Vous êtes sur la page 1sur 13

DOI: 10.1111/jbi.



Bayesian analyses detect a history of both vicariance and

geodispersal in Neotropical freshwater fishes

Carolina Barros Machado1 | Pedro Manoel Galetti Jr1 | Ana Carolina Carnaval2

Departamento de Gene tica e Evolucß~ao,
Universidade Federal de S~ao Carlos, S~ao Abstract
Carlos, SP, Brazil Aim: Among the main biogeographical hypotheses explaining the remarkable diver-
Department of Biology, City College of
sity of fishes in the Neotropics is the “palaeogeographical hypothesis”, focusing on
New York and The Graduate Center of the
City University of New York, New York, vicariance, and the “hydrogeological hypothesis”, focusing on geodispersal. Yet while
reflecting different processes, they may result in similar biogeographical patterns.
Correspondence We employed a model-based Bayesian approach to test these alternative hypothe-
Carolina Barros Machado, Departamento de
tica e Evolucß~
ses and determine which shaped the phylogeographical patterns observed in a
Gene ao, Universidade Federal
de S~ao Carlos, S~
ao Carlos, SP, Brazil. group of Neotropical fishes.
Email: carolbioms@gmail.com
Location: South America.
Funding information Taxon: Salminus.
Coordenacß~ao de Aperfeicßoamento de
Methods: We used mitochondrial and nuclear markers to infer phylogenetic rela-
Pessoal de Nıvel Superior; National Science
Foundation, Grant/Award Number: DEB tionships and estimate divergence times among Salminus species, associating the
1120487, 1343578; Conselho Nacional de
results with known geological events. We then employed approximate Bayesian
Desenvolvimento Cientıfico e Tecnolo
Grant/Award Number: CNPq 308385/2014- computation (ABC) to explore changes in population size over time, asking whether
4, 563299/2010-0; Fundacß~ao de Amparo a
vicariance or geodispersal events best explain the phylogeographical signature
Pesquisa do Estado de S~ao Paulo, Grant/
Award Number: BIOTA 2013/50297-0, observed in the data. Because geodispersal captures a few individuals from a paren-
FAPESP 10/52315-7
tal population, which can then expand and lead to a new lineage, we expect to find
Editor: Luiz Rocha genetic signatures of a founder event following population expansion under this
scenario, but not under vicariance.
Results: The analyses suggest that the diversification process in Salminus
began in Upper Miocene, and ABC indicates that it involved both vicariance
and geodispersal events: while a vicariance event better explains the
phylogeographical structure within S. brasiliensis and the genetic patterns of
differentiation between S. sp. Amazon and S. sp. Araguaia, geodispersal
appears to have been the most important event structuring lineages of Salminus
Main Conclusions: Both vicariance and geodispersal signatures were detected in
our biological model, inferring a complex yet realistic demographic history of
Salminus lineages. The correspondence between the ABC results and traditional
phylogeographical interpretations provide further confidence in the models
drawn and tested. This study reinforces the value of applying an ABC
framework in phylogeographical studies, particularly for those interested in testing
alternative and biologically plausible processes underlying similar biogeographical

Journal of Biogeography. 2018;1–13. wileyonlinelibrary.com/journal/jbi © 2018 John Wiley & Sons Ltd | 1

ABC analysis, fish diversification, historical biogeography, phylogeography, Salminus, South

1 | INTRODUCTION expected to be nested within a clade of fishes that occupy adjacent

basins (Hubert et al., 2007).
Multiple hypotheses have been invoked to explain the remarkable Although they reflect different processes, identifying whether
diversity and the complex process of diversification in the Neotropics allopatric sister species diverged through vicariance or geodispersal
(Leite & Rogers, 2013). Yet, most phylogeographical studies to date has been challenging, particularly when new barriers emerge after
have focused on terrestrial organisms (Beheregaray, 2008). Freshwa- headwater capture, or in the face of local extinctions (Burridge,
ter species, through their direct associations with hydrographic Craw, & Waters, 2006; Kodandaramaiah, 2010). Model-based analy-
basins, offer an unmatched opportunity to study the dynamics of the sis such as approximate Bayesian computation (ABC; Beaumont,
Neotropical hydrological landscape and the historical geomorphologi- Zhang, & Balding, 2002) has allowed the distinction between models
cal processes that shaped it (Hubert & Renno, 2006; Hubert et al., involving vicariance and dispersal processes in terrestrial organisms,
2007; Montoya-Burgos, 2003; Sivasundar, Bermingham, & Ortı, resulting in evidences of founder events to distinguish them (Perez,
2001; Tagliacollo, Roxo, Duke-Sylvester, Oliveira, & Albert, 2015). Bonatelli, Moraes, & Carstens, 2016; Prates, Rivera, Rodrigues, &
The major river systems in South America began their develop- Carnaval, 2016). ABC is a flexible class of Monte Carlo algorithms
ment with the breakup of Gondwana (c. 90 million years ago—Ma). that bypass the calculation of exact likelihoods using summary statis-
However, most of them reached their modern configuration recently, tics and simulation. Its use has increased in population genetics and
after being heavily influenced by the middle Miocene-Pleistocene phylogeographical studies, allowing a quantitative evaluation of the
phase of uplift of the Andes (Hoorn, Guerrero, Sarmiento, & Lorente, demographic and evolutionary history by strictly contrasting realistic
1995; Lundberg et al., 1998). Through compressional tectonics, models defined a priori and estimating relevant parameters (Inoue,
Andean orogeny is thought to have promoted both disruption and Lang, & Berg, 2015; Vanhaecke et al., 2015; Willis et al., 2015).
coalescence of drainage systems (Lundberg et al., 1998). The fluctu- Building on recent advances in how to test alternative demo-
ating sea levels and climate changes also largely impacted connec- graphic hypotheses, and on the availability of a user-friendly of ABC
tions across rivers (Albert, Petry, & Reis, 2011). This dynamism of plataform (Cornuet et al., 2014), we explored the use of this analyti-
hydrographic systems is the main promoter of diversification events cal tool to distinguish between phylogeographical signatures of
in freshwater species. vicariance (palaeogeographical hypothesis) versus geodispersal (hy-
Based on the transformations of the hydrological landscape, two drogeological hypothesis) in the evolutionary history of South Ameri-
main hypotheses have been employed to explain the (presumably) can fishes. For that, we focused on the genus Salminus. The
largely allopatric processes of diversification of Neotropical freshwa- monophyly of this group (Machado, Ishizuka, Freitas, Valiati, &
ter species (Hubert & Renno, 2006). On the one hand, the “palaeo- Galetti, 2017), the large migratory capacity of its species (Lima &
geographical hypothesis” has focused on the emergence of physical Britski, 2007), and their allopatric ranges along South America’s
barriers fragmenting an ancestral aquatic area, particularly resulting major hydrographic basins render it a biological model to clarify the
from tectonic readjustment, as major drivers of diversification; stud- role of vicariance and geodispersal processes on the diversification
ies of palaeoarches of the Amazon basin are one such example of South American fishes, and to ask how local landscape changes
(Albert, Lovejoy, & Crampton, 2006). This hypothesis invokes vicari- may have impacted the riverine biota.
ance scenarios related to the formation of impermeable barriers and To study how former landscape changes impacted the distribution
consequent interruption of gene flow, leading to speciation (Ho patterns of lineage diversity in Salminus species, we used mitochon-
et al., 2015). According to Hubert et al. (2007), it also predicts that drial and nuclear markers to reconstruct the phylogenetic relationships
in the taxon-area relationships, the lineages or species occurring in within the genus, including samples from all molecular operational tax-
hydrographic basins separated by geological barriers constitute recip- onomic unit defined previously by Machado et al. (2017). Following
rocally monophyletic clades. On the other hand, the “hydrogeological other phylogeographical studies of South American freshwater sys-
hypothesis” has emphasized headwater capture events between tems (Ramirez, Birindelli, & Galetti, 2017; Tagliacollo et al., 2015; Wil-
basins as an important trigger of colonization of new areas, and sub- lis et al., 2015), we used our trees to estimate divergence times among
sequent differentiation (Hubert & Renno, 2006; Montoya-Burgos, lineages, and tentatively associated them with geological events
2003). This scenario implies in the temporary removal of barriers known to have changed the hydrologic landscape (thus potentially
between adjacent basins, enabling mixing of local assemblages, gene flagging relevant processes in the group diversification). Going beyond
flow and range expansion—a process often referred to as geodisper- that, however, we employed an ABC statistical framework to specifi-
sal (Ho et al., 2015; Lieberman & Eldredge, 1996). Under this scenar- cally identify which biogeographical scenario (palaeogeographical or
io, lineages or species newly formed through geodispersal are hydrogeological hypotheses) better explains the genetic structure

