Journal of Biogeography (J. Biogeogr.) (2016)

ORIGINAL Colonizing the Caribbean: biogeography ARTICLE and evolution of livebearing fishes of the () Pablo F. Weaver1,2*, Alexander Cruz2, Steven Johnson1, Julia Dupin2 and Kathleen F. Weaver1

1Department of Biology, University of La ABSTRACT Verne, 1950 3rd St., La Verne, CA 91750, Aim We investigate the origin and colonization of the West Indian endemic USA, 2Department of Ecology and freshwater fish group Limia. We evaluate the leading hypotheses for the origins Evolutionary Biology, University of Colorado, Boulder, CO 80309-0334, USA of West Indian life, including trans-oceanic dispersal, late Cretaceous vicari- ance, and the GAARlandia land bridge at the Eocene/Oligocene boundary. Location Greater Antilles, with extensive sampling in the Dominican Republic. Methods We obtained DNA from wild sampling and the aquarium trade. We sequenced three mitochondrial (12S, ND2 and Cytb) and two nuclear genes (Rh, MYH6) for a combined molecular phylogenetic analysis to evaluate spe- cies relationships and the timing of divergence events between islands and the mainland. We used Bayesian and likelihood approaches to build phylogenies, a BEAST analysis to establish the timing of colonization, and R package BioGeo- BEARS to perform a historical biogeographical reconstruction.

Results Relaxed molecular clock results show that the ancestor to the West Indian clade, which includes the Limia and Hispaniolan , diverged from a South American ancestor at the Eocene/Oligocene boundary. The basal Jamaican species, L. melanogaster, split from the rest of Limia at the Oligocene/ Miocene boundary. Cuban and Cayman taxa are sister to a diverse species group from Hispaniola. Historical biogeographical reconstruction supported the GAARlandia DEC+j model as the best fitting model for colonization. Main conclusions Our results support a colonization model for Limia that is concordant with the timing of GAARlandia and climate change during the Eocene/Oligocene boundary. Limia colonization was most likely a result of facilitated dispersal during a period of lower sea levels and shorter passage along the Aves Ridge. These results are also consistent with other recent molec- ular clock studies of dispersal limited cichlids, toads and frogs, indicating a growing body of support for the significance of Eocene/Oligocene climate change for the historical biogeography of West Indian life. *Correspondence: Pablo Weaver, Biology Department, 1950 3rd St., La Verne, CA 91750, Keywords USA. colonization, GAARlandia, Hispaniola, historical biogeographical reconstruc- E-mails: [email protected]; [email protected] tion, Limia, relaxed molecular clocks, vicariance, West Indian biogeography

reaching over 90% in some groups (Hedges, 1996; Myers et al., INTRODUCTION 2000). However, the island ecosystems are relatively depauper- After nearly two centuries of debate, biogeographical recon- ate and some major mainland groups, including marsupials, struction in the West Indies continues to be a contentious carnivores, lagomorphs, salamanders and most families of issue, with wide disagreement over the relative importance of frogs, turtles, and snakes are missing (Hedges, 1996, 2001). dispersal–vicariance. The islands of the West Indies exhibit These patterns are typical of isolated oceanic islands, in which exceptional biodiversity and are ranked in the top five of the a few successful dispersers radiate in situ. Molecular clock world’s most important biodiversity hotspots, with endemism analyses across taxa, including some reptiles, amphibians, and

ª 2016 John Wiley & Sons Ltd http://wileyonlinelibrary.com/journal/jbi 1 doi:10.1111/jbi.12798 P. F. Weaver et al. mammals, show a pattern of random and recent divergences One of the groups for which colonization routes is still from mainland sources (mostly from South America), concor- unknown is Limia (family Poeciliidae), a genus of livebearing dant with a dispersal model of colonization (see Hedges, 1996, freshwater fishes endemic to the islands of the West Indies 2001 for figure showcasing random colonization patterns). (Rauchenberger, 1988; Burgess & Franz, 1989; Rodriguez, The primary mechanism of dispersal for these groups is 1997; Hamilton, 2001). With 17 described endemic species thought to be a combination of large flooding events in South on Hispaniola and one endemic species on each of the America, floating mats of debris, and the prevailing north-wes- islands of , and Grand Cayman, Limia is the terly oceanic currents (King, 1962; Vonhof et al., 1998; dominant freshwater fish group in the Greater Antilles (Bur- Hedges, 2001; Glor et al., 2005). Dispersal continues in the gess & Franz, 1989). However, the historical biogeography of present day, as evidenced by a colony of iguanas (Iguana Limia has not been critically assessed using molecular tech- iguana) arriving to the island of Anguilla from nearby Guada- niques and the timing and mechanism of colonization for lupe in the wake of Hurricane Luis in 1995 (Lawrence, 1998). Limia remains unresolved. Our hope is that by reconstruct- Alternatively, the vicariance argument highlights the pres- ing the colonization of Limia, we will gain a vital piece in ence of organisms with poor dispersal abilities and relies on the ever-evolving debate on the origins of West Indian life the correlation between evolutionary patterns and geological and gain insight into other organisms with limited long-dis- reconstructions. Advocates of West Indian vicariance bio- tance dispersal abilities. geography envision entire ecosystems, trapped on drifting Freshwater fish in the Limia genus provide an ideal oppor- island fragments, splitting from a proto-Central America in tunity to evaluate colonization hypotheses because of their the late Cretaceous (65 Ma) through plate tectonics (Rosen, diversity and presence on multiple islands. Using Limia spe- 1975, 1985; Nelson & Platnick, 1981; Guyer & Savage, 1986; cies from Hispaniola (Dominican Republic and Haiti), Page & Lydeard, 1994). The presence of freshwater fish on Jamaica, Cuba, as well as poeciliid species from the main- the islands of the West Indies has sparked much interest over land, evaluation of multiple divergence events is possible: the years because of their diversity and endemism, as well as West Indies from the mainland (all Limia), Jamaica from their presumed limited dispersal abilities (Rosen & Bailey, Cuba/Hispaniola (L. melanogaster split) and Cuba from His- 1963; Rosen, 1975; Rivas, 1986; Poeser, 2003), and they were paniola (L. vittata split). We test the above colonization sce- instrumental in the development of vicariance biogeography narios using a combined gene molecular phylogenetic as a discipline (Rosen, 1975). Previous taxonomic studies of approach with a robust sampling of poeciliids from the West Indian freshwater fish show that species on Hispaniola islands and the mainland, together with a relaxed molecular and Cuba are polyphyletic; northern Hispaniolan species are clock analysis and historical biogeographical reconstruction. more closely related to eastern Cuban and Cayman island The goal for this study was to test biogeographical hypothe- species than they are to southern Hispaniolan species ses of the West Indies through an updated multigene phy- (Rauchenberger, 1988; Burgess & Franz, 1989; Rodriguez, logeny of Limia. We examine three alternative scenarios, with 1997; Hamilton, 2001). These patterns seemed concordant regard to Limia colonization. First, long-distance transoceanic with proposed geological models of island formation and a dispersal models, which predict colonization across the Carib- late Cretaceous connection with the mainland (Pitman et al., bean sea by salt tolerant taxa from diverse mainland sources, 1993; Pindell, 1994). However, the late Cretaceous vicariance with subsequent evolution of the fresh-water clades (Myers, model has received much criticism, mainly because of a lack 1966; Briggs, 1984, 1987; Hedges, 1996, 2001, 2006). Second, of molecular clock support and the unlikely survival of life late Cretaceous vicariance models, which predict colonization in this region during and shortly after the late Cretaceous during the late Cretaceous (65–85 Ma) and persistence of bolide impact near the Yucatan peninsula (Hildebrand & biota on incipient island fragments (Rosen, 1975, 1985; Nelson Boynton, 1990). & Platnick, 1981; Guyer & Savage, 1986; Rauchenberger, 1988; An alternative geological reconstruction that shows a more Page & Lydeard, 1994). Finally, the GAARlandia model, which recent land connection between the mainland and the West postulates inundation of the incipient island fragments after Indies, is the GAARlandia model (Greater Antilles Aves the late Cretaceous and until the Middle Eocene (< 40 Ma), Ridge) (Iturralde-Vinent & MacPhee, 1999; Iturralde-Vinent, with range expansion of South American groups along a short- 2006). In this reconstruction, any late Cretaceous land for- lived, exposed Aves Ridge, which was created by sea level mations in the Caribbean would have been inundated by decline and Caribbean plate uplift during the Eocene/Oligo- subsequent sea level change, and permanent land was only cene boundary (33–35 Ma) (Iturralde-Vinent & MacPhee, present after the Middle Eocene (< 40 Ma). During the 1999; Iturralde-Vinent, 2006). Eocene/Oligocene transition (30–35 Ma), a rapidly cooling planet caused global declines in sea level upwards of 60 m METHODS (Haq et al., 1987; Miller et al., 2008). Coupled with general tectonic uplift in the region, this may have allowed for Sampling potential overland colonization of the West Indies from South America across the Aves Ridge (Iturralde-Vinent & For an analysis of Limia diversity within the West Indies, we MacPhee, 1999; Iturralde-Vinent, 2006). sampled poeciliids from across the region, including wild

