (Tanzania) and transiens are shown for well-studied field populations (Smith et al., 1997) at Athi River and Dar es Salaam in Table 6. At both locations, over many years of observation, and also at Kampala (Owen & Chanter, 1968; Smith et al., 1993), permanently and heavily female-biased sex ratios for both chrysippus s.s. (cc) and transiens (Cc) stand in marked contrast to undisturbed sex ratios in dorippus (CC). The female bias of transiens sex ratios supports other evidence, from mate selection in the field (Smith, 1984) and segregations for colour genes in laboratory-reared broods obtained from wild pairings (Smith et al., 1998), to suggest transiens inherits its male-killer cytoplasm, with its associated haplotypes, predominantly from cc (chrysippus s.s.+alcippus) matrilines. Introgression at the present time of ST1, ST2 and GH haplotypes, from chrysippus-o, chrysippus-b and alcippus, respectively, into dorippus in the hybrid zone is expected to be extensive. After successive generations of capture by dorippus males of ST/GH female haplotypes, the expected outcome is fixation of an alien haplotype in a new nuclear environment. This suggests a mechanism whereby the NB (dorippus-1) and/or DP (dorippus-2) haplotypes − the latter in particular − could, in the past, have been captured and assimilated by a proto-dorippus species with a nuclear genome ancestral to that of the present dorippus.

Inherited symbionts such as Spiroplasma lower the effective population size (Ne) to that of the infected population, with a concomitant decrease in mtDNA diversity. The mtDNA type associated with the initial infection increases in frequency by hitch-hiking with the symbiont (Hurst, Hurst & Majerus, 1997). Because vertical transmission of the symbiont is imperfect, its associated mtDNA quickly invades uninfected matrilines. Thus, the low haplotype diversity found in both chrysippus-o and chrysippus-b within the hybrid zone could result from selective sweeps on mtDNA by male-killers; this would also explain their female-biased sex ratios that are not found in any population of dorippus known to us (Smith, 1975b; Gordon, 1984; Smith et al., 1993). An important implication of such selective sweeps is that it is the evolution of the symbiont, rather than that of the butterfly, that will impact the history of mtDNA. Therefore, closely similar haplotypes among, for example, dorippus-2, chrysippus-o, chrysippus-b and alcippus (Table 2), and between D. eresimus and D. gilippus (Fig. 2), may result from recent selective sweeps on mtDNA which obfuscate their true phylogenetic affinities and understate genetic distances. Similarly, more ancient selective sweeps could account for the successive capture by an evolving dorippus lineage, from other lineages with which they hybridized, first, of the dorippus-1 haplotype and, more recently, dorippus-2. At the present time extensive 12S data from Kenya, Uganda and Tanxania (Lushai & Smith, unpublished) suggest that introgression of chrysippus-o, chrysippus-b and alcippus cytoplasm into dorippus occurs at high frequency wherever these lineages meet and hybridize, whereas the reverse process is undetected.

CONCLUSIONS

If the foregoing analysis is correct, the pervasive but paradoxical hybridism among sexually isolated, yet fully interfertile, lineages of D. chrysippus s.l. in East Africa is, for the most part, parasite driven and probably largely dysfunctional (Lushai et al., 2003a; but see Smith, 1981). If parasites that influence the sex ratio, by causing male death and, thus, female-dominated populations (Table 6), can enforce hybridism between otherwise behaviourally isolated populations, this may explain how a haplotype such as that described for dorippus-1, that may be considerably more ancient than the species in which it now resides, could have been captured from another species, presumed to be either extinct or subsumed. If, on the other hand, the present dorippus lineage is descended by anagenesis from a proto-dorippus species that was a close relative of proto-plexippus (Figs 2, 3), hybridization at various times with other chrysippus s.s. lineages can explain the haplotype polymorphisms we have observed in the dorippus phenotype. This argument is unaffected by the taxonomic issue, which we do not address here, i.e. whether the species, of which individuals carrying the dorippus-1, dorippus-2 and other haplotypes are members, should carry the label ‘D. dorippus’ or ‘D. chrysippus s.l.’. Furthermore, whatever the answer to this thorny taxonomic question, by highlighting the probable significance of a male-killer parasite to the evolution of D. chrysippus s.l. in Africa, we are proposing a novel mechanism whereby a polyphyletic species might originate through hybridism (Grant & Grant, 1989; Harrison, 1997; Arnold, 1997). Since monophyly is a sine qua non for a phylogenetic species (e.g. Rosen, 1978; Nelson & Platnick, 1981; Cracraft, 1989), whereas genetic isolation between populations of congeneric taxa in sympatry is the crucial qualifier for biological species (e.g. Dobzhansky, 1937; Mayr, 1942), by failing to satisfy key criteria for both species concepts, D. chrysippus s.l. and its component African subspecies raise difficult questions in regard to the origin, nature and boundaries of species.