observed within our focus group. Given the loci available, we were 1850]) and two undescribed species (Salminus sp. Amazon and Salmi-
able to investigate which process(es) was(were) influential in structur- nus sp. Araguaia) identified by Machado et al. (2017) (Figure 2, see
ing phylogeographical diversity within Salminus brasiliensis and Salmi- Appendix S1 in Supporting information for more sample details).
nus hilarii, as well as those underscoring the divergence between Partial sequences of three mitochondrial regions (cytochrome C
Salminus sp. Amazon and Salminus sp. Araguaia. oxidase I [COI], cytochrome b [CytB], and D-loop) and two nuclear
Throughout this analysis, we expected that diversification events markers (the recombination activating gene 2 [RAG2] and intron I of
triggered by geodispersal (that is, facilitated by headwater capture the S7 ribosomal protein gene [S7]), were amplified with primers
events that characterize the hydrogeological hypothesis) would leave described in Appendix S2. Polymerase chain reaction (PCR) condi-
a signature of genetic founder events following population expansions tions, PCR products purification, and sequencing were carried out as
in colonized areas. This reasoning is justified by the fact that headwa- described in Machado et al. (2017). All mitochondrial and nuclear
ter capture is expected to isolate a small number of individuals from sequences are available through GenBank (see Appendix S1 for
their parental population (Yeung et al., 2011). This is particularly evi- accession numbers). We also combined available COI and RAG2
dent when widely distributed large and medium sized species that sequences (Machado et al., 2017) with our dataset.
occupy first- and second-order rivers are used as biological model. Sequences were aligned and edited in GENEIOUS R7 6.1.6 (http://
Although many authors propose that the headwater captures can www.geneious.com, Kearse et al., 2012). Because insertion/deletion
involve both geodispersal and vicariance events (Albert et al., 2011), heterozygote individuals were found in S7 gene sequences from S.
we here adopt headwater captures as a geodispersal scenario because brasiliensis, S. hilarii and S. sp. Amazon, hence leading to overlapping
the genetic signature of founder event leaves strong footprint in the peaks, we used CODONCODE ALIGNER 4.2.7 (CodonCode Corporation,
genome. We expected that the severe population reduction associated Dedham, MA, USA) to find and process the heterozygous indels. For
with geodispersal would lead to a reduction in genetic variation that all nuclear sequences, we used GENEIOUS’ plugin Find heterozygotes
can be detected by the ABC analysis (Perez et al., 2016; Prates et al., with a 0.80 threshold to identify heterozygous positions and assign
2016). On the other hand, diversification caused by vicariance, as per ambiguity codes. Then, we reconstructed the haplotypic phase of
the palaeogeographical hypothesis, is not expected to be tied to foun- the heterozygotes with PHASE 2.1 (Stephens & Donnelly, 2003), using
der bottlenecks (Figure 1). To verify which signature is prevalent SEQPHASE to format the input files (Flot, 2010).
within species of Salminus, we explicitly simulated alternative scenar-
ios of temporal shifts in population size concomitant with lineage
2.2 | Phylogenetic inference, divergence time
divergence, and compared them to our observed (empirical) data.
estimation and ancestral area estimation
Before conducting a phylogenetic analysis, we checked for possible
recombination in the nuclear markers using RDP 3.44 (Martin &
Rybicki, 2000) and setting the highest acceptable p value to infer
2.1 | Sampling of molecular data
recombination to 0.01. We also assessed substitution saturation for
We sampled 94 specimens of Salminus across distinct South America all markers through an Index of substitution saturation based on
basins, including representatives of four named species (Salminus affi- entropy in DAMBE (Xia, Xie, Salemi, Chen, & Wang, 2003). Models of
nis [Steindachner, 1880], Salminus brasiliensis [Cuvier, 1816], Salminus nucleotide evolution were detected with JMODELTEST 2.1.4 (Posada,
franciscanus (Lima & Britski, 2007), and Salminus hilarii [Valenciennes, 2008) using the Bayesian information criterion.

F I G U R E 1 Evolutionary scenarios drawn to test palaeogeographic and hydrogeological hypotheses. (1) Palaeogeographical hypothesis
shows a vicariance signature. In this scenario, we assumed one ancestral population not sampled, with effective size Na, that at time t2
suffered split and two new populations were formed with effective size lower than Na (N1 and N2). (2) The hydrogeological hypothesis has a
founder event signature following by population expansion in the receiver basin. We assumed a small number of founders (N1f or N2f, where
“f” means founder), originated from source-basin, colonized the adjacent basin after a geodispersal event, caused by headwater capture, at t2,
followed by population expansion at t1

F I G U R E 2 Sampling sites of Salminus in

South America. Salminus affinis (♦),
Salminus brasiliensis (■), Salminus
franciscanus (★), Salminus hilarii (▲),
Salminus sp. Amazon ( ), and Salminus sp.
Araguaia ( ). Colors represent the
hydrographic basins where samples were

For phylogenetic inference, we used the species tree method ancestor (MRCA), with offset of 11.8 Ma and standard deviation
implemented by the *BEAST tool in BEAST 1.8.1 package (Drummond, 2 Ma (95% of the highest posterior density [HPD] = 7.88 and
Suchard, Xie, & Rambaut, 2012). Brycon specimens were used to root 15.72), uniform prior distribution (interval = 0–1) of the mean rate of
the tree, following Oliveira et al. (2011). Because the D-loop data the molecular clock (ucld.mean parameter), Yule model as species
showed much higher levels of saturation relative to the other mark- tree prior, with an unlinked clock model across markers, and default
ers, precluding our dating analyses from convergence, we opted to setting of substitution rates and nucleotide frequency parameters.
exclude it from the phylogenetic study and the divergence time Three independent Markov chains were started from random
reconstruction, yet kept it in the ABC analysis. The species tree was trees, ran independently for 200 million generations, and sampled at
thus inferred with three independent loci: the two mitochondrial every 10,000 steps. We then combined all runs in LOGCOMBINER 1.8.1
genes combined (COI and CytB), and two nuclear loci (S7 and RAG2). with 25% burn-in, and summarized a maximum clade credibility tree
Because species tree method requires a priori assignment of indi- in TREEANNOTATOR 1.8.1 (BEAST 1.8.1 package). Proper mixing was veri-
viduals to species, we considered as putative species both the mito- fied with TRACER 1.5 (Rambaut & Drummond, 2007), and an effective
chondrial lineages defined by DNA barcoding (Machado et al., 2017) sample size (ESS) of 200 or higher was required for all parameters.
and the basin of occurrence. Thus, we included samples from the To facilitate our understanding about what historical processes
following lineages: S. affinis, S. franciscanus, S. sp. Amazon, S. sp. Ara- are important, mainly in the early evolutionary history of Salminus,
guaia, S. brasiliensis from the Upper Parana basin, S. brasiliensis from we reconstructed ancestral area through the maximum likelihood
the Paraguay and Uruguay basins and the Lagoa dos Patos drainage approach using the “BioGeoBEARS” package (Matzke, 2013) for R
(hereafter named PUL lineage), and three S. hilarii lineages (two lin- (http://www.R-project.org; detailed analysis in Appendix S3). After
eages from S~ao Francisco basin [hereafter named SF1 and SF2] and theses procedures, we compared the divergence times of the species
one from Upper Parana basin). and lineages with know geoclimatic events.
Divergence times were estimated under an uncorrelated lognor-
mal relaxed-clock model in *BEAST. We calibrated the phylogeny
2.3 | Testing biogeographical scenarios by ABC
using a biogeographical event, the uplift of the Eastern Andes Cordil-
lera (Hoorn et al., 1995), which separated the Magdalena and Ama- We used all mitochondrial (COI, CytB and D-loop) and nuclear
zon palaeobasins and is consistent with the divergence of S. affinis (RAG2 and S7) markers in an ABC statistical framework to test
from all other Salminus. The following features were implemented: which biogeographical scenario best matched the current phylogeo-
normal prior distribution of the time of the most recent common graphical structure of the following lineages: S. brasiliensis from the