2 Journal of Biogeography ª 2016 John Wiley & Sons Ltd Biogeography of Limia (Poeciliidae) collections and species we could obtain through the aquar- provide resolution at both the shallower and deeper ium trade. Wild collections were focused on the island of divergence nodes, we amplified both mitochondrial and Hispaniola to sample from the greatest diversity of poeciliids nuclear loci (Hrbek et al., 2007). Five gene fragments, three and to ensure that each of the major clades of Limia was mitochondrial (12S, ND2 and Cytb) and two nuclear (Rh, included. A total of 30 Dominican populations were sampled MYH6), were amplified and sequenced for the 68 ingroup using seine and dip nets during trips in 2000, 2003, 2004, individuals. Primer sequences for polymerase chain reaction 2010, 2012 and 2013 (Fig. 1, see Appendix S1 in Supporting (PCR) are listed in Appendix S1. Information). All appropriate permits for collection and Amplifications were performed in an Eppendorf Mastercy- transport were obtained through local wildlife agencies. In cler (Eppendorf, Westbury, NY, USA) under the thermal addition, we included the Haitian species (L. nigrofasciata), conditions listed in Appendix S1. PCR products were puri- the Jamaican species (L. melanogaster), the Cuban species fied using ExoSAP enzymes (USB Corp., Cleveland, OH, (L. vittata) and the Grand Cayman species (L. caymanensis), USA). Purified PCR products were cycle sequenced using Big which were obtained through aquarium stock at the Univer- Dye chemistry v.3.1 (Applied Biosystems Inc., Foster City, sity of Colorado (see Appendix S1). All specimens were pre- CA, USA) and visualized on an ABI 377XL (Life Technolo- served in 80% ethanol prior to DNA extraction, in gies, Grand Island, NY, USA). DNA sequences were edited accordance with IACUC approval at the University of La with Geneious Pro 5.6.6 and alignments were performed Verne. DNA sequences from additional outgroup taxa from with Clustal W (Thompson et al., 1997). the mainland genera of Pamphorichthys, , Poecil- ia and were obtained through Genbank (see Phylogenetic analyses Appendix S1). Phylogenetic analyses were performed using both Bayesian analyses (BA) and maximum likelihood (ML) methods. Sub- Molecular methods stantial discussion has revolved around the contrasting Genomic DNA was extracted from ethanol-preserved caudal approaches of multigene concatenated data sets versus con- peduncle muscle tissue following the manufacturer’s protocol sensus gene trees (see Gadagkar et al., 2005; Degnan & for the DNeasy Kit (Qiagen Inc., Valencia, CA, USA). To Rosenberg, 2009). To evaluate the utility of combining the

(a)

(b)

Figure 1 Locality map for populations sampled for this study. Limia is endemic to the West Indies, with one species occurring on each of the islands of Cuba, Grand Cayman and Jamaica; the majority of the species (possibly 17) are found on Hispaniola. We sampled heavily from the south-west and central Dominican Republic where biodiversity is high. Additional species were obtained from Cuba, Jamaica, and Grand Cayman thru aquarium stock.