ACKNOWLEDGEMENTS

We are grateful for financial assistance from Leverhulme Trust Grant F/180/AP and the Linnean Society of London. For help with collecting butterflies, we thank Andy Brower, Ian Gordon, Francis Jiggins and Myron Zalucki. We also thank Christopher and Elizabeth Wood (Nairobi) and Henry and Eliza Harford (Grand Cayman) for providing hospitality to DASS on location. The late Derek Whiteley painted the butterflies in Figure 1.

REFERENCES

Ackery
PR
,
Vane-Wright
RI
1984.
Milkweed butterflies
 .
London
:
British Museum (Natural History).
Arnold
ML
1997.
Natural hybridization and evolution
 .
Oxford
:
Oxford University Press
.
Avise
JC
2000.
Phylogeography: the history and formation of species
 .
CambridgeMA
:
Harvard University. Press
.
Brower
LP
,
Edmunds
M
,
Moffitt
CM
1975
.
Cardenolide content and palatability of a population of Danaus chrysippus butterflies from West Africa
.
Journal of Entomology A
 
49
:
183
196
.
Brower
LP
,
Gibson
DO
,
Moffitt
CM
,
Panchen
AL
1978
.
Cardenolide content of Danaus chrysippus butterflies from three areas of East Africa
.
Biological Journal of the Linnean Society
 
10
:
251
273
.
Clarke
CA
,
Sheppard
PM
,
Smith
AG
1973
.
The genetics of fore and hindwing colour in crosses between Danaus chrysippus from Australia and from Sierra Leone (Danaidae)
.
Journal of the Lepidopterists’ Society
 
27
:
73
77
.
Cracraft
J
1989
.
Speciation and its ontogeny: the empirical consequences of alternative species concepts for understanding patterns and processes of differentiation
. In:
Otte
D
Endler
JA
eds.
Speciation and its consequences
 .
SunderlandMA
:
Sinauer
,
28
59
.
Dobzhansky
T
1937.
Genetics and the origin of species
 .
NewYork
:
Columbia University Press
.
Eriksson
T
1999.
AutoDecay
 , Version 4.0. Computer program distributed by the author.
Stockholm
:
Bergius Foundation
,
Royal Swedish Academy of Sciences.
Felsenstein
J
1993.
PHYLIP (phylogeny inference package)
 , Version 3.5. Distributed by the author.
Seattle
:
Department of Genetics, University of Washington.
Gordon
IJ
1984
.
Polymorphism of the tropical butterfly Danaus chrysippus L. in Africa
.
Heredity
 
53
:
583
593
.
Grant
PR
,
Grant
BR
1989
.
Sympatric speciation and Darwin's finches
. In:
Otte
D
Endler
JA
eds.
Speciation and its consequences
 .
SunderlandMA
:
Sinauer
,
433
457
.
Harrison
RG
1997.
Hybrid zones and the evolutionary process
 .
Oxford
:
Oxford University Press.
Hurst
GDD
,
Hurst
LD
,
Majerus
MEN
1997
.
Cytoplasmic sex ratio distorters
. In:
O'Neill
SL
Hoffman
AA
Werren
JH
eds.
Influential passengers.
 
Oxford
:
Oxford University Press
,
125
164
.
Hurst
LD
2002
.
The Ka/Ks ratio: diagnosing the form of sequence evolution
.
Trends in Genetics
 
18
:
486
487
.
Jiggins
FM
,
Hurst
GDD
,
Jiggins
CD
,
Schulenburg JHGVD, Majerus
MEN
2000
.
The butterfly Danaus chrysippus is infected by a male-killing Spiroplasma bacterium
.
Parasitology
 
120
:
439
446
.
Knowlton
N
,
Weight
LA
,
Solorzano
LA
,
Mills
DK
,
Bermingham
E
1993
.
Divergence in proteins, mitochondrial DNA and reproductive compatibility across the Isthmus of Panama
.
Science
 