Upper Paran
a basin, S. brasiliensis from the Paraguay, Uruguay, and the vicariance scenario. This founder event is expected in Salminus
Lagoa dos Patos hydrographic systems (PUL lineage), S. hilarii from species because they occur in first and second-order rivers (Lima &
the Upper Paran
a basin, two monophyletic S. hilarii lineages from the Britski, 2007); a headwater capture event is thus able to move just a
S~ao Francisco basin, and the clade composed of S. sp. Amazon and few individuals.
S. sp. Araguaia. Because we used a small number of loci, we did not In the case of S. hilarii, in which three well-structured mitochon-
expect to find clear, strong signatures of old demographic history drial lineages are observed, we added more parameters to the model
among ancestral populations in deep nodes. Instead, we focused our and modified the scenarios to be tested. Because there are two S~ao
efforts on testing alternative scenarios that may explain recent diver- Francisco lineages, and given that the identity of which one emerged
gence processes. The chosen lineages are ideal to an ABC-based test first is important, the choice of the best scenario was made in two
of alternative biogeographical scenarios because both the vicariance steps. Initially, each biogeographical scenario was tested separately
and the geodispersal hypotheses appear to be plausible explanations (geodispersion with the S~ao Francisco as the source basin, geodis-
for the distribution patterns observed within Salminus, as well as the persion with the Upper Parana as the source basin, and a vicariance
divergence times associated with Salminus lineages. scenario class—Figure S4.2 and results are in Table S4.4,
For the ABC analysis we used the DIYABC 2.0.4 (Do It Yourself Appendix S4). The second round compared the models with highest
ABC—Cornuet et al., 2014). DIYABC considers four categories of posterior probabilities resulting from each biogeographical scenario
events: population divergence, discrete change in effective popula- class calculated in the first round (Figure 3c). Details of the prior dis-
tion size, admixture and sampling (to allow considering samples tribution of parameters, summary statistics, model comparison,
taken at different times). In this study, we evaluated if population parameters estimation and precision, and model checking are
divergence was followed by discrete change in effective population described in Appendix S4.
size to differentiate geodispersal (which implies a founder effect)
from vicariance (no founder effect). Thus, we tested two classes of
biogeographical scenarios independently for each lineage described
above (Figure 1): a geodispersal event with founder effect (simulat-
3.1 | Phylogenetic relationship among Salminus
ing a hydrogeological hypothesis) and a vicariance scenario (simulat-
ing a palaeogeographical hypothesis). For each scenario, present-day
lineages were set at time t0, and divergence was set occur at t2. In The phylogeny recovered S. affinis, the only trans-Andean species, as
the case of the two S. brasiliensis lineages, and of the clade com- an early divergent lineage that is sister to two main clades: one com-
posed of S. sp. Amazon and S. sp. Araguaia, the scenarios were posed of S. sp. Amazon and S. sp. Araguaia (hereafter called the
drawn as follows: the vicariance scenario assumed one (unsampled) Northwest clade), and other comprised of S. brasiliensis, S. francis-
ancestral population with an effective population size of Na, which canus and S. hilarii (hereafter called the Southeast clade; Figure 4). In
split into two populations of effective size lower than Na (N1 and the Northwest clade, two groups of S. sp. Amazon were identified,
N2) at t2. These daughter populations were then simulated to evolve each from unique localities (Upper Amazon and Madeira basins), yet
independently, and at t0 represent different lineages within a named with low support (<70%). For this reason, we pooled data from both
species (in the case of S. brasiliensis), or different species (in the case areas and consider S. sp. Amazon as a single lineage in our biogeo-
of S. sp. Amazon and S. sp. Araguaia). In turn, the geodispersal sce- graphical tests. In the Southeast clade, S. franciscanus was placed as
nario assumed that a small number of founders (N1f or N2f, where the sister group of S. brasiliensis and S. hilarii. The relationship
“f” means founder) originated from a sampled source-basin, colo- between S. brasiliensis and S. hilarii was weakly supported (node 6),
nized the adjacent basin after a headwater capture event at t2, and and the nuclear and mitochondrial markers supported different
underwent population expansion at t1. There are two scenarios topologies regarding this pair of taxa: S. brasiliensis was paraphyletic
within this class because we allowed the source-basin identity to in the mitochondrial tree, while the nuclear intron S7 recovered it as
vary. As with the vicariance scenario class, the populations were sim- a monophyletic species (data not shown).
ulated to evolve independently after the founder event, until t0 (Fig-
ure 3a, b). The major differences between the vicariance and
3.2 | Divergence times within Salminus and
geodispersal scenarios are: (1) an (unsampled) ancestral population is
ancestral area estimation
incorporated in the vicariance scenario, while it does not exist in the
geodispersal scenario given that one daughter population came from Dating analyses suggested that the diversification process in Salminus
the source-basin; (2) changes in the effective size of the daughter began in the Late Miocene (Figure 4). Our biogeographical calibration
populations at t2: while the vicariance scenario does not include sig- places the Salminus MRCA in the proto-Amazon region at about
nificant differences between the size of daughter populations, the 11.05 Ma (95% HPD = 7.02–14.98 Ma). The next divergence event
geodispersal model simulates a severe reduction in the effective suggested by the analysis, between the Northwest clade (Amazon and
population size of the lineage that is putatively colonizing a new area Araguaia basins) and the Southeast clade (La Plata basin—Paraguay,
due to the headwater capture event; (3) demographic expansion is Uruguay, Parana basins—Lagoa dos Patos drainage, and S~ao Francisco
simulated after colonization in geodispersal scenario, but not under basin), places the split at 7.76 Ma (95% HPD = 4.01—11.64 Ma). All