Journal of Biogeography 3 ª 2016 John Wiley & Sons Ltd P. F. Weaver et al. nuclear and mitochondrial data sets, we ran preliminary substitution model as the seven partition scheme we used for analyses on the mitochondrial and nuclear data sets indepen- the MrBayes analysis, except that base frequencies were esti- dently using Mr. Bayes 3.1.2 (Ronquist & Huelsenbeck, mated for all partitions. We used the uncorrelated lognormal 2003). Doubts about the utility of incongruence length dif- distribution model on the partitioned data set to account for ference tests (Barker & Lutzoni, 2002; Darlu & Lecointre, lineage specific rate heterogeneity. We used the Yule speciation 2002; Alonso et al., 2012) lead us to visually inspect the process as the tree prior. Because no fossil records were avail- mitochondrial and nuclear trees for congruence. The mito- able for our taxa, we calibrated the tree based on the opening chondrial and nuclear trees were overall similar in their of the Windward Passage, which split Cuba from Hispaniola topologies, with the mitochondrial tree showing better reso- (L. vittata from the Hispaniolan Limia clade) 14–17 Ma (Itur- lution near the tips and the nuclear tree showing better reso- ralde-Vinent, 2006). Additional analyses were run using an lution at deeper nodes (see Appendix S2). alternative uniform prior distribution of 20–25 Ma for the We used PartitionFinder (Lanfear et al., 2012) to evalu- Windward Passage (Pitman et al., 1993) to evaluate the influ- ate the appropriate nucleotide substitution models and parti- ence of this parameter (see Appendix S2). The estimation of tion scheme on our combined data set. PartitionFinder 14–17 Ma is supported by the geological reconstruction of allowed us to evaluate various partition schemes, including Iturralde-Vinent (2006) as well as independent molecular the fragment of 12S rDNA and each codon position indepen- clock estimates for toads (Alonso et al., 2012) and cichlids dently in all of the protein coding genes (ND2, Cytb, Rh and (Chakrabarty, 2006; Perez et al., 2007; Rıcan et al., 2012). This MYH6). The optimal partition scheme for the data, as out- calibration point has also been used successfully in examina- lined by PartitionFinder, was a 7-way partition strategy tion of the Cuban Girardinus group (Doadrio et al., 2009), as with the following models: subset 1 = K80+I for 12S and well as the Central American Poecilia sphenops (Alda et al., MYH6 pos 2, subset 2 = GTR+I+G for Cytb pos 1, ND2 2013). The analyses were performed with two independent pos 1, subset 3 = HKY+G for Cytb pos 2 and ND2 pos 2, runs for 30 million generations, with trees selected every subset 4 = HKY + I for Cytb pos 3, ND2 pos 3 and Rh 10,000 estimations. We checked for convergence diagnostics pos 1, subset 5 = JC+I for Rh pos 2, subset 6 = F81 for Rh with Tracer 1.5 (Rambaut & Drummond, 2007) and com- pos 3, subset 7 = HKY+I for MYH6 pos 1 and MYH6 pos 3. bined runs using Log-Combiner 1.7 (part of the beast pack- Bayesian analyses were then performed on the combined age). The data were summarized using TreeAnnotator 1.7.4. data set, using the seven partition scheme provided by Par- (part of the beast package) and visualized using FigTree titionFinder. The Markov chain Monte Carlo (MCMC) 1.4.2. (Rambaut, 2009). simulation was performed utilizing two independent runs, using eight chains and 10 million generations, sampling Biogeographical analyses every 1000 generations. Convergence and stability of the runs was detected using Tracer 1.5. (Rambaut & Drummond, To investigate the historical biogeography of Limia, ancestral 2007). The first 25% of each run was established as burn-in ranges were estimated using the R package BioGeoBEARS and discarded. The remaining trees were used to compute a (Matzke, 2013); this allowed direct tests of the fit of dispersal 50% majority rule consensus tree. Maximum likelihood trees and GAARlandia models. For all of the hypotheses listed, we were built using PhyML (Guindon & Gascuel, 2003) as applied the Dispersal-Extinction-Cladogenesis (DEC; Ree & implemented in Geneious Pro 5.6.6 and paup 4.0 (Swofford, Smith, 2008) model and its modified version, DEC+j, which 2003). Support for nodes for both methods was established allows for founder event speciation. Additionally, analyses with 1000 bootstrap replicates. For the PhyML analysis, we included a time-stratified approach that represents different employed a general GTR+I+G substitution model, optimizing dispersal rates across time slices (see Appendix S3). Finally, topology, length and rate, and used the BEST topology the relative likelihood of all resulting models were compared search strategy. In the Paup analysis, we used the seven par- using Akaike’s information criterion (AIC; Burnham & tition scheme as outlined by PartitionFinder. Tree search- Anderson, 2002). The best model was then used to estimate ing was performed with a branch and bound strategy, using the biogeographical history of Limia. the FURTHEST addition method and bootstrapping was per- On the basis of the dated chronogram for Limia, we formed using a random seed and branch and bound search pruned the tree to remove outgroups, but kept the sister type. Trees were rooted with the Central American livebear- group comprised of the Hispaniolan Poecilia species, as well ing species Xiphophorus hellerii after Hrbek et al. (2007). as the closest mainland relative Pamphorichthys. We defined five geographical areas occupied by Limia and its closest rela- tives, including South America (SA), Hispaniola (HI), Divergence time estimation Jamaica (JA), Cuba (CU) and Cayman Islands (CA) We chose single representatives from each of the main lineages (Table 1). We evaluated the likelihood of a dispersal model (34 in total) to maximize coverage of available gene sequences versus a GAARlandia model of colonization, using time slices and calculated divergence time estimates using a relaxed-clock that correspond with the proposed colonization pathway method implemented in beast 1.7.4 (Drummond & Rambaut, applied to each model (see Appendix S3). We did not test 2007). The data set was partitioned with the same nucleotide the late Cretaceous vicariance hypothesis, as the timing was

