Speciation Around the Indian Ocean Basin

doi:10.1111/j.1420-9101.2008.01588.x

Establishing a perimeter position: speciation around the Indian Ocean Basin

G. VOELKER & R. K. OUTLAW Department of Biology, University of Memphis, Memphis, TN, USA

Keywords: Abstract dispersal; Historical biological interactions among peripheral landmasses on the peri- historical ecology; phery of the Indian Ocean Basin (IOB) are generally poorly understood. While Indian Ocean; interactions based on early Gondwanan vicariance have been used to explain thrushes; present day lineage distributions, several recent studies have instead inferred vicariance. dispersal across the IOB. This inference is often advanced because lineages under study have species inhabiting IOB islands. Here we examine the roles of continental vicariance vs. trans-IOB dispersal in the distribution of an avian genus found around the perimeter of the IOB. A molecular phylogeny does reveal evidence of a relationship that would require the inference of trans-IOB dispersal between eastern Africa and Sri Lanka. However, molecular clock data, ancestral area analyses and paleoclimatic reconstructions suggest that vicariance related to tropical forest expansion and retraction is more likely to have facilitated African–Asian interchange, with an initial colonization of Africa from Asia quickly followed by a recolonization of Asia. Subsequent dispersal from Asia to Sri Lanka and islands east of the Sunda Shelf are inferred; these latter islands were colonized in a stepping-stone fashion that culminated in colonization of the Sunda Shelf, and a recolonization of mainland Asia. We propose that circum-IOB distributions, which post-date early Gondwanan breakup, are most likely the result of continent-based vicariant events, particularly those events related to large-scale habitat alterations, and not trans-IOB dispersals.

mammals (Vrba, 1985, 1993) and plants (Treńel et al., Introduction 2007), suggest shifting climatic cycles that post-date The geological and ecological history of landmasses Gondwanan breakup drive major ecological changes, around the Indian Ocean Basin (IOB) provides evi- and have played a considerable role in driving speci- dence with which to test the roles that vicariance and ation in vertebrate groups distributed in both Africa dispersal may have played in the establishment of and Eurasia. circum-IOB lineages. Historically, vicariant mechanisms However, strict vicariance has proven insufficient to related to the geologically well-documented breakup of explain many present-day IOB distributions. Biogeo- Gondwanaland have been invoked to explain the graphic analyses of diverse lineages, from vertebrates to establishment of IOB distributions (McLoughlin, plants, have shown patterns of over-water dispersal 2001). More recently, biogeographic patterns in diverse within the IOB system (e.g. Vences et al., 2001; Raxwor- lineages, including birds (Voelker, 1999, 2002; Outlaw thy et al., 2002; Morley, 2003; Warren et al., 2003, 2005; et al., 2007), lizards (Amer & Kumazawa, 2005), Yuan et al., 2005; Rocha et al., 2006; Treńel et al., 2007). Increasingly, over-water dispersal has been inferred as a mechanism of colonization and speciation, because Correspondence: Gary Voelker, Department of Biology, University of Memphis, 3700 Walker Avenue, Memphis, TN 38152, USA. molecular clock dates suggest that many lineages Tel.: +1 901 678 1386; fax: +1 901 678 4746; diverged too recently to have experienced early Gond- e-mail: [email protected] wanan rifting, or are not inferred to have dispersed

ª 2008 THE AUTHORS. J. EVOL. BIOL. 21 (2008) 1779–1788 JOURNAL COMPILATION ª 2008 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY 1779 1780 G. VOELKER AND R. K. OUTLAW