F I G U R E 3 Evolutionary scenarios tested using DIYABC in (a) Salminus brasiliensis, (b) Salminus sp. Amazon and S. sp. Araguaia, and (c)
Salminus hilarii from South America. Colors represent the hydrographic basins where samples were collected. N: effective population size; Nf:
founder effective population size; t: time. Time and effective population size are not strictly to scale

remaining diversification processes among species were placed in the Southeast clade was in the S~ao Francisco basin and later reached
Early Pliocene (Figure 4). We also estimated the diversification time of the other basins of the Brazilian shield (Parana, Paraguay, Uruguay,
the S. brasiliensis lineages (Upper Parana and PUL lineages diverged at and Lagoa dos Patos—Figure S3.1, Appendix S3).
0.36 Ma, 95% HPD = 0.11–0.72 Ma), and S. hilarii lineages (Upper
Parana and S~
ao Francisco lineages split at 1.36 Ma, 95% HPD = 0.51–
3.3 | Testing biogeographical hypotheses with ABC
2.53 Ma, and the sympatric lineages in the S~ao Francisco basin are esti-
mated to have diverged at 0.68 Ma, 95% HPD = 0.17–1.44 Ma; Fig- The demographic reconstruction resulting from the ABC analysis
ure 4). identified vicariance (palaeogeographical hypothesis) as the most
Of the six biogeographical models that we implemented in “Bio- highly supported event to explain the current phylogeographical
GeoBEARS”, DIVALIKEj (Disperal-vicariance plus founder-event spe- structure of S. brasiliensis, and the divergence process between S. sp.
ciation) provided the best statistical fit to our data (AIC weights Amazon and S. sp. Araguaia. In contrast, geodispersal (hydrogeologi-
(>0.90) and see Table S3.3, Appendix S3). The most probable distri- cal hypothesis) was the event more highly supported to explain
bution area of MRCA of all Salminus was in the proto-Amazon diversity within S. hilarii (Table 1).
region. The model also recovered that the MRCA of the Northwest All estimated parameters of the winning scenarios and their rela-
and Southeast clades was present in the palaeo-Amazon (Amazon tive median of the absolute error (RMAE) can be found in Table 2.
and Tocantins-Araguaia basins). The MRCA of the Northwest clade The RMAE values were moderately low, indicating reliable estimation
continued in the palaeo-Amazon where it diverged to form S. sp. of all parameters, except NUPf (0.835) and NSF2 (0.337) estimated
Araguaia and S. sp. Amazonas, while the common ancestor of the from the S. hilarii winning scenario. Because both parameters were

F I G U R E 4 Bayesian phylogenetic analysis and divergence time estimative in BEAST 1.8.1 using two mitochondrial (CytB and COI) and two
nuclear (RAG2 and S7) markers. We used one biogeography event calibration (red circle). Numbers below nodes represent divergence time
estimates, while numbers above nodes represent the posterior probability. Blue bars represent 95% highest posterior probability of the time

not used here, this result did not affect our evaluation about drawn effective population sizes at present were inferred as
and tested scenarios. Winning scenarios for each species showed NAM = 155,000 for S. sp. Amazon and NToA = 30100 for S. sp. Ara-
high type I error rates, yet type II error rates were low (Table S4.6— guaia.
Appendix S4). This provides greater confidence in the best scenario The winning scenario for S. hilarii modeled the Upper Parana
proposed by the ABC analysis. basin as having captured part of the S~ao Francisco basin (source-
The winning scenario for S. brasiliensis inferred the median basin). This headwater capture was inferred to have happened at
effective population sizes for the Upper Parana and PUL lin- t2 = 1.952 Ma. A small number of founders (NUPf = 5,040), origi-
eages as NUP = 85,300 and NPUL = 136,000, respectively. Both lin- nated from SF1 lineage, was recovered as having colonized the
eages were inferred to have diverged from an ancestral population Upper Parana basin and then expanded at 1.592 Ma, t1. Divergence
of median effective population size of 685,000, at t2 = 0.892 Ma. time between the two sympatric lineages in the S~ao Francisco basin
For the clade comprising S. sp. Amazon and S. sp. Araguaia, the was inferred as ts = 0.606 Ma. The median effective population
scenario with highest posterior probability recovered an ancestral sizes of these lineages at present were estimated at NSF1 = 446,000,
population (Na = 582,000) splitting at t2 = 2.36 Ma. Median NSF2 = 441,000, and NUP = 460,000.

T A B L E 1 Bayesian posterior probabilities for each tested scenario estimated using DIYABC for Salminus brasiliensis, Salminus sp. Amazon and
S. sp. Araguaia, and Salminus hilarii. The highest posterior probability is in bold. Colors in the winner scenario column represent the
hydrographic basins where samples were collected.
Posterior probability (95% CI) Winner scenario
Salminus brasiliensis
Scenario 1- Vicariance event 0.4255 (0.4184-0.4325)
Scenario 2- Paraguay and Uruguay basins, and Lagoa dos 0.3566 (0.3497-0.3634)
Patos drainage captures part of Upper Parana basin
(geodispersal event)
Scenario 3- Upper Parana basin captures part of Paraguay 0.2180 (0.2117-0.2242)
and Uruguay basins, and Lagoa dos Patos drainage
(geodispersal event)
Salminus sp. Amazon and Salminus sp. Araguaia
Scenario 1- Vicariance event 0.5044 (0.4929-0.5158)
Scenario 2- Tocantins-Araguaia basin captures part of 0.2063 (0.1975-0.2151)
Amazon basin (geodispersal event)
Scenario 3- Amazon basin captures part of Tocantins-Araguaia 0.2893 (0.2788-0.2998)
basin (geodispersal event)
Salminus hilarii (Round 2)
Scenario 1 - Upper Parana and S~ao Francisco_2 lineages 0.2695 (0.2376-0.3013)
diverge from ancestral population (Vicariance event)
Scenario 2 - Upper Parana basin captures part of S~ao Francisco 0.6596 (0.6212-0.6981)
(SF1) (geodispersal event)
Scenario 3 - S~ao Francisco (SF2) basin captures part of Upper 0.0709 (0.0000-0.1431)
Parana basin (geodispersal event)

from that study, built from sequences of 16SrRNA, Cytochrome B,

recombination activating gene 1 and 2, and the myosin heavy chain
6 cardiac muscle. Among those molecular markers, only RAG2
4.1 | Salminus phylogeny
showed a distinct topology, placing S. hilarii as the early divergent
Our multiloci phylogenetic reconstruction of Salminus detected the species. Gene trees built with all other molecular markers utilized in
intraspecific structure and solved most interspecific relationships the Abe et al. (2014) analysis, as well as with those used in the pre-
within the group. All nodes were strongly supported, except for sent study, recovered S. hilarii as phylogenetically closer to S. francis-
the relationship between S. brasiliensis and S. hilarii (node 6, canus and S. brasiliensis. Given that inconsistency, we ran a new
Figure 4). Poor resolution in that portion of the tree may result Bayesian Inference analysis with RAG2 sequences of all Bryconidae
from incongruence among the mitochondrial and nuclear markers, species available in GenBank (Abe et al., 2014; Oliveira et al., 2011),
which is commonly observed in other organisms and expected also including our own RAG2 dataset (data not shown). All Salminus
under the coalescent theory (Davalos, Cirranello, Geisler, & Sim- sequences except by one—the S. hilarii specimen used by Abe et al.
mons, 2012). (2014)—formed a monophyletic group with strong support. That S.
The inferred tree (Figure 4) conflicts with two other phylogenies hilarii sequence, in turn, clustered with Brycon specimens, suggesting
previously proposed for the group, which were based on morpholog- that it had been mislabeled at some point of the laboratory work
ical (F.C.T. Lima, personal communication) and molecular data (Abe, that led to the original Abe et al. (2014) study. This finding indicates
Mariguela, Avelino, Foresti, & Oliveira, 2014). Our analysis strongly that mistake led to a misinterpretation of the real evolutionary rela-
supports S. affinis as a sister species to all remaining taxa. In con- tionship among Salminus species; the molecular phylogeny obtained
trast, both previous phylogenies recovered S. affinis as sister to S. sp. here is coherent.
(a species composed of S. sp. Amazon and S. sp. Araguaia). More-
over, Abe et al. (2014) recovered S. hilarii as the early divergent spe-
4.2 | Possible reconstructions of the early
cies and sister of all remaining Salminus, a scenario completely
evolutionary history of Salminus
different from ours results.
To understand the reasons why Abe et al.’s (2014) molecular Our dating analysis placed the MRCA of all species of Salminus in the
data led to such distinct conclusions, we re-analysed the gene trees late Miocene (11.05 Ma, 95% HPD = 7.02–14.98 Ma; node 1—

