cryptic speciation in the white shouldered antshrike thamnophilus aethiops aves thamnophilidae the...

8/16/2019 Cryptic Speciation in the White Shouldered Antshrike Thamnophilus Aethiops Aves Thamnophilidae the Tale of a Tr…

1/16

Cryptic speciation in the white-shouldered antshrike (Thamnophilusaethiops, Aves – Thamnophilidae): The tale of a transcontinentalradiation across rivers in lowland Amazonia and the northeasternAtlantic Forest

Gregory Thom a, Alexandre Aleixo b,⇑

a Curso de Pós-graduação em Zoologia, Universidade Federal do Pará/Museu Paraense Emílio Goeldi, Caixa Postal 399, CEP 66040-170 Belém, PA, Brazilb Coordenação de Zoologia, Museu Paraense Emílio Goeldi, Caixa Postal 399, CEP 66040-170 Belém, PA, Brazil

a r t i c l e i n f o

Article history:

Received 23 June 2014

Revised 20 September 2014

Accepted 26 September 2014

Available online 5 October 2014

Keywords:

Diversification hypotheses

Historical biogeography

Phylogeography

Population genetics

Refuges

Taxonomy

a b s t r a c t

The growing knowledge on paleogeography and the recent applications of molecular biology and phylo-

geography to the study of the Amazonian biota have provideda framework for testing competinghypoth-

eses of biotic diversification in this region. Here, we reconstruct the spatio-temporal context of

diversification of a widespread understory polytypic Amazonian bird species (Thamnophilus aethiops)

and contrast it with different hypotheses of diversification and the taxonomy currently practiced in the

group. Sequences of mtDNA (cytochrome b and ND2) and nuclear (b-fibrinogen introns 5 and 7 and the

Z-liked Musk4) genes, addingup to 4093 bp of 89 individuals covering the Amazonian, Andean, andAtlan-

tic Forest populations of T. aethiops were analyzed. Phylogenetic and population genetics analyses

revealed ten reciprocally monophyletic and genetically isolated or nearly-isolated lineages in T. aethiops,

highlighting several inconsistencies between taxonomy and evolutionary history in this group. Our data

suggest that the diversification of T. aethiops started in the Andean highlands, and then proceeded into

the Amazonian lowlands probably after the consolidation of the modern Amazonian drainage. The main

cladogenetic events in T. aethiops may be related to the formation and structuring of large Amazonian riv-

ers during the Late Miocene–Early Pleistocene, coinciding with the dates proposed for other lineages of

Amazonian organisms. Population genetics data do not support climatic fluctuations as a major source

of diversification in T. aethiops. Even though not entirely concordant with paleobiogeographic models

derived from phylogenies of other vertebrate lineages, our results support a prominent role for rivers as

major drivers of diversification in Amazonia, while underscoring that different diversification scenarios

are probably related to the distinct evolutionary origins of groups being compared.

2014 Elsevier Inc. All rights reserved.

1. Introduction

Since the 19th century, the high species diversity of the Amazon

allied to multifaceted patterns of geographical distributions insti-gated naturalists and researchers to explain such complexity

(Wallace, 1852; Haffer, 1969; Bates, 2001).

Despite the large number of distinct hypotheses to explain

Amazonian biodiversity (e.g., refuge hypothesis, Haffer, 1969; riv-

erine hypothesis, Wallace, 1852; Ayres and Clutton-Brock, 1992;

riverine-refuge hypothesis, Ayres and Clutton-Brock, 1992;

Haffer, 1993, 2001; ecological gradients hypothesis, Endler, 1977;

‘‘museum’’ hypothesis, Roy et al., 1997; and marine incursions,

Bates, 2001), and the common sense that many causations have

operated for the formation of such diversity (Bush, 1994; Haffer,

2001; Miller et al., 2008), recent phylogeographic and paleobioge-

ographic reconstructions (Aleixo and Rossetti 2007; Patel et al.,2011; Weir and Price, 2011; Ribas et al., 2012) have postulated

the formation of the current Amazonian physical landscape as

the main source of cladogenetic events among the studied lineages.

The formation of the Amazon basin was ultimately driven by

the Andean uplift and related arches, which resulted in the forma-

tion of a large fluvio-lacustrine system in western Amazonia, dur-

ing the early Miocene (Espurt et al., 2010; Mora et al., 2010). With

the continuous rise of the Andes in the middle Miocene, this sys-

tem probably expanded to the Purus Arch, a tectonic structure

located 300 km west of Manaus, and which separates the Solimões

and Amazonas sedimentary basins (Figueiredo et al., 2009; Hoorn

http://dx.doi.org/10.1016/j.ympev.2014.09.023

1055-7903/ 2014 Elsevier Inc. All rights reserved.

⇑ Corresponding author.

E-mail address: [email protected] (A. Aleixo).

Molecular Phylogenetics and Evolution 82 (2015) 95–110

Contents lists available at ScienceDirect

Molecular Phylogenetics and Evolution

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / y m p e v

http://dx.doi.org/10.1016/j.ympev.2014.09.023mailto:[email protected]://dx.doi.org/10.1016/j.ympev.2014.09.023http://www.sciencedirect.com/science/journal/10557903http://www.elsevier.com/locate/ympevhttp://www.elsevier.com/locate/ympevhttp://www.sciencedirect.com/science/journal/10557903http://dx.doi.org/10.1016/j.ympev.2014.09.023mailto:[email protected]://dx.doi.org/10.1016/j.ympev.2014.09.023http://-/?-http://-/?-http://-/?-http://-/?-http://-/?-http://crossmark.crossref.org/dialog/?doi=10.1016/j.ympev.2014.09.023&domain=pdfhttp://-/?-


2/16

et al., 2010). The period when the Purus Arch was breached prob-

ably marked the onset of the development of the transcontinental

Amazon River, draining eastward the western fluvio-lacustrine

system, and enabling the formation of upland terra-firme forests

in the western Amazonia; however, the timing of this event is still

a matter of controversy, with two main proposed periods based on

different sources of evidence: late Miocene, between 6.5 and 10 Ma

(Hoorn et al., 2010; Latrubesse et al., 2010) and Pliocene (approx-imately 2.5 Ma; Campbell et al., 2006). Despite the fact that major

interfluves of the Amazonian region delineate the geographical

distribution of most endemic vertebrate taxa and that this has pro-

vided the basis for the recognition of the so called areas of ende-

mism (Cracraft, 1985; Silva et al., 2005), the process of river

dynamics driving population divergence is still poorly understood,

with multiple area-relationships patterns documented so far

(Aleixo, 2004; Fernandes et al., 2012; Ribas et al., 2012; d’Horta

et al., 2013). This suggests that although rivers have been barriers

to gene flow, other historical events probably contributed to the

overall diversification pattern. However, this knowledge is still

meager when compared with the rich and complex biodiversity

of Amazonia, preventing the recognition of a general model or sets

of heuristic models of diversification, corroborated by several

lineages.

The polytypic species Thamnophilus aethiops (Aves: Thamno-philidae) is a good model to study the paleobiogeography of the

Amazon basin due the following reasons: (1) it is widely distrib-

uted in Amazonia and neighboring areas such as the Andes and

the Atlantic Forest in eastern Brazil, two areas known to be histor-

ically connected to the development of modern Amazonia; (2) it is

restricted to the understory of upland terra-firme forest, rapidlyresponding to environmental change (Stotz et al., 1996); and (3)

subspecies distributions are mainly bounded by the major tributar-

ies of the Amazon River, with some important exceptions that can

provide insights into the circumstances whereby a single lineage

does and does not respond to rivers as barriers. Thus, we estimated

the spatio-temporal scenario of diversification of the polytypic T.

aethiops to address the following questions: (1) what are the evo-lutionary relationships among the populations/subspecies of T.aethiops? (2) what are the relationships between the events that

led to diversification in T. aethiops and the paleogeographic modelsproposed for the formation of Amazonia? (3) do the demographic

history of the studied populations support any particular diversifi-

cation hypothesis?

2. Material and methods

2.1. Samples, laboratory procedures, and data analyses

A total of 89 individuals from 55 localities were sampled

throughout the distribution of T. aethiops in Amazonia, northeast-ern Atlantic Forest and the Andean foothills, covering nine of ten

described subspecies (Zimmer and Isler, 2003; Fig. 1b, Table 1).

Only T. a. wetmorei, purportedly endemic to the Andean foothills

of Colombia, was not sampled. Nonetheless, we assume here that

that this taxon is closely related to the polionotus subspecies fromnorthwestern Amazonia, from which it is hardly differentiated

based on plumage characters (Zimmer and Isler, 2003). Subspecific

identification of our samples were made through the inspection of

voucher specimens and followed the most recent taxonomy pro-

posed for T. aethiops (Zimmer and Isler 2003). We used Thamnophi-lus aroyae, T. unicolor , and T. caerulescens as outgroups due theirclose relationship to T. aethiops (Brumfield and Edwards, 2007).

