Biogeographical Signature of River Capture Events in Amazonian

Journal of Biogeography (J. Biogeogr.) (2015)

ORIGINAL Biogeographical signature of river ARTICLE capture events in Amazonian lowlands Victor A. Tagliacollo1,2*, Fabio Fernandes Roxo1, Scott M. Duke-Sylvester2, Claudio Oliveira1 and James S. Albert2

1Instituto de Biociências de Botucatu, ABSTRACT Universidade Estadual Paulista – UNESP, Aim To investigate the effects of river capture on the biogeographical history Botucatu, SP 18618–970, Brazil, 2Biology of South American freshwater fishes. Department, University of Louisiana at Lafayette, Lafayette, LA 70504-2451, USA Location Western Amazon and La Plata basins, and adjacent river drainages. Methods We used a species-dense time-calibrated phylogeny of long- whiskered catfishes (Siluriformes, Pimelodidae) to calculate likelihoods for 16 biogeographical scenarios of river capture, each differing in details of (1) landscape evolution and/or (2) models of species range evolution. We designed eight alternative landscape evolution models (LEMs) to represent distinct palaeogeographical river capture histories between the Western Amazon and La Plata drainages during the formation of the Central Andean (Bolivian) orocline (43.0–15.0 Ma). The LEMs differed only in patterns of area-connectivity constraints through time, and otherwise had the same geographical areas, time durations and dispersal probabilities. We used the DEC and DECj models of species range evolution under these eight LEM constraints to calculate likelihood values for ancestral area estimates.

Results Divergence time estimates indicated that crown-group pimelodids emerged during the Late Cretaceous or Palaeogene (c. 72.9 20 Ma) and model-selection recovered a best-ﬁt palaeogeographical scenario with (1) a LEM with three river capture events, and (2) a DECj model of species range evolution. These results were quantitatively replicated using Lagrange and BayArea-like methods. Main conclusions The taxon–area chronogram of pimelodids exhibits the characteristic biogeographical signature of river capture; i.e. several non-monophyletic regional (basin-wide) species assemblages coupled with the presence of many species inhabiting more than one basin. These phylogenetic and biogeographical patterns are consistent with the effects of three large-scale river capture events during the formation of the Bolivian orocline.

*Correspondence: Victor A. Tagliacollo, Keywords Universidade Estadual Paulista – UNESP, Amazonian biodiversity, geographical range evolution, historical biogeography, Instituto de Bioci^encias de Botucatu, Botucatu, landscape evolution models, Neotropical ﬁshes, parametric biogeography, SP 18618–970, Brazil. E-mail: [email protected] Pimelodidae, river capture

barriers that separate geographical areas and divide ancestral INTRODUCTION population ranges (Humphries & Parenti, 1999), and (2) the Historical biogeography aims to understand how geomor- erosion of barriers that merge previously separated geograph- phological events and other landscape evolution processes ical areas and allow geographical range expansions among affect the geographical distributions of species, higher taxa these newly connected areas (Lieberman & Eldredge, 1996; and whole biotas. Two important geomorphological Lieberman, 2003a). Studies in vicariance biogeography exam- processes of landscape evolution are: (1) the emergence of ine the phylogenetic relationships of individual taxa, or

ª 2015 John Wiley & Sons Ltd http://wileyonlinelibrary.com/journal/jbi 1 doi:10.1111/jbi.12594 V. A. Tagliacollo et al. taxon–area relationships of multiple phylogenies, to docu- landscape evolution processes (Lieberman, 2003b; Pyron, ment the history of biotic range fragmentation among conti- 2014). A biogeographical signature implies that geomorpho- nents and other areas of endemism (Ronquist, 1997; Brooks logical events help to regulate species range evolution et al., 2001; Burridge et al., 2007; Sanmartın et al., 2008). (Rosen, 1978; Lieberman, 2005). The signature of river cap- Other studies document patterns of coordinated dispersal ture (Fig. 1) might sometimes be masked by stochastic between areas to understand how geographical merging events of long-distance dispersals (Cook & Crisp, 2005), events (i.e. geodispersal) contribute to the formation of regional or local extinctions (Lieberman, 2002), pseudocon- regional species assemblages (Lieberman & Eldredge, 1996; gruence in taxon–area relationships (Donoghue & Moore, Lieberman, 2003a). The goal of all these studies is to under- 2003), or uncertainties in phylogenetic or biogeographical stand the role of landscape evolution in the process of lin- reconstructions (Nylander et al., 2008; Ho & Phillips, 2009). eage diversification and formation of regional species With the development of molecular phylogenetic methods to assemblages (Aleixo & de Fatima Rossetti, 2007; Cowling estimate absolute clade ages (Lepage et al., 2007), and et al., 2009; Hoorn et al., 2010b; Badgley et al., 2014). model-based methodologies in biogeography to estimate spe- Vicariance and geodispersal are geomorphological pro- cies range evolution (Ree & Smith, 2008; Landis et al., 2013; cesses that affect range evolution by separating or connecting Matzke, 2014), these confounding contingencies can be mini- populations and biotas (Lieberman, 2003a; Wiens & Dono- mized (Ree & Sanmartın, 2009; Chacon & Renner, 2014). ghue, 2004; Buerki et al., 2011). These two processes usually One strategy is to evaluate the likelihood for alternative his- have opposite effects on marine and terrestrial taxa. For torical scenarios of species range evolution in the context of instance, the Plio-Pleistocene rise of the Panamanian isthmus alternative hypotheses of landscape evolution. separated some marine groups into populations in the Wes- In South America, biogeographical studies of river capture tern Atlantic and Eastern Pacific (Lessios, 2008), while simul- have mostly been conducted in geologically stable areas (e.g. taneously allowing range expansions of terrestrial organisms shields), where erosion is slow and the geophysical signals of between Central and South America (Marshall et al., 1982). palaeogeographical events are geologically persistent (Ribeiro, Similarly, the Late Cretaceous separation of South America 2006; Lujan et al., 2011; Roxo et al., 2014). On the other and Africa isolated terrestrial groups on either side of the hand, historical studies in lowland Amazonian river systems, emerging South Atlantic (Cracraft, 2001), while allowing including the Western Amazon and other portions of the range expansions of marine organisms between the Northern sub-Andean foreland, have received less attention (but and Southern Atlantic basins (Marshall et al., 1982). In both see Albert et al., 2006; Albert & Carvalho, 2011). The sub- these examples, some clades display a vicariant signature of Andean foreland is a retroarc depression located along the range splitting, while other clades display a geodispersal sig- eastern margin of the Andean Cordillera (Horton & DeCelles, nature of synchronized range expansions. 1997). Over the last 50 million years, the sub-Andean system River capture is a special case of landscape evolution that has been fragmented into semi-isolated sub-basins by a series results in both the formation and removal of barriers of geomorphic uplifts that reshaped the watershed bound- between portions of adjacent river drainages. River capture is aries (Lundberg et al., 1998; Hoorn et al., 2010a). The for- especially interesting from a biogeographical perspective mation of the Bolivian orocline during the late Eocene to because, by moving the physical location of watershed middle Miocene (43.0–15.0 Ma) contributed to the fragmen- boundaries, different portions of adjacent river basins tation of the sub-Andean foreland and resulted in large-scale become isolated and merged (Albert & Crampton, 2010). river capture events in the area of the headwaters of what is This predictable barrier displacement produces complex and today the Upper Madeira and Upper Paraguay basins (Lund- reticulated, but also predicable, patterns of taxon–area rela- berg et al., 1998; Albert & Reis, 2011). This area tionships (Albert & Carvalho, 2011; Roxo et al., 2014). River (~1,050,000 km2) is located in the department of Santa Cruz, capture can have great consequences for the geographical Bolivia, and it has a complex history of watershed move- range evolution and diversification of obligate aquatic taxa ments between the Western Amazon and La Plata basins. (e.g. freshwater fishes, some amphibians), and taxa special- Here, we investigate the effects of river capture on the bio- ized to riparian and floodplain habitats (e.g. some plants and geographical history of pimelodid catfishes (Ostariophysi, Sil- birds) in which dispersal is constrained by the geometry of uriformes) inhabiting the sub-Andean foreland basin, river drainage networks (Waters et al., 2000; Aleixo & de especially at the margins of the modern Western Amazon Fatima Rossetti, 2007; Grant et al., 2007; Albert et al., 2011). and La Plata basins. Pimelodids are mainly riverine fishes In the absence of confounding historical contingencies, that inhabit virtually all portions of the sub-Andean foreland river capture imprints a characteristic biogeographical basins, and which represent an interesting case to understand signature on taxon–area relationships: the presence of non- the effects of river capture on the spatial distribution of monophyletic regional (basin-wide) species assemblages fishes across the sub-Andean foreland basin. In this study, coupled with the presence of many species inhabiting more we designed eight alternative palaeogeographical scenarios than one basin (Fig. 1). A biogeographical signature is a (i.e. landscape evolution models – LEMs) of connections recurring pattern in the geographical distribution of one or among portions of the sub-Andean foreland, and used two more lineages that arises predictably from a particular set of model-based analytical methods in historical biogeography