T A B L E 2 Posterior parameter values for scenarios with high probability (winners) in Salminus brasiliensis, Salminus sp. Amazon and S. sp.
Araguaia, and Salminus hilarii estimated using DIYABC. The unit of time, counted in generation, was converted into 106 years before present
(Ma), based in an average generation time of 2 years (de Godoy, 1975)a
Posterior parameter estimation

Parameters Median 95% CI RMAE

Salminus brasiliensis
PUL lineage effective population size NPUL 136,000 61,400–332,000 0.265
Upper Parana lineage effective population size NUP 85,300 37,900–214,000 0.247
Ancestral effective population size Na 685,000 269,000–970,000 0.119
Time of divergence between PUL and Upper Parana lineages t2 0.892 Ma 0.416–1.17 Ma 0.269
Salminus sp. Amazon and Salminus sp. Araguaia
Salminus sp. Araguaia effective population size NToA 30,100 12,300–115,000 0.288
Salminus sp. Amazon effective population size NAm 155,000 57,800–442,000 0.279
Ancestral effective population size Na 582,000 204,000–956,000 0.136
Time of divergence between Salminus sp. Araguaia and Salminus sp. Amazon t2 2.36 Ma 1.178–4.54 Ma 0.190
Salminus hilarii
S~ao Francisco_1 lineage effective population size NSF1 446,000 156,000–882,000 0.267
S~ao Francisco_2 lineage effective population size NSF2 441,000 135,000–907,000 0.337
Upper Parana lineage effective population size NUP 460,000 179,000–832,000 0.254
Founder effective population size into Upper Parana basin NUPf 5,040 358–9,740 0.835
Time of divergence between S~ao Francisco_1 and S~ao Francisco_2 lineages ts 0.606 Ma 0.216–1.086 Ma 0.272
Time of expansion for Upper Parana lineage t1 1.592 Ma 0.77–2.58 Ma 0.255
Time of founder event between S~ao Francisco and Upper Parana lineages t2 1.952 Ma 1.046–3.24 Ma 0.135

N: effective population size; t: time in years; Nf: founder effective population size; RMAE: relative median of absolute error.
See Appendix S4.

Figure 4), hence prior to the uplift of the Eastern Andean Cordillera. Two possible palaeo-routes of geodispersal have been proposed
At that time, the modern hydrographic basins northwest of South between the Amazon/Tocantins-Araguaia and the S~ao Francisco
America were not individualized, and the region was covered by a palaeobasins. One of them involves the Serra Geral do Tocantins
lagoon system known as the Pebas lake (Lundberg et al., 1998). It is (SGT), a tectonic structure that dates to the Cretaceous and which
likely that the Salminus MRCA was then distributed across all proto- stands between the two basins, acting as physical barrier to their
Amazon, and that the appearance of the Eastern Cordillera divided headwaters (Lima & Caires, 2011). It has been proposed that former
ancestral populations into the Magdalena and Amazon/Tocantins-Ara- erosion processes in the SGT resulted in a contact area where both
guaia palaeobasins—with the population from Magdalena palaeobasin basins share their headwaters, creating a potential for faunal inter-
leading to the modern S. affinis, while that in the Amazon palaeobasin change (Lima & Caires, 2011). Another possible geodispersal palaeo-
becoming the MRCA of the remaining species. We were unable to use route between these two hydrographic systems may have been
ABC to test whether this vicariance scenario was more strongly sup- established through a Northern-Northeastearn Brazil coastal corridor.
ported than the geodispersal event. However, vicariant reconstruc- According to Hubert et al. (2007), two geodispersal events involving
tions as the one discussed above have been widely employed to these basins occurred in the Serrasalmus genus, and one of them
explain cis- and trans-Andean ichthyofauna divergence (Albert et al., coincides with our estimated date. Hubert et al. (2007) proposed
2006; Montoya-Burgos, 2003; Ramirez et al., 2017; Sivasundar et al., that Miocene marine regressions promoted coalescence at the
2001). mouths of several rivers, creating transient connections between the
The data suggests that the next cladogenetic event within Salmi- palaeobasins of the Amazon/Tocantins-Araguaia and the Brazilian
nus resulted in the isolation of a Southeastern clade species from Northeast. Although speculative, this event may have allowed the
the Northwestern clade (node 2—Figure 4; Figure S3.1, Appendix 3). MRCA of the Southeast clade of Salminus to reach the Brazilian
The MRCA of this Southeast+Northwest clade was estimated at Shield. Hubert and Renno (2006) also implied this route in historical
approximately 7.76 Ma (95% HPD = 4.01–11.64 Ma), and likely pre- reconstructions of other Characiformes.
sent in the Amazon/Tocantins-Araguaia. Species distribution pat- Within the Southeast clade, the cladogenetic events involving
terns, tied to the inferred order of events, suggests that an ancestral the Upper Parana and S~ao Francisco basins were not consistent with
group of individuals colonized the Brazilian Shield basins, probably speciation by vicariance but by successive geodispersal, because the
through a headwater capture event to the S~ao Francisco basin, and hydrographic basins do not form monophyletic clades (Hubert et al.,
gave rise to Southeast clade. 2007). In fact, the geological structure of the Brazilian Shield is