Total DNA was extracted from approximately 20 mg of muscle

tissue following a standard phenol/chloroform protocol(Sambrook et al., 1989) or with the aid of a DNeasy Quiagen

(Hilden, Germany) extraction kit. We amplified the mitochondrial

genes cytochrome b (cyt b, 992 bp, n = 88) and NADH dehydroge-nase subunit 2 (ND2, 1041 bp, n = 79), as well as the nuclear genesb-fibrinogen intron 7 (Bf7, 963 bp, n = 51), b-fibrinogen intron 5

(Bf5, 532 bp, n = 63), and the intron 4 of the skeletal muscle recep-tor of tyrosine kinase (Musk4, 565 bp, n = 64) linked to the Z chro-mosome (see Table 2 for the primers used). Polymerase chain

reaction (PCR) was performed with 25ll of final volume, andapproximately 50 ng of genomic DNA, 1.5–2.5 mM of MgCl2,200 mM of dNTPs and 0.1 U of Taq DNA polymerase Promega

(Madison, WI, USA). Reactions started with a denaturation step at

94 C for 5 min, followed by 35 cycles of three steps: (1) 94 C for

1 min; (2) annealing temperatures ranging from 50 C (cyt b andBf5) and 52 C (Musk4) to 59 C (ND2, Bf7) for 1 min; and (3)

72 C for 1 min. The last step, for the extension, was at 72 C for

5 min. PCR products were purified with the Polyethylene glycol

protocol (PEG). Sequencing was carried out on an ABI 3130 auto-

mated capillary sequencer (Applied Biosystems, Foster City, Cali-

fornia, USA) with the ABI Prism Big Dye terminator Kit. To

confirm observed mutations, both strands of each sample were

sequenced. The DNA strands were edited and aligned manually

in BIOEDIT 7.0.5 (Hall, 1999). Saturation of the nucleotide substitu-

tions in the mitochondrial DNA were evaluated using the software

DAMBE (Xia and Xie, 2001). For nuclear loci, heterozygous nucleo-

tide positions were inferred by the presence of double peaks of the

same size in the electropherogram. To obtain the gametic phase of

haplotypes of the heterozygous individuals, we used a Bayesian

approach as implemented in PHASE 2.1 1 (Stephens et al., 2001;

Stephens and Donnelly, 2003; Stephens and Scheet, 2005). The

threshold of 70% of posterior probability was assumed as a confi-

dence value for the analyzed haplotypes (see Harrigan et al.,

2008). Lower values were regarded as ambiguous. To obtain the

gametic phase for individuals heterozygous in size (presence of

indels in one of the strands), we used the program CHAMPURU

v1.0 (Flot et al., 2006; Flot, 2007; available online at http://

www.mnhn.fr/jfflot/champuru/). Females did not have the gametic

phase estimated for Musk4, since this gene is linked to de Z chro-mosome. To test the hypothesis of neutral evolution among inde-

pendent loci we performed the Tajima’s D test (Tajima, 1989)

using DNASP v.4.10.9 (Rozas et al., 2003). The significance for these

tests was assessed through 10.000 coalescence simulations. To

check for possible recombination among the nuclear loci we used

the phi test as implemented in the SPLITSTREE 4.10 software

(Huson and Bryant, 2006).

2.2. Phylogenetic analyses

2.2.1. Concatenated genes and gene treesThe phylogenetic analyses were performed using Bayesian

inference (BI) in MrBAYES 3.1.2 (Ronquist and Huelsenbeck,

2003) and Maximum likelihood (ML) in RaxML 7.0.3 (Stamatakis,2006). The evolutionary models which best explain the evolution

of each dataset were selected in JMODELTEST 0.1.1 (Posada,

2008), using the Bayesian information criterion (BIC) for BI and

the Akaike information criterion (AIC) for ML. For the BI analyses

with the concatenated dataset, a Bayes factor analysis (Brandley

et al., 2005) selected four partitions (mtDNA + one separate parti-

tion for each nuclear gene) as the best partitioning scheme. The

evolutionary models for each partition were: (1) mtDNA (GTR + G);

(2) Bf5 (GTR + I + G); (3) Bf7 (GTR + I); and (4) Musk4 (HKY + G).

The BI estimated based on mtDNA only used one model for the

two genes (GTR + G). BI analyses were generated through two

independent runs of 1 107 generations, each with four Markov

chains. The parameters of the chains were sampled every 1000

generations and the first 1000 trees were discarded as burn-in.The posterior probabilities for each estimated node were obtained

96 G. Thom, A. Aleixo / Molecular Phylogenetics and Evolution 82 (2015) 95–110

http://www.mnhn.fr/jfflot/champuru/http://www.mnhn.fr/jfflot/champuru/http://www.mnhn.fr/jfflot/champuru/http://www.mnhn.fr/jfflot/champuru/


3/16

Fig. 1. (a) Phylogenetic relationships among major clades of the polytypic Thamnophilus aethiops and outgroups derived from Bayesian and Maximum likelihood inferences

based on sequences of the mtDNA genes (cyt b and ND2; 2033 bp) and all loci combined (cyt b, ND2, Bf5, Bf7 and Musk4; 4093 bp). Nodal support values correspond to

posterior probabilities (above) and bootstrap pseudo-replicates (below) obtained, respectively, in the mtDNA and concatenated dataset analyses. (b) Map showing the current

distribution and sampled localities of major clades of the polytypic T. aethiops. Sample codes are the same used in Fig. 1a and Table 1. (c) Haplotype network of the five

sequenced genes estimated through HaploViewer (Salzburger et al., 2011). Clade colors are the same as in Fig. 1a and 1b. Numbers inside circles represent actual number of

individuals having a particular haplotype, with circle sizes being proportional to it, while numbers on connecting branches are nucleotide substitutions higher than one

separating haplotypes. Asterisks beside light blue circles represent the number of individuals from the Pernambuco area of endemism. (d) Bayesian Phylogenetics and

Phylogeography (BP&P) results. Numbers on braches represent the number of times a given node was recovered with a speciation probability P0.95 based on six analyses

performed using two algorithms (0 and 1) and three combinations of priors on ancestral population sizes and divergence times (see methods). (e) Overlapped Bayesian

estimatedchronogramsderived fromalternativedating analyses based on theconcatenated mtDNA (cyt b andND2; 2033 bp;branches and95% of confidence interval in dark

gray) and a multi-locus coalescent species tree (mtDNA, Bf5/Bf7 andMusk4; 4093 bp; branches and 95%confidence intervals in Black). Asterisks andcrosses represent values

of posterior probability P95% for the species tree and concatenated mtDNA analyses, respectively. (For interpretation of the references to color in this figure legend, the

reader is referred to the web version of this article.)

G. Thom, A. Aleixo / Molecular Phylogenetics and Evolution 82 (2015) 95–110 97

http://-/?-


4/16

Table 1

Code (asin Fig. 1a andb), taxon(speciesor subspecies followingthe taxonomy in Zimmer and Isler, 2003), lineage (namedaccordingto recognizedNeotropicalareas of endemism as

in Silvaet al., 2005, withsome modifications), N – vouchernumber, Inst.– institution(MPEG: MuseuParaense Emílio Goeldi; LSUMZ: Louisiana State UniversityMuseum of Natural

Science), and locality information (presented coordinates are plotted in Fig. 1b) associated with each sample of Thamnophilus aethiops and outgroups sequenced in this study.

Code Taxon Lineage N Inst. Locality Coordinates

T1 T. a. atriceps Tapajós 63,888 MPEG Brazil, Pará, Altamira, Floresta Nacional de Altamira 06S0401200

55W1901200

T2 T. a. atriceps Tapajós 70,861 MPEG Brazil, Pará, Itaituba, Miritituba 04S2602400

55W380

2400

T2 T. a. atriceps Tapajós 70,862 MPEG Brazil, Pará, Itaituba, Miritituba

T3 T. a. atriceps Tapajós 59,159 MPEG Brazil, Pará, Novo Progresso, 20 km SW 07S1102400

55W3000000

T3 T. a. atriceps Tapajós 59,160 MPEG Brazil, Pará, Novo Progresso, 20 km SW

T3 T. a. atriceps Tapajós 59,161 MPEG Brazil, Pará, Novo Progresso, 20 km SW

X1 T. a. atriceps Xingu 57,707 MPEG Brazil, Pará, Senador José Porfírio, Rio Xingu 02S3202400

51W3300000

X1 T. a. atriceps Xingu 57,708 MPEG Brazil, Pará, Senador José Porfírio, Rio Xingu

X1 T. a. atriceps Xingu 57,709 MPEG Brazil, Pará, Senador José Porfírio, Rio Xingu

X2 T. a. atriceps Xingu 70,635 MPEG Brazil, Pará, Carajás, Serra dos Carajás 06S0000000

51W1904800

X3 T. a. atriceps Xingu 70,658 MPEG Brazil, Pará, Carajás, Serra dos Carajás, Serra Norte 06S0900000

50W2400000

P1 T. a. distans Pernambuco 70,454 MPEG Brazil, Alagoas, Ibateguara, Engenho Coimbra 09S1001200

35W3202400

P1 T. a. distans Pernambuco 70,455 MPEG Brazil, Alagoas, Ibateguara, Engenho Coimbra

P1 T. a. distans Pernambuco 70,457 MPEG Brazil, Alagoas, Ibateguara, Engenho Coimbra

P2 T. a. distans Pernambuco 70,459 MPEG Brazil, Pernambuco, Barreiros, Cachoeira Linda 08S4800000

35W1901200

P2 T. a. distans Pernambuco 70,460 MPEG Brazil, Pernambuco, Barreiros, Cachoeira Linda

P2 T. a. distans Pernambuco 70,458 MPEG Brazil, Pernambuco, Barreiros, Cachoeira Linda

B1 T. a. incertus Belém 58,647 MPEG Brazil, Pará, Santa Bárbara, Parque Ecológico GUNMA 01S3303600

47W3903600

B2 T. a. incertus Belém 61,190 MPEG Brazil, Pará, Barcarena 02S1800000

48W3903600

B2 T. a. incertus Belém 61,192 MPEG Brazil, Pará, Barcarena

B3 T. a. incertus Belém 58,632 MPEG Brazil, Pará, Tomé-Açú, Quatro Bocas 02S3202400

48W0104800

B4 T. a. incertus Belém 70,559 MPEG Brazil, Pará, Curuça 01S0202400

47W2703600

EI1 T. a. injunctus Eastern Inambari 68,883 MPEG Brazil, Amazonas, Careiro, Br 319 km 04S0404800

60W3903600

EI2 T. a. injunctus Eastern Inambari 71,054 MPEG Brazil, Amazonas, Humaitá, Rio Ipixuna 07S3101200

63W2002400

EI2 T. a. injunctus Eastern Inambari 71,055 MPEG Brazil, Amazonas, Humaitá, Rio Ipixuna



EI3 T. a. injunctus Eastern Inambari 60,637 MPEG Brazil, Acre, Senador Guiomard, Ramal Nabor Júnior 09S4604800

67W1904800

WI1 T. a. kapouni Western Inambari 58,016 MPEG Brazil, Acre, Parque Nacional Serra do Divisor 08S2100000

