Introduction

The neotropical genus Heliconius has long fascinated evolutionary biologists because of puzzles presented by variation of wing pattern within and among species. The strikingly diverse wing patterns which characterize different races of a species apparently evolve rapidly (Brower 1994a) and at different rates from overall genomic divergence (Turner et al. 1979). Studies of gene flow in hybrid zones between races show that genes controlling conspicuous pattern elements exhibit much steeper clines of frequency than genes with little or no effects on pattern (Mallet 1986, Mallet et al. 1990, Jiggins et al. 1996, Linares 1989) and manipulative experiments (Benson 1972, Mallet and Barton 1989) support the idea that visual selection by predators exerts a strong conservative effect to preserve ruling phenotypes within the boundaries of a race. Perhaps the biggest puzzle then, is how such qualitatively distinct patterns arise, persist, and come to characterize an area of a species' range in the first place. Many aspects of this general problem have been thoroughly explored and convincingly explained by Turner and Mallet (1996) but the source of the novelty that fuels this adaptive radiation of mimetic forms remains an unexplored question.

A second interesting puzzle concerns coexisting species of Heliconius within a local community: why are the patterns among species so diverse? Although eight of nine local Heliconius spp. are distasteful to a common butterfly predator in Corcovado Park, Costa Rica (e.g., Chai 1990), at least three distinctive pattern themes (called mimicry rings) coexist there ( Figure 1, top) as is the case in any neotropical rainforest community of 8-10 Heliconius species. So why do so many color pattern mimicry rings coexist at once when, in theory, Müllerian mimicry should favor community wide convergence (see Joron and Mallet 1998)? A hint to the answer comes from how distinct pattern themes within a habitat are related phylogenetically (see Figure 9). For example, in Corcovado Park, the erato/sapho/sara (ESS) clade (see Brower 1994b, Penz 1999), a group which contains five pupal-mating species (all of which share the lack of a signa in the female bursa copulatrix [Brown 1981]), exhibits four distinct pattern themes. Similarly, each of four sympatric melpomene/cydno/silvaniform (MCS) clade species possesses a different pattern theme, and three of these participate in intimate Müllerian partnerships with a pupal-mating (ESS) comodel. In addition to sharing wing patterns, mimetic comodels from different clades also exhibit similar patterns of overall habitat use, including heights of nocturnal roosts (Mallet and Gilbert, 1995).

Reproductive Behavior and Community Assembly

Within the MCS clade, interspecific mating leads both to non-mimetic F1 phenotypes, vulnerable to differential sampling by bird predators (e.g., Benson 1972), and sterile F1 females (in keeping with Haldane's rule). Within the ESS clade, interspecific matings often result in complete inviability of F1 zygotes (personal obs.). ESS (pupal mating) group males are known to compete intensively for mates (Deinert et al. 1994 and Deinert this vol.), and will also attempt mating with heterospecific pupae or emerging adults discovered near their own pupae in greenhouses or in the field (personal obs.). This phenomenon has been suggested as one selective force driving host plant and/or micro-habitat separation within the Heliconius community (Gilbert 1984, Gilbert 1991) because not only do ESS males sterilize female pupae of other ESS species, they may disable and kill both sexes of MCS group teneral individuals found on or near host being searched for pupae (personal obs.).

Realistic scenarios for the assembly of species-rich local communities of Heliconius (e.g., Figure 1) will take account of the potential for such reproductive interference along with other aspects of interspecific interaction (Gilbert 1991). Thus, the probability that sibling species will be able to establish breeding populations and coexist after arrival in a suitable patch of habitat should depend on the degree to which each has evolved distinct microhabitat and/or host plant preferences while in geographically isolated (Jiggins et al. 1997, Mallet 1993, Mallet et al. 1998). Incipient niche divergence would reduce competitive and reproductive interference between a newly arrived species and a resident sibling, or any other Heliconius species which already inhabit the area.