conducive to headwater captures because it has several geological event represented the break-up of a connection between the Amazon
faults that are continually activated, leading to rearrangements and Tocantins-Araguaia palaeobasins. Geological and geomorphological
among adjacent basins and consequent faunistic interchange data point out that the intense tectonic reactivation of the geological
(Buckup, 2011; Ribeiro, 2006; Ribeiro & Menezes, 2015). The faults located in the Lower Amazon basin changed the course of the
appearance of both hydrographic systems is very old and often asso- Tocantins palaeoriver during the Plio-Pleistocene: the ancient Tocan-
ciated with the Gondwana break-up (Potter, 1997; Ribeiro, 2006). tins river discharged on the opposite course of the modern drainage,
Between their adjacent headwaters there is a geological fault system into the Amazon palaeoriver (Rossetti & Valeriano, 2007). The interrup-
denominated Upper S~ao Francisco River Crustal Discontinuity tion of direct discharge of the Tocantins palaeoriver into the Amazon
(DCARSF), which has been in tectonic activity since the Pre-Cam- palaeoriver was also suggested as the vicariance event responsible to
brian (Buckup, 2011). Due to continuous DCARSF reactivations, it lead speciation between Inia araguaiaensis and Inia geofrensis at
thus appears that headwater captures became important diversifica- 2.08 Ma (Hrbek et al., 2014).
tion events for Salminus, Hyspostomus (Montoya-Burgos, 2003) and Within the Southeastern clade, the ABC-based test supported
some Characiformes (Hubert & Renno, 2006). Our analyses placed vicariance as the main driver of diversification across lineages of S.
the MRCA of the Southeast clade in S~ao Francisco basin nearly at brasiliensis, while geodispersal was responsible for diversification
4.28 Ma (95% HPD = 1.89–7.24 Ma), when part of the lineage gave across lineages of S. hilarii. S. brasiliensis showed two well-defined
rise to S. franciscanus and part of it geodispersed to the Upper allopatric lineages (Upper Parana and PUL), and their diversification
Parana basin. The latter appears as the probable ancestral area for process was likely driven by the split of the hydrographic basins
the MRCA of S. brasiliensis and S. hilarii (node 5, Figure 4; Fig- where both occurred. Other co-distributed species also have
ure S3.1). marked genetic structure between the Upper Parana basin and the
Although we were unable to use the ABC approach to infer the ıcan,
remaining hydrographic systems of the La Plata basin (Pialek, R
demographic history of the early Salminus divergences, the estimated  n, & Zrzavy
Casciotta, Almiro , 2012; Sivasundar et al., 2001). In fact,
divergence times are in close agreement with previous studies the Upper Parana basin is considered a distinct ecoregion relative
(Albert et al., 2006; Hubert & Renno, 2006; Montoya-Burgos, 2003; to the other areas here considered (Abell et al., 2008). Its biogeo-
Ramirez et al., 2017; Sivasundar et al., 2001). graphical history is complex, and it is associated with successive
reactivation of geological faults during the Quaternary (Bartorelli,
2004). These neotectonic activities promoted drainage rearrange-
4.3 | ABC, vicariance vs. geodispersion, and the
ments (Ribeiro & Menezes, 2015) and the formation of geographical
recent diversification history of Salminus
, Iguacßu, Mocona and Sete Quedas falls, for example;
barriers (Apipe
Phylogeographical studies of the Neotropical ichthyofauna usually Bartorelli, 2004).
provide estimates of divergence times and correlate them with The best supported scenario in the DIYABC analysis supports an
known historical events. However, due to the complexity of histori- ancestral S. brasiliensis population that was widely distributed across
cal events, similar biogeographical patterns often can emerge from all the La Plata basin, followed by the appearance of an effective bar-
distinct diversification processes (Moritz, Patton, Schneider, & Smith, rier that promoted allopatric divergence between lineages approxi-
2000)—attesting the importance of incorporating statistical infer- mately 0.89 Ma (95% HPD = 0.416–1.17 Ma). The divergence time
ences into hypothesis testing. Among the few studies using this between the lineages of S. brasiliensis seems consistent with the Sete
approach in South America freshwater fishes are the reconstruction Quedas falls appearance. Prior to its inundation, caused by the con-
of potential historic colonization routes for Aplochiton zebra species struction of the Itaipu Reservoir construction, the Sete Quedas falls
(Vanhaecke et al., 2015), an investigation of the performance of dis- was an important impermeable barrier between the Upper Parana
tinct divergence pattern resulted by nuclear and mitochondrial mark- basin and the remaining hydrographic basins that composed the La
ers in Cichla temensis lineages (Willis et al., 2015), and the Plata. The absolute age of the falls is unknown. According to Bar-
identification of headwater captures events that promoted the diver- torelli (2004), tectonic reactivation events happening during the Med-
sification in species of the Pimelodidae family in Amazon West ium Pleistocene and until the Maximum Last Glacial (1.5–0.1 Ma)
(Tagliacollo et al., 2015). Our study is the first to implement a were responsible for the genesis of several falls in the Upper Parana
model-based Bayesian framework to test alternative biogeographical basin, including Sete Quedas. This vicariant event is also used to
scenarios potentially responsible for the diversification process in explain diversification process in Prochilodus lineatus (Sivasundar
Neotropical freshwater fishes. et al., 2001) and Crenicichla lacustris group (Pialek et al., 2012).
Within the Northwestern clade, our phylogenetic analysis sug- Allopatric divergence processes involving species in the S~ao
gested a cladogenetic event to have happened at 2.7 Ma (95% Francisco and the Upper Parana basins have been previously associ-
HPD = 0.98–5.17 Ma), splitting S. sp. Amazon and S. sp. Araguaia ated with geodispersal events promoted by successive headwater
(node 3, Figure 4). The ABC analysis supported a vicariance scenario as captures between basins (Hubert et al., 2007). Our ABC analysis
the most probable event responsible for the divergence between these supports this interpretation: there is strong support for a geodisper-
species that happened at 2.36 Ma (95% HPD = 1.178–4.54 Ma). This sal event being responsible for the diversification of the lineages of

S. hilarii in the Upper Parana and S~ao Francisco basins. The most Natural History) for their insights and discussions of the methods
probable scenario recovered the S~ao Francisco hydrographic system and results of this paper. This research was funded by Coordenacßa~o
as the source-basin in the first headwater capture event (not the de Aperfeicßoamento de Pessoal de Nıvel Superior (CAPES), Fundacß~ao
Upper Parana basin, as observed in species tree-node 8, Figure 4; de Amparo a Pesquisa do Estado de S~ao Paulo (FAPESP 10/52315-
Figure S3.1). The chosen scenario recovered that part of S~ao Fran- 7) and Conselho Nacional de Desenvolvimento Cientıfico e Tec-
cisco lineage (SF1) was captured by the Upper Parana basin approxi-  gico (CNPq 563299/2010-0; 308385/2014-4). A. Carnaval
mately at 1.95 Ma. Although it was possible to estimate the acknowledges support by the National Science Foundation (DEB
divergence time between sympatric lineages (0.606 Ma, 95% HPD = 1120487, 1343578), FAPESP (BIOTA 2013/50297-0) and NASA.
0.216–1.086 Ma, as per the ABC analysis, 0.67 Ma, 95% HPD = This study was developed in accordance with the Brazilian laws,
0.17–1.44 Ma, as per *BEAST), their underlying diversification pro- approved by the ethic committee on animal use at Universidade
cess remains unclear. Federal de S~ao Carlos (CEUA number 5765010416), and carried out
under a temporally scientific collection license number 32217-4 pro-
vided by ICMBio-SISBIO. We also thank the anonymous reviewers
5 | CONCLUSION and Editor for all comments and suggestions, which were very help-
ful in improving the final manuscript.
Through an ABC analysis, we were able to build and distinguish phy-
logeographical signatures of two main events that promote diversifi-
cation in freshwater fishes—namely, vicariance and geodispersal. The ORCID

hydrological hypothesis, detected through signatures of founder Carolina Barros Machado http://orcid.org/0000-0001-5195-5577
events in colonized areas following by expansion population size,
represents a geodispersal event. In contrast, vicariance events failed
to leave signatures of founder events, source-basin identity, and REFERENCES
expansion event. Detection of these sequence of events in the
Abe, K. T., Mariguela, T. C., Avelino, G. S., Foresti, F., & Oliveira, C.
geodispersal was only possible because the biological model (large
(2014). Systematic and historical biogeography of the Bryconidae
and medium-sized migratory fishes) used in this study is appropriate. (Ostariophysi: Characiformes) suggesting a new rearrangement of its
Had we studied fishes restricted to headwaters, in which most of genera and an old origin of Mesoamerican ichthyofauna. BMC Evolu-
the population may be moved in geographical space through head- tionary Biology, 14, 152. https://doi.org/10.1186/1471-2148-14-152
Abell, R., Thieme, M. L., Revenga, C., Bryer, M., Kottelat, M., Bogutskaya,
water captures, this unique signature left by a founder event would
N., . . . Petry, P. (2008). Freshwater ecoregions of the world: A new
likely be absent. This investigation is thus unique in Neotropical map of biogeographic units for freshwater biodiversity conservation.
freshwater species diversification studies: for the first time, at the BioScience, 58, 403–414. https://doi.org/10.1641/B580507
genetic level, vicariance and geodispersal events were distinguished Albert, J. S., Lovejoy, N. R., & Crampton, W. G. R. (2006). Miocene tec-
by phylogeographical signatures. tonism and the separation of cis- and trans-Andean river basins: Evi-
dence from Neotropical fishes. Journal of South American Earth
As with any approach to data analysis, phylogeographical model
Sciences, 21, 14–27. https://doi.org/10.1016/j.jsames.2005.07.010
choice using ABC has limitations, and decisions about which models to Albert, J. S., Petry, P., & Reis, R. E. (2011). Major biogeographic and phy-
include in the comparison set can be challenging. Overall, ABC loses logenetic patterns. In J. S. Albert & R. E. Reis (Eds.), Historical biogeog-
 & Car-
power to differentiate among large numbers of models (Thome raphy of Neotropical freshwater fishes (pp. 21–17). Berkeley: University
of California Press. https://doi.org/10.1525/california/
stens, 2016), similar candidate models increased the failure to recover
the true model (Cabrera & Palsbøll, 2017), and the choice insufficient Bartorelli, A. (2004). Origem da Grandes Cachoeiras do Planalto Basaltico
summary statistics led to a loss of information that can bias the calcu- da Bacia do Paran a: Evolucß~ao Quaternaria e Geomorfologia. In V.
lation of the relative posterior probability (Pelletier & Carstens, 2014). Mantesso-Neto, A. Bartorelli, C. D. R. Carneiro, & B. B. Brito-Neves
(Eds.), Geologia do Continente Sul-Americano: Evolucß~ ao da Obra de Fer-
None of these affected our simulations. Results obtained through tra-
nando Fla vio Marques de Almeida (pp. 95–111). S~ ao Paulo: Editora
ditional phylogeographical analyses (divergence time estimations, fol- Beca Producßo ~ es Culturais.
lowed by associations with known historical events) were consistent Beaumont, M. A., Zhang, W., & Balding, D. J. (2002). Approximate Baye-
with the model-based Bayesian approach. This is reassuring, and sian computation in population genetics. Genetics, 162, 2025–2035.
Beheregaray, L. B. (2008). Twenty years of phylogeography: The state of
speaks of the reliability and robustness of the biogeographical scenar-
the field and the challenges for the Southern Hemisphere. Molecular
ios here inferred and reinforces the value of an ABC framework in Ecology, 17, 3754–3774.
phylogeographical studies, particularly those interested in testing Buckup, P. A. (2011). The Eastern Brazilian shield. In J. S. Albert & R. E.
alternative processes underlying similar biogeographical patterns. Reis (Eds.), Historical biogeography of Neotropical freshwater fishes (pp.
203–210). Berkeley: University of California Press. https://doi.org/10.
ACKNOWLEDGEMENTS Burridge, C. P., Craw, D., & Waters, J. M. (2006). River capture, range
expansion, and cladogenesis: The genetic signature of freshwater
We are thankful to Dr. Jorge Ramirez (Universidade Federal de S~ao vicariance. Evolution, 60, 1038–1049. https://doi.org/10.1111/j.0014-
Carlos) and Dr. Ivan Prates (Smithsonian’s National Museum of 3820.2006.tb01181.x

