Danaus gilippus, ISSUE

Smith, David A. S., Lushai, Gugs & Allen, John A., 2005, A classification of Danaus butterflies (Lepidoptera: Nymphalidae) based upon data from morphology and DNA, Zoological Journal of the Linnean Society 144 (2), pp. 191-212 : 199

THE GILIPPUS ISSUE

This problem is intertwined with the preceding one. A & V-W found no structural apomorphies to distinguish gilippus from chrysippus s.l. and suggested they might constitute a single cosmopolitan species. Smith et al. (2002) tested this hypothesis by crossing female gilippus berenice from Grand Cayman with male chrysippus dorippus (of unknown haplotype) from Kenya. They obtained an F 1 that comprised sterile males (when backcrossed to dorippus females), and nonviable females, thus demonstrating that the parents belonged to separate species. [A substantial caveat is that Smith et al. (2002) had not then recognized dorippus as a species distinct from chrysippus ; it cannot therefore be assumed that gilippus ¥ chrysippus s.s. crosses would produce the same result]. However, our analysis ( Fig. 2A ) shows that the gilippus matriline is very distinct from those comprising D. chrysippus s.l., and is neither a sister cluster, nor even closely related ( Lushai et al., 2005b).

Tamura- Nei GDs, based upon mtDNA (12S + COI loci), between gilippus and the chrysippus s.l. subtaxa are as follows: dorippus-1, 6.3%; petilia , 5.2%; chrysippus s.s., 4.9%; these distances may be compared with a mere 0.1% that separates gilippus from eresimus . And yet, whereas gilippus and eresimus are reproductively isolated and morphologically distinct species that are sympatric over a huge geographical area, gilippus lacks observed structural apomorphies compared to the allopatric and genetically distant taxa that comprise chrysippus s.l. (A & V-W; Smith et al., 2002 for discussion). The similarity of gilippus and eresimus haplotypes implies that speciation has occurred only within the last 40 000 years or so ( Lushai et al., 2003b). As the two species are now sympatric over most of their combined geographical range, cladogenesis too may have been sympatric, though allopatric or parapatric scenarios are at least equally plausible ( Coyne & Orr, 2004).

If the gilippus-eresimus speciation was sympatric, prezygotic isolation through genitalic and species recognition markers (characters 36, 46, 54, Appendix 1A, B), may have been enhanced, either before speciation by reinforcement, or subsequently by reproductive character displacement ( Butlin, 1989). Moreover, if cladogenesis occurred only millennia ago ( Smith et al., 2002; Lushai et al., 2003b), the gilippus and eresimus clusters must have acquired prezygotic isolation within that short time.

Lushai et al. (2003b) applied molecular clocks for the COI gene in Alphaeus prawns ( Knowlton et al., 1993) and the 12S locus in Littorina (gastropod molluscs), Reid, Rumbak & Thomas (1996) to the gilippus + eresimus and chrysippus s.l. haplotypes. The mean of the two correlated clock rates suggests that cladogenesis between these groups occurred ~2.8 million years ago (Mya), while divergence of the gilippus + eresimus and dorippus-1 matrilines must have been even earlier, around 4.1 Mya. These calculations indicate that the divergence of gilippus + eresimus from all chrysippus s.l. taxa occurred in the Pliocene, 3–4 Myr before the gilippus-eresimus dichotomy.

Thus, the morphological features (characters 46, 54, 55, Appendix 1A, B) that distinguish the gilippus + chrysippus s.l. cluster from eresimus ( Fig. 2B ) are probably symplesiomorphic, not synapomorphies as believed by A & V-W. It follows that butterflies with gilippus + chrysippus s.l. structural morphology must have dispersed to the Americas from the Old World early in the history of the genus, whereas the distinctive morphological features of eresimus , in particular characters 46 and 54 (Appendix 1A, B), are apomorphic and of more recent Neotropical origin.

An alternative scenario is that gilippus and eresimus had originally evolved distinct mitochondrial genomes but subsequent hybridism, possibly a rare or localized event, has resulted in the introgression of cytoplasm from one species to the other ( Lushai et al., 2003b) and thus erased the matrilineal history of the introgressed species. The possibility that such an event might be unique to Grand Cayman, where the samples of both species were collected, has to be acknowledged. However, it is unlikely that two relatively large and, moreover, highly vagile species would remain isolated on a far from remote island for long. This hypothesis could, however, be tested by sequencing mtDNA from sympatric sample pairs in other parts of their shared range.

Lushai et al. (2003a, 2005b) have produced evidence to suggest that hybridism in East Africa among partially isolated subspecies of the D. chrysippus complex is catalysed by female-biased sex ratios that result from male-killer Spiroplasma infections ( Jiggins et al., 2000; see below). If similar events occurred in the recent history of gilippus and eresimus , it could account for the mutual convergence and low diversity of their haplotypes ( Hurst, Hurst & Majerus, 1997) in our samples (N = 8 for both taxa). Although Sperling (1993) has shown that mtDNA is unlikely to cross species boundaries in Papilio butterflies due to the Haldane effect in heterogametic females, he did find one apparent exception in crosses between Papilio multicaudatus and P. rutulus .

Figures

Figure 2. A, single most parsimonious tree based on amalgamated mtDNA 12S rRNA (344 bp†) and COI (676 bp†) genes (Appendix 2) for the genus Danaus, rooted by outgroups Tirumala septentrionis and T. limniace (†values before alignment). Tree statistics: characters in matrix 1025, variable informative characters 54, parsimony informative characters 139, heuristic search length 304, consistency index (CI) = 0.80, homoplasy index (HI) = 0.20, retention index (RI) = 0.81, rescaled consistency index (RCI) = 0.65. Bootstrap replicates/Bremer Support values are shown for each node where significant. Branch lengths are drawn proportional to nucleotide changes and indicated in parentheses. The MP topology depicted here is congruent with the MP strict consensus, NJ and ML algorithms not shown. ML branch significance is P <0.01 unless marked *. See Table 1 for provenances and sample sizes. B, single most parsimonious tree based on morphological, biochemical and colour gene characters (Appendix 1A, B) for the genus Danaus, rooted by outgroups T. septentrionis and T. limniace (characters from T. petiverana, T. hamata and T. ishmoides were used to fill gaps); †character alignment was fixed to character). Tree statistics: characters in matrix 67†, variable uninformative characters 8, parsimony informative characters 59, heuristic search length 95, CI = 0.84, HI = 0.16, RI = 0.90, RCI = 0.75. Other notes as in (A). C, one of 11 most parsimonious trees based on amalgamated nDNA 18S rRNA (525 bp†) and EF1-a (400 bp†) genes for the genus Danaus, rooted by outgroups Amauris niavius and A. ellioti (†values before alignment). Tree statistics: characters in matrix 926, variable uninformative characters 19, parsimony informative characters 24, heuristic search length 52, CI = 0.83, HI = 0.17, RI = 0.72, RCI = 0.60. Other notes as in (A).