73W1803600

WI2 T. a. kapouni Western Inambari 59,837 MPEG Brazil, Acre, Assis Brasil, ESEC Rio Acre 11S0303600

70W1601200

WI3 T. a. kapouni Western Inambari 60,171 MPEG Brazil, Amazonas, RDS Cujubim, Rio Jutaí 04S3900000

68W1904800

WI4 T. a. kapouni Western Inambari 60,631 MPEG Brazil, Acre, Tarauacá, Br 364 km 40, Rio Liberdade 07S5302400

71W3903600

WI4 T. a. kapouni Western Inambari 60,632 MPEG Brazil, Acre, Tarauacá, Br 364 km 40, Rio Liberdade

WI5 T. a. kapouni Western Inambari 62,071 MPEG Brazil, Acre, Porto Walter, Igarapé Cruzeiro do Vale 08S3300000

72W5201200

WI6 T. a. kapouni Western Inambari 40,722 LSUMZ Peru, Loreto, 86 km SE Juanjui 07S3301900

75W5403200

WI6 T. a. kapouni Western Inambari 40,661 LSUMZ Peru, Loreto, 86 km SE Juanjui

WI7 T. a. kapouni Western Inambari 11,235 LSUMZ Peru, Ucayali, SE Cerro Tahuayo, ENE Pucallpa 08S2503300

74W2202600

WI8 T. a. juruanus Western Inambari 59,966 MPEG Brazil, Acre, Rio Branco, Fazenda Experimental Catuaba 02S1904800

67W3600000

WI9 T. a. juruanus Western Inambari 60,634 MPEG Brazil Acre, Bujari, Floresta Estadual do Antimary 09S2100000

68W0600000

WI9 T. a. juruanus Western Inambari 60,636 MPEG Brazil Acre, Bujari, Floresta Estadual do Antimary

WI10 T. a. juruanus Western Inambari 60,638 MPEG Brazil, Acre, Plácido de Castro, Ramal Novo Horizonte 10S0704800

67W1904800

WI11 T. a. juruanus Western Inambari 61,297 MPEG Brazil, Acre, Rio Branco, Ramal Jarinal 09S5403600

68W1901200

WI11 T. a. juruanus Western Inambari 61,298 MPEG Brazil, Acre, Rio Branco, Ramal Jarinal

WI11 T. a. juruanus Western Inambari 61,300 MPEG Brazil, Acre, Rio Branco, Ramal Jarinal



5/16

Table 1 (continued)

Code Taxon Lineage N Inst. Locality Coordinates

WI12 T. a. juruanus Western Inambari 62,069 MPEG Brazil, Acre, Mâncio Lima, Estrada do Barão 07S3300000

72W3603600

WI12 T. a. juruanus Western Inambari 62,070 MPEG Brazil, Acre, Mâncio Lima, Estrada do Barão

WI13 T. a. juruanus Western Inambari 59,965 MPEG Brazil, Acre, Porto Acre, Reserva Humaitá 09S4503600

67W4001200

WI14 T. a. juruanus Western Inambari 63,336 MPEG Brazil, Acre, Santa Rosa, Rio Purus 09S0701200

69W4904800

WI15 T. a. juruanus Western Inambari 64,546 MPEG Brazil, Acre, Senador Guiomard, Ramal Oco do Mundo 09S5002400

67W4901200

WI16 T. a. juruanus Western Inambari 8984 LSUMZ Bolivia, Pando, Nicolás Suarez, 12km by road of Cobija 11S1000200

68W2601800

WI17 T. a. juruanus Western Inambari 1164 LSUMZ Bolivia, La Paz, 20 k m by Rio Beni N. Puerto Linares 15S1400600

67W3701100

EN1 T. a. polionotus Eastern Negro 59,521 MPEG Brazil, Amazonas, Barcelos, Rio Aracá 00S2501200

62W5602400

EN1 T. a. polionotus Eastern Negro 59, 522 MPEG Brazil, Amazonas, Barcelos, Rio Aracá

EN1 T. a. polionotus Eastern Negro 59, 523 MPEG Brazil, Amazonas, Barcelos, Rio Aracá

WN1 T. a. polionotus Western Negro 59,524 MPEG Brazil, Amazonas, Barcelos, Rio Cuiuni 00S4604800

63W1604800

WN2 T. a. polionotus Western Negro 59,525 MPEG Brazil, Amazonas, Novo Airão, Igarapé-Açu 02S5100000

60W5100000

WN3 T. a. polionotus Western Negro 62,770 MPEG Brazil, Amazonas, Maraã, Lago Cumapi 01S4304800

65W520

4800

WN3 T. a. polionotus Western Negro 62, 771 MPEG Brazil, Amazonas, Maraã, Lago Cumapi

NR1 T. a. punctuliger Northern Rondônia 62,215 MPEG Brazil, Amazonas, Juruti, Base Capiranga 02S2801200

56W0000000

NR1 T. a. punctuliger Northern Rondônia 62,216 MPEG Brazil, Amazonas, Juruti, Base Capiranga

NR1 T. a. punctuliger Northern Rondônia 58,100 MPEG Brazil, Amazonas, Juruti, Base Capiranga

NR2 T. a. punctuliger Northern Rondônia 68,944 MPEG Brazil, Pará, Jacareacanga, Terra do Burandir 06S1500000

57W5201200

NR2 T. a. punctuliger Northern Rondônia 68,965 MPEG Brazil, Pará, Jacareacanga, Terra do Burandir

NR3 T. a. punctuliger Northern Rondônia 66,127 MPEG Brazil, Pará, Santarém, Retiro 02S2400000

55W4702400

NR4 T. a. punctuliger Northern Rondônia 67,048 MPEG Brazil, Amazonas, Maués, Flona do Pau Rosa 03S5403600

58W2400000

SR1 T. a. punctuliger Southern Rondônia 54,953 MPEG Brazil, Rondônia, Guajará-Mirim, REBIO Ouro Preto 10S4904800

64W4500000

SR1 T. a. punctuliger Southern Rondônia 54,955 MPEG Brazil, Rondônia, Guajará-Mirim, REBIO Ouro Preto

SR2 T. a. punctuliger Southern Rondônia 61, 575 MPEG Brazil, Mato Grosso, Querência, Fazenda Tanguro 12S5302400

52W2201200

SR2 T. a. punctuliger Southern Rondônia 61,576 MPEG Brazil, Mato Grosso, Querência, Fazenda Tanguro

SR2 T. a. punctuliger Southern Rondônia 70,226 MPEG Brazil, Mato Grosso, Querência, Fazenda Tanguro

SR3 T. a. punctuliger Southern Rondônia 67,397 MPEG Brazil, Mato Grosso, Paranaíta, Rio Teles Pires 09S3300000

56W3401200

SR3 T. a. punctuliger Southern Rondônia 67,396 MPEG Brazil, Mato Grosso, Paranaíta, Rio Teles Pires

SR3 T. a. punctuliger Southern Rondônia 67,398 MPEG Brazil, Mato Grosso, Paranaíta, Rio Teles Pires

SR4 T. a. punctuliger Southern Rondônia 67, 442 MPEG Brazil, Mato Grosso, Paranaíta, Fazenda Aliança 09S4004800

56W5103600

SR4 T. a. punctuliger Southern Rondônia 67,443 MPEG Brazil, Mato Grosso, Paranaíta, Fazenda Aliança

SR5 T. a. punctuliger Southern Rondônia 69, 215 MPEG Brazil, Mato Grosso, Paranaíta, Fazenda Paranaíta 09S3404800

56W4301200

SR6 T. a. punctuliger Southern Rondônia 58,704 MPEG Brazil, Amazonas, Humaitá, T. Indígena Parintintin 07S5700000

62W2201200

SR6 T. a. punctuliger Southern Rondônia 58,705 MPEG Brazil, Amazonas, Humaitá, T. Indígena Parintintin

SR6 T. a. punctuliger Southern Rondônia 58,707 MPEG Brazil, Amazonas, Humaitá, T. Indígena Parintintin

SR7 T. a. punctuliger Southern Rondônia 14,649 LSUMZ Bolivia, Santa Cruz, Serrania de Huanchaca 14S3104800

60W410

2400

A1 T. a. aethiops 5500 LSUMZ Peru, San Martín, 28 k m by road NE Tarapoto 06S2600600

76W1900400

T aroyae 38,993 LSUMZ Bolivia, Cochabamba, Chapare

T aroyae 38,985 LSUMZ Bolivia, Cochabamba, Chapare

T aroyae 22,836 LSUMZ Bolivia, La Paz

T. unicolor 12,144 LSUMZ Ecuador, Pichincha, Mindo

T. unicolor 6196 LSUMZ Morona-Santiago Province

T. unicolor 7741 LSUMZ Ecuador, Caóar

T. unicolor 32,602 L SUMZ P eru, Cajamarca

T. unicolor 43,515 LSUMZ Peru, San Martín


http://-/?-


6/16

through a majority rule consensus of the remaining MCMC sam-

ples. The ML analyses were not partitioned and the evolutionary

models selected were GTR + I + G for the concatenated (all loci)

and GTR + G for the mtDNA datasets. Nodal support was estimated

with 1000 bootstrap pseudo-replicates. Both BI and ML analyses

with the concatenated dataset (all loci) were performed using

the nuclear genes with both gametic phases and the mitochondrial

genes duplicated. To visualize genealogical relationships among

individuals of each sequenced gene, haplotype networks were con-

structed in HaploViewer (Salzburger et al., 2011) based on the

topologies recovered in the species tree analysis. Mean pairwise

p-distances (Nei, 1987) within and among populations were calcu-lated using the mtDNA dataset only in MEGA 4.0 ( Tamura et al.,

2007).

