ORIGINS OF DIVERSITY:

THE EVOLUTIONARY GENETICS OF CARIBBEAN BUTTERFLIES.

A thesis submitted in fiilfilment of the requirements for the degree of

Doctor of Philosophy

by

Neil Davies

London, July 1995 ProQuest Number: 10055864

All rights reserved

INFORMATION TO ALL USERS The quality of this reproduction is dependent upon the quality of the copy submitted.

In the unlikely event that the author did not send a complete manuscript and there are missing pages, these will be noted. Also, if material had to be removed, a note will indicate the deletion. uest.

ProQuest 10055864

Published by ProQuest LLC(2016). Copyright of the Dissertation is held by the Author.

All rights reserved. This work is protected against unauthorized copying under Title 17, United States Code. Microform Edition © ProQuest LLC.

ProQuest LLC 789 East Eisenhower Parkway P.O. Box 1346 Ann Arbor, Ml 48106-1346 I do not know wiiat I may appear to the world, but to myself I seem to have been only like a boy playing on the seashore, and diverting myself in now and then finding a smoother pebble or a prettier shell than ordinary, whilst the great ocean of truth lay all undiscovered before me.

Isaac Newton TO MY PARENTS TABLE OF CONTENTS

A bstract 6 Acknowledgements 8 1. Introduction 10 1.1. Bibliography 14 2. The species richness of West Indian butterfly faunas 2.1. Abstract 16 2.2. Introduction 17 2.3. Methods 21 2.4. Results 29 2.5. Discussion 34 2.6. Bibliography 52 2.7. Tables 66 2.8. Figures 70 3. Genetic differentiation in four species of West Indian butterfly 3.1. Abstract 74 3.2. Introduction 75 3.3. Methods 80 3.4. Results 84 3.5. Discussion 89 3.6. Bibhography 100 3.7. Tables 107 3.8. Figures 111 3.9. Appendices 120 4. The historical biogeography of the West Indian butterfly Dtyas iulia (Lepidoptera: Heliconiidae) 4.1. Abstract 134 4.2. Introduction 135 4.3. Methods 137 4.4. Results 140 4.5. Discussion 142 4.6. Bibliography 146 4.7. Tables 150 4.8. Figures 152 5. The island biogeography of genetic variation 5.1. Abstract 156 5.2. Introduction 157 5.3. Methods 160 5.4. Results 162 5.5. Discussion 164 5.6. Bibhography 174 5.7. Tables 178 5.8. Figures 179 6. Classifying island populations 6.1. Abstract 187 6.2. Bibhography 196 7. Fuzzy species 7.1. Abstract 200 7.2. Bibhography 217 7.3. Figures 222 8. Reproductive isolation in a butterfly hybrid zone: a subtle Haldane effect 8.1. Abstract 224 8.2. Introduction 225 8.3. Methods 227 8.4. Results 233 8.5. Discussion 237 8.6. Bibhography 242 8.7. Tables 246 9. Haldane’s rule: old theories are the best 9.1. Introduction 258 9.2. Bibhography 264 9.3. Figure 266 ABSTRACT

Darwinian evolution requires and generates biodiversity. The biodiversity of West

Indian butterflies is investigated at the population and species level The latest butterfly species numbers for West Indian islands are presented and the fectors wdiich influence community richness are determined. Area is by far the most significant physical variable. The nature of the species-area relationship is affected by community size: smaller communities seem more vagile, flattening the species- area curve. The genetic structure of four West Indian butterflies is surveyed using allozyme electrophoresis. A phenotypic tendency towards subspeciation is corroborated by the allozyme data. A long-standing controversy in West Indian biogeography is whether the 6una arose by dispersal from the continent, or is the remnants of a continental fruna isolated by ancient geological events. Whilst the action of vicariance cannot be conq)letely ruled out, any lasting inq)ression has certainly been masked by more recent dispersal. One of the species, Dryas iulia, has a particularly complex genetic structure in the West Indies. A survey of mtDNA variation confirms the major phylogenetic breaks identified by the allozyme data. Apphcation of the molecular clock indicates a major separation between West Indian Dryas iulia and continental populations around 2.5 million years ago. This is too recent for the vicariance hypothesis.

Classification is inevitably difficult with insular populations, some of which are on the borderline between differentiated subpopulation and separate species. Such populations, although anticipated by evolutionary biology, represent a challenge for taxonomy. The difficulties are discussed with particular reference to phylogeny reconstruction. The species-subspecies interface is also examined empirically. Pre and postzygotic isolation between two hybridising Central American butterflies are evaluated. The results are discussed in the context of spéciation genetics, particularly Haldane’s rule. Muller’s dominance theory is assessed as an explanation for Haldane’s rule. ACKNOWLEDGEMENTS

Many people have contributed to this project and the following deserve special thanks. My supervisor, Andrew Pomiankowski, allowed me the freedom to leam from my own mistakes whilst making sure I didn’t stray too far off track. Ifis comments and criticisms, along with those of Jim Mallet, have added greatly to this thesis. Much of my time was spent with the University of Puerto Rico (UPR) at

Mayaguez and the Smithsonian Tropical Research Institute (STRI) in Panama. I am very grateful to both institutions for their support. In particular I would like to thank Stuart Ramos of the UPR for providing space in his laboratory and sharing his great knowledge of West Indian butterflies. Annette Aiello and Eldredge

Bermingham, from STRI, provided invaluable advise and assistance, for which I will always be grateful. Collecting specimens was made more pleasurable and productive thanks to numerous people, only a few of which are mentioned here:

Luis Roberto Hernandez, Jesus Bretado and Behsario Cepeda, Lee and Jaquie

Miller, Tom Turner, Kelvin Antonio, Dennis Knowles, Peter Evans, Grilles Seutin,

Adela Olivardia, and Messrs. Bonnet, Fennec and Tanasi. I would also like to thank Igor Emelianov, Peter King, Owen McMillan and Stephen Stokes. This project would never have happened had it not been for David Spencer Smith, who introduced me to West Indian butterflies. He has continued to play an important and valued role in my studies. I thank him, and his wife Sylvia, for their he^ and friendship over the past four years. Finally I have to thank my frmily - for everything really, but not least their tolerance during the writing of this thesis. Financial support was provided by a Science and Engineering Research Council

Ph.D. studentship, a short-term fellowship from the Smithsonian Institution and a

Galton bursaiy from the Department of Genetics and Biometry, University College

London. CHAPTER 1

INTRODUCTION

This project began with the 1989 Oxford University Grenadines E?q)edition, wiiich set out to describe the butterflies of the tiny Grenadine islands in the West Indies.

West Indian butterflies have a proud, if little recognised, place in the history of island biogeography. Eugene Munroe was patron to the OU Grenadines

Expedition and it was in his doctoral thesis (Munroe, 1948) that the modem theory of island biogeography was first formulated (Gilbert, 1984; Brown and Lomolino,

1989; Wilkinson, 1993). In 1948 West Indies butterflies were inconq)letely known and since then many new species have been described and new island records added.

Norman Riley was the first to bring together the whole West Indian butterfly fauna

(Riley, 1975). This has recently been updated in an even more comprehensive account of the region’s butterflies (Smith et al., 1994). These studies, and the many investigations of individual islands and butterflies on which they are based, form a solid foundation for the work presented here. With the application of genetic techniques, this thesis begins a new era in the study of West Indian butterflies. Such genetic studies might finally resolve some of the questions which have long vexed West Indian lepidopterists. Maybe West Indian butterflies, which

10 were at the forefront of classical island biogeography, will also be in the vanguard

of genetical island biogeography.

This thesis is rooted in classical island biogeography and opens with an updated

examination of the species richness of West Indian butterflies. Based on the data

from his book (Smith et ai, 1994) David Spencer Smith and I put together hsts of

species numbers and physical variables for many of the West Indies. I present and

analyse this data in Chapter 2. Essentially, this reworks Munroe’s seminal study

(Munroe, 1948) and provides an insight into how our perception of the species-

area relationship has changed with more data.

I examine the influence of various factors on species richness, and ask vdiether

island biogeography can be treated as a stochastic process. The results suggest that, in general, the number of butterfly species found on West Indian islands is

strongly correlated with their area. This is consistent with the stochastic hypothesis that smaller populations face an increased probabihty o f ‘random’

extinction. However, the species-area relationship could also result from selective

(deterministic) extinction or immigration affecting individual species in a variety of

different ways. Without an examination of actual species distributions (presence-

absence data is not yet available), it is not possible to exclude this alternative.

Chapter 2 provides the framework for the following three chapters which deal with island biogeography at a genetic level.

11 In Chapter 3 four butterfly species are examined using allozyme electrophoresis.

The survey is set in the context of the vicariance versus dispersal debate v ^ c h has long defined West Indian biogeography (Liebherr, 1988; Woods, 1989). Genetic data provides a new, and potentially powerful tool with which to resolve this controversy. For this study I was assisted in the laboratory by Igor Emelianov and

Peter King. The results show that population genetic structure varies greatly among species and that dispersal has been the most important factor in their historical biogeography. Although vicariance cannot be con^letely ruled out, it seems improbable given the levels of genetic divergence.

One of the species in the allozyme survey, Dryas iulia, was particularly interesting, having a marked and somewdiat surprising genetic structure. This species was therefore chosen for further genetic analyses, this time using mtDNA sequence variation (Chapter 4). This study, which is in many ways a pilot for future work, provides a useflil comparison with the allozyme data and a more accurate means of dating the major phylogenetic breaks. Assuming that mtDNA evolves in a roughly clocklike fashion, and even with the potentially large error associated with such an assumption, it appears unlikely that Dryas iulia travelled with the West Indies as they drifted into the Caribbean. This work was done with Eldredge Bermingham at the Naos laboratories of the Smithsonian Tropical Research Institute (STRI). I carried out all the data analysis, the DNA extractions and some of the primer trials, however, the actual sequencing was mainly conducted by Connie Colman of the

Bermingham laboratory.

12 Chapter 5 brings together species richness (Chapter 2) and genetic variation

(Chapter 3) in an integrated approach to the study of insular biodiversity. It is argued that species richness and intraspecific genetic variation should be positively correlated if stochastic neutral processes determine overall biodiversity

(MacArthur and Wilson, 1967; Kimura, 1983). This expectation is fulfilled in three of the four species. The exception is thought to result fi'om differences in dispersal

abihty among West Indian butterflies rather than selection or other deterministic forces.

Anyone studying intraspecific variation among insular populations will sooner or later wonder about the nature of species. Chapter 6 addresses some of the difficulties faced when classifying allopatric populations and asks whether genetic data can inçrove insular taxonomy. It is clear from this discussion that there are more general issues at stake and these are dealt with in more detail in the next chapter.

In Chapter 7 a concept of species is developed with reference to the feasibihty of phylogeny reconstruction. It is argued that cladistics is the best system of classification and that its basic units, species, are irrevocably isolated gene pools.

Such species, however, are only an ideal and real phylogenies will be fuzzy.

Although irreversible isolation can occur through ecological or geographical means, it is more likely to be permanently achieved through pre or postzygotic barriers to reproduction.

13 The evolution of such reproductive barriers is examined in Chapter 8, an ençirical study of the species boundary. In the late 1970s Robert Silberglied and Annette

Aiello of the Smithsonian Tropical Research Institute conducted a series of breeding experiments to investigate reproductive isolation in two hybridising

Panamanian butterflies. Robert Silberglied was tragically killed in a plane crash before any of the data could be analysed and it was left untouched for the next fifteen years. Annette Aiello kindly agreed to let me analyse the data. The results are presented in Chapter 8 where they are discussed in relation to spéciation theory. The data are of particular relevance to Haldane’s rule (Haldane, 1922) which states that vsdien one sex of hybrid is absent, sterile or inviable it is usually the heterogametic sex. In Chapter 9 Muller’s dominance theory (Muller, 1940) is assessed as an explanation for Haldane’s rule. This theory is convincing and is supported by some of the data in Chapter 8. Chapter 9 has been submitted for publication with Andrew Pomiankowski as co-author.

BIBLIOGRAPHY

Brown, J.H. and Lomolino, M. V. 1969. Independent discovery of the equihbrium theory of island biogeography. Ecology 70:1954-1957.

Gilbert, L.E. 1984. The biology of butterfly communities. In: The Biology of Butterflies, Symposium of the Royal Entomological Society of London, No. 11. Vane-Wright, RI. and Ackery, P R. (eds ). Academic Press, London, pp.41-54.

Haldane, J.B.S. 1922. Sex-ratio and unisexual sterihty in hybrid animals. J. Genetics 12: 101-109.

14 Kimura, M. 1983. The Neutral Theory of Molecular Evolution. Cambridge University Press, N.Y., U.S.A.

Liebherr, J.K (ed.) 1988. Zoogeography of Caribbean Insects. Cornell University Press, Ithaca N.Y.

MacArthur, R.H. and Wilson, E.O. 1967. The Theory of Island Biogeography. Princeton University Press, NJ.

Muller, H.J. 1940. Bearings of the Drosophila work on systematics. In: The New Systematics. Huxley, J.H. (ed). Clarendon Press, Oxford, pp. 185-268.

Munroe, E.G. 1948 . The Geographical Distribution of Butterflies in the West Indies. Dissertation, Cornell University, Ithaca, NY.

Riley, N.D. 1975. A field guide to the butterflies of the West Indies. Collins, London.

Smith, D.S., Miller L , and Miller, J. 1994. The Butterflies of the West Indies and South Florida. Oxford University Press, Oxford.

Wilkinson, D M. 1993 . Equilibrium island biogeography: its independent invention and the marketing of scientific theories. Global Ecol. and Biogeog. 3:65-66.

Woods, C A (ed.) 1989. Biogeography of the West Indies. Sandhill Crane Press, Inc. Gainesville FL. pp. 229-262.

15 CHAPTER 2

THE SPECffiS RICHNESS OF WEST INDIAN BUTTERFLY FAUNAS.

ABSTRACT

Numbers of butterfly and plant species on West Indian islands are presented. The specific limitations on these lists and the general difficulties in conq)iling such data sets are discussed. Species numbers are probably under-estimates because West

Indian islands have not been the subject of long-term intensive surveying.

Although difficulties in recognising extinction and the inclusion of vagrant species both risk the over-estimation of species numbers, this is considered to be of lesser importance. The species-area relationshÿ is also examined, in particular its stability with respect to the accumulation of data. Over the last fifty years the species-area regression line for West Indian butterflies has flattened slightly whilst the intercept has risen. The scatter about the line, however, has increased and it is also heterogeneous in terms of slope: smaller islands have flatter regression lines than larger islands. Possible explanations for these observations are discussed and the relative importance of stochastic and deterministic influences considered.

16 INTRODUCTION

Ecological, biogeographical and evolutionary change are all expected to contribute to local biodiversity (Ricklefs, 1987). Archipelagos allow these processes to be

corcpared in a single system. Islands are ‘natural laboratories’: they vary in geological history, size, height, and remoteness, but also have definite boundaries

and relatively single communities (Diamond and May, 1976). The ecological

study of archipelagos is of wider significance because all ecosystems are heterogeneous and fi’agmented. Although less clearly defined, ‘continental islands’ can still be recognised by their common physical, biotic and historical characteristics. Island biogeography is therefore vital to the understanding of community structure (Begon et al., 1986).

This study presents the numbers of butterfly species recorded fi'om sixty seven

West Indian islands and south Florida. A list of plant species numbers is also presented for those islands wiiose floras are well known. Although less exhaustive than the butterfly data, this nevertheless represents a usefiil first step. Species hsts are the raw material of ecological and biogeographical study and it is important to understand their limitations; the inherent problems of such data sets are therefore discussed. The data is then used to investigate the generation of species richness in

West Indian butterflies: what determines the size of butterfly faunas?

One of the oldest and strongest rules in community ecology is the existence of a linear relationship between the number of species (species richness) and the area in

17 wiiich they are found (Arrhenius, 1921; Gleason, 1922). The mechanics of this relationship are unclear, as are the rules of community assembly in general (Cody,

1989). One particular controversy is the extent to which communities are open or closed (Roughgarden, 1989). Closed communities have finite membership (Elton,

1933) and species richness is limited either by competition (Hutchinson, 1959) or

‘ecological impoverishment’ (Lack, 1976). Open communities, conversely, have unlimited membership and species richness is determined by the stochastic processes of colonisation and extinction (Gleason, 1926; Preston, 1962;

MacArthur and Wilson 1963, 1967).

MacArthur and Wilson (1963, 1967) combined elements of both the open and closed schools in their equilibrium theory of island biogeography. In its smq)lest form, however, theirs is essentially a stochastic model (Whittaker, 1992). The species-area relationship results fi'om higher extinction rates on small islands. This rests upon two assurcptions: population sizes must be lower on smaller islands, and small populations must be more susceptible to random extinction (Williamson,

1981). Deviations fi'om the species-area relationship are explained by differential isolation: less isolated islands will receive more immigrants and so support more species. At a certain species richness (determined by the area and isolation of the island) immigration and extinction balance. At this point a dynamic equihbrium is attained: species composition is in a state of flux (temporal turnover) but species number remains constant.

18 This theory was actually first developed by Eugene Munroe in a doctoral thesis on

West Indian butterflies (Gilbert, 1984; Brown and Lomolino, 1989; Wilkinson,

1993). Munroe (1948, 1953) discovered the relationship between butterfly species richness and island area in the West Indies, but his seminal study is rarely cited in reviews of the species-area relationship. Munroe did not publish his resuhs, except for an equation describing his model which was only pubhshed in abstract

(Munroe, 1953). Moreover, he failed to stress the general apphcabihty of his ideas

(Wilkinson, 1993). Here two facets of the relationship are of interest: firstly, how resihent has it been to the addition of new data, and secondly, what can it tell us about the forces governing community structure?

In 1948 butterfly lists for most of the West Indies were very incomplete. Munroe was therefore obliged to group many of the smaller islands. Out of his eleven data points, five of the seven smallest represented pooled data for the northern and southern Lesser Antilles, the Cayman islands, the Virgin islands, and the entire

Bahamas archipelago. Some twenty years later small islands still had to be amalgamated for biogeographic analysis (Scott, 1972). Since then, however, many island hsts have been greatly augmented (Riley, 1975; Scott, 1986; Smith et al.,

1994a). How have these new data changed our perception of the species-area relationship for West Indian butterflies? This is inçortant as many data sets

(particularly those of invertebrates) are still at an early stage of compilation. To what extent can we use such incomplete data in ecological and biogeographical study?

19 The data presented here are still incomplete, certainly in conq)arison with the bird or mammal data sets. However, inasmuch as any tropical invertebrate fauna can be used to address questions of island biogeography. West Indian butterflies are as good as any and better than most. Following Connor and McCoy (1970) a purely stochastic model forms the null hypothesis for investigating species richness. I assume that habitat diversity, interactions among species, immigration, and spéciation are all insignificant influences on West Indian butterfly feunas.

Furthermore island feunas are at equilibrium Le. they are not affected by historical events such as land-bridges. In this clearly unrealistic situation, species richness is solely determined by extinction and population size.

If population size increases linearly with area, species richness is e?q)ected to correlate strongly with island area (Diamond and May, 1976 p. 169). Preston

(1962) showed that when individual abundance is distributed among species in a log-normal canonical fashion (as is common in natural populations) then species will be related to area according to the power function, logS = z log A + c, where

S is the species number, A is island area, c the intercept and z the slope of the regression line. So long as the species-abundance distribution is canonical (see

Rosenweig, 1995 p.268-272) then z will be 0.26 (Preston, 1962). Even if the relationship is not canonical, its plausible form still restricts the range of z to between 0.15 and 0.39 (May, 1975).

20 If species richness is not related to area in this way then various explanations could follow: i) immigration may be inq)ortant, ii) island feunas may not be at equilibrium, or iii) 6unas may be closed.

METHODS

Determining species richness

The butterfly numbers are based on the recent survey of West Indian butterflies by

Smith et a l (1994a), other published sources, and unpublished records. A conservative view has been taken, assuming that species, once reliably recorded,

are still extant even if they have not been seen for many years. All the island records supported by a museum specimen are included. Such records are only

excluded when there is reasonable doubt over their legitimacy due to mislabelhng.

In most cases, rejected records are old and the species has not been recorded on

any West Indian island before or since. Every effort was made to avoid the introduction of personal biological biases as constraints on the list. Further caveats

concerning the interpretation of island feunal Hsts are considered below.

The islands

Historically the West Indies have included all the Caribbean islands, the Bahamas

and even parts of mainland South America (see figure 1, Chapter 3 p. 111).

FaunisticaUy, however, mainland South and Central America, together with their coastal islands, are usuaUy excluded (Smith et al., 1994a). This convention is

21 followed here. The most notable exclusion is Trinidad, lying only 15 km of the coast of Venezuela. Although relatively small (4820 km^), this island supports over 600 butterfly species (Barcant, 1970) compared with only some 350 in all the

West Indies (Smith et al. 1994a). Essentially the butterfly fauna of Trinidad is a virtually unmodified subset of the adjacent South American mainland, to which it has been recently linked; only some 10% of Trinidad’s butterflies are ‘West

Indian’.

South Florida is here treated as part of the West Indies. Whilst part of mainland

North America, the southern tip of Florida (Dade and Munroe Counties) with the

Florida Keys, may be regarded ecologically as a West Indian island, surrounded by ocean and with tenq)erate land to the north. In this area well over half the butterflies are of Neotropical, rather than Nearctic afiSnities (Smith et al, 1994a)

In addition to analysing the entire West Indies, islands are also subdivided into three classes: (1) islands larger than Guadeloupe, the largest of the Lesser Antilles,

(2) the Lesser Antilles, and (3) islands with fewer than twenty butterfly species

(Table 1). For convenience I will refer to Class 1 as the Greater Antilles, although it also includes two Bahamian islands and south Florida. Our object is to determine whether the relative in^ortance of factors influencing species richness is scale-dependent. Furthermore the classes allow the influence of geological history to be assessed: islands in Classes 2 and 3 tend to be young and oceanic, whilst those in Class 1 are older and may have been linked to one another and/or the continent (WiUiams, 1989).

22 Estimating environmental variables

Area might increase species richness because larger islands tend to have more individuals (Preston, 1960, 1962) or because they have more habitats (Wilhams,

1964). Separating these two factors is difficult in natural systems, but both are known to be important (Rosenweig, 1995 p.217-228). Habitat diversity can be estimated indirectly by its association with physical heterogeneity. An attençt to measure the influence of habitat diversity on butterfly faunas is made by assessing the correlation between species richness and physical heterogeneity with area held constant. Notice that the absence of a correlation does not mean that species richness is determined solely by population size and is unaffected by habitat diversity. It may sircç)ly reflect a lack of variation in habitat diversity independent of area; they may act as ‘surrogates’ of one another (Rosenweig, 1995 p.208).

1. Physical heterogeneity

Physical heterogeneity enhances the suitability of islands to a greater range of species by increasing habitat diversity. The maximum elevation of islands (see below) is used as an estimate of physical heterogeneity. This is a common procedure in island ecology (Power, 1971; Diamond and Mayr, 1976). All of the

West Indies are tropical or subtropical and the regional climate is fairly homogenous. Local climatic conditions, however, are greatly affected by the topology of individual islands. Mountains cause variability in tenq)erature, oreographic rainfall and produce rainshadows. This creates a number of distinct climatic zones on an island, increasing the range of possible habitats.

23 Physical heterogeneity can also increase the spéciation rate, since populations are more likely to become geographically isolated on more conçlex heterogeneous islands. Mountainous areas are associated with mammahan species richness in

North America (Sinçson, 1964), and the radiation of the Calisto (Satyridae) butterflies on Hispaniola seems best explained by the great topographic diversity of the island (Smith et ai, 1994a).

2. Area and Elevation

Area and maximum elevation are the most straightforward variables to measure.

Geographical data on most of the West Indies are published and generally the most recent figures are used (Evans, 1973; Hunter, 1994). It should be noted, however, that figures vary slightly fi'om one source to the next.

3. Latitude

There are many exan^les of how species richness declines with distance from the equator (Fischer, 1961). Terborgh (1973) explained this latitude gradient as merely an area efrect: there is more land-area in the tropics. If this is correct then we would not expect latitude per se to effect species richness in the West Indian islands, the vast majority of which are tropical. However, primary productivity also decreases at higher latitudes and this might cause a decline in species richness

(Rosenweig, 1995 p.295-6). The influence of latitude on species richness is therefore investigated. Latitudes are taken at the longest (latitudinal) axis of an

24 island. For the plant data, where some islands are grouped, the latitude of the largest island in the group is used.

4. Isolation

The isolation of an island is not singly the distance between it and the closest mainland point. Any land mass supporting a butterfly fauna is a potential source of butterflies, although some sources may be more significant than others. Islands close to poor sources may be more isolated than islands fiirther away flrom rich

sources. Island isolation may be considered a resultant of the interaction between the remoteness and richness of source(s). There are many ways to estimate isolation, for exarr^le. Power (1971) used nine different measures. The problem is that no single measure of isolation is appropriate for all taxa, or even for the same taxa in different archipelagos.

The inçortance of isolation can be shown when a simple relationship exists between islands and the source of their fauna. Diamond (1972) demonstrated that bird species richness on islands in the south-west Pacific declines with distance fi'om Papua New Guinea (Rosenweig, 1995 p. 237-241). The inq)ortance of this

'distance effect’ is much harder to determine in a cort^lex geographical situation.

Atten^ts to estimate isolation become inevitably subjective when several potential sources are present. The only criterion for choosing one index over another is whether it explains more variation in species richness. This is clearly circular and we cannot distinguish the ‘right’ index fi'om a chance correlation.

25 There is no single source of West Indian butterflies (Brown and Heineman, 1972 p.76; Scott, 1972). South America largely supplies the Lesser Antilles, but its influence wanes in the northern islands where the Greater Antilles become progressively more in^ortant sources. Cuba, Hispaniola and Florida are inçortant sources of the Bahamian fauna, whilst the Greater Antilles show aflSnities to

Central and South America as well as displaying a remarkable level of endemism

(Smithet al., 1994a). Inter-island colonisations have clearly occurred within the

West Indies. This, and the need to weight sources according to their richness, yields an almost infinite number of possible isolation indices. Indeed there is yet another conq)lication: the size of islands and the distances between them have varied greatly throughout their geological history (Wilhams, 1989). Historical biogeography can have an inq)ortant intact on current community structure

(Wilcox, 1978; Bermingham and Avise, 1986).

Despite all these problems I decided, a priori, to use a single, simple measure of isolation for butterfly faunas. The index chosen was the minimum great circle distance fi'om an island to the nearest land mass (island or continent) which supports about twice the number of butterfly species. Islands further fi’om species- rich sources are expected to have fewer species. I have not attempted to determine an isolation index for plants because many of the data points are island groups, further complicatiag the geographical situation.

26 5. Plant species richness

Estimates of plant species number were obtained from island floras: for the

Bahamas, Turks and Caicos Islands (Correll and Correll, 1982), for Cuba (Léon and Alain, 1946-1963), for the Isle of Pines (Jennings, 1917; Gort et al, 1994), for the Cayman Islands (Proctor, 1984), for Jamaica (Adams, 1972), for Puerto Rico and adjacent islands (Liogier and Martorrel, 1982), for Mona (Woodbury et al,

1977), and for Anegada (D’Arcy, 1975). No conq>rehensive account of the flora of the Lesser Antilles is available, but estimates for several of these islands are given in Davis et al (1986). The figures used generally include spermatophytes

(gymnosperms and angiosperms), but should be regarded as only a first approximation since some of the floras are more con^rehensively known than others, and estimates of introduced plants are not uniformly presented.

6. Disturbance

All of the West Indies are often hit by hurricanes. The impact of hurricanes on both dispersal and extinction is the subject of considerable speculation but little hard evidence. I have not, therefore, attempted to evaluate the influence of such disturbance on species richness. Another, if much more recent, source of disturbance is man. Many of the islands have been subject to massive changes in land-use over the last few centuries; for exang)le Barbados, and many other islands, were largely converted to sugar cane plantations. Again we do not know the in^)act such changes may have had (and are undoubtedly having) on the flora or fauna. I have therefore not attempted to incorporate them into the analyses.

27 One of the main problems in assessing the importance of disturbance on these island feunas is the lack of knowledge with respect to their ecology and species conçosition before the disturbance event. Hopefully surveys such as this will help in the future study of this mq)ortant question.

Data analysis

The traditional expectation of a linear correlation between island area and species number is tested. The power function model (Arrhenius, 1921) is used, log- transforming both species number and island area. This has become the standard approach in species-area studies and is therefore of most conq)arative value

(Connor and McCoy, 1979). Other transformations include the e?q)onential function (species number versus log area) which was used by Munroe (1948). The e?q)onential function and untransformed data occasionally yield stronger correlation coefficients than the power function, so these are also calculated.

The influence of elevation and isolation are tested in a manner similar to that used for area. Step-wise regressions are performed to determine the independent influence of factors. For example, area is held constant to determine the independent influence of elevation (the residuals of the species-area regression line are regressed against elevation).

28 RESULTS

The number of butterfly species for each island, together with the island’s physical data are shown in Table 1. Plant species numbers and the con^arable butterfly figures are hsted in Table 2.

Species-area relationship

Preston (1962) explained the species-area relationship in terms of the power function: logS = z log A + c, vdiich is expected to give a slope ofz = 0.26. Table

3 shows how for the last fifty years data on West Indian butterflies has generally supported this prediction. Over three surveys (Munroe, 1948; Scott, 1986; and this) the slope has flattened somewhat from z = 0.26 to z = 0.20, whilst the intercept has risen from c = 0.80 to c = 1.06. The change in slope could not have been readily predicted, although the intercept is expected to rise as island faunas become better known (Wilson, 1988). The power function, however, has not always given the best linear fit to the data. Munroe (1948) found that the exponential function (species number versus log area) gave a stronger correlation, whilst for Scott (1986) the choice of function made httle difference.

In our survey butterfly species richness is best correlated with island area according to the power function (r^ = 0.64, F = 118.08, p<0.001) (Figure 1). The additional data, however, has increased the scatter about the regression line (r^ has fallen) and flattened the slope. Although the slope (z = 0.20) remains within the predicted range of the stochastic model (Preston, 1962; MacArthur and Wilson, 1967; May,

29 1975) it is flat conq)ared with most West Indian taxa, including: ants, z = 0.28

(Wilson, 1988), beetles, z = 0.34 (Darlington, 1943), birds, z = 0.24 (Hamilton et al, 1964), the herpetofauna, z = 0.30 (Darlington, 1957), and plants, z = 0.37 (this paper).

A very similar species-area relationship is found when the Lesser Antilles are treated separately (r^ = 0,60, F = 38.00, p<0.001). For the Greater Antilles, however, the e?q)onential function gives a better fit (r^ = 0.83, F = 33.66, p<0.001) than the power function (r^ = 0.69, F = 15.81, p<0.01). The regression line is also significantly steeper (F = 8.09; d.f = 1,32; p<0.01) in the Greater Antilles (z =

0.37) than in the Lesser Antilles (z = 0.20) (Figure 2). For the 19 islands with feunas of less than twenty recorded butterflies, the correlation between species richness and area is only just significant (r^ = 0.23, F = 4.99, p<0.05) and the slope is very flat (z = 0.06) (Figure 2).

The scatter about the regression line, and the heterogeneity of its slope, suggest that deterministic factors may contribute to butterfly species richness in the West

Indies. In particular, the flat relationship for small faunas is contrary to the theoretical expectations of the stochastic model (Diamond and May, 1976).

Influence of elevation

For the entire West Indies, the Greater Antilles, and islands with small butterfly faunas, species richness and maximum elevation are not correlated once the effect of area is removed. This suggests that physical heterogeneity, and hence habitat

30 diversity, are unimportant determinants of fauna size. Hov^ever, in the Lesser

Antilles species richness was more closely correlated with elevation (r^ = 0.68, F =

53.60, p<0.001) than with area (r^ = 0.60, F = 38.00, p<0.001). Elevation also e?q)lained a significant amount of the residual variation in species richness with area held constant (r^ = 0.16, F = 4.81, p<0.05).

Physical heterogeneity may therefore influence species richness, at least in the

Lesser Antilles. In the Greater Antilles this influence might be hidden because area and maximum elevation are more strongly correlated (r^ = 0.67, F = 14.33, p<0.01) than in the Lesser Antilles (r^ = 0.48, F = 23.41, p<0.001).

Elevation is therefore correlated with species richness, independently of area. This effect is most noticeable in the Lesser Antilles, possibly suggesting that habitat diversity may be a more inq)ortant determinant of species richness in these islands.

A lack of habitat diversity may explain the flat slope of the species-area regression lines in the smaller islands (see below).

Influence of latitude

With area held constant latitude was not correlated with species richness, reflecting the climatic homogeneity of the region. Local climatic conditions on individual islands are probably a greater source of variation in primary productivity among

West Indian islands.

31 Influence of isolation

Isolation was not significantly correlated with species richness once the effects of area and elevation have been removed. In the entire West Indies isolation was correlated with area (r^ = 0.39, F = 41.83, p<0.001) and with elevation (r^ = 0.38,

F = 40.15, p<0.001) i.e. small fiat islands were less isolated. This suggests that greater distance fi'om colonisation sources, traditionally regarded as a barrier to dispersal, does not strongly limit butterfly species richness in the West Indies. The more inq)ortant limiting factor may be estabhshment and survival on islands.

Plants and butterflies

Plant species richness is strongly correlated with area (Figure 3) on a log-log scale

(r^ = 0.81; F = 96.15; p<0.001) and the regression line is steep (z = 0.38). Indeed it is significantly steeper (F = 65.16; d.f. = 1,48; p<0.01) than the butterfly regression (z = 0.24) for the same islands. The strongest correlation between plant species richness and island area is with untransformed data (r^ = 0.87; F = 151.32; p<0.001).

The plant data was grouped into large and small size classes for analysis. The small class contained islands of area smaller than or equal to Guadeloupe, with the large class being all islands bigger than Guadeloupe. As with the butterflies, the species-area regression (log-log plot) is significantly steeper (F = 141.67; d.f. =

1,22; p<0.01) in the large islands (z = 0.57) than in the small islands (z = 0.40).

32 Elevation explains a substantial amount of the residual variation in plant species

richness when area is held constant (r^ = 0.44; F = 18.40; p<0.001). This suggests

that physical heterogeneity may be a more mq)ortant determinant of species

richness in plants than in butterflies. There was no observed latitudinal effect once

the influence of area was held constant.

Butterfly species richness was more strongly correlated with plant richness than

with any physical variable (r^ = 0.84; F = 121.55; p<0.001) (Figure 4). This

suggests that similar forces shape plant and butterfly faunas, although it does not

necessarily imply a causal link. Such a link is, however, plausible as butterflies

depend upon the presence of certain plants as food sources, whilst some plants

may depend on butterflies as pollinators.

There is circumstantial evidence in the West Indies that the absence of suitable

food-plants has precluded the establishment of some butterflies. Papilionid

butterflies of the genus Battus are known to feed exclusively on Aristolochia. In

the West Indies there are three species ofBattus, the most widespread being Battus polydamas which occurs on virtually every island. It does not, however, appear to

be resident on the Cayman Islands, Anegada, Mona, or the southern Bahamas and

the Turks and Caicos Islands, none of which supports Aristolochia. We know that

Battus polydamas is capable of surviving on such small islands (it is found in the

Grenadines for example) and we know that it has reached at least Grand Cayman

(Carpenter and Lewis, 1943) and the Turks and Caicos Islands (St.Leger, 1991).

33 DISCUSSION

Compiling the butterfly list

Regional species numbers are important data for many biogeographical and

ecological studies. In some ways the task is sin^lified on islands because the

region in question is clearly demarcated. Nevertheless, many difficulties, both

biological and artifectual, must be resolved to arrive at a meaningfiil and consistent

list. For a primary reference it is best to include all records and, ideally, the count

would recognise regular or irregular vagrants. These could then be included or

excluded at the choice of individual investigators, depending upon the purpose of

their particular analysis. This level of documentation is achieved, for exanqjle, in

the British butterfly fauna, from records of almost two centuries with constant and

intensive surveying. Several factors must be considered to arrive at figures for the

West Indian island faunas, and I discuss these below. It is important to avoid

ambiguity and to assist others who may take differing views of conq>arabiHty iu

assessing faunal size. All rehable records have therefore been included, regardless

of date and evidence of residence, resisting the temptation to sieve the list which would inevitably introduce a degree of subjectivity.

Doubtful and old records. It is unsurprismg that the collection data associated with

some early museum specimens is unrehable, especially as communications were often difficult between the West Indies and the major sanq)ling agencies of the time

(usually European and American museums or private collections). Furthermore, during the last century major European private collections were assembled largely

34 by purchase and occasionally locality data may have been manufactured to suit the

market. Ramos (1982 and pers. comm) suspects this to be the case for supposed

Puerto Rican specimens oîPhocides pigmalion, Pyrrhocalles anti qua, and

Ephyriades zephodes\ they have been omitted accordingly.

Following Brown and Heineman (1972) eight continental hesperiids {Urbanus

teleus, U.tarma, U. albimargo, Cogia chalcas, Nisoniades bessus, Antigonus

nearchus, Helopetes arsalte, and Oileus fridericus) have also been omitted from

the Jamaican count. These had been attributed to the island on the basis of

specimens in the British Museum (Natural History) which are not acconq)anied by

collecting data and, other than an uncertain record of O. fridericus from Cuba

(Ayalo and Hernandez, 1987), they have never been recorded elsewhere in the

West Indies. O. fridericus has been omitted together with the similarly

unsubstantiated early records of Callimormus radiola, from the Cuban hst (see

Ayalo and Hernandez, 1987).

Scattered through the hterature are examples of island records that are clearly

unintentional errors: Riley’s (1975) illustration of the Hispaniolan endemic

Heraclides machaonides from Puerto Rico, Hall’s (1925) inclusion of the Cuban

Phoebis avellaneda on the Hispaniolan hst, Schaus’ misidentification of the Puerto

Rican Choranthus borincona as Hipaniolan Choranthus haitensis recognised by

Ramos (Smith et al., 1994a) and so on. Although in^ortant for individual island hsts, these corrections are very minor in the overaU context of the statistical treatment of the area’s butterflies.

35 Taxonomie errors and species definitions. On occasion species have been

described as new but later recognised as synonyms of already described taxa. This

is particularly problematic in archipelagos where taxonomists often expect (perhaps

even hope) to find endemic taxa. Holland (1916) described the new hesperiid

species Telegonns geronae and Amblyscirtes insulaepinorum from the Isle of

Pines. These were later recognised to be synonyms for Astraptes cassander and

Euphyes c. Cornelius respectively (Ayalo and Hernandez, 1987). While the

species richness of the Isle of Pines is unaffected by the taxonomic correction,

without it the number of species in the West Indies as a whole would have been

spuriously inflated by two.

Such taxonomic errors are sometimes not really mistakes but a consequence of the

confusion surrounding the term ‘species’. Probably the most widely accepted

species definition is the biological species concept (Mayr, 1942) Wiich enq)hasises the importance of reproductive isolation. The difficulty in applying such a criterion to allopatric populations is obvious (Mallet, 1995). Assigning specific rank to island populations is often subjective because usually we can only guess at levels of reproductive isolation.

Consequently several West Indian butterflies have been described by some authors as subspecies but by others as fiiU species. Smith et al. (1994a) regarded the island populations of Anaea troglodyta and Wallengrenia otho as specifically distinct, although they had previously been considered subspecies (Riley, 1975). Again this

36 does not change the overall species number for individual islands (for example,

W.misera merely replaces W.otho misera in Cuba and W.drury replaces W.otho drury in Hispaniola). It does, however, affect the regional count because there are now two species {W.misera and W.drury) wdiereas previously there was just one

{W.otho).

Subspecies often reflect the intuition of those who describe them as much as their true taxonomic status. Undoubtedly some subspecies will be described as species and vice versa. ‘Lumpers’, taking a conservative view toward spéciation, might underestimate the real species number for a group of islands, whilst ‘splitters’ might tend toward overestimation. This can be a serious problem when data sets are inconq)lete because islands are often grouped e.g. Munroe (1948) and Scott

(1972). I have avoided this source of error by treating islands individually. There are very few instances of two subspecies occurring on the same island, eastern and western populations of the Cuban Papilionid Parides gundlachianus providing a rare exan^le (Hernandez et (%/., 1995).

The species richness figures presented here are only estimates of the number of biological species {sensu Mayr, 1942). Different species definitions could well yield different numbers. Our estimates of species richness are, in fact, measures of phenotypic differentiation: a butterfly diversity index. HopefiiUy they are internally consistent because lepidopterists have developed standards by which they evaluate the level of differentiation appropriate to specific classification. They are not.