Cabrera, A. A., & Palsbøll, P. J. (2017). Inferring past demographic changes observacßo ~ es sobre as implicacßo ~ es biogeograficas das “aguas emen-
from contemporary genetic data: A simulation-based evaluation of the dadas” dos rios Sap~ ao e Galheiros. Biota Neotropica, 11, 1–20.
ABC methods implemented in DIYABC. Molecular Ecology Resources, Lundberg, J. G., Marshall, L. G., Guerrero, J., Horton, B., Malabarba, M. C.
17 (6) e94–e110. https://doi.org/10.1111/1755-0998.12696 S. L., & Wesselingh, F. (1998). The stage for Neotropical fish diversifi-
Cornuet, J. M., Pudlo, P., Veyssier, J., Dehne-Garcia, A., Gautier, M., cation: A history of tropical South American Rivers. In L. R. Mal-
Leblois, R., . . . Estoup, A. (2014). DIYABC v2.0: A software to make abarba, R. E. Reis, R. P. Vari, Z. M. S. Lucena, & C. A. S. Lucena (Eds.),
approximate Bayesian computation inferences about population his- Phylogeny and classification of Neotropical fishes (pp. 13–48). Porto
tory using single nucleotide polymorphism. DNA sequence and Alegre, Brazil: Edipucrs.
microsatellite data. Bioinformatics, 30, 1187–1189. https://doi.org/10. Machado, C. B., Ishizuka, T. K., Freitas, P. D., Valiati, V. H., & Galetti, P.
1093/bioinformatics/btt763 M. Jr (2017). DNA barcoding reveals taxonomic uncertainty in Salmi-
Davalos, L. M., Cirranello, A. L., Geisler, J. H., & Simmons, N. B. (2012). nus (Characidae). Systematics and Biodiversity, 15, 372–382. https://d
Understanding phylogenetic incongruence: Lessons from phyllostomid oi.org/10.1080/14772000.2016.1254390
bats. Biological Reviews, 87, 991–1024. https://doi.org/10.1111/j. Martin, D., & Rybicki, E. (2000). RDP: Detection of recombination
1469-185X.2012.00240.x amongst aligned sequences. Bioinformatics, 15, 562–563.
Drummond, A. J., Suchard, M. A., Xie, D., & Rambaut, A. (2012). Bayesian Matzke, N. J. (2013). Probabilistic historical biogeography: New models for
phylogenetics with BEAUti and the BEAST 1.7. Molecular Biology and founder event speciation, imperfect detection, and fossils allow improved
Evolution, 29, 1969–1973. https://doi.org/10.1093/molbev/mss075 accuracy and model-testing. Frontier of Biogeography, 5, 242–248.
Flot, J.-F. (2010). SeqPHASE: A web tool for interconverting PHASE Montoya-Burgos, J. I. (2003). Historical biogeography of catfish genus
input/output files and FASTA sequence alignments. Molecular Ecology Hypostomus (Siluriformes: Loricariidae), with implications on the
Resource, 10, 162–166. https://doi.org/10.1111/j.1755-0998.2009. diversification of Neotropical ichthyofauna. Molecular Ecology, 12,
02732.x 1855–1867. https://doi.org/10.1046/j.1365-294X.2003.01857.x
Ho, S. Y. W., Tong, K. J., Foster, C. S. P., Ritchie, A. M., Lo, N., & Crisp, Moritz, C., Patton, J. L., Schneider, C. J., & Smith, T. B. (2000). Diversifi-
M. D. (2015). Biogeographic calibrations for the molecular clock. Biol- cation of rainforest faunas: An integrated molecular approach. Annual
ogy Letters, 11, 20150194. https://doi.org/10.1098/rsbl.2015.0194 Review of Ecology and Systematics, 31, 533–563. https://doi.org/10.
Hoorn, C., Guerrero, J., Sarmiento, G. A., & Lorente, M. A. (1995). 1146/annurev.ecolsys.31.1.533
Andean tectonic as a cause for changing drainage patterns in Mio- Oliveira, C., Avelino, G. S., Abe, K. T., Mariguela, T. C., Benine, R. C., Ortı, G.,
cene northern South America. Geology, 23, 237–240. . . . Castro, R. M. C. (2011). Phylogenetic relationships within the speciose
Hrbek, T., da Silva, V. M. F., Dutra, N., Gravena, W., Martin, A. R., & Far- Family Characidae (Teleostei: Ostariophysi: Characiformes) based on mul-
ias, I. P. (2014). A new species of river Dolphin from Brazil or: How tilocus analysis and extensive ingroup sampling. BMC Evolutionary Biology,
little do we know our biodiversity. PLoS ONE, 9, e83623. https://doi. 11, 275. https://doi.org/10.1186/1471-2148-11-275
org/10.1371/journal.pone.0083623 Pelletier, T. A., & Carstens, B. C. (2014). Model choice for phylogeo-
Hubert, N., Duponchelle, F., Nun ~ez, J., Garcia-Davila, C., Paugy, D., & graphic inference using a large set of models. Molecular Ecology, 23,
Renno, J. F. (2007). Phylogeography of the piranha genera Ser- 3028–3043. https://doi.org/10.1111/mec.12722
rasalmus and Pygocentrus: Implications for the diversification of the Perez, M. F., Bonatelli, I. A. S., Moraes, E. M., & Carstens, B. C. (2016).
Neotropical ichthyofauna. Molecular Ecology, 16, 2115–2136. Model-based analysis supports interglacial refugia over long-dispersal
https://doi.org/10.1111/j.1365-294X.2007.03267.x events in the diversification of two South American cactus species.
Hubert, N., & Renno, J. F. (2006). Historical biogeography of South Heredity, 116, 550–557. https://doi.org/10.1038/hdy.2016.17
American freshwater fishes. Journal of Biogeography, 33, 1414–1436. Pi
alek, L., Rıcan, O., Casciotta, J., Almiro n, A., & Zrzav
y, J. (2012). Multilo-
https://doi.org/10.1111/j.1365-2699.2006.01518.x cus phylogeny of Crenicichla (Teleostei: Cichlidae), with biogeography
Inoue, K., Lang, B. K., & Berg, D. J. (2015). Past climate change drives of the C. lacustris group: Species flocks as a model for sympatric spe-
current genetic structure of an endangered freshwater mussel spe- ciation in rivers. Molecular Phylogenetics and Evolution, 62, 46–61.
cies. Molecular Ecology, 24, 1910–1926. https://doi.org/10.1111/mec. https://doi.org/10.1016/j.ympev.2011.09.006
13156 Posada, D. (2008). jModelTest: Phylogenetic model averaging. Molecular
Kearse, M., Moir, R., Wilson, A., Stones-Havas, S., Cheung, M., Sturrock, Biology and Evolution, 25, 1253–1256. https://doi.org/10.1093/molbe
S., . . . Drummond, A. (2012). Geneious Basic: An integrated and v/msn083
extendable desktop software platform for the organization and analy- Potter, P. E. (1997). The Mesozoic and Cenozoic paleodrainage of South
sis of sequence data. Bioinformatics, 28, 1647–1649. https://doi.org/ America: A natural history. Journal of South American Earth Sciences,
10.1093/bioinformatics/bts199 10, 331–344. https://doi.org/10.1016/S0895-9811(97)00031-X
Kodandaramaiah, U. (2010). Use of dispersal-vicariance analysis in bio- Prates, I., Rivera, D., Rodrigues, M. T., & Carnaval, A. C. (2016). A mid-
geography – a critique. Journal of Biogeography, 37, 3–11. Pleistocene rainforest corridor enabled synchronous invasions of the
Leite, R. F., & Rogers, D. S. (2013). Revisiting Amazonian phylogeography: Atlantic Forest by Amazonian anole lizards. Molecular Ecology, 25,
Insights into diversification hypotheses and novel perspectives. 5174–5186. https://doi.org/10.1111/mec.13821
Organisms, Diversity & Evolution, 13, 639–664. https://doi.org/10. Rambaut, A., & Drummond, A. J. (2007). Tracer v1.4. Retrieved from:
1007/s13127-013-0140-8 http://tree.bio.ed.ac.uk/software/tracer/
Lieberman, B. S., & Eldredge, N. (1996). Trilobite biogeography in the Ramirez, J. L., Birindelli, J. L., & Galetti, P. M. (2017). A new genus of
Middle Devonian: Geological process and analytical methods. Anostomidae (Ostariophysi: Characiformes): Diversity, phylogeny and
Paleobiology, 22, 66–79. https://doi.org/10.1017/S009483730001 biogeography based on cytogenetic, molecular and morphological
602X data. Molecular Phylogenetics and Evolution, 107, 308–323. https://d
Lima, F. C. T., & Britski, H. A. (2007). Salminus franciscanus, a new species oi.org/10.1016/j.ympev.2016.11.012
from the rio S~ ao Francisco basin, Brazil (Ostariophysi: Characiformes: Ribeiro, A. C. (2006). Tectonic history and the biogeography of the fresh-
Characidae). Neotropical Ichthyology, 5, 237–244. https://doi.org/10. water fishes from the coastal drainages of eastern Brazil: An example
1590/S1679-62252007000300001 of faunal evolution associated with a divergent continental margin.
Lima, F. C. T., & Caires, R. A. (2011). Peixes da Estacß~ao Ecolo  gica Serra Neotropical Ichthyology, 4, 225–246. https://doi.org/10.1590/S1679-
Geral do Tocantins, bacias dos rios Tocantins e S~ao Francisco, com 62252006000200009