2 Journal of Biogeography ª 2015 John Wiley & Sons Ltd River capture signature in pimelodids

(a) (b)

Figure 1 Biogeographical signature of river capture on the formation of regional species assemblages. (a) A landscape evolutionary model (LEM) illustrating the effects of river capture on landscape processes of geodispersal and vicariance. The formation of a vicariant at time interval 1 (T1) separates areas AB and C (horizontal black bar). The subsequent displacement of the barrier northwards at T2 separates areas A and B resulting in the species 1 and 2, and simultaneously connected areas B and C (i.e. BC) allowing dispersal (or range expansions) of species 2 to area C, and of species 3 to area B (red and blue arrows, respectively). Note the species assemblages of areas B and C are not monophyletic, and that species 2 and 3 occupy more than one area. (b) A single simulation run of interrelationships and geographical range evolution consistent with the hypothetical LEM of panel (a), under the following assumptions: (1) geographical barriers always cause instantaneous speciation (k = 1.0); (2) no extinctions (l = 0.0), and (3) dispersal is instantaneous and constrained by barriers only, not by geographical distance (dTi = 1.0). A–C = geographical areas; T0–T3 = time intervals; sp. 1– 3 = hypothetical species; black squares = speciation events.

(i.e. Lagrange and BayArea-like) to estimate likelihoods of 1. Platysilurus sp. – (c. 10.0–8.0 Ma), late Miocene, Urumaco pimelodid geographical range evolution on each of the Formation – Venezuela; see in Lundberg et al. (2010). LEMs. We demonstrate that pimelodids exhibit the charac- Platysilurus sp. was used to calibrate the clade P. malarmo + teristic biogeographical signature of river capture (Fig. 1), P. mucosus. and that their modern biogeographical distributions are con- 2. Brachyplatystoma promagdalena – (13.0–12.8 Ma), middle sistent with the effects of three large-scale and short-lived Miocene Villavieja Formation – Colombia (Lundberg, 2005). river capture events concomitant with the formation of the Brachyplatystoma promagdalena was used to calibrate the Bolivian orocline. clade composed of B. ﬁlamentosum, B. capapretum and B. rousseauxii. 3. Pimelodus group – (c. 40.0–30.0 Ma), Eocene–Oligocene, MATERIALS AND METHODS Yahuarango Formation – Peru; see in Sullivan et al. (2013). This taxon was used to calibrate the clade P. ornatus– Divergence time estimates Calophysus–Pimelodus (OCP). We used the most updated and complete molecular dataset Divergence time estimates were composed of two indepen- for South American Pimelodidae (Lundberg et al., 2011) to dent runs of Markov chain Monte Carlo (MCMC), each one estimate absolute lineage divergence times. This comprehen- comprising 5.0 9 107 generations. Parameter values and sin- sive dataset of 7362 base pairs (six genes) was originally gle trees were sampled every 5.0 9 103 generations. We assembled by Lundberg et al. (2011), and includes 57 employed the Bayesian phylogram of Lundberg et al. (2011) pimelodid species representing 27 of the 31 valid genera. We as the starting tree, and birth–death prior for estimates of re-aligned sequences in mafft (Katoh et al., 2002, 2005) and branching times (Gernhard, 2008). MCMC runs were com- time-calibrated the phylogeny using fossil age dates. The bined using LogCombiner 1.7.5. All parameter estimates were Lundberg et al. (2011) dataset contains suitable taxon sam- inspected for stationary convergence after the burn-in period. pling for each portion of the sub-Andean foreland basin used in this analysis. Palaeogeographical scenarios: landscape evolution We estimated lineage divergence times and phylogenetic models (LEMs) relationships in beast 1.7.5 (Drummond & Rambaut, 2007) using the same models of nucleotide substitutions suggested We designed eight LEMs to represent alternative histories of by Lundberg et al. (2011), and a lognormal relaxed- area-connectivity between the Western Amazon (WA), Upper molecular clock calibrated with palaeontologically based Madeira (UM), and La Plata (LP) basins. These LEMs repre- prior constraints. Lognormal prior constraints were based on sent distinct palaeogeographical scenarios regulating the geo- the youngest stratigraphic ages of the following pimelodid graphical evolution of species range distributions. All LEMs fossil taxa: possess the same set of (1) geographical units (river basins)

Journal of Biogeography 3 ª 2015 John Wiley & Sons Ltd V. A. Tagliacollo et al. to assign species distributions, (2) time intervals bounded by et al., 1998; Garzione et al., 2008). Each time interval was geomorphological events and (3) area-dispersal rate matrix bounded by large-scale geomorphological events potentially among geographical units. However, each LEM contains a resulting in river capture between adjacent river drainages unique set of area-connectivity constraints (i.e. adjacency (Table 2). The physical locations of the geographical barriers matrices) regulating the species geographical range evolution (watershed boundaries) between drainages were held con- through time (Fig. 2). stant through time intervals T0, T4 and T5, but were vari-

able at time intervals T1, T2 and T3, which correspond to the formation of the Bolivian orocline. Pattern of watershed Geographical units: species distributions displacements were combined in all possible ways among We designed ﬁve geographical units (areas) to assign pimelo- the geographical units LP, UM and WA. Each of these did species distributions. Species ranges were assigned to one watershed displacement patterns serves as an alternative sce- or more of these geographical units using data from pub- nario of river capture in the Santa Cruz region, Bolivia lished papers and museum specimens. The geographical units (Table 2). were composed of at least three conjoined freshwater ecoregions (Abell et al., 2008), and each geographical unit was Area-dispersal rate matrix quantitatively characterized by its geographical area (km2) and geometric centre (i.e. GPS coordinates; Table 1). We constructed an area-dispersal matrix that contains pairwise dispersal rates between geographical units (see Appendix S1 in Supporting Information). These dispersal rates are Time intervals: geomorphological events calculated using the spatial areas (km2) of the geographical We divided the palaeogeographical history of the sub- units, and the shortest, stepping-stone-like linear distances

Andean foreland into six time intervals (T0–T5) (Lundberg between the centre points (GPS coordinates) of the

Figure 2 Landscape evolution models (LEMs 1–8) representing eight river capture scenarios between the Western Amazon (WA), Upper Madeira (UM), and La Plata (LP) basins during the Bolivian orocline at around the late Eocene–middle Miocene (43.0– 15.0 Ma). These eight LEMs are embedded within the larger biogeographical history of the sub-Andean foreland (panel upper left), which also includes the Trans-Andean (TR) and Orinoco (OR) basins, and which extended throughout whole of the Maastrichtian and Cenozoic (73.0–0.0 Ma). The LEMs are composed of nine time-area units of similar temporal and areal dimensions, with identical dispersal rate matrices between time–area units when areas are connected, and different patterns of connectivity in space and time (the time–area adjacency matrix). Time T0 represents an open corridor to species dispersal along the length of the sub-Andean foreland. Horizontal solid lines indicate impermeable barriers to species dispersal. Horizontal dashed lines exclusively at T5 between LP and UM, and between WA and OR indicate semi-permeable barriers to dispersal (see section ‘Area-connectivity constraints: adjacency matrices’). Additional abbreviations: MA, Michicola Arch; CB, Chapare Buttress; BA, Bolivian Altiplano. T0 = 73.0–43.0 Ma, T1 = 43.0–32.0 Ma, T2 = 32.0–23.0 Ma, T3 = 23.0–15.0 Ma, T4 = 15.0–11.0 Ma, T5 = 11.0–0.0 Ma.