37 however, necessarily conq)arable with other such data sets, involving taxa A^diere different criteria may be used to assess species level differentiation.

A conservative approach has been taken in not considering subspecies. Specific classification usually requires a substantial and consistent level of differentiation.

Subspecies, however, are less reliable indicators of independent evolution among islands. Widespread polymorphisms are often mistaken for evidence of separate

subspecies until a long series and especially reared specimens are considered, for exanq)le in Ascia momiste (Ramos and Mieres, 1993).

Differential collecting intensity. Differences in species richness among islands may reflect the number of individuals sanq)led rather than actual differences in species number (Wilhams, 1964). The number of species found on an island is directly related to the time spent collecting. For exanq)le, Holland (1916) pubhshed a list of 65 species from the Isle of Pines, Cuba. In 1975-1976 Hernandez and colleagues added 25 species to the list, and as a result of four visits from 1993-

1995 the species total has now reached 111 (Smithet al, in prep ). Another example is provided by the history of butterfly recording on Mona Island (Puerto

Rico) where the recorded species number has risen from 21 (Ramos, 1946) to 53

(Smith et al, 1994b).

Ideally islands should be intensively and continuously monitored. This has been the case for a long period in the British Isles. Many of the West Indies, however, have been surveyed neither continuously nor intensively. Underestimates of species

38 richness due to infrequent sançling are exacerbated in the tropics vsdiere phonological patterns are far less well understood than in temperate systems. By no means are all West Indian butterflies multivoltine (continuously brooded): some are certainly or probably uni- or bivoltine, with one or two generations a year. The period when adults are flying may be very short and highly variable. This probably depends upon local climatic conditions, but at the moment there is very little data and the flight periods of most species remain unpredictable. Many species will therefore be overlooked during short surveys when some will not be flying. Islands must therefore be sarq>led each month of the year to reduce this potential source of underestimation.

Most of the larger West Indian islands, especially those with resident entomologists, have been quite well documented over many years (Smith et a/.,

1994a). Changes in their known species richness are therefore proportionately small when conq)ared with historically poorly documented islands, such as the Isle of Pines or Mona. Indeed recent work, although relatively extensive, has not added greatly to the species counts of the Greater Antilles. The Cuban list, for example, has received only three additions since 1987: Ministrymon azia (Smith and Hernandez, 1994a), and the newly described Ministrymon hernàndezi and

Leptotes hedgesi (Schwartz and Johnson, 1992). The Jfrspaniolan list has grown mainly through continuing descriptions of new Calisto species. Only one species,

Rhinthon cubana, has been added to the Puerto Rican count during the last few years (Smith et al, 1994a). Of these islands, perhaps Jamaica has benefited most from recent field work, with seven species added (Vhymeister, 1980; Turner and

39 Parnell, 1985; Vane-Wright et al., 1992; Johnson and Smith, 1993; Iftner et al,

1993) since the study of Brown and Heineman (1972).

It should be noted, however, that this does not necessarily inq)ly that all the smaller island estimates are unreliable. Once a small island has been the subject of intensive and prolonged study, as is now the case for Mona, estimates of species richness may be very accurate. Indeed it is probably easier, in practical terms, to establish the species richness of a small island than a large one.

Recognising extinction. A butterfly species may be recorded on an island at one time but then not seen for a considerable period during repeated visits. Do we take this as evidence of extinction? Extinction is extremely hard to prove as ‘it is impossible to be certain that any species is absent at any particular time” (Askew,

1988). Only when there has been continuous and intensive study, along with the careful monitoring of remnant populations, can we be sure that extinction has occurred e.g. the extinction of Maculinea arion in Britain (Thomas and

Lewington, 1991 p. 105). Most studies of West Indian butterflies, however, only sanq)le the adult population and recording is often carried out during sporadic, or even single visits, rather than in the course of long-term surveys. Field observers often miss species that have been found previously on an island (often at the same site and at the same time of year) and yet we cannot tell whether this is because the colony is extinct or because the population is present but at an immature stage

(eggs, larvae or pupae).

40 Nor can we assume the extinction of butterflies which have not been recorded for

generations. A single specimen of the skipper Espargyreus spanna was collected

in Hispaniola in 1855 and the next in 1983 (Gah and Schwartz, 1993). The Cuban

skipper Chiodes marmorosa had not been recorded this century until its

rediscovery in a by no means remote part of the island (Roque et al., in prep).

These species would certainly have been discounted if we had set a statute of

limitations on species records. We must therefore assume that a member of an

incompletely documented fauna is still extant even if it has not been seen for many

years e.g. Atalopedes carteri on New Providence Island in the Bahamas (Smith et

al., 1994a).

What constitutes immigration? If a single individual of a species is found on an

island it is clearly present but should it be added to the list without evidence of a

breeding, resident population? MacArthur and Wilson (1967) recognised this

problem: “is a migrating bird, passing through an island called [an] immigrant? If

so it goes extinct as it leaves”. Their solution was to ignore immigrants that do not

colonise with the potential of reproducing. Simberlofif ( 1969) clarified this

definition of immigration as “the arrival of a propagule on an island unoccupied by that species” a propagule being “the minimum number of individuals of a species

capable of breeding and population increase, under ideal conditions for that

species”. For butterflies a single gravid female is capable of breeding, so any record of a female butterfly should be scored as an immigrant.

41 Establishing that there is actually a breeding population on an island is often much more difficult and for museum specimens it is inq)ossible. If we required proof of such breeding status then the species lists would be very small indeed. There is another problem with any breeding criterion: how do we score species that regularly colonise islands but then go extinct after a couple of generations, only to reappear a few years later as another tençorary colony is estabhshed? For example, Hamadryas amphinome mexicana was common in western Cuba in

1930, then disappeared until it was found again in 1976. An arrival may be unsuccessful on one occasion but succeed the next; colonisation may even be transient in one region of an island but long-lasting in another. Within the last three decades butterflies reaching south Florida jfrom Cuba and/or the Bahamas show a spectrum of colonisation success fi'om estabhshed breeding populations to colonies surviving only very briefly (Smith et al., 1994a).

In island feunas, species may commonly range from permanent breeding populations to single generation vagrant colonies. With long-term and detailed surveying such metapopulations could be categorised, however, this remains a distant goal in the West Indies. At the moment colonisation success is more readily recognised than failure.

How accurate are the hsts? Ah rehable species records are included, on the assumption that no species has gone extinct. This inevitably risks overestimating the true species number because vagrants are scored as residents and extinct species are included. This error is reduced, however, by the number of species

42 overlooked. This is probably a more significant source of error. It is therefore likely that the species numbers are still minimum estimates, although as field recording continues we are able to use the lists with growing confidence.

Species-area relationship

The species-area relationship was explained by Preston (1962) as a stochastic sampling phenomenon based on the distribution of abundance into species. For island faunas at equilibrium the q)ecies-area regression is expected to give the best linear fit on a log-log plot with a slope of 0.15-0.39 (MacArthur and Wilson, 1967;

May, 1975). If West Indian faunas are built stochastically, and if they are at equilibrium, we expect taxa to fit this pattern. Wilson (1988) found support for this among West Indian ants. He suggested that although future research might augment species numbers, the slope (z = 0.28) would probably remain fairly constant whilst the intercept would increase. The evidence from West Indian butterflies is also consistent with a stochastic equilibrium model and supports

Wilson’s (1988) prediction for the future stability of the ant species-area relationship. The accumulation of butterfly data has flattened the slope only slightly, whilst the intercept has indeed risen. It is interesting to note that linearity is now best obtained by the power function, whereas in 1948 it was the e?q)onential function which gave the best fit (Munroe, 1948; Table 3). The exponential fimction is expected to give the best fit when the species-area relationship merely reflects san^ling properties, with area being a measure of samq)le size (May, 1975 p.85). The progression from an exponential function to a power function is therefore to be expected as a data set becomes more complete.

43 The use of the species-area relationship to infer rules of Aunal assembly, however, has been criticised by Connor and McCoy (1979) who point out that a slope of 0.2

- 0.4 is more parsimoniously explained by the characteristics of the regression system Such a narrow range is often found wften there are high correlation coefficients and a small range in the dependent variable relative to the independent variable. Connor and McCoy (1979) suggest that only slope values deviating from this range are biologically significant. Our results show that in the Greater Antilles the slope approaches the upper limit of this range (z = 0.37), in the Lesser Antilles it is towards the lower boundary (z = 0.20), and in faunas of less than 20 species it is well below it (z = 0.06). Furthermore there is quite a lot of scatter around the species-area regression lines. This indicates that population size and extinction rate do not fully explain the richness of these butterfly faunas.

The most obvious reason why the null model should breakdown is the influence of isolation (MacArthur and Wilson, 1963, 1967). Increased isolation is expected to reduce immigration rates and hence pull species richness below the species-area regression line. There are some obvious cases where high immigration rates might be inflating species richness e.g. the Isle of Pines and Cayo Coco off the coast of

Cuba, Lignumvitae Key off Florida, and Mona between Puerto Rico and

Hispaniola. There are fewer examples where isolation might be expected to depress species richness, e.g. Barbados. However, the results showed that there is no overall distance effect on the species richness of West Indian butterfly faunas.

44 Instead the data suggest that differences in habitat diversity probably explain much

of the scatter around the regression line. But why should the slope be shallower in

the small feunas and the Lesser Antilles conçared with the Greater Antilles? Three

hypotheses can explain divergent z-values among island groups: faunas may be

closed, immigration rates may differ, or the fauna may not be at equihbrium

Closed faunas

Given the relative richness of the mainland butterfly fauna, it seems unlikely that

interspecific conyetition constrains richness in the West Indies. It is hard to see how anything other than historical biogeography could explain why the larger and physically more heterogeneous Hispaniola should have far fewer butterflies (about two hundred) than Trinidad (over six hundred).

Terborgh and Faaborg (1980) in their study of West Indian birds, however, proposed that ‘conq)etitive abihty is a coevolved property of regional faunas’.

Conqjetition can therefore close island faunas, despite the fact that they are

depauperate in comparison with the mainland fauna, and even when many of the

species involved are the same. Terborgh and Faaborg (1980) argue that insular bird faunas have a lower tolerance to conq)etition because of the ecological release associated with colonisation. Species expand to fill enq)ty niches on islands (Cox and Ricklefs, 1977) and so new immigrants are increasingly likely to find their niche occupied. Island habitats may therefore become saturated at lower levels of species richness than their continental counterparts.

45 We do not have sufiBcient information on the ecology of West Indian (or mainland) butterflies to determine whether niche width is broader in the islands. Nor can we assess vdiether habitats are saturated as we do not have adequate data on habitat utiHsation. The bird data, however, is at a more advanced stage and suggests that interspecific competition may be inq)ortant. In the West Indies the number of birds found in any given area of similar habitat is much lower than on the mainland, but remarkably constant among the islands, even the small ones. This appears to be due to competition and not the exhaustion of potential colonists (Terborgh and

Faaborg, 1980).

Saturation might also occur on small islands because of ecological impoverishment

(Lack, 1976). Small islands tend to support fewer habitats than large islands.

Take the extreme case that small islands have only one habitat, say xeric scrub.

Species richness is then limited to those species Wiich are capable of surviving in xeric scrub. Xeric scrub species might have very low (effectively zero) extinction rates on a xeric scrub island, whilst the extinction rate for other species might be effectively infinite. Once all the xeric scrub species have colonised small islands species richness will therefore remain constant, irrespective of island area. Again we need more ecological data on butterfly habitat utilisation to assess this hypothesis in the West Indies.

Immigration rates

Immigration rate increases with the proximity of an island to the mainland. Minno and Emmel (1993) found that island area had no effect on species richness in the

46 gmall islands of the Florida Keys. MacArthur and Wilson (1967) predicted flat regression lines in such proximate archÿelagos because they include species unable to maintain a resident population, so called ‘sink’ species (Shmida and EUner,

1984; Pulham, 1988). This inches that the immigration rate affects the extinction

rate. Species experiencing a critical trough in population size can be saved by the

arrival of conspecifics. This is known as the rescue effect (Brown and Kodric-

Brown, 1977). The Florida Keys are indeed relatively species-rich for such small

islands. They are more like continental islands, supporting many sink species and

having flat species-area regression lines (Rosenweig, 1995).

High immigration rates therefore reduce the inq)ortance of extinction through the

rescue effect, and hence dilute the influence of area. This is clearly demonstrated by endemic species. Endemics presumably have low dispersal abihty; the rescue

effect should be largely irrelevant. Island area wih therefore be strongly indicative

of endemic species richness because of its important influence on population size

and hence extinction. Indeed the species-area regression is typicaUy steeper for

endemic species than for whole faunas e.g. West Indian butterflies (Scott, 1972)

and ants (Wilson, 1988).

Curiously, increased isolation can also flatten species-area regression lines.

Isolated archipelagos have flatter species-area regression lines than less isolated

archipelagos (Connor and McCoy, 1979). At first this might seem to contradict the above conclusions. However, isolated archipelagos are only likely to be reached by good dispersers. This means that, on average, isolated archipelagos

47 have more vagile faunas and hence high rates of inter-island immigration. Within

the archipelago islands are less isolated and therefore species-area regression lines

will be flat.

Small West Indian islands may be more isolated because they represent a smaller

target for dispersing butterflies. Sedentary butterflies are therefore less likely to

reach small islands and so their faunas will tend to be more vagile (on average)

than a random sanqile of species in the source fauna. Even wdien sedentary species

do manage to become established on a small island they will have a higher

probability of extinction because of the limits on population size. Extinction is

more of a threat on small islands and sedentary species are less likely to be rescued

by arriving conspeciflcs. Small islands will therefore quickly lose sedentary

species. Large islands, on the other hand, are easier to reach and once estabhshed

sedentary species are less hkely to go extinct because large population sizes can be

attained. Large islands will therefore accumulate proportionately more sedentary

faunas.

An alternative, or complementary, explanation might be that species adapted to

small island habitats (xeric scrub for example) are singly better than average

dispersers. Species in different habitats acquire ecological adaptations which allow them to succeed in those habitats. Some of these characters may affect dispersal

abihty, either by ‘design’ or as a by-product of unrelated adaptation: dispersal

abihty might therefore vary among species according to their habitat types. It is

48 possible that the few habitats of a small island might support good dispersers

compared with the average across the many habitats of a large island.

Wilson (1961) proposed that colonising species tend to be adapted to coastal

habitats, which for small islands is usually the dominant habitat type. Such species

are thought to be good dispersers. When these species colonise larger islands

selection acts as populations move to fill enq)ty niches, particularly montane

forests. This ecological shift is thought to coincidentally reduce their dispersal

ability and accelerate their extinction. The ‘taxon cycle’ begins again as new

colonising species arrive on the island. This hypothesis is still very speculative

(even in well studied groups like birds) and again it cannot be tested for West

Indian butterflies without more knowledge of their ecology and historical

biogeography.

Equilibrium

A long-running controversy in West Indian biogeography is whether the fauna is a

result of vicariance (Schuchert, 1935; Rosen, 1976, 1985) or dispersal (Sinçson,

1952; Darlington, 1957). Are West Indian faunas remnants of a continental fauna

reduced by extinction as the islands drifted into the Caribbean? Or is it a truly

insular fauna, built by dispersal from the continent? The geological evidence

suggests that these plate-tectonic events could only be a factor in the large islands

of the Greater Antilles and Bahamas. Whilst some of these islands may have been variously linked to one another and/or the mainland, the small islands (including the

Lesser Antilles) are probably oceanic (Wilhams, 1989).

49 The Greater Antilles may therefore be supersaturated, retaining elements of a continental fauna which is now ‘relaxing’ to an equilibrium island level. Some have suggested that plate-tectonic events may be inqjortant in the biogeography of

Greater Antillean butterflies (Miller and Müler, 1989). The remnants of a continental 6una are likely to include sedentary species incapable of reaching oceanic islands. This therefore provides an alternative means by which Greater

Antillean butterflies could, on average, be more sedentary than those in the Lesser

Antilles.

The geological history of the West Indies is complex and controversial but most agree that the islands have been in their current positions for at least 5 million years

(Smith et al., 1994a). I suspect that relaxation times are probably considerably shorter than 5 million years for most butterfly taxa, even though some elements of the fauna may be considerably older e.g. Calisto. Any influence of such ancient vicariance would therefore be limited to a small number of species with exceptionally low extinction rates and colonisation abilities.

In fact the small island may be better candidates for vicariance. Although oceanic, some small islands may have been larger and/or linked during the glacial maxima of the Pleistocene. This might mean that some of the small islands are now supersaturated. Wilcox (1978) demonstrated that time since isolation has an important influence on the number of lizards on post-Pleistocene land-bridge islands off California. In the West Indies this effect will be restricted mainly to the

50 Bahamas, Virgin Islands, some of the northern Lesser Antilles and the Grenadines.

Miller et al. (1992) invoked such events in the biogeography of some Bahamian butterflies. Unfortunately it is unclear just how much land was exposed and for how long; it is therefore diflScult to quantify the importance of this effect.

Furthermore many of the islands potentially involved do not seem to be particularly species-rich.

The contribution of vicariance (both ‘ancient’ and Pleistocene) to the overall species richness of West Indian butterflies remains an open question which is perhaps best resolved through genetic analyses (Chapters 3 and 4).

Summary

Definite conclusions as to how insular assemblages are constructed cannot be drawn fi-om the species-area relationship alone. It is, however, a useful departure point for future ecological and evolutionary study. Area itself is an important factor, probably through a combination of habitat diversity and population size.

Dispersal also seems inq)ortant despite the fact that the isolation of an island has little direct influence on its species richness. Immigration rates may depend on island area and this could cause scale dependent variation in the slope of the species-area relationship. Conqjetition might also cause heterogeneity in slope.

More work is clearly needed on the ecology and historical biogeography of individual butterfly species. In particular, a matrix of individual species distributions will allow an analysis of structure, including orderedness (Ryti and

51 Gilpin, 1987) and nestedness (Patterson and Atmar, 1986; Cutler, 1991; Kadmon,

1995). It is possible that the species-area relationship does not resuh from mere

random extinction determined by population size. The presence or absence of

species on islands may be determined by the ecological characteristics of each

species. Extinction may be selective, or in some other way deterministic, not

random

There is some evidence of structure in the distribution of West Indian butterflies.

Cuba and Hispaniola have two species of Anartia and Battus, but there is only a

single species of these genera on the smaller islands. However, some distributions

appear unstructured. Of the three insular Libytheid species in the West Indies, one

occurs on Cuba, one on Hispaniola, Jamaica and Puerto Rico, and the other is

endemic to Dominica. Similarly, Jamaica and Puerto Rico have only one species of

Satyrid, but so does tiny Anegada. A treatment of the entire fauna is necessary to

determine the extent and nature of any structure, and only then might we be in a

position to understand the species-area relationship more fully.

BIBLIOGRAPHY

Adams, C D 1972. Flowering plants of Jamaica. University of the West Indies, Mona, Jamaica.

Albelo, L.K, Hernandez, L.K, and Smith, D.S. 1995. Rediscovery of Chiodes marmorosa in Cuba (Lepidoptera: Hesperiidae). Trop. Lep. 6:99-102.

52 Askew, KR. 1988. Butterflies of Grand Cayman, a dynamic island faima. J. Nat. Hist. 22:875-881.

Arrhenius, O. 1921. Species and area. J.Ecol. 9:95-99.

Alayo, P.D. and Hernandez, L.R. 1987. Atlas de las mariposas diumas de Cuba (Lepidoptera: Rhopalocera). Editorial Cientifico-Tecnica, Ciudad de la Habana.

Barcant, M. 1970. Butterflies of Trinidad and Tobago. Collins, London.

Becker, V O. and Miller, S.E. 1992. The butterflies of Guana Island, British Vbgin Islands. Bull. Allyn Mus., No. 136, 1-9.

Begon, M., Harper, J.L., and Townsend, C.R. 1986. Ecology: individuals, populations and communities. Blackwell, pp.876.

Bermingbam, E. and Avise, J.C. 1986. Molecular zoogeography of ifresbwater fishes in the southeastern United States. Genetics 113:939-965.

Brown, F.M. and Heineman, B. 1972. Jamaica and its butterflies. Classey, London.

Brown, J.H. and Kodric-Brown, A. 1977. Turnover rates in insular biogeograpby: effect of immigration on extinction. Ecology 58:445-449.

Brown, J.H. and Lomolino, M.V. 1969. Independent discovery of the equibbrium theory of island biogeograpby. Ecology 70:1954-1957.

Carpenter, G.D.H. and Lewis, C.B. 1943. A collection of Lepidoptera (Rhopalocera) fi'om the Cayman Islands. Annals Carnegie Mus. 29:371-396.

53 Clench, H .K 1977. A list of butterflies of Andros, Bahamas. Ann. Carnegie Mus. 46:173-94.

Clench, H.K. and Bjomdal, K.A. 1980. Butterflies of Great and Little Inagua, Bahamas. Ann. Carnegie Mus. 49:1-30.

Cody, M L. 1989. Discussion: Structure and assembly of communities. In: Perspectives in ecological theory. Roughgarden, J., May, KM. and Levin, S.A. (eds) Princeton University Press, pp.227-242.

Connor, E.F. and McCoy, E D. 1970. The statistics and biology of the species- area relationship. Amer. Nat. 113:791-833.

Correfl, D E and CorreU, H.B. 1982. Flora of the Bahamas Archipelago (including the Turks and Caicos Islands). Cramer, Vaduz.

Cox, G.W. and Ricklefs, KE. 1977. Species diversity, ecological release, and community structure in Caribbean land bird faunas. Oikos 29:675-678.

Cutler, A. 1991. Nested 6unas and extinction in fi’agmented habitats. Conservation Biology 5:496-505.

Darlington, P.J. 1943. Carabidae of mountains and islands: data on the evolution of isolated faunas, and on atrophy of wings. Ecol. Mongr. 13:37-61.

. 1957. Zoogeography. John Wiley and Sons, New York.

D’Arcy, W.G. 1975. Anegada Island: vegetation and flora. Atoll Res. Bull. Smithsonian Inst., Washington D C , No. 188, 1-40.

Davies, N. 1989. Oxford University Grenadines Expedition: Final Report. Oxford University, Oxford.

54 Davis, S.D., Droop, S.J.M., Gregerson, P., Henson, L., Leon, C.J., Villa-Lobos et al. 1986. Plants in danger. What do we know? IUCN, Gland and Cambridge.

Diamond, J.M. 1972. Biogeographic kinetics: estimation of relaxation times for avifaunas of southwest Pacific islands. Proc. Natl Acad. Sci. U.S.A. 69:3199- 3203.

Diamond, J.M. and May, RM. 1976. Island biogeography and the design of natural reseives. In Theoretical Ecology: Principles and Applications. Blackwell pp.317.

Diamond, J.M. and Mayr, E. 1976. The species-area relation for birds of the Solomon Archipelago. Proc. Natl. Acad. Sci. U.S.A. 73:262-266.

Elliott, N.B., D. Riley and Clench, HK 1980. Annotated list of the butterflies of San Salvador Island, Bahamas. J. Lepid. Soc. 34: 120-6.

Elton, C. 1933. The Ecology of Animals. Reprinted 1966 by Science Paperbacks and Methuen, London.

Evans, F.C. 1973. The West Indies. Cambridge University Press, Columbus Publications, Trinidad.

Fischer, A.G. 1961. Latitudinal variations in organic diversity. Amer. Sci. 49:50- 74.

Gali, F. and Schwartz, A. 1983. The second specimen ofEpargyreus spanna (Hesperiidae). J. Lep. Soc. 37:170-171.

Gaud, S.M. and L.F. Martorrel. 1974. The insects ofCaja de Muertos Island, Puerto Rico. J. Agric. Univ. Puerto Rico. 58:244-78.

55 Gilbert, L.E. 1984. The biology of butterfly communities. In: The Biology of Butterflies, Symposium of the Royal Entomological Society of London, No. 11. Vane-Wright, KI. and Ackery, P R (eds). Academic Press, London, pp.41-54.

Gleason, H.A. 1922. On the relation between species and area. Ecology 3:158- 162.

Gleason, H.A. 1926. The individualistic concept of the plant association. Bull. Torrey Bot. Club 53:1-20.

Gort, S.A., Escobar, T., Izquierdo, J., Correoso, M. and Singh, N. 1994. Isla de Lesser Antilles Juventud. Su naturaleza. Editorial Cientifico-Tecnica, Ciudad de la Habana.

Hall, A. 1925. List of butterflies of Hispaniola. Entomologist, 58:161-165, 171- 174.

Hamilton, T.H., Barth, RH . and Rubinoff^ I. 1964. The environmental control of insular variation in bird species abundance. Proc. Nat. Acad. Sci. Philadelphia 105:1-12.

Harvey, D.J. and Peacock, J.W. 1989. New Records of Butterflies fi'om North Andros, Bahamas. Ent. News. 100:86-8.

Hernandez, L.R., Alayon Garcia, G. and Smith, D.S. 1995. A new subspecies of Parides gundlachianus fi'om Cuba (Lepidoptera: Papilionidae). Trop. Lep. 6:15-

20.

Holland, W.J. 1916. The Lepidoptera of the Isle of Pines, being a list of the species collected on the island by Mr. J.L. Graf and Mr. G.A. Link, Sr, in 1910 and 1912-1913. Ann. Carnegie Mus. 10:487-518.

56 Hutchinson, G.E. 1959. Homage to Santa Rosalia, or vAiy there are so many kinds of animals? Amer. Natur. 93:145-159.

Hunt, D. and Mitchell, G. 1979. A recognition guide to the insects of St. Lucia. 1. Butterflies (LepiHesperioidea & Papihonoidea). WINBAN and the St. Lucia Naturahsts' Society.

Hunter, B. (ed.) 1994. The Statesman’s Year-Book. Statistical and Historical Annual of the States of The World for the Year 1994-1995. 131st edit. Macmillan, London.

Iflner, D C , Shuey, J.A. and Calhoun, J.V. 1993. Hemiargus ammon (Lucas), a new butterfly for Jamaica. Trop. Lep. 4:37-38.

Jennings, O.E. 1917. A contribution to the botany of the Isle of Pines, Cuba, based on the specimens of plants fi’om that island contained in the herbarium of the Carnegie Museum under date of October, 1916. Ann. Carnegie Mus. 11:19-290.

Johnson, K. and Smith, D.S. 1993. A remarkable new butterfly species fi^om Jamaica (Lepidoptera: Lycaenidae), with notes on Jamaican endemics and their sister species. Rep. Mus. Nat. Hist. Univ. Wisconsin, Stevens Point. 24:1-14.

Kadmon, R 1995. Nested species subsets and geographic isolation: a case study. Ecology 76:458-465.

Knowles, D O. and D.S. Smith. 1995. of Parachoranthns magdalia (Hesperiidae) fi'om the Bahamas, and extension of the Bahamian range of Battus (Papihonidae). J. Lep. Soc. 49:91-4.

Lack, D. 1976. Island Biology, Illustrated by the Land Birds of Jamaica. Blackwell.

57 Léon, H and Alain, H. [1946-1963]. Flora de Cuba. Vols. 1-4 reprint (1974) Koeltz, Koenigstein. Vol. 5 (ed. Alain, H.) (1962) Editorial Universitaria Puerto Rico, Rio Piedras.

Leston, D , Smith, D.S. and Lenczewski, B. 1982. Habitat, diversity and immigration in a tropical island fauna: the butterflies of Lignumvitae Key, Florida. J. Lepid. Soc. 36:241-55.

Liogier, A H and Martorrel, L.F. 1982. Flora of Puerto Rico and adjacent islands; a systematic synopsis. Editorial Universitaria Puerto Rico, Rio Piedras.

MacArthur, R.H and Wilson E.O. 1963. An equilibrium theory of insular zoogeography. Evolution 17:373-387.

______. 1967. The Theory of Island Biogeography. Princeton University Press, NJ.

Mallet, J. 1995. Species definition for the modem synthesis. Trends Ecol. Evol. 10, in press.

May, KM. 1975. Patterns of species abundance and diversity. In Ecology and Evolution of Communities. Cody, M L. and Diamond, J.M. (eds) Belknap Press of Harvard University Press pp. 81-120.

Mayr, E. 1942. Systematics and the Origin of Species. Columbia University Press, New York.

Miller L.D. and Miller J.Y. 1989. The biogeography of West Indian butterflies. In: Biogeography of the West Indies. Woods, C A. (ed). Sandhill Crane Press, Inc. Gainesville FL. pp. 229-262.

58 Miller, L.D. and Steinhauser, S. 1992. Butterflies of the Cayman Islands, with the description of a new subspecies. J. Lepid. Soc. 46:119-27.

Miller, L.D., Simon, M.J. and Harvey, D.J. 1992. The butterflies (Insecta: Lepidoptera) of Crooked, Acklins, and Mayaguana Islands, Bahamas, with a discussion of the biogeographical aflSnities of the southwestern Bahamas and a description of a new subspecies by H K. Clench. Ann. Carnegie Mus. 61:1-31.

Minno, M.C. and Emmel, T.C. 1993. Butterflies of the Florida Keys. Scientific Publishers, Gainesville.

Miskimen, G.W. and R.M. Bond. 1970. N.Y. Acad. Sci. Scientific Survey of Porto Rico and the Virgin Islands. 13:54-8.

Munroe, E.G. 1948 . The Geographical Distribution of Butterflies in the West Indies. Dissertation, Cornell University, Ithaca, NY.

. 1953. The size of island feunas. 7th Pacific Sci. Congr. 4:52-53 (Abstract)

Patterson, B.D. and Atmar, U. 1986. Nested subsets and the structure of insular mammahan faunas and archipelagos. Biol. J. Linn. Soc. 28:65-82.

Pearce, E.J. 1969. An attençted re-appraisal of the butterflies of Barbados, with reference to certain weather phenomena. J. Barbados Mus. Nat. Hist. Soc. 33:76- 84.

Pinchon, P R. and Enrico, P. 1969. Faune des Antilles Françaises, Les Papillons. M.M. Ozanne et Cie , Fort de France, Martinique.

Power, DM . 1971. Numbers of bird species on the Californian Islands. Evolution 26:451-463.

59 Preston, F W 1960. Time and space and the variation of species. Ecology 41:785-790.

. 1962. The canonical distribution of commonness and rarity. Ecology 43:185-215,410-432.

Proctor, G.K 1984. Flora of the Cayman Islands. Kew Bulletin Additional Series XI.

PulHam, H.R. 1988. Sources, sinks and population regulation. Amer. Natur. 132:652-661.

Ramos, J.A. 1946. The insects of Mona Island (West Indies). J. Agric. Univ. Puerto Rico 30:1-74.

Ramos, S.J. 1982. Checklist of the butterflies of Puerto Rico (Lepidoptera, Rhopalocera, West Indies). Caribb. J. Sci. 17:59-68.

Ramos, S.J. and Mieres, D.I. 1993. The biology of Ascia monuste in Puerto Rico. 1. The taxonomic situation. Caribb. J. Sci. 29:4-11.

Ricklefs, R.E. 1987. Community diversity: relative roles of local and regional processes. Science 235:161-171.

Riley, N.D. 1975. A field guide to the butterflies of the West Indies. Collins, London, pp. 224

Rosen, D.E. 1976. A vicariance model of Caribbean biogeograpby. Syst. Zool. 24:431-464.

60 . 1985. Geological hierarchies and biogeographic congruence in the Caribbean. Missouri Bot. Gard. 71:636-659.

Rosenweig, M.L. 1995. Species Diversity in Space and Time. Cambridge University Press.

Roughgarden, J.M. 1989. The structure and assembly of communities. In: Perspectives in Ecological Theory. Roughgarden, J., May, RM . and Levin, S.A. (eds). Princeton University Press, pp. 203-227.

RusseU, A.B. 1992. Butterflies of Barbados 1988-1991. J. Barbados Mus. Nat. Hist. Soc. 40:48-54.

Ryti, RT. and Gilpin, M E. 1987. The con^arative analysis of q>ecies occurrence patterns in archipelagos. Oecologia 73:282-287.

St.Leger, RG.T. 1991. The butterflies and hawkmoths of the Turks and Caicos Islands. Bull. Amat. Ent. Soc., June 1991, 114-120.

Schuchert, G. 1935. Historical geology of the Antillean-Caiibbean region or lands bordering the Gulf of Mexico and the Caribbean Sea. John Wiley & Sons, New York, 811 pp.

Schwartz, A. 1989. The butterflies of Hispaniola. Univ. of Florida Press, Gainesville.

. 1990. Battus polydamas (Papihonidae) on Barbados, West Indies. J. Lepid. Soc. 44:290-1.

Schwartz, A. and Henderson, RW. 1990. The butterflies of Barbuda, West Indies. Caribb. J. Sci. 26:98-100.

61 Schwartz, A. and Johnson, K. 1992. Two new butterflies (Lepidoptera: Lycaenidae) fi'om Cuba. Caribb. J. Sci. 28:149-157.

Schwartz, A. and Jimenez, C.J. 1982. The butterflies of Montserrat, West Indies. BuU. Allyn Mus. No. 66, 1-18.

Schwartz, A., Gonzalez, F.L. and Henderson, KM. 1987. New records of butterflies fiomtbe West Indies. J Lepid. Soc. 41:145-50.

Scott, J.A. 1972. Biogeograpby of the Antillean butterflies. Biotropica 4:32-45.

. 1986. Distribution of Caribbean butterflies. Papilio, New Series 3:1-26.

Sbmida, A. and Ellner, S. 1984. Coexistance of plant species with similar niches. Vegatio 58:29-55.

Simberlofi) D.S. 1969. Experimental zoogeography of islands: a model for insular colonisation. Ecology 50:296-314.

Simon, M.D. and Miller, L.D. 1986. Observations on the butterflies of Great Inagua Island, Bahamas, with records of three species new to the islands. Bull. Allyn Mus. No. 105, 1-14.

Simpson, G.C. 1952. Probabibties of dispersal in geologic time. Bull. Am. Nat. Nat. Hist. 99:163-176.

______. 1964. Species density of North American recent mammals. Syst. Zool. 13:57-73.

Smith, D.S. and Hernandez, L.K 1992. New subspecies of Pseudochrysops bornoi (Lycaenidae) and Saliana esperi (Hesperiidae) fiom Cuba, with a new island record and observations on other butterflies. Caribb. J. Sci. 28:139-148.

62 Smith, D.S., Classey, E.W. and Ramos, S.J. 1989. Appiaspunctifera d’Almeida (Pieridae) in the Dominican Repubhc and Puerto Rico. J. Lepid. Soc. 43:333-6.

Smith, D.S., Hernandez, L.R. and Davies, N. The butterflies of the Isle of Pines, Cuba: eighty years on. Bull. Carnegie Mus. Nat. Hist, (submitted).

Smith, D.S., Miller, L.D. and McKenzie, F. 1991. The butterflies of Anegada, British Virgin Islands, with descriptions of a new Calisto (Satyridae) and a new Copaeodes (Hesperiidae) endemic to the island. Bull. Allyn Mus. No. 133, 1-25.

Smith, D.S., Miller L., and Miller, J. 1994a. The Butterflies of the West Indies and South Florida. Oxford University Press, Oxford.

Smith, D.S., Ramos, S.J., McKenzie, F. 1994b. The butterflies of Mona Island (Puerto Rico) and an approach to their origins and phenology. Caribbean J. Sci. 30:95-103.

Terborgh, J. 1973. On the notion of favorableness in plant ecology. Amer. Natur. 107:481-501.

Terborgh, J. and Faaborg, J. 1980. Saturation of bird communities in the West Indies. Amer. Natur. 116:178-195.

Thomas, J. and Lewington, R. 1991. The butterflies of Britain and Ireland. Dorling Kindersley, London.

Turner, T.W. and Parnell, JR. 1985. The identification of two species o f Junonia Hiibner (Lepidoptera: Nynqjhahdae): J. evarete and J. genoveva in Jamaica. J. Res. Lep. 24:142-153.

63 Vane-Wright, KL, Ackery, P.K and Turner, T.W. 1992. Anetiajaegeri, Danaus cleophile and Lycorea cleobaea from Jamaica (Nynçhalidae: Danainae). J. Lepid. Soc. 46:273-279.

Vhymeister, G. 1980. Tmolus azia in Jamaica: a new record for the West Indies (Lycaenidae). J. Lepid. Soc. 34:60.

West, B.K 1966. Butterflies of New Providence Island, Bahamas. Ent. Rec. 78:174-9, 206-10.

Whittaker, KJ. 1992. Stochasticism and determinism in island ecology. J. Biogeogr. 19:587-591.

Wilcox, B.A. 1978. Supersaturated island faunas: a species-age relationship for lizards on post-Pleistocene land-bridge islands. Science 199:996-998.

Wilkinson, D.M. 1993 . Equilibrium island biogeography: its independent invention and the marketing of scientific theories. Global Ecol. and Biogeog. 3:65-66.

Wilson, E.O. 1961. The nature of the taxon cycle in the Melanesian ant fauna. Amer. Natur. 95:169-193.

. 1988. The biogeography of the West Indian Ants (Hymenoptera: Formicidae). In Zoogeography of Caribbean Insects. Liebherr, J.K (ed.) Cornell University Press, Ithaca N.Y.

Williams, C.B. 1964. Patterns in the Balance of Nature. Academic Press, London.

64 Williams, E E 1989. Old problems and new opportunities in West Indian biogeography. In Biogeograpby of the West Indies. Sandhill Crane Press, Inc. Gainesville FL. Woods, G.A. (Ed), pp. 1-46

WilUamson, M.H. 1981, Island Populations. Oxford University Press.

Woodbury, R.C , Martorrel, L.F. and Garcia- Tudurri, J G 1977. The flora of Mona and Monito Islands, Puerto Rico (West Indies). Agric. Exptl. Stn., Rio Piedras, Puerto Rico, Bull. No. 252, 1-60.