during periods ecologically suitable for range expansion. intervening habitats between distributional areas were Thus, robust historical biogeographic analyses need to ecologically unsuitable for the group (i.e. requiring an explore the possibility of both scenarios as viable expla- inference of dispersal over those habitats). Evidence of nations for the observed, often allopatric distributions of either dispersal scenario was lacking in those previous lineages around the IOB. studies. Avian distributions in and around the IOB suggest that In this paper, we use a molecular phylogeny of a dispersal may indeed have played an important role in songbird genus in conjunction with ancestral area recon- establishing those distributions; birds fly, making long- struction methods and molecular clock and paleoecolog- distance over-water crossings more plausible than infer- ical data to empirically test vicariance vs. dispersal ring similar sweepstake dispersals for other vertebrate hypotheses with respect to circum-IOB distributions. lineages. This is particularly true with respect to highly Our focal genus is currently recognized as a constituent mobile avian lineages such as seabirds. While songbirds of Zoothera thrushes and has a distributional presence are typically considered less likely to cross large water around the margins of the IOB (Fig. 1). Zoothera is a barriers than are seabirds, numerous songbird lineages member of the avian family Turdidae, which has a have circum-IOB distributions (or nearly so). Although cosmopolitan distribution (Collar, 2005), thereby indicat- long-distance sweepstake dispersals (Simpson, 1953) ing a tendency for inter-continental movements that is across the IOB possibly explain these patterns, an reflected in both inter- and intra-generic biogeographic alternative would involve stepping-stone colonizations patterns. Many species within several Turdidae genera of land-masses in the IOB. Fuchs et al. (2006) and (particularly Myadestes and Turdus) have proven very Jønnson & Fjeldsa˚ (2006) have relied on an explicit successful in achieving over-water dispersal to islands and trans-IOB scenario directly linking avian lineages on the across the Atlantic (Collar, 2005; G. Voelker, S. Rohwer, eastern margin of the IOB (southeast Asia and Australia), D.C. Outlaw & R.C.K. Bowie, unpublished data), and with avian lineages in Africa. This scenario relies on Zoothera has clearly dispersed to Southeast Asian islands stepping-stone dispersal across Eocene–Oligocene land (see below). Therefore, while many songbird lineages are bridges, and currently submerged plateaus and sea- geographically restricted to single continents and as such mounts (see also Steenis, 1962), but the scenario is not are unlikely to cross ocean barriers, members of Turdidae unique to birds. A similar trans-IOB dispersal pattern appear far more likely to disperse across such barriers. connecting southeast Asia and Australia with Madagascar Furthermore, Zoothera thrushes and their close rela- and Africa has been proposed to explain distributions in tives are too young to have experienced initial Gond- bees (Fuller et al., 2005), lizards (Rocha et al., 2006) and wanaland rifting (Klicka et al., 2005; and below), other lineages (Morley, 2003; Sanmartıń & Ronquist, phylogenetic relationships directly linking species in 2004; Treńel et al., 2007). Africa with those in India, Sri Lanka or the islands of Elsewhere in the IOB, the Lemurian stepping-stones, Southeast Asia would strongly support trans-IOB dis- which connect India and Sri Lanka to Madagascar via the persal events (e.g. Raxworthy et al., 2002; Fuchs et al., Seychelles, are considered a likely IOB dispersal channel 2006). This is true even if the modern-day absence of for animals and plants (Yuan et al., 2005), and there is Zoothera on major IOB islands is interpreted as an clear phylogenetic evidence that a variety of lineages indication that IOB islands were never colonized by this have utilized them to some extent (e.g. Raxworthy et al., lineage. However, the absence of distributions could be a 2002; Warren et al., 2003, 2005; Rocha et al., 2006). function of extinction. While we are unaware of fossil However, the large number of avian (and other) evidence of Zoothera species from IOB islands, it is clear lineages distributed around the IOB suggests that vicar- that larger songbirds can quickly go extinct after human iant events post-dating Gondwanan breakup might better colonization (Steadman, 2006). Regardless, if Zoothera explain many of those distributions. The few phylogeny- historically utilized or dispersed over IOB islands such based studies that have dealt with this possibility in that they established distributions on multiple conti- songbirds (e.g. Voelker, 1999, 2002; Outlaw et al., 2007) nents, evidence of this should still be reflected in the have implicated climate change and commensurate phylogeny and dispersal could again be supported. habitat alterations as primary vicariant events driving Specifically, this historical utilization would directly link lineage divergence between continents surrounding the species in Africa with those in India or the islands of IOB. In each case, biogeographic reconstructions and Southeast Asia. molecular clock data supported continent-based move- Narrow ranges of species divergences, as reflected by ment in response to major paleoecological shifts. Phylo- very short internode distances, can also be indicative of genetic support for the alternative biogeographic scenario dispersal events. In such instances, barring repeated long- of dispersal could have been invoked if species with term migration of genes, colonists (dispersers) to new geographically disjunct IOB distributions (e.g. southern areas will rapidly differentiate from source populations Africa and Southeast Asia) were shown to be sisters. given their reduced genetic variability (Clegg et al., Dispersal could also have been invoked if clock dates 2002). Such a pattern would not generally be expected indicated that land-based movements occurred when in response to vicariant events, which typically affect

ª 2008 THE AUTHORS. J. EVOL. BIOL. 21 (2008) 1779–1788 JOURNAL COMPILATION ª 2008 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY Speciation around the Indian Ocean Basin 1781

Fig. 1 Generalized distribution of focal species around the Indian Ocean Basin. Stippled areas show broad multi-species distributions. Black arrows indicate vicariant movements; grey areas indicate dispersal events. The head of each grey arrow, and of all black arrows in the Southeast Asian Islands, indicates an island that is occupied by at least one focal species. The Lemurian Stepping-stone and African– Australian interchange over-water dispersal routes have been used to explain trans-Indian Ocean Basin distribution patterns in a variety of lineages (see text). population divergences over longer evolutionary periods, taxonomy (Sibley & Monroe, 1990). These taxa formed thus resulting in longer internode distances. two distinct clades (see below), with 18 taxa forming a Alternatively, the absence of Zoothera species on IOB clade (the sibirica clade) that is distributed around the islands could suggest that continental vicariant events are IOB (Table S1; Collar, 2005). We focus on this clade here. equally plausible alternatives that might better explain In general, we did not use multiple samples from species, circum-IOB distributions. In this instance, we would as detecting genetic breaks within species was not the expect that phylogenetic relationships would show a focus of the study. The major exception is citrina, which pattern of movement between adjacent areas of Africa has a broad distribution, several major allopatric popu- and Asia indicative of range expansion (biotic dispersion) lations and significant morphological diversity across rather than dispersal between distant areas (e.g. Voelker, those populations (Collar, 2005). We therefore sampled 1999, 2002; Outlaw et al., 2007). Using molecular clock from major citrina distributions to test for monophyly data and ancestral area reconstruction methods, we (Table S1). We extracted whole genomic DNA from each examined the roles that vicariance and dispersal have sample using DNeasy (Qiagen, Valencia, CA, USA) based played in the development of Zoothera distributions on the manufacturer’s protocols for tissue and blood. around the IOB. The NADH subunit 2 (ND2) and cytochrome b (cyt b) mitochondrial genes were isolated and amplified using Materials and methods standard PCR protocols. Gene amplification and sequencing were performed using primers designed for other thrushes (Voelker & Spellman, 2004; Outlaw et al., 2007) Taxon sampling and molecular methods and the following thermocycling parameters were used: We sampled 40 taxa comprising 29 species of Eurasian an initial 2-min denaturation at 94 C followed by 35–40 and African Zoothera, based on previously published cycles of: 94 C for 45 s (denature), 50–54 C for 45 s

ª 2008 THE AUTHORS. J. EVOL. BIOL. 21 (2008) 1779–1788 JOURNAL COMPILATION ª 2008 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY 1782 G. VOELKER AND R. K. OUTLAW