4 Journal of Biogeography ª 2016 John Wiley & Sons Ltd Biogeography of Limia (Poeciliidae)

Table 1 Current species distributions used to reconstruct the biogeographical history of Limia in the West Indies. The species Divergence time estimation included in this analysis were: Pamphorichthys hollandi, from the The West Indies clade of Limia and Poecilia is shown to split South America Biogeographic region, Limia dominicensis, Limia from the South American Pamphorichthys group during the perugiae, Limia nigrofasciata, Limia zonata, Limia versicolor, Poecilia elegans, Poecilia dominicensis and Poecilia hispaniolana Eocene/Oligocene transition approximately 32.9 Ma (95% from Hispaniola, Limia melanogaster from Jamaica, Limia vittata highest posterior densities (HPD) = 21.6–48.8) (Fig. 3, node from Cuba and Limia caymanensis from Grand Cayman. A). The Jamaican species L. melanogaster split from other Limia species near the Oligocene/Miocene transition 22.8 Ma Biogeographical (95% HPD = 17.2–32.0) (Fig. 3, node B). The tree was cali- region Species occurrences brated by the splitting of Cuban L. vittata and Grand Cay- South America (SA) Pamphorichthys hollandi man L. caymanensis from the Hispaniolan Limia clade Hispaniola (HI) L. dominicensis, L. perugiae, L. nigrofasciata, (Fig. 3, node C), which corresponds with the opening of the L. zonata, L. versicolor, Poecilia elegans, Windward Passage and separation of the proto-island land- Poecilia dominicensis, Poecilia hispaniolana masses of eastern Cuba and northern Hispaniola 14–17 Ma. Jamaica (JA) L. melanogaster The Grand Cayman species, L. caymanensis split from its sis- Cuba (CU) L. vittata Grand Cayman (CA) L. caymanensis ter taxon, the Cuban L. vittata at 3.64 Ma (95% HPD = 1.28–7.85) (Fig. 3, node D). The large clade of His- paniolan lineages, which include several putative L. perugiae populations (see Appendix S2 for further discussion), as well not supported by the divergence time analysis. For the dis- as a sympatric population of L. sulphurophila from La Zurza, persal model, our probability matrices reflect no land con- and several unidentified groups, is shown to be a recent nection between the Greater Antilles and South America and diversification event at 1.76 Ma (95% HPD = 0.69–3.10) the emergence of Jamaica during the mid-Miocene (after (Fig. 3, node E). The correlation of our molecular clock Robinson, 1994). For the GAARlandia model, time slices analyses with the GAARlandia model is shown in Fig. 4. The reflect the reconstructions of Iturralde-Vinent & MacPhee timing of the West Indies clade divergence (A, Fig. 4), the (1999) and Iturralde-Vinent (2006) and include facilitated isolation of Blue Mountains terrane of Jamaica from south- dispersal via a land connection between South America and ern Hispaniola (B, Fig. 4) and the opening of the Windward the landmass that included the proto-island fragments of Passage separating eastern Cuba from Hispaniola (C, Fig. 4) Cuba, Hispaniola and the Jamaican Blue Mountain terranes correlate with estimates from the Beast analysis. from 35 to 30 Ma, followed by flooding and separation of the Greater Antilles from the mainland from 30 Ma to the Biogeographical analyses present. Both models assume the opening of the Windward Passage separating Eastern Cuba from Hispaniola at 15 Ma The AIC model comparison (Table 2) supported the (Iturralde-Vinent, 2006), as well as the formation of the Cay- GAARlandia DEC+j model as the best fitting model. This man Islands during late Miocene uplift, with a short-lived model was 4.8 AIC units lower than the second best model, connection to Cuba from 5 to 3 Ma (Jones, 1994). no-GAARlandia DEC. This result shows that a model that includes a land connection between South America and the proto-island fragments during 35 to 30 Ma is better fitted for RESULTS the clade comprised of Limia, Poecilia and Pamphorichthys than a model where the colonization of the Caribbean islands Phylogenetic analysis happened only through dispersal. Moreover, the model with Our combined and concatenated mitochondrial (12S, ND2, the founder event parameter (j) shows a significant lower like- Cytb) and nuclear (Rh, MYH6) data sets yielded 2661 charac- lihood score than its simpler version, which suggests that range ters (1133 mitochondrial, 1528 nuclear). The resulting gene expansions alone are not sufficient to explain movements to sequences are available in GenBank, accession #s KX023907- new areas. For example, the results show that one jump disper- KX024245. The 7-way partitioned data set, which included sal (or, founder event) is estimated to explain the current both nuclear and mitochondrial data, provided good resolu- range of L. caymanensis (Fig. 5). tion at both the shallower and deeper nodes of our ingroup taxa with Bayesian posterior probabilities (PP) of 0.90–1.0 and DISCUSSION maximum likelihood bootstrap (MLB) support of 80–100% (Fig. 2, Appendix S2). Overall tree topologies of the Bayesian, Biogeography of Limia as well as ML reconstructions from both PhyML and PAUP were similar at all nodes. Support for nodes was very similar West Indian fauna likely colonized the islands during many across ML methods. A discussion of Limia evolution and spe- events and through multiple mechanisms. Here, we show cies relationships within the genus are included in evidence that an ancestor to the West Indies clade of live- Appendix S2. bearing freshwater fishes (including all Limia and the

Journal of Biogeography 5 ª 2016 John Wiley & Sons Ltd P. F. Weaver et al.

Figure 2 Phylogenetic reconstruction based on a Bayesian analysis of a combined data set for 88 individuals and five genes (three mitochondrial: 12S, ND2, Cytb and two nuclear: MYH6, Rh) with seven data partitions. Numbers above nodes represent Bayesian Posterior Probabilities and numbers below nodes represent maximum likelihood bootstrap estimates from ML analysis. A version of this phylogeny with labelled terminal nodes is available in Appendix S2.

6 Journal of Biogeography ª 2016 John Wiley & Sons Ltd Biogeography of Limia (Poeciliidae)

Figure 3 Chronogram of Limia and related groups derived from a relaxed molecular clock analysis of a combined data set (three mitochondrial: 12S, ND2, Cytb, and two nuclear: MYH6, Rh). Bars indicate the 95% highest posterior densities (HPD). Branch lengths represent time since divergence and major nodes are indicated by lettered circles. Nodes A, B, and C represent significant splitting events during the colonization of the West Indian livebearers and correspond with the geological timeline as illustrated in Fig. 4.

Hispaniolan Poecilia) diverged from mainland taxa at the caps and substantial sea level drops on the order of 60 m Eocene/Oligocene boundary (32.9 Ma) (Fig. 3). The timing (Haq et al., 1987; Miller et al., 2008). Supplemented by tec- of the split between the West Indian clade and the mainland tonic uplift of the Caribbean plate, these conditions may precludes a late Cretaceous vicariant origin or recent transo- have resulted in either shallow banks or a land bridge con- ceanic dispersal. In addition, our historical biogeographical nection between proto-Antillean islands and the South reconstruction showed the most likely model for colonization American mainland (Iturralde-Vinent & MacPhee, 1999; is a GAARlandia model, including the colonization of Jamai- Iturralde-Vinent, 2006). During this time frame, much of the can and species. Greater Antilles existed as a single landmass, with connec- The GAARlandia model predicts a short-lived (~3 Myr) tions between eastern Cuba, north and south Hispaniola, the subaerial land connection between South America and the Blue Mountains of Jamaica, Puerto Rico, and the Aves Ridge incipient landmasses of the West Indies across the Aves south to mainland South America. The paleogeographical Ridge during the Eocene/Oligocene transition (30–35 Ma) model of Iturralde-Vinent & MacPhee (1999) and Iturralde- (Iturralde-Vinent & MacPhee, 1999; Iturralde-Vinent, 2006). Vinent (2006) also predicts that after the lowest sea levels of This time frame marks an important transitional period in the Oligocene, the Aves Ridge land span was subsequently the evolution of life on Earth, characterized by global climate inundated by rising sea levels before the Miocene (23.5 Ma), cooling and related changes in sea level. Transitioning from severing the mainland connection and isolating several island the warm climate of the Eocene, the Earth experienced rapid fragments, including the Blue Mountains of Jamaica (Fig. 4). cooling of over 5 °C from the Eocene optimum to the start Further movement of Greater Antillean island fragments of the Oligocene, accompanied by the formation of polar ice acted to separate eastern Cuba from northern Hispaniola,

Journal of Biogeography 7 ª 2016 John Wiley & Sons Ltd P. F. Weaver et al.