65 Table 1. Butterfly species richness and island data.

Island Area Elevation Isolation Latitude Class Species Refs. (Km*) (m) (miles) Acklins (Bah) 497 36 100 22*20' 28 1 Andros (Bah) 5957 36 97 24*35' 1 66 2.3,4,5 Anegada (B.V.I.) 39 9 61 18*45' 24 6 Anguilla 91 61 4 18*12' 2.3 11 7 Antigua 280 402 240 17*03' 2 31 8.9 Barbados 430 340 79 13*09' 2 25 10,11.12 Barbuda 161 62 47 17*36' 2,3 16 43 Bequia (St.V.&Gr.) 18 268 4 13*01' 2 21 13,14 Biminis (Bah) 22 10 45 25*45' 3 20 8 Caille (Gr.) 1 74 4 12*17 2.3 11 13 Caja de Muertos (P.R.) 1.6 70 5 17*54' 3 11 15 Canouan (St.V.&Gr.) 7 260 23 12*43' 2.3 10 13 Carriacou (Gr.) 34 299 14 12*28' 2 23 13 Cat (Bah) 388 63 108 24*15' 25 8,16 Cayman Brae 36 43 105 19*4' 30 17,18 Cayo Coco (Cuba) 370 5 10 22*30' 50 42 Cayo Lobo (P.R.) 0.3 41 2 18*20' 3 8 19 Cayo Luis Pena (P.R.) 1.3 146 0.5 18*20' 3 16 19 Cayo Norte (P.R.) 1.4 104 1 18*21' 3 12 19 Crooked (Bah) 181 47 124 22*45' 36 1 Cuba 108660 1994 106 20*15' 1 182 20,21,22 Culebra (P.R.) 26 198 16 18*20' 30 19 Culebrita (P.R.) 1 94 22 18*20' 3 18 19 Dominica 751 1447 265 15*31' 2 52 8,23 Eleuthera (Bah) 518 18 68 25* 38 8,24 Florida (Dade/Munroe Co. 8500 6 385 25*50' 1 110 7 Grand Bahama (Bah) 1373 6 58 26*37 23 8 Grand Cayman 197 150 155 19*20' 46 17.18,25 Grand Turk 18 20 95 21*30' 28 26 Grenada 345 840 75 12*08' 2 47 7.8,27 Gt Abaco (Bah) 1681 41 60 26*21' 1 26 8 Gt Exuma (Bah) 210 38 84 23*38' 33 8.24 Gt Inagua (Bah) 1500 45 46 21*50' 1 37 28,29 Guadeloupe 1438 1480 270 16*16' 2 44 27 Guana Island (B.V.I.) 2.8 266 65 18*30' 31 30 Hispaniola 76190 3087 310 18*30' 1 202 7,31,32 lies des Saintes (Guad.) 13 309 6 15*51' 2.3 11 27 Isle of Pines (Cuba) 2200 310 203 21*30' 1 111 33 Jamaica 11424 2255 335 18*05' 1 126 34-37 La Desirade (Guad.) 20 278 5 16*19' 2.3 10 27 Lignumvitae (FL) 1 5 15 24*45' 41 38,39 Little Cayman 26 20 118 19*40' 23 17 Little Inagua (Bah) 127 18 92 21*30' 3 19 28 Long (Bah) 448 54 115 23*15' 31 8 Marie Galante (Guad.) 158 205 14 15*56' 2.3 17 27

66 Table 1. Butterfly species richness and island data.

Island Area Elevation Isolation Latitude Class Species Refe. (Km*) (m) (miles) Martinique 1079 1397 213 14°31' 2 38 27 Mayaguana (Bah) 285 40 135 22°20' 24 1 Mayreau (St.V.&Gr.) 3 108 27 12°39' 2,3 15 13 Mona (P.R.) 62 85 32 18°05' 53 40 Montserrat 106 914 210 16°43' 2 39 41 Mustique (St.V.&Gr.) 5 151 13 12°52' 2,3 16 13 Nevis 93 985 2 17“08* 2,3 16 8,17 New Providence (Bah) 207 370 175 25°05' 60 4,7,44,45 North Caicos 121 30 98 21»47 28 26 Palm (StV.&Gr.) 0.5 50 1 12°35' 2,3 9 13 Puerto Rico 8866 1341 384 18°21' 1 97 8,46 Rum Cay (Bah) 78 30 140 23"42' 26 47 Saba (D.A.) 13 880 25 17®38’ 2 22 8,17 San Salvador (Bah) 163 43 160 24»05' 48 8,24,48 St Croix (U.S.V.l.) 212 355 49 17*46' 41 49 St John (U.S.V.l.) 53 389 46 18*22' 24 8 St Kitts 168 1156 162 17*21’ 2 42 7,8,50 St Lucia 617 950 173 13*52' 2 48 51 St Thomas (U.S.V.I.) 80 474 139 18*23' 33 8 St Vincent 344 1234 33 13*14' 2 41 8 St. Bartholomew (Guad. 21 281 27 17*54' 2 21 8,17 St. Eustatius (D.A.) 21 600 7 17*3' 2,3 14 8,17 Union (St.V.&Gr.) 11 308 23 12*36' 2 22 13

Key to rrferences:

1. MUer et al. (1992) 2 Cl each (1977) 3. Harvey and Peacock (1989) 4. Knovies and Smith (1995) 5, Smith, D.S. (unpuU. data) 6. Smith et al. (1991) 7. Smith et a/.(1994a) 8. Scott (1986) 9. MUer, L.D. and J.Y. Miller (pers. comm.) 10. Pearce (1969) 11. Schwartz. (1990) 12 RusseU (1992) 13.Davies (1989) 14. Smith, D.S. and N. Davies (unpuU. data) 15. Gaud and Martorrd (1974) 16. Clendi, H.K. In (7) p.228. 17. Schwartz et o/.(1987) 18. MUer and Steinhauser (1992) 19. Smith, D.S. and F. McKenzie (unpuU. data) 20. Alayo and Hernandez (1987) 21. Schwartz and Jchnson (1992) 22 Smith and Hernandez (1992) 23. Evans, P. (pers. comm.) 24. Knowles, D.O. (pers. comm.) 25. Askew (1988) 26. St.Legw (1991) 27. Pinchon and Enrico (1969) 28. Clendr and Bjomdal (1980) 29. Simon and MUer (1986) 30. Bedcer and MUer (1992) 31. Schwartz (1989) 32 Smith et al. (1989) 33. Smithet a l .^ prep.) 34. Brown and Heineman (1972) 35. Vane-Wright et al. (1992) 36. Johnson and Smith (1993) 37. Iftner et al. (1993) 38. Leston et al. (1982) 39. Mnno and Emmel (1993) 40. Smithet al. (1994b) 41. Schwartz and Æmœez. (1982) 42 Hanândez et al. (in prq).) 43. Schwartz and Henderson (1990) 44. West (1966) 45. Knowles, D.O. and D.S. Smith. (ut^uU. data) 46. Ramos (1982) 47. (8) p.213 48. EUiott e/o/. (1980) 49. Mskimen and Bond (1970) 50. (7) p. 116. 51. Hunt and MtdieU (1979)

67 Table 2. Plant species richness and island data.

Area Elevation Island (Km*) (m) Latitude Butterflies Plants Acklins, Crooked and Mayaguana 963 47 22*20' 41 466 Anegada 39 9 18*45' 24 210 Antigua and Barbuda 441 402 17*03' 33 724 Barbados 430 340 13*09' 25 700 Biminis and Andros 6979 36 24*35' 67 764 Cat 388 63 24*15' 25 448 Cayman brae 36 43 19*4' 30 273 Cuba 108660 1994 20*15' 182 7000 Dominica 751 1447 15*31' 52 1600 Grand Bahama and Abacos 354 41 26*21' 35 737 Grand Cayman 197 150 19*20' 46 502 Great Exuma 210 38 23*38" 33 517 Great and Little Inagua 1627 45 21*50' 37 450 Guadeloupe 1438 1480 16*16' 44 1700 Hispaniola 76190 3087 18*30' 202 5000 Isle of Pines 2200 310 21*30' 111 1400 Jamaica 11424 2255 18*05' 126 3003 Little Cayman 26 20 19*40' 23 237 Long 448 54 23*15' 31 468 Mona 62 85 18*05' 53 417 New Providence and Eleuthera 725 37 25* 61 907 Puerto Rico 8866 1341 18*21' 97 2700 San Salvador and Rum Cay 241 43 24*05' 49 469 St. Vincent 344 1234 13*14' 41 1150 Turks and Caicos 430 30 21*47 38 477

68 Table 3. History of the species-area relationship for West Indian butterflies.

Exponential function Power function Power function* Survey N r* r* F p 2 c Munroe (1948) 11 0.92 0.75 27.32 <0.001 0.26 0.80 Scott (1986) 24 0.83 0.85 124.15 <0.001 0.26 0.91 Davies and Smith (1995 68 0.54 0.64 118.08 <0.001 0.20 1.06

* All regression statistics based on log-transformed for both species richness and island area. N = number of data points, z = slope c = intercept

69 Figure 1. Species-area relationship; West Indian butterflies.

2.50

2.00 -

0) ♦ n ♦ ♦ E 1.50 - ♦ ♦ 3 ♦ 4» C 0)if> ♦ % 'ü ♦ ♦ ♦ ♦ & (O y = 0.20x + 1.06 = 0.64 p<0.001

0.50

^ : 60- -1.00 0.00 1.00 2.00 3.00 4.00 5.00 6.00 Area (Km ) Figure 2. Species-area relationship for three island classes: West Indian butterflies.

Greater Antilles: y = 0.37X + 0.489 = 0.69 p<0.001

Lesser Antilles: y = 0.20x + 0.98 R^ = 0.60 p<0.01

È <20 species: I oo c y = 0.06X + 1.06 Ui Q) = 0.23 p<0.05 'o d> a

oL esser Antilles (2)

■G reater Antilles (1)

ACommunities <20 spp.(3)

0 1 2 3 4 5 6

Area (Km^) Figure 3. Species-area relationship: West Indian plants.

4.00

3.80

3.60 -

^ 3.40 o> ■§ 3.20 ♦ 3 I 3.00 ♦ y = 0.38X + 1.81 .£ B 2.80 = 0.80 p<0.001 ♦ ♦ -2.80

2.40

2.20

2.00 1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00 5.50

Area (km ) Figure 4. Relationship between plant and butterfly species richness in the West Indies.

1000 T

0) n

E 100 - 3 C v> .22 'o 0) a (/) ♦ ♦ >» «E y = 0.027x + 25.11 10 - ë 3 R^ = 0.84 p<0.001 CD

100 1000 10000 Plant species number CHAPTERS

GENETIC DIFFERENTIATION IN FOUR SPECIES OF WEST INDIAN BUTTERFLY.

ABSTRACT

Four species of neotropical butterfly: Battus polydamas^ Heliconius charitonia, Dryas iulia and Anartia jatrophae^ were surveyed using allozyme electrophoresis. Phenotypic studies have shown that these species vary in their tendency to form island races in the West Indies. Those observations are generally supported by this study, with all the species, except Anartia jatrophae, showing high levels of population subdivision. A long-standing debate in West Indian biogeograpby is the relative inçortance of dispersal versus vicariance. Low levels of genetic divergence between islands clearly demonstrate recent and/or continuing gene flow among some populations of all four species. Dispersal must, therefore, have some role in the historical biogeography of these butterflies. Although genetic distances between some of the Greater Antilles and the continent are large, particularly for Dryas iulia and

Battus polydamas, they still seem insufiBcient to invoke the ancient separation of the West Indies and continental America. The subspeciation of island populations could result from either local adaptation or purely demographic causes, such as random genetic drift and founder effects. It is argued that the correlation between geography, genetic and phenotypic structure is less likely

74 to be caused by local adaptation than by historical biogeography. More data, however, are needed on phenotypic variation and ecological parameters

(particularly host-plant phylogeography) before local adaptation can be excluded.

BmiODUCTION

Islands have played an inq)ortant role in the study of evolution ever since Darwin first visited the Galapagos. Although the tag 'natural laboratory* is now something of a chché, recent advances in genetic technology have made these laboratories even more productive. Many see local populations as the units of evolutionary importance and great significance is ofien attached to the discontinuity of natural populations, as in the shifting balance (Wright, 1977) and spéciation theories

(Mayr, 1963; Carson and Templeton, 1984). Studies based on real islands are particularly usefiil as they enable discrete demes to be clearly defined, thus providing a fi*amework in which to test evolutionary hypotheses. Here I examine genetic variation in four West Indian butterflies and test predictions concerning the evolution of island populations.

Woods (1989a) noted the paucity of information on the biogeography of West

Indian invertebrates. This is doubly true for genetic studies. Although many West

Indian taxa are well described at the phenotypic level (see Liebherr, 1988 and

Woods 1989b for reviews), only vertebrates have been the subject of extensive

75 genetic study (Hedges, 1989; Phillips et al., 1989; Seutin et al., 1993, 1994).

Butterflies are probably the best documented of aU West Indian invertebrates

(Riley, 1976; Scott, 1972, 1986; Smith et al., 1994a), yet genetic data are still

lacking. This study takes the first steps to rectify the situation.

There are two major hypotheses as to how the existing West Indian fauna came

into being: dispersal (Wallace, 1876; Mathew, 1915; Sinq)son, 1952; Darlington,

1957) and vicariance (Schuchert, 1935; Rosen, 1976; 1985). The dispersal model

argues that the West Indies were once virgin territory, subsequently colonised by

waif dispersal from the continent. Differences in colonisation ability explain vAiy

the modem fauna only contains a limited number of taxa. Extreme proponents of vicariance, on the other hand, beheve that the islands were once populated by a

continental fauna (like modern-day Trinidad) and that extinction events have led to

the current, depauperate fauna (Trinidad has about twice as many butterflies as the

rest of the West Indies put together). The two hypotheses are not mutually

exclusive. It can be argued that the distributions of more mobile taxa are the result

of dispersal, whilst older, more sedentary groups reflect the influence of vicariant

events. In addition, some parts of a species’ distribution may result from vicariance, whilst others dispersal.

Butterflies are old relative to the West Indies and so ancient vicariance may have a role in their biogeography (Miller and Miller, 1989). Lepidoptera originated in the

Cretaceous (Smart and Hughes, 1973; Raven and Axelrod, 1974; Whalley, 1986) with most groups probably in existence by the Palaeocene (Miller and Miller,

76 1989). The West Indies are also of Cretaceous origin, when a proto-Antillean chain of islands linked North and South America. In the Palaeocene, about 60 milHon years before present (myrBP), this chain (the future Greater Antilles) began to move eastwards, finally breaking from the Yucatan in the Eocene, some 40 myrBP. This vicariant event is invoked for some butterflies (Miller and Miller,

1989) and anole lizards (Guyer and Savage, 1986). Jamaica and the ‘south island’ of Hispaniola did not move with the main chain, remaining close to Central

America until the early Miocene, about 20 myrBP. During the Ohgocene (37-24 myrBP) Hispaniola and Cuba were broken up into several fragments, with central

Hispaniola linked to Puerto Rico, and eastern Cuba to north Hispaniola. Puerto

Rico separated from Hispaniola by the early Miocene, and the Greater Antilles reached their current locations by the end of the PHocene, at least 5 myrBP. The

Lesser Antilles originated further to the west of their current position, and there is no evidence that they were ever linked to the continent nor to one another. Falling sea levels associated with Pleistocene glaciations, however, may have caused some reduction in the distances between some of the Lesser Antilles. Most notably the

Grenadine and Puerto Rican banks were probably exposed.

This interpretation of the region’s geological history broadly follows Pindell and

Dewey (1982) and Perfit and Wilhams (1989). However, this should not be taken as the consensus view of West Indian geologists. Unfortunately “so much evidence of complexity in geologic history has led to controversy that is opaque to the outsider” (Wilhams, 1989). Whilst the potential geological constraints should be noted, they should not drive our biogeographical hypotheses.

77 If we accept that genetic divergence increases with time and with the strength of barriers to dispersal, then it is possible to test hypotheses concerning the origin of island taxa (Seutin et al., 1994). If dispersal is important then genetic divergence and geographic isolation should be correlated, although different species might

show different patterns depending upon their dispersal mode and aptitude. In

Drosophila it has been shown that genetic divergence is more rapid between isolated populations than between those in closer proximity (Singh and Rhomberg,

1987). By contrast, if vicariance shapes population structure of populations, genetic divergence is not necessarily correlated with current geography but more with the length of time since the vicariant event. Moreover, all the taxa within a completelv vicariant fauna should show similar patterns, irrespective of their ecologies.

I chose four species to test the origin of West Indian butterflies: Dryas iulia

Fabricius (Heliconiidae), Heliconius charitonia Linnaeus (Heliconiidae), Anartia jatrophae Linnaeus (Nyrq)halidae), and Batins polydamas Linnaeus (Papilionidae).

These were selected because they are common and (phenotypically) they span the spectrum of intraspecific differentiation in West Indies butterflies. Whilst there are better candidates for vicariance, such as Calisto (Smith et al., 1994a), most of these have a limited distribution and differentiation is at the species or even genus level. Since I am also interested in spéciation I prefered to investigate mtraspecific variation.

78 Evidence of vicariance among widspread conspecific populations, and not just

between a few isolated endemic species, would clearly demonstrate an inq)ortant

role for vicariance in the historical biogeography of West Indian butterflies.

Concordant population genetic patterns among species have been used to infer the

historical biogeography of faunas (Bermingham and Avise, 1986).

Battus polydamas is highly differentiated and has a fi'agmented distribution that

may result fi'om the extinction of some island populations. Almost every island has

its own subspecies, there are thirteen in all, recognisable by wing colour pattern

and juvenile morphology (Smith et al., 1994a). Dryas iulia is present on virtually

all the islands and is also highly differentiated, having twelve subspecies.

Heliconius charitonia shows less striking phenotypic differentiation, although six

subspecies have been described. Its West Indian distribution extends fi*om Cuba to

Montserrat; it is absent fi*om the southern Lesser Antilles and Trinidad. Anartia jatrophae is distributed throughout the West Indies and is the least differentiated of

the four species surveyed. It has described forms on each of the Greater Antilles,

St. Croix, and one that covers all of the Lesser Antilles and the continent. The

subspecific classifications of Anartia jatrophae, however, have been strongly

criticised on phenotypic grounds (Gillham, 1957).

According to Miller and Miller (1989), Dryas iulia, Heliconius charitonia and

Anartia jatrophae, probably reached the West Indies by dispersal fi^om the

continent. The Greater Antillean subspecies oîBattus polydamas, however, could

be a result of late Ohgocene-early Miocene geological events. Allozyme

79 electrophoresis was used to determine levels of genetic divergence among insular populations of the four species. If vicariance has been important in the historical biogeography of Battus polydamas (or any of the others) we e?q>ect high levels of

divergence among the Greater Antilles and particularly between the Greater

Antilles and the continent. If) however, dispersal has been more important we

expect genetic distance to be relatively low and to increase with geographic

isolation.

METHODS

Sandies were collected from the West Indies on three trips during 1991 - 1994

(Table 1, Figure 1). Specimens were collected using a hand-net and kept alive in

glassine envelopes on wet ice before being placed in hquid nitrogen for transportation. Prior to freezing the wings were removed and put into small zip- lock bags which were labelled with the island, collecting site, date and an individual number. The corresponding body was similarly coded and wrapped in aluminium foil. In the laboratory specimens were stored at -80^C in cryotubes.

Once a productive site was located on an island up to thirty individuals were collected. Sites were of similar size on all the islands: typically less than 0.5 km^.

This increases the risk of sampling related individuals, and so every effort was made to collect from at least two such sites on each island (preferably some distance apart and/or in different habitats). Sanq)les from either end of the largest

80 island, Cuba, were analysed first to assess the extent of within-island variation.

East and west Cuba (over 600 miles apart) were treated as separate islands for most of the analyses.

Cellulose acetate plates (Helena laboratories) were used for the electrophoresis.

Preparation of specimens, buffer systems and electrophoretic procedure were as in

Richardson et al. (1986). Each species was screened for thirty eight different enzyme systems. The recipes for the stains were minor modifications of those described in Richardson et al. (1986) and Mallet et al. (1993). Twenty five enzyme systems gave consistently scorable results in at least one of the species

(abbreviated name of enzyme and the E C. number are shown in parentheses): phosphoglucose mutase (PGM 2.7.5.1); glutamateoxaloacetate transaminase (GOT

2.6.1.1);phe-pro peptidase (PP 3.4.13.9); leu-ala peptidase (LA 3.4.11 or 13); mafic enzyme (ME 1.1.1.40); P-hydroxybutarate dehydrogenase (HBDH

1.1.1.30); alcohol dehydrogenase (ADH 1.1.1.1); fiuctose diphosphate (FDP

3.1.3.11); glutathione reductase (GR 1.6.4.2); sorbitol dehydrogenase (SDH

1.1.1.14); fiunarase (FUM 4.2.1.2); superoxide dismutase (SOD 1.15.1.1); phosphoglucose isomerase (PGI 5.3.1.9); glucose-6-phosphate dehydrogenase

(G6PD 1.1.1.49); 6-phosphogluconic acid (6PGD 1.1.1.44); glyceraldehyde-3- phophate dehydrogenase (GAPDH 1.2.1.12); phosphoglycerate dehydrogenase

(G3PD 1.1.1.95); isocitrate dehydrogenase (IDH 1.1.1.42); adenylate kinase (AK

2.7.4.3); enolase (ENO 4.2.1.11); malate dehydrogenase (MDH 1.1.1.37); aconitase (ACO 4.2.1.3); adenosine deaminase (ADA 3.5.4.4); mannose-6- phosphate isomerase (MPI 5.3.1.8); esterase (ES 3.1.1.1).

81 The anterior half of the abdomen and half of the thorax were at first analysed separately. Based on those results it was decided to concentrate on the abdomen since it contained more activity for most loci. The thorax was saved for future

DNA analyses, whilst the posterior half of the abdomen (including the genitalia) was kept for future phenotypic character analysis, or fiuther allozyme work.

All scorable loci were run for at least ten individuals (when available) fi'om each island for each species. Subsequently only polymorphic loci were run. Specimens fi'om six islands were usually run on each plate (twelve individuals per plate: two individuals per island). Running individuals fiom difierent islands next to one another on the same plate reduced the chance of mis-scoring alleles.

The data obtained was processed using the software package Biosys-1.7 (Swofford and Selander, 1989). Deviations fiom Hardy-Weinberg equilibrium were tested using a test with Levene’s (1949) correction for small sanq)le sizes. Due to relatively small sample sizes the test was repeated with genotypes pooled into three classes: heterozygotes for the commonest allele, homozygotes for the commonest allele, and other genotypes. Three measures of genetic variation were determined: an unbiased estimate of heterozygosity {He) based on Hardy Weinberg expectations (Levene, 1949; Nei, 1978), percentage of loci polymorphic, and average number of alleles per locus. Genetic variation will be dealt with in more detail elsewhere (Chapter 5).

82 Unbiased genetic distances, D, (Nei, 1978) were calculated. Nei's D is probably the most widely used measure of genetic distance. It is based on the probability that a randomly chosen allele from each of two populations will be identical, relative to the probabihty that two randomly chosen alleles from the same population will be identical. It is zero when there is no genetic divergence and has no upper boundary. Sample sizes and, more importantly, number of loci affect the significance of D. The sanq)lmg properties of D are discussed by Nei (1978) and following this up to thirty individuals were run for as many loci as possible.

Roger’s (1972) genetic distances, modified following Wright (1978), were also calculated. This was in order to use a parsimony method (distance Wagner) of phylogenetic analysis (Farris, 1972) which cannot be applied to nonmetric measures such as Nei's D. The resulting phenograms, known as distance Wagner trees, were rooted to a continental population, which is assumed to be an outgroup to the island fauna.

Wright’s F-statistics: Fis, Fit and Fst (Wright, 1951) were calculated. Fis measures the reduction in heterozygosity of individuals due to nonrandom mating in subpopulations; Fst measures the gene frequency variance in the different sites

(the effect of population subdivision and consequent drift or differential selection);

Fit includes both the previous measures to give the overall reduction in the individuals heterozygosity relative to the total meta-population. For species which showed distinct clustering of islands, these clusters were used in the calculation of hierarchical F-statistics (Wright, 1978).

83 Matrices of geographic distances were constructed based on the minimum great- circle distances separating islands (Admiralty Charts 4400, 956 and 2600). Where two populations were analysed for one island, such as east and west Cuba, these were treated as one population in forming the genetic distance matrix. The effect of geographic isolation on genetic differentiation was assessed by con^aring the matrices of geographic isolation with matrices of genetic distances using the matrix conçarison test (Mantel, 1967) in the software package NT-SYS (Rohlf^ 1988).

This test compares the matrices to see if they have a common structure and was run through 5000 permutations to assess significance.

RESULTS

Sample sizes and genetic diversity data are shown in Table 1. Observed heterozygosity did not differ significantly from the unbiased estimate. He, in any of the species. Wagner and UPGMA trees were constructed, however, as these did not differ greatly only the UPGMA trees are illustrated (Figures 2-7)

Battus polydamas:

Battus polydamas populations were not in Hardy-Weinberg equilibrium for 3 out of 46 comparisons. There was no apparent pattern to the direction of these deviations nor to which loci were out of equihbrium (all 3 were different), nor in the direction the deviation occurred. Of the 20 scorable loci, 13 were polymorphic in at least one population (Appendix 1.1)

84 The average F-statistics over all loci showed that the species was very subdivided in the Antilles: Fis = - 0.183; F/Y = 0.411 ; Fst = 0.503. For conçarison,

Euphydryas editha, a known sedentary species in California, has Fst = 0.118

(McKechnie et al. 1975), whilst the very mobile Heliothis virescens hasFst =

0.002 across the USA (Korman et a i, 1993).

Geographic and genetic distance were unrelated according to the Mantel test

(r = -0.31; t = -1.58; N.S.). This is not surprising given the structure of the

UPGMA tree (Figure 2).

The average Nei's D between populations (Appendix 2.1) ranged from 0.00

(Florida-Martinique) to 0.149 (Dominica-Jamaica). Whilst the close relationship of

Florida and Martinique may reflect a real evolutionary relationship, the absence of any genetic variability in the Florida population may have biased the analysis. All the alleles found in Florida were also common in Cuba; it is therefore a strong possibility that the Florida population originated from proximate Cuba and not distant Martinique.

Another long-distance association must also be treated with caution: Cuba and

Barbados appeared very similar (D = 0.009) but there were only two individuals in the sarq)le from Barbados. Nevertheless one of them did possess the ‘B’ allele of

ME, which was rare throughout the West Indies except in Cuba, where it was the commonest allele.

85 These two coiq)arisons, however, were not solely responsible for the lack of correlation between genetic and geographic distance. Cuba was quite distinct from its Greater Antillean neighbours, Hispaniola and Jamaica (D = 0.116 and 0.136 respectively), whilst they were very similar to one another (D = 0.006), This is so despite the fact that the three islands are roughly equidistant. The Puerto Rican population, meanwhile, was quite idiosyncratic: distinct from both the Greater and

Lesser Antilles, possessing a unique PP allele.

It is interesting that Cuba was quite differentiated from the rest of the Greater

Antilles, but shared some alleles with the Lesser Antilles, Florida and Panama (in particular allele ‘D’ of PP). This suggests that the Lesser Antilles, Florida and

Cuba may be united through relatively recent colonisation, and/or persistent gene flow via the continent.

Anartia Jatrophae:

Anartia Jatrophae populations were out of Hardy-Weinberg equilibrium m 10 out of 118 conq)arisons. There was no apparent pattern to the direction of these deviations nor to which loci were out of equilibrium Of the 22 scorable loci, 19 were polymorphic in at least 1 population (Appendix 1.2).

The average F-statistics over all loci showed that the species is subdivided in the

West Indies as a whole: Fis = 0.007; Fit = 0.195; Fst = 0.189. The average

86 Nei's genetic distances were low (Appendix 2.2) ranging from 0.00 (Cuba-Isle of

Pines; Cuba-Florida, Florida,R-Florida,F) to 0.049 (Hispaniola-Dominica).

Whilst the UPGMA (Figure 3) suggested some limited geographical cohesion, there was no significant correlation between genetic and geographic distance (r =

0.02; t = 0.15 N.S.). This, and the low levels of genetic differentiation, suggest strong and persistent gene flow among West Indian Anartia jatrophae.

Heliconius charitonia:

Heliconius charitonia populations were not in Hardy-Weinberg equilibrium for 12 out of 92 conq)arisons. There was no apparent pattern to the direction of these deviations nor to which loci were out of equihbrium Of the 25 loci that were

scorable, 19 were polymorphic in one population or more (Appendix 1.3).

The average Nei's genetic distance between populations (Appendix 2.3) ranged from 0.00 (Mona-Puerto Rico) to 0.113 (Key Largo-Montserrat). The average F- statistics over all loci showed that the species is subdivided in the West Indies; Fis

= 0.128; Fit = 0.502; Fst = 0.430.

Heliconius charitonia populations showed a significant correlation between genetic and geographic distance (r = 0.78; t = 3.30; p<0.001). Genetic distances generally increased from west to east and the structure of the UPGMA tree

(Figures 4 and 5) suggests that colonisation occurred in this direction. The ‘D’ aUele of MPI was absent in most of the western islands but was common in Mona,

87 Puerto Rico and the northern Lesser Antilles. Similarly the ‘D’ allele of IDH-2 was absent from the eastern islands (except at low frequency in Mona) but was common in the west.

Dryas iulia:

Dryas iulia populations were out of Hardy-Weinberg equilibrium in 15 out of 116 conq)arisons. There was no apparent pattern to the direction of these deviations nor to which loci were out of equihbrium. Of the 21 scorable loci, 17 were polymorphic in at least one population (Appendix 1.4).

There was no significant correlation between genetic and geographic distance according to the Mantel test (r = 0.02; t = 0.19; N.S.). (The single individual from

Florida was excluded from the analysis.) However, the UPGMA tree clearly showed strong geographic cohesion among four genetic groups (Figures 6 and 7): the Greater Antilles (including the Bahamas), the North Lesser Antilles

(Montserrat, St.Kitts, Guadeloupe, and Dominica) the Central Lesser Antilles

(St.Lucia and Martinique) and the continent/South Lesser Antilles (Panama,

Trinidad, Grenada and St. Vincent). These groups were genetically homogenous and distinct from each other. Genetic distances within groups were significantly lower than those between groups (t = 1.98, p<0.001; Table 2).

Determining the phylogenetic relationships among the groups is not straightforward. For example, some loci suggested that the North Lesser Antilles

(NLA) fall in with the Greater Antilles (GA) and Central Lesser Antilles (CLA)

88 groups e.g. PGM. LAL-1, however, showed strong differentiation in the NLA, wMst other loci grouped the NLA with the continent/South Lesser Antilles

(CO/SLA) e.g. GOT-1.

The GA and CO/SLA clades, however, did appear to be very distinct. Overall the

NLA seemed more closely ahgned with the GA than the CO/SLA, whilst the CLA fell in the CO/SLA clade (Figure 6).

The average Nei's D between populations (Appendix 2.4) ranged from 0.00

(Cuba,west-Isle of Pines; Guadeloupe-Dominica-Montserrat ) to a massive 0.482

(Guadeloupe-Martinique). Generally, the level of divergence between the continental and Greater Antillean clades was not much lower than those observed between Dryas iulia and seven other Hehconiid species (Turner et al., 1979).

The average F-statistics over all loci showed that the species is very subdivided in the Antilles: Fis = 0.025; Fit = 0.730; Fst = 0.723. The hierarchical F-statistics

(Table 3) show that much of the subdivision can be explained by the differentiation among island groups.

DISCUSSION

The geographic structure of genetic and phenotypic variation can reveal the historical biogeography of island taxa, allowing us to assess the relative inq)ortance

89 of vicariance versus dispersal. In this survey the relationship between geography

and genetic differentiation varied among species. Geographic and genetic

distances were only significantly correlated for Heliconius charitonia. Dryas iulia

also showed geographically cohesive structure, however, this was not sinq)ly

related to the distance between islands. No geographic coherence was evident in

the phylogenetic structure of either Anartia jatrophae or Battus polydamas.

The apparent independence of geography and genetic structure in Anartia jatrophae can be explained by either recent isolation and/or gene flow.

Populations will move toward fixed allele differences only when less than one

individual migrates between populations each generation (Wright, 1969). With a

wide distribution and low levels of phenotypic and genetic differentiation, Anartia jatrophae must be very mobile. High levels of gene flow seem to have masked any

influence geography may have had on genetic structure. Levels of genetic

divergence in Anartia jatrophae are clearly inconsistent with the ancient, vicariant

origin of any single insular population or West Indian Anartia jatrophae in general.

The genetic data supports Gdham (1975) who suggested that there is little phenotypic justification for the West Indian subspecies of Anartia jatrophae. It is interesting to note, however, that the genetic diversity of Lesser Antillean Anartia jatrophae is similar to that found in Trinidad. Such high genetic diversity in these

small islands is probably due to gene flow from the continent. Phenotypically,

Lesser Antillean Anartia jatrophae has been classified as the South American

90 subspecies, A. j. jatrophae, I therefore agree with Smithet al. (1994a) that, whilst the subspecies are very weak, they may have some merit.

The genetic structure of Battus polydamas also appears unrelated to geography.

But unlike Anartia jatrophae, Battiis polydamas shows strong differentiation

among some islands. In the dispersal model, one expects geographic distance to

reduce the probability of inter-island movement, thus it should increase genetic

divergence. When divergence is unrelated to geographic distance, this might

suggest vicariance. Indeed vicariance has been invoked, at least as a possibility, for the Greater Antillean subspecies of Battus polydamas (Miller and Miller, 1989).

The genetic distances, however, seem too low since the relevant plate-tectonic

events occurred over 5 myrBP. A dispersal explanation seems more likely.

Genetic distances are not expected to correlate with geographic distance if distance

does not represent a barrier to dispersal. The difficulties associated with becoming

established on an island might have a greater bearing on immigration than problems

in reaching islands (Chapter 2). Colonisation among islands could be rare and

stochastic, and all islands might be potential sources.

Battus polydamas is a strong flying butterfly, capable of at least 6.56 m/s (Srygley

and Dudley, 1993), and there is good evidence for relatively recent, long distance

over-water colonisations. Firstly, Schwartz ( 1990) identified Battus polydamas on

Barbados as Battus polydamas polydamas, the continental subspecies. This would

explain the high genetic diversity observed despite the tiny size (N=2) of the

sample fi’om Barbados. Battus polydamas must have crossed the sea to establish

91 this population since Barbados has never been connected to the continent.

Furthermore Barbados is upwind of Trinidad.

Secondly, the two subspecies of Battus polydamas from Jamaica and Hispaniola are very similar, as was noted by Rothschild and Jordan (1906) in their original descriptions. Genetically they are almost indistinguishable, suggesting a recent colonisation and/or persistent gene flow (the two islands have been separated for at least 5 million years). The lower genetic diversity of the Jamaican population might indicate a recent colonisation from Hispaniola, Wiich raises the possibility thatBattus polydamas did not previously occur on Jamaica, or had gone extinct.

(Interestingly, whilst Hispaniola and Cuba have two Battus species, Jamaica has only Battus polydamas.)

With the notable exception of Jamaica and Hispaniola, the Greater Antillean populations are all quite different from one another. Although this is possible evidence of a vicariant origin, if so it has been masked by recent dispersal. The relationship between Cuba, the Lesser Antilles and the continent suggests persistent and/or recent gene flow since the initial colonisation/vicariance of the

Greater Antillean populations. The rest of the Greater Antilles, however, does not seem to have exchanged genes with the continent or the Lesser Antilles for a considerable period of time. More sanq)ling of this species from the Lesser

Antilles and the continent is necessary to resolve its historical biogeography.

92 Heliconius charitonia has a limited distribution (it is absent from the southern

Lesser Antilles and north-eastern South America), it shows a significant correlation between genetic and geographic distance, and has relatively low levels of genetic

divergence. Together these characteristics suggest a history of relatively recent

dispersal The most hkely source of colonists is Central America, with an

e?q)ansion from west to east. It is unlikely that this species spread from St.Kitts or

Montserrat to the Greater Antilles but not south to the other Lesser Antilles. The

structure of the UPGMA tree (Figures 4 and 5) suggests that the Greater Antilles were colonised in stepping stone fashion from the continent to Jamaica, Jamaica to

Hispaniola, and rinally Hispaniola to Cuba and the Bahamas. Puerto Rico appears to have been part of a separate, or earlier wave of colonisation, also from the

continent, which included Mona and extended into the northern Lesser Antilles.

Unfortunately the continental sanq)le is Ecuadorean, He. peruviana, a different

subspecies from Central American Heliconius charitonia. If Central American

Heliconius charitonia are as distinct from the West Indian populations as are

H.c.peruviana, then vicariance may have had a role in the origin of the West Indian population. Irrespective of their origin, it appears that Heliconius charitonia is

still expanding its range in the West Indies since it must have recently crossed

water to colonise the northern Lesser Antilles. It seems hkely that Heliconius

charitonia will eventuaUy colonise the southern Lesser Antilles. That these islands have not already been colonised surely reflects the absence of Heliconius

charitonia from north-eastern South America and highlights its in^ortance as a

source of Lesser Antihean butterflies.

93 Heliconius charitonia also provides evidence for the inq)ortant role wind may play in butterfly dispersal. Mona is a small island, wdiich emerged midway between

Hispaniola and Puerto Rico during the Pleistocene. Its butterflies have been surveyed by Smith et a l (1988, 1994b), who found Heliconius charitonia on

Mona to be the Puerto Rican subspecies, Heliconius charitonia charitonia. This is supported by the allozyme data. Smith et al. (1988, 1994b) noted that the influence of Puerto Rico extends to most of Mona’s butterflies. This is surprising given that Hispaniola is nearly ten times the size of Puerto Rico with more than twice the number of species (it is also a few miles closer). Island biogeography theory predicts that larger, closer species pools will supply a correspondingly larger proportion of species (MacArthur and Wilson, 1967). The prevailing east/north-easterly winds are the most likely explanation for the counter-intuitive origm of Mona’s butterfly fauna.

Like Heliconius charitonia the genetic structure of Dryas iulia is related to geography, although the situation is much more complex. This may reflect the fact that there are two possible sources of West Indian Dryas iulia. South America and

Central America/Mexico, or that it has sinq)ly had a longer history in the West

Indies.

A great deal of the differentiation among Dryas iulia populations has occurred between four distinct island clusters: the Greater Antilles and Bahamas (GA), the

North Lesser Antilles (NLA), the Central Lesser Antilles (CLA), and the South

94 Lesser Antilles/continent (SLA/CO). This structure is not entirely missed by the

current phenotypic classification. The subspecies found in the northern Lesser

Antilles strongly resemble one another and to a lesser extent those in the Greater

Antilles. The Martinique and St.Lucia subspecies are more distinct, whilst

St. Vincent and Grenada share a subspecies that resembles the continental form

(Smith et al., 1994a).

The low levels of genetic differentiation among Greater Antillean populations

would seem to rule out vicariance. However, there is strong evidence for a major

phylogenetic break between Greater Antillean and continental Dryas iulia. This

might be the result of ancient vicariance, vsfien the West Indies drifted away from

the continent and into the Caribbean. A chance colonisation of the Greater

Antilles, followed by little subsequent gene flow, however, is just as consistent

with the evidence. Indeed dispersal is more likely given the genetic distance Wiich,

though substantial, is less than one would expect if isolation occurred >5 myrBP.

However the Greater Antilles were initially colonised, dispersal must have

continued among them to explain their current homogeneity. But why didn’t this

dispersal extend to the continent? As with Heliconius charitonia, the prevailing winds may inhibit the eastward dispersal of Dryas iulia. This would explain a lack

of gene flow from Central America/Mexico to the Greater Antilles. Although gene flow might have occurred from the Greater Antilles to the continent, the shortest

distance is still relatively large (105 miles across the Yucatan channel).

Furthermore any migrants which make it from Cuba to the Yucatan would

95 probably be swamped in the relatively massive continental population (though evidence of gene flow might be found in a sançle fi-om the Yucatan).

Gene flow between the Greater Antilles and the North Lesser Antilles also seems to have been restricted. The Greater and Lesser Antilles are separated by the wide

Anegada passage (>100 miles). Colonisation is unlikely from the Lesser Antilles to

Puerto Rico, because of the small pool of potential colonists. If emigrants form a constant proportion of populations, smaller islands with smaller populations will supply fewer emigrants. Most dispersal should, therefore, occur from the relatively massive Puerto Rico, eastwards to the small islands of the North Lesser Antilles.

Again, as demonstrated by Heliconius charitonia^ dispersal to the east may be rare given the prevailing easterly winds. Thus the width of the Anegada passage, combined with opposing winds, may explain the reduction of gene flow between the Greater and Lesser Antilles.

The North Lesser Antilles may only have been colonised recently from the Greater

Antilles. This explanation seems plausible since the North Lesser Antilles became more accessible to the Greater Antilles during the Pleistocene, when sea levels were occasionally much lower than today. During these periods the Puerto Rican bank would have been exposed and some of the North Lesser Antilles would have been larger. This would have diminished the Anegada passage as a barrier to dispersal. Changes in sea-level have been similarly invoked to explain the biogeography of some Bahamian butterflies (Miller et al., 1992).

96 The South and Central Lesser Antilles were probably colonised by dispersal from

South America. The presence of three genetically distinct groups of Dryas iulia within the Lesser Antilles was possibly the most surprising result of this survey.

The islands are approximately equidistant (typically about 30 miles apart), they have similar physical and ecological characteristics, and there is no obvious barrier to gene flow between Grenada and the Anegada passage

Given the ability of Dryas iulia to move between Trinidad and the Lesser Antilles, how could three such distinct genetic groups be maintained? It is possible that we are not dealing with one species and that reproductive isolation explains the lack of gene flow. The levels of divergence between West Indian groups are not very different from those found between Dryas iulia and other species of Heliconiid

(Turner et a l, 1979). However, reproductive isolation cannot be assumed on the basis of genetic data alone (Chapter 6). Whether or not spéciation has occurred, geographic isolation probably helped to reduce gene flow among the groups.

Indeed spéciation might not have occurred without such a reduction.

As discussed above, the North Lesser Antillean population probably originated from the Greater Antilles and it has either not yet reached Martinique, or is reproductively isolated from the southern groups. The genetic similarity of the

South Lesser Antilles and continental populations is indicative of either recent colonisation and/or sustained gene flow. I suspect that the Central Lesser Antilles represents an older South American clade that has been replaced in the South

97 Lesser Antilles. Again it cannot be ruled out that reproductive isolation maintains these two groups.

Dryas iulia therefore appears to have colonised the Lesser Antilles in a pincer movement, with one wave from the Greater Antilles and two waves from South

America (Figure 8). However, it is equally possible that colonisation has occurred solely from South America (Figure 9). Although we cannot be certain for Dryas iulia, other butterflies have almost certainly colonised the North Lesser Antilles from the Greater Antilles, probably the best exart^le being Heliconius charitonia.

This hypothesis is supported by Scott (1972) who plotted faunal identities for all butterfly species in the West Indies and concluded that colonisation occurred both from the west and the south, with the north and central Lesser Antilles showing afiBnities to both the Greater Antilles and South America.