Ribeiro, A. C., & Menezes, N. A. (2015). Phylogenetic relationship of the

species and biogeography of the characid genus Oligosarcus Gu €nther, BIOSKETCH
1864 (Ostariophysi, Characiformes, Characidae). Zootaxa, 3949, 041–
081. https://doi.org/10.11646/zootaxa.3949.1.2 Carolina Machado is developing her Postdoc project at Universi-
Rossetti, D. F., & Valeriano, M. M. (2007). Evolution of the lowest ama- dade Federal de S~ao Carlos. She studies phylogenetics and phy-
zon basin modeled from the integration of geological and SRTM logeography of Neotropical freshwater fishes. Her aim is to
topographic data. Catena, 70, 253–265. https://doi.org/10.1016/j.ca
know which processes (ecological and/or historical) are responsi-
Sivasundar, A., Bermingham, E., & Ortı, G. (2001). Population structure ble for actual ichthyofauna distribution pattern in South America.
and biogeography of migratory freshwater fishes (Prochilodus: Pedro Galetti is professor at Universidade Federal de S~ao Carlos.
Characiformes) in major South American Rivers. Molecular Ecology, His research is focused on Genetic Conservation, involving aqua-
10, 407–417.
tic and terrestrial animals. Studies related to forensic genetic,
Stephens, M., & Donnelly, P. (2003). A comparison of Bayesian methods
for haplotype reconstruction. The American Journal of Human Genet- systematic, taxonomy phylogeny and phylogeography are also
ics, 73, 1162–1169. https://doi.org/10.1086/379378 part of his research scope.
Tagliacollo, V. A., Roxo, F. F., Duke-Sylvester, S. M., Oliveira, C., & Albert, Ana Carnaval is assistant professor at City University of New York
J. S. (2015). Biogeographical signature of river capture events in
and researcher associate at The American Museum of Natural His-
Amazonian lowlands. Journal of Biogeography, 42, 2349–2363.
https://doi.org/10.1111/jbi.12594 tory. She studies spatial patterns of biodiversity and their underly-
Thome , M. T. C., & Carstens, B. C. (2016). Phylogeographic model selec- ing evolutionary and ecological processes, with the explicit aim of
tion leads to insight into the evolutionary history of four-eyed frogs. improving diversity prediction and conservation in tropical regions.
Proceedings of the National Academy of Sciences, 113, 8010–8017.
https://doi.org/10.1073/pnas.1601064113 Author contributions: C. B. M., P. M. G. J. and A. C. C. conceived
Vanhaecke, D., Garcia de Leaniz, C., Gajardo, G., Dunham, J., Giannico,
the ideas; C. B. M. collected and analysed the data; C. B. M. led
G., & Consuegra, S. (2015). Genetic signatures of historical dispersal
of fish threatened by biological invasions: The case of galaxiids in the writing with assistance from P. M. G. J. and A. C. C.
South America. Journal of Biogeography, 42, 1942–1952. https://doi.
Willis, S. C., Winemiller, K. O., Montan ~a, C. G., Macrander, J., Reiss, P.,
Farias, I. P., & Ortı, G. (2015). Population genetics of the speckled SUPPORTING INFORMATION
peacock bass (Cichla temensis), South America’s most important inland
sport fishery. Conservation Genetics, 16, 1345–1357. https://doi.org/ Additional Supporting Information may be found online in the sup-
10.1007/s10592-015-0744-y porting information tab for this article.
Xia, X., Xie, Z., Salemi, M., Chen, L., & Wang, Y. (2003). An index of
substitution saturation and its application. Molecular Phylogenetics
and Evolution, 26, 1–7. https://doi.org/10.1016/S1055-7903(02)
00326-3 How to cite this article: Machado CB, Galetti Jr PM, Carnaval
Yeung, C. K., Tsai, P. W., Chesser, R. T., Lin, R. C., Yao, C. T., Tian, X. H.,
AC. Bayesian analyses detect a history of both vicariance and
& Li, S. H. (2011). Testing founder effect speciation: Divergence pop-
ulation genetics of the spoonbills Platalea regia and Pl. minor geodispersal in Neotropical freshwater fishes. J Biogeogr.
(Threskiornithidae, Aves). Molecular Biology and Evolution, 28, 473– 2018;00:1–13. https://doi.org/10.1111/jbi.13207
482. https://doi.org/10.1093/molbev/msq210