2.2.2. Species treeRecent approaches have shown that multi-locus analyses with

concatenated datasets can result in well-supported topologies thatare incongruent with the true species tree, particularly in recent

scenarios of diversification (Degnan and Rosenberg, 2009;

Kubatko and Degnan, 2007). Thus, we used a Bayesian hierarchical

model implemented in *BEAST 1.6.1 (Drummond and Rambaut,

2007; Heled and Drummond, 2010) to obtain an estimate of the

species tree for the different lineages of T. aethiops. For thisapproach, we assumed three independent loci: (1) mtDNA (cyt band ND2); (2) Bf5 and Bf7; and (3) Musk4. The evolutionary models

were the same used in the BI for the concatenated dataset, with the

addition of the locus joining the Bf5 and Bf7 genes, for which the

GTR + I + G model was selected. We performed an initial run of

5 107 generations to optimize the analysis operators. Afterwards,

five independent runs of 5 107 generations were performed,

sampling the parameters of the Markov chain every 1000 genera-

tions, until the effective sample sizes (ESS) of all parameters were

equal or higher than 200 (Kuhner 2008). Four independent runs

were performed and combined in a majority rule consensus tree

with posterior probability support. TRACER 1.5 (Rambaut and

Drummond, 2007) was used to access ESS values and verify when

the MCMC reached stable and convergent values, allowing for a

definition of burn-in values. Based on the obtained results and

under a conservative approach, the first 5000 samples of each

run were discarded as burn-in. The purported species in this anal-

ysis were the reciprocally monophyletic populations of T. aethiopsrevealed by the BI and ML analyses (Fig. 1a).

2.2.3. Species delimitation

We implemented a coalescent-based bayesian modelingapproach to generate speciation probabilities of closely related

taxa from multi locus sequence data using the program Bayesian

Phylogenetics and Phylogeography (BP&P v.2.0b; Yang and

Rannala 2010), considered the most accurate coalescent-based

method for species delimitation evaluated by Camargo et al.

(2012). The model implemented in BP&P assumes no gene flow

among species, no recombination, and takes into account gene tree

uncertainty and lineage sorting. BP&P incorporates a model that

include the species divergence times (s), and the population sizeparameters h = 4N l, where N is the effective population size andl is the substitution rate per site per generation. However, reason-able prior distributions of population parameters can be difficult to

determine, potentially affecting the posterior probability of species

delimitations (Yang and Rannala 2010). Thus, we implemented the

approach of Leaché and Fujita (2010), performing analyses using

three combination of priors representing different population sizes

and ages for the root of the species tree as follows: (1) large ances-

tral population size and deep divergences (h and s gamma priorsG(1, 10) and G(1, 10)); (2) small ancestral population size and shal-

low divergences among species (h and s gamma priors G(2, 2000)and G(2, 2000)); and 3) large ancestral population size and shallow

divergences among species (h and s gamma priors G(1, 10) andG(2, 2000)). We ran the analyses for 1 106 generations, sampling

every five and discarding the first 5 104 generations as burn-in,

using the algorithms 0 and 1 with different fine tuning parameters

(e = 5 for algorithm 0 and a = 2 and m = 1 for algorithm 1). We usedthe populations and the topology obtained through the species tree

analyses as the guide tree (Fig. 1e). The population occurring dis-

junctly in the Pernambuco region (Atlantic Forest) was considered

a distinct lineage, sister to that of the Belém area of endemism

population.

2.3. Molecular dating

To estimate the divergence times among T. aethiops populationsrevealed by the BI and ML phylogeny estimates, a second BI was

carried out using only the mtDNA dataset (cyt b and ND2). Afterthe selection of the model of molecular evolution for this dataset

with JMODELTEST 0.1.1 (Posada, 2008), a preliminary run of

1 107 generations was performed in BEAST 1.6.1 (Drummond

and Rambaut, 2007) fixing the rate of nucleotide mutations at

1.0 per millions years for the cyt b (for which a robust calibrationis available; Weir and Schluter, 2008) and estimating that for the

ND2 under the options relaxed clock and uncorrelated lognormal.

This analysis was performed to evaluate whether both mitochon-

drial genes were evolving at similar rates and the cyt b calibration

could be applied to ND2. After the confirmation that these twogenes were evolving at similar rates, the overall molecular clock

Table 2

List of primers used in PCR amplifications and in DNA cycles sequencing reactions of samples of Thamnophilus aethiops and outgroups sequenced in this study.

Gene Primer Sequence 50–30 Source

Cyt b L14990 CCA TCC AAC ATC TCA GCA TGA TGA AA Brumfield et al. (2007)

H16065 AAC TGC AGT CAT CTC CGG TTT ACA AGA C Brumfield et al. (2007)

ND2 L5215 TAT CGG GCC CAT ACC CCG AAA AT Brumfield et al. (2007)

H5677 DGA DGA RAA DGC YAR RAY YTT DCG Brumfield et al. (2007)

L5758 GGN GGN TGR RBH GGN YTD AAY CAR AC Brumfield et al. (2007)

H6313 CTC TTA TTT AAG GCT TTG AAG GC Brumfield et al. (2007)

bf7 Fib-B17L TCC CCA GTA GTA TCT GCC ATT AGG GTT Prychitko and Moore (1997)

Fib-B17U GGA GAA AAC AGG ACA ATG ACA ATT CAC Prychitko and Moore (1997)

bf5 Fib5L CGC CAT ACA GAG TAT ACT GTG ACA T Brumfield et al. (2007)

Fib5H GCC ATC CTG GCG ATC TGA A Brumfield et al. (2007)

Musk4 Musk13R CTCTGAACATTGTGGATCCTCAA Clark and Witt (2)006

Musk13F CTTCCATGCACTACAATGGGAAA Clark and Witt (2006)



7/16

calibration used was 2.1% of nucleotide substitutions per million

years (Weir and Schluter, 2008), under the relaxed clock and

uncorrelated lognormal options. We performed an initial run of

1 107 generations to optimize the analysis operators. Afterwards,

two independent runs of 1 107 generations were performed,

sampling the parameters of the Markov chain every 1000 genera-

tions, until the ESS of all parameters were equal or higher than

200. Based on the obtained results and under a conservativeapproach, the first 800 samples of each run were discarded as

burn-in. Values of posterior probability and divergence times were

estimated through a majority rule consensus of the remaining

samples. For this analysis we used only one individual from each

populations using the Yule process as a speciation prior.

Analyses based on a multi-locus coalescent approach enable a

more accurate estimate of divergence processes, since gene diver-

gence times normally occur before true speciation events

(Maddison and Knowles, 2006; Carstens and Knowles, 2007;

Edwards et al., 2007; Heled and Drummond, 2010; McCormack

et al., 2010). Thus, during the species tree analysis, we set for the

mtDNA dataset (cyt b and ND2 genes) the mutation rate of 2.1%substitutions per million years (Weir and Schluter, 2008), whereas

those of the nuclear loci (Bf5/Bf7 and Musk4) were estimated in

comparison to that of the mtDNA under a Yule process speciation

prior.

2.4. Population genetics, historical demography, and gene flow

The recognized groups for the population genetics analyses

were the reciprocally monophyletic populations revealed by the

BI, ML, and species tree analyses (Fig. 1a and e). Even though pop-

ulations of the Belém and Pernambuco areas of endemism did not

show significant genetic differentiation and have not acquired reci-

procal monophyly, they were recognized as independent popula-

tions in the population genetics and network analyses, since they

are currently separated by a long stretch of non-forest environ-

ments (Cerrado and Caatinga, Ab’Saber, 1977), without possibility

of current gene flow.We analyzed sequences of the three sampled loci (mtDNA, Bf5/

Bf7 and Musk4) and populations with a minimum sampling of at

least five individuals. The nuclear loci were used with both gametic

phases, except females for Musk4. To estimate changes in historical

demography through non-neutral models of evolution, we per-

formed the Tajima’s D (Tajima, 1989), Fu’s Fs (Fu, 1997), and R2

Ramos-Onsins and Rozas (2002) tests. Significance was evaluated

through 10,000 coalescent simulations. We also calculated nucleo-

tide diversity (p) and haplotype diversity (h) for all populations. Allanalyses were performed in DNASP v.4.10.9 (Rozas et al., 2003).

The dynamics of effective population sizes through time of the

recognized populations of T. aethiops was estimated throughExtended Bayesian Skyline Plots (EBSP; Heled and Drummond,

2008), using a linear model in BEAST 1.6.1 (Drummond andRambaut, 2007). EBSP estimate changes in effective population

sizes over time under a multi-locus approach by using the times

of coalescent events among gene trees. We used the rate of 2.1%

of nucleotide substitutions per million years (Weir and Schluter,

2008) for the mtDNA with the relaxed clock and uncorrelated log-

normal priors. The evolutionary rates of the nuclear loci (Bf5/Bf7

and Musk4) were estimated in comparison to that the mtDNA.

The MCMC parameters were the same used in the species tree

analysis. The evolutionary models for each locus of each popula-

tion were estimated in JMODELTEST 0.1.1 (Posada, 2008).

To corroborate the obtained results in the phylogenetic analyses

and test for population structuring across recognized geographical

barriers (rivers and non-forest environments) we performed an

analysis of molecular variance (AMOVA) using the mtDNA datasetin ARLEQUIN 3.1 (Excoffier et al., 2006).