TheDryas iulia pattern is not unusual at the phenotypic level. Another Hehconiid,

Agraulis vanillae, shows a similar distribution. Agraulis vanillae is present in the

West Indies as two subspecies: A.v.insularis from the Greater AntiUes south to

Dominica, and A.v.vanillae from South America north to Dominica. Apparently on Dominica the two subspecies freely intergrade (Smith et al., 1994a).

Taking the evidence of ah four species in the study, dispersal has clearly played a major role in the historical biogeography of these butterflies. Vicariance remains plausible, if improbable, for some of the Greater Antihean populations of Dryas iulia and Battus polydamas. However, even here, the very different pattern

98 between the two species, and the low levels of genetic divergence among some of the islands, suggest that dispersal has blurred any underlying vicariant origins.

The genetic structure of butterfly populations generally seems to concur with phenotypic variation and subspecific classification (Scott, 1972; Miller and Mdler,

1989; Smithet al. 1994a). Notable exceptions to this are the Lesser Antillean populations q ïBattus polydamas, which show httle genetic differentiation, but are phenotypically the most strongly demarcated (Smith et al. 1994a).

Irrespective of how the species came to be in the islands, once they arrived it is clear that differentiation, even spéciation, has occurred. This may result from either local adaptation, or neutral evolution propagated by genetic drift and founder effects. If island forms only differ at neutral loci, phenotypic variation (on which sub specific distributions are based) will match differentiation at neutral genetic markers. If either the genetic or the phenotypic characters are influenced by selection, they will not necessarily be congruent. Furthermore, selection is only likely to lead to geographically cohesive variation if the irrq)ortant ecological parameters also show some geographic coherence. For exanq)le, if host-plant distributions are determined by dispersal, and hence correlated with geography, then butterflies adapting to local host-plant races would reflect the plants’ phylogeographic history.

Whilst the data are still inadequate to make detailed conçarisons of genetic and phenotypic variation, the general congruence of subspecies, genetic structure and

99 geography (with Lesser Antillean Battus polydamas being a significant exception)

suggests that subspeciation results from diflFerential neutral evolution, not local

adaptation.

Clearly we are unable at the moment to draw firm conclusions concerning the

historical biogeography of these four species, let alone of West Indian butterflies in

general As Williams (1989) points out: “the data behind systematic and

phylogenetic hypotheses in biology is defective as regards detail, sometimes also

erroneous in part or entirety, inadequate in its sanq)ling, irregular in its analysis,

and insufficient to the claims made”. Nevertheless working hypotheses should be

suggested, although we make our “biased statements with lowered voices”.

BIBLIOGRAPHY

Bermingham, E. and Avise, J.C. 1986. Molecular zoogeography of freshwater fishes in the southeastern United States. Genetics 113:939-965.

Carson, H.L., and Tenq)leton, A R. 1984. Genetic revolutions in relation to spéciation phenomena: the founding of new populations. Ann. Rev. Ecol Syst. 15: 97-131.

Darlington, P.J. 1957. Zoogeography: The Geographical Distribution of Animals. John Wiley, New York, 675 pp.

Farris, J.S. 1972. Estimating phylogenetic trees from distance matrices. Am. Nat. 106:645-668.

100 Gillham, N.W. 1957. Subspecies versus geographic variation in Caribbean populations o f Anartia jatrophae Johansson (Lepidoptera, Nynq)hahdae). Amer. Mus. Novit. No. 1845. 22p.

Guyer, C. and Savage, J.M. 1986. Cladistic relationship among anoles (Sauria: Iguanidae). Syst. Zool. 35:509-531.

Hedges, S.B. 1989. Evolution and biogeography of West Indian frogs of the genus Eleutherodactylus'. slow evolving loci and the major groups. In: Biogeography of the West Indies. C.A. Woods ed. Sandhill Crane Press, Inc. Gainesville FL. pp.305-370.

Korman, A .K , Mallet, J., Goodenough, J.L., Graves, J.B., Hayes, J.L., Hendricks, D E , LuttreU, K , Pair, S.D. and Wall, M. 1993. Population structure in Heliothis virescens (Lepidoptera: Noctuidae): an estimate of gene flow. Ann. Ent. Soc. 86:182-188.

Levene, H. 1949. On a matching problem arising in genetics. Ann. Math. Stat. 20:91-94.

Liebherr, J.K. (ed.) 1988. Zoogeography of Caribbean Insects. Cornell University Press, Ithaca and London.

MacArthur, R.H. and Wilson, E.G. 1967. The Theory of Island Biogeography. Princeton University Press, NJ.

Mallet, J., Korman, A., Heckel, D G , and King, P. 1993. Biochemical genetics of Heliothis and Helicoverpa (Lepidoptera: Noctuidae) and evidence for a founder event m Helicoverpa zea. Ann. Entomol. Soc. Amer. 86: 189-197.

Mantel, N. 1967. The detection of disease clustering and a general regression approach. Cancer Research 27:209-220.

101 Mathew, G. A. 1915. Climate and evolution. Annals of the New York Academy of Science 24: 171-318.

Mayr, E. 1963. Animal Species and Evolution. Harvard University Press, Cambridge, MA.

McKetchnie, S.W., Ehrhch, W. and White, R.B. 1975. Population genetics of Euphydryas butterflies I. Genetic variation and the neutrahty hypothesis. Genetics 81:571-594.

Miller, L.D. and Miller J.Y. 1989. The biogeography of West Indian butterflies. In: Biogeography of the West Indies. C.A. Woods ed. Sandhill Crane Press, Inc. Gainesville FL. pp. 229-262.

Miller, L D , Simon, M.J. and Harvey, D.J. 1992. The butterflies (Insecta: Lepidoptera) of Crooked, Acklins, and Mayaguana Islands, Bahamas, with a discussion of the biogeographical afiBnities of the southwestern Bahamas and a d escription of a new subspecies by H.K. Clench. Ann. Carnegie Mus. 61:1-31.

Nei, M. 1978. Estimation of average heterozygosity and genetic distance fi*om a small number of individuals. Genetics 89:583-590.

Perfit, M R., and Wilhams, E.E. 1989. Geological constraints and biological retrodictions in the evolution of the Caribbean sea and its islands. In: Biogeography of the West Indies. Woods, C.A. (ed.) Sandhill Crane Press, Inc. Gainesville FL. pp. 47-102.

102 Phillips, C J., Pumo, D E., Genoways, H.H. and Ray, P.E. 1989. Caribbean island zoogeography: a new approach using mitochondrial DNA to study Neotropical bats. In: Biogeography of the West Indies. C.A. Woods ed. Sandhill Crane Press, Inc. Gainesville FL. pp. 661-684.

Pindell, J.L. and Dewey, I.E. 1982. Permo-Triassic reconstruction of western Pangea and the evolution of the Gulf of Mexico/Caribbean region. Tectonics 1:179-211.

Raven, P. H. and Axehod, D.I. 1974. Angiosperm biogeography and past continental movements. Annals of the Missouri Hot. Gdn. 61: 539-673.

Richardson, B.J. Baverstock, P R. and Adams, M. 1986. AUozyme Electrophoresis. A Handbook for Animal Systematics and Population Studies. Academic Press, London.

Ricklefs, R.E. and Cox, G.W. 1972. Taxon cycle in the West Indian avi&una. Am Natur. 106:195-219.

Riley, N.D. 1975. A field guide to the butterflies of the West Indies. Collins, London.

Rogers, J.S. 1972. Measures of genetic similarity and genetic distance. Studies in Genet. Vn. Univ. Texas Publ. 7213:145-153.

Rohlf) F.J. 1988. NTSYS: Numerical Taxonomy and Multivariate Analysis System, Version 1.40. Setauket, New York: Exeter Publ., Ltd.

Rosen, D E 1976. A vicariance model of Caribbean biogeography. Systematic Zoology 24: 431-464.

103 . 1985. Geological hierarchies and biogeographic congruence in the Caribbean. Ann. Missouri. Bot. Gdn. 72:636-659.

Rothschild, W. and Jordan, K 1906. A revision of the American Papihos. Novitates Zoologicae 13, No. 3.

Scott, J.A. 1972. Biogeography of Antillean butterflies. Biotropica 4: 32-45.

. 1986. Distribution of Caribbean butterflies. Papiho, New Series 3:1-26.

Schuchert, G. 1935. Historical geology of the Antillean-Caribbbean region or lands bordering the Gulf of Mexico and the Caribbean Sea. John Wiley & Sons, New York.

Schwarz, A. 1990. Battus polydamas (Papihonidae) on Barbados, West Indies. J. Lep. Soc. 44: 290-291.

Seutin, G , Brawn, J., Ricklefs, R.E. and Bermingham, E. 1993. Genetic divergence among populations of a tropical passerine, the streaked saltator {Saltator albicollis). Auk 110: 117-126.

Seutin, G , Klein, N .K , Ricklefs, RE., and Bermingham, E. 1994. Historical biogeography of the bananaquit (Coereba flaveold) in the Caribbean region: a mitochondrial DNA assessment. Evol. 48: 1041-1061.

Sinq)son, G.C. 1952. Probabihties of despersal in geologic time. Bull, of A.M.N.H. 99:163-176.

Singh, R S. and Rhomberg, L.R 1987. A conçrehensive study of genic variation in natural populations of Drosophila melanogaster. n. Estimates of heterozygosity and patterns of geographic differentiation. Genetics 117:255-271.

104 Smart, J. and Hughes, N.F. 1973. The insect and the plant: progressive palaeoecological integration. Pp. 143-155. In The Great American Biotic Interchance. F.G.Stehli and S.D. Webb (eds.). Plenum Press, New York.

Smith, D.S., Ramos, S.J. McKenzie, F., Munroe, E. and Miller, L.D. 1988. Biogeographical afiBnities of the butterflies of a ‘forgotten’ island: Mona (Puerto Rico). Bull. AUyn Mus. No. 121: 1-31.

Smith, D.S., Miller, L.D. and Miller, J.Y. 1994a. The Butterflies of the West Indies and South Florida. Oxford University Press, Oxford.

Smith, D.S., Ramos, S.J. McKenzie, F. 1994b. The butterflies of Mona Island (Puerto Rico) and an approach to their origins and phenology. Caribb. J. Sci. 30:95-103.

Srygley, KB. and Dudley, K 1993. Correlations of the position of center of body mass with butterfly escape tactics. J. Exp. Biol. 174:155-166.

SwofiFord, D.L. and Selander, KB. 1989. BIOSYS-1: A Computer Program for the Analysis of Allelic Variation in Population Genetics and Biochemical Systematics. Release 1.7 Chanq)aign, Illinois: Dl. Natural History Survey.

Turner, J.KG., Johnson, M.S. and Eanes, W.F. 1979. Contrasted modes of evolution in the same genome: allozymes and adaptive change in Heliconius. Proc. Natl. Acad. Sci. USA 76:1924-1928.

Wallace, A K 1876. The Geographical Distribution of Animals. Harper, New York. vol. 1, 503 pp. and vol. 2, 607 pp.

Whalley, P. 1986. A review of the current fossil evidence of Lepidoptera in the Mesozoic. Biol. J. Linn. Soc. 28: 253-271.

105 Williams, E.E. 1989. Old problems and new opportunities in West Indian biogeography. In Biogeography of the West Indies. Sandhill Crane Press, Inc. Gainesville FL. Woods, C.A. (Ed), pp. 1-46

Woods, C.A. 1989a. Introduction to West Indian biogeography. In Biogeography of the West Indies. Sandhill Crane Press, Inc. Gainesville FL. Woods, C.A (ed). pp. ix-xviii.

. (ed.) 1989b. Biogeography of the West Indies. Sandhill Crane Press, Inc. Gainesville FL.

Wright, S. 1951. The genetical structure of populations. Ann. Eug. 15: 323-354.

. 1969. Evolution and the Genetics of Populations, vol. 2. The Theory of Gene Frequencies. Chicago: University of Chicago Press.

. 1977. Evolution and the Genetics of Populations, vol. 3. Experimental results and evolutionary deductions. Chicago: University of Chicago Press.

. 1978. Evolution and the Genetics of Populations, vol. 4. Variabihty within and among natural populations. Chicago: University of Chicago Press.

106 Table 1. Sanq)ling data and genetic diversity.

Allele % of loci Species Locale N diversity* polymorphic** He Battus polydamasBarbados 2 1.2 20 0.125 Cuba west 33 1.6 25 0.093 Dominica 2 1 5 0.033 Florida 21 1 0 0 Guadeloupe 20 1.1 10 0.02 Hispaniola 30 1.7 35 0.082 Jamaica 29 1.4 10 0.053 Martinique 10 1.1 10 0.01 Panama 4 1.4 30 0.13 Puerto Rico 9 1.3 25 0.094 Anartia jatrophaeCuba east 16 1.4 13.6 0.063 Cuba west 21 1.5 22.7 0.062 Dominica 16 1.5 36.4 0.108 Florida, Fairchild 30 1.3 22.7 0.049 Florida, Redlands 20 1.2 18.2 0.048 Grenada 16 1.5 27.3 0.099 Guadeloupe 16 1.4 27.3 0.114 Hispaniola 19 1.4 22.7 0.047 Isle of Pines 18 1.5 18.2 0.06 Jamaica 26 1.6 27.3 0.091 Montserrat 18 1.5 36.4 0.156 Puerto Rico 20 1.1 9.1 0.036 St. Lucia 20 1.5 36.4 0.127 Trinidad 20 2 54.5 0.152

* average number of alleles per locus ** locus polymorphic when frequency of most common allele < 0.95

107 Table 1. Sampling data and genetic diversity.

Allele % of loci Species Locale N diversity* polymorphic** He Heliconius charHonia Cuba east 24 1.6 24 0.114 Cuba west 24 1.9 32 0.136 Florida 24 1.5 28 0.122 Hispaniola 22 1.5 20 0.082 Isle of Pines 24 1.9 36 0.131 Jamaica 24 1.8 32 0.144 Key Largo 5 1.3 24 0.128 Mona 24 1.5 20 0.056 Montserrat 25 1.2 8 0.02 New Providence 5 1.3 24 0.111 Puerto Rico 24 1.2 12 0.049 St. Kitts 20 1.2 12 0.034 Dryas iuiia Cuba east 13 1.5 23.8 0.081 Cuba west 17 1.7 28.6 0.082 Dominica 24 1.1 0 0.004 Florida 1 1.1 9.5 0.095 Grenada 6 1.2 23.8 0.05 Guadeloupe 6 1 0 0 Hispaniola 14 1.7 33.3 0.082 Isle of Pines 20 1.9 38.1 0.095 Jamaica 21 1.6 28.6 0.054 Martinique 11 1.2 4.8 0.031 Montserrat 22 1.2 4.8 0.015 New Providence 20 1.3 19 0.048 Panama 19 2.1 52.4 0.204 Puerto Rico 18 1.5 19 0.057 St. Kitts 19 1.3 14.3 0.037 St. Lucia 16 1.4 19 0.075 St. Vincent 9 1.1 14.3 0.042 Trinidad 7 1.5 33.3 0.15

* average number of alleles per locus ** locus polymorphic when frequency of most common allele < 0.95

108 Table 2. Genetic distances among and between Dryas iulia groups.

No.of GROUP pops. GA NLA OLA CO/SLA

Greater Antilles (GA) 8 0.028 ( .000- .089)

North Lesser 4 0.197 0.052 Antilles (NLA) (.127-.375) (.000- .104)

Central Lesser 2 0.309 0.449 0.068 Antilles (OLA) ( .264- .355) (.417- 482) ( .068- .068)

Continent/South 4 0.243 0.296 0.183 0.079 Lesser Antilles (CO/SLA) (.133-.367) (.180- 434) (.144-.242) (.034- .119)

109 Table 3. IfierarchicalF statistics for Dryas iulia groups.

Comparison Variance F X Y component XV Island Group 0.85 0.38 Island Total 3.21 0.70 Group Total 2.37 0.52

110 Figure 1. The West Indies.

Florida

P ro v id e n c e ."*K*y Largo

Bahamas ATLANTIC OCEAN

Yucatan (Mexico) Cuba

Hispaniola

GREATER ANTILLES LESSER ^ ANTILLES

CENTRAL AMERICA

Trinidad Panama PACIFIC OCEAN 500 KM SOUTH AMERICA Figure 2. UPGMA tree: Battus polydamas

DisUoœ, D (Nei,1978)

.10 .07

Cuba, west Barbados

Guadeloupe Florida I Martinique

Dominica Panama Puerto Rico Hispaniola

Jamaica

112 Figure 3. UPGMA tree: Anartia jatrophae

DUt*Dce.D(Nei, 1978)

.03 .02

Cuba, west

Florida, F È Florida, R 'Isle of Pines

' Cuba, east I Dominica

Guadeloupe Trinidad

StLucia

I Montserrat

Hispaniola

Puerto Rico

Jamaica Grenada

113 Figure 4. UPGMA tree: Heliconius charitonia

Dùtance. D (Nei. 1978)

.20 .13

Cuba,west Isle of Pines Cuba, east

Hispaniola

New Providence Jamaica

Mona

Puerto Rico

StKitts Montserrat

Florida Key Largo

H.peruviana (Peru)

114 Heliconius charitonia Figure 5. UPGMA tree and geographic distribution of clades. Cuba, west - 1 Isle of Pines - 2 Cuba, east - 3

Hispaniola - 4

New Providence - 5

Jamaica • 6

Mona - 7

Pumo Rko -

St. Kitts - 9

Florida - A

Key Largo - B

H.c.peruviana Figure 6. UPGMA tree: Dryas iulia

Distance, D (Nei, 1978)

:n .20 .13

Cuba, west

Isle of Pines New Providence

Cuba, east Florida

Puerto Rico Hispaniola Jamaica

Guadeloupe Dominica Montserrat

StKitts St.Lucia Martinique Grenada Panama StVincent

Trinidad

116 Figure 7. UPGMA tree and geographic distribution of Dryas iulia clades.

NU. 1 N IA 2 NLA 3 NLA 4 > ^ CLA 1 CLA 2 CO/SLA CO/SLA 2 CO/SLA 3 CO/SLA 4

NLA 4 NLA 2 NLA 1 NLA 2

CL4 2 ^

CO/SLA 3 0

CO/SLA 1

CO/SLA A ^

CO/SLA 2 Figure 8. Colonisation of the West Indies by Dryas iulia: pincer scenario.

\

1 = Colonisation of Greater Antilles directly from continent. 2 = Colonisation of North Lesser Antilles from Puerto Rico. 3,4 = Successive colonisation of the Lesser Antilles from South America. Figure 9. Colonisation of the West Indies by Dryas iulia: South America scenario.

\

la, 2 and 3 = successive colonisations of tresser Antilles from South America lb = colonisation of Greater Antilles from North Lesser Antilles. Appendix 1.1 Allele frequencies: Battus polydamas.

0 g X 2 t- o ? I 1 1 ? a s 1 Dl 3 1 1 L o c u s 1 1 1 1 s i 2 % 1« Dl MEN-1 (N) 21 33 29 9 28 20 2 10 2 4 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 B 0.00 0.79 0.03 0.06 0.04 0.00 0.00 0.00 0.50 0.00 C 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 D 1.00 0.21 0.97 0.94 0.96 1.00 1.00 1.00 0.50 1.00 PGM-1 (N) 20 33 30 9 29 20 2 10 2 4 A 0.00 0.02 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.13 B 0.00 0.00 0.30 0.22 0.62 0.00 0.00 0.00 0.00 0.38 C 1.00 0.82 0.53 0.61 0.28 1.00 1.00 0.95 0.50 0.50 D 0.00 0.17 0.17 0.17 0.10 0.00 0.00 0.05 0.00 0.00 E 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.50 0.00 GOT-1 (N) 20 33 30 9 28 20 2 10 2 4 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 B 0.00 0.12 0.02 0.00 0.00 0.00 0.00 0.00 0.00 0.00 C 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 D 1.00 0.88 0.93 1.00 1.00 0.95 1.00 1.00 1.00 1.00 E 0.00 0.00 005 0.00 0.00 0.05 0.00 0.00 0.00 0.00 IDH-1 (N) 21 31 28 9 29 20 2 10 2 4 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 B 0.00 0.00 0.05 0.72 0.00 0.00 0.00 0.00 0.00 0.00 C 1.00 0.08 0.95 0.28 1.00 1.00 1.00 1.00 1.00 0.88 D 0.00 0.02 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.13 LAL-1 (N) 21 33 30 9 29 20 2 10 2 4 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 B 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 C 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 0.88 D 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.13 Appendix 1.1 Allele frequencies: Battus polydamas (N

A i f a I 1 1 \ 1 Locus I 1 B" 1 1 1 1 t 1 PPR-1 (N) 21 33 29 9 29 20 2 10 1 4 A 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.00 0.00 0.00 B 0.00 0.06 0.98 0.00 1.00 0.00 0.00 0.00 0.00 0.13 C 0.00 0.03 0.00 0.83 0.00 0.00 0.00 0.00 0.00 0.00 D 1.00 0.91 0.00 0.17 0.00 1.00 1.00 1.00 1.00 0.88 EST-1 (N) 19 27 28 8 26 20 2 10 2 4 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 B 1.00 0.46 0.00 0.69 0.00 1.00 1.00 1.00 0.50 0.75 C 0.00 0.13 0.93 0.31 1.00 0.00 0.00 0.00 0.50 0.25 D 0.00 0.41 0.07 0.00 0.00 0.00 0.00 0.00 0.00 0.00 GPI-1 (N) 21 31 27 9 29 20 2 10 2 4 A 0.00 0.00 0.11 0.00 0.02 0.03 0.00 0.00 0.00 0.00 B 1.00 0.95 0.89 1.00 0.79 0.98 1.00 1.00 1.00 1.00 C 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 D 0 0.048 0 0 0.19 0 0 0 0 0 ACO-1 (N) 21 31 30 9 29 20 2 10 2 4 A 0.00 0.00 0.03 0.00 0.02 0.00 0.00 0.00 0.00 0.00 B 0.00 0.03 0.05 0.00 0.00 0.15 0.50 0.00 0.25 0.00 C 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 D 1.00 0.97 0.90 1.00 0.98 0.85 0.50 1.00 0.75 1.00 E 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.00 0.00 0.00 HKI-1 (N) 21 31 30 9 29 20 2 10 2 4 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.25 B 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.38 C 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 0.38 Appendix 1.2 Allele frequencies: Anartia jatrophae.

O 1 S. 1 i 2 s 1 8 I T1 1 1 f 1 1 ■s 1 1 1 7 i 1 % I « I f 1 G O T -1 (N) 20 30 20 18 18 17 26 20 18 16 16 20 16 20 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 B 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.28 0.16 0.09 0.38 0.00 0.03 C 1.00 1.00 1.00 1.00 1.00 0.97 1.00 1.00 0.72 0.84 0.91 0.63 1.00 0.78 0 0.00 0.00 0.00 0.00 0.00 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.20 PGM (N) 20 30 20 16 18 19 26 20 18 16 16 20 10 20 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.03 0.00 0.03 B 0.08 0.00 0.00 0.00 0.00 0.08 0.35 0.00 0.00 0.00 0.00 0.00 0.05 0.08 C 0.00 0.08 0.10 0.38 0.19 0.92 0.40 0.83 0.44 0.19 0.28 0.33 0.25 0.50 D 0.93 0.02 0.90 0.63 0.81 0.00 0.25 0.18 0.53 0.81 0.72 0.65 0.50 0.40 E 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 F 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.03 0.00 0.00 0.00 0.20 0.00 LA (N) 20 30 18 16 15 19 26 19 17 14 13 20 12 20 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 B 0.00 0.02 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.03 0.13 0.13 C 1.00 0.98 0.94 0.94 1.00 1.00 1.00 1.00 1.00 1.00 1.00 0.98 0.88 0.88 D 0.00 0.00 0.06 0.06 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 PR (N) 20 30 20 16 18 19 26 20 18 14 15 20 16 20 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 B 0.00 0.00 0.00 0.00 0.00 0.03 0.06 0.00 0.61 0.21 0.40 0.08 0.06 0.13 C 1.00 1.00 1.00 1.00 1.00 0.97 0.92 1.00 0.39 0.75 0.60 0.93 0.94 0.88 0 0.00 0.00 0.00 0.00 0.00 0.00 0.02 0.00 0.00 0.04 0.00 0.00 0.00 0.00 ME (N) 20 29 18 16 17 19 26 19 18 14 16 20 16 20 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.03 0.00 0.03 B 0.48 0.47 0.44 0.28 0.32 0.76 0.65 0.68 0.58 0.61 0.16 0.60 1.00 0.55 C 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 D 0.53 0.53 0.56 0.63 0.68 0.24 0.35 0.32 0.42 0.39 0.84 0.38 0.00 0.43 E 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 F 0.00 0.00 0.00 0.09 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Appendix 1.2 Gene frequencies/IAiar/za jatrophae

2 2 o 1 o O ? a 1 1 1 Ï 1 1 f I 1 » 70 Tl I A 1 1 1 f Î ■S f I 1 1 FUM (N) 1 2 20 18 18 19 2 19 16 15 16 20 16 20 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 B 1.00 1.00 1.00 1.00 0.97 1.00 1.00 1.00 1.00 1.00 1.00 1.00 0.84 0.95 C 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0 0.00 0.00 0.00 0.00 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.16 0.05 HBDH (N) 1 2 20 16 18 19 2 20 18 15 16 20 12 20 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 8 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.13 0.03 C 1.00 1.00 0.98 0.97 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 0.88 0.98 D 0.00 0.00 0.03 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 G3PD (N) 20 30 20 16 18 19 26 20 18 16 16 20 16 20 A 0.00 0.00 0.03 0.00 0.00 0.00 0.00 0.00 0.03 0.00 0.00 0.13 0.00 0.08 B 1.00 1.00 0.98 1.00 1.00 1.00 1.00 1.00 0.97 1.00 1.00 0.88 1.00 0.93 6PGD (N) 20 29 20 16 18 14 26 19 18 14 16 19 11 20 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 B 0.1S 0.09 0.00 0.00 0.06 0.00 0.08 0.03 0.14 0.00 0.03 0.00 0.00 0.03 C 0.00 0.00 0.05 0.00 0.03 0.00 0.00 0.00 0.00 0.00 0.03 0.00 0.00 0.00 D 0 05 0.91 0.95 1.00 0.92 1.00 0.92 0.97 0.86 1.00 0.94 1.00 1.00 0.98 IDH-1 (N) 20 29 18 16 18 10 26 20 18 15 16 20 16 20 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 B 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.03 C 0.00 0.00 0.00 0.00 0.03 0.05 0.02 0.00 0.22 0.30 0.13 0.20 0.03 0.20 D 1.00 1.00 0.97 1.00 0.97 0.95 0.98 1.00 0.58 0.70 0.88 0.80 0.97 0.70 E 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 F 0.00 0.00 0.03 0.00 0.00 0.00 0.00 0.00 0.19 0.00 0.00 0.00 0.00 0.08 GPI (N) 20 29 20 16 18 19 28 20 18 16 16 20 16 20 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 B 0.03 0.05 0.03 0.03 0.06 0.08 0.12 0.00 0.31 0.53 0.28 0.18 0.06 0.03 C 0.00 0.95 0.98 0.97 0.89 0.92 0.89 1.00 0.69 0.47 0.72 0.70 0.72 0.90 0 0.00 0.00 0.00 0.00 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.13 0.22 0.05 E 0.00 0.00 0.00 0.00 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.03 Appendix 1.3 Heliconius charitonia

z s 0 c 1 2 S' s, 1 f 1 f 2 1 1 2 1 Locus 1 Î ; 1 1 1 1 t f I 1 i 1 GPI-1 (N) 24 24 24 24 4 24 21 5 23 24 23 20 2 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 B 0.08 0.02 0.15 0.00 0.00 0.02 0.05 0.00 0.00 0.00 0.00 0.00 0.00 C 0.25 0.40 0.40 0.15 0.25 0.25 0.48 0.10 0.80 0.88 0.00 0.48 0.00 D 0.42 0.40 0.31 0.81 0.25 0.56 0.48 0.70 0.20 0.13 1.00 0.53 1.00 E 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 F 0.25 0.19 0.15 0.04 0.50 0.17 0.00 0.20 0.00 0.00 0.00 0.00 0.00 IDH-1 (N) 24 24 24 24 5 24 21 5 24 24 24 20 2 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.75 B 0.06 0.04 0.00 0.00 0.00 0.19 0.00 0.00 0.00 0.00 0.00 0.00 0.00 C 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0 0.94 0.96 1.00 1.00 1.00 0.81 1.00 1.00 1.00 1.00 1.00 1.00 0.25 IDH-2 (N) 24 19 19 23 5 21 20 5 24 24 23 20 2 A 0.00 0.00 0.00 0.00 0.00 0.02 0.00 0.00 0.00 0.02 0.00 0.00 0.00 B 0.42 0.26 0.34 0.28 0.40 0.33 0.65 0.60 1.00 0.88 1.00 1.00 0.00 C 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0 0.58 0.74 0.66 0.72 0.60 0.64 0.35 0.40 0.00 0.08 0.00 0.00 1.00 E 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.02 0.00 0.00 0.00 PGM-1 (N) 24 23 24 24 5 24 20 5 24 24 24 20 2 A 0.00 0.00 0.10 0.00 0.00 0.00 0.00 0.00 0.00 0.02 0.00 0.00 0.00 B 0.02 0.02 0.02 0.29 0.10 0.00 0.00 0.50 0.00 0.02 0.00 0.03 0.00 C 0.75 0.67 0.77 0.67 0.60 0.19 0.88 0.40 0.96 0.94 1.00 0.98 0.00 D 0.15 0.28 0.08 0.04 0.30 0.69 0.13 0.10 0.02 0.02 0.00 0.00 1.00 E 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 F 0.08 0.02 0.02 0.00 0.00 0.13 0.00 0.00 0.02 0.00 0.00 0.00 0.00 G6P-1 (N) 24 15 18 22 5 18 21 5 23 23 22 20 2 A 0.04 0.00 0.00 0.00 0.00 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.00 B 0.96 1.00 1.00 0.59 0.40 0.97 1.00 1.00 1.00 1.00 1.00 1.00 1.00 C 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 D 0.00 0.00 0.00 0.41 0.60 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 Appendix 1.3 Allele frequencies; Heliconius charttonia c s 1 0 1 X ? s , Î I f t g 3 1 2 L o c u s 1 1 ; 1 1 1 I 1 1 I ( i 1 G O T -1 (N) 24 24 24 24 5 24 19 5 24 24 21 20 2 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 B 0.00 0.00 0.02 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.00 0.00 0.00 C 0.96 1.00 0.98 1.00 1.00 0.96 1.00 1.00 1.00 1.00 1.00 1.00 0.75 D 0.04 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.25 E 0 0 0 0 0 0 0 0 0 0 0 0 0 GOT-2 (N) 24 24 24 23 5 24 21 s 24 24 23 20 2 A 0.00 0.00 0.00 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.00 0.00 1.00 B 1.00 1.00 1.00 1.00 1.00 0.98 0.98 1.00 1.00 1.00 1.00 1.00 0.00 C 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 D 0.00 0.00 0.00 0.00 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.00 0.00 MEN-1 (N) 24 24 24 24 5 24 21 5 24 24 24 20 2 A 0.00 0.06 0.04 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 B 0.38 0.58 0.42 0.44 0.80 0.06 0.19 0.30 0.21 0.10 0.08 0.05 0.25 C 080 0.35 0.54 0.56 0.20 0.46 0.76 0.70 0.79 0.88 0.85 0.95 0.75 0 0.00 0.00 0.00 0.00 0.00 0.48 0.05 0.00 0.00 0.02 0.06 0.00 0.00 E 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 F 0.02 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 G3P-1 (N) 24 23 24 23 5 24 21 5 24 24 24 20 2 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 B 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 C 1.00 1.00 1.00 1.00 1.00 0.81 1.00 1.00 1.00 0.96 1.00 1.00 1.00 D 0.00 0.00 0.00 0.00 0.00 0.19 0.00 0.00 0.00 0.04 0.00 0.00 0.00 ADH-1 (N) 23 22 23 22 5 22 19 4 21 21 24 18 2 A 0.00 0.00 0.00 0.00 0.00 0.00 0.03 0.00 0.00 0.00 0.00 0.00 0.00 B 0.91 0.98 0.85 0.59 0.70 0.98 0.97 0.75 1.00 1.00 1.00 1.00 1.00 C 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 D 009 0.02 0.15 0.41 0.30 0.02 0.00 0.25 0.00 0.00 0.00 0.00 0.00 Appendix 1.3 Allele frequencies: Heliconius charitonia

z S O X c 1 ■b 2" 1 & Î 1 1 2 o Locus { \ I i 1 1 1 ÔT 1 1 2 [ i 1 M PI-1 (N) 24 23 23 24 5 22 21 5 22 23 23 20 1 A 0.02 0.04 0.02 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 1.00 B 0.65 0.72 0.85 0.04 0.00 0.75 0.91 0.90 0.39 0.33 0.09 0.00 0.00 C 0.29 0.24 0.13 0.96 1.00 0.25 0.05 0.10 0.00 0.04 0.04 0.10 0.00 D 0.04 0.00 0.00 0.00 0.00 0.00 0.05 0.00 0.61 0.63 0.87 0.90 0.00 LAL-1 (N) 24 24 24 24 5 24 21 5 24 24 24 20 2 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 1.00 B 0.02 0.00 0.02 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0 0.00 000 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0 0.08 1.00 0.98 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 0.00 LAL-2 (N) 18 16 18 14 5 18 16 5 17 18 12 16 2 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 B 0.06 0.06 0.08 0.04 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.25 C 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 D 0.94 0.94 0.92 0.96 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 0.75 LAL-3 (N) 24 24 24 24 5 24 21 4 24 24 24 20 2 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 B 1.00 0.98 0.98 1.00 1.00 1.00 0.95 1.00 1.00 1.00 1.00 1.00 0.25 C 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 D 0.00 0.02 0.02 0.00 0.00 0.00 0.05 0.00 0.00 0.00 0.00 0.00 0.75 Appendix 1.4 Allele frequencies: Dry as iulia.

1 0 2 g o 1 X c 1 o 0 a a 12 5! 5 ¥ ¥ 1 1 « 1 s I 5 ‘ L o c u s 1 1 I % 1 I 1 1 Î 1 2 I 1 1 1 1 1 1 01 G O T -1 (N) 1 17 12 20 20 14 20 18 6 24 19 22 11 16 9 6 7 19 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 B 1.00 1.00 0.06 0.68 1.00 0.96 0.95 0.97 0.00 0.00 0.00 0.00 0.96 0.00 0.00 0.00 0.00 0.05 C 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.03 0.00 0.00 0.00 0.00 D 0.00 0.00 0.04 0.00 0.00 0.04 0.05 0.00 1.00 1.00 1.00 1.00 0.05 0.09 1.00 1.00 1.00 0.95 E 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 F 0.00 0.00 0.00 0.03 0.00 0.00 0.00 0.03 0.00 0.00 0.00 0.00 0.00 0.88 0.00 0.00 0.00 0.00 GOT-2 (N) 1 17 12 20 20 14 20 18 6 24 19 22 11 16 9 6 7 20 A 0.00 0.00 0.08 0.00 0.00 0.04 0.00 0.03 0.00 0.00 0.00 0.00 1.00 0.38 0.67 0.00 0.07 0.08 8 1.00 0.07 0.75 0.08 0.93 0.06 1.00 0.97 1.00 1.00 1.00 1.00 0.00 0.63 0.33 0.92 0.86 0.75 C 0.00 0.00 0.00 0.00 0.08 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 D 0.00 0.03 0.17 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.08 0.00 0.08 E 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 F 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.07 0.10 PGM-1 (N) 1 17 12 20 20 14 20 18 6 24 19 22 11 16 9 6 7 19 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.17 0.00 0.03 B 0.00 0.18 0.08 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.05 0.00 1.00 0.83 0.64 0.47 C 1.00 0.77 0.92 0.93 0.98 0.89 0.03 0.97 1.00 1.00 1.00 0.96 0.96 1.00 0.00 0.00 0.36 0.50 D 0.00 0.03 0.00 0.05 0.00 0.11 0.98 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 E 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.05 0.00 0.00 0.00 0.00 0.00 0.00 F 0.00 0.03 0.00 0.00 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 HBD-1 (N) 1 17 12 20 20 14 20 18 6 24 19 22 11 16 9 6 7 5 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 B 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.09 0.00 0.00 0.00 0.00 C 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 0.84 1.00 1.00 1.00 0.90 0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.06 0.00 0.00 0.00 0.10 LAL-1 (N) 1 IS 12 16 20 14 17 17 4 24 19 22 11 16 9 6 7 17 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.06 B 0.00 0.00 0.00 0.00 0.00 0.07 0.00 0.00 0.00 0.02 0.00 0.07 1.00 0.97 0.89 1.00 0.93 0.77 C 0.00 0.07 0.00 0.09 0.00 0.14 0.06 0.03 1.00 0.98 0.87 0.93 0.00 0.03 0.11 0.00 0.07 0.18 D 1.00 0.03 1.00 0.88 1.00 0.79 0.94 0.97 0.00 0.00 0.13 0.00 0.00 0.00 0.00 0.00 0.00 0.00 E 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 F 0.00 0.00 0.00 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 00 CS Appendix 1.4 Allele frequencies: Dryas iulia

z ? 0 g 1 s a 1 ? f [ 1 "O 1 1 3 I Locus 1 1 1 1 [! 1 1 1 i 1 Î 1 1 1 È 91 LAL>3 (N) 1 10 12 17 20 11 16 18 6 22 19 22 11 16 9 6 7 17 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 B 0.00 0.00 0.00 0.00 0.00 0.09 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 C 0.50 0.85 0.75 0.85 0.98 0.91 0.97 1.00 0.00 0.00 0.00 0.00 1.00 1.00 1.00 1.00 1.00 1.00 D 0.50 0.15 0.25 0.15 0.03 0.00 0.03 0.00 1.00 1.00 1.00 1.00 0.00 0.00 0.00 0.00 0.00 0.00 ADH-1 (N) 1 17 12 19 20 12 19 16 6 24 13 21 8 12 3 6 1 10 A 0 0 0 0.026 0 0 0 0 0 0 0 0 0 0 0 0 0 0.00 B 1 0.041 0.958 0.921 1 1 0.921 1 1 1 1 1 0 0 0 0.083 0.5 0.45 C 0.00 0.03 0.00 0.03 0.00 0.00 0.05 0.00 0.00 0.00 0.00 0.00 1.00 1.00 1.00 0.92 0.50 0.50 D 0.00 0.03 0.04 0.03 0.00 0.00 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.05 GPI-1 (N) 1 16 12 20 20 14 18 18 6 24 19 22 11 16 9 6 7 18 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 B 0.00 0.00 0.00 0.00 0.15 0.07 0.06 0.14 0.00 0.00 0.08 0.00 0.23 0.03 0.00 0.00 0.21 0.08 C 1.00 0.94 1.00 0.88 0.85 0.93 0.89 0.86 1.00 1.00 0.90 1.00 0.77 0.63 0.89 1.00 0.36 0.47 0 0 0.031 0 0.075 0 0 0.056 0 0 ' 0 0.026 0 0 0.344 0.111 0 0.429 0.39 E 0.00 0.03 0.00 0.05 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.03 F 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.03 G6P-1 (N) 1 16 12 17 20 14 20 17 6 24 19 22 11 16 9 6 7 17 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 B 0.00 0.00 0.00 0.03 0.00 0.07 0.10 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.24 C 1.00 0.97 1.00 0.97 1.00 0.93 0.90 0.97 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 0.65 D 0.00 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.12 GAP-1 (N) 1 17 12 20 20 14 20 18 6 24 19 22 11 16 9 6 7 6 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 B 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.07 0.00 C 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 0.86 1.00 D 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.07 0.00 Appendix 1.4 Allele frequencies: Dryas iulia

z i 0 X 2 2 g 1 1 o ? a 1 I 1 \ 2 3 1 1 3 1 1 L o c u s 2 % I01 1 1 1 ? 1 1 1 f I % 1 I 1 G 3 P -1 s (N) 1 15 12 16 17 14 20 14 6 24 19 22 11 16 9 6 5 12 A 0.00 0.00 0.04 0.00 0.00 0.00 0.00 0.04 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.04 B 1.00 1.00 0.92 1.00 1.00 0.96 0.98 0.96 1.00 1.00 0.00 1.00 0.00 0.00 1.00 1.00 0.00 0.96 C 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 1.00 0.00 0.05 0.00 0.00 0.00 1.00 0.00 D 0.00 0.00 0.04 0.00 0.00 0.04 0.03 0.00 0.00 0.00 0.00 0.00 0.96 1.00 0.00 0.00 0.00 0.00 IDH-2 (N) 1 16 12 20 20 14 18 18 6 24 19 21 11 16 9 6 7 16 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 B 0.00 0.03 0.00 0.00 0.00 0.04 0.00 0.11 0.00 0.00 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.00 C 0.50 0.31 0.13 0.33 0.28 0.96 0.06 0.89 1.00 0.98 0.97 1.00 0.00 0.00 0.00 0.00 0.00 0.13 D 050 0.66 0.88 0.68 0.73 0.00 092 0.00 0.00 0.00 0.00 0.00 1.00 1.00 1.00 0.92 1.00 0.88 E 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 F 0.00 0.00 0.00 0.00 0.00 0.00 0.03 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.08 0.00 0.00 ENO-1 (N) 1 16 12 20 20 14 18 18 6 24 19 22 11 16 9 6 7 18 A 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.17 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 B 1.00 1.00 0.96 0.93 1.00 1.00 1.00 0.83 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 C 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 D 0.00 0.00 0.04 0.08 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 SOH-1 (N) 1 12 8 14 20 9 13 18 6 22 19 22 11 16 9 4 7 14 A 0.00 0.00 0.00 0.00 0.05 0.06 0.00 0.00 0.00 0.00 0.00 0.02 0.00 0.00 0.00 0.00 0.00 0.04 B 1.00 0.96 1.00 0.96 0.95 0.89 1.00 1.00 1.00 1.00 0.00 0.98 0.00 0.00 0.00 0.88 0.00 0.68 0 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00 1.00 0.00 0.00 0.00 0.00 0.00 0.14 0.00 D 0.00 0.04 0.00 0.04 0.00 0.06 0.00 0.00 0.00 0.00 0.00 0.00 1.00 1.00 1.00 0.13 0.86 0.29 o Appendix 2.1 Genetic distances: Battus polydamas.