4 Journal of Biogeography ª 2015 John Wiley & Sons Ltd River capture signature in pimelodids

Table 1 Overall characteristics of geographical units. Geographical units are discrete conjoined freshwater ecoregions to assign species distributions, each characterized by geographical areas and geometric centres. Spatial area sizes and respective centroid GPS coordinates were used to construct a pairwise area-dispersal matrix between geographical units.

Geographical unit Abr. Ecoregions Area (km2) GPS coordinates

Trans-Andean TR 301, 302, 303 731,632 3.19° N 75.14° W Orinoco-Guiana lowland OR 304, 305, 306, 307, 308, 309, 310, 311 1,900,902 4.29° N 67.84° W Western Amazon lowlands WA 312, 313, 314, 315, 316, 317, 320, 322, 323, 324 6,094,940 6.55° S 75.08° W Guapore-Mamore-Madeira UM 318, 319, 321 1,053,630 17.92° S 63.00° W La Plata basins LP 332, 333, 342, 343, 344, 345, 346, 347 2,989,542 26.50° S 58.10° W

Table 2 Summary of time intervals used in landscape evolution models (LEMs). The LEMs are composed of six time intervals, each associated with a geomorphological event of river capture between adjacent geographical units. Note that all barriers are impermeable to species dispersal, except at T5 to indicate existing headwater river interchanges. Patterns of area-connectivity across LEMs vary only at T1, T2 and T3 resulting in all possible combinations of river capture events between the geographical units WA, UM and LP during the Bolivian orocline (c. 43.0–15.0 Ma).

Time intervals Ages (Ma) Geological uplifts Barrier permeability Connectivity LEMs Geographical barriers

T0 73–43 – Permeable Invariable – T1 43–32 Michicola Arch (MA) Impermeable Variable LP//UM T2 32–23 Chapare Buttress (CB) Impermeable Variable UM//WA T3 23–15 Bolivian Altiplano (BA) Impermeable Variable LP//UM T4 15–11 Eastern Cordillera (EC) Impermeable Invariable TR//WA-OR T5 11–0 Vaupes Arch (VA) Semi-impermeable Invariable WA//OR

geographical units (Table 1). Pairwise area-dispersal rates between geographical units at Ti time intervals. Alternative were calculated according to: patterns in these adjacency matrices represent different possi-

2 ble histories of river capture regulating the geographical range Aj=r D ¼ P ij (1) evolution. In adjacency matrices, values of zero (0) indicate ij n A =r2 1 j ij impermeable barriers to species dispersal, and values of one where Dij indicates the probability of dispersal from geo- (1) indicate permeable or semi-permeable barriers between graphical unit i to geographical unit j, Aj indicates the area adjacent geographical units. Note that differences between 2 2 fully permeable and semi-permeable barriers were given size (in km ) of the target geographical unit j, and rij indicates the linear distance (in km) between the centre points by rates of area-dispersal between geographical units, not by of geographical units i and j respectively (Table 1). This the binary values of the adjacency matrix (see section ‘Area- area-dispersal rate matrix is asymmetrical, where rows always dispersal rate matrix’). sum to 1.0 and diagonals are always 0.0. Pairwise area-dispersal rates between geographical units were Estimates of global likelihoods and ancestral ranges similar in all six time intervals and across all eight LEMs, except during time interval T5 between adjacent geographical units LP We estimated likelihoods of ancestral range evolution using and UM, and between WA and OR. Dispersal rates between the dispersal–extinction–cladogenesis (DEC) and DECj mod- these geographical units were adjusted in all eight LEMs to els of geographical (species) range evolution (Ree & Smith, model dispersal through semipermeable barriers across the Izo- 2008; Matzke, 2014). These DEC models are composed of zog swamp and Casiquiare canal, respectively (Wilkinson et al., two (DEC) or three (DECj) parameters including: (1) disper- 2006; Willis et al., 2010). These adjustments were made by sal (D), where species expand ancestral ranges by adding new dividing the original dispersal rates by two, to distinguish river- geographical units, (2) extinction (E), where species reduce ine connections with a higher probability of biological dispersal ancestral ranges by extirpation of geographical units and (3) from palaeogeographical conditions where species dispersal was ‘jumping’ events (j), where j speciﬁes the weight of ‘jumping’ limited by stream size and seasonality (e.g. Izozog swamp) or by events beyond an ancestral range (Matzke, 2014). The two rapids or water chemistry (e.g. Casiquiare canal). models of range evolution (i.e. DEC and DECj) are implemented in the R package BiogeoBears 0.2.1 (Matzke, 2014) in three different analytical procedures: DIVA-like (Ronquist, Area-connectivity constraints: adjacency matrices 1997), Lagrange (Ree & Smith, 2008) and BayArea-like We built pairwise adjacency matrices (see Appendix S2) to (Landis et al., 2013). We estimated likelihoods of ancestral constrain species dispersal or to allow species range expansions range evolution using the Lagrange and BayArea-like

Journal of Biogeography 5 ª 2015 John Wiley & Sons Ltd V. A. Tagliacollo et al.

6 Journal of Biogeography ª 2015 John Wiley & Sons Ltd River capture signature in pimelodids

Figure 3 Time-calibrated phylogeny containing mean divergence time estimates of the clade Pimelodidae. Chronogram estimated from 7362 aligned base pairs including six genes of 57 species representing 27 of the 31 valid pimelodid genera. Red numbers below branches indicate posterior probability values. Black numbers near to ancestral nodes indicate mean ages in Ma. Closed circles (•) represent palaeontological calibration points used to estimate divergence times. Clades: LN, Leiarius–neopimelodines Clade; N, neopimelodines; S, sorubimines; OCP, Pimelodus ornatus–Calophysus–Pimelodus Clade; CP, Calophysus–Pimelodus Clade; C, calophysines; P, pimelodines; EA, Exallodontus–Pimelodus altissimus Group; PI, Pimelodus Group (Lundberg et al., 2011). Symbol * indicates relationships among Hypophthalmus, Platystomatichthys, and Platysilurus sensu Lundberg et al. (2011). analytical procedures. BayArea-like is a simpliﬁed likelihood were calculated to provide measures of relative likelihoods interpretation of the program BayArea (Landis et al., 2013). across all 16 biogeographical scenarios. BayArea-like in BiogeoBears includes the same two parameters of the DEC model implemented in Lagrange (i.e. dis- RESULTS persal and extinction) and cladogenesis assumptions of the Bayesian model implemented in BayArea. Global likelihoods Time-calibrated tree at the root node of the 16 biogeographical scenarios (eight LEMs each with two models of species range evolution) were The time-calibrated tree of pimelodids is almost identical to compared using the Akaike information criterion (AIC; that proposed by Lundberg et al. (2011), differing only in Akaike, 1973). AIC weights (Wagenmakers & Farrell, 2004) the sister relationships between the clades Hypophthalmus

Table 3 Maximum likelihood estimates for 16 biogeography scenarios indicates the LEM 2 and DECj model range evolution best optimize (i.e. maximize likelihood) the geographical distribution of pimelodid clades along the sub-Andean foreland. Each biogeographical scenario represents a distinct combination between eight alternative LEMs (of river capture in the Central Andean), and two models of geographical range evolution (GRE; i.e. DEC or DECj). Similar likelihood results were quantitatively replicated with Lagrange. Abbreviations: LEM, landscape evolution model; GRE, geographical range evolution; DEC, dispersal–extinction–cladogenesis; ln L, log likelihood; K, parameters; d, dispersal; l, extinction; j, cladogenesis per-event weights; AIC, Akaike information criterion; DAIC, delta AIC; D AICw, AIC weighted.