(anneal) and 72 C for 2 min (extension). A final Remaining trees were combined yielding a total of extension of 72 C for 10 min and a 4 C soak were 45 000 topologies from which a 50% majority rule used to insure reaction termination. Products were consensus tree was reconstructed. Nodes having posterior purified using the Qiaquick PCR purification kit (Qiagen) probability values of 95% or greater were deemed signif- based on the manufacturer’s protocols. icantly supported. We further assessed node support using We performed 20-lL cycle-sequencing reactions on ML bootstrap analysis in TREEFINDER (Jobb, 2005), using purified PCR products using BigDye dye-labelled term- the GTR + I + G model with 500 pseudoreplicates. inators (ABI, Foster City, California, USA) according to Subsequent analyses focused on the ‘sibirica’ clade (see standard protocols. Reactions were purified and target below); sibirica was identified in the above analyses as the DNA precipitated with Isopropanol. Cycle-sequencing basal (albeit distantly related) member of the focal clade, reactions were conducted on an ABI377-automated and it was used as an outgroup. This clade is distributed sequencer using acrylamide gels. Each sample was around the perimeter of the IOB (Fig. 1). In assessing sequenced for light and heavy strands. relationships within this clade, we again used the All sequence data (1041 bp ND2 and 998 bp cyt b) GTR + I + G for ML searches, based on new MODELTEST were aligned using Sequencher v. 4.5 (Gene Codes, Ann (Posada & Crandall, 1998) results. All ML and support Arbor, MI, USA) and by eye. For toepad samples, no analyses for this clade followed the protocols described fewer than two separate DNA extractions and sequencing above. reactions were performed, at intervals of up to several months apart. Repeated toepad extractions yielded iden- Molecular clock tical sequences. Relevant sequences (Table S1) are deposited in GenBank under the following accession In applying a molecular clock, many avian studies utilize a numbers: EU874398-EU874429. 2% sequence divergence per million year value, around which most calibrations for cyt b cluster (Garcıá-Moreno, 2004; Ho et al., 2005). Despite this, studies (Arbogast & Phylogenetic analysis Slowinski, 1998; Garcıá-Moreno, 2004) have questioned Several initial analyses were conducted on the combined the validity of this rate even when applied to lineages ND2 and cyt b dataset. Zoothera has previously been related to those initially calibrated (songbirds). An addi- suggested to be a polyphyletic genus based on limited tional problem is that most songbird lineages, including sampling (Klicka et al., 2005). Therefore, our first analysis Zoothera, do not have a deep fossil record to aid in applying determined to which clade each species belonged. We did temporal calibration points to phylogenies. In the absence this by including previously published sequence data of fossils, Barker et al. (2004) applied a calibration date (Klicka et al., 2005; Voelker et al., 2007) from most other from a geological event to the basal passeriform diver- Turdidae genera in a maximum likelihood (ML) search. gence, with the assumption that one divergent lineage MODELTEST 3.04 (Posada & Crandall, 1998) was used to was present on New Zealand during its rift from Antarc- select the most appropriate model of sequence evolution. tica. This method eliminates some of the issues surround- Hierarchical likelihood ratio tests (LRTs) and the Akaike ing the application of a DNA-based molecular clock. Information Criterion identified GTR + I + G as the best We employed this concept, relying on the historical fit model for our data. PAUP* (Swofford, 2000) was used ecology of the equatorial tropical forest of Africa, which for all ML searches. reached its maximum eastward extent (to the Kenyan An initial ML search was started from a HKY-85 coast) about 3 Mya (Hamilton & Taylor, 1991; Feakins neighbour-joining topology using MODELTEST para- et al., 2005; Sepulchre et al., 2006). Subsequent forest meters. After 5000 rearrangements without a change in retraction to the west from 3 Mya was rapid, as indicated likelihood score, parameters were re-estimated on the by the loss of east African equatorial high forest mam- resulting topology, and a subsequent search was initiated malian fauna from the fossil record 2.95–2.52 Mya using this topology and the re-estimated parameters. This (Wesselman, 1995). This would suggest that the estab- process was repeated a second time, running 10 000 lishment of east African montane refugia to the south rearrangements without a change in likelihood score or was also rapid. We therefore used 3 My to date the parameters. divergence between gurneyi and its sister clade (see Bayesian inference was implemented using MRBAYES below), all of which occupy montane areas around the 3.0 (Huelsenbeck & Ronquist, 2001) to assess nodal East African Rift system (Collar, 2005). This assumes support. We used the GTR + I + G model of sequence niche conservatism of species through time. Calibrating evolution, and each gene was allowed to vary in parameter this same node using a standard rate applied to songbirds estimates. Three Bayesian analyses were initiated from (2% divergence per My; Garcıá-Moreno, 2004; Ho et al., random starting trees, with four MCMC chains running for 2005) provides a congruent date of 3.2 My. two million generations and sampled every 100 genera- Divergence times were estimated by incorporating the tions, yielding 20 000 trees. The first 5000 trees from each calibration date into a nonparametric rate-smoothing analysis were discarded to ensure chain stationarity. analysis (Sanderson, 2002) using likelihood branch

ª 2008 THE AUTHORS. J. EVOL. BIOL. 21 (2008) 1779–1788 JOURNAL COMPILATION ª 2008 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY Speciation around the Indian Ocean Basin 1783