■0 0 0 X c c c m S or c. to ? u I' to 1 . ? u 3 3" 3. 1 3 3 3 ? 1 to & 5 ‘ Jo' to c 6 s Ô c 3 1 ■O ir to to S 2 f A to Cuba, west ***** 0.048 0.114 0.047 0.135 0.009 0.056 0.103 0.046 0.061 Guadeloupe 0.215 0.116 0.001 0.139 0.023 0.002 0.076 0.001 0.024 Hispaniola 0.316 0.325 ***** 0.115 0.006 0.077 0.123 0.097 0.114 0.065 Florida 0.214 0.036 0.325 ***** 0.137 0.025 0.008 0.074 0 0.022 Jamaica 0.344 0.355 0.081 0.354 ***** 0.091 0.149 0.112 0.135 0.091 Barbados 0.157 0.195 0.288 0.202 0.309 ***** 0.022 0.077 0.024 0.026 Dominica 0.238 0.079 0.338 0.112 0.37 0.202 ***** 0.086 0.008 0.033 Puerto Rico 0.303 0.27 0.296 0.267 0.319 0.29 0.29 ***** 0.073 0.069 Martinique 0.212 0.039 0.323 0.016 0.351 0.199 0.113 0.265 ***** 0.021 Panama 0.249 0.178 0.287 0.175 0.297 0.215 0.207 0.266 0.172 *****

Below diagonal: Modified Roger's (1972) genetic distance (Wright, 1978) Above diagonal: Unbiased genetic distance (Nei, 1978) Appendix 2.2 Genetic distances: Anartia jatrophae.

0 O 5T ■0 c 0 c 5* X S n •n O' e 55’ o c. tr & 2. •D ? O o It 3» A 1 2 3 % O TX 3 s â 3 r- c 5' 1 5' 1It It 3. c 5' & 1 1 3 8 ST f o. : 7i T1 2 g' Cuba, west ***** , 0.003 0.02 0 0.041 0.027 0.015 0.038 0.016 0.024 0 0 0.018 0.014 Cuba, east 0.069 ***** 0.024 0.001 0.025 0.015 0.009 0.035 0.012 0.027 0.007 0.005 0.011 0.014 Guadeloupe 0.145 0.158 ***** 0.019 0.045 0.04 0.019 0.013 0.014 0.025 0.021 0.019 0.026 0.008 l^le of Pines 0.043 0.051 0.142 ***** 0.037 0.025 0.013 0.035 0.01 0.028 0.002 0.001 0.016 0.014 Hispaniola 0.198 0.16 0.200 0.191 0.001 0.013 0.04 0.049 0.023 0.046 0.042 0.009 0.028 Puerto Rico 0.105 0.125 0.198 0.157 0.045 ***** 0.009 0.038 0.039 0.02 0.032 0.028 0.007 0.022 Trinidad 0.131 0.109 0.143 0.125 0.125 0.111 0.018 0.02 0.017 0.019 0.016 0.007 0.007 Montserrat 0.193 0.187 0.124 0.186 0.2 0.195 0.139 0.015 0.039 0.039 0.037 0.03 0.017 Dominica 0.132 0.119 0.127 0.11 0.218 0.196 0.144 0.131 ***** 0.048 0.019 0.017 0.028 0.019 Grenada 0.157 0.166 0.162 0.168 0.156 0.146 0.138 0.194 0.213 ***** 0.025 0.023 0.014 0.018 Florida, R 0.041 0.09 0.15 0.056 0.21 0.177 0.144 0.198 0.141 0.16 ***** 0 0.018 0.017 Florida, F 0.028 0.077 0.142 0.042 0.2 0.166 0.136 0.191 0.133 0.155 0.029 ***** 0.018 0.014 Jamaica 0.134 0.111 0.16 0.129 0.102 0.091 0.099 0.172 0.165 0.124 0.136 0.133 ***** 0.016 StLucIa 0.126 0.127 0.101 0.124 0.169 0.153 0.098 0.136 0.143 0.141 0.135 0.126 0.131 *****

Below diagonal: Modified Roger's (1972) genetic distance (Wright, 1978) Above diagonal: Unbiased genetic disance (Nei, 1978) Appendix 2.3 Genetic distances: Heliconius charitonia

Z s 0 E "0 ■D c O e X- S c . 0 1 2 . •n 1 B) I' ? II 1 1 f 2 3 » S 5" : 3 7J o 1 I o' 3 1 8 u o 8 ir s 8 II 1 1 1 Cuba, west ***** 0.002 0.001 0.037 0.042 0.023 0.008 0.009 0.036 0.037 0.052 0.049 0.235 Cuba, east 0.07 ***** 0.002 0.045 0.043 0.023 0.016 0.017 0.05 0.051 0.076 0.071 0.227 Isle of Pines 0.062 0.07 ***** 0.06 0.057 0.03 0.007 0.012 0.04 0.04 0.069 0.062 0.25 Florida 0.185 0.204 0.214 ***** 0.015 0.068 0.066 0.044 0.004 0.092 0.082 0.086 0.258 Key Largo 0.21 0.212 0.237 0.145 ***** 0.075 0.082 0.067 0.097 0.097 0.113 0.105 0.288 Jamaica 0.152 0.15 0.168 0.245 0.266 ***** 0.033 0.026 0.076 0.074 0.087 0.089 0.207 Hispaniola 0.006 0.129 0.097 0.244 0.279 0.179 ***** 0.011 0.021 0.023 0.045 0.039 0.258 New Providence 0.122 0.15 0.132 0.213 0.264 0.172 0.131 ***** 0.047 0.051 0.052 0.058 0.242 Puerto Rico 0.187 0.218 0.197 0.29 0.304 0.263 0.145 0.222 ***** 0 0.03 0.01 0.316 Mona 0.188 0.219 0.197 0.287 0.304 0.26 0.15 0.228 0.035 0.035 0.01 0.312 Montserrat 0.223 0.266 0.255 0.275 0.328 0.283 0.207 0.232 0.172 0.183 0.009 0.285 SLKitts 0.217 0.258 0.242 0.28 0.316 0.285 0.194 0.244 0.101 0.104 0.096 0.301 H.peruvlana 0.446 0.442 0.458 0.465 0.491 0.422 0.47 0.450 0.514 0.51 0.496 0.506 *****

Below diagonal: Modified Roger's (1972) genetic distances (Wright, 1978) Above diagonal: Unbiased genetic distance (Nei, 1978) Appendix 2.4 Genetic distances: Dryas iulia.

Z

O cQ ■0 S g c % X 5 & C- ? 0 ai ? z A % S? ? 5! 6 3 r 3 3 I I 31 3 2 e ÔT 1 c 3* f 3 c 5' I s I 2. & 5' c 1 % ■g I) 2 % §■ f a* 1 1 F 1 1 a. A I Cuba, west 0.003 0.164 0.009 0.041 0 0.299 0.163 0.023 0.182 0.133 0.162 0.024 0.002 0.279 0.266 0.281 0.277 Cuba, east 0.078 ***** 0.177 0.012 0.05 0.004 0.291 0.175 0.041 0.195 0.142 0.174 0.044 0.005 0.288 0.275 0.264 0.287 Guadeloupe 0.384 0.397 ***** 0.13 0.252 0.159 0.424 0 0.157 0.254 0.186 0 0.134 0.187 0.36 0.338 0.482 0.104 Florida 0.1 0.114 0.345 ***** 0.071 0.009 0.336 0.128 0.025 0.232 0.175 0.127 0.025 0.015 0.344 0.319 0.318 0.238 Jamaica 0.2 0.219 0.466 0.257 ***** 0.047 0.345 0.25 0.069 0.187 0.157 0.248 0.084 0.05 0.289 0.284 0.329 0.375 Isle of Pines 0.045 0.079 0.377 0.096 0.211 ***** 0.29 0.157 0.022 0.194 0.134 0.157 0.023 0.001 0.295 0.271 0.278 0.273 StLucia 0.49 0.486 0.578 0.519 0.524 0.482 ***** 0.421 0.355 0.203 0.165 0.417 0.336 0.298 0.159 0.144 0.068 0.422 Dominica 0.382 0.395 0.006 0.343 0.465 0.375 0.576 ***** 0.156 0.251 0.184 0 0.133 0.185 0.357 0.335 0.478 0.104 Puerto Rico 0.154 0.2 0.377 0.156 0.286 0.15 0.53 0.375 ***** 0.254 0.185 0.154 0.002 0.024 0.367 0.336 0.337 0.265 Grenada 0.399 0.412 0.47 0.447 0.406 0.409 0.42 0.467 0.464 ***** 0.034 0.245 0.236 0.206 0.069 0.119 0.242 0.38 Panama 0.337 0.348 0.401 0.383 0.366 0.336 0.371 0.396 0.392 0.195 ***** 0.18 0.168 0.143 0.076 0.084 0.209 0.294 Montserrat 0.38 0.393 0.019 0.341 0.461 0.373 0.572 0.017 0.373 0.461 0.393 ***** 0.131 0.184 0.351 0.329 0.474 0.103 Hispaniola 0.156 0.207 0.35 0.158 0.279 0.154 0.515 0.348 0.067 0.448 0.374 0.345 ***** 0.027 0.338 0.314 0.316 0.237 New Providence 0.065 0.085 0.409 0.121 0.219 0.057 0.494 0.407 0.154 0.424 0.352 0.405 0.166 ***** 0.306 0.276 0.278 0.297 St.Vincent 0.481 0.487 0.545 0.528 0.491 0.491 0.376 0.543 0.542 0.242 0.269 0.537 0.522 0.504 ***** 0.1 0.155 0.434 Trinidad 0.463 0.47 0.524 0.504 0.479 0.465 0.355 0.521 0.514 0.33 0.274 0.516 0.497 0.475 0.304 ***** 0.189 0.253 Martinique 0.483 0.471 0.614 0.512 0.52 0.479 0.253 0.612 0.525 0.458 0.417 0.608 0.506 0.484 0.374 0.405 0.477 StKitts 0.48 0.487 0.312 0.451 0.547 0.474 0.572 0.312 0.473 0.552 0.481 0.31 0.448 0.497 0.583 0.458 0.606 *****

Below diagonal: Modified Roger’s genetic distance (Wright, 1978) Above diagonal: Unbiased genetic distances (Nei, 1978) CHAPTER 4

THE HISTORICAL BIOGEOGRAPHY OF THE WEST INDIAN

BUTTERFLY DRYAS IULL4 (LEPIDOPTERA: HELICONHDAE)

ABSTRACT

Genetic differences among sequences of a 609 base pair region of mitochondrial c o n reveal a strong phylogenetic break in Dryas iulia. Eleven substitutions,

>2.5% sequence divergence, separate the North West Indies from the continent/South West Indies. Variation within the two clades is low, <1.5% sequence divergence, but some structure can be discerned. The North West Indies are spht into two groups: a single transition is shared by Greater Antillean populations, whilst the North Lesser Antilles share a single transversion. Within the continental/South West Indian clade the Central Lesser Antilles are distinguished by three transitions. A previous allozyme study identified the same phylogenetic pattern. The evidence suggests that Dryas iulia colonised the Lesser

Antilles relatively recently from South America on at least two separate occasions.

The Greater Antilles were probably colonised some 2.5 myrBP, either directly from the continent, or indirectly via the Lesser Antilles from South America. This level of divergence is too low to have been caused by the geological separation of the

West Indies and the continent. The phylogenetic structure of Dryas iulia in the

West Indies seems to be a result of dispersal not ancient vicariance.

134 INTRODUCTION

Islands are usually considered evolutionary dead-ends; they are merely passive

sinks dependent upon mainland sources for their faunas. The isolation of islands

may lead to spéciation; endemics, however, are often seen as mere evolutionary

curiosities which are vulnerable to invasion by fitter continental forms (Berry,

1986). Many island species may therefore have a limited life-expectancy. These

assunq)tions, imphcit in much of the biogeographical hterature (MacArthur and

Wilson, 1967; Ricklefs and Cox, 1972; Lack, 1976) have rarely been tested

en^irically.

There are reasons for expecting archipelagos to be evolutionarily active. In his

shifting balance theory, Wright (1977) proposed that population subdivision is necessary for adaptive evolution. Large continuous populations will get stuck on

adaptive peaks and only small subpopulations can reach new peaks as genetic drift

takes them off the old one. Equipped with fresh adaptations, such subpopulations

spread through the whole population. Subdivided populations, including

archipelagos, may therefore be the engines of evolution.

Indeed, recent genetic studies of the West Indian avifauna reveal a corrq)lex

evolutionary situation (Klein, 1992; Seutin etal., 1993, 1994). Seutm et al. (1994)

suggested that populations may go through periods of invasiveness and relative geographical quiescence, that different populations of the same species may be in

135 différent phases of colonising activity at any given time, and that species might sometimes originate in the islands and then invade the mainland.

Another volant West Indian group, butterflies, has also been the subject of genetic investigation. In Chapter 3 allozyme variation was surveyed in four butterfly species throughout the West Indies. One of these butterflies, Dryas iulia, showed complex structure and here I continue to investigate its historical biogeography, this time using mtDNA sequence analysis. Such analyses permit finer resolution of phylogenetic patterns and intraspecific historical biogeography (Bermingham and

Avise, 1986; Avise et ai, 1987). Furthermore, through the molecular clock it is possible to estimate the actual time scale involved (Zuckerkandl and Pauling,

1965). Such information might reveal the direction and sequence of colonisation, and whether ancient geological events or more recent dispersal have shaped the structure of West Indian Dryas iulia.

Dryas iulia is distributed throughout the continental neotropics, south Florida and the West Indies. Phenotypic studies have identified twelve island races in the West

Indies (Smith et a l, 1994). Allozyme data (Chapter 3) revealed additional structure, with four genetic groups characterised by differences in allele fi*equency:

Greater Antilles (GA), North Lesser Antilles (NLA), Central Lesser Antilles

(CLA), and continent/South Lesser Antilles (SLA/CO). Average genetic distances,

D (Nei, 1978) were much greater between these groups (0.18

136 In Chapter 3 it was suggested that Dryas iulia might have colonised the West

Indies in a ‘pincer movement’ from Central America eastwards through the Greater

Antilles, and from South America northwards into the Lesser Antilles. Under this scenario, the South American population is now in a phase of expansion. The

CLA clade represents the remnants of a wave of colonisation by an earher continental population which is now in a period of geographical quiescence. The current continental population has already invaded the South Lesser Antilles and is replacing the CLA clade as it progresses up the Lesser Antillean chain. The

Greater Antilles were colonised from Central America and this population eventually spread to the North Lesser Antilles and may still be progressing southward.

In this preliminary study I use mtDNA sequence data to address two main questions. First, does the mtDNA variation confirm the allozyme pattern?

Second, does the molecular clock suggest separation times indicative of ancient vicariance?

METHODS

Samples

A total of 22 individuals were sequenced from twelve island and two continental locahties (Table 1, Figure 1). An additional two sequences (from Ecuador and

Costa Rica) were obtained from Brower (1994a) for use in the phylogenetic

137 analysis. The specimens were collected and stored as described in Chapter 3.

Thoraxes were ground in liquid nitrogen and genomic DNA was extracted using the protocol of Harrison et ai, (1987).

PCR and Sequencing

A 929 base pair section of mitochondrial DNA was anq)lified from individual genomic DNA via PCR (Saiki et a l, 1988). This contained part of the cytochrome oxidase subunit I (COI) protein coding gene, the entire tRNA-leu coding region and most of the cytochrome oxidase subunit II (COE) gene. Oligonucleotide primers used to initiate PCR were S2792 and A3772 (Table 2) (Brower, 1994a).

Double stranded PCR products were sequenced directly with the Taq DyeDeoxy

Terminator Cycle Sequencing system (Applied Bio systems). The sequence reaction was initiated using the flanking primers S2792 and A3772 and an internal primer A3568 (Table 2). A 609 base pair section was sequenced. This section corresponds to positions 3120-3729 in the Drosophilayakuba sequence (Clary and

Wolstenhohne, 1985) and contains most of the protein coding region of the COE gene.

Phylogenetic analysis

Unrooted phylogenetic trees were produced by Neighbour-joining using the software package MEGA (Kumar et al, 1993). Pairwise difference matrices were constructed using the Jukes-Cantor (Jukes and Cantor, 1969) and Tamura

(Tamura, 1992) methods. The Tamura distance corrects for a bias in the number of transitions conq)ared to transversions, which is common in animal mtDNA

138 (Brown et al, 1982), and also for sequences, such as insect mtDNA, which are typically A+T rich (Liu and Beckenbach, 1992). Confidence levels for the nodes of the Neighbour-joining tree were estimated via 500 bootstrap rephcations of the original data matrix using MEGA (Felsenstein, 1985; Kumar et al., 1993).

Maximum parsimony trees were also constructed using the software package

PAUP (3.0s, SwoflFord, 1993). A subset of all the possible trees was determined by an heuristic search. The tree bisection and reconnection (TBR) branch swapping algorithm was then used to search for the optimal trees.

The distribution of sequence variation with respect to geographical region was explored using Gst analysis, where Gst =1- Hw/Hb, Hw is the average of all pairwise difterences among individuals from the same subpopulation, and Hb is the average of all pairwise differences among individuals from different subpopulations. The significance of the observed Gst value (G^/obs) was assessed by conq)arison with 1000 randomly generated Gst values (Gs^ran). The original data matrix was randomly divided into demes, equivalent in size and number to the deme structure which was tested to give Gstohs- The randomly generated structure was then analysed and the resulting Gj^an conq)ared with G^^obs (program available from S.R. Palumbi, University of Hawaii). Genetic heterogeneity among regions was considered to be significant if G^/obs was greater than 95% of the G st^ values.

139 RESULTS

An overlapping 609 base pair region of mitochondrial COE varied at 32 different sites in 24 individuals (Figure 2). The majority of these differences (28) were silent third position changes at two- or four-fold degenerate sites. The remaining 4 differences were at first (3) and second (1) positions and led to amino acid substitutions (Figure 3). All the amino acid substitutions occurred in single individuals.

A+T content was high (76%), and there was a strong transition/transversion bias

(8:1). Of the 5 sites at which transversions occurred, 4 were between A and T, as might be expected in an A+T rich region. Such results are typical for insect COE with low levels of divergence between individuals (Liu and Beckenbach, 1992;

Sperling and Hickey, 1994; but see Brower, 1994a).

Maximum parsimony and nei^bour-joining methods (the latter using either Jukes-

Cantor or Tamura distances) both produced identical trees (the neighbour-joining tree is shown in Figure 4). The major feature of the tree is a >2.5% phylogenetic break between the North West Indies and the continent/South West Indies. This node is strongly supported in all 500 bootstrap trees and was characterised by eleven substitutions (including one transversion).

Within each clade mtDNA variation was low (<1.5%). There was, however, evidence of phylogenetic structure. Within the North West Indies clade,

140 individuals from the North Lesser Antilles (NLA) all shared a single transversion, whilst those from the Greater Antilles (GA) shared a single transition. Individuals from New Providence (Bahamas) had neither of these substitutions. Though bootstrap support for these clades was low, Gst analysis (with GA and NLA populations as separate demes) indicated significant structure {Gst = 0.5; p<0.001) with more variation occurring between demes (d= 0.0059) than within them

(ûf<0.0036).

There was also some evidence of phylogenetic structure within the Greater

Antilles. Hispaniola, Cuba and Jamaica share a transition which distinguishes them from the Isle of Pines and Puerto Rico. Furthermore, within this clade the two

Cuba individuals (one from either end of the island) share another transition separating them from Hispaniola and Jamaica.

More marked structure was found in the South West Indies/continental clade, where the two individuals from St.Lucia, in the Central Lesser Antilles (CLA), shared 3 transitions. The individual from Panama was also quite distinct from the rest of the continent/South West Indies clade, sharing 5 transitions with the North

West Indies clade.

141 DISCUSSION

The results clearly show a major phylogenetic break between Dryas iulia populations in the North West Indies and those in the continent/South West Indies.

There is also evidence of additional structure within these clades. Four groups can be identified: the North Lesser Antilles, Greater Antilles, Central Lesser Antilles and continent/South Lesser Antilles. This phylogenetic structure exactly matches the pattern found with the aUozymes in Chapter 3 (Figure 4).

Separation times can be estimated from mtDNA divergence assuming the action of a molecular clock. Comparing data on mtDNA evolution between various closely related arthropod taxa, Brower (1994b) estimated a constant mutation rate of 1.1-

1.2% per million years for silent sites. Dryas iulia in the North West Indies, appears to have been isolated from the continent/South West Indies for around 2.5 myrBP. This is nowhere near long enough to have been caused by the geological separation of the West Indies from the continent, which is sometimes cited in the biogeography of West Indian butterflies (Miller and Miller, 1989). Furthermore, the low levels of divergence within the major clades indicate relatively recent dispersal events rather than vicariance caused by ancient island movements.

The mtDNA data is consistent with the colonisation hypotheses put forward in

Chapter 3. The Lesser Antilles, appear to have been colonised in at least two waves from South America. The first wave of colonisation reached at least

Martinique and was subsequently replaced in the southern islands (St.Vincent and

142 Grenada) by the second wave. Such lineage turnover can occur in two ways: new

lineages may colonise islands aheady vacated by extinction events, or they may replace lineages through introgression and subsequent lineage sorting. In the latter

case we would expect to find a cline at the leading edge of the new lineage’s range.

The analysis of more individuals fi*om St. Lucia and St.Vincent will test this hypothesis.

Unfortunately the resolution is still inadequate to discern directionahty or phylogenetic relationships within the major clades. We cannot tell whether the

Greater Antilles were colonised directly fi*om the continent or via the Lesser

Antilles; indeed the data would also be consistent with Dryas iulia originating in the Greater Antilles and then colonising the continent.

Whichever direction colonisation occurred the evidence does not support a vicaiiant origin of West Indian Dryas iulia. The sequence divergence between the two major clades indicates a separation some 2.5 myrBP. This is much too recent to be explained by the ancient movement of the Greater Antilles into the Caribbean

(Miller and Miler, 1989). Below the major break, another period of ‘phylogenetic activity’ seems to have occurred <1 myrBP, when the North Lesser Antilles spht fi'om the Greater Antilles and the Central Lesser Antilles spht fi’om the continent/South Lesser AntiUes.

This differentiation may have resulted fi’om the climate changes and fluctuations in sea level associated with Pleistocene glaciations (MiUer et al., 1992). Although it

143 is unclear exactly how individual islands were affected, the Puerto Rican bank was probably exposed at various times and some of the North Lesser Antilles might have been somewhat larger. This may have enabled Greater Antillean Dryas iulia to spread to the North Lesser Antilles (or vice versa). Similarly the Grenadine bank was also largely exposed increasing the population size of Dryas iulia in the

South Lesser Antilles, and possibly precipitating a wave of colonisation into the

Central Lesser Antilles.

The genetic structure of Dryas iulia, although still relatively sketchy, shows some similarity to patterns observed in the West Indian aviûuna (Klein, 1992; Seutin et al., 1993, 1994). For exanq)le, in the bananaquit {Coereba flaveold) there is a strong phylogenetic break between Jamaica and populations in the eastern West

Indies/continent, as well as a less marked break between populations in the southern and central Lesser Antilles (Seutin et al, 1994). Whilst the variation might not be exactly concordant, the population genetic structure of both birds and butterflies is conq)lex and suggests phases of range expansion, contraction, or quiescence (particularly during the Pleistocene). Such patterns were anticipated in taxon cycle theory (Wilson, 1961, Ricklefs and Cox, 1972) although at the species level rather than among intraspecific populations.

Investigations into the population genetics of these species may help to elucidate microevolutionary processes caused by population subdivision. One hypothesis is that local adaptation and selection drives this 'lineage turnover’. Conq)aring patterns of mitochondrial and nuclear (allozyme) variation tests the likelihood that

144 selection influences the historical biogeography of these species. It is inq)robable that selection would act in a similar manner on nuclear and mitochondrial loci.

Thus concordance in geographical variation between the two measures supports their selective neutrahty and suggests demographic causes. However, it might also be evidence that the major clades are actually separate species i.e. reproductively isolated.

Future work

With the analysis of more individuals from each population and the sequencing of a more variable region I hope to clarify the phylogenetic relationships among West

Indian populations of Dryas iulia. Whilst a more variable region might reveal the order in which colonisation occurred, the problem is distinguishing whether Dryas iulia originated in the Greater Antilles or the continent i.e. determining directionahty. This is difihcult without a rehable root to the tree. Unfortunately

Dryas is a monospecific genus and Dryas iulia is quite distinct from the other

Hehconiidae (Brower, 1994a). Outgroups are, therefore, likely to be ineffective roots because of the high probabihty of multiple hits.

The relatively large sample sizes available, however, aUow the appHcation of techniques which estimate the relative ages of populations. Rogers and

Harp ending (1992) showed that the distribution of pairwise genetic differences between individuals can be used to infer past episodes of rapid population growth and decline. Such episodes are characteristic of colonisation events which are usually associated with population bottlenecks followed by rapid expansions. This

145 technique provides a means for estimating the relative ages of island populations and deducing ancestor-descendent relationships.

BIBLIOGRAPHY

Avise, J.C., Arnold, J., Ball, R.M., Bermingham, E., Lamb, T., Neigel, I.E., Reeb, C.A. and Saunders, N.C. 1987. Intraspecific phylogeny: the mitochondrial DNA bridge between population genetics and systematics. Ann. Rev. Ecol. Syst. 18:489-522.

Bermingham, E. and Avise, J.C. 1986. Molecular zoogeography of fi'eshwater fishes in the southeastern United States. Genetics 113:939-965.

Berry, RJ. 1986. Genetics of insular populations o f mammals, with particular reference to differentiation and founder effects in British small mammals Biological J. Linn. Soc. 28:205-230.

Brower, A.V.Z. 1994a. Phylogeny of Heliconius butterflies inferred fi*om mitochondrial DNA sequences (Lepidoptera: Nymphahdae). Mol. Phylogenet. Evol. 3:159-174.

. 1994b. Rapid morphological radiation and convergence among races of the butterfly Heliconius erato inferred from patterns of mitochondrial DNA evolution. Proc. Natl. Acad. Sci. U.S.A. 91:6491-6495.

Brown, W.M., Prager, E.M., Wang, A. and Wilson, A C 1982. Mitochondrial DNA sequences of primates: tempo and mode of evolution. J. Mol. Evol. 18: 225- 239.

146 Clary, D O and Wolstenhohne, D R. 1985. The mitochondrial DNA molecule of Drosophila yakuba'. nucleotide sequence, gene organisation, and genetic code. J. Mol. Evol. 22:252-271.

Felsenstein, J. 1985. Phylogenies and the con^arative method. Am. Nat. 125:1- 15.

Harrison, R G , Rand, D M. and Wheeler, W.C. 1987. Mitochondrial DNA variation in field crickets across a narrow hybrid zone. Mol. Biol. Evol. 4:144- 158.

Jukes, J.H. and Cantor, C.R. 1969. Evolution of protein molecules. In Munro, H.N. (ed.) Mammalian Protein Metabohsm. Academic Press, NY. pp. 21-132.

Klein, N.K. 1992. Historical processes and the evolution of geographic variation patterns in the Yellow Warbler {Dendroica petechia). Ph D. dissertation. University of Michigan, Ann Arbor, Michigan.

Lack, D. 1976. Island Biology, Illustrated by the Land Birds of Jamaica. Blackwell.

Kumar, S., Tamura, K. and Nei, M. 1993. MEGA: Molecular Evolutionary Genetics Analysis, version 1.01. The Pennsylvania State University, University Park, PA 16802.

Liu, H. and Beckenbach, A T 1992. Evolution of the mitochondrial cytochrome oxidase II gene among 10 orders of insects. Mol. Phylogenet. Evol. 1:41-52.

MacArthur, RH. and Wilson E.O. 1967. The Theory of Island Biogeography. Princeton University Press, NJ.

147 Miller, L D and Miller J.Y. 1989. The biogeography of West Indian butterflies. In: Biogeography of the West Indies. Woods, C.A. (ed.). Sandhill Crane Press, Inc. Gainesville FL. pp. 229-262.

Miller, L.D., Simon, M.J. and Harvey, D.J. 1992. The butterflies (Insecta: Lepidoptera) of Crooked, Acklins, and Mayaguana Islands, Bahamas, with a discussion of the biogeographical afGnities of the southwestern Bahamas and a description of a new subspecies by H.K. Clench. Ann. Carnegie Mus. 61:1-31.

Nei, M 1978. Estimation of average heterozygosity and genetic distance from a small number of individuals. Genetics 89:583-590.

Ricklefs, R.E. and Cox, G.W. 1972. Taxon cycle in the West Indian avifauna. Am Natur. 106:195-219.

Rogers, A R. and Harp ending, H. 1992. Population growth makes waves in the distribution of pairwise genetic differences. Mol. Biol. Evol. 9(3): 552-569.

Saiki, R.K., Gelfand, D.H., Stoffel, S., Scharf^ S.J., Higuchi, G T , Horn, G T , MuUis, K.B. and Ehrlich, H.A. 1988. Primer-directed enzymatic an^lification of DNA with a thermostable DNA polymerase. Sci. 239:487-491.

Seutin, G , Brawn, J., Ricklefs, R.E and Bermingham, E. 1993. Genetic divergence among populations of a tropical passerine, the streaked saltator (Saltator albicollis). Auk 110: 117-126.

Seutin, G , Klein, N.K., Ricklefs, R.E , and Bermingham, E. 1994. Historical biogeography of the bananaquit {Coereba flaveold) in the Caribbean region: a mitochondrial DNA assessment. Evol. 48: 1041-1061.

Smith, D , Miller L., and Miller, J. 1994a. The Butterflies of the West Indies and South Florida. Oxford Universtiy Press.

148 Sperling, A.H. and Hickey, D.A 1994. Mitochondrial DNA sequence varitiation in the Spruce Budworm species complex (C/ïom/o«gMra:Lepidoptera). Mol. BioL Evol. 11(4): 656-665.

SwofiFord, D.L. 1993. “PAUP: Phylogenetic Analysis Using Parsimony,” Version 3.0s, Illinois Natural History Survey, Chanq)aign.

Tamura, K. 1992. The rate and pattern of nucleotide substitution in Drœophila mitochondrial DNA. Mol. Biol. Evol. 9:814-825.

Wilson, E.O. 1961. The nature of the taxon cycle in the Melanesian ant fauna. Am. Nat. 95:169-193.

Wright, S. 1977. Evolution and the Genetics of Populations, voL 3. Experimental results and evolutionary deductions. Chicago: University of Chicago Press.

Zuckerkandl, E. and Pauling, L. 1965. Evolutionary divergence and convergence in proteins. In Evolving Genes and Proteins. Bryanson, V., and Vogel, H.J. (eds). Academic Press, New York. pp.97-106.

149 Table 1. Sançling data.

Region Locale Sample size Greater Antilles Cuba, west 1 Cuba, east 1 Isle of Pines 2 Jamaica 2 Hispaniola 2 Puerto Rico 1 Bahamas New Providence 2 North Lesser Antilles St. Kitts 2 Montserrat 1 Guadeloupe 1 Dominica 2 Central Lesser Antilles St. Lucia 2 Grenada 1 Central America Panama 1 Costa Rica* 1 South America Brazil 1 Ecuador* 1

* from Brower (1994a)

150 Table 2. Oligonucleotide primers.

Name Primer Sequence

S 2 7 9 2 5 ' ATA CCT CGA CGT TAT TCA GA 3 '

A 3 5 6 8 5 ' CCT AAI GAA/T GGA ATA/T GTT CA 3'

A 3 7 7 2 5 ' GAG ACC ATT ACT TGC TTT CAG TCA TCT 3 '

Note: Names refer to sense (S) or antisense (A) and the position of the 3’ end in the complete Drosophila yakuba mtDNA sequence (Clary and Wolstenholme, 1985).

151 fN

Figure 1. The West Indies. Islands sampled are underlined and number of indivduals sequenced are shown in parentheses.

Bahamas ATLANTIC OCEAN

Yucatan (Mexico)

^VHispaniPla (,2)

M o n a Jamaica (2)

Monuerral rn * LESSER GREATER ANTILLES ^ ANTILLES f i i n r i f l n i in a f n CENTRAL AMERICA Dominic» ^ M » ltiju q u o

S L V in c e n lQ

Trinidad

PACIFIC OCEAN 200 500 KM SOUTH AMERICA (2) Figure 2. Nucleotide variation.

Variable rites 112 222 223 333 344 444 445 555 555 55 558 881 478 991 267 822 467 890 244 556 78 Reg on Locale 477 073 905 572 702 119 354 327 809 257 95 South Amedca Ecuador TAA GTT TCG TTC TCT TAG TTA c n CTC ATC CC B rsàl .A. . .T G.? Cmtral Amedca Costa Rica .A. A. . . .T Panama .A..T. .C..C. .CT T. South Lesser Aitilles Grenada .A. Central Lesser Antilles StLuda 1 .G. AA, .C . C. . . .T ? StLuda2 .G. AA. ? Greater Antilles Cuba, west ??? ??C CTA .CT .T.C.A . .G TCC TCT .C? TT Cuba, east C.C ?AC CTA .CT .T . C.A ..G TCC TCT .C? TT Ide of Pines 1 C.C AAC CTA .CT .T. . .A . .G TCC T.T .CT TT We cf Pines 2 C.C AAC CTA .CT .T . C.A ..G T-C T.T.C? T. Hspanida 1 C.C AAC CTA .CT .T . C.A . .G TCC T.T .CT TT Hsponola2 C.C AAC CTA CCT .T. C.A . .G TCC T.T .CT TT Jamaica 1 C.C AAC CTA .CT .T . C.A ..G TCC T.T .CT TT Jamaica 2 C.C ????????? .T . C.A . .G TCC T.T .CT TT Puerto Rico C.C ??? CTA .CT .T . . .A . .G TCC T.T .CT T. NewProwdaice 1 C.C AAC CTA .CT ..A ..G TCC T.T .CT T. New Providence 2 C.C AAC CTA .CT .?A . .G TCC T.T .C? T. North Lesser Antilles Mxitserrat C.C AAC CTA .CT . .A . .A . .G TCC T.T .CT TT StKitts 1 C.C AAC CTA .CT . .A . .A . .G TCC T.T .C . TT SL Kitts 2 C.C AAC CTA .CT . .A .TA • *G TCC T.T .C? T. Guadeloupe C.C AAC CTA .CT . .A . .A . ,G TCC T.T.C? TT Dominica 1 C.C AAC CTA .CT . .A . .A . .G TCC T.T .c. TT Dominica 2 C.C AAC CTA .CT . .A . .A , .G TCC T.T .c. TT

153 Figure 3. Amino acid variation

Variable sites* 1244 8924 Region Locale 7513 South America Ecuador SYIV Brazil T. . . Central America Costa Rica T. .E Panama T. . . South Lesser Antilles Grenada T. . . Central Lesser Antilles St.Lucia 1 T. . . St.Lucia 2 T. . . Greater Antilles Cuba, west ?... Cuba, east T. . . Isle of Pines 1 T. . . Isle of Pines 2 T. . . Hispaniola 1 T. . . Hispaniola 2 TH. . Jamaica 1 T. . . Jamaica 2 ??. . Puerto Rico ??. . New Providence 1 T. . . New Providence 2 T. ?. North Lesser Antilles Montserrat T?. . St.Kitts 1 T. . . St.Kitts 2 T.F. Guadeloupe T. . . Dominica 1 T. . . Dominica 2 T. . .

♦site numbers correspond to nucleotide substitution site.

154 Figure 4. Neigbour-joining (mtDNA) and UPGMA (allozyme) trees for Dryas iulia.

Numbers in parentheses are mtDNA sample sizes. Numbers above branches represent bootstrap confidence levels.

73 Cuba, west Cuba (2) C l Isle of Pines

Hispaniola (2) I New Jamaica (2) Providence Cuba, east Isle of Pines I Florida Puerto Rico 100 isle of Pines Puerto Rico

New Hispaniola Providence (2)

Jamaica

Guadeloupe ? Guadeloupe Dominica (2) Dominica

Montserrat II Montserrat 45

StKitts (2) StKitts

98 Martimque StLucia (2)

StLucia if%

Grenada 96

Costa Rica Panama Brazil II StVincent ïl Ecuador Grenada Trinidad Panama ï |

0 1 .00 .07 .13 .20 .27 ^sequence divergence Distance (Nei, 1978) mtDNA AUozymes

Transversion Transition

155 CHAPTER 5

THE ISLAND BIOGEOGRAPHY OF GENETIC VARIATION

ABSTRACT

According to neutral theory, the genetic diversity of insular populations at equilibrium depends upon their effective population size. Island area usually determines the size of insular populations, so genetic diversity should be correlated with island area. Gene flow also increases effective population size.

Since gene flow decreases with isolation, more remote islands should be less diverse. Remote islands will also experience more severe bottlenecks at colonisation, further diminishing their current genetic diversity. Island area and isolation both therefore influence the genetic diversity of insular populations. According to the stochastic theory of island biogeography species richness is similarly determined by area and isolation. Species richness and genetic diversity are expected to be positively correlated if both depend upon neutral stochastic factors. Selection and deterministic fectors, however, can disrupt this relationship and even lead to a negative correlation. The genetic diversity of four species of West Indian butterfly is investigated using data from a recent allozyme survey. Species richness is significantly correlated with genetic diversity in three of the four species. This suggests that neutral stochastic forces shape biodiversity at both the species and the

156 population genetic level. Although the fourth species shows a negative correlation between species richness and genetic diversity, this does not seem to result from selection or other deterministic influences. It is more likely to be a consequence of the difrering dispersal abihties among West Indian butterflies.

INTRODUCTION

Biogeography seeks to explain the distribution of species. Biodiversity, however, is not restricted to species but is also found among conspecific populations - as genetic diversity. The increasing sophistication of genetic technology has made it easier to assess this intraspecifrc variation and it is now possible to investigate the biogeography of genetic diversity. Archipelagos are usefiil systems for studying both inter- and intraspecific biogeography. Islands form discrete entities allowing physical and biotic parameters to be clearly defined. This chapter investigates the nature and causes of genetic variation among insular populations.

The simplest model predicting the genetic diversity of populations is the neutral theory (Kimura, 1983). When selection has no significant influence, genetic diversity reaches an equihbrium value which is determined by effective population size {Né) and mutation rate (//) according to the relationship: H = 4Ne/j. MNefj, +J whereH is heterozygosity and // the mutation rate for neutral alleles per locus per generation (from Kimura, 1983 p.320). It can be assumed that // is approximately

157 constant for a given set of loci in conspecific populations. Under these conditions the uneven distribution of genetic diversity among insular populations results from differences in Ne.

Island area determines Ne, by limiting available resources. The neutral theory therefore predicts a positive correlation between island area and genetic diversity, since larger islands support larger populations. Ne is also affected by the rate of gene flow: high levels of gene flow lead to greater Ne. If distance is a barrier to dispersal, gene flow from source populations wiU be lower on more remote islands.