We applied the isolation with migration model (Hey and

Nielsen, 2004) implemented in the IMa program (Hey and

Nielsen, 2007) to estimate whether gene flow has occurred among

those parapatric populations potentially in contact (Fig. 1b). For

each pair of populations being compared, IMa estimates values of

H = 4N l (where l is the mutation rate per year per marker) forpopulations 1 and 2, and their ancestor, as well as the t time

(where t = t l, the time of population splitting at t generations inthe past) in which they diverged in the presence of gene flow(where m = m/l, given in the coalescent). We structured the anal-yses assuming as populations the reciprocal monophyletic lineages

reveled in the phylogenetic and species tree analyses. IMa esti-

mates compared populations potentially in contact, without an

apparent physical barrier separating them and parapatric popula-

tions occurring on opposite banks of major Amazonian tributaries

(see Fig. 1b). Our main objective with IMa analyzes was to test if

the observed absence of phylogenetic signal in the nuclear markers

(see Fig. 1c) is better explained by incomplete lineage sorting as a

result of recent population splitting (m = 0) or by the presence of

gene flow (m > 0). We assumed the Hasegawa–Kishino–Yano(HKY) model of evolution for all analyses and markers. Initial runs

in MCMC mode were performed to establish the best priors for

effective population sizes, divergence times, and migration param-

eters for each pair of populations being compared. After prior

selection, three independent final runs in MCMC mode were per-

formed changing the seed number to observe the convergence of

parameter values. Every run was performed using 1,000,000 gener-

ations of burn-in, sampling 100,000 trees in 5,000,000 generations

with 20 chains with a geometric heating. We performed a likeli-

hood ratio test (LRT, Nielsen and Wakeley, 2001), using the Load-

trees mode of IMa, sampling 100,000 generations and the results

of the final MCMC runs, to reject or accept scenarios with migra-

tion different from 0. If the LRT rejected a scenario of m > 0, weassumed that the absence of phylogenetic signal in the nuclear

markers was due to incomplete lineage sorting. On the other hand,

if the LRT rejected a scenario with m = 0, we performed another

MCMC run fixing m = 0 to evaluate whether a model withoutmigration adjusted better to the observed data than the full model

with migration. For these comparison we applied the Akaike infor-

mation Criterion (AIC), where the model that minimized

AIC = 2[log (L) d] was chosen as the best, where d is the numberof parameters (Nielsen and Wakeley, 2001). If the model with

migration had a better fit to the data, we assumed that the phylog-

eographic pattern may have been affected by gene flow. To convert

the model parameter estimations of m in demographic parameterswe applied the rate of 2.1% of nucleotide substitutions per million

years based on the mtDNA calibration adopted (Weir and Schluter,

2008), and a generation time of 1 year.

3. Results

3.1. Description of DNA sequences

Results of Tajima’s D was not significant for the mtDNA and

Musk4 (P > 0.05), but did reject neutrality significantly for Bf5/Bf7(P < 0.05; Table 3), suggesting that this locus does not evolve in aclock-like fashion. Wedid notdetect signsof saturation forthe faster

evolving mitochondrial genes (cyt b and ND2). The phi’stest did notrejectthe null hypothesis of a dataset without recombination forthe

nuclear genes (Bf5, Bf7 and Musk4; P > 0.1). Were identified 57 dis-tinct haplotypes for cyt b (992pb, n = 82), 38 for the ND2 (1041pb,

n = 72), 42 for Bf5 (532pb, n = 118), 50 for Bf7 (963pb, n = 98), and30 for Musk4(565pb, n = 97; Table 3). All sequences weredeposited

in Gen Bank under the accession numbers KF685940-KF686022

(cyt b), KF664029-KF664102 (ND2), KF686023-KF686081 (Bf5),KF686141-KF686186 (Bf7) and KF686081-KF686140 (Musk4).


http://-/?-


8/16

3.2. Area and phylogenetic relationships

BI and ML analyses based on the mtDNA and the concatenated

dataset (total of four analyses; Fig. 1a) recovered identical topolo-

gies showing with high statistical support the reciprocal mono-

phyly of 10 populations of T. aethiops, with T. aroyae grouping asthe sister species. Two main clades were recovered in T. aethiops,splitting apparently parapatric and attitudinally segregated

Andean foothill from lowland Amazonian populations, which

diverge by about 5% (uncorrected p genetic distance-value;Table 4). A subsequent major split involved lowland populations

separated by the Tapajós River in south-central Amazonia, which

diverge by roughly 4% and are hereafter referred to as the eastern

and western clades (populations found to the east and west of the

Tapajós River), respectively. The eastern clade grouped reciprocally

monophyletic populations endemic to the Tapajós, Xingu, and

Belém Amazonian areas of endemism (Silva et al., 2005), as well

as the isolated Pernambuco population in the northern sector of

the Atlantic Forest biome in northeastern Brazil (Fig. 1a and b).

The latter population was nested within the Belém population

despite these two being located in different biomes separated by

more than 1000 km of inhospitable habitat to T. aethiops (extensivenon-forest vegetation). The western clade grouped populations

from the Madeira, Inambari, Napo, Imeri, and Guiana areas of ende-

mism (Silva et al., 2005), but only the Napo/Imeri (west of the

Negro River) and Guiana (east of the Negro River to west Branco

River) areas were occupied by reciprocally monophyletic popula-

tions (Fig. 1a and b). Although we did not sample populations from

the Napo area of endemism (Silva et al., 2005), we tentatively

include them in the Imeri population based on their morphological

similarity and placement in the same subspecies polionotus(Zimmer and Isler, 2003). In contrast, the Inambari and Rondônia

Table 3

Diversity and demographic parameters of lineages of the polytypic T. aethiops for which at least five individuals were sequenced for mitochondrial (cyt b and ND2) and nuclear

(Bf5/Bf7, and musk4) genes. Lin. = lineages/clades; N = number of phased alleles; n H = number of haplotypes; H = haplotype diversity; p = nucleotide diversity; D = Tajima’s Dtest (Tajima,1989); F s = Fu’s Fs (Fu, 1997); R2 = Ramos-Onsins, Rozas test (2002). Significance levels for Tajima’s D an d R2 = *P < 0.05; **P < 0.01. Significance levels for Fu’s

F s = *P < 0.02.

Lin. Gene N n H H p D F s R2

T. a. aethiops mtDNA 72 64 0.996 0.0261 0.3502

Bf5/Bf7 118 42 0.935 0.0054 1.68371*

musk4 97 30 0.921 0.0051 1.37883

Tapajós mtDNA 6 6 1.000 0.0018 0.4155 1.013 0.1446

Bf5/Bf7 6 5 0.933 0.0061 0.7612 1.009 0.211

musk4 5 3 0.700 0.0023 0.1747 0.0607 0.2848

Xingu mtDNA 5 4 0.900 0.0014 0.1909 0.4448 0.2186

Bf5/Bf7 6 4 0.867 0.0083 0.1057 1.0749 0.2073

musk4 5 3 0.700 0.0033 0.2735 0.6436 0.1936

Pernambuco mtDNA 6 4 0.800 0.0006 0.4474 1.4544 0.1805

Bf5/Bf7 8 3 0.714 0.0037 0.5044 1.7263 0.2143

musk4 6 3 0.600 0.0032 0.5617 0.957 0.2206

Belém mtDNA 5 5 1.000 0.0042 0.4208 0.5418 0.1597

Bf5/Bf7 12 6 0.864 0.0028 0.7877 1.8982 0.2027

musk4 8 3 0.714 0.0024 0.9657 0.8748 0.2381

Eastern IInambari mtDNA 6 6 1.000 0.0021 0.6915 2.4843 0.1446

Bf5/Bf7 6 6 1.000 0.0077 0.9902 2.4199 0.1023**

musk4 6 5 0.933 0.0044 1.0484 1.5652 0.2386

Western Inambari mtDNA 15 13 0.981 0.0021 1.5705* 7.6834* 0.0700*

Bf5/Bf7 32 14 0.897 0.0034 1.6007* 8.3201* 0.0582**

musk4 25 11 0.847 0.0031 0.6854 5.3341* 0.0958

Northern Rondônia mtDNA 7 7 1.000 0.0021 0.3606 3.4573* 0.1267*

Bf5/Bf7 14 8 0.901 0.0040 0.768 2.8844 0.1174

musk4 11 9 0.945 0.0045 0.9471 5.1712* 0.1247

Southern Rondônia mtDNA 15 12 0.971 0.0047 0.4438 2.11639 0.1208

Bf5/Bf7 24 6 0.757 0.0023 1.2115 1.1177 0.1042

musk4 20 7 0.800 0.0036 0.5208 0.9804 0.1166

Table 4

Average pairwise uncorrected p-distances (mtDNA) between and within major clades of the polytypic Thamnophilus aethiops and outgroups recovered by Bayesian and Maximum

Likelihood phylogenies.

Clade (average pairwise p-distances within clades) 1 2 3 4 5 6 7 8 9 10 11

1 – Western Negro (0.1)

2 – Eastern Negro (0) 0.9

3 – Western Inambari (0.2) 1.6 1.4

4 – Southern Rondônia (0.6) 2.2 2.1 2.7

5 – Eastern Inambari (0.2) 2.2 2.3 2.9 1.7

6 – Northern Rondônia (0.2) 2.7 2.8 3.1 3.0 3.4

7 – Tapajós (0.1) 3.6 3.7 4.0 3.8 3.9 3.9

8 – Xingu (0.2) 3.4 3.4 3.6 3.6 3.7 3.8 1.9

9 – Belém (0.4) 3.1 3.1 3.5 3.2 3.2 3.4 1.9 0.7

10 – Pernambuco (0.1) 3.1 3.2 3.5 3.3 3.4 3.3 1.8 0.6 0.3

11 – T. a. aethiops 5.0 5.1 5.7 5.2 5.3 6.0 4.3 4.8 4.5 4.7

12 – T. aroyae 5.2 5.5 6.2 5.2 5.5 5.5 5.2 5.8 5.5 5.6 6.0



9/16

areas of endemism were inhabited by four distinct non-sister pop-

ulations with the following distributions: (1) northern portion of

the Rondônia area of endemism in the Tapajós–Madeira interfluve,

probably limited to the south by the Aripuanã River; (2) southern

part of the Rondônia area of endemism, probably limited to the

north by the Aripuanã River, crossing the Tapajós–Madeira inter-

fluve eastwards into the headwaters of the Tapajós and Xingu Riv-

ers, and hence reaching the southernmost parts of the Tapajós andXingu areas of endemism; (3) eastern Inambari area of endemism

in the lower–central portion of the Purus–Madeira interfluve; and

(4) western Inambari area of endemism west of the Purus River

and south of the Marañon/Huallaga/Ucayali/Solimões River in

Peru, eastward reaching the headwaters of the Purus–Madeira

interfluve into northeastern Bolivia (Pando and La Paz Depart-

ments). Uncorrected genetic p-distances among populations of T.aethiops ranged from 0.6% (between Belém/Pernambuco and Xingupopulations) to 6.0% between Andean foothill and northern Rondô-

nia area of endemism populations, with the average p-distancesbetween reciprocally monophyletic sisters populations around

1.6%.