lengths estimated from our cyt b dataset. The Langley– The second major clade (guttata–cinerea) is comprised of Fitch method was applied under a truncated Newton a plumage-diverse assemblage of species that are distrib- algorithm for this analysis. uted around the IOB (Figs 1 and 2). Although nodal support for the placement of guttata and spiloptera is lacking (Fig. 2), all ML analyses placed them at the base Ancestral area reconstruction of this clade. These species are clearly not part of the For biogeographic analysis, we used both dispersal– African spot-winged thrush clade or of the clade to which vicariance analysis (DIVA; Ronquist, 1997) and weighted they are basal (the citrina clade); both of these clades are ancestral area analysis (WAAA; Hausdorf, 1998). Our well-supported (Fig. 2). focus was to determine the most likely ancestral area(s) ‘Zoothera’ citrina diverged early from a common ances- for the sibirica clade and to infer movement patterns tor with other Southeast Asian species (Fig. 2). The between major areas. Movements at finer scales are more cinerea clade is sister to citrina and comprises at least six readily interpreted from distributions mapped on the species, all of which occur primarily on Southeast Asian phylogeny. In both analyses, we coded each species as Islands (Figs 1 and 2). Although support is lacking for present or absent in each of four major areas: Asia, three nodes in this clade, molecular clock dates (and India + Sri Lanka (India), Africa and Southeast Asian distributions; see below) suggest rapid, and thus difficult Islands (Islands). To minimize the number of areas to support, divergences. recovered as potentially ancestral, we set the maximum Our analyses concur with the previous suggestion areas allowed at any node to two (the lowest possible (Klicka et al., 2005) that Zoothera is polyphyletic, and the setting) in DIVA, and the minimum Probability Index type (Zoothera monticola) falls in an assemblage of threshold in WAAA to 0.2. Austral-Asian species that is not sister to our focal clade (see Klicka et al., 2005; Voelker & Klicka, 2008). Results Following taxonomic priority for Zoothera (see Mayr & Paynter, 1964), the next available genus name is Geokichla. The type, by original designation, is Turdus Phylogenetic analyses and taxonomic considerations citrinus, which is included in our focal clade (Geokichla Klicka et al. (2005) have suggested that Zoothera com- citrina;Mu¨ ller, 1835, in Mayr & Paynter, 1964). We prises up to three distinct lineages. Our results, which therefore recognize Geokichla as the valid taxonomic incorporated additional species sampling, confirm designation for all members of the sibirica clade, and Zoothera polyphyly, with all species sampled falling into apply it below. two distantly related (nonsister) clades (not shown). ML analyses identified 18 species as belonging unambigu- Biogeography ously to a derived clade of Afro-Asian ‘Zoothera’, and we focus on this clade here. Both methods of ancestral area reconstruction identify Initial ML analyses identified sibirica as the basal extant Asia as the most likely area for the origin of Geokichla taxon in the focal clade, with wardii being the next to (Fig. 2, Table S2). Geokichla sibirica and wardii are distrib- diverge (not shown). Subsequent analysis with sibirica as uted in Asia and both are inferred to have arisen in the an outgroup confirms the position of wardii (Fig. 2). late Miocene (Fig. 2). The sister clade to Geokichla Bayesian analysis indicates posterior probabilities of 1.0 comprises the genera Psophocichla (monotypic) and for the position of these taxa relative to the remaining Turdus (65 species) (Voelker et al., 2007). While the species, and ML bootstrap support is similarly high former is distributed in Africa, the ancestral area for (Fig. 2). Turdus is clearly Asia, and several purely Eurasian species Results of ML analyses indicate that the next divergence are the basal members of this genus (Voelker et al., 2007). resulted in two distinct clades to which all remaining focal Therefore Asia would remain the most likely ancestral ‘Zoothera’ belong (Fig. 2). The first of these clades (camer- area for Geokichla in a broader multi-genus analysis. onensis–gurneyi) comprises the African spot-winged Despite arising approximately 8.5 Mya, just two thrushes, all of which have distributions that include the speciation events appear to have occurred in Asia over East African Rift system. Our results confirm a prediction the next 4.5 My (Fig. 2). From 4–3.4 Mya, lineage from morphology (Klicka et al., 2005) that crossleyi and diversification increased, following the colonization of oberlaenderi are members of this clade (Fig. 2). The phylo- Africa (Figs 1 and 2). Most African species form a single genetic placement of oberlaenderi suggests that it is best clade, the spot-winged thrushes, and their diversity is recognized as a distinct species; it has previously been centred around the East African Rift system (Collar, recognized as a race of both piaggiae and gurneyi. The ML 2005). As all have distributions including the Rift system, topology further indicates that crossleyi is not a race of this is the most likely ancestral area for this clade. gurneyi, where it has been placed in the past despite Additional sampling of races and isolated populations substantial plumage differences (Collar, 2005). in the Rift system is necessary to resolve taxonomic

ª 2008 THE AUTHORS. J. EVOL. BIOL. 21 (2008) 1779–1788 JOURNAL COMPILATION ª 2008 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY 1784 G. VOELKER AND R. K. OUTLAW

cameronenensis Africa 2.49 princei Africa crossleyi Africa 3.58 2.49 oberlaenderi Africa AFRICA 2.57 piaggiae Africa 2.12 3.0 tanganjicae Africa gurneyi Africa cinerea Philippines interpres Sundas, Borneo,Asia 1.74 2.46 dohertyi Lesser Sundas(W) ISLANDS 4.14 1.7 8 AFRICA peronii Lesser Sundas(E) 1.98 schistacea Tanimbar 2.09 ISLANDS 3.37 erythronota Sulawesi citrina Borneo Borneo 1.60 5.45 citrina China Asia ASIA INDIA INDIA 2.74 3.87 citrina Ghats India 1.07 AFRICA 4.07 citrina Nepal Asia ASIA 8.42 spiloptera Sri Lanka guttata Africa wardii Asia, India sibirica Asia 0.05