LEM GRE ln LKd l j AIC DAIC DAICw

BayArea-like LEM 2 DECj 214.8 3 0.069 0.009 0.306 435.6 0.0 1.00 9 10+00 LEM 8 DECj 230.3 3 0.054 0.010 0.345 466.6 31.0 1.86 9 10 07 LEM 4 DECj 243.5 3 0.082 0.013 0.501 493.0 57.4 3.43 9 10 13 LEM 1 DECj 268.1 3 0.079 0.026 0.067 542.2 106.6 7.11 9 10 24 LEM 6 DECj 269.9 3 0.095 0.022 0.067 545.8 110.2 1.18 9 10 24 LEM 8 DEC 300.7 2 0.225 0.067 0.000 605.4 169.8 1.34 9 10 37 LEM 4 DEC 301.8 2 0.278 0.073 0.000 607.6 172.0 4.47 9 10 38 LEM 2 DEC 304.4 2 0.192 0.061 0.000 612.8 177.2 3.32 9 10 39 LEM 6 DEC 305.8 2 0.165 0.064 0.000 615.6 180.0 8.19 9 10 40 LEM 5 DEC 307.0 2 0.208 0.064 0.000 618.0 182.4 2.47 9 10 40 LEM 5 DECj 307.0 3 0.208 0.064 0.000 620.0 184.4 9.08 9 10 41 LEM 1 DEC 308.4 2 0.162 0.058 0.000 620.8 185.2 6.09 9 10 41 LEM 3 DEC 308.6 2 0.238 0.070 0.000 621.2 185.6 4.98 9 10 41 LEM 3 DECj 308.6 3 0.238 0.070 0.000 623.2 187.6 1.83 9 10 41 LEM 7 DEC 311.0 2 0.185 0.060 0.000 626.0 190.4 4.52 9 10 42 LEM 7 DECj 311.0 3 0.185 0.060 0.000 628.0 192.4 1.66 9 10 42 Lagrange LEM 2 DECj 227.2 3 0.076 0.008 0.254 460.4 0.0 1.00 9 10+00 LEM 6 DECj 252.8 3 0.082 0.009 0.092 511.6 51.2 7.62 9 10 12 LEM 1 DECj 256.4 3 0.116 0.016 0.054 518.8 58.4 2.08 9 10 13 LEM 2 DEC 259.6 2 0.136 0.017 0.000 523.2 62.8 2.31 9 10 14 LEM 8 DECj 259.7 3 0.144 0.022 0.019 525.4 65.0 7.68 9 10 15 LEM 5 DEC 263.0 2 0.139 0.018 0.000 530.0 69.6 7.70 9 10 16 LEM 6 DEC 263.1 2 0.129 0.018 0.000 530.2 69.8 6.97 9 10 16 LEM 5 DECj 263.0 3 0.139 0.018 0.000 532.0 71.6 2.83 9 10 16 LEM 7 DEC 264.7 2 0.130 0.017 0.000 533.4 73.0 1.41 9 10 16 LEM 1 DEC 265.1 2 0.116 0.020 0.000 534.2 73.8 9.43 9 10 17 LEM 7 DECj 264.7 3 0.130 0.017 0.000 535.4 75.0 5.18 9 10 17 LEM 4 DEC 266.8 2 0.175 0.025 0.000 537.6 77.2 1.72 9 10 17 LEM 4 DECj 266.8 3 0.175 0.025 0.000 539.6 79.2 6.34 9 10 18 LEM 3 DEC 269.7 2 0.154 0.021 0.000 543.4 83.0 9.48 9 10 19 LEM 3 DECj 269.7 3 0.154 0.021 0.000 545.4 85.0 3.49 9 10 19 LEM 8 DEC 271.5 2 0.084 0.014 0.000 547.0 86.6 1.57 9 10 19

Journal of Biogeography 7 ª 2015 John Wiley & Sons Ltd V. A. Tagliacollo et al.

8 Journal of Biogeography ª 2015 John Wiley & Sons Ltd River capture signature in pimelodids

Figure 4 Ancestral range estimates of the clade Pimelodidae using LEM 2 and DECj model of range evolution under the BayArea-like analytical procedure. Colour changes between stem (squares) and crown (circles) groups indicate dispersal and/or extinction events. Note that multiple colours represent species present in multiple geographical areas, not probabilities of alternative ancestral area estimates. Ancestral ranges were estimated using (1) LEM 2 palaeogeographical scenario, (2) DECj model of geographical range evolution, and (3) BayArea-like analytical procedure. Time intervals: T0 = open corridor to species dispersal throughout the whole sub- Andean foreland basin; T1 = Michicola Arch (MA), T2 = Chapare Buttress (CB), T3 = Bolivian Altiplano (BA), T4 = Eastern Cordillera (EC), T5 = Vaupes Arch (VA). Solid and dashed parallel bars represent impermeable and semi-permeable barriers to species dispersal, respectively. RC1 = 1st river capture event at start of T1: rivers of Upper Madeira (UM) basin flowing northwards allowing species dispersal to Western Amazon (WA), Orinoco (OR), and/or Trans-Andean (TR) regions. RC2 = 2nd river capture event at start of T2: UM flowing southwards allowing species dispersal to the La Plata basin (LP). RC3 = 3rd river capture event at start of T3:UM geographical unit flowing northwards again allowing species dispersal to the WA, OR and TR regions. Symbol * indicates relationships among Hypophthalmus, Platystomatichthys and Platysilurus sensu Lundberg et al. (2011).

and Platystomatichthys + Platysilurus. Here, we estimate that 39.9 10 Ma; the Calophysus-Pimelodus Clade (CP) emerged Hypophthalmus is the sister group to all other sorubimines, around 36.9 9 Ma; calophysines (C) emerged around except Zungaro zungaro, while in Lundberg et al. (2011) the 35.2 9 Ma; pimelodines (P) emerged around 30.5 8 Ma; clade Hypophthalmus is sister group to the Platystomatichthys the Pimelodus Group (PI) emerged around 22.2 6 Ma, and + Platysilurus (Fig. 3). This divergence might be due to the the Exallodontus–Pimelodus altissimus Group (EA) emerged stochasticity of MCMC runs that sometimes do not converge around 17.8 5 Ma. The absolute age values (in Ma) indicate to the same global likelihood space (Huelsenbeck et al., posterior mean and their respective credibility intervals. 2002). We inspected the posterior density distributions of the two independent MCMC runs to ensure that global Palaeogeographical scenarios optima were achieved for all parameter estimates. Relation- ships between the clades Hypophthalmus and Platystom- LEM 2 (see Fig. 2) is the best-ﬁt palaeogeographical scenario atichthys + Platysilurus are statistically poorly supported by to optimize the estimates of ancestral range evolution of posterior values in our time-calibrated phylogeny (Fig. 3) pimelodids in the sub-Andean foreland basin (Table 3). LEM and the Bayesian phylogram proposed by Lundberg et al. 2 is the best landscape scenario regardless of which analytical (2011). This divergence between hypotheses did not have procedure (BayArea-like, Lagrange) was used to calculate strong effects on the quantitative results of ancestral range likelihoods of ancestral range estimates (Table 3). AIC estimates of pimelodid clades inhabiting the sub-Andean weights indicated that LEM 2 average more than 99% across foreland. all other LEMs (Table 3). This result suggests that Pimelodi- dae was exposed to three river capture events in the Santa Cruz region of southern Bolivia during the formation of the Model parameter estimates Bolivian orocline. Effective sample sizes for each MCMC run suggested that the model parameters were all above the threshold of 200, indi- Ancestral range estimates cating an adequate MCMC chain mixing and global opti- mum parameter estimates. Examining the two stationary Ancestral range estimates using LEM 2 and BayArea-like MCMC convergence curves, we discarded the 1.0 9 107 (1st indicated the Pimelodidae was present in the Santa Cruz run) and 1.5 9 107 (2nd run) initial generations as a burn- region, Bolivia, in what is now the headwater region of the in procedure. Posterior topologies were summarized as 95% Upper Madeira and La Plata basins (Fig. 4); i.e. pimelodids maximum clade credibility. had to be in that region to have been affected by river cap-