The coallescent species tree recovered an identical topology to

those of BI and ML analyses based on the concatenated dataset

(Fig. 1e). The species tree also recovered with high statistical sup-

port the monophyly of the polytypic T. aethiops, as well as that of the Amazonian populations, the eastern and western clades, and

the internal relationships within them. Only the sister relationship

between the Western Inambari and the Northern Solimões River

populations was not statistically supported (posterior probability

less than 0.95; Fig. 1e).

The individual topologies of each nuclear gene (Bf5, Bf7 and

Musk4, results not shown) do not support any particular well-

resolved relationship among individuals of T. aethiops, except forMusk4, which supported the monophyly of lowland Amazonian

and Atlantic Forest populations.

3.3. Species delimitation

We obtained equal results among all six BP&P analyses, using

both algorithms and the three different combinations of s and hvalues, assuming a cutoff of 95% of posterior probability. Species

delimitations results were robust to the choice of prior distribu-

tions, supporting most of the reciprocally monophyletic popula-

tions recovered by the BI, ML, and species tree estimates as

distinct species, except for the Belém/Pernambuco and Eastern/

Western Negro sister population pairs (Fig. 1d). The purported spe-

ciation event between Belém and Pernambuco populations had

higher speciation probabilities when we used priors representing

large ancestral populations and deep divergences (PP = 0.90) or

shallow divergences (PP = 0.89), but very low values with priors

representing small ancestral populations and shallow divergences

(PP = 0.10), using algorithm 0 with the fine-tuning parameter setto 5. The split between Eastern and Western Negro populations

had very low speciation probabilities in all analyses (PP 6 0.20).

3.4. Molecular dating

According to the molecular dating based on the concatenated

mtDNA dataset (Fig. 1e), the split between T. aethiops and T. aroyaetook place during the Pliocene between 4.07 and 2.85 Mya (mean

3.45 Mya). The divergence between the Andean (T. a. aethiops) andthe ancestor of the Amazonian and Atlantic Forest populations took

place between 3.54 and 2.44 Mya (mean 2.97 Mya).

The earliest split within Amazonian populations occurred in the

Late Pliocene–Early Pleistocene (2.70–1.86 Mya, mean 2.27 Mya),

with separation of the eastern and western clades across theTapajós River. The first split in the western clade separated

populations from the northern part of the Rondônia area of ende-

mism (east of Aripuanã River) from the remaining ones found to

the west of this river and across the Madeira, Solimões, and Negro

rivers, during the Early–Middle Pleistocene (1.84–1.19 Mya, mean

1.50 Mya). Other cladogenetic events in the western clade

stretched from the Middle to Late Pleistocene between 1.4–0.29

Mya (mean 0.84 Mya), with one of them (involving southern

Rondônia + eastern Inambari versus western Inambari + northernSolimões populations) coinciding to some extent with the modern

course of the Purus River (Fig. 1b and e). The splits in the eastern

clade occurred between the Middle (1.27–0.66 Mya, mean 0.95

Mya) and Late Pleistocene (0.49–0.15 Mya, mean 0.31 Mya) and

separated populations of the Tapajós, Xingu, and Belém/Pernam-

buco areas of endemism across the Xingu and Tocantins rivers.

The splitting dates estimated by the multi-locus coalescent

approach (species tree) were in general more recent than those

obtained by the single locus mtDNA dataset, particularly for those

younger than 1.0 Mya. However, both coalescent and concatenated

mtDNA timing estimates overlapped to various degrees in their

respective 95% confidence intervals, and thus cannot be considered

statistically different.

3.5. Population genetics and historical demography

Haplotype networks recovered for cyt b and ND2 mirrored thephylogenetic analyses based on BI and ML (Fig. 1a and c). For cyt

b, populations with the highest number of sequenced individuals

differed dramatically in structure: the western Inambari popula-

tion presented a dominant haplotype shared by six individuals

and 12 other haplotypes represented by one or two individuals,

generally differing from the commonest haplotype by a single

mutation. On the other hand, in the southern Rondônia population,

10 different haplotypes distributed in three groups separated by

two to eight mutations were recovered, with only three haplotypes

present in more than one individual (Fig. 1c). Networks of the

nuclear makers recovered the absence of phylogeographic struc-

ture and haplotype sharing among the recognized populations(Fig. 1c).

Haplotype diversity was generally high and similar among the

sampled Amazonian populations (Table 3). Estimated nucleotide

diversity varied widely among populations depending on the locus

considered, while within a single locus values obtained for the dif-

ferent populations were similar (Table 3), suggesting an overall

demographic stability. The Tajima’s D test (1989) showed negative

and significant values for the mtDNA and Bf5/Bf7 loci in the wes-

tern Inambari population, consistent with demographic fluctua-

tions (Table 3). The Fu’s Fs (1997) showed significant and

negative values for the mtDNA and musk4 loci in the northern

Rondônia population and in all three sampled loci in the western

Inambari, also supporting demographic instability. The R2 (2002)

test results showed low and significant values for the same popu-lations as the Fu’s Fs (1997), suggesting a similar demographic sce-

nario regardless of the test considered.

The EBSP’s (Fig. 2) were in part congruent with the neutrality

tests, inferring tendencies of recent demographic expansions in

both populations of the Inambari and the northern Rondônia areas

of endemism during the Late Pleistocene and Holocene. In contrast,

EBSP’s favored scenarios of demographic stability in the Belém,

Pernambuco, Xingu, Tapajós, and southern Rondônia populations

(Fig. 2).

An analysis of molecular variation (AMOVA; Table 5) confirmed

that most of the genetic variation in the Amazonian and Atlantic

Forest populations of T. aethiops is partitioned among the geo-

graphically structured clades recovered by the phylogenetic analy-

ses, and not within these groups. Similarly, these resultscorroborate that the large tributaries of the Amazon River are at


http://-/?-http://-/?-


10/16

least contemporary barriers to gene flow between populations

from opposite banks.

3.6. Gene flow among populations

IMa analyses favored scenarios of very low levels or absence of

gene flow among all populations of T. aethiops potentially in con-

tact (Table 6). Migration events were not rejected by LRT tests

among most pairwise comparisons, with models predicting migra-tion having better AIC scores than models predicting no gene flow

between the following clades: Xingu/Tapajós, Northern Rondônia/

Southern Rondônia, Northern Rondônia/Eastern Inambari, South-

ern Rondônia/Western Inambari, and Eastern Inambari/Western

Inambari. In all these instances, the estimated migration rate was

very low in both directions (with confidence intervals’ lower bonds

reaching zero except in one comparison), and involved non-sister

populations (Fig. 1, Table 6). The highest migration rate detected

and the only one not including a lower bond likelihood of zero,

involved the Eastern Inambari and Western Inambari populations(Maximum likelihood = 0.0015; 90% HPD interval = 0.001–0.004).

Fig. 2. Demographic histories of major clades of Thamnophilus aethiops for which at least five individuals were sampled, inferred through Extended Bayesian Skyline Plots

based on mtDNA (cyt b and ND2), Bf5/Bf7, and Musk4 sequences. Black solid lines represent median values, while the dashed lines corresponds to 95% confidence intervals.

The X axis corresponds to time in million years before present. The Y axis represents N es in a logarithmic scale.



11/16

4. Discussion

4.1. Species limits and taxonomy

Although our analyses support the monophyly of the polytypic

T. aethiops, of the 10 reciprocally monophyletic populations recov-ered only two are entirely consistent with subspecies limits cur-

rently recognized in this group: nominate aethiops from thefoothills of the Andes and T. a. injunctus, which corresponds tothe eastern Inambari population (Zimmer and Isler, 2003; Fig. 1a,

d and e). In fact, high statistical support for key nodes in the

multi-locus coalescent species tree and high speciation probabili-

ties recovered by the species delimitation analysis (BP&P), support

the recognition of nine reciprocally monophyletic lineages fitting

the definition of evolutionary species (see Fujita et al., 2012;Fig. 1d) in T. aethiops.

These lineages are as follows (taxon names are assigned torecovered populations based on type locality and nomenclatural

priority following Peters 1951 and Zimmer and Isler 2003): (1)

Thamnophilus aethiops Sclater, 1858 (foothills of the Andes); (2)Thamnophilus atriceps Todd, 1927 (Tapajós population, distributedthroughout most of the Tapajós area of endemism); (3) Thamnophi-lus incertus Pelzeln, 1869 (Belém and Pernambuco populations;

given their lack of reciprocal monophyly and low speciation prob-

abilities, the former Thamnophilus aethiops incertus from the Belémarea of endemism and Thamnophilus aethiops distans Pinto 1954from the Atlantic Forest should be regarded as subspecies of a sin-

gle species, T. incertus, the name with nomenclatural precedence);(4) Thamnophilus punctuliger Pelzeln 1869 (northern Rondôniapopulation distributed east of the Aripuanã River in the Madeira–

Tapajós interfluve); (5) Thamnophilus injunctus Zimmer 1933(eastern Inambari population, distributed through most of the

Table 5

Results of an Analysis of Molecular Variation (AMOVA) based exclusively on the mtDNA among major clades of the polytypic Thamnophilus aethiops for which at least five

individuals were sampled, separated by potential modern barriers (rivers and non-forest habitats). Results regarded as significant if P < 0.001.