Consequently, genetic diversity should be negatively correlated with insular isolation. The neutral theory therefore predicts that the genetic diversity of insular populations, at equihbrium, will depend on both island area and isolation.

Many insular populations, however, might not be at equihbrium Historical bottlenecks in population size can diminish current levels of genetic diversity.

Bottlenecks reduce the average effective population size, Ne_. As ^ is the harmonic mean of the actual numbers present in each generation, past troughs in population size have a disproportionately large effect (Haiti and Clark, 1989 p.83-

84). Although genetic diversity is graduaUy regained foUowing a bottleneck, extreme bottlenecks can result in a prolonged loss of variation (Nei, 1975;

Chakraborty and Nei, 1977). The extent of this loss depends on the bottleneck’s severity and how recently it occurred. The size of colonisation bottlenecks increases with isolation. The colonisation of distant islands tends to involve smaUer propagules (colonising individuals) and hence larger bottlenecks. New

158 populations on remote islands therefore sample less of the continental variation and retain less genetic diversity.

Area will best explain differing genetic diversity among insular populations if gene flow and historical demography are unimportant. When these 6ctors are significant, an estimate of isolation should account for some of the residual variation left unexplained by island area. Isolation, however, cannot predict genetic diversity if differences among islands are mainly determined by the time elapsed since colonisation. Isolation only corresponds with the severity of colonisation bottlenecks, not with their age.

Isolation, unlike area, is extremely diflScult to estimate (Chapter 2). For most archipelagos there are numerous distance measures that could be used: distance to mainland, to closest island, to closest larger island etc. Even then one should transform those distances to account for the source’s diversity: being close to a small island is not the same as being close to the mainland. Instead of this a biological index, species richness (the number of species found on an island) can be used to provide the isolation component. According to the theory of island biogeography (MacArthur and Wilson, 1967) species richness is determined by the stochastic processes of immigration and extinction, which depend on area and isolation.

159 West Indian butterflies

When investigating the biodiversity of islands, be it species or genetic diversity, several characteristics are important for the study system One needs a region with many islands, which vary in their size and degree of isolation, and a taxonomic group that is both speciose and well known. West Indian butterflies are approaching such a situation. Butterflies have received unparalleled attention among West Indian insects and we can now use the island species hsts with some confidence (Chapter 2).

Here the results of a recent allozyme survey (Chapter 3) are used to test the mq)ortance of gene flow and historical demography on the genetic diversity of four species of West Indian butterfly. If gene flow or colonisation bottlenecks are important then species richness should be a better indicator than area of genetic diversity. If neutral stochastic factors determine insular biodiversity then species richness and genetic diversity should be positively correlated.

METHODS

Four species, Battus polydamas Linnaeus (Papilionidae), Anartia jatrophae

Linnaeus (Nymphalidae), Heliconius charitonia Linnaeus (Heliconiidae), and

Dryas iulia Fabricius (Heliconiidae) were surveyed using allozyme electrophoresis (Chapter 3). The four species span the spectrum of intra-specific phenotypic variation in West Indian butterflies and are distributed widely

160 throughout the neotropics. Battus polydamas and Dryas iulia are highly differentiated in the West Indies and present on virtually all the islands. Heliconius charitonia shows less differentiation and its West Indian distribution only extends from Cuba to Montserrat, being absent from the southern Lesser Antilles and

Trinidad. Anartia jatrophae is relatively undifferentiated but has described forms on each of the Greater Antilles, St. Croix, and one that stretches from the continent through the Lesser Antilles (see Smithet al., 1994 and references therein).

Mainly for logistical reasons, samples were collected from an area of less than 1 km^ on each island (Table 1). Although this increases the risk of san^ling related individuals, it is important that collecting sites were of similar size as differences among site areas confound the influence of island area on genetic diversity. It would not be legitimate to compare the genetic diversity of a sarcple collected from all over Cuba with a sample from a single field on Montserrat. When two sanq)les were analysed for the same island, e.g. east and west Cuba, only the larger sangle was used in this analysis.

Two measures of genetic diversity were calculated: unbiased heterozygosity

(Levene, 1949; Nei, 1978) and allele diversity (average number of alleles per locus) using the Biosys-1.7 software package (Swofford and Selander, 1989).

Heterozygosity should not be biased if sufBcient loci are examined (Nei, 1978); allele diversity, however, is particularly sensitive to sançle size. When examining allele diversity, therefore, sample size was factored out in a step-wise regression.

Allele diversity was first regressed against sample size and then the residuals were

161 regressed against the tested variables (species richness, and area). The partial correlation coefficient thus obtained is thought to reflect the contribution of the tested variable to allele diversity.

The statistics used in this analysis assume linearity. The null hypothesis is that eflfective population size influences genetic diversity as predicted by the neutral theory. Heterozygosity should increase linearly until reaching a plateau at a large eflfective population size (Kimura, 1983). Continental sanq)les of the species examined tended to be more diverse than those fi'om the islands (Table 1). It is therefore assumed that the plateau has not been reached in the island populations and hence a linear relationship is expected.

RESULTS

Battus polydamas heterozygosity was not significantly correlated with species richness (Figure 1) or with island area. Battus polydamas fi*om Florida, however, was remarkable in being conq)letely monomorphic. If this population was excluded then both species richness (r^ = 0.75, F = 11.88, p<0.05) and area (r^ =

0.74, F = 11.55, p<0.05) were significantly correlated with heterozygosity.

Species richness did not, however, explain any residual variation in heterozygosity when area was held constant. No significant relationship was found between allele diversity and either species richness or area, whether Florida was included or excluded fi'om the analysis.

162 Heterozygosity was correlated with species richness (r^ = 0.67, F = 18.19, p<0.05) in Anartia jatrophae^ although this relationship was negative (Figure 2).

Heterozygosity was also negatively correlated with island area (r^ = 0.63, F =

15.18, N. S p<0.05) (Figure 6). Species richness did not explain heterozygosity significantly better than area. Allele diversity was correlated with neither area not species richness.

The heterozygosity of Heliconius charitonia was correlated with species richness

(r^ = 0.58, F = 8.81, p<0.05) (Figure 3) but not island area. Again species richness did not explain any of the residual variation once the influence of area had been removed. Allele diversity was correlated with neither area nor species richness.

Area was significantly correlated with heterozygosity (r^ = 0.30, F = 5.07, p<0.05)

(Figure 6) in Dryas iulia, as was species richness (r^ = 0.50, F = 11.91, p<0.01)

(Figure 4). Again, when area was held constant species richness failed to explain any variation in heterozygosity. Dryas iulia was the only species wdiich showed a significant correlation involving allele diversity (having factored out sançle size).

As with heterozygosity, the correlation between allele diversity and species richness (Figure 5) appeared stronger (r^ = 0.54, F = 14.05, p<0.01) than that with area (r^ = 0.36, F = 6.65, p<0.05) but in fact did not explain any of the residual variation when area was held constant.

163 DISCUSSION

Most island studies show that insular populations are genetically depauperate compared with those on the mainland (Van Valen, 1962; Soule etal., 1973; Patton et al., 1975; Berry and Peters, 1976; Gill, 1980; Berry, 1986; Ashley and Wills,

1987; Peterson, 1990; Wayne et al., 1991; Peterson and Heaney, 1993; but see

Seutin et al., 1994). Such studies, however, often fail to demonstrate a significant relationship between area and genetic variation within island groups. This may be because most populations are strongly influenced by gene flow and/or historical demography.

Neutral theory predicts that the genetic diversity of populations at equilibrium is determined by their effective size and the mutation rate (Kimura, 1983). Island area and isolation (through its effect on gene flow) influence the current effective population size of island species. Isolation also affects the severity of historical bottlenecks in population size, such as those associated with colonisation.

Bottlenecks reduce current levels of genetic diversity by lowering the average effective population size.

The stochastic theory of island biogeography predicts that the species richness of insular communities at equilibrium is similarly determined by the area and isolation of islands. Average population size determines the extinction rate, whilst isolation determines the immigration rate (MacArthur and Wilson, 1967). Combining the neutral theory of population genetics and the stochastic theory of island

164 biogeography, we expect the genetic diversity of insular populations to be correlated with the species richness of insular communities.

The results presented here show that species richness, particularly heterozygosity, is indeed significantly correlated with genetic diversity. An unexpected result of this study was that heterozygosity appeared to be a better indicator than allele diversity. Allele diversity is usually expected to be more sensitive to population bottlenecks (Nei et al., 1975), and hence more likely to be correlated with island area or species richness. As expected, allele diversity and heterozygosity were low in the West Indies compared with the continent. However, having lost the rarest alleles in the initial colonistion of the archipelago, variation in allele diversity among islands may have been somewhat more restricted than heterozygosity.

Genetic diversity is not a significantly better indicator of species richness than island area. The isolation of islands may therefore be unimportant at one, or both, levels of biodiversity. The results presented in Chapter 2 suggest that isolation does not influence the overall species richness of West Indian butterfly communities. Heterogeneity in the slope of the species-area curve, however, might reflect differences in the average dispersal abihty of communities. Communities in the Lesser Antilles appeared more vagile and relatively speciose con^ared with those in the Greater Antilles.

The slope of the heterozygosity-area curves vary among the four species. The relative importance of isolation over area probably depends upon the biogeography

165 and dispersal propensities of individual species (Chapter 2). Most notably the heterozygosity-area curve oï Anartia jatrophae has a significantly negative slope whilst that ofDryas iulia is significantly positive (Figure 6). Species with flat or even negative heterozygosity-area curves are probably more vagile, whilst species with steeper positive curves are more sedentary; the inq)ortance of area declines with gene flow. This hypothesis appears to match the population genetic structure of Dryas iulia and Anartia jatrophae (Chapter 3). Dryas iulia is capable of widespread dispersal but is highly differentiated in the West Indies {Fst = 0.72) indicating low levels of gene flow. Anartia jatrophae, on the other hand, appears to disperse readily throughout the West Indies and is relatively undifferentiated

(Fst = 0.189).

Although species richness (hence area and isolation) explain some of the variation in genetic diversity, a substantial amount remains unresolved. This might be due to the problems of estimating species richness (Chapter 2) or the inherent inaccuracy of allozyme data. Archie (1985) pointed out that alleles can behave exactly as predicted by theory but the high variance associated with measures of genetic diversity makes it impossible (without unpractically large sanq)le sizes and number of loci) to resolve this concordance.

Alternatively there may be a biological explanation. Species richness and genetic diversity are only expected to correlate under a neutral stochastic model of island biogeography. This assunq)tion may be invalid: genetic diversity and/or species richness might depend on deterministic factors.

166 Manipulations of population size have generally supported models based on allele neutrality. The consequences of bottlenecking laboratory and ‘quasi-natural’ populations has been tested on fruit flies, Drosophila melanogaster (Sing et al.,

1973), house flies, Musca domestica (McCommas and Bryant, 1990) and eastern mosquitofish, Gambusia holbrooki (Leberg, 1992). Reviewing their own laboratory study and that of Sing et al. (1973), McCommas and Bryant (1990) concluded that: "the decline in average heterozygosity over all loci in both cases closely followed neutral expectation." Under what was close to natural conditions,

Leberg (1992) also found patterns consistent with neutrality. The neutral theory, however, has been criticised on the grounds that it doesn't adequately explain the genetic diversity observed in nature (Ayala, 1972; Lewontin, 1974). Selection has been invoked to explain genetic similarities at allozyme loci in supposedly isolated populations (Goulson, 1993).

Johnson (1973) tested the idea that unpredictable environments support fewer species but promote polymorphic gene variation. In other words that species richness should be negatively correlated with genetic diversity (the exact opposite to the neutral stochastic prediction). Johnson calculated an index of genetic diversity for each of the Hawaiian islands using fifty one species of picture-wing

Drosophila', this was then plotted against the number of picture-wing species on each island. The result was indeed a linear and negative relationship between species number and genetic diversity.

167 Studies on Darwin’s finches in the Galapagos Islands have provided further evidence for the importance of selection on spéciation and genetic diversity.

Spéciation is inhibited wiien environments are unstable as specialisations are sometimes eradicated by environmental fluctuations (Grant and Grant, 1989).

Such environmental heterogeneiity, however, might favour alternative genotypes: genetic diversity could be high where selection favours flexibihty.

Spéciation, when it does occur, may lead to the increased partitioning of limited resources. The resulting, more speciahsed, species will probably have smaller population sizes and hence reduced genetic diversity. Islands with high spéciation rates might therefore support populations with, on average, low levels of genetic diversity. Immigration could have a similar effect (MacArthur and Wilson, 1967).

As more species colonise an island (and equihbrium is approached) conçetition increases, reducing average population sizes and accelerating the extinction rate.

Such resource partitioning effects can also be achieved indirectly. Species richness might be limited by ecological inqjoverishment (Lack, 1976). Islands with more habitats are able to support a wider range of specialised species. However, although the island has greater species richness, each species has less of its preferred habitat. Species will therefore tend to have lower population sizes and reduced genetic diversity. If species richness is partly determined by habitat diversity, and not by population size, then the correlation between species richness and genetic diversity will be reduced accordingly.

168 To summarise, deterministic factors can upset the neutral stochastic predictions of population genetics and island biogeography. Indeed wiiere these factors are important there may be a negative correlation between species richness and genetic diversity. Resource partitioning and conq>etition (associated with high levels of speciation/immigration) can lead to more species hut smaller average population sizes and lower genetic diversities. Greater habitat diversity may also lead to increased species richness but similarly partitions islands into smaller units. Finally selection in unstable environments can inhibit spéciation whilst favouring genetic diversity.

In Hawaiian Drosophila deterministic factors appear to be significant because species richness is negatively correlated with genetic diversity (Johnson; 1973). In

West Indian butterflies, however, stochastic neutral forces appear more important: species richness and genetic diversity are positively correlated. Why should different taxa in different archipelagos show such divergent patterns?

There are some obvious differences between Hawaiian Drosophila and West

Indian butterflies. First and foremost Hawaii is an oceanic archipelago, \\^ereas the West Indies are what might be called ‘offshore’ islands. The difference between oceanic and offshore islands is the relative contribution of immigration and spéciation to species richness. In oceanic archipelagos spéciation dominates since the islands are so remote. In offshore islands, however, the immigration rate is more significant.

169 Spéciation, particularly adaptive radiation, is more likely to lead to a very closely related fauna, in which conq)etition and resource partitioning are likely to be in^ortant. Hawaiian Drosophila are much more closely related to one another than are West Indian butterflies. Most islands in the West Indies only support a few butterfly species of the same genus and adaptive radiations appear to have been rare e.g. Calisto (Smith et al., 1994). The stochastic model, therefore, seems to best e?q)lain the biodiversity of West Indian butterflies, whilst selection or resource partitioning may explain the biodiversity of Hawaiian Drosophila.

The correlation between species richness and genetic diversity, however, is not consistent among West Indian butterflies. In all four species studied here the Isle of Pines has more genetic diversity than predicted by either its area or species richness. Lesser Antillean populations of Anartia jatrophae also have ‘too much’ genetic diversity.

Yet deterministic factors do not appear to have depressed species richness. The

Isle of Pines and the Lesser Antilles are relatively species rich, falling on or above the species-area curve for West Indian butterflies (Chapter 2). A more probable e?q)lanation for the inflated genetic diversity is gene flow from rich and proximate sources (Cuba for the Isle of Pines, South America for the Lesser Antilles).

Species richness, however, should account for such a distance effect. If the Isle of

Pines and the Lesser Antilles are more accessible and experience greater gene flow, then they should also have a greater immigration rate and inflated species richness.

170 Species richness and genetic diversity should increase in concert. Although these islands are quite species rich, they are not rich enough to explain the genetic diversity.

Isolation is a relative term, it depends on the species concerned. An island may be very isolated for a weak flying butterfly but quite accessible for a migratory species. Isolation affects the genetic diversity of populations through gene flow and species richness through immigration (the rate at which new species colonise an island). The relationship between gene flow and isolation for some species may be quite different to the relationship between isolation and immigration for the community.

The dispersal ability of individual butterflies has a particular distribution. Clearly these curves will not be the same for aU species. Individuals of some species will tend to disperse further than individuals of other species. The summation of these species curves forms the community curve (Figure 7). The community curve determines the immigration rate, whilst the individual species curves determine gene flow.

Some species curves will be very much out of phase with the overall community curve. Anartia jatrophae is a very good disperser. This means that gene flow for

Anartia jatrophae will reach saturation (where all individuals of the species are capable of reaching an island) sooner than the community immigration rate (where all species are capable of reaching the island).

171 The effect of island isolation on gene flow and immigration is shown in Figure 8.

At low isolation gene flow is high for most species, maintaining their eflfective population size and hence genetic diversity. Similarly the immigration rate for the community is also high, raising species richness. With increased isolation gene flow and immigration fall off. The rate of decrease, however, varies among species.

This could e>q)lain the great disparity between the genetic diversity of Anartia jatrophae and species richness in the Lesser Antilles. The distance between

Grenada and Trinidad is still within the dispersal range of Anartia jatrophae but is well out of range for most species. Dryas iulia, however, appears to approximate well to the community curve; its gene flow closely mirrors the community immigration rate. Dryas iulia reflects the average dispersal ability of West Indian butterflies. The genetic diversity of Dryas iulia is therefore highly correlated with species richness. The Isle of Pines, however, represents the degree of isolation at which theDryas iulia and community curve part, and hence genetic diversity and species richness also diverge.

There is still controversy over the origin of the West Indian butterfly fauna

(Chapter 3; Miller and Miller, 1989). The most prevalent view is that West Indian butterflies arrived by dispersal Jfrom the continent (Scott, 1972). This hypothesis predicts that West Indian butterflies are not a random sanq)le of continental butterflies; they are a particularly mobile subset. Their dispersal capabilities are

172 presumably greater than the community norm, where the community is continental

Caribbean butterflies. As such we might expect the island species to have higher genetic diversity than one would predict from their species richness. If the West

Indian butterflies were a random sample of the continental community then we should be able to extrapolate the continental species richness from the relationship between island species richness and genetic diversity.

Dryas iulia appears to represent the average dispersal capability of West Indian butterflies. Using the heterozygosity of Dryas iulia on Trinidad we predict that the continent will have about 300 species. In fact Trinidad has about 600. Clearly the island fauna is far more genetically diverse than one would e?q)ect from its species richness. It is possible that genetic diversity has become saturated on the continent. The heterozygosity of Dryas iulia population in Trinidad, however, is substantially lower than that found in Panama (Table 1) suggesting that this is not the case. The data therefore supports the dispersal hypothesis that West Indian butterflies are a particularly vagile subset of the continental fauna.

In conclusion, species richness does seem to be a good predictor of genetic diversity. This is evidence that both levels of biodiversity are determined by stochastic neutral factors in West Indian butterflies. The occasional break-down of this relationship does not result from deterministic factors but is more likely to reflect difrering dispersal propensities.

173 BIBLIOGRAPHY

Archie, J.W. 1985. Statistical analysis of heterosygosity data: independent sanq)le comparisons. Evolution 39: 623-637.

Ashley, M. and Wills, C. 1987. Analysis of mitochondrial DNA polymorphisms among channel island deer mice. Evolution 41: 845-863.

Ayala, F. J. 1972. Darwinian vs non-Darwinian evolution in natural populations of Drosophila. In Proc. 6th Berkeley Symp. Math. Stat. Prob. vol. 5. Le Cam, L.M., Neyman, J. and Scott, E.L. (Eds.) University of Cahfomia Press, pp.211-236.

Berry, R.J. 1986. Genetics of insular populations of mammals, with particular reference to differentiation and founder effects in British small mammals Biological J. Linn. Soc. 28:205-230.

Berry, R.J. and Peters, J. 1976. Genes, survival and adjustment in an island population of house mouse, pp. 23-48. In: Population Genetics and Ecology. Karlin, S. and Nevo, E. (eds ). Academic Press, N.Y.

Chakraborty, R. and Nei, M. 1977. Bottleneck effects on average heterozygosity and genetic distance with stepwise mutation model. Evol. 31:347-356.

Gill, A.E. 1980. Evolutionary genetics of California Islands Peromysciis, pp. 719- 744. In D M. Power (ed.). The California Island: Proceeding of a multidisciplinary symposium Santa Barbara Museum of Natural History, Santa Barbara, CA.

Goulson, D. 1993. Allozyme variation in the butterfly, Maniola jurtina (Lepidoptera: Satyrinae) (L.): evidence for selection. Heredity 71:386-393.

174 Grant, P R. and Grant, B.R 1989. Synq)atric spéciation and Darwin’s finches. In Spéciation and its Consequences, Otte D. and Endler, J.A. Eds. Sinauer Associates Inc. Sunderland MA.

Hartl, D.L. and Clark, AG. 1989. Principles of Population Genetics. 2nd Ed. Sinauer Associates, Sunderland, MA.

Johnson, G.B. 1973. Relationship of enzyme polymorphism to species diversity. Nature 242:193-194

Kimura, M. 1983. The Neutral Theory of Molecular Evolution. Cambridge University Press, N.Y., U.S.A.

Lack, D. 1976. Island Biology, Illustrated by the Land Birds of Jamaica. Blackwell.

Leberg, PH . 1992. EfiFects of population bottlenecks on genetic diversity as measured by allozyme electrophoresis. Evolution 46:477-494.

Levene, H. 1949. On a matching problem arising in genetics. Ann. Math. Stat. 20:91-94.

Lewontin, RC . 1974. The Genetic Basis of Evolutionary Change. Columbia University Press.

MacArthur, RH . and Wilson, E.G. 1967. The Theory of Island Biogeography. Princeton University Press, NJ.

McCommas, S. A. and Bryant, E.H. 1990. Loss of electrophoretic variation in serially bottlenecked populations. Heredity 64:315-321.

175 Miller L D and Miller J.Y. 1989. The biogeography of West Indian butterflies. In: Biogeography of the West Indies. Woods, C A (ed.). Sandhill Crane Press, Inc. Gainesville FL. pp. 229-262.

Nei, M. 1975. The bottleneck effect and genetic variability in populations. Evol. 29:1-10.

______. 1978. Estimation of average heterozygosity and genetic distance from a small number of individuals. Genetics 89:583-590.

Patton, J.L., Yang, S.Y. and Myers, P. 1975. Genetic and morphologic divergence among introduced rat populations {Rattus rattus) of the Galapagos Arhcipelago, Ecuador. Syst. Zool. 24:296-310.

Peterson, AT. 1990. Evolutionary relationships of the Aphelocoma jays. Unpublished Ph.D. dissertation of the University of Chicago.

Peterson, AT. and Heaney, L.K 1993. Genetic differentiation in Phillipine bats of the genera Cynopterus and Haplonycteris. Biological J. Linn. Soc. 49:203-218.

Scott, J.A. 1972. Biogeography of Antillean butterflies. Biotropica 4: 32-45.

Seutin, G., Klein, N.K., Ricklefs, R.E., and Bermingham, E. 1994. Historical biogeography of the bananaquit (Coereba flaveola) in the Caribbean region: a mitochondrial DNA assessment. Evol. 48: 1041-1061.

Sing, C.F., Brewer, G.J. and Thirtle, B. 1973. Inherited biochemical variation in Drosophila melanogaster: noise or signal? I. Single-locus analysis. Genetics 75:381-404.

Soule, M E., Yang, S.Y. and Weiler, M.G.W. 1973. Island lizards: the genetic- phenetic variation correlation. Nature 242:193-193.

176 Smith, D S., Miller L , and Miller, J. 1994. The Butterflies of the West Indies and South Florida. Oxford Universtiy Press.

SwofiFord, D.L. and Selander, KB. 1981. BIOSYS-1: A FORTRAN program for the conq)rehensive analysis of electrophorestic data in population genetics and systematics. J. Heredity 72:281-283.

Van Valen, L. 1962. A study of fluctuating asymmetry. Evolution 10:139-146.

Wayne, K K , George, S B., Gilbert, D.A., Collins, P.W., Kovach, S.D., Girman, D. and Lehman, N. 1991. A morphologic and genetic study of the island fox Urocyon littoralis. Evolution 45:1849-1868.

177 Table 1. Genetic diversity and island data.

Area Butterfly Alleles Species Island (Km') species richness per locus He N Battus Cuba 108660 182 1.6 0.09 33 polydamasFlorida 8500 110 1 0 21 Guadeloupe 1438 44 1.1 0.02 20 Hispaniola 76190 202 1.7 0.08 30 Jamaica 11424 126 1.4 0.05 29 Martinique 1079 38 1.1 0.01 10 Panama* n/a 600+ 1.4 0.13 4 Puerto Rico 8866 97 1.3 0.09 9 Anartia Cuba 108660 182 1.5 0.06 21 jatrophaeDominica 751 52 1.5 0.11 16 Florida 8500 110 1.3 0.05 30 Grenada 345 47 1.5 0.1 16 Guadeloupe 1438 44 1.4 0.11 16 Hispaniola 76190 202 1.4 0.05 19 Isle of Pines 2200 111 1.5 0.06 18 Jamaica 11424 126 1.6 0.09 26 Montserrat 106 39 1.5 0.16 18 Puerto Rico 8866 97 1.1 0.04 20 St Lucia 617 48 1.5 0.13 20 Trinidad* n/a 600+ 2 0.15 20 Heliconius Cuba 108660 182 1.9 0.13 24 charitoniaFlorida 8500 110 1.5 0.12 24 Hispaniola 76190 202 1.5 0.08 22 Isle of Pines 2200 111 1.9 0.13 24 Jamaica 11424 126 1.8 0.14 24 Mona 62 53 1.5 0.06 24 Montserrat 106 39 1.2 0.02 25 Puerto Rico 8866 97 1.2 0.05 24 St Kitts 168 42 1.2 0.03 20 Dryas iuliaCuba 108660 182 1.7 0.08 17 Dominica 751 52 1.1 0 24 Grenada 345 47 1.2 0.05 6 Guadeloupe 1438 44 1 0 6 Hispaniola 76190 202 1.7 0.08 14 Isle of Pines 2200 111 1.9 0.1 20 Jamaica 11424 126 1.6 0.05 21 Martinique 1079 38 1.2 0.03 11 Montserrat 106 39 1.2 0.02 22 New Providence 207 60 1.3 0.05 20 Panama* n/a 600+ 2.1 0.2 19 Puerto Rico 8866 97 1.5 0.06 18 St Kitts 168 42 1.3 0.04 19 St Lucia 617 48 1.4 0.08 16 St Vincent 344 41 1.1 0.04 9 Trinidad* n/a 600+ 1.5 0.15 7

* continental samples not included in analyses.

178 Figure I. Heterozygosity vs species richness: B attus p o ly d a m a s . os

0 .1

0.09

0.08 -

0.07 -

0.06

^(U 0.05

0.04 y = 0.097X - 0.14 0.03 = 0.46 N.S. 0.02

0.01 f Florida 0 --- 4— ' 1.5 1.6 1.7 1.8 1.9 2 2.1 2.2 2.3 2.4

Species number Figure 2. Heterozygosity vs species richness: Atiartia jatrophae . §

0.16 J

0.14 -

0.12 -

0.1 -

0.08 -

0.06 - ♦

♦ 0.04 - y = -0.12x + 0.32 = 0.67 p<0.05

0.02 -

- f —i- 1.5 1.6 1.7 1.8 1.9 2.1 2.2 2.3 2.4

Species number Figure 3. Heterozygosity vs species richness: Heliconius charitonia.

0.16 J

0.14 - ♦ ♦ 0.12 -

0.1 -

0.08 -

0.06 -

y = 0.13x-0.18 0.04 - = 0.56 p<0.05

0.02 -

— j— 0 - 1.5 1.6 1.7 1.8 1.9 2 2.1 2.2 2.3 2.4 Species number Dryas iulia. Figure 4. Heterozygosity vs species richness: (N oo

0.1 J

0.09 -

0.08

0.07 -

0.06 --

^ 0.05

0.04

0.03 - y = O.OSx - 0.099

0.02 - = 0.50 p<0.01

0.01 -

0 1.5 1.6 1.7 1.8 1 9 2 2.1 2.2 2.3 2.4 Species number oo Figure 5, Average number of alleles per locus (allele diversity) vs species richness: Dry as iulia.

0.5 J

0.4 -

0.3 -

0.2 - > 12 (D 0.1 - > ■ D 0) 0) ■ -f-- 4 1. 1.6 1.7 1.8 1.9 2 2.1 2.2 2.3 2.4

- 0.1 y = 0.68x- 1.25 - 0.2 - = 0.54 p<0.01

-0.3 -

-0.4 -

Species number Figure 6. Heterozygosity-area curves: Dryas iulia and Anartia jatrophae . Tf- OO

0.16 T A. jatrophae y = -0.032x + 0.198 0.14 - R^ = 0.63 p<0.01

0.12 -

D . iulia

0.1 - y = 0.017x-0.0056 R^ = 0.30 p<0.05

0) 0.08 - o ô

0.06 - oDi

0.04 - ■Aj

0.02 -

—H ------.— f— ---- 1 1.5 2.5 3 3.5 4.5 5 5.5

A rea (km ) Figure 7. Dispersal ability distributions.

N

Distance

Species a, b, c, d have progressively better dispersal abilities. TTie sunmialion of these curves gives the average dispersal curve for the community.

185 Figure 8. Isolation, gene flow and immigration.

C = community immigration

Di = D. iulia gene flow

Aj = A.jalrophae gene flow

Distance

a = Distance separating Cuba and the Isle of Pines, b = Distance separating Trinidad and Grenada.

Anartia jatrophae and Dryas iulia are easily able to reach the Isle of Pines from Cuba. Gene flow in these species is high compared with the community immigration rate.

Grenada is quite remote from Trinidad. Gene flow between D.iulia populations and the community immigration rate are both low at this distance. A.jatrophae, however, is a more vagile species and gene flow remains high.

186 CHAPTER 6

CLASSIFYING ISLAND POPULATIONS

ABSTRACT

Most attempts to define species have come unstuck when confi*onted with differentiated island populations. Here it is argued that recent advances in genetic technology have not eased this problem The genetic basis of spéciation is still poorly understood; no simple relationship between reproductive isolation and genetic divergence has been discovered. Spéciation is discussed with reference to island populations and the concept of subspecies explored. Subspecies are distinct

(ideally monophyletic) groups whose evolutionary independence is unclear. Whilst they may evolve independently, they retain the potential to draw on a common genetic reservoir.

Islands and the indeterminate species concept

Nowhere is the inadequacy of species definitions more glaringly obvious than in the classification of island populations. As Mallet (1995) points out: ‘The arbitrariness of allop atiic races and species is a consequence of the lack of reality and cohesion

187 of species over long distances, rather than any problem with definitions’. For island populations, it seems we have to settle for an indeterminate species concept.

Many view the argument over species definitions as merely an interesting, if inconsequential, philosophical conundrum To take a trivial exanq)le, consider two island populations separated by thousands of miles and evolving as conq)letely independent gene pools. Separate species some might cry. I am, however, talking about rabbits in Britain and Austraha. By almost any definition these populations are conspecific. Of course this is obvious, but only because we know how rabbits got to Australia and because we can easily put them in a hutch and get perfectly healthy Anglo-Austrahan bunnies. The question becomes much more difficult vdien one looks at organisms of unknown history, confusing phylogeny and poorly understood dispersal capabilities.

Taxonomy is given a bad name when populations are recognised as species, merely because somebody describes them and subjectively decides that they merit specific rank The blame is not with the taxonomist; each must act according to their own criteria and experience. They have no choice for there is no universally accepted standard. Unfortunately this generates inconsistency wèich makes any work based on species problematic. Some coherence has been obtained within taxa; butterflies for exan^le are often judged on their genitaha, mammals on bone structure etc.

Basing species classification on one set of characteristics, however, is unsatisfactory. Some butterflies, whilst clearly the same species, have wildly variable genitalic moiphology, e.g. Maniola jurtina (Goulson, 1993).

188 More serious difficulties arise when one tries to make coirçarisons across taxa.

Here the level of differentiation considered worthy of species status can vary greatly. Bird species might appear less genetically differentiated than butterfly species but would this reflect quicker spéciation or merely the sharper eyes of ornithologists? This is a crucial question, with ramifications throughout biology.

In ecology, does the comparison of bird and butterfly species richness have any legitimacy, in evolution does spéciation really occur at lower levels of genetic divergence in birds than butterflies, and in conservation would a national park protecting 100 bird species conserve as much ‘biodiversity’ as one that protected

100 butterfly species? Species is a critical term which underpins biology.

W hat is a species?

This paper deals with population genetics and systematics. Spéciation is ‘fully within the province of population genetics’ (Terrq)leton, 1989) and species form the ‘ultimate taxon’ in systematics (Loevtrup, 1987). This immediately opens a

Pandora’s box of definitions and schools of thought. I believe that cladistics

(Hennig, 1966) is the best system of classification (Ridley, 1986) and that it requires the acceptance of at least a general theory of evolution (to justify the search for a branching hierarchy in biodiversity, see Chapter 7). There are many species definitions but probably the most widely quoted, and the one I will deal with here, is the biological species concept (Mayr, 1942). The essential property

189 of species so defined is reproductive isolation: biological species are groups of individuals which can breed with each other but not with individuals of different species. This ‘breeding criterion’ is all very well in synq)atry but how can it be applied to allopatric populations?

Applying the biological species concept to island populations

The evolutionary status of many island races remains unclear. Island populations may be geographically isolated but plate tectonics has shown that this need not be permanent. Even when populations are on separate continents they may still end up synçatric, as in the ‘great American biotic interchange’ (Stehli and Webb,

1985). Allopatric populations may therefore interbreed one day unless they have evolved intrinsic barriers to reproduction.

Testing for intrinsic reproductive isolation requires breeding experiments. This, however, is not a very practical way of resolving taxonomic rank. Breeding experiments are usually time consuming, technically difficult and expensive.

Furthermore, despite all the effort, one can never be sure of the result. It can always be argued that the experimental conditions were unnatural and that under different circumstances the two populations would (or wouldn’t) hybridise.

Recent advances in genetic technology hold out the hope of a sinq)le solution. It is much easier to directly assess genetic divergence than reproductive isolation.

190 Where we know the relationship between genetic divergence and reproductive isolation within a group, we might use this as a yardstick to infer reproductive isolation in related island taxa. Unfortunately this approach does not seem very satisfactory either.

There are several difficulties with using genetic divergence as a species test. We are not yet sure how many genetic changes lead to spéciation; indeed it may be more a question of which rather than how many. Reproductive isolation might be caused by the accumulation of many genetic differences of small effect, or by a relatively small number of genes with large effects (Coyne, 1992). In the latter case genetic distance would still be correlated with spéciation but overall divergence of genomes per se would not be the cause.

There is some evidence, however, that genetic divergence might be used to predict reproductive isolation. Among Drosophila species there exists a roughly linear relationship between allozyme genetic distance (Nei, 1978) and reproductive isolation (Coyne and Orr, 1989). But such a relationship might be unusual and restricted only to very closely related species. In other groups there seems to be great heterogeneity in spéciation rates as measured by genetic divergence. In the

Lepidoptera, for exatq)le, butterfly species in ihQ Papilio glauctis species group are separated by Nei's D (Nei, 1978) ranging from 0.15 to 0.4 (Hagen and Scriber,

1991) whilst species of the spruce budworm moth differ by as httle as D = 0.06

(Castrovillo, 1982; Harvey, 1985).

191 Another problem is that reproductive isolation can be influenced by selection, vdiich might lead to breeding barriers independent of genetic divergence. This seems particularly likely in prezygotic isolation which may evolve rapidly with respect to genetic distance through sexual selection (Fisher, 1930; Lande, 1981) or reinforcement (Dobzhansky, 1937). For exanq)le, Coyne and Orr (1989) proposed that reinforcement has accelerated prezygotic isolation among synçatric populations of Drosophila.

The use of genetic distance as an indicator of spéciation is further comphcated in islands. Bottlenecks in population size can lead to a loss of genetic variation (Nei et a/., 1975). Genetic distance can be severely affected by such demographic factors. Bottlenecks, which usually occur at colonisation, can rapidly increase the genetic distance observed between island populations (Chakraborty and Nei,

1977). Such ‘genetic revolutions’ might lead to spéciation (Mayr, 1954; but see

Barton, 1989).

The final problem with genetic divergence is that distance estimates sometimes vary between genetic markers. For exanq)le allozyme variation yielded much lower genetic distances than mtDNA among sea urchins (Bermingham and Lessios, 1993) and among fi’eshwater fish (Ovenden et al., 1993).

One is forced to conclude that, without more knowledge of the genetic mechanisms underlying spéciation, we still cannot draw confident conclusions on the specific (i.e. reproductive) status of island populations based solely on genetic

192 data. If we could determine which genetic changes cause reproductive isolation, and if there is a consistency in this pattern across taxa, then genetic data might be used to diagnose specific rank. For the time being, however, if one wants to describe island species in accordance with the biological species concept, breeding experiments are a necessary evil.

So what about subspecies?

Typically when one looks at island populations, some appear to have unique characteristics which identify them as coming fi*om a particular island. These island races are often described as subspecies. A working definition of subspecies might be that if one can successfully assign 75% of individuals Jfrom two islands, then you have two subspecies. Recent genetic studies of West Indian birds (Seutin et al.

1993, 1994) and butterflies (Chapters 3 and 4) have shown that insular subspecies are often separated by a wide range of genetic distances. Whilst subspecific classification reflects genetic and phenotypic diflFerentiation, all subspecies are not created equal.

According to Mayr (1982 p.289) subspecies are not units of evolution, they are units of convenience. Subspecies are a recognition of pattern with no clear reference to process; they may be monophyletic groups but it is unclear whether they have achieved evolutionary independence. Many subspecies will not be evolving as fully separate units, most will be only partially autonomous. The

193 majority of genes may be independently evolving from one subspecies to the next but advantageous genes might still spread through all subpopulations.

Spéciation in island populations

I propose two possible states preceding spéciation in island populations. This is not a definition, merely a way of compartmentahsing allopatric spéciation. The term state is used instead of stage, as the latter would inçly a progression that is not necessarily required. In the first state geographically isolated subpopulations, such as those on islands, undergo divergence despite continued gene flow.

Random genetic drift, founder effects and selection can lead to fixation of different alleles in isolated populations (Sokal and Wartenberg, 1983).

Island populations will diverge at neutral loci despite gene flow, so long as the rate of gene flow is low (less than one migrant per generation) (Wright, 1969). This is because ‘introduced’ alleles are likely to be swanq)ed by native alleles, and will drift to extinction. If alleles arriving through gene flow are advantageous, however, they can become fixed even if the rate of gene flow is low.

Consequently, insular populations can diverge at neutral loci vriiilst still contributing to, and drawing from, the whole population’s genetic reservoir.

Populations at state one can also diverge through selection. If conditions differ among islands, some insular populations will evolve characteristics which are only

194 locally adaptive. Locally adaptive genes can spread wherever dispersing individuals take them (if those individuals breed successfully). They will only become estabhshed on other islands, however, if their selective advantage is retained under the new conditions and if this advantage is sufficient to overcome their initially low frequency in the new population.

In the second state, gene flow is sufficiently inhibited that even adaptive genes do not spread between islands. This does not necessarily mean that insular populations have evolved intrinsic barriers to reproduction. An ecological shift could reduce the tendency to disperse, rendering other islands inaccessible. A similar outcome could be achieved by increasing the physical isolation of islands, through rising sea levels, changes in wind direction, plate tectonics etc. State two does not necessarily follow from state one. It can be reached instantaneously without any evolutionary differentiation. A state two island population could be created suddenly through rare events: a hurricane blowing propagules out to oceanic islands or an introduction by man e.g. rabbits in Australia.

Populations at states one and two cannot easily be distinguished and are often considered subspecies. Subspecific status is achieved when individuals can be consistently assigned to their island of origin and where the insular population is considered monophyletic. Subspecies have no clear status with respect to evolution, since they vary greatly in their degree of differentiation and their evolutionary independence.

195 At some point state two subspecies should be recognised as species. According to the biological species concept, spéciation occurs when intrinsic barriers to reproduction have evolved. Only then is the separation of gene pools irrevocable.

Whilst populations in state two are evolving on separate evolutionary tracks, in both adaptive and neutral terms, they are not species wdiilst there is still the potential for gene flow. Sub^ecies may fuse if returned to sympatry; separate species, however, will never again share the same evolutionary destiny. As we have seen this definition of species has considerable problems when applied to insular populations, and genetic data does not provide any easy answers.