(a) also demonstrate support for the separation of Jamaica near the early Miocene (22.8 Ma). Divergence of the Jamaican L. melanogaster corresponds well with the proposed timing of rising sea levels and the separation of the Blue Mountain block of Jamaica from southern Hispaniola (Iturralde-Vinent & MacPhee, 1999; Fig 4). This result is interesting in itself, as many palaeogeographical models debate the independence of the eastern Blue Mountain block from a large western block (part of the Nicaraguan rise) (see Lewis et al., 1990; Robinson, 1994). In addition, orthodox views of Jamaican geology find little evidence for persistence of any aerial Jamaican terranes prior to the mid-Miocene (Robinson, 1965, 1994). Other studies evaluating the relationships (b) between Jamaican and Hispaniolan taxa may help to solidify this pattern and better explain the unique geological history of the Jamaican terranes.

Implications and future directions While our divergence timing results are concordant with the predictions of the GAARlandia model, we do not believe there is sufficient evidence to rule out alternative methods of colonization, such as a combination of dispersal and vicari- ant events, or short distance, stepping stone dispersal. As dis- cussed by Ali (2012), there is little geological evidence to (c) support the notion of a contiguous land bridge from South America into the West Indies during the Eocene/Oligocene transition. Furthermore, rather than crossing a land bridge, any of the taxa used for support of the GAARlandia model could have also used short distance dispersal to colonize the West Indies. In the case of West Indian freshwater fish fauna, the ability of poeciliids and cichlids to tolerate, and even thrive, in saltwater has been widely demonstrated both on the islands and in mainland taxa (Myers, 1938; Burgess & Franz, 1989). Other taxa that may have used the GAARlandia land span for colonization, including megalonychid sloths and other mammals, were also physically capable of crossing Figure 4 Inferred colonization by the West Indies (WI) clade over a saltwater barrier, especially with the potentially shorter (represented by Limia and the Hispaniolan Poecilia). Most distances between landmasses that would have occurred dur- recent common ancestor (MRCA) estimates correspond with key ing the Eocene/Oligocene transition. Coupled with the nodes from the molecular clock analysis in Fig. 3. Divergence absence of less tolerant groups on the islands, including pri- events are in concordance with a three-part model of Caribbean mary division freshwater fish and caecilians (Hedges, 1996, palaeogeography presented by Iturralde-Vinent & MacPhee (1999) and Iturralde-Vinent (2006). (a) The WI clade diverged 2001), it is difficult to rule out ancient dispersal, as argued from the South American Pamphorichthys group and reached by Heinicke et al. (2007) and their divergence time estimate the islands via the GAARlandia land span. (b) Rising sea levels for West Indian Eleutherodactylus lineage at 29–47 Ma. The separated southern Hispaniola from the Blue Mountains terrane alternative, in which fish were confined to the freshwater of Jamaica, isolating the Jamaican species L. melanogaster. (c) connections of a GAARlandia land span seems less likely and Opening of the Windward Passage separated eastern Cuba from would not have restricted the colonization of primary divi- the Hispaniolan species. sion freshwater fishes. Whether GAARlandia was a true land bridge or not, the opening the Windward Passage in the mid-Miocene predictions of the GAARlandia model as a corridor for dis- (14–17 Ma). persal should not be discounted. The main historical critique The results of our analyses follow rather closely several of the GAARlandia model was the absence of a clustering of events predicted by the GAARlandia model. Our best model divergence time estimates around 30 to 35 Ma. Until very of colonization supports the split of the West Indian/South recently, DNA sequencing and molecular clock techniques American clade during the Eocene/Oligocene (32.9 Ma). We did not allow for fine scale resolution of divergence time

8 Journal of Biogeography ª 2016 John Wiley & Sons Ltd Biogeography of Limia (Poeciliidae)

Table 2 Biogeographical reconstruction inferred in BIOGEOBEARS and the relative probabilities of each model calculated from the AIC weights. Models include dispersal-extinction-cladogenesis (DEC) for both GAARlandia reconstruction and the overwater dispersal models, as well as complementary models that allow for founder event speciation (+J).

Model LnL No. of parameters dejAIC D AIC AIC_weight Relative LL

GAARlandia DEC 15.1 2 0.0097 1E-12 0 34.297 6.574 0.032 0.037 À GAARlandia DEC+J 10.9 3 1E-12 1E-12 0.5534 27.723 0 0.862 1 À No-GAARlandia DEC 15.3 2 1E-12 1E-12 0 34.605 6.881 0.028 0.032 À No-GAARlandia DEC+J 13.3 3 2.01E-10 2.1E-11 0.0751 32.514 4.791 0.079 0.091 À LnL, log-likelihood; d, rate of dispersal; e, rate of extinction; j, relative probability of founder event speciation at cladogenesis; AIC, Akaike’s information criterion; D AIC, AIC-min(AIC); AIC weight, Normalized relative model likelihood; Relative LL, Model relative likelihood.

Figure 5 Maximum likelihood, ancestral range estimations of historical biogeography of Limia under the best model, GAARlandia DEC+j. Geographical codes: SA: South America, HI: Hispaniola. JA: Jamaica, CU: Cuba, CA: Grand Cayman. estimates. A poor fossil record and lack of reliable calibration (Alonso et al., 2012), some frogs, for example, Syrrhophus points lead to further inconsistency. As such, those estimates (Crawford & Smith, 2005) and Osteopilus (Moen & Wiens, that were available were often crude approximations relying 2009), Polistinae wasps (Silva & Noll, 2015), and dispersal on percentage sequence divergence and seemed to show a limited spiders (Crews & Gillespie, 2010) and butterflies scattering of divergence events through time (Hedges, 2001). (Wahlberg, 2006). Additional earlier work using fossil and However, many of these interpretations have changed with phylogenetic evidence also support GAARlandia colonization, recent phylogenetic work utilizing a combined mitochondrial including extinct primates (Horovitz & MacPhee, 1999; and nuclear DNA approach, in combination with broader Davalos, 2004) and megalonychid sloths (Macphee & Itur- sampling, more accurate calibration and relaxed molecular ralde-Vinent, 2000; White & MacPhee, 2001; Davalos, 2004), clock methods (see Hedges, 1996 and the updated study, several bat groups (Davalos, 2004), as well as hystricognath Heinicke et al., 2007). rodents (Woods, 2001; MacPhee et al., 2003; Davalos, 2004; Many recent re-evaluations of West Indian taxa give esti- but see Fabre et al. (2014) advocating mid-Miocene disper- mates for colonization that are now consistent with Eocene/ sal). Even without concrete evidence for a subaerial land Oligocene colonization, including cichlids (Hulsey et al., span, it seems clear that the time frame of the Eocene/Oligo- 2011; Rıcan et al., 2012), bufonids of the Peltophryne genus cene transition and its associated climate and