ture between the LP and UM at T1. Ancestral range estimates also indicated that pimelodids were exposed to the effects of Divergence time estimates three river capture events between the Western Amazon and

Divergence time estimates (Fig. 3) indicated that the ancestor La Plata basins. The 1st river capture (T1 = 43.0–32.0 Ma) of the Pimelodidae emerged during the Late Cretaceous or reshaped watershed boundaries by separating the LP and Palaeogene (c. 72.9 20 Ma). The primary lineage-split UM and, simultaneously, connecting the UM and WA between Steindachneridion scriptum and all other pimelodids (Fig. 4 – RC1). This event divided the ancestral pimelodid was estimated at around 56.4 15 Ma, and subsequent crown-group species between Steindachneridion, and other lineage-splitting between Phractocephalus hemioliopterus and pimelodids (Fig. 4 – RC1). the Leiarius–neopimelodines Clade (LN) at around The 2nd river capture (T2 = 32.0–23.0 Ma) reshaped 52.6 12 Ma. The age estimates also indicated that: watershed boundaries by separating the WA and UM and, neopimelodines (N) emerged around 44.4 12 Ma; soru- simultaneously, connecting the UM and LP (Fig. 4 – RC2). bimines (S) emerged around 38.2 10 Ma; the Pimelodus This event resulted in the separation of ﬁve pimelodid ornatus–Calophysus–Pimelodus Clade (OCP) emerged around groups: (1) Hemisorubim platyrhynchos and Brachyplatystoma

Journal of Biogeography 9 ª 2015 John Wiley & Sons Ltd V. A. Tagliacollo et al. spp. + Platynematichthys notatus; (2) the two major clades of adjacent river basins allows taxa inhabiting the newly cap- pimelodines – clade P; (3) Pimelodina ﬂavipinnis and its sis- tured drainage to disperse outside the area of its ancestral ter-clade the Calophysus, Luciopimelodus and Pinirampus range, while species in a formerly isolated basin may expand spp.; (4) Cheirocerus + Megalonema clades; (5) and Leiarius their range to the newly captured river. Within monophyletic marmoratus and L. pictus + Perrunichthys (Fig. 4 – RC2). clades, such processes of range fragmentation and expansion

The 3rd river capture (T3 = 23.0–15.0 Ma) reshaped result in the formation of non-monophyletic regional watershed boundaries by separating once again the LP and assemblages in the basin containing the newly captured UM and, simultaneously, connecting the UM and WA drainage. (Fig. 4 – RC3). This event reveals pseudocongruence in taxon–area relationships of pimelodids inhabiting the Wes- Landscape evolution of the sub-Andean foreland tern Amazon and La Plata basins. This 3rd event resulted in the separation of four pimelodid groups: (1) Pseudoplatys- The best-fit palaeogeographical scenario (LEM) indicated toma corruscans and P. magdaleniatum; (2) the clades Pimelo- that the Pimelodidae was affected by at least three river cap- dus blochii + P. coprophagus and P. maculatus + P. albicans; ture events (RC1 – 3), spatially located in the area of what is (3) Megalonema platinum and M. platycephalum; and (4) today the Santa Cruz region of southern Bolivia, and tempo- Calophysus macropterus and Luciopimelodus pati (Fig. 4 – rally coinciding with the geological formation of the Bolivian RC3). Finally, three pimelodid species have expanded their orocline in the time frame 43.0–15.0 Ma (during the late range via semi-permeable barriers between the LP and UM Eocene to middle Miocene) (Table 3, Figs 4 & 5). These at T5. The species Sorubim lima, Hemisorubim platyrhynchos river capture events moved the watershed of the UM and LP and Pinirampus pirinampu expanded their ranges southwards basins northwards (RC1) into the Western Amazon, then to the La Plata basin (Fig. 4). southwards (RC2) into the La Plata basin, and then again northwards (RC3) into the Western Amazon (Fig. 2 – LEM 2; Figs 4 & 5). These three river capture events are hypothe- DISCUSSION sized to be spatially and temporally coincident with three geomorphological events in the Santa Cruz region: (1) the Biogeographical signature of river capture rise of the Michicola Arch at c. 43.0–32.0 Ma, (2) the rise of We uncovered several examples of pimelodid taxon–area the Chapare Buttress at c. 28.0–23.0 Ma, and (3) the initial relationships that can be attributed to river capture events in rise of the Bolivian Altiplano at c. 23.0–15.0 Ma (Lundberg the sub-Andean foreland. These river capture events et al., 1998; Allmendinger et al., 2005; Garzione et al., 2008; imprinted consistent biogeographical patterns on clades Hoorn et al., 2010b). Other hypothetical capture events, inhabiting the Western Amazon, Upper Madeira, and La implemented as alternative LEMs (Fig. 2), did not increase Plata basins (Fig. 4). Patterns in the spatial and phylogenetic likelihood values of ancestral area estimates under the two distributions of these pimelodid taxa exhibit the characteris- models (DEC, DECj) of range evolution (Table 3). tic biogeographical signature of river capture; i.e. non-mono- Likelihood estimates indicated (with 99% of AIC weights) phyletic regional (basin-wide) species assemblages coupled that the biogeographical history of pimelodids was affected with the presence of many species inhabiting more than one by the following hypothesized sequence of events resulting in basin (Fig. 1). the fragmentation of the sub-Andean foreland (see in Lund- The biogeographical signature of river capture observed berg et al., 1998). The first river capture event (RC1, Fig. 4) here for Pimelodidae has also been revealed in several other separated an extensive sub-Andean foreland into two por- taxon–area relationships of tropical freshwater fishes, includ- tions, one draining south to what is now the La Plata basin, ing: Gymnotus (Albert et al., 2005; Lovejoy et al., 2010), and the other draining north to what is now the Western Pseudoplatystoma (Torrico et al., 2009), Aphyocharax (Taglia- Amazon basin, also referred to as the proto-Amazon- collo et al., 2012) Serranochromis (Musilova et al., 2013) and Orinoco River (Lundberg et al., 1998; Albert et al., 2006,