Populations Potential barrier Among p opulations o n opposite s ides o f

the barrier

Within the population in the same side

of the barrier

Ust p-

Value

Pernambuco/Belém Caatinga/Cerrado (non-forest

landscapes)

18.77 81.23 0.18 0.026

Belém/Xingu Tocantins R. 60.42 39.58 0.60 0.004

Xingu/Tapajós Xingu R. 91.08 8.92 0.91 0.002Xingu/Southern Rondônia Unknown 84.78 15.22 0.84 0.000

Tapajós/Northern Rondônia Lower Tapajós R. 71.25 28.75 0.71 0.000

Tapajós/Southern Rondônia Teles Pires R. (part) 46.49 53.51 0.46 0.000

Northern/Southern Rondônia Aripuanã R. (part) 86.05 13.95 0.86 0.000

Northern Rondônia/Eastern

Inambari

Lower Madeira R. 93.39 6.61 0.93 0.000

Southern Rondônia/Eastern

Inambari

Middle Madeira R. 73.82 26.18 0.73 0.000

Southern Rondônia/Western

Inambari

Upper Madeira R. 84.84 15.16 0.84 0.000

Eastern/Western Inambari Purus R. (part) 92.26 7.74 0.92 0.000

Table 6

Summary of IMa results of migration estimates between parapatric population pairs of T. aethiops potentially in contact (see Fig. 1b). Posterior distribution values of migration

were converted to demographic units, where the values of m1 and m2 represent the average number of migrations per 1000 generations per gene copy on the coalescent

(m1 = rate of migration from population 2 to population 1; m2 = rate of migration from population 1 to population 2). Confidence intervals represent HPD90. All maximum values

for migration reached likelihood zero, indicating the posterior distribution ends.

Pairwise analyses m1 m2 AIC

Belém (5)/Xingu (5) 0 (0–0.0011) 0.0027 (0–0.0063) 9569.71

Forced to be zero Forced to be zero 9548.03

Xingu (5)/Tapajós (6) 0 (0–0.0033) 0 (0–0.003) 9586.18


Xingu (5)/Southern Rondônia (15)* 0 0 –

Tapajós (6)/Northern Rondônia (7)* 0 0 –

Tapajós (6)/Southern Rondônia (15)* 0 0 –

Northern Rondônia (7)/Southern Rondônia (15) 0 (0–0.0012) 0 (0–0.0013) 10366.19


Northern Rondônia (7)/Eastern Inambari (6) 0 (0–0.0013) 0 (0–0.002) 9996.38Forced to be zero Forced to be zero 10012.88

Southern Rondônia (15)/Eastern Inambari (6) 0 (0–0.0016) 0.0013 (0–0.0032) 9920,21


Southern Rondônia (15)/Western Inambari (9) 0 (0–0.0019) 0 (0–0.0012) 10963.26


Eastern Inambari (6)/Western Inambari (9) 0 (0–0.0032) 0.0015 (0.001–0.004) 10369.91


Western Inambari (9)/Northern Solimões (7)* 0 0 –

Asterisks indicate pairwise analyses where the likelihood ratio test rejects a scenario with m > 0. For pairwise comparisons where the likelihood ratio test supports a scenario

with m > 0, models with and without migration were contrasted with the Akaike information criterion (AIC). The model that minimized AIC was chosen as the best scenario

(in bold). Numbers in parenthesis next to population names denote the number of individuals of the respective population used in the comparison. See Table 1 for detailed

specimen and locality information.


http://-/?-http://-/?-http://-/?-http://-/?-


12/16

Madeira–Purus interfluve); (6) Thamnophilus juruanus Iheringi1904 (western Inambari population, distributed through most of

the Inambari area of endemism west of the Purus River; the taxon

name Thamnophilus aethiops kapouni Seilern 1913 also applies to

this population, but it has no nomenclatural priority, and therefore

should be synonymized with juruanus); (7) Thamnophilus poliono-

tus Pelzeln 1869 (western and eastern Negro populations distrib-

uted north of the Amazon in the Branco–Solimões interfluve; asexplained above the taxon Thamnophilus aethiops wetmorei deSchauensee, 1945 is tentatively included in this lineage); and,

finally, two additional unnamed lineages (no known taxon

described previously from their ranges) corresponding respectively

to most of the Xingu area of endemism (assigned currently to T.atriceps, whose type locality is in the Tapajós area of endemism)and the southern part of the Rondônia, Tapajós, and Xingu areas

of endemism (assigned currently to T. punctuliger , whose typelocality lies in the northern part of the Rondônia area of

endemism).

By occupying mostly shaded environments in the understory of

tropical forests, the evolution of plumage color states traditionally

used in subspecies delimitations in T. aethiops probably proceededat a slower pace than at least one of the sequenced loci (mtDNA).

This process can generate the formation of morphologically cryptic

and sometimes polymorphic lineages, resulting in the taxonomic

inconsistencies cited above (Baker et al., 1995; Irwin et al., 2001;

Fernandes et al., 2012). In contrast, vocal characters are thought

to be directly involved in the reproductive isolation of forest inte-

rior suboscine lineages (Carneiro et al., 2012). We do not report on

the vocal variation within the polytypic T. aethiops in this paper,but will present in a separate publication a detailed taxonomic

overhaul of the T. aethiops complex based on a multi character

approach, where the unnamed evolutionary lineages mentioned

above will be described.

The only sister populations of T. aethiops for which low specia-tion probabilities were recovered are the parapatric eastern/wes-

tern Negro populations (T. polionotus) and the allopatric Belém/

Pernambuco populations (T. i. incertus and T. i. distans). Eventhough the species tree recovered the former pair as distinct spe-

cies, the species delimitation analyses favored a scenario of very

shallow divergence between them, not supporting their status as

distinct evolutionary species (Fig. 1d and e). Indeed, the lack of

reciprocal monophyly in the mtDNA between the Belém (T. i. incer-tus) and Pernambuco (T. i. distans) populations supports a very

recent divergence between them, despite their current separation

by over 1000 km.

The overall lack or very low rate of gene flow estimated among

most parapatric populations of T. aethiops are also indicative of

their status as independent species (Table 6). Interestingly, when-

ever lack of gene flow (i.e., zero migration rates) could be statisti-

cally rejected between any population pair potentially in contact, it

always involved non-sister lineages, implying in secondary contactrather than primary intergradation (Table 6). Hence, it is possible

that these very low rates of gene flow are explained by a significant

degree of reproductive isolation already acquired by these popula-

tions that have diverged in allopatry for a relatively long time

before coming in contact. Estimates of gene flow between non-sis-

ter populations or populations that exchange migrants with a third

unsampled population go against a main assumption of the IMa

model and can generate several biases in the estimation of the

main parameters, although they have little effect when low levels

of gene flow are detected (see Strasburg and Rieseberg 2010), as

documented herein. IMa analyses supported absolute no gene flow

among all populations of the two main Amazonian clades of T.aethiops separated by the Tapajós River (eastern clade: Xingu and

Tapajós populations; and western clade: northern Rondônia andsouthern Rondônia populations), which likely come in contact

across a wide area in the headwaters of the Tapajós, Xingu, and

Tocantins rivers (Fig. 1b, Table 6). Despite sampling limitations,

this study provides the best evidence available so far that interflu-

vial populations of terra-firme avian lineages in contact across the

headwaters of southern Amazonian tributaries can retain their

genetic diagnosabilities, probably as a result of an advanced degree

of reproductive isolation acquired in allopatry before coming in

contact in more recent times, as suggested by several other studiesbased solely on mtDNA datasets (Carneiro et al., 2012; Batista

et al., 2013; Rodrigues et al., 2013). In contrast, gene flow was pres-

ent, yet at a very small rate, between the non-sister Tapajós and

Xingu populations belonging to the same eastern clade, as well

as between the southern Rondônia and western Inambari popula-

tions of the western clade across the upper Madeira River ( Fig. 1,

Table 6). Therefore, comparative levels of gene flow across the

headwaters of major Amazonian tributaries are likely to be more

correlated with the phylogenetic distance of the lineages/popula-

tions in contact, rather than the smaller barrier effect of rivers in

these areas (Haffer 1992), which result in a more random pattern

of gene flow.

4.2. The diversification of the Thamnophilus aethiops complex in Cis- Andean South America

Populations of T. aethiops distributed in the Amazon and Atlan-tic Forests are in a clade with several basal Andean taxa typically

occurring above 1000 m (T. unicolor, T. aroyae, and T. a. aethiops),

supporting an ancestor of Andean origin for these lowland forest

populations (see also Brumfield and Edwards, 2007). Diversifica-

tion events along Andean altitudinal gradients have been proposed

for some avian lineages (Bates and Zink, 1994; Weir, 2006; Isler

et al., 2012; d’Horta et al., 2013), and are probably related to allo-

patric speciation during altitudinal shifts in habitats due to cli-

matic oscillations (Graham et al., 2004). Under this diversification

model, the dispersal of an Andean ancestor into the Amazonian

lowlands would only be possible after the draining of the lacus-

trine system of the Solimões Basin in western Amazonia (the socalled Pebas and Amazonas lakes), which led to the formation of

the modern Amazon drainage (Hoorn et al., 2010) and the estab-

lishment of upland terra-firme forests in this region. Paleogeo-graphic hypotheses differ in the time frame when these events

occurred (Campbell et al., 2006; Figueiredo et al., 2009;

Latrubesse et al., 2010), even though they coincide in postulating

a favorable environment for the establishment of upland terra- firme forests in western Amazonia between the late Pliocene andearly Pleistocene, when the diversification of the T. aethiops com-plex in Amazonia began (Fig. 1e).

The genetically poorly-differentiated yet isolated Atlantic Forest

population (T. i. distans) may have originated from a recent dis-persal event of the Belém population (T. i. incertus) or through iso-

lation caused by climatic fluctuations leading to the disruption of the forest corridor linking eastern Amazonia with the northeastern

Atlantic Forest during the Late Quaternary (Batalha-Filho et al.,

2013). In contrast, other Amazonian and Atlantic Forest avian sister

lineages exhibited reciprocal monophyly and high levels of genetic

differentiation estimated to date back to a time frame stretching

from the Late Pliocene to the Early Pleistocene (Aleixo, 2002;

Cabanne et al., 2008; Weir and Price, 2011). These different phylog-

eographic patterns are consistent with multiple and cyclical con-

nection and isolation episodes between eastern Amazonia and

the northern sector of the Atlantic Forest until at least the Late

Quaternary (Batalha-Filho et al., 2013).