Journal of Biogeography 9 ª 2016 John Wiley & Sons Ltd P. F. Weaver et al. palaeogeography have played a major part in West Indian (Bufonidae: Peltophryne) based on mitochondrial and colonization and in shaping the remarkable present-day bio- nuclear genes. Journal of Biogeography, 39, 434–451. diversity. Barker, F.K. & Lutzoni, F. (2002) The utility of the incongru- ence length difference test. Systematic Biology, 51, 625–637. Briggs, J.C. (1984) Freshwater fishes and biogeography of A note on the of Limia Central America and the Antilles. Systematic Zoology, 33, As expected, we find support for the monophyly of Limia as a 428–435. group. Interestingly, however, we also show the closest sister Briggs, J.C. (1987) The Caribbean connection. Biogeography group of Limia to be the clade represented by the three His- and plate tectonics, pp. 33–44. Elsevier, London. paniolan Poecilia species. Earlier studies have demonstrated Burgess, G.H. & Franz, R. (1989) Zoogeography of the Antil- the sister group relationship between Limia and the Pam- lean freshwater fish fauna. Biogeography of the West Indies: phorichthys group (Hamilton, 2001; Hrbek et al., 2007; Mered- past, present, and future (ed. by C.A. Woods), pp. 263– ith et al., 2010, 2011) but did not include the Hispaniolan 304. Sand Hill Crane Press, Gainesville, Florida. Poecilia species (P. elegans, P. dominicensis and P. hispan- Burnham, K.P. & Anderson, D. R. (2002) Model selection and iolana) in their analyses. One consequence of this new phylo- multimodel inference: a practical information-theoretic genetic arrangement is to further add to the confusion in the approach. Springer Science & Business Media, New York. taxonomic designations of Poecilia. Several studies have advo- Chakrabarty, P. (2006) Systematics and historical biogeogra- cated generic rankings for the taxa Poecilia, Limia, Mollienesia, phy of Greater Antillean Cichlidae. Molecular Phylogenetics Pamphorichthys, Acanthophacelus and Micropoecilia (Rodri- and Evolution, 39, 619–627. guez, 1997; Hamilton, 2001; Poeser, 2003), while others use Crawford, A.J. & Smith, E.N. (2005) Cenozoic biogeography them as subgenera within the genus Poecilia (Meredith et al., and evolution in direct-developing frogs of Central Amer- 2010, 2011). In either case, the designation of the Hispaniolan ica (Leptodactylidae: Eleutherodactylus) as inferred from a Poecilia (P. elegans, P. dominicensis and P. hispaniolana) and phylogenetic analysis of nuclear and mitochondrial genes. South American Poecilia (P. vivipara) as a group does not Molecular Phylogenetics and Evolution, 35, 536–555. accurately reflect evolutionary history and needs revision. Crews, S.C. & Gillespie, R.G. (2010) Molecular systematics of Selenops spiders (Araneae: Selenopidae) from North and ACKNOWLEDGEMENTS Central America: implications for Caribbean biogeography. Biological Journal of the Linnean Society, 101, 288–322. We give special thanks to colleagues at both the University Darlu, P. & Lecointre, G. (2002) When does the incongru- of La Verne and the University of Colorado for their input ence length difference test fail? Molecular Biology and Evo- and guidance. In particular, we thank Jerome Garcia, lution, 19, 432–437. Vanessa Morales, Roshan Gamage, Andrew Martin, David Davalos, L.M. (2004) Phylogeny and biogeography of Carib- Stock, Robert Guralnick and Dena Smith for the helpful sug- bean mammals. Biological Journal of the Linnean Society, gestions and support. We also thank our Dominican col- 81, 373–394. leagues, Carlos Rodriguez, Arlen Marmolejo, Miguel Degnan, J.H. & Rosenberg, N.A. (2009) Gene tree discor- Landestoy and Marcos Rodriguez for assistance in the field dance, phylogenetic inference and the multispecies coales- and in the lab, as well as the Ministerio de Medio Ambiente cent. Trends in Ecology & Evolution, 24, 332–340. y Recursos Naturales for permits. Additional thanks to the Doadrio, I., Perea, S., Alcaraz, L. & Hernandez, N. (2009) referees of this manuscript for their constructive insight and Molecular phylogeny and biogeography of the Cuban suggestions. Funding for this project was made possible by genus Girardinus (Poey, 1854) and relationships within the the CU EBIO departmental graduate student grant, the CU tribe Girardinini (, Poeciliidae). Molecular Museum Research grant and the ULV Faculty Research Phylogenetics and Evolution, 50, 16–30. grant. Drummond, A.J. & Rambaut, A. (2007) Beast: Bayesian evolutionary analysis by sampling trees. BMC Evolutionary Biology, 7, 214. REFERENCES Fabre, P.H., Vilstrup, J.T., Raghavan, M., Der Sarkissian, C., Alda, F., Reina, R.G., Doadrio, I. & Bermingham, E. (2013) Willerslev, E., Douzery, E.J. & Orlando, L. (2014) Rodents Phylogeny and biogeography of the Poecilia sphenops species of the Caribbean: origin and diversification of hutias complex (Actinopterygii, Poeciliidae) in Central America. unraveled by next-generation museomics. Biology letters, Molecular Phylogenetics and Evolution, 66, 1011–1026. 10, 20140266. Ali, J.R. (2012) Colonizing the Caribbean: is the GAARlandia Gadagkar, S.R., Rosenberg, M.S. & Kumar, S. (2005) Infer- land-bridge hypothesis gaining a foothold? Journal of Bio- ring species phylogenies from multiple genes: concatenated geography, 39, 431–433. sequence tree versus consensus gene tree. Journal of Experi- Alonso, R., Crawford, A.J. & Bermingham, E. (2012) Molec- mental Zoology Part B: Molecular and Developmental Evolu- ular phylogeny of an endemic radiation of Cuban toads tion, 304, 64–74.