Otothyrinae (Roxo et al., 2014). In these groups, geographi- Fig. 2, LEM 2 at T1). We estimate the RC1 event occurred at cal range evolution is usually constrained by watershed c. 43.0–32.0 Ma (late Eocene) in association with the uplift boundaries and dispersal to adjacent drainages depends of Michicola Arch (Lundberg et al., 1998; Jacques, 2003). exclusively on physical connections between river basins The full impact of this river capture event on the biogeo- (Bishop, 1995; Albert & Crampton, 2010). For most freshwa- graphical history of lowland Amazonian ﬁshes is incom- ter ﬁshes river capture modulates clade range evolution by pletely known. In pimelodids, only a single clade including isolating ancestral populations between river basins and, the ancestral node of all Pimelodidae (RC1 – Fig. 4) was almost simultaneously, allowing synchronized dispersals to affected by the RC1 event, limiting inferences about the mag- newly captured drainages. The synchronized dispersal of nitude of its spatial scale. monophyletic groups is a central concept to understand The second river capture event (RC2, Fig. 4) separated how river capture, or watershed displacement, leads to the some portion of the Upper Madeira basin formerly con- formation of the characteristic river capture biogeographical nected to Western Amazon, and connected this portion with signature (Fig. 1). The displacement of a barrier between a portion of the rivers formerly connected to the La Plata

10 Journal of Biogeography ª 2015 John Wiley & Sons Ltd River capture signature in pimelodids

Figure 5 Three river capture events in the Santa Cruz region of southern Bolivia during the late Eocene to middle Miocene. Note these capture events moved the watershed of the Upper Madeira (UM) basin southwards into the Western Amazon, then northwards into the La Plata basin, and then again southwards into the Western Amazon. TR, Trans-Andean; OR, Orinoco; WA, Western Amazon; UM, Upper Madeira; LP, La Plata. RC1–3 = river capture events. T1 = 43.0–32.0 Ma, T2 = 32.0–23.0 Ma, T3 = 23.0–15.0 Ma.

basin (Fig. 2, LEM 2 at T2). We estimate that the RC2 event The RC3 event reveals a recurrent biogeographical pattern occurred at c. 32.0–23.0 Ma (late Oligocene) in association in the taxon–area relationships of pimelodids inhabiting the with the rise of the Chapare Buttress (Sempere et al., 1990). sub-Andean foreland: the non-monophyly of regional assem- The RC2 event was probably a large-scale river capture, and blages, and sister relationships between taxa inhabiting the it affected several pimelodid clades originally evolving in the La Plata and Western Amazon basins (Figs 3 & 4). Similar proto-Amazon-Orinoco River (see Lundberg et al., 1998). In spatial patterns have been reported in other ostariophysan pimelodids, this event imprinted a similar biogeographical clades (Montoya-Burgos, 2003; Torrico et al., 2009; Musilova signature on ﬁve clades that have representatives in the Santa et al., 2013). An illustrative example in pimelodids is the Cruz region (RC2 – Fig. 4). Similarities in clade age esti- clade Pimelodus group (PI), in which the clade P. albicans mates among these groups suggest the RC2 event was rela- (La Plata) + P. maculatus (La Plata) is the sister group of tively rapid, occurring during the interval 31.6–26.8 Ma the clade P. blochii (Western Amazon) + P. coprophagus (Fig. 3). This brief interval of 4.8 Ma may indicate the for- (Western Amazon), instead of the clade Iheringichthys mation of a strong riverine connection, allowing species labrosus (La Plata) + Parapimelodus ssp. (La Plata) in the range expansions between the La Plata and Western Amazon Western Amazon basin (Figs 3 & 4). This spatial pattern of basins in the Santa Cruz region. taxon–area relationships in pimelodids represents a readily The third river capture event (RC3 – Fig. 4) separated a recognized biogeographical signature (Fig. 1), in which closely portion of the Upper Madeira basin formerly connected to related taxa in the La Plata basin do form a non-monophyletic the La Plata basin, and connected this region with a portion regional assemblage. Another example of this biogeographical of the Western Amazon; i.e. the RC3 event captured a signature in pimelodids is observed in the clade Megalonema palaeodrainage of the Upper Madeira river and redirected it ssp., in which M. platycephalum (Western Amazon) is the to ﬂow northwards into the Western Amazon basin (Fig. 2, sister taxon to M. platanum (La Plata) instead of its sympatric

LEM 2 at T3). We estimate this event occurred at c. 23.0– congener M. amaxanthum (Western Amazon); i.e. in this case 15.0 Ma (late to middle Miocene) in association with the most closely related taxa in the Western Amazon do not form initial uplift of the Bolivian Altiplano (Garzione et al., a monophyletic group (Fig. 4). 2008). RC3 was probably a large-scale river capture event From a biogeographical perspective RC1 and RC3 are affecting a wide geographical range and the biogeographical pseudocongruent events (Donoghue & Moore, 2003), distributions of several freshwater clades, including at least exhibiting similar spatial patterns of taxon–area relationships, four pimelodid clades originally evolving in south-ﬂowing but exhibiting dissimilar lineage divergence times. The reason (La Plata) rivers (Fig. 4 – RC3). RC3 was also a relatively for this pseudocongruence is that rivers of the Western Ama- short-lived event, occurring between 19.1 and 14.9 Ma zon twice and independently captured palaeodrainages of the (Fig. 3). Upper Madeira basin that had been in each case previously

Journal of Biogeography 11 ª 2015 John Wiley & Sons Ltd V. A. Tagliacollo et al. connected to the La Plata basin (Fig. 2 – LEM 2; Fig. 4). Albert, J.S. & Reis, R.E. (2011) Historical biogeography of Pseudocongruent biogeographical events such as these illus- Neotropical freshwater fishes. University of California Press, trate the importance of high-resolution geophysical data Berkeley, CA. regarding the timing of landscape evolution events on our Albert, J.S., Crampton, W.G.R., Thorsen, D.H. & Lovejoy, understanding of the processes of biotic diversification. N.R. (2005) Phylogenetic systematics and historical biogeography of the Neotropical electric fish Gymnotus (Teleostei: Gymnotidae). Systematics and Biodiversity, 2, CONCLUSIONS 375–417. Several clades of pimelodid catfishes exhibit biogeographical Albert, J.S., Lovejoy, N.R. & Crampton, W.G.R. (2006) Miocene signatures of river capture in their taxon–area relationships: tectonism and the separation of cis- and trans-Andean river i.e. non-monophyletic regional (basin-wide) species assem- basins: evidence from Neotropical fishes. Journal of South blages coupled with the presence of many species inhabiting American Earth Sciences, 21,14–27. more than one basin. Likelihood-based estimates of pimelo- Albert, J.S., Petry, P. & Reis, R.E. (2011) Major biogeo- did biogeographical evolution (Table 3) suggest a palaeogeo- graphic and phylogenetic patterns. Historical biogeography graphical scenario with three river capture events, or of neotropical freshwater fishes (ed. by J.S. Albert and R.E. watershed-barrier displacements, in the Santa Cruz region of Reis), pp. 21–57. University of California Press, Berkeley, southern Bolivia during formation of the Bolivian orocline CA. (c. 43.0–15.0 Ma). We suggest that the resulting complex Aleixo, A. & de Fatima Rossetti, D. (2007) Avian gene trees, taxon–area relationships between the La Plata and Western landscape evolution, and geology: towards a modern syn- Amazon are a consequence of three large-scale and relatively thesis of Amazonian historical biogeography? Journal of short-lived river capture events in the region of what is now Ornithology, 148, S443–S453. the Upper Madeira basin. Allmendinger, R.W., Smalley, R., Bevis, M., Caprio, H. & Brooks, B. (2005) Bending the Bolivian orocline in real 33 – ACKNOWLEDGEMENTS time. Geology, , 905 908. Badgley, C., Smiley, T.M. & Finarelli, J.A. (2014) Great Basin We thank Joseph Neigel, Tiago Carvalho, Emmanuel Max- mammal diversity in relation to landscape history. Journal ime, Gisela Farinelli Tagliacollo, Max Bernt, Kory-Evans, of Mammalogy, 95, 1090–1106. Brandon Waltz, Jack Craig and Damian Green for their Bishop, P. (1995) Drainage rearrangement by river capture, comments and suggestions. Special thanks to John Lundberg beheading and diversion. Progress in Physical Geography, for his contributions to the understanding of South America 19, 449–473. ichthyology and biogeography. This work was supported by Brooks, D.R., van Veller, M.G.P. & McLennan, D.A. (2001) the United States National Science Foundation (NSF DEB How to do BPA, really. Journal of Biogeography, 28, 0614334, 0741450 and 1354511 to J.S.A.), the Fundacß~ao de 345–358. Amparo a Pesquisa do Estado de S~ao Paulo (FAPESP 2012/ Buerki, S., Forest, F., Alvarez, N., Nylander, J.A.A., Arrigo, 09990-0 to V.A.T, FAPESP 2014/05051-5 to F.F.R.), the N. & Sanmartın, I. (2011) An evaluation of new parsi- Conselho Nacional de Desenvolvimento Cientıfico e Tec- mony-based versus parametric inference methods in nologico (CNPq 309632/2007-2 to C.O.) and the Louisiana biogeography: a case study using the globally distributed Education Quality Support Fund (2011-14-RD-A-27 to plant family Sapindaceae. Journal of Biogeography, 38, S.M.D-S.). 531–550. Burridge, C.P., Craw, D. & Waters, J.M. (2007) An empirical test of freshwater vicariance via river capture. Molecular REFERENCES Ecology, 16, 1883–1895. Abell, R., Thieme, M.L., Revenga, C. et al. (2008) Freshwater Chacon, J. & Renner, S.S. (2014) Assessing model sensitivity ecoregions of the world: a new map of biogeographic units in ancestral area reconstruction using Lagrange: a case for freshwater biodiversity conservation. BioScience, 58, study using the Colchicaceae family. Journal of Biogeogra- 403–414. phy, 41, 1414–1427. Akaike, H. (1973) Maximum likelihood identification of Cook, L.G. & Crisp, M.D. (2005) Directional asymmetry of Gaussian autoregressive moving average models. Biome- long-distance dispersal and colonization could mislead trika, 60, 255–265. reconstructions of biogeography. Journal of Biogeography, Albert, J.S. & Carvalho, T.P. (2011) Neogene assembly of 32, 741–754. modern faunas. Historical biogeography of Neotropical fresh- Cowling, R.M., Proches, Sß. & Partridge, T.C. (2009) Explain- water fishes (ed. by J.S. Albert and R.E. Reis), pp. 119–136. ing the uniqueness of the Cape flora: incorporating geomor- University of California Press, Berkeley, CA. phic evolution as a factor for explaining its diversification. Albert, J.S. & Crampton, W.G.R. (2010) The geography and Molecular Phylogenetics and Evolution, 51,64–74. ecology of diversification in Neotropical freshwaters. Cracraft, J. (2001) Avian evolution, Gondwana biogeography Nature Education Knowledge, 1,13–19. and the Cretaceous–Tertiary mass extinction event.