Wing Patterns and Species Packing

Does possessing a highly distinct wing pattern assist a new arrival's establishment in a local community? If it does, then local species diversity of Heliconius could be at least partly due to its capacity to generate different wing patterns. Wing-pattern diversity seems most causally connected to local species diversity in groups like the MCS clade, which appear to rely heavily on visual signals in courtship. This is because probability that an additional species can be packed into a local community would depend upon its wing pattern being sufficiently distinct from close relatives to minimize confusion in courtship which might lead to hybridization. By the same token, its degree of mimicry of a well-established non-MCS aposematic wing pattern (not necessarily that of other Heliconius) should promote a new arrival's chances of establishment.

Such mimetic convergence of wing pattern in a local community of Heliconius (see Figure 1), tends to be found between species from different clades which are able to share micro-habitat (e.g., Mallet and Gilbert 1995) by virtue of having distinctly different mating and oviposition habits. Such cross-clade combinations may coevolve (in its micro-evolutionary sense) towards identical wing patterns as Müllerian mutualists (Gilbert 1983) in spite of also being competitors for resources (Templeton and Gilbert 1985). Excluding the highly polymorphic H. numata, whose various forms mimic different ithomiine butterflies (Brown and Benson 1975), the intra-generic mimicry we find in Heliconius insures that the number of different wing pattern themes coexisting locally will be always be somewhat less than the total number of Heliconius species present.

Wing Patterns in Space and Time

From region to region quantum jumps with respect to wing pattern are taken in concert by each species of a Müllerian partnership. Müllerian pairs such as H. erato and H. melpomene exhibit concordant racial changes and parallel hybrid zones (Turner 1981) and provide comparative systems for the ecological and evolutionary genetics of mimetic adaptation (e.g., Mallet et al. 1990, Jiggins and McMillian 1997), behavioral mimicry (Srygley and Ellington 1999), coevolution and phylogeny of mimetic races (Brower 1996b), and speciation (e.g., Mallet 1993).

It is also striking how novel, superficially unrelated wing patterns "punctuate" racial variation within species like H. erato and comodel H. melpomene (see Turner 1983) no less than they characterize interspecific variation across the entire genus Heliconius. This generality is most interesting in terms of pattern evolution within clades of Heliconius for two reasons: (1) underlying mechanisms of pattern development (see "tool box" below) are likely to be shared by all members of a clade and (2) genes that determine pattern are known to pass across species boundaries in some clades such as MCS (see below).

How do pattern novelties arise within a lineage? To date, the approach to this question has been at the level of evolutionary mechanisms (e.g., Turner and Mallet 1996), such as "biotic drift" (Brown et al. 1974), or "shifting balance" ( Mallet 1993, but see Coyne et al. 1997); i.e., processes that would filter, concentrate, and preserve pattern novelty within species. Implicit in this literature is the assumption that mutation is sufficient to provide the variation necessary for novelty-selecting mechanisms to work. But if that's all there is to it, one has to ask why Heliconius should be so extraordinary in its display of intra-specific and intra-generic wing pattern novelty. Linares (1989 1997b) has implicated hybridization as a process generating pattern novelty, but introgression, while of major importance in this system, is not sufficient to explain why Heliconius, but not related genera, displays such variety of pattern, so rapidly evolved.

In this chapter I attempt to account for the origin of novel wing pattern in Heliconius by proposing developmental/genetic mechanisms which enhance the capacity of introgression to generate qualitative genetic variation. I then attempt to explain the diversification of the genus in light of these mechanisms and our knowledge of its natural history and ecology.

This chapter will address the following questions:

1. Why are so many wing patterns available to Heliconius?

2. Why are so few wing patterns successful in nature?

3. What are the likely circumstances for cross-species recombination in nature?

4. What are the relative contributions of mutation and introgression to pattern evolution?

5. How might evolving Heliconius jump from one adaptive peak to another without crossing fitness valleys below?

6. Does one clade typically drive the evolution of novel patterns?

7. Do Heliconius and particularly its MCS clade possess unique developmental genetic mechanisms for generating wing pattern?

8. How did these developmental mechanisms i.e., "the tool box," originate?