260
:
1629
1632
.
Kumar
S
,
Tamura
K
,
Jakobsen
IB
,
Nei
M
2001.
MEGA2: molecular evolutionary genetics analysis software
 .
Tempe, AZ
:
Arizona State University
.
Lewontin
RC
1964
.
The interaction of selection and linkage. I. General considerations; heterotic models
.
Genetics
 
49
:
49
67
.
Lewontin
RC
1974.
The genetic basis for evolutionary change
 .
NewYork
:
Columbia University Press
.
Lushai
G
,
Gordon
IJ
,
Smith
DAS
2003b
.
Evidence from mitochondrial DNA supports earlier records of. African queen butterflies (Danaus chrysippus) migrating in East Africa
.
Journal of East African Natural History
 
92
:
119
125
.
Lushai
G
,
Smith
DAS
,
Gordon
IJ
,
Goulson
D
,
Allen
JA
,
Maclean
N
2003a
.
Incomplete sexual isolation in sympatry between subspecies of the butterfly Danaus chrysippus (L.) and the creation of a hybrid zone
.
Heredity
 
90
:
236
246
.
Lushai
G
,
Smith
DAS
,
Goulson
D
,
Allen
JA
,
Maclean
N
2003c
.
Mitochondrial DNA clocks and the phylogeny of Danaus butterflies
.
Insect Science and its Applications
 
23
:
309
315
.
Magurran
AE
1988.
Ecological diversity and its measurement
 .
London
:
Croom-Helm
.
Mayr
E
1942.
Systematics and the origin of species
 .
NewYork
:
Columbia University Press
.
Nelson
GJ
,
Platnick
NI
1981.
Systematics and biogeography: cladistics and vicariance
 .
NewYork
:
Columbia University Press
.
Owen
DF
,
Chanter
DO
1968
.
Population biology of tropical African butterflies. 2. Sex ratio and polymorphism in Danaus chrysippus L
.
Revues Zoologiques et Botaniques Africaines
 
78
:
81
97
.
Owen
DF
,
Smith
DAS
1993.
Danaus chrysippus and its polymorphic Müllerian mimics in tropical Africa. (Lepidoptera: Nymphalidae: Danainae)
.
Tropical Lepidoptera
 
4
:
1
5
.
Owen
DF
,
Smith
DAS
,
Gordon
IJ
,
Owiny
AM
1994
.
Polymorphic Müllerian mimicry in a group of African butterflies: a re-assessment of the relationship between Danaus chrysippus, Acraea encedon and Acraea encedana (Lepidoptera: Nymphalidae)
.
Journal of Zoology, London
 
232
:
93
108
.
Rosen
DE
1978
.
Vicariant patterns and historical explanation in biogeography
.
Systematic Zoology
 
27
:
159
188
.
Rothschild
M
,
Von Euw
J
,
Reichstein
J
,
Smith
DAS
,
Pierre
J
1975
.
Cardenolide storage in Danaus chrysippus with additional notes on D. plexippus
.
Proceedings of the Royal Society of London B
 
190
:
1
31
.
Rumbak
E
,
Reid
DG
,
Thomas
RH
1994.
Reconstruction of the phylogeny of 11 species of Littorina (Gastropoda: Littorinidae) using mitochondrial sequence data
.
Nautilus
 
2
:
91
97
.
Sheppard
PM
1967.
Natural selection and heredity
 ,
3rd edn
.
London
:
Hutchinson
.
Smith
DAS
1973a
.
Batesian mimicry between Danaus chrysippus and Hypolimnas misippus, Tanzania
.
Nature
 
242
:
129
131
.
Smith
DAS
1973b
.
Negative non-random mating in the polymorphic butterfly Danaus chrysippus in Tanzania
.
Nature
 
242
:
131
132
.
Smith
DAS
1975a.
Genetics of some polymorphic forms of the African butterfly Danaus chrysippus L. (Lepidoptera: Danaidae)
.
Entomologica scandinavica
 
6
:
134
144
.
Smith
DAS
1975b
.
All-female broods in Danaus chrysippus L. and their ecological significance
.
Heredity
 
34
:
363
371
.
Smith
DAS
1975c
.
Sexual selection in a wild population of the butterfly Danaus chrysippus L
.
Science
 
187
:
664
665
.
Smith
DAS
1976
.
Phenotypic diversity, mimicry and natural selection in the African butterfly Hypolimnas misippus
.
Biological Journal of the Linnean Society
 