10 Journal of Biogeography ª 2016 John Wiley & Sons Ltd Biogeography of Limia (Poeciliidae)

Glor, R.E., Losos, J.B. & Larson, A. (2005) Out of Cuba: over- Lanfear, R., Calcott, B., Ho, S.Y. & Guindon, S. (2012) Par- water dispersal and speciation among lizards in the Anolis titionFinder: combined selection of partitioning schemes carolinensis subgroup. Molecular Ecology, 14, 2419–2432. and substitution models for phylogenetic analyses. Molecu- Guindon, S. & Gascuel, O. (2003) A simple, fast, and accu- lar Biology and Evolution, 29, 1695–1701. rate algorithm to estimate large phylogenies by maximum Lawrence, E. (1998) Iguanas Ride the Waves. Nature News, likelihood. Systematic Biology, 52, 696–704. October 15. http://www.nature.com/news/1998/981015/full/ Guyer, C. & Savage, J.M. (1986) Cladistic relationships among news981015-3.html anoles (Sauria: Iguanidae). Systematic Biology, 35, 509–531. Lewis, J.F., Draper, G., Bourdon, C., Bowin, C., Mattson, Hamilton, A. (2001) Phylogeny of Limia (Teleostei: Poecili- P.O., Maurrasse, F., Nagle, F. & Pardo, G. (1990) Geology idae) based on NADH Dehydrogenase Subunit 2 Sequences. and tectonic evolution of the northern Caribbean margin. Molecular Phylogenetics and Evolution, 19, 277–289. The geology of North America, volume H, the Caribbean Haq, B.U., Hardenbol, J. & Vail, P.R. (1987) Chronology of region (ed. by G. Dengo and J.E. Case), pp. 77–140. Geo- fluctuating sea levels since the Triassic. Science, 235, 1156– logical Society of America, Boulder, Colorado. 1167. Macphee, R.D. & Iturralde-Vinent, M.A. (2000) A short history of Hedges, S.B. (1996) Historical biogeography of West Indian Greater Antillean land mammals:biogeography,paleogeogra- vertebrates. Annual Review of Ecology and Systematics, 27, phy, radiations, and extinctions. Tropics, 10,145–154. 163–196. MacPhee, R.D., Iturralde-Vinent, M.A. & Gaffney, E.S. Hedges, S.B. (2001) Biogeography of the West Indies: an (2003) Domo de Zaza, an Early Miocene Vertebrate Local- overview. Biogeography of the West Indies: patterns and per- ity in South-Central Cuba, with Notes on the Tectonic spectives (ed. by C.A. Woods and F.E. Sergile), pp. 15–33. Evolution of Puerto Rico and the Mona Passage 1. Ameri- CRC Press, Boca Raton, Florida. can Museum Novitates, 3394,1–42. Hedges, S.B. (2006) Paleogeography of the Antilles and ori- Matzke, N. J. (2013) BioGeoBears: BioGeography with Bayes- gin of West Indian terrestrial vertebrates. Annals of the ian (and likelihood) Evolutionary Analysis in R Scripts. Missouri Botanical Garden, 93, 231–244. Available at: http://cran.r-project.org/web/packages/BioGeo Heinicke, M.P., Duellman, W.E. & Hedges, S.B. (2007) BEARS/ (accessed 17 January 2016). Major Caribbean and Central American frog faunas origi- Meredith, R.W., Pires, M.N., Reznick, D.N. & Springer, M.S. nated by ancient oceanic dispersal. Proceedings of the (2010) Molecular phylogenetic relationships and the evolu- National Academy of Sciences USA, 104, 10092–10097. tion of the placenta in Poecilia (Micropoecilia) (Poeciliidae: Hildebrand, A.R. & Boynton, W.V. (1990) Proximal Creta- ). Molecular Phylogenetics and Evolu- ceous-Tertiary boundary impact deposits in the Caribbean. tion, 55, 631–639. Science, 248, 843–847. Meredith, R.W., Pires, M.N., Reznick, D.N. & Springer, M.S. Horovitz, I. & MacPhee, R.D.E. (1999) The quaternary Cuban (2011) Molecular phylogenetic relationships and the platyrrhinae Paralouatta varonai and the origin of Antillean coevolution of placentotrophy and superfetation in Poecilia monkeys. Journal of Human Evolution, 36, 33–68. (Poeciliidae: Cyprinodontiformes). Molecular Phylogenetics Hrbek, T., Seckinger, J. & Meyer, A. (2007) A phylogenetic and Evolution, 59, 148–157. and biogeographic perspective on the evolution of Poe- Miller, K.G., Browning, J.V., Aubry, M.P., Wade, B.S., Katz, ciliid fishes. Molecular Phylogenetics and Evolution, 43, M.E., Kulpecz, A.A. & Wright, J.D. (2008) Eocene-Oligocene 986–998. global climate and sea-level changes: St. Stephens Quarry, Hulsey, C.D., Keck, B.P. & Hollingsworth, P.R., Jr. (2011) Alabama. Geological Society of America Bulletin, 120, 34–53. Species tree estimation and the historical biogeography of Moen, D.S. & Wiens, J.J. (2009) Phylogenetic evidence for heroine cichlids. Molecular Phylogenetics and Evolution, 58, competitively driven divergence: body-size evolution in Car- 124–131. ibbean tree frogs (Hylidae: Osteopilus). Evolution, 63, 195– Iturralde-Vinent, M.A. (2006) Meso-Cenozoic Caribbean pale- 214. ogeography: implications for the historical biogeography of Myers, G.S. (1938) Fresh-water fishes and West Indian zoogeo- the region. International Geology Review, 48, 791–827. graphy. Smithsonian Report for 1937, 339–364. Iturralde-Vinent, M. & MacPhee, R.D.E. (1999) Paleogeogra- Myers, G.S. (1966) Derivation of the freshwater fish fauna of phy of the Caribbean region: implications for Cenozoic Central America. Copeia, 1966, 766–773. biogeography. Bulletin of the American Museum of Natural Myers, N., Mittermeier, R.A., Mittermeier, C.G., Da Fonseca, History, 238,1–95. G.A. & Kent, J. (2000) Biodiversity hotspots for conserva- Jones, B. (1994) Geology of the Cayman Islands. The Cay- tion priorities. Nature, 403, 853–858. man Islands: natural history and biogeography (ed. by M.A. Nelson, G. & Platnick, N.I. (1981) Systematics and biogeography: Brunt and J.E. Davies), pp. 13–49. Kluwer Academic Pub- cladistics and vicariance.ColumbiaUniversityPress,NewYork. lishers, The Netherlands. Page, R.D. & Lydeard, C. (1994) Towards a cladistic biogeog- King, W. (1962) The occurrence of rafts for dispersal of land raphy of the Caribbean. Cladistics, 10, 21–41. into the West Indies. Quarterly Journal of the Flor- Perez, G.A.C., Rıcan, O., Ortı, G., Bermingham, E., Doadrio, ida Academy of Sciences, 25, 46–52. I. & Zardoya, R. (2007) Phylogeny and biogeography of 91