The data presented herein support that distinct events of vicar-

iance related to river formation and reorganization mostly during

the Pleistocene have influenced the diversification of the T. aethiopscomplex, providing another body of evidence that different



13/16

modern interfluves in Amazonia harbor distinct evolutionary lin-

eages, with little or no gene flow among them (Table 6). In the T.

aethiops complex, all splits involving the recognized lineages coin-cide with at least part of the course of a major Amazonian river,

indicating that at some point in time these rivers might have acted

as primary diversification barriers. Interestingly, the absence of a

physical barrier separating the eastern and western Inambari

clades where they meet in the middle Purus–Madeira interfluveis associated with the highest gene flow rate recovered herein

among lineages of the T. aethiops complex, yet still very low(0.0015; Table 6).

The phylogeographic data presented herein is consistent with at

least two independent splits across different portions of the mod-

ern course of the Madeira River as also suggested for several other

avian lineages (Aleixo 2004; Patané et al., 2009; Fernandes et al.,

2013; Sousa-Neves et al., 2013). Latrubesse (2002) postulated that

the Madeira paleodrainage was broader and more complex than

today until the late Pleistocene, reporting the presence of two

mega-fans related to the Aripuanã and Jiparaná rivers, both of

which currently flow into the Madeira River. Thus, it is possible

that the initial isolation of the northern Rondônia lineage in the

western clade is related to a former larger barrier (Aripuanã

mega-fan), which decreased in size due to a drainage-capture

event (Wilkinson et al., 2010), promoting the secondary contact

between different non-sister lineages that diverged in allopatry,

as postulated for other terra-firme avian lineages (Fernandes2013). The absence or very low levels of gene flow detected among

populations of the T. aethiops complex suggest a fickle scenario of migration across Amazonian rivers, which is probably best

explained by an overall inability to cross open water, as demon-

strated for other understory Neotropical avian lineages (Moore

et al., 2008). Thus, multiple and independent splitting events along

a single modern river course are more likely explained by the dis-

ruption of river drainages, whereby ‘‘barrier effects’’ move across

neighbouring interfluvia.

Likewise, the estimated dates for the splits across the Solimões

and Negro rivers long after the formation of these barriers, suggesta similar scenario of river lateral change or course capture. The for-

mation of the Solimões River is closely related to the formation of

the transcontinental Amazonian drainage during the Late Plio-

cene–Early Pleistocene (Campbell et al., 2006; Hoorn et al., 2010;

Latrubesse et al., 2010). Thus, if these time estimates for the estab-

lishment of the Solimões River are correct, the relatively recent

split of T. aethiops populations across this river (0.68–0.26 Mya)is explained by vicariance following dispersal after the formation

of the current Solimões River. As observed for clades separated

by the Solimões, the formation of the Negro River (see Almeida-

Filho and Miranda 2007) probably predates the cladogenetic event

separating the Imeri/Napo and western Guianan clades of T. polion-otus (0.23–0.02 Mya). Unlike widely reported local instances of

river-channel migration due to sedimentation involving ‘‘white-water’’ rivers in western Amazonia (Salo et al., 1986; Patton and

Silva, 1998), some shifts in Amazonian rivers courses take place

over long periods of time and are probably related to drainage cap-

ture events mediated by tectonics (Almeida-Filho and Miranda

2007; Rossetti and Valeriano, 2007; Shephard et al., 2010). The

growing body of phylogeographic evidence available now for Ama-

zonian lineages suggest that river courses, particularly those of the

Negro, Madeira, Tapajós and Tocantins, have been stable for long

periods of time until changing part of their courses, and then

remaining stable for an additional time (Ribas et al., 2012;

d’Horta et al., 2013; Fernandes et al., 2012, 2013; Maldonado-

Coelho et al., 2013; Sousa-Neves et al., 2013). Hence, a dynamic

drainage landscape, over long periods of time, can act like a speci-

ation pump rather than prevent diversification, as suggested previ-ously (Haffer 1993; Colwell 2000; Gascon et al., 2000).

When compared to other upland terra-firme avian lineageswhich were also shown to have their diversification mediated by

Amazonian rivers (Fernandes et al., 2012, 2013; Ribas et al.,

2012; d’Horta et al., 2013; Sousa-Neves et al., 2013), the order

and timing in which different rivers accounted for splits of lineages

of the T. aethiops complex differ significantly from most of them,indicating that distinct lineages responded differently to the pro-

cess of drainage evolution in Amazonia. Perhaps the most similarcases to that of the T. aethiops complex are those reported for the Xiphorhynchus spixii/elegans (Dendrocolaptidae; Aleixo 2004) and

Phlegopsis nigromaculata (Thamnophilidae; Aleixo et al., 2009),whereby the Tapajós River also accounted for the first split in time.

One hypothesis is that these differences in the relative order and

timing in which different rivers account for splits throughout the

Amazon are related to the geografic origins of the ancestral popu-

lations of these widespread lineages (Aleixo and Rossetti 2007). In

contrast to the T. aethiops complex, X. spixii/elegans and Sclerurus

mexicanus (d’Horta et al., 2013), whose sister groups in Amazoniainclude widespread Andes lineages, a different but pervasive phy-

logeographic pattern in the Amazon is that where earliest splits

usually involve lineages endemic to the Guianan shield and coin-

cide either with the present course of the lower Amazon or Negro

rivers (Carneiro et al., 2012; Batista et al., 2013; d’Horta et al.,

2013; Rodrigues et al., 2013; Sousa-Neves et al., 2013). This alter-

native pattern imply a Guianan shield origin for these lineages,

whereas those reported for T. aethiops, P. nigromaculata, and X. spixii/elegans (which are absent or nearly absent from the Guianan

shield) support a Brazilian shield origin (see also Aleixo and

Rossetti 2007). Therefore, we hypothesize that the structuring of

the modern Amazon drainage and the drying out of the western

Amazonian sedimentary basins from the late Pliocene to the late

Pleistocene were keystone events to promote concomitantly: (1)

an increase in the dispersal rates of both Brazilian and Guianan

shield lineages across the more recent upland terra-firme forestsof the upper Amazonia (Solimões sedimentary basin); and (2) an

increase in cladogenetic events involving upland terra-firme forest

lineages due to the latest cycle of drainage evolution throughoutAmazonia, which shaped the modern transcontinental river sys-

tem. Under this scenario, depending on the geographical origins

of any given lineage before this latest cycle of widespread drainage

structuring (late Pliocene–Early Pleistocene), their response to riv-

erine barriers can differ widely, as verified by the phylogeographic

studies conducted so far.

When climate change is evaluated as a possible driver of diver-

sificationin the T. aethiops complex, neutrality tests and EBSP’s sup-port demographic expansions in lineages distributed in the

Inambari and northern Rondônia areas of endemism since the last

interglacial cycle (approximately 0.13 Mya) to the present (Fig. 2).

Recent demographic expansions have been postulated for other lin-

eages of birds in western Amazonia (Aleixo, 2004; Fernandes et al.,

2012; Ribas et al., 2012), but in all cases there was no evidence thatpurported refuges were correlated with cladogenetic events in

these lineages. The dates estimated for the origin of the main recip-

rocally monophyletic lineages of the T. aethiops complex based on

the multi-locus species tree precede the coalescent time of the

locus used in EBSP’s analyzes as well as the estimated time of the

demographic expansions. Thus, if climatic oscillations during the

last 0.2 Mya affected populations of the T. aethiops complex, theyapparently were not strong enough to generate the formation of

new clades, hence not supporting the refuge hypothesis as an

important cause of diversification in the T. aethiops complex(Haffer, 2001). Furthermore, differences in the historical population

demography among T. aethiops complex lineages (Fig. 1c and 2),

with some lineages exhibiting relatively stable demographic histo-

ries, suggest that climate oscillations during the Late Pleistocenedid not impact uniformly all sectors of Amazonia. Even though



14/16

the sparse sampling of some T. aethiops complex lineages may yieldunreliable results that underestimate the effect of climatic oscilla-

tions on ancestral population sizes during the last glacial cycles,

our data is in agreement with that recovered for several other

linages of Amazonian organisms where no significant changes in

historical population structures were found since the Last Glacial

Maximum (Lessa et al., 2003; Fernandes et al., 2012; Ribas et al.,

2012; Maldonado-Coelho et al., 2013). These results have ignitedan intense debate on the applicability of the refuge hypothesis in

explaining patterns of diversification in Amazonia (Bush and de

Oliveira, 2006), along with paleo-environmental uncertainties con-

cerning the existence and location of refuges (Bush, 1994; van der

Hammen and Hooghiemstra, 2000; Haffer, 2001).

Acknowledgments

We thank the curator and curatorial assistants of the Museum

of Natural Science, Louisiana State University (LSUMZ), for allow-

ing us to sequence tissues and study specimens under their care.

We also grateful to the tenacious effort of many specimen collec-

tors working for or together with the Museu Paraense Emílio Goe-

ldi (MPEG), who along years of intense field work in the Amazon,

amassed the tissues necessary for the analyzes contained in this

paper, with permits provided by ICMBIO (Instituto Chico Mendes

de Conservação da Biodiversidade). Field and laboratory work

related to this paper were generously funded by the Brazilian

Research Council (CNPq; grants # 310593/2009-3; ‘‘INCT em Bio-

diversidade e Uso da Terra da Amazônia’’ # 574008/2008-0; #

563236/2010-8; and # 471342/ 2011-4) and the Pará State Funding

Agency (ICAAF 023/2011). Support to GT’s graduate research was

provided by a CNPq Master’s fellowship (# 131000/2010-1). AA

is supported by a CNPq research productivity fellowship (#

310880/2012-2). We thank an anonymous referee as well as Mar-

cos Maldonado-Coelho, Renato Caparroz, Luciano N. Naka, Camila

Ribas, and Péricles S. Rego for reviewing earlier versions of this

manuscript.

References

Ab’Saber, A.N., 1977. Os domínios morfoclimáticos da América do Sul. Primeiraaproximação. Geomorfologia 53, 1–23.

Aleixo, A., 2002. Molecular systematics and the role of the ‘‘várzea’�

cryptic speciation in the white shouldered antshrike thamnophilus aethiops aves thamnophilidae the...

Documents