8
:
183
204
.
Smith
DAS
1979
.
The significance of beak marks on the wings of an aposematic, distasteful and polymorphic butterfly
.
Nature
 
281
:
215
216
.
Smith
DAS
1980
.
Heterosis, epistasis and linkage disequilibrium in a wild population of the polymorphic butterfly Danaus chrysippus
.
Zoological Journal of the Linnean Society
 
69
:
87
109
.
Smith
DAS
1981
.
Heterozygous advantage expressed through sexual selection in a polymorphic African butterfly
.
Nature
 
289
:
174
175
.
Smith
DAS
1984
.
Mate selection in butterflies: competition, coyness, choice and chauvinism
. In:
Vane-Wright
RI
Ackery
PR
eds.
The biology of butterflies.
  Symposia of the Royal Entomological Society of London 11.
London
:
Academic Press
,
225
244
.
Smith
DAS
1998
.
Non-Mendelian segregation and variable penetrance of colour genes in the polymorphic butterfly Danaus chrysippus (L.)
.
Heredity
 
80
:
474
480
.
Smith
DAS
,
Gordon
IJ
,
Depew
LA
,
Owen
DF
1998
.
Genetics of the butterfly Danaus chrysippus (L.) in a broad hybrid zone, with special reference to sex ratio, polymorphism and intragenomic conflict
.
Biological Journal of the Linnean Society
 
65
:
1
40
.
Smith
DAS
,
Gordon
IJ
,
Lushai
G
,
Goulson
D
,
Allen
JA
,
Maclean
N
2002
.
Hybrid queen butterflies from the cross Danaus chrysippus (L.) ×D. gilippus (Cramer): confirmation of species status for the parents and further support for Haldane's Rule
.
Biological Journal of the Linnean Society
 
76
:
535
544
.
Smith
DAS
,
Lushai
G
,
Allen
JA
2005
.
A classification of Danaus butterflies (Lepidoptera: Nymphalidae) based upon data from morphology and DNA
.
Zoological Journal of the Linnean Society
 
144
:
191
212
.
Smith
DAS
,
Owen
DF
1997
.
Colour genes as markers for migratory activity: the butterfly Danaus chrysippus in Africa
.
Oikos
 
78
:
127
135
.
Smith
DAS
,
Owen
DF
,
Gordon
IJ
,
Lowis
NK
1997
.
The butterfly Danaus chrysippus (L.) in East Africa: polymorphism and morph-ratio clines within a complex, extensive and dynamic hybrid zone
.
Zoological Journal of the Linnean Society
 
120
:
51
78
.
Smith
DAS
,
Owen
DF
,
Gordon
IJ
,
Owiny
AM
1993
.
Polymorphism and evolution in the butterfly Danaus chrysippus (L.) (Lepidoptera: Danainae)
.
Heredity
 
71
:
242
251
.
Swofford
DL
1998.
PAUP*: phylogenetic analysis using parsimony (*and other methods)
 , Version 4.
SunderlandMA
:
Sinauer
.
Talbot
G
1943
.
Revisional notes on the genus Danaus Kluk (Lep. Rhop. Danaidae)
.
Transactions of the Royal Entomological Society
 
93
:
115
148
.
Thomaz
D
,
Guiller
A
,
Clarke
B
1996
.
Extreme divergence of mitochondrial DNA within species of pulmonate land snails
.
Proceedings of the Royal Society of London B
 
263
:
363
368
.
Thompson
JD
,
Higgins
DG
,
Gibson
TJ
1994
.
CLUSTAL X: improving the sensitivity of progressive multiple sequence alignment through sequence weighting, position-specific gap penalties and weight matrix choice
.
Nucleic Acids Research
 
22
:
4673
4690
.
Wahlberg
N
,
Oliveira
R
,
Scott
JA
2003a.
Phylogenetic relationships of Phycoides butterfly species. (Lepidoptera: Nymphalidae): complex mtDNA variation and species delimitations
.
Systematic Entomology
 
28
:
257
274
.
Wahlberg
N
,
Weingartner
E
,
Nylin
S
2003b
.
Towards a better understanding of the higher systematics of Nymphalidae (Lepidoptera: Papilionoidea)
.
Molecular Phylogenetics and Evolution
 
28
:
473
484
.
Zhang
J
,
Kumar
S
,
Nei
M
1997
.
Small-sample tests of episodic adaptive evolution: a case study of primate lysozymes
.
Molecular Biology and Evolution
 
14
:
1335
1338
.