BIBLIOGRAPHY

Barton, N.H. 1989. Founder effect spéciation. In: Spéciation and Its Consequences. Otte, D. and Endler, J.A. (eds.). Sinauer, Sunderland, MA. pp. 158-179

Bermingham, E. and Lessios, H.A. 1993. Rate variation and mitochondrial DNA evolution as revealed by sea urchins separated by the Isthmus of Panama. Proc. Natl. Acad. Sci. 90:2734-2738.

Castrovillo, P.J. 1982. Interspecific and intraspecific genetic comparisons of North American spruce budworms {Chohstoneura ssp.). Ph D. dissertation. University of Idaho, Moscow.

Chakraborty, R. and Nei, M. 1977. Bottleneck effects on average heterozygosity and genetic distance with stepwise mutation model. Evol. 31:347-356.

196 Coyne, J.A. 1992. Genetics and spéciation. Nature 355:511-515.

Coyne, J.A. and Orr, H.A. 1989. Patterns of spéciation in Drosophila. Evolution 43: 362-381.

Dobzhansky, T., 1937. Genetics and the Origin of Species. Columbia University Press, NY.

Fisher, KA. 1930. The Genetical Theory of Natural Selection. Clarendon, Oxford.

Goulson, D. 1993. Variation in the genitaha of the butterfly Mûrmo/a jurtina (Lepidoptera: Satyrinae). ZooL J. Linn. Soc. 107:65-71.

Hagen, R.H. and Scriber J.M. 1991 Systematics of the Papiho glaucus and P. troilus species groups (Lepidoptera: Papihonidae): inferences from allozymes. Ann. Entomol. Soc. Amer. 84: 380-395.

Harvey, G.T. 1985. The taxonomy of the coniferophagous Choristoneura (Lepidoptera Tortricidae): a review. Pp. 16-48 In: Recent Advances in Spruce Budworm Research. Sanders, C.J., Stark, R.W., Mullins, E.J. and Murphy, J. (eds.) Proc. CANUS A Spruce Budworm Research Symposium, Bangor, Sept. 16- 20, 1984. Canadian Forestry Service, Ottawa.

Hennig, W. 1966. Phylogenetic Systematics. University of Illinois Press, Urbana.

Lande, R. 1981. Models of spéciation by sexual selection on polygenic traits. Proc. Natl. Acad. Sci. USA. 3721-3725.

Loevtrup, S. 1987. On species and other taxa. Cladistics 3:157-177.

197 Mallet, J. 1995. Species definition for the modem synthesis. Trends Ecol. Evol. 10, in press.

Mayr, E. 1942. Systematics and the Origin of Species. Columbia University Press, NY.

. 1954. Change of genetic environment and evolution. In Evolution as a Process. Allen and Unwin, London, pp. 157-180.

1982. The Growth of Biological Thought. Harvard University Press, Cambridge, MA.

Nei, M. 1978. Estimation of average heterozygosity and genetic distance fi’om a small number of individuals. Genetics 89:583-590.

Nei, M., Mamyama, T., and Chakraborty, R. 1975. The bottleneck effect and genetic variabihty in populations. Evol. 29: 1-10.

Ovenden, J R., White, R W .G , and Adams, M. 1993. Mitochondrial and allozyme genetics of two Tasmanian galaxiids {Galaxias auratus and G. tanycephalus, Pisces: Galaxiidae) with restricted lacustrine distributions. Heredity 70:223-230.

Ridley, M. 1986. Evolution and Classification: the Reformation of Cladism Longman, NY.

Seutin, G , Brawn, J., Ricklefs, RE. and Bermingham, E. 1993. Genetic divergence among populations of a tropical passerine, the streaked saltator {Saltator albicollis). Auk 110: 117-126.

198 Seutin, G , Klein, N.K., Ricklefs, R.E., and Bermingham, E. 1994. Historical biogeography of the bananaquit {Coereba flaveola) in the Caribbean region: a mitochondrial DNA assessment. Evol. 48: 1041-1061.

Sokal, R.K and Wartenberg, D E 1983. A test of spatial autocorrelation analysis using an isolation by distance model. Genetics 105:219-237.

Stehli, E.G. and Webb, S.D. (eds.) 1985. The Great American Biotic Interchange. Plenum Press, New York.

Templeton, A.R. 1989. The meaning of species and spéciation: a genetic perspective. In: Spéciation and Its Consequences. Otte, D. and Endler, J.A. (eds ). Sinauer, Sunderland, MA. pp.3-27.

Wright, S. 1969. Evolution and the Genetics of Populations, vol. 2. The Theory of Gene Frequencies. Chicago: University of Chicago Press.

199 CHAPTER 7

FUZZY SPECIES

ABSTRACT

The philosophical basis of scientific terminology is discussed, in particular the purpose and meaning of scientific definitions. It is argued that definitions describing the natural world will always be unsatisfactory and vague. This is because there are no essential truths in natural science, only hypotheses. Systems of classification, particularly cladistics, are discussed in this context. Cladistics is based on the expectation that evolution creates an hierarchical pattern.

Intraspecifically, however, a reticulate pattern is expected. Spéciation is the transition firom the reticulate tokogenetic level to the hierarchical phylogenetic level. This transition is probably not sharp but gradual; consequently it is inq)ossible to determine at what point spéciation actually occurs. Trying to classify populations below the species level will founder on higher and higher levels of homoplasy and thus irresolvable relationships. The classification of recently divergent populations is always likely to be problematic because spéciation is a fuzzy process.

200 Introduction

Aristotle was not only a founding father of Western science but also one of the first people to define species. He believed that there was an essential truth in nature and the task of science was to find and describe those truths. This view, however, has come under increasing attack. The scientific philosopher Karl Popper argued that science does not provide truths, only best guesses that can be refuted with more evidence. Definitions are mere symbols created by scientists to smq)lify life:

“in modem science, only nominalist definitions occur, that is to say, shorthand symbols or labels are introduced in order to cut a long story short” (Popper, 1945 p. 14). To illustrate this Popper treats definitions as equations. On the left is the set to be described and on the right are the describing parameters, the definition.

Aristotle believed that the set on the left was some natural tmth; it had essential properties that, once discovered, could be defined. The set was tme and sharply focused. Popper, however, sees the set as artificial. Science does not work from left to right but from right to left; the set is merely short-hand for the describing parameters.

If sets are not real, if they have no essential tmth m nature, then attempts to define natural phenomena precisely are futile. Natural sets will be fiizzy and sharp sets will only exist in theory (Kosko, 1994). A perfect definition inevitably leaves nature behind and is fit only for abstract entities. Scientific terms, however, need not be exact nor definitions perfect: “the precision of a language depends...upon the fact that it takes care not to burden its terms with the task of being precise”

201 Popper (1945 p. 19). In other words we should not waste time in futile arguments over definitions; they are merely useful, if in^erfect, tools.

Evolution exemplifies the Popperian view. In a system that is constantly changing, definitions are necessarily transient. Aristotle might be forgiven for believing that species were either A or not-A. For all he knew species were immutable groups created by God. Since Darwin, however, biologists have had httle excuse for seeking the definitive essence of natural populations. Species do not miraculously arise ready formed and distinct firom one another, they evolve and gradually spht into new species. Evolution is not a teleological process. It does not reach a preordained target then stop, it is continuous. Species boundaries are always going to be unclear; species are fuzzy.

The term fuzzy has become symbohc of the late 20th century’s rejection of pure

Aristotehan logic. The new science of'fuzzy logic’ (Zadeh, 1965; Kosko, 1994) is spreading beyond its roots in electrical engineering. Fuzziness, however, is nothing new to biology. Despite a tendency towards reductionism and computer simulations, biologists recognise that life is not digital: nature is not black and white, it is grey. Indeed the spectacular common sense of fuzzy logic may condemn it to the status of a truism. Whether a useful advance, or a just trendy name for what we knew already, fuzzy logic provides a philosophical alternative to

Aristotle.

202 But surely such a lack of precision cannot be accepted in classification? Vagueness creates ambiguity and undermines universality. Ornithologists, for exarrq)le, might use ‘species’ in a very different Avay from lepidopterists, and they would be quite justified in doing so. They would smq)ly enq)loy divergent, yet legitimate, interpretations of a vague species definition. Given the fuzziness of nature, how can taxonomy be consistent?

The answer hes in the fact that precision can be attained in abstract: species concepts can be precisely defined in theory, even if their apphcation remains somewhat vague. With an ideal species in mind biologists would mean the same thing, whether referring to the virtual species of a computer simulation or the real species of an ecosystem In nature, however, the populations known as species would only be best approximations based on the available evidence i.e. hypotheses.

So even though ornithologists and lepidopterists might have differing success in estimating species, at least they would be working towards a common end. The essence of their species would be the same in theory. Bird and butterfly species could then be compared legitimately, both being based on the same premise.

Pattern versus process

Biology is fortunate in having a u n if^ g theory, evolution. Having been extensively tested, evolution is almost unanimously accepted by biologists

(Dawkins, 1986). Biology therefore has a ready built standard on which to base its

203 classificatory system. However, the use of evolution in classification is rejected by one of the major schools of taxonomy, phenetics (Sokal and Sneath, 1963).

Phenetic classification is based solely on the pattern created by evolution. Some of its adherents, the transformed cladists, argue that it is inherently circular to test evolution using taxa generated from an evolutionary perspective (Platnick, 1979).

Classification, however, is not the primary evidence for evolution (Cain, 1962) and even if it were the argument is not circular but one o f ‘successive approximation’

(Ridley, 1986 p. 123). As more data accumulates supporting evolution and an evolutionary classification our confidence in both the theory and the taxonomy grows.

Aristotle might have hoped to find the essential pattern of life but evolutionary biology has shown that the pattern is ever changing. Species are not fixed, they cannot be captured in taxa because evolution always frees them from the phenetic straightjacket. Any classification which attempts to ignore evolution is thus doomed to failure. Phenetics would only work perfectly in the absence of evolution. Yet all this is no reason to reject phenetics; any classificatory system must expect practical problems in apphcation.

But is phenetics the best system available? Classification may be inevitably arbitrary to some extent, but that does not mean we should accept any old system

A taxonomy will only be universally accepted if it can be demonstrated in a scientific manner. The tragic flaw of phenetics is in theory not practise (it is

204 relatively easy to apply). Phenetics relies on cluster statistics to define taxa. But there is no objective measure of differential divergence (Ridley, 1986 p. 13). One man’s cluster is another man’s smudge, and consequently the phenetic school is ofl;en accused of science’s cardinal sin, subjectivity. Cluster statistics cannot be falsified. Phenograms are internally cohesive but lack external justification.

The cladistic school of taxonomy (Hennig, 1966), on the other hand, makes exphcit use of evolution. Cladism produces a hierarchy o f ‘natural’ groups based on homologies, which can be displayed m a branching diagram - a tree or cladogram (Eggleton and Vane-Wright, 1994). The central problem for cladists is recognising homology. Apparently homologous characters sometimes give conflicting signals (homoplasy) and consequently there may be several possible trees to choose fi'om. Distinguishing the ‘right’ tree (i.e. the one most likely to reflect the true phylogeny) may seem dangerously similar to choosing the ‘right’ cluster statistic, i.e. subjective. Although superficially appealing, this criticism is wrong.

Cladistics, unlike phenetics, is based on the scientific theory that evolution (descent with modification) produces a branching hierarchy. The practical implementation of cladistics ofl;en leads to several competing hypotheses i.e. several possible trees.

To discover a single ‘best’ tree, or set of trees, more specialised theories of evolution are sometimes invoked, e.g. maximum parsimony, maximum likelihood or neighbour-joining (Swofiford and Olsen, 1990). Special theories of evolution

205 are not universally accepted and may be falsified; as we learn more about evolution, new tree building algorithms generate more accurate trees.

Phenograms can also change with the addition of new characters, but the cluster statistic itself cannot be falsified. Whilst special theories of evolution can be tested and rejected, cluster statistics are untestable since they make no theoretical claims in the first place. Cladograms are hypotheses, phenograms are self-declared truths.

Science works through hypothesis testing, not acts of faith. Nevertheless, phenetic methodologv can often yield results very similar to those of cladistics. This seems particularly true of molecular studies, which is no surprise to the advocates of a neutral theory of molecular evolution (Kimura, 1983).

Cladistics seems to provide the basis for a universal system of classification, although it will be harder to apply than phenetics. Pheneticists just need to agree on a cluster statistic and stick with it. Some might welcome the imposition of such a statutory classification system. Evolutionary biologists, however, would prefer cladistics if practically possible. Phenetics clumps things according to their similarity. This produces a measure of phenotypic divergence which may be intuitive but does not allow evolutionary hypotheses to be tested (Harvey and

Pagel, 1991). The evolution of traits cannot be investigated and explained without some knowledge of their pedigree. Phenetics involves the subjective choice of a cluster statistic, yet even if it were objective the resulting phenogram would still be less usefiil in evolutionary biology than a cladogram.

206 The usefulness of cladograms in evolutionary biology (along with a more consistent theoretical basis) is the principal advantage of cladistics over another school of taxonomy, evolutionary taxonomy. Pure cladists (Hennig, 1966) prefer groups which contain all the descendants of a common ancestor. This is considered unnecessary by evolutionary taxonomists (Mayr, 1969) who exclude the longer branches of a monophyletic group, where evolution has increased phenetic divergence. Consequently, birds are excluded from the Reptilia by evolutionary taxonomists but included by cladists (Ridley, 1986 p.31-34).

Whilst paraphyletic groups such as Reptilia may be intuitive, they contain little for evolutionary biologists who are interested m phylogeny. However, it is the tree which provides the information; taxa are just symbols and “all defmitions can be omitted without loss to the information imparted.” (Popper, 1945 p. 18). How one names the branches of a tree does not really matter. What is important is that the tree represents an accurate reconstruction of phylogeny.

Problems with evolutionary systematics

Cladistic classification is based upon the hypothesis that evolution creates a hierarchical phylogeny. The existence of evolution is not in doubt (see Dawkins,

1986), and the practical problems of cladistics are not insurmountable. The real difficulty faced by cladistics occurs at the transition between the phylogenetic and tokogenetic levels (Hennig, 1966). At the phylogenetic level all characters are

207 shared vertically through common descent. At the tokogenetic level, however, trees become reticulate as characters are shared through both recombination

(horizontally) and common descent (vertically). This is the case with human culture. Cultural evolution is similar to organic evolution but its units, ideas, are easily transferred horizontally by imitation. Without the help of history books it would be impossible to distinguish homologous characters.

Units of evolution

Units of evolution are necessary in order to reconstruct phylogeny. The term ‘unit’ implies permanence; a unit is a fixed quantity. In a dynamic evolving system such permanence cannot mean immutabihty, instead it is achieved through common descent. The units of evolution change but retain their integrity through time.

Such units might be populations (species) or genes. Recombination breaks-up evolutionary units and confuses relationships. Hybridisation leads to the horizontal transmission of characters among species, whilst crossing-over leads to horizontal transmission of genetic code among genes.

At lower taxonomic levels cladistic relationships among groups become irresolvable because horizontal transmission is common, phylogeny is replaced by tokogeny. Only the gene phylogenies of non-recombining parts of the genome, such as mtDNA, remain sharp. However, these gene phylogenies do not necessarily reflect the population phylogeny (Pamilo and Nei, 1988). With more

208 evolutionary divergence it becomes easier to resolve relationships because horizontal transmission tends to zero between higher taxa. This is the ‘resolution effect’ shown in figure 1. (Unfortunately, at higher taxonomic levels it also becomes harder to recognise homologous characters, e.g. it is more difficult to ahgn homologous gene sequences.)

Ideally, evolutionary units are irrevocably isolated populations which form the terminal taxa of cladograms. Such ‘cladistic species’ are the lowest level at which cladism, given its assumption of a hierarchical pattern, is expected to fimction.

Species, so defined, are the smallest monophyletic units cohesive over time, or groups which share a common evolutionary fate. Only conspecific individuals can contribute directly to the genetic constitution of each other’s descendants.

Cladistic species

Species are constantly evolving, ever splitting into new species. Consequently it will be difficult to recognise recently formed species. In fact there is no reason why spéciation should ever be conq)leted: species might remain in limbo as semi­ species (hence hybrid zones, ring species etc.). Taxonomists trying to decide the status of two differentiated populations are like philosophers arguing over whether a glass of water is half full or half empty. Unfortunately this is an inevitable consequence of evolution; only creationists can hope to define species easily.

209 Nevertheless cladistic species are unambiguous in theory. Their existence can be falsified by the refutation of evolution. The power of the cladistic species definition is tested by examining how it handles situations which have traditionally fiustrated most attempts at a universal species concept. A few of these are discussed in detail below; the cladistic species can be apphed in all cases.

1. Asexuals

Bacteria range from being completely clonal e.g. Escherichia coli, to panmictic sexual populations e.g. Neisseria gonorrhoeae (Maynard Smith et al, 1993). In E. coli characters are only transferred between individuals vertically by descent. The absence of sex makes it easier to estimate phylogeny: horizontal transmission is inq)ossibIe both at the population level (no hybridisation) or at the gene level (no crossing-over). E. coli produces hierarchical phylogenies, whilst relationships among N. gonorrhoeae are reticulate (Maynard-Smith, 1993 and Figure 1).

E. coli individuals never fuse, they are irrevocably isolated. The smallest monophyletic population of individuals is therefore a singleton; each individual

E.coli is a cladistic species. Many would cite this as evidence that the cladistic species concept does not work. It is clearly ridiculous to give specific names to every E. coli individual! But why should all species be named? Names, after all, are only tools. In the case of E. coli it is not necessary to know the phylogeny of every individual (although it could be determined) so why bother to name them?

We retain the current specific name, although it does not describe a cladistic species but a superspecies. This is still an independent monophyletic group, but

210 the term superspecies recognises the potential to resolve an unambiguous phylogeny below this level.

2. Rampant hvbridisation: svngameons

Templeton (1989 p. 12) argues that species must be recognised by the complex interaction of genetic and ecological replaceabihty: “the most inclusive population of individuals having the potential for phenotypic cohesion through intrinsic cohesion mechanisms”. Cohesive species are exen^li&ed by cottonwoods and poplars. These plants are capable of producing perfectly fertile hybrids and they have been doing so frequently for a very long time. Yet they are real biological units because they have maintained their genetic, phenotypic and ecological integrity for about 12 milhon years (Eckenwalder, 1984). A similar example is found in the animal kingdom: coyotes and wolves hybridise in nature but are obviously real biological groups whose species status is beyond doubt (Hall, 1978).

These real biological groups are not strictly cladistic species. They share a common genetic reservoir. They may one day fuse. Templeton (1989 p. 18) provides a perfect example of this potential: Red and Black Oaks live together and cross-pollination is common although the hybrids do badly in the forest shade and fail to become estabhshed. The pure parental offspring flourish and the polymorphism is maintained. If^ however, the forest is thinned then the hybrids actually do better than the parental types: the two species merge. This clearly demonstrates the potentially ephemeral nature of ecological isolation when gene flow is still possible.

211 Red and Black Oaks, Cottonwoods and Poplars, Wolves and Coyotes may all be

‘real biological groups’ and as such they can be named. To call them species, however, is dangerous because it imphes potentially unwarranted evolutionary properties. In particular species are often assumed to be independent. Such an assumption is risky whilst the horizontal transmission of characters is still possible and can lead to serious misconceptions (Knowlton and Jackson, 1994).

4. Paraphvletic species

Cladistic species may sometimes be paraphyletic (nonexclusive) according to de

Quieroz and Donoghue (1988). This is because some individuals in recently separated clades may be more closely related to each other that they are to some members of their own clade. Similarly, if an individual of one species successfully hybridises and becomes integrated into another species, then the species is rendered paraphyletic. I would argue that this is merely the type of noise which must be expected when attempting cladistic systematics at the interface between phylogeny and tokogeny.

The biological species concept (Mayr, 1942) has also been criticised for creating paraphyletic species (Cracraft, 1989 p.46). Indeed paraphyletic biological species may be very common (Hafiier et al., 1987; Ackery and Vane-Wright, 1984).

Paraphyletic biological species are populations which have evolved intrinsic barriers to reproduction with respect to closely related taxa, yet remain capable of hybridisation with more distant taxa (Figure 2). In such a situation horizontal

212 transmission is possible among what appear to be monophyletic groups. Derived populations could therefore lose derived characters whilst ancestral populations might acquire them. Characters would not necessarily be identical by descent and relationships are more likely to be ambiguous.

Direct reproductive isolation is not a mystic property allocated to species by divine edict, it is just another character (or by-product of many characters). Paraphyletic biological species are therefore expected to occur frequently. However, if trees are well supported (if there is httle evidence of homoplasy), paraphyletic hybridisations have probably been rare in the past, although they may occur in the future. Such a situation could result from ecological or geographical barriers to gene exchange

(indirect reproductive isolation). With increasing divergence the potential for paraphyletic hybridisation is likely to be eliminated as both direct (intrinsic) and indirect barriers are strengthened.

If biological species are often paraphyletic, then direct reproductive isolation is not a useful character for determining cladistic species (Rosen, 1979). Strong and extended ecological or geographical isolation can also define species e.g. poplars and cottonwoods, wolves and coyotes. Cladistic spéciation requires permanent reproductive isolation which on occasion is provided by indirect means, such as ecology or geography. Often both indirect and direct reproductive isolation will act in concert. This is the essence of the cohesive species concept (Templeton,

1989).

213 In general, however, direct mechanisms seem safer criteria for determining specific rank because they often require complex evolutionary change (Coyne, 1992) wdiich is presumably hard to reverse. However, indirect reproductive isolation may sometimes be just as complex and difficult to reverse, e.g. Poplars and

Cottonwoods, whilst direct reproductive isolation occasionally has a very single genetic basis which might easily be reversed. This seems particularly true of prezygotic isolation; for example, single mutations which change the direction of shell whorl in snails can induce immediate and total prezygotic isolation (Orr,

1991). It is harder to see how post zygotic isolation could be reversed as this appears to involve complex genetic changes, for example, hybrid male sterihty in theDrosophila simulans clade (Cabot et al., 1994). These are unlikely to be overcome by neutral or selective evolution, although ‘hybrid rescue genes’ do occur (Rutter and Ashbumer, 1987).

In many ways cladistic species are similar to Cracraft’s phylogenetic species: “an irreducible (basal) cluster of organisms, diagnosably distinct from other such clusters, and within which there is a parental pattern of ancestry and descent”

(Cracraft, 1989 p.34-35). Phylogenetic species are any population that can be cladistically identified.

Cladistic species are a more conservative version of the phylogenetic species concept. The difference between them is that cladistic species are supposedly permanent. This is an important distinction. In practise species are often assumed to be evolutionarily independent. Cladistic species, monophyletic groups

214 characterised by complete pre- or postzygotic isolation and/or long-term geographical or ecological barriers to hybridisation, are usually independent.

However, a Hberal use of the species term (e.g. the phylogenetic species concept) erodes this useful property. Subspecies should be used where non-independence remains a distinct possibihty. The phylogenetic species concept would consider such subspecies full species, even though they may still be exchanging characters and have certainly done so in the very recent past. The border between subspecies and species is of course somewhat arbitrary, nevertheless it is a usefiil indication of confidence in the independence of the taxa.

Conclusion

Hierarchical phylogenetic trees can be reconstructed when groups do not recombine. Birds do not hybridise with fish, mammals, or butterflies, so we should be able to estimate their phylogeny. N. gonorrhoeae, by contrast, forms a virtually panmictic sexual population and below the species level there is no phylogeny to be resolved (Maynard-Smith, 1993; figure 1). Many populations, however, are likely to fall somewhere between these two extremes. As we try to resolve relationships among more closely related groups, the phylogeny becomes fuzzier as the probabihty of horizontal transmission rises (we approach the N. gonorrhoeae situation). This is not true of asexual organisms, such as E. coli, where the phylogeny remains sharply focused right down to the individual. Nor is it the case

215 with non-recombining genomes, such as those in asexual individuals or mtDNA, where the phylogeny is similarly sharp.

So at what point do we abandon trying to find a phylogeny; when is the fiizz factor

(amount of homoplasy) too high to go any further? There is no right or wrong answer, it depends on the level of confidence one requires for the independence of taxa. A cladistic subspecies is the smallest, identifiably monophyletic, population - the phylogenetic species of Cracraft (1989). Deciding whether a population diould be called a subspecies or not depends upon the confidence one has in its monophyly. A subspecies is therefore an hypothesis; the question is at what level of confidence should the hypothesis be accepted. Although it is still difficult to estimate confidence levels for phylogenetic nodes, there are methods for doing so, such as the bootstrap (Felsenstein, 1985). A subspecies might be accepted when there is at least 95% confidence in the node. Many island races are considered subspecies because geography imphes monophyly.

Specific rank, however, requires a higher level of confidence because of the important role species have throughout biology. Species should be largely independent evolutionary units. Our confidence that two species are independent inevitably rises with evolutionary divergence. Subspecies may differ at a single locus, species rarely do so. A cladistic species is the smallest groups of individual organisms which we beheve to be permanently monophyletic. They are ‘the largest integrating entitites immediately below the level of nonintegrating clade’ (Frost and

Kluge, 1994).

216 The difl&culty is therefore distinguishing nonintegrating clades (species) from integrating entities (demes). In an evolving system, species arise from integrating populations and the point o f ‘spéciation’ is unclear. In an Aristotelian world, the transition from integrating population to species (tokogeny to phylogeny) would be sharp, instantaneous. In the real world, however, species are fuzzy.

Bibliography

Ackery, P R. and Vane-Wright, RI. 1984. Milkweed Butterflies: Their Cladistics and Biology. BMNH/Comell Univ. Press, London and New York.

Cabot, E L , Davis, A.W., Johnson, N.A., Wu C-1, 1994. Genetics of reproductive isolation in the Drosophila simulans clade: complex epistasis underlying hybrid male sterility. Genetics 137:175-189.

Cain, A.J. 1962. The evolution of taxanomic principles. Symposia of the Society for General Microbiology 12:1-13.

Coyne, LA. 1992. Genetics and spéciation. Nature 355:511-515.

217 Cracraft, J. 1989. Spéciation and its ontology: the enq)irical consequences of alternative species concepts for understanding patterns and processes of differentiation. In: Spéciation and its Consequences, Otte D. and Endler, J.A. Eds.

Sinauer Associates Inc. Sunderland MA. pp. 28-59.

Dawkins, K 1986. The Blind Watchmaker. Longman Scientific and Technical,

Essex.

De Queiroz, K and Donoghue, M.J. 1988. Phylogenetic systematics and the species problem. Cladistics 4:317-338.

Eckenwalder, I.E. 1984. Natural intersectional hybridization between North

American species of Populns (Sahcaceae) in stcXions Aigeiros and Tacamahaca. m . Paleobatany and Evolution. Can. J. Hot. 62:336-342.

Eggleton, P. and Vane-Wright, KI. 1994. Some principles of phylogenetics and their inçlications for comparative biology. In: Phylogenetics and Ecology.

Eggleton, P. and Van e-Wright, KI. Eds. Academic Press, London, pp. 345-366.

Felsenstein, J. 1985. Phylogenies and the comparative method. Am. Nat. 125:1- 15.

Frost, D.K and Kluge, A G. 1994. A consideration of epistemology in systematic biology, with special reference to species. Cladistics 10:259-294.

218 Hafiier, M S., Hafiier, J.C., Patton, J.L., and Smith, M.F. 1987. Macrogeographic patterns of genetic differentiation in the pocket gopher Thomomys umbrintis. Syst.

Zool. 36:66-78.

Hall, KL. 1978. Variability and spéciation in canids and hominids. In: Wolf and

Man: Evolution in Parallel. Hall, KL. and Sharp, H.S. (eds.). Academic Press,

New York. pp. 153-177

Harvey, P H. and Pagel, M.D. 1991. The comparative method in evolutionary biology. Oxford University Press, Oxford.

Hennig, W. 1966. Phylogenetic Systematics. University of Illinois Press, Urbana.

Hutter, P. and Ashbumer, M. 1987. Genetic rescue of inviable hybrids between

Drosophila melanogaster and its sibling species. Nature 327:331-333.

Kimura, M. 1983. The Neutral Theory of Molecular Evolution. Cambridge University Press, N.Y., U.S.A.

Kosko, B. 1994. Fuzzy Thinking. HarperCollins, London.

Knowlton, N. and Jackson, J.B.C. 1994. New taxonomy and niche width partitioning on coral reefs: jack of all trades or master of some? Trends Ecol.

Evol. 9:7-9.

219 Maynard-Smith, J., Smith, N.H., O’Rourke, M. and Spratt, B.G. 1993. How clonal are bacteria? Proc. Natl. Acad. Sci. USA. 90:4384-4388.

Mayr, E. 1942. Systematics and the Origin of Species. Columbia University

Press, New York.

Pamilo, P. and Nei, M. 1988. Relationships between gene trees and species trees.

Mol. Biol. Evol. 5:568-583.

Platnick, N.I. 1979. Philosophy and the transformation of cladistics. Systematic

Zoology 26, 537-546.

Popper, K.R 1945. The Open Society and its Enemies. Vol. n The TEgh Tide of

Prophecy: Hegel, Marx, and the Aftermath. Routledge and Kegan Paul Ltd,

London.

Ridley, M. 1986. Evolution and Classification: the Reformation of Cladism.

Longman, New York.

Rosen, D.E. 1979. Fishes from the uplands and intermontane basins of

Guatemala: revisionary studies and comparative biogeography. Bulletin of the

American Museum of Natural History. 162:267-376.

220 Sokal, R.R. and Sneath, P.H.A. 1963. The Principles of Numerical Taxonomy.

W.H. Freeman, San Fransisco.

Tenq)leton, A.K 1989. The meaning of species and spéciation: a genetic perspective. In: Spéciation and its Consequences, Otte D. and Endler, J.A. Eds.

Sinauer Associates Inc. Sunderland MA. pp. 3-27.

Zadeh, L A 1965. Fuzzy sets. Information and Control 8:338-353.

221 Figure 1. The resolution effect (following Maynard-Smith, 1993)

A. Phylogeny for ideal cladistic species, E.coli or mtDNA gene. B. Phylogeny for real species and higher taxa. The blown-up section shows how phylogeny becomes net-like at lower taxonomic levels. C. Net-phylogeny of a panmictic sexual population, such as N. gonorrhoeae.

222 Figure 2. Paraphyletic hybridisation.

A* B D'

xyz

Populations A and D are monophyletic but retain the potential to hybridise with each other. Neither B nor C can hybridise either with each other, or with A or D.

ADB AD X

A and D have hybridised and exhanged phylogeneticallyinformative characters, such as z and z'. Depending upon which characters are examined population, AD can appear either ancestral or derived. The phylogeny is ambiguous.

223 CHAPTER 8

REPRODUCTIVE ISOLATION IN A BUTTERFLY HYBRID ZONE;

A SUBTLE HALDANE EFFECT

ABSTRACT

Anartia amathea and Anartia fatima form a hybrid zone in eastern Panama,

Interpopulation crosses were carried out to determine the extent and nature of any reproductive isolation. Substantial prezygotic isolation was evident in both directions of cross but there was no evidence of reinforcement. There was no sex ratio bias in any of the hybrid crosses but FI hybrid females in one direction of cross had less mating vigour. This supports Haldane's rule since females are the heterogametic sex in butterflies. The FI hybrids showed a tendency towards hybrid vigour, whilst the backcross and especially the F2 hybrids have reduced survivorship. Some of the F2 crosses involved siblings but inbreeding depression does not seem to explain the reduced survivorship of the F2 hybrids. All of the hybrid crosses produced some fertile offspring. The implications of these results for theories concerning the genetic basis of Haldane’s rule are discussed. This study emphasises the care that must be taken in assessing postzygotic isolation.

Attention has historically concentrated on hybrid survival and fertihty; this may significantly underestimate the true evolutionary costs of hybridisation.

224 INTRODUCTION

Despite the plethora o f ‘species’ definitions, the development of reproductive isolation is widely accepted to be a crucial step in spéciation (Mayr, 1942). There is a single generahsation concerning spéciation so defined: Haldane's rule (Haldane,

1922). This is the observation that where one sex of hybrid is missing, sterile, or inviable, it is usually the heterogametic sex.

Technological advances have sparked a resurgence of interest in the genetic basis of postzygotic isolation in general and Haldane's rule in particular (Charlesworth et al., 1987; Coyne & Orr, 1989a,b; Frank, 1991; Hurst and Pomiankowski, 1991;

Coyne, 1992; Wu, 1992; Wu & Davis, 1993; Orr, 1993; Turelh and Orr, 1995;

Davies and Pomiankowski, 1995). Increasingly sophisticated genetic markers, combined with hybridisation experiments, can reveal the genetic mechanisms that define species (Ten^leton, 1981).

Most investigations into postzygotic isolation have concentrated on taxa where the heterogametic sex is male, such as Drosophila and mammals (Coyne and Orr,

1989a; Wu and Davis, 1993). These studies alone cannot distinguish whether postzygotic isolation mostly afiOicts males because of their gender or because they are heterogametic. It was evidence fi’om birds and butterflies, where the heterogametic sex is female, that first demonstrated the significance of heterogamety (Haldane, 1922). The evidence, however, remains sparse. Wu and

Davis (1993) fisted 225 cases of postzygotic isolation recorded in Drosophila but

225 only 53 in birds and 55 in Lepidoptera. More hybridisation experiments in birds and butterflies are needed to establish the vahdity of Haldane’s rule and to determine its causes.

More such data would also help resolve a controversy surrounding the evolution of prezygotic isolation. Reinforcement describes the process whereby selection strengthens prezygotic isolation in order to reduce deleterious hybridisations

(Dobzhansky, 1937). Reinforcement was long accepted as an integral part of spéciation, however, mathematical models and a lack of empirical evidence have led many to question its inq)ortance in spéciation (Paterson, 1978; Butlin, 1989).

One crucial prediction of the reinforcement hypothesis is that hybridising populations will show stronger prezygotic isolation in areas where they overlap; the evidence, however, is limited and controversial (Wasserman and Koepfer,

1977; Ehrman, 1965; Hibino and Iwahashi, 1991).

In the late 1970s Robert Silberglied and Annette Aiello began a study to investigate spéciation in the neotropical butterflies, Anartia amathea and Anartia fatima. These common butterflies form a hybrid zone in eastern Panama. It is important to emphasise that these are tropical species because most hybrid zone studies, and investigations into reproductive isolation, have concentrated on temperate zone species (Hewitt, 1988). Spéciation must also be investigated in tropical regions, not least because it is here that most species are found. Species richness is a balance of spéciation (plus immigration) and extinction. If spéciation

226 is more prolific in the tropics this may not be merely a question of rate, but also

potentially of mechanism

The aim of this study is to determine the strength and nature of any reproductive

isolation between Anartia fatima and Anartia amathea. In particular does

Haldane's rule apply and is there evidence of reinforcement?

METHODS

Notation

To shrq)lify the text and tables Anartia amathea is abbreviated to ^A ’ and Anartia fatima to ‘F ’. FoUowmg convention the female is always shown first in the cross descriptions. The hybrids resulting from the cross: female Anartia amathea x male

Anartia fatima are hence coded the backcross o f ‘/IF ’ females to Anartia amathea males being 'AF/A \

Collecting sites

The wild-types were collected from the hybrid zone in eastern Panama where they are synq)atric, and from sites in Colombia (A) and the Panama Canal area (F) where they are allopatric.

227 Prezygotic isolation

Females are able to mate as soon as they emerge but resist mating more than once.

Males do not appear capable of mating in the first three days after eclosion although they will mate repeatedly. It was therefore vital that all the males were of the same age i.e. 3 days or more, and that all the females were virgins. Males were kept in holding cages until the 4th day and were then placed into an experimental cage with 1 day old virgin females. Two large experimental cages were used, each approximately 3m long x 3m wide x 2.5m high. These were located out-of-doors on Barro Colorado Island, Panama.

Six multiple-mate choice e?q)eriments were conducted: i) A (allopatric) and F

(allopatric), ü) A (sympatric) and F (sympatric), iii) A and AF, iv) F and AF, v) A and FA, vi) F and FA. Although there were sometimes different numbers of males and females in the experimental cage, each sex always consisted of equal numbers of each type e.g. in experiment i) half the males were A, half wereF; similarly half the females wereA and half were F. With the cage ‘balanced’ in this manner each of the four possible crosses should occur with equal frequency (0.25) if mating is random

Each experiment was run over several days. Butterflies were transferred from holding cages into the experimental cage on the morning of the first day. The cage was checked for matings every 10 minutes from dawn to 10 am and from 3 pm to dusk (matings usually lasted at least 20 minutes). Between 10 am and 3 pm they were constantly monitored because this is the period when the butterflies are most

228 active. The butterflies were left in the experimental cage overnight. Anartia becomes inactive when light levels are low, even when the sun goes behind a cloud, and so it was assumed that no matings would occur between dusk and dawn.

The cages were always kept ‘balanced’. Butterflies were transferred from the holding cages to the experimental cage each morning, and throughout the day, in order to correct for individuals which had died. When a mating was observed it was monitored and once the pair had separated the female was removed and placed in an oviposition cage. She was immediately replaced with another virgin female of the same type from a holding cage. Mated males were left in the experimental cage to mate again.

The precise number of butterflies in the cage at any one time varied (between about

30 and 100) throughout the course of each experiment, as butterflies died and the cages were kept balanced.

The strength of prezygotic isolation was initially tested separately for allopatric and sympatric populations of the parental species. The sympatric experiment was run from 8-15 July 1980 and involved a total of 726 butterflies, including \9\ A males,

189 F males, \13 A females, and 173 F females. The allopatric experiment took place from 30 July to 3 August 1980. A total of 325 butterflies were involved, including 80 A males, 77 F males, 85 A females, and 83 F females.

229 The backcross to A experiment was carried out 17-24 July 1976. It involved

264 butterflies (72 A males, 66 A females, 74 AF males, and 52 AF females). The v4F backcross to F experiment, which took place 9-12 July 1976, involved 180 butterflies (33 F males, 64 F females, 29 AF males, and 54 AF females). TheFA backcross to A experiment, which was carried out 1-4 August 1976, involved 93 butterflies (31 A males, 15 ^ females, 33 FA males, and 14 F4 females). The FA backcross to F experiment took place 9-12 August 1976 and involved 170 butterflies (45 F males, 42 F females, 36 FA males and 47 FA females).

Gilbert and Starmer (1985) recommended the statistic as a robust test for assortative mating. I have used the G-test which has some advantages over in particular it is additive (Sokal and Rohlf, 1973). With cages balanced for each type, random mating should produce each of the four possible crosses with equal fi*equency. Deviations from this ratio, however, do not necessarily iroply mate choice. If for example, A males were more vigorous in courtship than F males, then A males might take part in over half the total matings, even if mate-choice were completely random

The various cross types were therefore tested for differences in mating vigour. If mating vigour is the same there should be no difference in the total number of matings for each type, even if there are mating preferences e.g. A males should mate as often as F males, and A females should mate as often as F females. Again

G tests were used to detect deviations from the null hypothesis that mating vigour

230 did not difiFer between the species. Significant deviations suggest differential mating vigour between species.

If mated females had been lefi; in the cage then the opportunity to mate would have reduced with time (females seem to mate only once) which is likely to have a disproportionate effect on a ‘choosy’ species, reducing its overall mating fi-equency and hence its apparent vigour. By removing and replacing mated females this potential ambiguity was avoided. Less choosy species, however might also mate more often simply because it is easier for them to find a mate. This is possibly a problem in these experiments, although locating mates should not be a limiting factor in the relatively confined environment of a cage.

Mating vigour and mating preference can be confounding factors. Where there is evidence of both, it is not possible to determine (from data such as this) which makes the largest contribution to non-random mating.

Postzygotic isolation

Mated females were placed in oviposition chambers with a sprig of the larval food plant, Blechum brcfwneii. Blechum browneii is widely distributed throughout the neotropics and is used as a larval food plant by both A and F (Silberghed et a/.,

1979). Eggs were removed and divided into lots of twenty, which were placed in separate petri dishes. Each day the larvae were fed with freshBlechum browneii leaves and the dishes were cleaned. All broods (offspring of the same female) received the same treatment and the division into separate lots makes it unlikely

231 that any one brood would have been disproportionately affected by a viral infection or other external factors.

Nested G tests were performed to determine whether sex ratio differed significantly among cross types. The X-chromosome is often in^ortant in post­ zygotic isolation (Coyne and Orr, 1989b; Grula and Taylor, 1980a) and so sex ratio differences were also tested between hybrid classes grouped according to the specific origin of their ^^chromosomes. For some hybrids (those with a hybrid father) the origin could not be determined and so these were treated together as a separate class.