Fig. 2 Left: maximum likelihood phylogeny of Geokichla species. Where no major genetic or biogeographical differences exist, species represented by multiple samples were collapsed to single exemplars. Circles at nodes indicate support values, with Bayesian posterior probabilities in the top hemisphere (‡ 95% in black, 85–94% in grey), and maximum likelihood bootstrap support in the lower hemisphere (‡ 90% in black, 75–89% in grey, otherwise white). Numbers at nodes are lineage divergences estimates, in millions of years. Centre: distributional area for each species. Right: ancestral area reconstructions based on dispersal–vicariance analysis and weighted ancestral area analysis. affinities and biogeographic reconstructions of the spot- suggest a stepping-stone pattern of colonizations through winged thrushes. Southeast Asian Islands (Figs 1 and 2). Five islands must Although guttata is distributed in Africa, it is clearly have been colonized via dispersal, as they fall east of the sister to an Asian clade, and not to the African spot- Sunda Shelf (Fig. 1) and therefore have no historical winged thrushes (Fig. 2). Thus, we infer from ancestral interconnection between them or the mainland. Rapid area analysis a recolonization of Asia, via India (Figs 1 dispersal across islands, with subsequent loss of gene and 2; Table S2), which accounts for the presence of flow, would account for rapid divergences evident in the spiloptera on Sri Lanka, and citrina throughout much of phylogeny (Clegg et al., 2002). Southeast Asia. Two movements to Southeast Asian Islands follow. One of these movements can explain the Discussion distribution of citrina aurata on Borneo, while a second movement would explain the distribution of the cinerea Which process, dispersal or vicariance, explains more of clade throughout Southeast Asian Islands (Fig. 2). the variance in Geokichla distribution? Recalling the ability Within the cinerea clade, most species distributions of Turdidae (and other) lineages to disperse over water reflect a high degree of island endemism (or nearly so if barriers, and more specifically the distribution of some small satellite islands are considered; Figs 1 and 2). Only Geokichla species on oceanic islands (Fig. 2), it is imprac- interpres has a broad distribution, which includes overlap ticable to a priori hypothesize an inability to disperse on Borneo with citrina, and a recolonization of mainland across the IOB. Indeed, we note that barring any corre- Asia (Figs 1 and 2). Distribution across, and speciation in, lation between divergence dates, paleoecology and hab- the Southeast Asian Islands occurred rapidly. Molecular itat adaptation, the structure of the phylogeny (Fig. 2, left) clock dates suggest that initial diversification in this would clearly support a dispersal event between south- lineage occurred in the late Pliocene, with a pulse of eastern Africa (guttata) and Sri Lanka (spiloptera). speciation occurring from 2.09–1.74 Mya (Fig. 2) near Despite phylogenetic structure, the most parsimonious the Plio–Pleistocene boundary. Distributions in this clade explanation for African–Asian interchange in Geokichla is

ª 2008 THE AUTHORS. J. EVOL. BIOL. 21 (2008) 1779–1788 JOURNAL COMPILATION ª 2008 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY Speciation around the Indian Ocean Basin 1785

vicariance. The colonization of Africa is dated at 4.1 Mya, formations formed aerial land-masses (Fuchs et al., 2006; and recolonization of Asia occurred almost simulta- Jønnson & Fjeldsa˚, 2006) at temporal intervals consistent neously (Fig. 2). From 5–3 Mya, African tropical forests with Geokichla evolution. Finally, the lineage divergence extended to coastal Kenya and through Ethiopia (Ham- dates inferred in connection with these potential plateau- ilton & Taylor, 1991; Cane & Molnar, 2001; Feakins et al., facilitated dispersals are far older (ca. 45 Mya) than those 2005; Sepulchre et al., 2006). Divergence dates suggest we consider here (Fuchs et al., 2006; Jønnson & Fjeldsa˚, that ancestral movements between Asia and Africa 2006). We therefore conclude that trans-IOB dispersal occurred during this period, when forest connectivity by Geokichla is unlikely to have taken place, and that was high. Virtually all Geokichla are (now) forest adapted, vicariant mechanisms better explain IOB perimeter making intercontinental movements during periods of distributions in Africa and continental Asia. forest expansion a likely event. Similar ‘suitable paleo- Dispersal can, however, explain Geokichla distributions ecology’ arguments have, when temporally associated on islands located on the IOB perimeter. Geokichla with lineage divergences, been used to explain African– spiloptera, endemic to Sri Lanka, diverged from its sister Asian interchange in lizards (Amer & Kumazawa, 2005), clade of derived Asian Geokichla in the early Pliocene. and nonforest-adapted birds (Voelker, 1999, 2002) and During this period, it appears that northern Sri Lanka had mammals (Vrba, 1985, 1993). subsided below sea level, isolating southern Sri Lanka Following intercontinental movements by Geokichla, from India (Jacob, 1949). Dispersal may therefore African tropical forests began to rapidly retract westward account for the distribution of spiloptera. between 3–2.5 Mya (Hamilton & Taylor, 1991; Feakins Dispersal is inferred to explain the distributions of et al., 2005; Sepulchre et al., 2006). This retraction served Geokichla species on the northern Philippines, Sulawesi, as a driving vicariant event that eliminated Asian–African Tanimbar and the Lesser Sunda Islands. These islands are forest connectivity, and montane forest connectivity in not part of the Sunda Shelf system, which precludes the Eastern Africa. The resulting retraction greatly impacted possibility of historical movements across the exposed small mammal distributions (Wesselman, 1995), and shelf during periods of low sea level. Most Southeast vicariance-driven speciation in forest refugia would Asian Islands have been in their current geographical explain the major radiation of African thrushes, in which position for 10 My (Hall, 1998), which, when considered most speciation events occurred after 3.0 Mya. This with molecular clock dates, eliminates the possibility of pattern of lineage diversification via forest fragmentation Geokichla distributions resulting from tectonic rifting. The has been proposed for other thrushes, which, like most distribution of interpres on Sunda Shelf islands and African Geokichla, are isolated in montane forest refugia mainland Asia (Figs 1 and 2), and a race of citrina on around the African Rift system (Bowie et al., 2005). Borneo, is likely due to vicariant isolation following Although a vicariant explanation is plausible to explain movement onto the islands during periods of low sea divergence between Asian and African lineages, the level. Short-range dispersal is necessary to account for ranges of guttata (southeast coast of Africa), spiloptera (Sri the interpres–dohertyi sister relationship; dohertyi is distrib- Lanka) and the cinerea clade (southeast Asian islands) uted east of the Sunda Shelf. A similar pattern of require exploring the possibility of dispersal across the dispersal across nonshelf islands, followed by recoloniza- IOB. Avian lineages have clearly used the Lemurian tion of the mainland or Sunda Shelf islands, has also stepping-stones to colonize IOB islands from both India been documented in the avian genus Ficedula (Outlaw & and Africa (Warren et al., 2003, 2005), and lineages far Voelker, 2008). The question of why Geokichla skipped less mobile than birds, such as chameleons, have over the Sunda Shelf islands early in its biogeographic dispersed from Madagascar to India (Raxworthy et al., history remains. Given that Geokichla subsequently and 2002). The trans-IOB distributions of bees (Fuller et al., rapidly established distributions on Sunda Shelf islands, 2005), Cryptoblepharus lizards (Rocha et al., 2006) and extinction of ancestral Sunda Shelf lineages seems plants (Morley, 2003; Sanmartıń & Ronquist, 2004; unlikely. Treńel et al., 2007) have been explained by dispersal, The narrow range of divergence dates (ca. 0.75 My) often times with molecular clock support for that across the cinerea clade suggests rapid island coloniza- conclusion, rather than by land-based vicariant explana- tions. The initial divergence of this clade from its sister tions. (citrina) is coincident with the closing of the Indonesian Although there is clear evidence that diverse lineages seaway 3–4 Mya (Cane & Molnar, 2001); associated have dispersed across the IOB, there is no evidence to regional climate and habitat changes might explain rapid support trans-IOB dispersal by Geokichla from Africa to colonization of islands. These alterations positively either southern Asia or Southeast Asian Islands. No impacted the ability of vertebrates to expand from Asia Geokichla species are distributed on Madagascar or other onto Sunda Shelf islands and beyond (Tedford, 1985; potential Africa-to-Asia stepping-stone islands; phylo- Voelker, 1999; Gorog et al., 2004). Divergence dates genetic evidence suggests no relationships that might between members of the cinerea clade also imply rapid have required utilization of these islands. Further, evi- speciation, which is expected unless continued dispersal dence is lacking that submerged plateaus and seamount from source areas maintains gene flow (Clegg et al.,