Journal of Biogeography 11 ª 2016 John Wiley & Sons Ltd P. F. Weaver et al.

species of heroine cichlids (Teleostei: Cichlidae) based on Thompson, J.D., Gibson, T.J., Plewniak, F., Jeanmougin, F. & sequences of the cytochrome b gene. Molecular Phylogenet- Higgins, D.G. (1997) The Clustal_X windows interface: ics and Evolution, 43, 91–110. flexible strategies for multiple sequence alignment aided by Pindell, J.L. (1994) Evolution of the Gulf of Mexico and the quality analyses tools. Nucleic Acid Research, 25, 4876–4882. Caribbean. Caribbean geology: an introduction (ed. by S.K. Vonhof, H.B., Wesselingh, F.P. & Ganssen, G.M. (1998) Donovan and T.A. Jackson), pp. 13–39. University of the Reconstruction of the Miocene western Amazonian aquatic West Indies Publishers’ Association, Kingston, Jamaica. system using molluscan isotopic signatures. Palaeogeogra- Pitman, W.C., III, Cande, S.C., LaBrecque, J. & Pindell, J.L. phy, Palaeoclimatology, Palaeoecology, 141, 85–93. (1993) Fragmentation of Gondwana: the separation of Wahlberg, N. (2006) That awkward age for butterflies: insights Africa from South America. Biological relationships between from the age of the butterfly subfamily Nymphalinae (Lepi- Africa and South America (ed. by P. Goldblatt), pp. 15–34. doptera: Nymphalidae). Systematic Biology, 55, 703–714. Yale University Press, New Haven. White, J.L. & MacPhee, R.D.E. (2001) The sloths of the West Poeser, F. (2003) From the Amazon River to the Amazon Indies: a systematic and phylogenetic review. Biogeography molly and back again. PhD dissertation. IBED, Amsterdam of the West Indies: patterns and perspectives (ed. by C.A. University, Amsterdam, NL. Woods and F.E. Sergile), pp. 201. CRC Press, Boca Raton, Rambaut, A. (2009) FigTree v.1.4.2. Available at: http:// Florida. tree.bio.ed.ac.uk/software/figtree/ (accessed 9 July 2014). Woods, C.A. (2001) Introduction and historical overview of Rambaut, A. & Drummond, A.J. (2007) Tracer. Available at: patterns of West Indian biogeography. Biogeography of the http://beast.bio.edu.ac.uk/Tracer/). West Indies: patterns and perspectives (ed. by C.A. Woods Rauchenberger, M. (1988) Historical biogeography of poeciliid and F.E. Sergile), pp. 1–14. CRC Press, Boca Raton, Florida. fishes in the Caribbean. Systematic zoology, 37, 356–365. Ree, R.H. & Smith, S.A. (2008) Maximum likelihood infer- SUPPORTING INFORMATION ence of geographic range evolution by dispersal, local extinction, and cladogenesis. Systematic Biology, 57,4–14. Additional Supporting Information may be found in the Rıcan, O., Pialek, L., Zardoya, R., Doadrio, I. & Zrzavy, J. (2012) online version of this article: Biogeography of the MesoamericanCichlidae(Teleostei:Her- Appendix S1 Additional details regarding localities and oini): colonization through the GAARlandia land bridge and protocols. early diversification. Journal of Biogeography, 40,579–593. Appendix S2 Phylogenetic reconstructions and calibration Rivas, L.R. 1986. Comments on Briggs (1984): Freshwater schemes. fishes and biogeography of Central America and the Antil- Appendix S3 Matrices used in the historical biogeographi- les. Systematic Zoology, 35, 633–639. cal reconstructions. Robinson, E. (1965) Tertiary rocks of the Yallahs area. Jour- nal of the Geological Society of Jamaica, 7, 18–27. BIOSKETCH Robinson, E. (1994) Jamaica. Caribbean geology: an introduc- tion, 111, 127. Pablo Weaver is a researcher and instructor at the Univer- Rodriguez, C.M. (1997) Phylogenetic analysis of the tribe Poe- sity of La Verne in southern California and recently received ciliini (Cyprinodontiformes: Poeciliidae). Copeia, 1997, 663. his doctoral degree from the Ecology and Evolutionary Biol- Ronquist, F. & Huelsenbeck, J.P. (2003) MrBayes 3: Baye- ogy Department at the University of Colorado, Boulder. His sian phylogenetic inference under mixed models. Bioinfor- interests include ecology and evolution of livebearing fishes, matics, 19, 1572–1574. especially West Indian groups. Rosen, D.E. (1975) A vicariance model of Caribbean bio- geography. Systematic Biology, 24, 431–464. The main interests of all authors include evolution, phyloge- Rosen, D.E. (1985) Geological hierarchies and biogeographic netics and biogeography, with a focus on island systems. congruence in the Caribbean. Annals of the Missouri Author contributions: P.W. and A.C. conceived of the study. – Botanical Garden, 72, 636 659. P.W., A.C., and S.J. collected specimens from the field, and Rosen, D.E. & Bailey, R.M. (1963) The poeciliid fishes along with K.W., P.W. and S.J. conducted the gene sequenc- (Cyprinodontiformes): their structure, zoogeography, and ing and phylogenetic analyses. P.W. and S.J. led the writing systematics. Bulletin of the American Museum of Natural and provided the artwork. J.D. conducted the biogeographi- History, 126,1–176. cal analyses. A.C. and K.W. provided laboratory space and Silva, M. & Noll, F.B. (2015) Biogeography of the social wasp financial assistance. All authors contributed to the writing of genus Brachygastra (Hymenoptera: Vespidade: Polistinae). the manuscript. Journal of Biogeography, 42, 833–842. Swofford, D.L. (2003) PAUP*: phylogenetic analysis using parsimony, version 4.0 b10. Available at: http://paup.csit.f- Editor: Liliana Katinas su.edu/index.html (accessed 3 June 2015).

12 Journal of Biogeography ª 2016 John Wiley & Sons Ltd

View publication stats