12 Journal of Biogeography ª 2015 John Wiley & Sons Ltd River capture signature in pimelodids

Proceedings of the Royal Society B: Biological Sciences, 268, Lepage, T., Bryant, D., Philippe, H. & Lartillot, N. (2007) A 459–469. general comparison of relaxed molecular clock models. Donoghue, M.J. & Moore, B.R. (2003) Toward an integrative Molecular Biology and Evolution, 24, 2669–2680. historical biogeography. Integrative and Comparative Biol- Lessios, H.A. (2008) The great American schism: divergence ogy, 43, 261–270. of marine organisms after the rise of the Central American Drummond, A.J. & Rambaut, A. (2007) BEAST: Bayesian Isthmus. Annual Review of Ecology, Evolution, and System- evolutionary analysis by sampling trees. BMC Evolutionary atics, 39,63–91. Biology, 7, 214. Lieberman, B.S. (2002) Phylogenetic biogeography with and Garzione, C.N., Hoke, G.D., Libarkin, J.C., Withers, S., without the fossil record: gauging the effects of extinction MacFadden, B., Eiler, J., Ghosh, P. & Mulch, A. (2008) and paleontological incompleteness. Palaeogeography, Rise of the Andes. Science, 320, 1304–1307. Palaeoclimatology, Palaeoecology, 178,39–52. Gernhard, T. (2008) The conditioned reconstructed process. Lieberman, B.S. (2003a) Paleobiogeography: the relevance of Journal of Theoretical Biology, 253, 769–778. fossils to biogeography. Annual Review of Ecology, Evolu- Grant, E.H.C., Lowe, W.H. & Fagan, W.F. (2007) Living in tion, and Systematics, 34,51–69. the branches: population dynamics and ecological Lieberman, B.S. (2003b) Taking the pulse of the Cambrian processes in dendritic networks. Ecology Letters, 10, 165– radiation. Integrative and Comparative Biology, 43, 175. 229–237. Ho, S.Y.W. & Phillips, M.J. (2009) Accounting for calibra- Lieberman, B.S. (2005) Geobiology and paleobiogeography: tion uncertainty in phylogenetic estimation of evolutionary tracking the coevolution of the Earth and its biota. Palaeo- divergence times. Systematic Biology, 58, 367–380. geography, Palaeoclimatology, Palaeoecology, 219,23–33. Hoorn, C., Wesselingh, F.P., Hovikoski, J. & Guerrero, J. Lieberman, B.S. & Eldredge, N. (1996) Trilobite biogeogra- (2010a) The development of the amazonian mega-wetland phy in the Middle Devonian: geological processes and ana- (Miocene; Brazil, Colombia, Peru, Bolivia). Amazonia, lytical methods. Paleobiology, 22,66–79. landscape and species evolution: a look into the past (ed. by Lovejoy, N.R., Lester, K., Crampton, W.G.R., Marques, F.P.L. C. Hoorn and F. Wesselingh), pp. 123–142. Blackwell- & Albert, J.S. (2010) Phylogeny, biogeography, and electric Wiley, Hoboken, NJ. signal evolution of Neotropical knifefishes of the genus Hoorn, C., Wesselingh, F.P., ter Steege, H., Bermudez, M.A., Gymnotus (Osteichthyes: Gymnotidae). Molecular Phyloge- Mora, A., Sevink, J., Sanmartın, I., Sanchez-Meseguer, A., netics and Evolution, 54, 278–290. Anderson, C.L., Figueiredo, J.P., Jaramillo, C., Riff, D., Lujan, N.K., Armbruster, J.W., Albert, J. & Reis, R. (2011) Negri, F.R., Hooghiemstra, H., Lundberg, J., Stadler, T., The Guiana Shield. Historical biogeography of Neotropical S€arkinen, T. & Antonelli, A. (2010b) Amazonia through freshwater fishes (ed. by J.S. Albert and R.E. Reis), pp. time: Andean uplift, climate change, landscape evolution, 211–224. University of California Press, Berkeley, CA. and biodiversity. Science, 330, 927–931. Lundberg, J.G. (2005) Brachyplatystoma promagdalena, new Horton, B.K. & DeCelles, P.G. (1997) The modern foreland species, a fossil goliath catfish (Siluriformes: Pimelodidae) basin system adjacent to the Central Andes. Geology, 25, from the Miocene of Colombia, South America. Neotropi- 895–898. cal Ichthyology, 3, 597–605. Huelsenbeck, J.P., Larget, B., Miller, R.E. & Ronquist, F. Lundberg, J.G., Marshall, L.G., Guerrero, J., Horton, B., Mal- (2002) Potential applications and pitfalls of Bayesian infer- abarba, M. & Wesselingh, F. (1998) The stage for ence of phylogeny. Systematic Biology, 51, 673–688. Neotropical fish diversification: a history of tropical South Humphries, C.J. & Parenti, L.R. (1999) Cladistic biogeogra- American rivers. Phylogeny and classification of Neotropical phy. Oxford University Press, Oxford. fishes (ed. by L.R. Malabarba, R.E. Reis, R.P. Vari, Z.M. Jacques, J.M. (2003) A tectonostratigraphic synthesis of the Lucena and C.A.S. Lucena), pp. 13–48. EDIPUCRS, Porto Sub-Andean basins: implications for the geotectonic seg- Alegre. mentation of the Andean Belt. Journal of the Geological Lundberg, J.G., Sabaj Perez, M.H., Dahdul, W.M. & Aguilera, Society, 160, 687–701. O.A. (2010) The Amazonian Neogene fish fauna. Amazo- Katoh, K., Misawa, K., Kuma, K. & Miyata, T. (2002) nia, landscape and species evolution: a look into the past MAFFT: a novel method for rapid multiple sequence (ed. by C. Hoorn and F.P. Wesselingh), pp. 281–301. alignment based on fast Fourier transform. Nucleic Acids Blackwell-Wiley, Hoboken, NJ. Research, 30, 3059–3066. Lundberg, J.G., Sullivan, J.P. & Hardman, M. (2011) Phylo- Katoh, K., Kuma, K., Toh, H. & Miyata, T. (2005) MAFFT genetics of the South American catfish family Pimelodidae version 5: improvement in accuracy of multiple sequence (Teleostei: Siluriformes) using nuclear and mitochondrial alignment. Nucleic Acids Research, 33, 511–518. gene sequences. Proceedings of the Academy of Natural Landis, M.J., Matzke, N.J., Moore, B.R. & Huelsenbeck, Sciences of Philadelphia, 161, 153–189. J.P. (2013) Bayesian analysis of biogeography when the Marshall, L.G., Webb, S.D., Sepkoski, J.J. & Raup, D.M. number of areas is large. Systematic Biology, 62, (1982) Mammalian evolution and the Great American 789–804. Interchange. Science, 215, 1351–1357.