ª 2008 THE AUTHORS. J. EVOL. BIOL. 21 (2008) 1779–1788 JOURNAL COMPILATION ª 2008 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY 1786 G. VOELKER AND R. K. OUTLAW

2002). Species relationships within the clade suggest that patterns, we suggest that, barring extant or fossil taxa on distributions were achieved by a series of stepping-stone IOB islands or future evidence of seamount exposure that movements, beginning with colonization of the Philip- is temporally relevant, vicariant speciation related to pines, and ending with colonization of islands on the continental paleoecological shifts should be considered eastern perimeter of the IOB (Figs 1 and 2). Relationships the null hypothesis for any nonseagoing lineage distrib- and distributions within the citrina clade indicate a uted around the perimeter of the IOB. complex pattern of mainland–Sunda Shelf–nonshelf island dispersals similarly reflected in other lineages Acknowledgments (Evans et al., 2003; Filardi & Moyle, 2005). It is evident that vicariance and dispersal have both We thank the Curators and staff of those museums that operated in generating Geokichla distributions around the provided us with samples. We thank E. Dickinson and perimeter of the IOB, and that they have done so at S. Gregory for clarifying the taxonomic vagaries sur- different temporal and geographical scales. Vicariance via rounding Geokichla vs. Geocichla. Two anonymous referees ecological discontinuity is evident early in Geokichla made constructive comments. This work was supported in evolution, and involved African–Asian interchange and part by NSF DEB-0613668 to GV and R. C. K. Bowie. subsequent divergence. This interchange is similar to historical geographic patterns found in similarly distrib- References uted songbird lineages (Voelker, 1999, 2002), yet temporally dissimilar from those lineages due to differing Amer, S.A.M. & Kumazawa, Y. 2005. Mitochondrial DNA ecological associations. Similar temporal ecological asso- sequences of the Afro-Arabian spiny-tailed lizards (genus ciations, driven by repeated climatic shifts (Hamilton & Uromastyx; family Agamidae): phylogenetic analyses and Taylor, 1991; Feakins et al., 2005; Sepulchre et al., 2006), evolution of gene arrangements. Biol. J. Linn. Soc. 85: 247–260. have also proven important to the establishment of inter- Arbogast, B.S. & Slowinski, J.B. 1998. Pleistocene speciation and the mitochondrial DNA clock. Science 282: 1955a. and intra-continental distributions in other vertebrate Barker, F.K., Cibois, A., Schikler, P., Feinstein, J. & Cracraft, J. lineages (Vrba, 1985, 1993; Wesselman, 1995; Amer & 2004. Phylogeny and diversification of the largest avian Kumazawa, 2005). radiation. Proc. Natl Acad. Sci. U.S.A. 101: 11040–11045. Such ecological associations allow a general temporal Bowie, R.C.K., Voelker, G., Fjelsda˚, J., Lens, L., Hackett, S.J. & prediction for African–Asian interchange of lineages: Crowe, T.M. 2005. Systematics of the Olive Thrush Turdus forest-adapted species (e.g. sunbirds; R.C.K. Bowie, olivaceus species complex with reference to the taxonomic unpublished data) and thrushes (this study) should status of the endangered Taita Thrush T. helleri. J. Avian Biol. colonize during periods of forest expansion (5–3 My), 36: 391–404. whereas nonforest-adapted species (e.g. pipits, Voelker, Cane, M.A. & Molnar, P. 2001. Closing of the Indonesian 1999; wagtails, Voelker, 2002; rock-thrushes, Outlaw Seaway as a precursor to East African aridification around 3–4 million years ago. Nature 411: 157–162. , 2007) should colonize during periods of forest et al. Clegg, S.M., Degnan, S.M., Kikkawa, J., Moritz, C., Estoup, A. & retraction. Further, forest-adapted lineages should expe- Owens, I.P.F. 2002. Genetic consequences of sequential rience increased cladogenesis during periods of forest founder events by an island-colonizing bird. Proc. Natl Acad. retraction, and arid species should do the same during Sci. USA 99: 8127–8132. periods of forest expansion. Thus far, these predictions Collar, N.J. 2005. Family Turdidae (Thrushes). In: Handbook of seem to hold for songbird lineages and fossil mammal Birds of the World, Vol. 10 (J. del Hoyo, A. Elliott & D.A. lineages; analyses of other lineages are clearly necessary Christie, eds), pp. 514–807. Lynx Edicions, Barcelona. to test whether this prediction applies more broadly Evans, B.J., Brown, R.M., McGuire, J.A., Supriatna, J., Andayani, across other vertebrate lineages. Another potential vicar- N., Diesmos, A., Iskandar, D., Melnick, D.J. & Cannatella, D.C. 2003. Phylogenetics of fanged frogs: testing biogeographical iant signature is evident late in Geokichla evolution as hypotheses at the interface of the Asian and Australian faunal well, but is limited to a few movements presumably zones. Syst. Biol. 52: 794–819. linked to low sea levels across the Sunda Shelf. Feakins, S.J., de Menocal, P.B. & Eglinton, T.I. 2005. Biomarker By comparison, dispersal has played a predominant records of late Neogene changes in northeast African vegeta- role in establishing distributions at the northern and tion. Geology 33: 977–980. eastern margin of the IOB, and through the islands of Filardi, C.E. & Moyle, R.G. 2005. Single origin of a pan-Pacific Wallacea. The possibility of derived Asian lineages having bird group and upstream colonization of Australasia. Nature stemmed from a single trans-IOB dispersal, or via 438: 216–219. stepping-stone movements across the IOB for Geokichla, Fuchs, J., Fjeldsa˚, J., Bowie, R.C.K., Voelker, G. & Pasquet, E. seems unlikely. Yet, it is clear from studies of a variety 2006. The African warbler genus Hyliota as a lost lineage in the Oscine songbird tree: evidence from mitochondrial and of lineages (Raxworthy et al., 2002; Morley, 2003; nuclear sequence data. Mol. Phylogenet. Evol. 39: 186–197. Warren , 2003, 2005; Sanmartıń & Ronquist, 2004; et al. Fuller, S., Schwarz, M. & Tierney, S. 2005. Phylogenetics of the Fuller et al., 2005; Yuan et al., 2005; Rocha et al., 2006; allodapine bee genus Braunsapis: historical biogeography and Treńel et al., 2007) that such dispersal events cannot be long-range dispersal over water. J. Biogeogr. 32: 2135–2144. discounted out of hand. Despite evidence of IOB dispersal