Survival rates were calculated for each dam at each stage of development (egg- larvae, larvae-pupa, and pupa-adult) and the sex ratio recorded. Nested G tests were performed to determine whether there was significant heterogeneity in survivorship depending upon cross type. Survivorship is presented for each brood as the number of ecolosions (individuals surviving to emerge as adults) and the number of non-éclosions (individuals dying before emerging as adults). This was done first treating crosses individually and then grouping the crosses into classes according to the specific origin of their chromo some or the nature of the cross

(parental, FI, backcross or F2).

232 RESULTS

Prezygotic isolation

Females refused to mate twice, but males mated would mate repeatedly.

Copulation time, however, increased with subsequent matings. The first mating usually lasted about 20 minutes, the second 30 - 60 minutes and subsequent matings for several hours. On several occasions 3rd and 4th matings lasted 12 hours.

There was no significant difference in the levels of assortative mating between individuals from the hybrid zone and those from allopatric regions (Table 1).

Without evidence of reinforcement, allopatric and sympatric populations were pooled for all subsequent analyses.

Significant deviation from random mating was observed between A and F in both directions of cross: both species mated assortatively (Table 2). The behavioural data is insufficient to determine whether male or female choice is responsible for this assortative mating. In courtship males follow females, settle next to them and then approach from behind or the side. Contact is estabhshed with the male’s antennae and legs probing the female’s abdomen or thorax. Males then position themselves parallel with the female and bend their abdomen around in an attenq)t to clasp the female’s abdomen and copulate. Females can reject these advances by pushing their abdomens up in the air, often protecting them between closed wings.

Females were seen to reject courtship and copulation attempts advanced by both

233 heterospecijBc and conspecific males. Males regularly courted females of both species, and they even atten^ted to copulate with heterospecific males. These observations suggest that female choice may be responsible for the assortative mating.

A and F females showed no significant difference in mating vigour but A males mated significantly more often than F males (Table 3). This does not change the fact that mating was assortative {A and F males mated significantly more often with conspecific females) however, the differences in mating vigour make it harder to compare the strength of that preference. At first sight A females (or maybe the males) appear more choosy than F females (or males). This, however, could result fi^om the lower mating vigour of F males. It cannot be determined which species has the stronger mating preference.

There was no evidence of assortative mating in the backcross experiments involving ^47^ hybrids (Table 2). However, there were significant differences in mating vigour. F and A females mated more often than AF females, whilstAF males mated more frequently than F males and as often as A males (Table 3).

No significant assortative mating or differences in mating vigour were observed in the experiments involving FA backcrosses (Table 2 and 3). Although FA males mated as frequently as A males and more often than F males, the sanq)le size was small and this difference failed to be statistically significant.

234 Postzygotic isolation

There was little heterogeneity in sex ratio among broods of the same cross type

(Table 4). Neither was there evidence of any deviation from 1:1 in the overall sex ratio for each cross type, nor of heterogeneity in sex ratio between cross types

(Table 5). There was still no evidence of a sex ratio bias among hybrids grouped according to the specific origin of their vY-chromosome (Table 6).

None of the hybrids was completely sterile; aU produced offspring except for the

F2 hybrids which were not mated.

There was a great deal of heterogeneity in survivorship among broods within each cross type, and between cross types (Table 7). There was also evidence for differing survival rates among four classes of hybrid cross (parental, FI hybrids, backcross hybrids and F2 hybrids) (Table 8). Within class heterogeneity was large

(G = 397.59; p<0.005) but between class heterogeneity was even greater (G =

885.90; p<0.005). FI hybrids actually survived slightly better than the parental types, wfrilst backcross and especially F2 hybrids had lower survival rates.

Unfortunately half of the hybridisations which yielded the F2 hybrids (FA/FA) were between siblings. Inbreeding depression may therefore have reduced F2 survivorship. The survival of FA/FA hybrids which resulted from sibling matings were compared with those that were between distant or completely unrelated individuals (Table 9). There was no significant effect of the sibling matings on

235 brood survivorship. A more rigorous assessment of the effect of inbreeding on offspring viabihty will be assessed in a future study.

Survivorship was also tested among the vY-chromo some classes (Table 6) and found much more heterogeneity within classes (G = 1004.96; p<0.005) than between them (G = 271.59; p<0.005).

Significance of results

Assessing the significance of hybridisation data must take into account the number of hybridisations tested i.e. the number of dams. Haldane (1922) used ten dams as his criterion of significance and decided that there was a significant bias in sex ratio only vdien one sex was twice as numerous of the other.

The survivorship of crosses represented by ten or more dams were therefore tested separately (Table 10). Again there was evidence of F2 hybrid inviabihty but the evidence for FI hybrid vigour and backcross hybrid inviabihty was somewhat weaker.

The /4F hybrids resulted from only 4 hybridisations and so, strictly speaking, we might reject their diminished mating vigour as insignificant. The reduction in mating vigour, however, was so great that it should probably not be discounted as merely a statistical fluke.

236 DISCUSSION

Evolution of prezygotic isolation

Coyne and Orr (1989a) conjectured that reinforcement probably explains why prezygotic isolation in Drosophila appears to evolve more quickly between sympatric species pairs than between those in allopatry. The reinforcement hypothesis predicts that we should find increased prezygotic isolation in areas where two closely related species overlap (but see Butlin, 1989). This is not the case in our study.

Prezygotic isolation between Anartia amathea and Anartia fatima is just as strong whether the populations are allopatric or sympatric. This is further evidence that hybrid zones rarely support the reinforcement hypothesis (Barton and Hewitt,

1981). Prezygotic isolation probably evolved between Anartia amathea and

Anartia fatima before they came into contact.

Evolution of postzygotic isolation

Hybridisation studies of butterflies are particularly useful tests for Haldane’s rule and associated theories of postzygotic isolation. This is because females are the heterogametic sex in butterflies. The results presented here show that Haldane's rule is obeyed but the effect is only subtle. In one direction of parental cross female mating vigour is inq)aired.

237 This reduced mating vigour would not count as postzygotic isolation according to the criteria of Coyne and Orr (1989a) where “a hybrid sex was considered to be viable if any adults of that sex appeared, even rarely. Similarly, it was considered fertile if any individuals of that sex were ever fertile.” An equally conservative approach was also followed by Wu and Davis (1993). Although understandable for comparative consistency, these criteria are very strict for studies seeking to investigate spéciation. It is not unreasonable to assume that in the early stages of spéciation hybrid fitness is only partially reduced.

Coyne and Orr (1989a) found that postzygotic isolation is rarely asymmetric in

Drosophila: only 5 out of 81 cases. This could either be evidence for the non­ linear evolution of postzygotic isolation or that the conservative criteria used in most reviews excludes populations at an early stage of spéciation. The genes that cause postzygotic isolation are different in each direction of cross (Wu and

Beckenbach, 1983). It is therefore unlikely that two genetic pathways should cause postzygotic isolation at exactly the same time, and so asymmetric postzygotic isolation should be a common first step in spéciation. If postzygotic isolation evolves non-linearly, however, a ‘snowballing’ effect may lead to reciprocal postzygotic isolation very soon after asymmetric isolation evolves (Orr,

1995).

Given that X-linked genes are important in postzygotic isolation (Coyne and Orr,

1989b), it follows that taxa with smaller X-chromosomes ought to take longer to evolve Haldane’s rule because of the reduced probability of X-linked mutations.

238 Indeed, Haldane’s rule might even be rare because inviability evolves almost as quickly in the homogametic sex as it does in the heterogametic sex. Where

Haldane’s rule has evolved in small-X taxa, it may take a considerable time before another mutation causes a Haldane effect in the reciprocal cross; inviability affecting both sexes may well evolve first (Turelh and Orr, 1995).

In butterflies the X-chromosome constitutes about 5% of the genome, compared with 25% in Drosophila. We therefore might expect asymmetric postzygotic isolation to occur more frequently and at larger genetic distances in butterfly hybridisations. Two of the most intensive investigations into reproductive isolation in the Lepidoptera both reported asymmetric postzygotic isolation.

Hagen and Scriber (1992) described the hybridisation of various species in the

Papilio glaucus species group. They found that Haldane’s rule is obeyed (females are rarer than males) but only in one direction of the cross. Although females are never entirely absent from any of the crosses, their frequency decreases as genetic distance increases between the parental species. This suggests that postzygotic isolation increases incrementaUy as taxa diverge, such a result is not expected if the evolution of postzygotic isolation quickly snowballs. Grula and Taylor (1980a,b) also reported asymmetry. They found that FI female hybrids of Coliasphilodice and Colias eurytheme were partially inviable and sterile in one direction of the cross. As for Anartia, neither theColias nor Papilio glaucus experiments would have qualified for postzygotic isolation in the Coyne and Orr review.

239 It is not clear whether the decreased mating vigour observed in AF females involved a sterihty effect or was solely due to reduced hybrid viabihty. A failure to produce fertile eggs might cause abnormal development of the reproductive tract which could somehow reduce sex-drive. Some of the female hybrids concerned may have been sterile (only 2 out of the 11 matings yielded offspring). The reduced mating vigour, however, seems most likely to be a result of generally low activity caused by abnormahties not related to egg formation. Mating vigour may be a useful way to assess the overaU viabihty of hybrids, in addition to conventional survivorship measures. In Colias butterflies Grula and Taylor

(1980b) also found a large reduction in FI hybrid female mating vigour.

The reduced survivorship of the backcross and particularly the F2 hybrids between

Anartia amathea and Anartia fatima provides support for the dominance theory as an explanation of Haldane’s rule (Muher, 1940; Orr, 1993; Turehi & Orr, 1995;

Davies and Pomiankowski, 1995; Chapter 9). This theory states that the mutations which cause postzygotic isolation tend to be recessive. Most of these wih be autosomal, and so wih not be noticed in FI hybrids (which are heterozygous at ah loci); recessive aheles are only expressed in FI hybrids when A^linked and in the heterogametic sex (where the chromo some is hemizygous). This explains both

Haldane’s rule and the large A^eflect.

The dominance theory makes some testable predictions, one being that postzygotic isolation should usuahy afflict both sexes of backcross and F2 hybrids (Coyne,

1994). In these hybrids loci of different species can be homozygous interact with

240 one another even if they are recessive. Such interactions among heterospecific loci may be incompatible and reduce hybrid fitness. There is some evidence for this from Drosophila (Davis et al. , 1994) and this study provides an exanç)le from the

Lepidoptera.

To conclude, Anartia amathea and Anartia fatima are reproductively isolated.

Both species mate assortatively, although there is no evidence of reinforcement.

The area of overlap probably represents a secondary contact tension zone (Key,

1986). When hybridisation does occur gene flow is impeded by a loss of female mating vigour in one direction of the parental cross, and by the relative inviabihty of backcross and F2 hybrids. The postzygotic isolation, however, is quite subtle and the reduction of an important fitness component in hybrid females would have been missed without hybrid mate-choice experiments. According to the criteria typicaUy used in reviews of reproductive isolation, Anartia amathea and Anartia fatima are not postzygoticaUy isolated. Whilst it is safest to be conservative in assigning reproductive isolation (and is often necessary in conq)arative studies) it is important to enqjhasise that these are minimum estimates. Hybrid mating behaviour and ecological characteristics are rarely tested adequately in hybridisation studies, despite the fact that they may be of great evolutionary significance.

241 BIBLIOGRAPHY

Barton, N.H. and Hewitt, G.M. 1981. Hybrid zones and spéciation. In :. Evolution and Spéciation, Essays in Honor of M.J.D. White. Atchley, W.K and Woodrufi^ D.S. (eds.). Cambridge University Press, Cambridge, pp. 109-145.

Butlin, R.K, 1989. Reinforcement of premating isolation. In: Spéciation and its Consequences, Otte D. and Endler, J. A. Eds. Sinauer Associates Inc. Sunderland MA.pp.158-179.

Charlesworth, B., Coyne, J.A. and Barton, N.H. 1987. The relative rates of evolution of sex chromosomes and autosomes. Am Nat. 130: 113-146.

Coyne, J.A., 1992. Genetics and spéciation. Nature 355: 511-515.

. 1994. Rules for Haldane’s rule. Nature 369: 189-190.

Coyne, J.A. and Orr, H. A. 1989a Patterns of spéciation in Drosophila. Evolution 43: 362-381

______. 1989b. Two rules of spéciation. In J.A. Endler and D. Otte (eds.). Spéciation and Its Consequences, pp. 180-207, Sinauer, Sunderland, MA.

Davis, A W., Noonburg, E.G. and Wu, C-I. 1994. Evidence for complex genic interactions between conspecific chromosomes underlying hybrid female sterility in ihQ Drosophila simulans cladQ. Genetics 137: 191-199.

Davies, N. and Pomiankowski, A. 1995. Haldane’s rule: old theories are the best. Trends Ecol. Evol. 10: in press.

242 Dobzhansky, T., 1937. Genetics and the Origin of Species. Columbia University Press, NY.

Ehrman, L. 1965. Direct observation of sexual isolation between allopatric and between sympatric strains of the different Drosophila paulistorum races. Evol. 19: 459-464.

Frank, S.A., 1991. Divergence of meiotic drive-suppression systems as an e?q)lanation for sex-biased hybrid sterility and inviability. Evol. 45: 262-267.

Gilbert, D.G and Starmer, W.T. 1985. Statistics of sexual isolation. Evol. 39: 1380-1383.

Grula, J.W. and Taylor, OR. 1980a. Some characteristics of hybrids derived from the sulfur butterflies, Colias eurytheme and Colias philodice: phenotypic effects of the chromo some. Evol. 34: 673-687.

. 1980b. The effect of A-chromo some inheritance on mate selection behavior in the sulfiir butterflies, Colias eurytheme and Colias philodice. Evol. 34: 688-695.

Hagen, RH ., and Scriber, J.M. 1992. Sex chromosomes and spéciation in tiger swallowtails. In: The Ecology and Evolutionary Biology of the Papilionidae. Scriber, J.M., Tsubaki, Y. and Lederhouse RC. (eds.) Scientific Publishers, Gainesville, FL.

Haldane, J.B.S. 1922. Sex-ratio and unisexual sterility in hybrid animals. J. Genetics 12: 101-109.

Hewitt, G.M. 1988. Hybrid zones - Natural laboratories for evolutionary studies. TREE 7: 158-167.

243 Hibino, Y., and Iwahashi, O. 1991. Appearance of wild females imreceptive to steriHsed males on Okinawa Is. in the eradication program of the Melon Fly, Dactis cucurbitae Coquillett (Diptera: Tephritidae). Appl. Ent. Zool. 26:265-270.

Hurst, L.D. and Pomiankowski, A. 1991. Causes of sex ratio bias may account for unisexuahty m hybrids: a new explanation of Haldane’s rule and related phenomena. Genetics 128: 909-920.

Key, K-H.L. 1986. Species, parapatry and the morabine grasshoppers. Syst. Zool. 17: 14-22.

Mayr, R. 1942. Sytematics and the Origin of Species. Columbia University Press, NY

Muller, H. J. 1940. Bearings of the Drosophila work on systematics. In: The New Systematics. Huxley, J.H.(ed.). Clarendon Press, Oxford, pp. 185-268.

Orr H.A., 1993. A mathematical model of Haldane’s rule. Evol. 47: 1606-1611.

. 1995. The population genetics of spéciation: the evolution of hybrid inconq)atibihties. Genetics 139: 139:1085-1813.

Paterson, H.E.H., 1978. More evidence against spéciation by reinforcement. S. Afi-. J. Sci. 74: 369-371.

Silberghed, R E., Aiello, A. and Lamas, G. 1979. Neotropical butterflies of the genus Anartia: systematics, life histories, and general biology (Lepidoptera: Nynq)hahdae). Psyche 86: 219-260.

Sokal, R R and Rohlf^ F.J. 1973. Introduction to Biostatistics. W.H. Freeman & Co., San Francisco.

244 Ten^leton, A R. 1981. Mechanisms of spéciation - a population genetic approach. Ann. Rev. Ecol. Syst. 12: 23 - 48.

Turelli, M. and Orr, H.A. (1995) The dominance theory of Haldane’s rule. Genetics 140:389-402.

Wasserman, M. and Koepfer, H. R. 1977. Character displacement for sexual isolation between Drosophila mojavensis and Drosophila arizonensis. Evol. 31: 812-823.

Wu, C-I. 1992. A note on Haldane’s rule: hybrid inviabihty \s hybrid sterihty. Evol. 46: 1584-1587.

Wu, C-I, and Beckenbach, A T. 1983. Evidence for extensive genetic differentiation between the sex-ratio and the standard arrangement of Drosophila pseudoobscura and Drosophila per similis and identif cation of hyrid sterihty factors. Genetics 105: 71-86.

Wu C-I, and Davis, A.W. 1993. Evolution of postmating reproductive isolation: the composite nature of Haldane’s rule and its genetic bases. Am.Nat. 142: 187-

212.

245 Table 1. Test for reinforcement.

Cross Sympatric Allopatric G d.f. P A/A 92 44 6.9996 3 N.S. A/F 16 14 F/F 64 36 F/A 40 10

246 Table 2. Assortative mating.

AxF* A male F male total G overall d.f. P A female 136 30 166 80.27 1 <0.005 F female 50 100 150 total 186 130 316

A X AF* A male AF male total G overall d.f. P A female 22 25 47 0.002 1 N.S. AF female 3 3 6 total 25 28 53

F X AF* F male AF male total G overall d.f. P F female 2 10 12 1.24 1 N.S. AF female 0 4 4 total 2 14 16

A X FA* A male FA male total G overall d.f. P A female 5 3 8 1.18 1 N.S. FA female 2 4 6 total 7 7 14

F X FA* F male FA male total G overall d.f. P F female 5 7 12 0.001 1 N.S. FA female 7 9 16 total 12 16 28

Experiment type shown in top left hand comer of each table.

247 Table 3. Mating vigour.

A x F* A F G 1:1 P Female 166 150 0.81 N.S. Male 186 130 9.98 <0.005

A x AF* A AF G 1:1 P Female 47 6 36.04 <0.005 Male 25 28 0.17 N.S.

Fx AF* F AF G 1:1 P Female 12 4 4.19 <0.005 Male 2 14 10.12 <0.005

A x FA* A FA G 1:1 P Female 8 6 0.29 N.S. Male 7 7 0.00 N.S.

FxFA* F FA G 1:1 P Female 12 16 0.57 N.S. Male 12 16 0.57 N.S.

G1:1 d.f. P Sum of female 41.9 5 <0.005 Sum of male 230.84 5 <0.005 Overall 62.74 10 <0.005

Experiment type shown in top left hand comer of each table.

248 Table 4. Sex ratio data.

Type Dam Maies Females Eclosions G (1:1) d.f. P G within cross d.f. P A 1 0 1 1 1.39 1 N.S. 14.45 13 N.S. 2 11 6 17 1.49 1 N.S. 46 10 21 31 3.99 1 N.S. 63 52 45 99 0.25 1 <0.05 65 34 33 68 0.00 1 N.S. 67 88 89 177 0.01 1 N.S. 75 74 72 150 0.00 1 N.S. 252 55 51 106 0.15 1 N.S. 280 1 0 1 1.39 1 N.S. 282 9 4 13 1.97 1 N.S. 304 4 9 13 1.97 1 N.S. 308 1 1 2 0.00 1 N.S. 316 3 5 8 0.51 1 N.S. 319 18 12 30 1.21 1 N.S. total 14 360 349 716 14.33 14 N.S. F 67 12 14 27 0.33 1 N.S. 8.75 11 N.S. 121 40 50 90 1.11 1 N.S. 165 14 11 25 0.36 1 N.S. 177 22 15 37 1.33 1 N.S. 292 2 1 3 0.34 1 N.S. 297 9 5 14 1.16 1 N.S. 327 63 51 115 1.05 1 N.S. 347 2 0 2 2.77 1 N.S. 354 12 8 21 0.43 1 N.S. 356 27 30 57 0.16 1 N.S. 499 3 2 5 0.20 1 N.S. 549 1 1 2 0.00 1 N.S. total 12 207 188 398 9.26 12 N.S. AF 32 139 153 292 0.67 1 N.S. 12.907 3 <0.005 124 99 114 213 1.06 1 N.S. 234 27 8 35 10.89 1 p<0.005 268 6 4 10 0.40 1 N.S. total 658 271 279 550 13.02 4 p<0.025 FA 5 57 40 97 2.99 1 N.S. 15.274 10 N.S. 12 6 3 9 1.02 1 N.S. 16 50 43 93 0.53 1 N.S. 18 15 12 27 0.33 1 N.S. 36 10 2 12 5.82 1 p<0.025 51 15 16 31 3.23 1 N.S. 54 2 4 6 0.68 1 N.S. 57 11 14 25 0.36 1 N.S. 579 8 1 9 6.20 1 N.S. 587 4 6 10 0.40 1 N.S. 591 2 4 6 0.68 1 N.S. total 11 180 145 325 22.24 11 p<0.025 AF/A 11 17 13 30 0.53 1 N.S. 0.002 1 N.S. 16 3 2 5 0.20 1 N.S. total 27 20 15 35 0.74 2 N.S. FA/A 3 1 0 1 1.39 1 N.S. 1.687 3 N.S. 6 20 13 33 1.50 1 N.S. 7 2 1 3 0.34 1 N.S. 14 8 8 16 0.00 1 N.S. total 4 31 22 53 3.22 4 N.S.

249 Table 4. Sex ratio data.

Type Dam Males Females Eclosions G (1:1) d.f. p G within cross d.f. p FA/F 212 0 1 1 1.39 1 N.S. 13.852 5 <0.025 219 1 5 6 2.91 1 N.8. 225 6 3 9 1.02 1 N.S. 243 0 4 4 5.55 1 p<0.025 266 7 5 12 0.33 1 N.S. 272 2 0 2 2.77 1 N.S. total 6 16 18 34 13.97 6 p<0.05 A/AF 256 13 12 25 0.00 1 N.S. 5.868 8 N.S. 313 1 2 3 0.34 1 N.S. 256 3 2 5 0.20 1 N.S. 276 3 3 6 0.00 1 N.S. 279 3 0 3 4.16 1 <0.05 311 8 5 13 0.70 1 N.S. 335 8 4 12 1.36 1 N.S. 298 16 14 30 0.13 1 N.S. 362 3 4 7 0.14 1 N.S. total 9 58 46 104 7.04 9 N.S. A/FA 398 3 4 7 0.14 1 N.S. n/a F/AF 76 1 2 3 0.34 1 N.S. 8.313 8 N.S. 253 0 1 1 1.39 1 N.S. 293 5 3 8 0.51 1 N.S. 362 35 52 87 3.34 1 N.S. 402 25 28 53 0.17 1 N.S. 424 7 14 21 2.38 1 N.S. 425 21 29 50 1.29 1 N.S. 435 26 18 44 1.46 1 N.S. 441 3 5 8 0.51 1 N.S. total 9 123 152 275 11.38 9 N.S. F/FA 616 3 2 5 0.20 1 N.S. 0.004 1 N.S. 650 2 1 3 0.34 1 N.S. total 2 5 3 8 0.54 2 N.S. FA/FA 8 5 6 11 0.01 1 N.S. 1.790 6 N.S. 24 1 0 1 1.39 1 N.S. 27 2 2 4 0.00 1 N.S. 30 1 1 2 0.00 1 N.S. 37 7 7 14 0.00 1 N.S. 222 1 1 2 0.00 1 N.S. 258 1 2 3 0.34 1 N.S. total 7 18 19 37 1.74 7 N.S. FA/AF 5 15 12 27 0.33 1 N.S.

250 Table 5. Sex ratio totals for each cross type.

Type Dams Male Female G (1:1) d.f. P A 14 360 349 0.17 1 N.S. F 12 207 188 0.91 1 N.S. AF 4 271 279 0.12 1 N.S. FA 11 180 145 3.78 1 N.S. AF/A 2 20 15 0.72 1 N.S. FA/A 5 31 22 1.54 1 N.S. FA/F 6 16 18 0.12 1 N.S. A/AF 11 58 46 1.39 N.S. A/FA 1 3 4 0.14 1 N.S. F/AF 12 123 152 3.06 1 N.S. F/FA 2 5 3 0.51 1 N.S. FA/AF 1 15 12 0.33 1 N.S. FAÆA 10 18 19 0.03 1 N.S. Total 91 1307 1252 12.81 13 sum of within cross heterogeneity from 1:1 12.81 13 N.S. between cross heterogeneity 11.63 12 N.S. overall heterogeneity from 1:1 1.18 1 N.S.

251 Table 6. Sex ratio: the influence of X-chromosome origin.

Origin of X-chromosome Type Dams Male Female G (1:1) d.f. P F AF 4 271 279 0.12 1 N.S. FA/F 6 16 18 0.12 1 N.S. A FA 11 180 145 3.78 1 N.S. AF/A 2 20 15 0.72 1 N.S. FA/A 5 31 22 1.54 1 N.S. AF A/AF 11 58 46 1.39 1 N.S. A/FA 1 3 4 0.14 1 N.S. F/AF 12 123 152 3.06 1 N.S. F/FA 2 5 3 0.51 1 N.S. F/VAF 1 15 12 0.33 1 N.S. FA/FA 10 18 19 0.03 1 N.S. Total 65 740 715 11.72 11 N.S.

within A heterogeneity 0.20 2 N.S. within F heterogeneity 0.06 1 N.S. within AF heterogeneity 5.03 5 N.S. sum of within class heterogeneity 5.30 8 N.S. between class heterogeneity 5.99 2 N.S. overall heterogeneity by cross 11.29 10 N.S.

252 Table 7. Heterogeneity in survivorship within and among cross types. Type ______Eggs Adults on-eclosion within G d.f.______p 2 1 1 73.13 14 p<0.005 20 17 3 66 31 35 280 97 181 176 67 108 1 0 1 300 177 123 333 146 183 204 106 98 6 1 5 27 13 14 23 13 10 15 2 13 34 8 26 70 30 40 108 26 81 243.13 12 p<0.005 118 90 28 81 25 56 50 37 13 9 3 6 128 14 114 290 114 175 38 2 36 86 20 65 130 57 73 32 5 27 39 0 39 28 2 26 AF 509 292 217 17.021 3 p<0.005 436 213 223 63 35 28 37 10 27 FA 179 97 82 48.699 10 p<0.005 13 9 4 215 93 122 42 27 15 26 12 14 80 31 49 6 6 0 51 25 26 41 9 32 21 10 11 41 6 35 AF/A 247 3030 217 0.9492 1 N.S. 26 5 21 FA/A 80 1 79 79 43.423 4 p<0.005 171 33 138 93 3 90 105 16 89 51 0 51

253 Table 7. Heterogeneity in survivorship within and among cross types.

Type Eggs Adults on-eclosion within G d.f. P FA/F 36 1 35 13.373 5 p<0.025 117 6 111 55 9 46 77 4 73 60 12 48 30 2 28 A/AF 67 25 42 91.649 10 p<0.005 23 0 23 4 3 1 1 0 1 6 5 1 21 6 15 7 3 4 20 13 7 13 12 1 138 30 108 88 7 81 M=A 44 7 37 n/a F/AF 20 0 20 197.97 11 p<0.005 1 0 1 31 3 28 29 1 28 66 8 58 116 87 29 6 0 6 72 53 19 40 21 19 83 50 33 63 44 19 23 8 15 F/FA 5 5 0 0 1 N.S. 3 3 0 FA/FA 173 11 162 25.986 9 p<0.005 31 1 30 18 4 14 166 2 164 297 14 283 86 0 86 26 2 24 20 0 20 13 0 13 111 3 108 FA/AF 211 27 184 n/a Total 7744 2559 5175

G d.f. P Sum of within cross heterogeneity 755.33 80 p<0.005 Between cross heterogeneity 1288.57 12 p<0.005 Overall heterogeneity among broods 2043.90 92 p<0.005

254 Table 8. Survivorship by cross type and hybrid classes. «n (S Origin of X chromosome class Cross Type Adults Non-eclosions n/a Parental A 709 841 n/a F 395 739 A F1 hybrid AF 550 495 F FA 325 390 A Backcross hybrid AF/A 35 238 A FA/A 53 447 F FA/F 34 341 A or F A/AF 104 284 A or F A/FA 7 37 A or F F/AF 275 275 A or F F/FA 8 0 A or F F2 hybrid FA/FA 37 904 A or F FA/AF 27 184

Class Adults Non-eclosed Survivorship within G d.f. P Parentals 1104 1580 0.41 32.19 1 <0.005 FI hybrids 875 885 0.50 8.76 1 <0.005 Back crosses 516 1622 0.24 335.70 6 <0.005 F2 hybrids 64 1088 0.06 20.94 1 <0.005 sum of within ciass heterogeneity 397.59 9 <0.005 between ciass heterogeneity 885.90 3 <0.005 Parentals 1104 1580 0.41 32.19 1 <0.005 Hybrids (A X-chromosome) 413 1075 0.28 225.26 1 <0.005 Hybrids (F X-chromosome) 584 836 0.41 249.74 6 <0.005 Hybrids (A or F X-chromosome) 458 1684 0.21 497.77 1 <0.005 sum of within class heterogeneity 1004.96 9 <0.005 between class heterogeneity 271.59 3 <0.005 Table 9. Test for inbreeding depression in FA/FA sibling matings.

Mating Dam Eggs Adults Non-eclosions within G d.f. P Sib 8 173 11 162 15.45 4 p<0.005 24 31 1 30 27 18 4 14 30 166 2 164 233 13 0 13 total 5 401 18 383 Non-si b 37 297 14 283 9.97 4 p<0.05 218 86 0 86 222 26 2 24 232 20 0 20 258 111 3 108 total 5 540 19 521

G d.f. P Sum of within type heterogeneity 25.42 8 p<0.005 Between type heterogeneity 0.57 1 N.S. Overali heterogeneity among broods 25.99 9 p<0.005

256 Table 10. Survivorship (crosses < 10 dams)

Class Adults Non-eclosed Survivorship within G d.f. p Parentals 1104 1580 0.41 32.19 1 <0.005 F1 hybrids 325 390 0.45 n/a 0 <0.005 Back crosses 379 665 0.36 52.04 1 <0.005 F2 hybrids 37 904 0.04 n/a 0 <0.005 sum of within class heterogeneity 84.23 2 <0.005 between class heterogeneity 620.49 3 <0.005

257 CHAPTER 9

HALDANE’S RULE: OLD THEORffiS ARE THE BEST

Introduction

Haldane's rule (Haldane, 1922) is the only regularity in spéciation. It states that when one sex of an interspecific hybrid is inviable, sterile or absent, it is usually the heterogametic sex (i.e. the XY sex). The remarkable consistency of this rule holds out the promise that spéciation might have a simple genetic basis, appHcable to a wide range of taxa from butterflies to mammals. Despite numerous hypotheses there is no satisfactory general explanation of Haldane's rule (Virdee, 1983) although Turelh and Orr (1995) present a theory which might finally solve this part of the spéciation puzzle.

Turelh and Orr’s (1995) theory is in fact not new at all, but continues the resuscitation began by Orr (1993a) of Muller's dominance hypothesis (Muller,

1940). Muller proposed that Haldane's rule results from the expression of recessive alleles in heterogametic hybrids. Homogametic hybrids are unaffected because all their loci are heterozygous. This idea, however, seemed dead in the water. It was experimentally tested using ‘unbalanced’ Drosophila females carrying two Xs from the same species. These females are fertile even though they are genetically equivalent to sterile FI hybrid males (Coyne and Orr, 1989).

258 Orr (1993a) and Wu and Davis (1993) recently pointed out a short coming in these experiments. If oogenesis and spermatogenesis genes are sex-specific then this might explain why unbalanced females are fertile because there is then no reason to e?q)ect hybrid male sterility to affect unbalanced females, even if they are genetically equivalent. In fact unbalanced female tests should only apply to inviability genes which are expressed in both sexes. Only two tests have been performed but both show unbalanced female inviability (Wu and Davis, 1993; Orr,

1993b). This insight removes a big obstacle to serious consideration of the dominance hypothesis.

The dominance model

Genes often act epistatically to fulfil physiological functions. Genes evolve together in coadapted complexes. When two species diverge new advantageous alleles are fixed. This may disturb co-adaptation and lead to a cascade of evolutionary change at epistatically linked loci. Eventually some alleles will not be interchangeable with their heterospecific counterparts because of their separate evolutionary histories. These will tend to be derived alleles (Orr, 1995) and their incompatibility will cause malfunction in the physiological pathways for which they are responsible i.e. cause a loss of function.

259 The central idea behind the dominance hypothesis is that coadapted gene conq)lexes function normally so long as a haploid set of genes is expressed in a hybrid. So even after non-interchangeable alleles have evolved, FI homogametic hybrids can be perfectly healthy because they retain a full haploid set of genes fi'om both species (see Figure). Things are intrinsically worse for the heterogametic sex because it lacks a paternal species X chromosome (assuming males are XY). The only X-linked genes present in the hybrid are therefore fi'om the maternal species.

A loss of function can arise with just a single dominant paternal mutation at a locus epistatically involved with X-linked genes (which are maternal in the hybrid). If maternal X-linked allele(s) and the dominant paternal allele are incorqpatible then

Haldane's rule will result.

X-linked alleles are vital to the dominance theory, but they need not be dominant and they are not really the ‘cause’ of Haldane's rule. In the heterogametic sex, X- hnked alleles act as ‘dominants’ by virtue of their hemizygosity. In the homogametic sex it is better to think of them as codominant. The unusual evolutionary event causing hybrid disruption is the appearance of an autosomal dominant allele that epistatically interacts with X-linked alleles. When the genes causing Haldane's rule are mapped, they will always appear on the X chromosome.

This produces the widely observed large X effect on hybrid disruption (Coyne and

Orr, 1989). But X-linked alleles are not sufficient for Haldane's rule. They always require other epistaticaly linked loci which will map to the autosomes.

260 Homogametic FI hybrids will not suffer equivalent fitness loss until the occurrence

of two dominant mutations, one in each species and both affecting the same

epistatically linked pathway. This is an extremely unlikely event if dominant

mutations are rare and pathways are usually short. If there are many pathways

epistatically involved with X-linked genes, heterogametic inviability in both

directions will evolve very quickly but there will be a long time lag before

homogametic isolation also appears.

Indeed it might be unlikely for two dominant incompatible mutations to ever occur

in a single pathway. This leads to the possibility that hybrid fitness loss in the

homogametic sex has a different cause. With time, pathways may diverge

sufficiently to acquire distinct fimctions in the diverging species. Brought together

in hybrids, these pathways may be distinct enough to make them inconq)atible.

This form of dysfimction will apply to both sexes and not show greater

involvement of X-linked alleles.

A unitary explanation?

An apparently serious obstacle to a general explanation of Haldane's rule is that hybrid sterility and inviability evolve at different rates depending upon the gender

of the heterogametic sex (Wu and Davis, 1993). Haldane's rule in taxa with heterogametic males (Drosophila and mammals) generally involves sterility, whilst in taxa with heterogametic females (birds and butterflies) Haldane's rule is more

261 evenly distributed between inviability and sterility. Remarkably the dominance theory might even explain this pattern.

There are no cases of mammahan hybrid inviabihty following Haldane's rule (Wu and Davis, 1993). This is expected because one X chromosome is inactivated in female mammals to achieve dosage compensation (Migeon, 1994). Females are effectively hemizygous for the X and the same as males. Both X chromosomes are, however, briefly active during oogenesis. As predicted Haldane's rule appUes to sterihty in mammals

Such an explanation cannot apply to Drosophila species as they dosage compensate by down regulating both Xs in females. But the same bias is seen, cases of Haldane's rule sterihty outnumbering inviabihty by 199:14 (Wu and Davis,

1993). Turelh and Orr (1995) put this down to the influence of the Y chromosome. The Drosophila V has no known somatic function but contains several fertihty factors (Hennig, 1985). The active involvement of the Y in gametogenesis and its permanent hemizygosity thus increase the number of potential epistatic dominance routes to sterihty. This is not a whoUy convincing explanation. Even though some cases of sterihty involve interactions with the Y chromosome (Coyne and Orr, 1989), it seems unlikely that this can explain the 14- fold excess of sterihty over inviabihty in Drosophila. It is possible that other explanations may contribute to this bias. Male reproductive characters may diverge more quickly as a result of sexual selection (Eberhard, 1985; Thomas and

Singh, 1992; Wu and Davis, 1993) or heterogametic spermatogenesis may be

262 particularly susceptible to evolutionary change (Wu and Davis, 1993). The problem with both these explanations is that if maleness or spermatogenesis are particularly unstable, why aren’t male butterflies and birds similarly affected?

Some data defnitely do not fit the dominance theory. X-linked genes seem to be important for homogametic sterility and inviabihty even though X and autosomes are predicted to have the same effect (Coyne and Orr, 1989). Turelh and Orr

(1995) optimisticaUy put this down to a ‘statistical fluke’. Indeed there are only two examples and both are in the Drosophila where the X chromosome constitutes about 1/4 of the genome. If similar data is obtained jfrom taxa where the X chromosome makes up only a smaU portion of the genome, such as

Lepidoptera (<5%), then we might have to go back to the drawing board.

There are some other interesting ways of testing the hypothesis. It seems likely that most derived aheles are co-dominant and wih have no deleterious effect in FI hybrids. Only dominant mutations are visible in the FI hybrids. But in F2 hybrids many co-dominant alleles become homozygous and can stop the expression of a full haploid set of genes. F2 sterility and invaibihty should appear equally fi'equently in both sexes and may even appear before Haldane's rule (ie, any FI disruption). Such F2 effects have been recorded in Drosophila (Coyne, 1994) and in butterflies (Chapter 8).

After being ignored for the last decade Muller's dominance theory turns out to be surprisingly good at explaining Haldane's rule, the large X effect and even

263 exceptions to the rule. Old theories often have merit even when they appear to have been refuted. Haldane's original explanation, that hybrid viabihty and fertihty require a conq)lete haploid genome from each species (Haldane, 1922), looks uncanily accurate as well. Substitute ‘either’ for ‘each’ and Haldane had it right all

along.

Bibliography

Coyne, J.A., 1994. Rules for Haldane’s rule. Nature 369: 189-190.

Coyne, J.A. and Orr, H.A. 1989. Two rules of spéciation. In: Spéciation and Its Consequences. Otte, D. and Endler, J.A. (eds ). Sinauer, Sunderland, MA. pp. 180-207.

Eberhard, W.G. 1985. Sexual selection and animal genitaha. Harvard University Press, Cambridge, MA.

Haldane, J.B.S. 1922. Sex-ratio and unisexual sterihty in hybrid animals. J. Genetics 12: 101-109.

Hennig, W. 1985. Y chromosome function and spermatogenesis in Drosophila hydei. Adv. Genet. 23, 179-234.

Migeon, B. 1994. chromo some inactivation: molecular mechanisms and genetic consequences. Trends Genet. 10,230-235

Muher, H.J. 1940. Bearings of the Drosophila work on systematics. In: The New Systematics. Huxley, J.H. (ed ). Clarendon Press, Oxford, pp. 185-268. 264 Orr H A , 1993. A mathematical model of Haldane’s rule. Evol. 47: 1606-1611.

. 1993b. Haldane’s rule has multiple genetic causes. Nature 361, 532-533

. 1995. The population genetics of spéciation: the evolution of hybrid incompatibihties. Genetics 139: 139:1085-1813.

Thomas, S. and Singh, R.S. 1992. A comprehensive study of genic variation in natural populations of Drosophila melanogaster. VII. Varying rates of genic divergence as revealed by by 2-dimensional electrophoresis. Mol. Biol. Evol. 9:507-525.

Turelli, M. and Orr, H A 1995. The dominance theory of Haldane’s rule. Genetics 140:389-402.

Virdee, S.R. 1983. Unravelling Haldane’s rule. Trends Ecol. Evol. 6, 71-72

Wu, C-I. and Davis, A.W. 1993. Evolution of postmating reproductive isolation: the composite nature of Haldane’s rule and its genetic bases. Am.Nat. 142: 187-

212.

265 Figure.

F1 male F1 female

X y X X

A a A a

B bdom B bdom

C c C c

inviable viable

Fl hybrids receive gaies from two species denoted by upper case (X, Y, A, B, C) and lower case (x, y, a ,b, c). Imagine that there is a physiological pathway X-A-B-C q)istatically linking an X-linked allele (X or x) to several autosomal genes (A B C or a b c). Hybrid males (heterogametic sex) inherit a full maternal haploid set of gaies but lack a paternal species X chromosome. If any of the paternal autosomal gmes act as dominants in the hybrid (b in this example) thaï hybrids lack expression of a full haploid set from either species. Under Muller's hypothesis these males are inviable. Hybrid females (homogametic sex) do not suffer fitness loss from single dominant alleles because they inherit an X diromosome from both species.

266