Journal of Biogeography 13 ª 2015 John Wiley & Sons Ltd V. A. Tagliacollo et al.

Matzke, N.J. (2014) Model selection in historical biogeogra- Tagliacollo, V.A., Souza-Lima, R., Benine, R.C. & Oliveira, phy reveals that founder-event speciation is a crucial pro- C. (2012) Molecular phylogeny of Aphyocharacinae cess in island clades. Systematic Biology, 63, 951–970. (Characiformes, Characidae) with morphological diagnoses Montoya-Burgos, J.I. (2003) Historical biogeography of the for the subfamily and recognized genera. Molecular Phylo- catfish genus Hypostomus (Siluriformes: Loricariidae), with genetics and Evolution, 64, 297–307. implications on the diversification of Neotropical ichthy- Torrico, J.P., Hubert, N., Desmarais, E., Duponchelle, F., ofauna. Molecular Ecology, 12, 1855–1867. Nunez~ Rodriguez, J., Montoya-Burgos, J., Garcia Davila, Musilova, Z., Kalous, L., Petrtyl, M. & Chaloupkova, P. C., Carvajal-Vallejos, F.M., Grajales, A.A., Bonhomme, F. (2013) Cichlid fishes in the Angolan headwaters region: & Renno, J.F. (2009) Molecular phylogeny of the genus molecular evidence of the ichthyofaunal contact between Pseudoplatystoma (Bleeker, 1862): biogeographic and evo- the Cuanza and Okavango-Zambezi Systems. PLoS ONE, lutionary implications. Molecular Phylogenetics and Evolu- 8, e65047. tion, 51, 588–594. Nylander, J.A.A., Olsson, U., Alstrom, P. & Sanmartın, I. Wagenmakers, E.J. & Farrell, S. (2004) AIC model selection (2008) Accounting for phylogenetic uncertainty in bio- using Akaike weights. Psychonomic Bulletin & Review, 11, geography: a Bayesian approach to dispersal–vicariance 192–196. analysis of the thrushes (Aves: Turdus). Systematic Biology, Waters, J.M., Dijkstra, L.H. & Wallis, G.P. (2000) Biogeogra- 57, 257–268. phy of a southern hemisphere freshwater fish: how impor- Pyron, R.A. (2014) Biogeographic analysis reveals ancient tant is marine dispersal? Molecular Ecology, 9, 1815–1821. continental vicariance and recent oceanic dispersal in Wiens, J.J. & Donoghue, M.J. (2004) Historical biogeogra- amphibians. Systematic Biology, 63, 779–797. phy, ecology and species richness. Trends in Ecology and Ree, R.H. & Sanmartın, I. (2009) Prospects and challenges Evolution, 19, 639–644. for parametric models in historical biogeographical infer- Wilkinson, M.J., Marshall, L.G. & Lundberg, J.G. (2006) ence. Journal of Biogeography, 36, 1211–1220. River behavior on megafans and potential influences on Ree, R.H. & Smith, S.A. (2008) Maximum likelihood infer- diversification and distribution of aquatic organisms. Jour- ence of geographic range evolution by dispersal, local nal of South American Earth Sciences, 21, 151–172. extinction, and cladogenesis. Systematic Biology, 57,4–14. Willis, S.C., Nunes, M., Montana, C.G., Farias, I.P., Orti, G. Ribeiro, A.C. (2006) Tectonic history and the biogeography & Lovejoy, N.R. (2010) The Casiquiare river acts as a cor- of the freshwater fishes from the coastal drainages of east- ridor between the Amazonas and Orinoco river basins: ern Brazil: an example of faunal evolution associated with biogeographic analysis of the genus Cichla. Molecular Ecol- a divergent continental margin. Neotropical Ichthyology, 4, ogy, 19, 1014–1030. 225–246. Ronquist, F. (1997) Dispersal–vicariance analysis: a new SUPPORTING INFORMATION approach to the quantification of historical biogeography. Systematic Biology, 46, 195–203. Additional Supporting Information may be found in the Rosen, D.E. (1978) Vicariant patterns and historical explana- online version of this article: tion in biogeography. Systematic Biology, 27, 159–188. Appendix S1 Area-dispersal rate matrices. Roxo, F.F., Albert, J.S., Silva, G.S.C., Zawadzki, C.H., Foresti, Appendix S2 Adjacency matrices (area-connectivity constraints). F. & Oliveira, C. (2014) Molecular phylogeny and biogeographic history of the armored Neotropical catfish BIOSKETCH subfamilies Hypoptopomatinae, Neoplecostominae and 9 Otothyrinae (Siluriformes: Loricariidae). PLoS ONE, , Victor Alberto Tagliacollo has had a long-lasting interest e105564. in the phylogenetics and biogeography of Neotropical fresh- Sanmartın, I., van der Mark, P. & Ronquist, F. (2008) Infer- water fishes. He is especially interested in the evolutionary ring dispersal: a Bayesian approach to phylogeny-based and ecological forces underlying the formation of species-rich island biogeography, with special reference to the Canary tropical freshwater ecosystems. Islands. Journal of Biogeography, 35, 428–449. Sempere, T., Herail, G., Oller, J. & Bonhomme, M.G. (1990) Author contributions: V.A.T. and J.S.A. conceived the ideas; Late Oligocene–early Miocene major tectonic crisis and V.A.T. collected the data; V.A.T. and F.F.R. analysed the related basins in Bolivia. Geology, 18, 946–949. data; all authors discussed the data; V.A.T. and J.S.A. led the Sullivan, J.P., Muriel-Cunha, J. & Lundberg, J.G. (2013) writing. Phylogenetic relationships and molecular dating of the major groups of catfishes of the Neotropical superfamily Editor: Jim Provan Pimelodoidea (Teleostei, Siluriformes). Proceedings of the Academy of Natural Sciences of Philadelphia, 162,89–110.

14 Journal of Biogeography ª 2015 John Wiley & Sons Ltd