ª 2008 THE AUTHORS. J. EVOL. BIOL. 21 (2008) 1779–1788 JOURNAL COMPILATION ª 2008 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY Speciation around the Indian Ocean Basin 1787

Garcıá-Moreno, J. 2004. Is there a universal mtDNA clock for Sanderson, M. 2002. Estimating absolute rates of molecular birds? J. Avian Biol. 35: 465–468. evolution and divergence times: a penalized likelihood Gorog, A.J., Sinaga, M.H. & Engstrom, M.D. 2004. Vicariance or approach. Mol. Biol. Evol. 19: 101–109. dispersal? Historical biogeography of three Sunda shelf murine Sepulchre, P., Ramstein, G., Fluteau, F., Schuster, M., Tiercelin, rodents (Maxomys surifer, Leopoldamys sabanus and Maxomys J.-J. & Brunet, M. 2006. Tectonic uplift and eastern African whiteheadi). Biol. J. Linn. Soc. 81: 91–109. aridification. Science 313: 1419–1423. Hall, R. 1998. The plate tectonics of Cenozoic SE Asia and the Sibley, C.G. & Monroe, B.L. Jr 1990. Distribution and Taxonomy of distribution of land and sea. In: Biogeography and Geological Birds of the World. Yale University Press, New Haven. Evolution of SE Asia (R. Hall & J.D. Holloway, eds), pp. 99–124. Simpson, G.G. 1953. The Major Features of Evolution. University of Backhuys Publishers, Leiden. Colombia Press, New York. Hamilton, A.C. & Taylor, D. 1991. History of climate and forests Steadman, D.W. 2006. Extinction and Biogeography of Tropical in tropical Africa during the last 8 million years. Climatic Pacific Birds. University of Chicago Press, Chicago. Change 19: 65–78. Steenis, C.G.G.J. 1962. The land-bridge theory in botany. Blumea Hausdorf, B. 1998. Weighted ancestral area analysis and a 11: 235–372. solution of the redundant distribution problem. Syst. Biol. 47: Swofford, D.L. 2000. PAUP*. Phylogenetic Analysis Using Parsi- 445–456. mony (*and Other Methods). Version 4.0b2. Sinauer, Sunder- Ho, S.Y.W., Phillips, M.J., Cooper, A. & Drummond, A.J. land, MA. 2005. Time dependency of molecular rate estimates and Tedford, R.H. 1985. Late Miocene turn-over of the Australian systematic overestimation of recent divergence times. Mol. mammal fauna. South Afr. J. Sci. 81: 262–263. Biol. Evol. 22: 1561–1568. Treńel, P., Gustafsson, M.H.G., Baker, W.J., Asmussen-Lange, Huelsenbeck, J.P. & Ronquist, F. 2001. MRBAYES: Bayesian C.B., Dransfield, J. & Borchsenius, F. 2007. Mid-Tertiary inference of phylogenetic trees. Bioinformatics 17: 754–755. dispersal, not Gondwanan vicariance explains distribution Jacob, K. 1949. Land connections between Ceylon and penin- patterns in the wax palm subfamily (Ceroxyloideae: Areca- sular India. Proc. Natl Inst. Sci. India 15: 341–343. ceae). Mol. Phylogenet. Evol. 45: 272–288. Jobb, G. 2005. TREEFINDER, Version of June 2005. Distributed by Vences, M., Freyhof, J., Sonnenberg, R., Kosuch, J. & Veith, M. the author at http://www.treefinder.de. Munich, Germany. 2001. Reconciling fossils and molecules: cenozoic divergence Jønnson, K.A. & Fjeldsa˚, J. 2006. Determining biogeographical of cichlid fishes and the biogeography of Madagascar. patterns of dispersal and diversification in oscine passerine J. Biogeogr. 28: 1091–1099. birds in Australia, Southeast Asia and Africa. J. Biogeogr. 33: Voelker, G. 1999. Dispersal, vicariance and clocks: historical 1155–1165. biogeography and speciation in a cosmopolitan passerine Klicka, J., Voelker, G. & Spellman, G.M. 2005. A molecular genus (Anthus: Motacillidae). Evolution 53: 1536–1552. systematic revision of the true thrushes (Turdinae). Mol. Voelker, G. 2002. Systematics and historical biogeography of Phylogenet. Evol. 34: 486–500. wagtails (Aves: Motacilla): dispersal versus vicariance revisited. Mayr, E. & Paynter, R.A. Jr (eds) 1964. Check-list of Birds of the Condor 104: 725–739. World, Vol. 10. Museum of Comparative Zoology, Cambridge. Voelker, G. & Spellman, G.M. 2004. Nuclear and mitochondrial McLoughlin, S. 2001. The breakup history of Gondwana and its DNA evidence of polyphyly in the avian superfamily Musci- impact on pre-Cenozoic floristic provincialism. Aust. J. Bot. 49: capoidea. Mol. Phylogenet. Evol. 30: 386–394. 271–300. Voelker, G. & Klicka, J. 2008. Systematics of Zoothera thrushes, Morley, R. 2003. Interplate dispersal paths for megathermal and a synthesis of true thrush molecular systematic relation- angiosperms. Perspect. Plant Ecol. Evol. Syst. 6: 5–20. ships. Mol. Phylogenet. Evol., in press. Outlaw, D.C. & Voelker, G. 2008. Pliocene climatic change in Voelker, G., Rohwer, S., Bowie, R.C.K. & Outlaw, D.C. 2007. insular Southeast Asia as an engine of diversification in Molecular systematics of a speciose, cosmopolitan songbird Ficedula flycatchers. J. Biogeogr. 35: 739–752. genus: defining the limits of, and relationships among, the Outlaw, R.K., Voelker, G. & Outlaw, D.C. 2007. Molecular Turdus thrushes. Mol. Phylogenet. Evol. 42: 422–434. systematics and historical biogeography of the Rock-Thrushes Vrba, E.S. 1985. African Bovidae: evolutionary events since the (Muscicapidae: Monticola). Auk 124: 561–577. Miocene. South Afr. J. Sci. 81: 263–266. Posada, D. & Crandall, K.A. 1998. MODELTEST: testing the Vrba, E.S. 1993. Mammal evolution in the African Neogene model of DNA substitution. Bioinformatics 14: 817–818. and a new look at the Great American Interchange. Raxworthy, C.R., Forstner, M.R.J. & Nussbaum, R.A. 2002. In: Biological Relationships between Africa and South America Chameleon radiation by oceanic dispersal. Nature 415: 784– (P. Goldblatt, ed.), pp. 393–434. Yale University Press, New 787. Haven. Rocha, S., Carretero, M.A., Vences, M., Glaw, F. & Harris, D.J. Warren, B.H., Bermingham, E., Bowie, R.C.K., Prys-Jones, R.P. 2006. Deciphering patterns of transoceanic dispersal: the & Thebaud, C. 2003. Molecular phylogeography reveals island evolutionary origin and biogeography of coastal lizards (Cryp- colonization history and diversification of western Indian toblepharus) in the Western Indian Ocean region. J. Biogeogr. Ocean sunbirds (Nectarinia: Nectariniidae). Mol. Phylogenet. 33: 13–22. Evol. 29: 67–85. Ronquist, F. 1997. Dispersal-vicariance analysis: a new approach Warren, B.H., Bermingham, E., Prys-Jones, R.P. & Thebaud, C. to the quantification of historical biogeography. Syst. Biol. 46: 2005. Tracking island colonization history and phenotypic 195–203. shifts in Indian Ocean bulbuls (Hypsipetes: Pycnonotidae). Biol. Sanmartıń, I. & Ronquist, F. 2004. Southern Hemisphere J. Linn. Soc. 85: 271–287. biogeography inferred by event-based models: plant versus Wesselman, H.B. 1995. Of mice and almost-men: regional animal patterns. Syst. Biol. 53: 216–243. paleoecology and human evolution in the Turkana Basin.

ª 2008 THE AUTHORS. J. EVOL. BIOL. 21 (2008) 1779–1788 JOURNAL COMPILATION ª 2008 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY 1788 G. VOELKER AND R. K. OUTLAW

In: Paleoclimate and Evolution with Emphasis on Human Origins Table S1 Specimens used, sample sources and locality (E.S. Vrba, G.H. Denton, T.C. Partridge & L.H. Burkle, eds), pp. information for species in the sibirica clade. 356–368. Yale University Press, New Haven. Table S2 Results of weighted ancestral-area analysis. Yuan, Y.-M., Wohlhauser, S., Mo¨ ller, M., Klackenberg, J., Please note: Wiley-Blackwell are not responsible for Callmander, M.W. & Ku¨ pfer, P. 2005. Phylogeny and bioge- the content or functionality of any supporting materials ography of Exacum (Gentianaceae): a disjunctive distribution supplied by the authors. Any queries (other than missing in the Indian Ocean Basin resulted from long distance dispersal and extensive radiation. Syst. Biol. 54: 21–34. material) should be directed to the corresponding author for the article.

Supporting information Received 7 December 2007; revised 6 June 2008; accepted 13 June 2008 Additional supporting information may be found in the online version of this article:

ª 2008 THE AUTHORS. J. EVOL. BIOL. 21 (2008) 1779–1788 JOURNAL COMPILATION ª 2008 EUROPEAN SOCIETY FOR EVOLUTIONARY BIOLOGY